WEBVTT 00:00:00.000 --> 00:00:01.968 align:middle line:90% [SQUEAKING] 00:00:01.968 --> 00:00:02.468 align:middle line:90% 00:00:02.468 --> 00:00:03.444 align:middle line:90% [RUSTLING] 00:00:03.444 --> 00:00:03.944 align:middle line:90% 00:00:03.944 --> 00:00:07.388 align:middle line:90% [CLICKING] 00:00:07.388 --> 00:00:12.800 align:middle line:90% 00:00:12.800 --> 00:00:14.960 align:middle line:84% NELSON REPENNING: What do you think you get out 00:00:14.960 --> 00:00:18.080 align:middle line:90% of this if you do it well? 00:00:18.080 --> 00:00:22.240 align:middle line:84% So if we have good model that gets published in a good place, 00:00:22.240 --> 00:00:25.000 align:middle line:84% what can we reasonably expect that a system dynamics 00:00:25.000 --> 00:00:26.640 align:middle line:90% model could produce? 00:00:26.640 --> 00:00:30.820 align:middle line:84% What would be the upshot of this type of analysis? 00:00:30.820 --> 00:00:33.920 align:middle line:90% 00:00:33.920 --> 00:00:35.203 align:middle line:90% Yeah, go ahead. 00:00:35.203 --> 00:00:36.620 align:middle line:84% STUDENT: Like you were saying, you 00:00:36.620 --> 00:00:38.720 align:middle line:84% can discover good policy that could 00:00:38.720 --> 00:00:41.220 align:middle line:84% help fix any problems you're encountering. 00:00:41.220 --> 00:00:44.200 align:middle line:84% But also, at least I find in the process of actually writing 00:00:44.200 --> 00:00:46.320 align:middle line:84% the entire model out, you usually 00:00:46.320 --> 00:00:50.820 align:middle line:84% uncover parts of the problem that aren't immediately visible. 00:00:50.820 --> 00:00:53.340 align:middle line:84% So I guess that goes back to policy. 00:00:53.340 --> 00:00:55.920 align:middle line:84% Like, you actually start to robustly understand 00:00:55.920 --> 00:00:59.090 align:middle line:84% the entire issue rather than just the very small, like, 00:00:59.090 --> 00:01:02.290 align:middle line:84% first-order ideal kind of things. 00:01:02.290 --> 00:01:03.650 align:middle line:90% NELSON REPENNING: OK. 00:01:03.650 --> 00:01:06.970 align:middle line:84% STUDENT: I would say also testing mental models. 00:01:06.970 --> 00:01:13.510 align:middle line:84% So some of the mental models are actually [INAUDIBLE]. 00:01:13.510 --> 00:01:22.852 align:middle line:90% 00:01:22.852 --> 00:01:23.810 align:middle line:90% NELSON REPENNING: Good. 00:01:23.810 --> 00:01:25.853 align:middle line:90% [INAUDIBLE] 00:01:25.853 --> 00:01:26.770 align:middle line:90% STUDENT: I don't know. 00:01:26.770 --> 00:01:29.690 align:middle line:90% This one really bothers me. 00:01:29.690 --> 00:01:31.370 align:middle line:84% My first-order answer is you can learn 00:01:31.370 --> 00:01:33.370 align:middle line:84% about the consequences of how things 00:01:33.370 --> 00:01:36.310 align:middle line:84% are organized if you're asking the organizational question. 00:01:36.310 --> 00:01:40.010 align:middle line:90% 00:01:40.010 --> 00:01:40.590 align:middle line:90% That's it. 00:01:40.590 --> 00:01:42.890 align:middle line:90% NELSON REPENNING: OK. 00:01:42.890 --> 00:01:44.450 align:middle line:84% STUDENT: You understand much better 00:01:44.450 --> 00:01:48.630 align:middle line:84% what is kind of the nature of what's driving the system. 00:01:48.630 --> 00:01:52.130 align:middle line:84% You feel you understand better what's happening. 00:01:52.130 --> 00:01:53.970 align:middle line:84% It's really helpful for me, I think. 00:01:53.970 --> 00:01:54.845 align:middle line:90% NELSON REPENNING: OK. 00:01:54.845 --> 00:01:57.740 align:middle line:90% 00:01:57.740 --> 00:01:59.780 align:middle line:84% STUDENT: I think you can understand the feedback 00:01:59.780 --> 00:02:00.700 align:middle line:90% loops that in-- 00:02:00.700 --> 00:02:03.000 align:middle line:84% that are driving the behavior of the system, 00:02:03.000 --> 00:02:05.700 align:middle line:84% and also the leverage points where 00:02:05.700 --> 00:02:09.020 align:middle line:84% you can apply the policy, more strict policies, in order 00:02:09.020 --> 00:02:12.940 align:middle line:84% to change the behavior of the system in the way you want. 00:02:12.940 --> 00:02:14.100 align:middle line:90% NELSON REPENNING: OK. 00:02:14.100 --> 00:02:17.020 align:middle line:84% STUDENT: Yeah, I think sometimes, it can have a-- 00:02:17.020 --> 00:02:18.140 align:middle line:84% NELSON REPENNING: You're going to have to speak up. 00:02:18.140 --> 00:02:20.015 align:middle line:84% STUDENT: Sometimes, it can have an organizing 00:02:20.015 --> 00:02:22.460 align:middle line:84% and a calming effect on the conversation with the people 00:02:22.460 --> 00:02:24.500 align:middle line:90% whose problem it is. 00:02:24.500 --> 00:02:27.300 align:middle line:84% And so I think when you're talking with people about, 00:02:27.300 --> 00:02:30.493 align:middle line:84% oh, these things are horrible, what's happening, 00:02:30.493 --> 00:02:32.160 align:middle line:84% and this, and that, and the other thing, 00:02:32.160 --> 00:02:34.800 align:middle line:84% and it's everything, and it's totally unsolvable. 00:02:34.800 --> 00:02:37.380 align:middle line:84% And then what you get out of distilling it into a model 00:02:37.380 --> 00:02:40.660 align:middle line:84% is, like, you get to validate what 00:02:40.660 --> 00:02:42.520 align:middle line:90% they're talking about to them. 00:02:42.520 --> 00:02:45.460 align:middle line:90% And it kind of is-- 00:02:45.460 --> 00:02:47.100 align:middle line:90% it makes sense. 00:02:47.100 --> 00:02:49.920 align:middle line:84% Their own problem makes sense in a different way to them. 00:02:49.920 --> 00:02:52.820 align:middle line:84% And then they can kind of come up with different things 00:02:52.820 --> 00:02:56.813 align:middle line:84% and stop getting stuck in the same thought patterns. 00:02:56.813 --> 00:02:58.230 align:middle line:84% NELSON REPENNING: Yeah, so I think 00:02:58.230 --> 00:03:00.990 align:middle line:90% you're all on the right track. 00:03:00.990 --> 00:03:03.110 align:middle line:84% For my tastes, and again, I think my view of this 00:03:03.110 --> 00:03:07.630 align:middle line:84% probably differs slightly from Hasier and John, is in the end, 00:03:07.630 --> 00:03:11.110 align:middle line:84% all we're testing is the sort of logical consequences 00:03:11.110 --> 00:03:14.530 align:middle line:84% and consistency of the theory that's embodied in the model. 00:03:14.530 --> 00:03:16.830 align:middle line:84% It's not a statement about the real world. 00:03:16.830 --> 00:03:18.750 align:middle line:84% It's a statement about the internal logic 00:03:18.750 --> 00:03:22.430 align:middle line:84% of the assumptions that you have made. 00:03:22.430 --> 00:03:26.450 align:middle line:84% If you then, which I know Hasier has a lot of expertise, 00:03:26.450 --> 00:03:28.970 align:middle line:84% sort of validate it, and estimate it, and compare it, 00:03:28.970 --> 00:03:33.430 align:middle line:84% then you are, to some degree, establishing some connection 00:03:33.430 --> 00:03:36.310 align:middle line:84% to the sort of external phenomenon 00:03:36.310 --> 00:03:37.710 align:middle line:90% that you're trying to model. 00:03:37.710 --> 00:03:39.550 align:middle line:84% But always keep in mind, right, these 00:03:39.550 --> 00:03:41.410 align:middle line:90% are not randomized experiments. 00:03:41.410 --> 00:03:44.190 align:middle line:84% So the fact that your model fits the data 00:03:44.190 --> 00:03:47.685 align:middle line:84% does not necessarily make it true. 00:03:47.685 --> 00:03:49.310 align:middle line:84% But the thing that I want to highlight, 00:03:49.310 --> 00:03:54.313 align:middle line:84% and what really came out for me from reading 00:03:54.313 --> 00:03:56.980 align:middle line:84% the overviews of the problem you have is, at the end of the day, 00:03:56.980 --> 00:04:01.500 align:middle line:84% I would say that we either are looking for policies, as many 00:04:01.500 --> 00:04:02.500 align:middle line:90% of you mentioned, right? 00:04:02.500 --> 00:04:05.200 align:middle line:84% We need to figure out where to sort of intervene. 00:04:05.200 --> 00:04:07.040 align:middle line:84% Or I think for those of you that aspire 00:04:07.040 --> 00:04:10.680 align:middle line:84% to be academics, the kind of ivory-tower framing of that 00:04:10.680 --> 00:04:14.480 align:middle line:84% is, we're looking to explain puzzles. 00:04:14.480 --> 00:04:19.000 align:middle line:84% So why is it that the conventional theory would 00:04:19.000 --> 00:04:20.720 align:middle line:84% predict x, but in fact, we get y, 00:04:20.720 --> 00:04:23.680 align:middle line:84% or we get not x, or something like that? 00:04:23.680 --> 00:04:25.840 align:middle line:84% And that's really a sort of, at least 00:04:25.840 --> 00:04:28.060 align:middle line:84% originally, a theoretical question of, 00:04:28.060 --> 00:04:30.582 align:middle line:84% OK, if the conventional theory says x and we're seeing y, 00:04:30.582 --> 00:04:33.040 align:middle line:84% we need a set of assumptions that seem reasonable that then 00:04:33.040 --> 00:04:36.240 align:middle line:84% add up to y, then there's the sort of testing validation 00:04:36.240 --> 00:04:38.760 align:middle line:84% question that I think comes second in terms of, 00:04:38.760 --> 00:04:41.960 align:middle line:84% OK, now we have an internally consistent theory of why 00:04:41.960 --> 00:04:43.400 align:middle line:90% we're getting y. 00:04:43.400 --> 00:04:45.880 align:middle line:84% Do we think this is sort of remotely in the ballpark 00:04:45.880 --> 00:04:48.260 align:middle line:90% of a reasonable thing to do? 00:04:48.260 --> 00:04:52.090 align:middle line:90% And so I think-- 00:04:52.090 --> 00:04:55.190 align:middle line:84% I understand that these are just assignments, but for my taste, 00:04:55.190 --> 00:04:57.250 align:middle line:84% many of your overviews are a little muddled 00:04:57.250 --> 00:05:01.370 align:middle line:84% in terms of what you wanted to get out of your exercise. 00:05:01.370 --> 00:05:02.750 align:middle line:90% And I don't want to pick on one. 00:05:02.750 --> 00:05:04.875 align:middle line:84% I don't even remember who it was, but someone said, 00:05:04.875 --> 00:05:07.850 align:middle line:84% I want to study the effects of x on this system. 00:05:07.850 --> 00:05:11.250 align:middle line:84% And I don't think that's a great framing because it doesn't give 00:05:11.250 --> 00:05:14.050 align:middle line:84% you a kind of yardstick to decide 00:05:14.050 --> 00:05:16.790 align:middle line:84% what's going in your model and what's outside the model. 00:05:16.790 --> 00:05:20.490 align:middle line:84% So I would very much urge you to be much more precise 00:05:20.490 --> 00:05:24.690 align:middle line:84% about, there's either a specific phenomenon I want to understand 00:05:24.690 --> 00:05:28.290 align:middle line:84% that I don't understand, or there's a phenomenon out there 00:05:28.290 --> 00:05:31.090 align:middle line:84% that maybe we don't like, and we want to do something about it, 00:05:31.090 --> 00:05:31.590 align:middle line:90% right? 00:05:31.590 --> 00:05:34.863 align:middle line:84% So Jake, if I looked at yours, and we'll come back 00:05:34.863 --> 00:05:37.530 align:middle line:84% to that, because I think it's a very interesting topic about how 00:05:37.530 --> 00:05:39.110 align:middle line:84% the sort of-- if I understood it correctly, 00:05:39.110 --> 00:05:40.568 align:middle line:84% the epicenter of the movie business 00:05:40.568 --> 00:05:46.250 align:middle line:84% is now no longer in sort of Hollywood, California. 00:05:46.250 --> 00:05:50.970 align:middle line:84% But I think you have both a dynamic hypothesis 00:05:50.970 --> 00:05:53.748 align:middle line:84% kind of as to why, which is it's now cheaper, 00:05:53.748 --> 00:05:56.290 align:middle line:84% and there's tax incentives, and other cities have figured out 00:05:56.290 --> 00:05:57.750 align:middle line:90% this is a good thing. 00:05:57.750 --> 00:06:02.590 align:middle line:84% That doesn't seem that complicated a problem. 00:06:02.590 --> 00:06:05.770 align:middle line:84% But then I think there's a much more interesting question of, 00:06:05.770 --> 00:06:08.537 align:middle line:84% if you're the Chamber of Commerce of Hollywood, 00:06:08.537 --> 00:06:09.870 align:middle line:90% what do you want to do about it? 00:06:09.870 --> 00:06:12.570 align:middle line:84% Or if you're an investor, where should I put my chips? 00:06:12.570 --> 00:06:15.270 align:middle line:84% But I think in a lot of cases, in some of your other ones, 00:06:15.270 --> 00:06:17.370 align:middle line:84% it was just a little unclear sort 00:06:17.370 --> 00:06:19.150 align:middle line:90% of to what end you're directed. 00:06:19.150 --> 00:06:21.530 align:middle line:84% So I would really urge you, and we can go through these 00:06:21.530 --> 00:06:24.810 align:middle line:84% as we go through them to be super clear on either there's 00:06:24.810 --> 00:06:28.370 align:middle line:84% a puzzle I'm trying to solve, so if I go back to my work 00:06:28.370 --> 00:06:30.170 align:middle line:84% a million years ago, everybody knows 00:06:30.170 --> 00:06:32.130 align:middle line:84% product development should be done one way, 00:06:32.130 --> 00:06:33.932 align:middle line:84% and everybody does it some different way. 00:06:33.932 --> 00:06:36.390 align:middle line:84% That just doesn't kind of make sense to first order, right? 00:06:36.390 --> 00:06:40.810 align:middle line:84% It seems a very dysfunctional sort of way of behaving. 00:06:40.810 --> 00:06:42.930 align:middle line:84% Or I have a system where maybe we 00:06:42.930 --> 00:06:46.270 align:middle line:84% have some understanding of why it works, 00:06:46.270 --> 00:06:47.700 align:middle line:84% but people aren't doing the things 00:06:47.700 --> 00:06:48.980 align:middle line:84% that we think that they should do. 00:06:48.980 --> 00:06:50.355 align:middle line:84% So now, we have to figure out how 00:06:50.355 --> 00:06:52.540 align:middle line:90% to put our scarce resources in. 00:06:52.540 --> 00:06:54.460 align:middle line:84% So another way of putting that is, 00:06:54.460 --> 00:06:56.060 align:middle line:84% if your problem statement or model 00:06:56.060 --> 00:07:00.020 align:middle line:84% motivation is, I want to understand the x system, 00:07:00.020 --> 00:07:04.540 align:middle line:84% I think you're already headed down a troubled road. 00:07:04.540 --> 00:07:07.380 align:middle line:84% Now John and I have had this debate for more years 00:07:07.380 --> 00:07:09.860 align:middle line:84% than probably some of you have been alive. 00:07:09.860 --> 00:07:12.900 align:middle line:84% When I started in the field, we used to always say, 00:07:12.900 --> 00:07:15.300 align:middle line:90% we model problems, not systems. 00:07:15.300 --> 00:07:17.480 align:middle line:84% But I will say that I think some of my colleagues, 00:07:17.480 --> 00:07:22.040 align:middle line:84% their work often lapses pretty close to modeling systems, 00:07:22.040 --> 00:07:25.780 align:middle line:84% and they sort of throw in everything. 00:07:25.780 --> 00:07:28.160 align:middle line:84% I'm not sure there's a resolution to that question. 00:07:28.160 --> 00:07:30.260 align:middle line:84% So I just want you to kind of be aware 00:07:30.260 --> 00:07:33.220 align:middle line:84% of the different philosophies that are out there in terms 00:07:33.220 --> 00:07:34.980 align:middle line:90% of thinking about it. 00:07:34.980 --> 00:07:39.700 align:middle line:84% But I think it does often lead to a model that's 00:07:39.700 --> 00:07:42.840 align:middle line:84% very hard to sell, at least from a scholarly point of view, 00:07:42.840 --> 00:07:45.650 align:middle line:84% is, I have built a model of the x system, 00:07:45.650 --> 00:07:48.590 align:middle line:84% and it fits the data from the x system very well. 00:07:48.590 --> 00:07:49.490 align:middle line:90% Full stop. 00:07:49.490 --> 00:07:49.990 align:middle line:90% Right? 00:07:49.990 --> 00:07:55.170 align:middle line:84% And I think the journals tend to be like, OK. 00:07:55.170 --> 00:07:57.190 align:middle line:90% What do I do with that? 00:07:57.190 --> 00:07:59.170 align:middle line:84% Is there implications for this for policy? 00:07:59.170 --> 00:08:01.510 align:middle line:84% Or have we learned something new from a theory? 00:08:01.510 --> 00:08:05.390 align:middle line:84% Or does it challenge our understanding of this? 00:08:05.390 --> 00:08:07.430 align:middle line:84% And I think it's very easy, and the field is 00:08:07.430 --> 00:08:09.390 align:middle line:84% very sloppy in general on this, which 00:08:09.390 --> 00:08:11.550 align:middle line:84% is you go to the conference, half the papers 00:08:11.550 --> 00:08:14.190 align:middle line:90% are system dynamics model of x. 00:08:14.190 --> 00:08:18.590 align:middle line:84% And I would urge you to be very wary of that 00:08:18.590 --> 00:08:21.060 align:middle line:84% because you're going to have a very hard time writing kind 00:08:21.060 --> 00:08:23.310 align:middle line:84% of the discussion and conclusion section of your paper 00:08:23.310 --> 00:08:25.590 align:middle line:84% because, OK, so now I have a model that 00:08:25.590 --> 00:08:28.910 align:middle line:90% replicates the behavior of x. 00:08:28.910 --> 00:08:30.170 align:middle line:90% What are the implications? 00:08:30.170 --> 00:08:33.169 align:middle line:84% Should I either think about this social phenomenon different? 00:08:33.169 --> 00:08:35.090 align:middle line:84% Should we intervene in this differently? 00:08:35.090 --> 00:08:38.169 align:middle line:84% Should we allocate our resources differently, or whatever? 00:08:38.169 --> 00:08:40.635 align:middle line:84% So that was kind of the big thing I got from just kind 00:08:40.635 --> 00:08:42.510 align:middle line:84% of browsing the problem statements is I think 00:08:42.510 --> 00:08:44.260 align:middle line:84% it's-- and you're not alone in this. 00:08:44.260 --> 00:08:46.960 align:middle line:84% I think everybody lapses into this if you're not careful. 00:08:46.960 --> 00:08:49.200 align:middle line:84% But then it makes it very hard to finish your project 00:08:49.200 --> 00:08:53.840 align:middle line:84% because you sort of have no idea what you're going to work on. 00:08:53.840 --> 00:08:54.920 align:middle line:90% So does that make sense? 00:08:54.920 --> 00:08:56.940 align:middle line:84% Or do you have questions or comments about that? 00:08:56.940 --> 00:08:58.420 align:middle line:84% I mean, again, it's a little bit philosophical, 00:08:58.420 --> 00:09:00.140 align:middle line:84% but I think it's quite important. 00:09:00.140 --> 00:09:02.098 align:middle line:84% And I've certainly watched enough students kind 00:09:02.098 --> 00:09:06.240 align:middle line:84% of wander off into enormous models, 00:09:06.240 --> 00:09:08.870 align:middle line:84% and then they're not really quite sure what they have 00:09:08.870 --> 00:09:09.620 align:middle line:90% when they're done. 00:09:09.620 --> 00:09:11.703 align:middle line:84% And there can be a lot of rework to kind of circle 00:09:11.703 --> 00:09:16.920 align:middle line:84% back to have some finding that would be a job market 00:09:16.920 --> 00:09:19.700 align:middle line:84% paper, or a consulting project, or whatever it might be. 00:09:19.700 --> 00:09:20.200 align:middle line:90% Yeah. 00:09:20.200 --> 00:09:20.867 align:middle line:90% Go ahead, Jason. 00:09:20.867 --> 00:09:24.560 align:middle line:84% STUDENT: I think this is hard because you have to decide early 00:09:24.560 --> 00:09:29.200 align:middle line:84% and shut off other potentially interesting avenues that you 00:09:29.200 --> 00:09:31.410 align:middle line:90% might want to hang on to. 00:09:31.410 --> 00:09:33.910 align:middle line:84% Of course, there's a trade off in hanging on to those things 00:09:33.910 --> 00:09:35.440 align:middle line:90% and you might end up nowhere. 00:09:35.440 --> 00:09:36.960 align:middle line:90% NELSON REPENNING: Yep. 00:09:36.960 --> 00:09:41.930 align:middle line:84% STUDENT: So how do you deal with that foreclosure and willingness 00:09:41.930 --> 00:09:42.570 align:middle line:90% to commit? 00:09:42.570 --> 00:09:45.070 align:middle line:84% NELSON REPENNING: So I think this is a really good question. 00:09:45.070 --> 00:09:48.730 align:middle line:84% I thought the other thing that, as I read through the overviews, 00:09:48.730 --> 00:09:51.423 align:middle line:84% that I thought, I don't know if I'd 00:09:51.423 --> 00:09:53.090 align:middle line:84% say it was missing, but could have used, 00:09:53.090 --> 00:09:56.350 align:middle line:84% is I think you all have instincts for your projects 00:09:56.350 --> 00:09:58.512 align:middle line:84% why this is a good dynamic problem to look at. 00:09:58.512 --> 00:09:59.970 align:middle line:84% And so that's why I asked you, what 00:09:59.970 --> 00:10:01.595 align:middle line:84% do you think these models are good for? 00:10:01.595 --> 00:10:05.030 align:middle line:84% So I don't know if we still use this language in class, 00:10:05.030 --> 00:10:07.670 align:middle line:84% but we used to talk a lot about the dynamic hypothesis. 00:10:07.670 --> 00:10:10.665 align:middle line:84% And there's this endless and never-to-be-resolved debate 00:10:10.665 --> 00:10:13.290 align:middle line:84% in the field about whether you start with causal loop diagrams, 00:10:13.290 --> 00:10:14.957 align:middle line:84% or stock-and-flow diagrams, or whatever, 00:10:14.957 --> 00:10:18.530 align:middle line:84% which I think is an enormous waste of time. 00:10:18.530 --> 00:10:21.290 align:middle line:84% But I think there is a kernel of wisdom in there, which 00:10:21.290 --> 00:10:26.050 align:middle line:84% is if you think this is a good area for dynamic modeling, 00:10:26.050 --> 00:10:28.370 align:middle line:84% or a good puzzle or problem, chances 00:10:28.370 --> 00:10:32.130 align:middle line:84% are you have some instinct about why. 00:10:32.130 --> 00:10:34.630 align:middle line:84% You have a dynamic hypothesis underlying that. 00:10:34.630 --> 00:10:37.390 align:middle line:84% And I think in many cases for your proposals, 00:10:37.390 --> 00:10:39.330 align:middle line:84% I think it would have been helpful just 00:10:39.330 --> 00:10:41.553 align:middle line:90% to be more explicit, right? 00:10:41.553 --> 00:10:43.220 align:middle line:84% I think there's a reinforcing feedback-- 00:10:43.220 --> 00:10:44.340 align:middle line:84% I'm going to pick on you, Jake, because I 00:10:44.340 --> 00:10:45.382 align:middle line:90% happen to remember yours. 00:10:45.382 --> 00:10:46.580 align:middle line:90% Sorry. 00:10:46.580 --> 00:10:49.020 align:middle line:84% I think there's a reinforcing feedback loop that's 00:10:49.020 --> 00:10:52.140 align:middle line:84% driving the epicenter of movie making to Vancouver or Atlanta 00:10:52.140 --> 00:10:52.890 align:middle line:90% or whatever it is. 00:10:52.890 --> 00:10:53.390 align:middle line:90% Right? 00:10:53.390 --> 00:10:55.520 align:middle line:84% And that was kind of implicit in what you had. 00:10:55.520 --> 00:10:58.780 align:middle line:84% But I think it's much better, and actually better science, 00:10:58.780 --> 00:11:03.260 align:middle line:84% to state, this is my hypothesis about what's out there, 00:11:03.260 --> 00:11:06.260 align:middle line:84% rather than a kind of much more amorphous, 00:11:06.260 --> 00:11:09.700 align:middle line:84% I'm going to build a model about the x system. 00:11:09.700 --> 00:11:14.118 align:middle line:84% So I think, Jason, to answer your question, 00:11:14.118 --> 00:11:15.660 align:middle line:84% the first answer is, I think you just 00:11:15.660 --> 00:11:18.822 align:middle line:84% have to be kind of honest with yourself about your priors. 00:11:18.822 --> 00:11:21.280 align:middle line:84% And whether it's a causal loop diagram or stocks and flows, 00:11:21.280 --> 00:11:23.540 align:middle line:84% I don't actually really care about it. 00:11:23.540 --> 00:11:27.763 align:middle line:84% But you have some instinct that this is a good thing to work on. 00:11:27.763 --> 00:11:29.180 align:middle line:84% And then I think, as you all know, 00:11:29.180 --> 00:11:33.700 align:middle line:84% but it can't be reinforced, it pays to proceed incrementally. 00:11:33.700 --> 00:11:36.200 align:middle line:84% So you build a little model, you test it, and understand it, 00:11:36.200 --> 00:11:38.890 align:middle line:84% you add, so you want to be agile or whatever flavor. 00:11:38.890 --> 00:11:42.510 align:middle line:84% And so yeah, you have to make trade offs about what 00:11:42.510 --> 00:11:45.350 align:middle line:84% is the next most interesting feedback, or structure, 00:11:45.350 --> 00:11:48.190 align:middle line:90% or disaggregation to do. 00:11:48.190 --> 00:11:49.750 align:middle line:84% And that's always a judgment call 00:11:49.750 --> 00:11:52.890 align:middle line:84% in terms kind of model complexity. 00:11:52.890 --> 00:11:54.990 align:middle line:84% I always tease John, he's the patron saint 00:11:54.990 --> 00:11:56.210 align:middle line:90% of the broad model boundary. 00:11:56.210 --> 00:12:00.470 align:middle line:84% He never met a feedback loop he didn't want to add to a model. 00:12:00.470 --> 00:12:02.530 align:middle line:84% I tend to be kind of the opposite. 00:12:02.530 --> 00:12:05.008 align:middle line:84% I'd much prefer the sort of much smaller, more 00:12:05.008 --> 00:12:06.050 align:middle line:90% aggregate representation. 00:12:06.050 --> 00:12:08.550 align:middle line:84% I don't think there's a right or wrong sort of answer there. 00:12:08.550 --> 00:12:11.950 align:middle line:84% But you do have to make choices about what's 00:12:11.950 --> 00:12:16.437 align:middle line:84% the next thing to go at, or the next disaggregation to work on. 00:12:16.437 --> 00:12:17.270 align:middle line:90% And it's just-- yep. 00:12:17.270 --> 00:12:18.988 align:middle line:90% STUDENT: [INAUDIBLE] 00:12:18.988 --> 00:12:21.030 align:middle line:84% NELSON REPENNING: There's no algorithm that says, 00:12:21.030 --> 00:12:22.250 align:middle line:90% this is the next feedback loop. 00:12:22.250 --> 00:12:23.890 align:middle line:84% And that's kind of where the art of this. 00:12:23.890 --> 00:12:25.848 align:middle line:84% And I'm not sure even AI is ever going to solve 00:12:25.848 --> 00:12:28.430 align:middle line:90% that particular problem. 00:12:28.430 --> 00:12:29.270 align:middle line:90% STUDENT: Thank you. 00:12:29.270 --> 00:12:30.810 align:middle line:84% NELSON REPENNING: So other thoughts or comments? 00:12:30.810 --> 00:12:32.530 align:middle line:84% I actually think this is a really important point. 00:12:32.530 --> 00:12:34.863 align:middle line:84% So I want to kind of make sure it's seared in your heads 00:12:34.863 --> 00:12:38.060 align:middle line:84% before we jump into equations and details. 00:12:38.060 --> 00:12:38.640 align:middle line:90% Yeah. 00:12:38.640 --> 00:12:42.600 align:middle line:84% STUDENT: Yeah, I think I have a problem definition that I 00:12:42.600 --> 00:12:44.340 align:middle line:90% thought was pretty like-- 00:12:44.340 --> 00:12:46.898 align:middle line:90% 00:12:46.898 --> 00:12:47.440 align:middle line:90% I don't know. 00:12:47.440 --> 00:12:50.080 align:middle line:84% It was pretty, like, well designed in my mind 00:12:50.080 --> 00:12:51.100 align:middle line:90% when I wrote it down. 00:12:51.100 --> 00:12:54.080 align:middle line:84% And then I found myself more so than with the problem 00:12:54.080 --> 00:12:56.760 align:middle line:84% definitions that we've got in the past 00:12:56.760 --> 00:13:00.600 align:middle line:84% from faculty, where it's very obvious what 00:13:00.600 --> 00:13:02.180 align:middle line:90% the structures should be. 00:13:02.180 --> 00:13:06.400 align:middle line:84% I found myself with the blank-page syndrome, 00:13:06.400 --> 00:13:08.060 align:middle line:90% with my own problem definition. 00:13:08.060 --> 00:13:11.920 align:middle line:84% And but I couldn't identify what was wrong with it 00:13:11.920 --> 00:13:14.480 align:middle line:90% that was making it so-- 00:13:14.480 --> 00:13:17.960 align:middle line:84% I don't know if it's the case that it's because I made it 00:13:17.960 --> 00:13:21.160 align:middle line:84% that I'm kind of getting stuck versus when I see one 00:13:21.160 --> 00:13:24.560 align:middle line:84% that someone else made, it's easier to look at it a little 00:13:24.560 --> 00:13:27.580 align:middle line:90% differently or what. 00:13:27.580 --> 00:13:28.340 align:middle line:90% But I don't know. 00:13:28.340 --> 00:13:30.920 align:middle line:84% I had a lot of trouble kind of deciding 00:13:30.920 --> 00:13:34.850 align:middle line:84% what structures should go in the model that I proposed. 00:13:34.850 --> 00:13:36.950 align:middle line:84% NELSON REPENNING: OK, so let's start with yours. 00:13:36.950 --> 00:13:40.750 align:middle line:84% So you're going to regret probably speaking in class. 00:13:40.750 --> 00:13:43.630 align:middle line:84% But so tell us about your problem. 00:13:43.630 --> 00:13:45.450 align:middle line:90% What's the area? 00:13:45.450 --> 00:13:49.350 align:middle line:84% STUDENT: OK, so it's kind of the puzzle idea. 00:13:49.350 --> 00:13:52.770 align:middle line:84% The puzzle is that book sales have really 00:13:52.770 --> 00:13:56.390 align:middle line:84% increased a lot after many kind of years of decline. 00:13:56.390 --> 00:14:00.690 align:middle line:84% They increased a lot starting around 2019, 2020. 00:14:00.690 --> 00:14:03.370 align:middle line:84% However, reading rates, like the number 00:14:03.370 --> 00:14:05.690 align:middle line:84% of adults who have read a book in the last year, 00:14:05.690 --> 00:14:09.170 align:middle line:84% and the number of books that adults are reading 00:14:09.170 --> 00:14:12.190 align:middle line:90% have both stayed on decline. 00:14:12.190 --> 00:14:15.705 align:middle line:84% And one of the hypotheses that I had was-- 00:14:15.705 --> 00:14:16.830 align:middle line:90% NELSON REPENNING: Time out. 00:14:16.830 --> 00:14:19.430 align:middle line:84% So just before we get to the explanation, 00:14:19.430 --> 00:14:21.650 align:middle line:84% so just in one sentence, the puzzle 00:14:21.650 --> 00:14:23.930 align:middle line:84% that you're trying to resolve is? 00:14:23.930 --> 00:14:25.990 align:middle line:84% STUDENT: Book sales up, reading down. 00:14:25.990 --> 00:14:28.450 align:middle line:84% NELSON REPENNING: OK, and the kind of casual intuition 00:14:28.450 --> 00:14:31.427 align:middle line:84% would be that the more books we would sell, right? 00:14:31.427 --> 00:14:33.260 align:middle line:84% So we wouldn't expect those two graphs to go 00:14:33.260 --> 00:14:35.640 align:middle line:90% in the opposite direction. 00:14:35.640 --> 00:14:36.140 align:middle line:90% OK? 00:14:36.140 --> 00:14:39.260 align:middle line:90% 00:14:39.260 --> 00:14:40.800 align:middle line:84% This is a little bit of a tangent, 00:14:40.800 --> 00:14:41.820 align:middle line:84% but it would be relevant if you wanted 00:14:41.820 --> 00:14:42.903 align:middle line:90% to turn this into a paper. 00:14:42.903 --> 00:14:46.820 align:middle line:84% I mean, is that a puzzle that other literature acknowledges 00:14:46.820 --> 00:14:47.320 align:middle line:90% as such? 00:14:47.320 --> 00:14:48.860 align:middle line:84% Or is this something that you've come across? 00:14:48.860 --> 00:14:49.420 align:middle line:90% Or-- 00:14:49.420 --> 00:14:51.212 align:middle line:84% STUDENT: It's something that I came across. 00:14:51.212 --> 00:14:54.940 align:middle line:84% I think it's relatively new phenomenon. 00:14:54.940 --> 00:14:59.260 align:middle line:90% Like, from 2020, 2021. 00:14:59.260 --> 00:15:04.240 align:middle line:84% And no, I haven't seen it in other literature. 00:15:04.240 --> 00:15:06.840 align:middle line:84% I've seen people, I mean, people are talking a lot. 00:15:06.840 --> 00:15:12.300 align:middle line:84% I think where I kind of found it, or came up with it, 00:15:12.300 --> 00:15:15.100 align:middle line:84% is people are talking a lot about the crisis of literacy, 00:15:15.100 --> 00:15:16.820 align:middle line:90% and how people are not reading. 00:15:16.820 --> 00:15:19.700 align:middle line:84% And then I had remembered reading headlines about, 00:15:19.700 --> 00:15:23.960 align:middle line:84% oh, Barnes and Noble, like, was going to shut down, 00:15:23.960 --> 00:15:25.660 align:middle line:84% but suddenly they've had a resurgence, 00:15:25.660 --> 00:15:29.020 align:middle line:84% and they're opening new stores, and things are looking up. 00:15:29.020 --> 00:15:30.820 align:middle line:90% And then. 00:15:30.820 --> 00:15:32.840 align:middle line:84% I thought, like, oh, that's interesting. 00:15:32.840 --> 00:15:37.100 align:middle line:84% It's not clear why Barnes and Noble is suddenly 00:15:37.100 --> 00:15:40.080 align:middle line:84% doing so great in the last few years, 00:15:40.080 --> 00:15:42.825 align:middle line:84% and literacy rates among adults are plummeting. 00:15:42.825 --> 00:15:43.700 align:middle line:90% NELSON REPENNING: OK. 00:15:43.700 --> 00:15:44.283 align:middle line:90% STUDENT: Yeah. 00:15:44.283 --> 00:15:45.960 align:middle line:90% NELSON REPENNING: All right. 00:15:45.960 --> 00:15:48.220 align:middle line:84% Did those two things move together 00:15:48.220 --> 00:15:50.520 align:middle line:84% at some point in earlier history? 00:15:50.520 --> 00:15:53.900 align:middle line:84% STUDENT: Yeah, they've moved together until, like, 00:15:53.900 --> 00:15:55.520 align:middle line:90% the last four or five years. 00:15:55.520 --> 00:15:59.820 align:middle line:84% NELSON REPENNING: OK, so over what time horizon? 00:15:59.820 --> 00:16:00.520 align:middle line:90% I mean, roughly. 00:16:00.520 --> 00:16:02.620 align:middle line:84% Is it the last 50 years or the last hundred years? 00:16:02.620 --> 00:16:03.260 align:middle line:90% Or-- 00:16:03.260 --> 00:16:05.120 align:middle line:90% STUDENT: No, nothing like that. 00:16:05.120 --> 00:16:06.800 align:middle line:90% I think 30. 00:16:06.800 --> 00:16:11.380 align:middle line:84% I think the data that I have on reading goes back 00:16:11.380 --> 00:16:16.380 align:middle line:84% to the '80s, but it's pretty sparse, like every few years. 00:16:16.380 --> 00:16:20.640 align:middle line:84% And then the book sales data is annual, but not as far back. 00:16:20.640 --> 00:16:24.398 align:middle line:90% I think I have 20 years of-- 00:16:24.398 --> 00:16:25.940 align:middle line:84% NELSON REPENNING: OK, I mean, I think 00:16:25.940 --> 00:16:28.527 align:middle line:84% that's a perfectly fine problem statement, right? 00:16:28.527 --> 00:16:31.110 align:middle line:84% We have two things that used to go together, that one of which 00:16:31.110 --> 00:16:35.408 align:middle line:84% seems quite important towards polite and civic society. 00:16:35.408 --> 00:16:36.950 align:middle line:84% And now, they've diverged for reasons 00:16:36.950 --> 00:16:38.075 align:middle line:90% that aren't entirely clear. 00:16:38.075 --> 00:16:39.932 align:middle line:84% OK, I mean, I think that's a great setup. 00:16:39.932 --> 00:16:41.390 align:middle line:84% I mean, there's a separate question 00:16:41.390 --> 00:16:43.265 align:middle line:84% we can talk about offline about if you wanted 00:16:43.265 --> 00:16:46.350 align:middle line:84% to publish this in a journal, making links 00:16:46.350 --> 00:16:49.203 align:middle line:84% to broader phenomena, things like that, 00:16:49.203 --> 00:16:50.870 align:middle line:84% and connecting to the value of literacy, 00:16:50.870 --> 00:16:51.570 align:middle line:90% and so on, and so forth. 00:16:51.570 --> 00:16:53.990 align:middle line:84% But I think for a kind of initial modeling project, that 00:16:53.990 --> 00:16:54.730 align:middle line:90% seems fine. 00:16:54.730 --> 00:16:56.470 align:middle line:84% STUDENT: And you say, connecting to what? 00:16:56.470 --> 00:16:57.928 align:middle line:84% NELSON REPENNING: I mean, I think-- 00:16:57.928 --> 00:17:01.230 align:middle line:90% you might want to move down. 00:17:01.230 --> 00:17:04.589 align:middle line:84% I mean, I think if you want to publish it, 00:17:04.589 --> 00:17:07.710 align:middle line:84% you might need to connect it to sort of bigger questions 00:17:07.710 --> 00:17:12.319 align:middle line:84% that the journal of choice cared about. 00:17:12.319 --> 00:17:14.069 align:middle line:84% And in that case, I think there's probably 00:17:14.069 --> 00:17:16.510 align:middle line:84% a variety of different directions you could go. 00:17:16.510 --> 00:17:19.329 align:middle line:84% We could spend a few minutes talking about that as well. 00:17:19.329 --> 00:17:23.187 align:middle line:84% I mean, it doesn't obviously scream to me 00:17:23.187 --> 00:17:25.270 align:middle line:84% at least that there's one literature in which that 00:17:25.270 --> 00:17:25.932 align:middle line:90% would go in. 00:17:25.932 --> 00:17:27.640 align:middle line:84% But literacy is something that's probably 00:17:27.640 --> 00:17:31.826 align:middle line:84% important to a variety of constituencies. 00:17:31.826 --> 00:17:34.038 align:middle line:90% Or you could-- 00:17:34.038 --> 00:17:35.580 align:middle line:84% I'm making this up now as I go along. 00:17:35.580 --> 00:17:37.800 align:middle line:84% I mean, it could be a strategy paper. 00:17:37.800 --> 00:17:41.560 align:middle line:84% If I thought that we normally believe that as more people use 00:17:41.560 --> 00:17:44.240 align:middle line:84% a product, demand for that product goes up, 00:17:44.240 --> 00:17:45.780 align:middle line:84% but in fact, that's not the case, 00:17:45.780 --> 00:17:47.608 align:middle line:84% but yet it seems to be good for business. 00:17:47.608 --> 00:17:49.400 align:middle line:84% This would be an important thing for people 00:17:49.400 --> 00:17:50.760 align:middle line:90% to know on the strategy side. 00:17:50.760 --> 00:17:54.240 align:middle line:84% Or it could be more publicly focused. 00:17:54.240 --> 00:18:00.080 align:middle line:84% STUDENT: There's a story related to one of my hypotheses 00:18:00.080 --> 00:18:04.560 align:middle line:84% around, like, exaptation, like the purpose of buying 00:18:04.560 --> 00:18:08.440 align:middle line:90% and consuming goods has changed. 00:18:08.440 --> 00:18:13.660 align:middle line:84% I think my first brush with, oh, why is this happening? 00:18:13.660 --> 00:18:17.920 align:middle line:84% I was thinking about how social media has kind of blown up 00:18:17.920 --> 00:18:22.520 align:middle line:84% a lot of books online, and people 00:18:22.520 --> 00:18:25.310 align:middle line:84% flock towards single titles by them 00:18:25.310 --> 00:18:27.170 align:middle line:90% and then talk about them online. 00:18:27.170 --> 00:18:35.290 align:middle line:84% [INAUDIBLE] that don't-- so then sales are inflated, 00:18:35.290 --> 00:18:37.470 align:middle line:84% and things look good for bookstores, 00:18:37.470 --> 00:18:41.130 align:middle line:84% but people aren't really reading a variety or widely vary 00:18:41.130 --> 00:18:45.250 align:middle line:84% in the books that they're talking about. 00:18:45.250 --> 00:18:47.730 align:middle line:84% It's more so like a social signaling thing. 00:18:47.730 --> 00:18:51.550 align:middle line:84% But that was what I was like, oh, I don't know about all 00:18:51.550 --> 00:18:52.050 align:middle line:90% of that. 00:18:52.050 --> 00:18:52.530 align:middle line:90% But-- 00:18:52.530 --> 00:18:54.322 align:middle line:84% NELSON REPENNING: Well, so without worrying 00:18:54.322 --> 00:18:56.190 align:middle line:84% about how to formulate it, right, 00:18:56.190 --> 00:18:58.790 align:middle line:84% you had the seeds of a dynamic hypothesis there. 00:18:58.790 --> 00:19:02.530 align:middle line:84% So just take it three more sentences in terms of, 00:19:02.530 --> 00:19:03.690 align:middle line:90% what's the story? 00:19:03.690 --> 00:19:04.990 align:middle line:90% I mean, you had all the pieces. 00:19:04.990 --> 00:19:06.970 align:middle line:84% So just link them together a little bit. 00:19:06.970 --> 00:19:09.210 align:middle line:84% STUDENT: Yeah, so I think the story is 00:19:09.210 --> 00:19:13.770 align:middle line:84% that you get two kinds of capital, 00:19:13.770 --> 00:19:16.150 align:middle line:84% maybe, from buying a book, right? 00:19:16.150 --> 00:19:18.530 align:middle line:84% You get the human capital from reading 00:19:18.530 --> 00:19:22.520 align:middle line:84% the book, just the entertainment, knowledge, 00:19:22.520 --> 00:19:23.840 align:middle line:90% and et cetera that you have. 00:19:23.840 --> 00:19:25.340 align:middle line:84% And then you have the social capital 00:19:25.340 --> 00:19:27.700 align:middle line:84% that you get from signaling, I'm reading this, 00:19:27.700 --> 00:19:30.240 align:middle line:84% this says something about me that I'm reading this, 00:19:30.240 --> 00:19:34.300 align:middle line:84% and I'm going to show everybody, oh, I care about this topic, 00:19:34.300 --> 00:19:37.160 align:middle line:84% or and am knowledgeable in this area. 00:19:37.160 --> 00:19:40.460 align:middle line:90% 00:19:40.460 --> 00:19:45.820 align:middle line:84% So my hypothesis is that, on balance, the social capital has 00:19:45.820 --> 00:19:47.840 align:middle line:90% outweighed the human capital. 00:19:47.840 --> 00:19:51.940 align:middle line:84% And so it's more valuable than it used to be, 00:19:51.940 --> 00:19:55.220 align:middle line:84% relatively, to own and signal the book 00:19:55.220 --> 00:19:57.940 align:middle line:84% than it is to have read the book because you 00:19:57.940 --> 00:20:01.620 align:middle line:84% can get the CliffsNotes version pretty easily, 00:20:01.620 --> 00:20:08.420 align:middle line:84% and from a wide variety of mostly social media sources 00:20:08.420 --> 00:20:11.685 align:middle line:84% to kind of don't need to read it to get the big picture. 00:20:11.685 --> 00:20:13.060 align:middle line:84% Of course, you would need to read 00:20:13.060 --> 00:20:15.720 align:middle line:84% it to get the little details, but those don't really matter 00:20:15.720 --> 00:20:20.190 align:middle line:84% and aren't really valued in the same way. 00:20:20.190 --> 00:20:23.910 align:middle line:84% And so yeah, the decoupling happens 00:20:23.910 --> 00:20:30.110 align:middle line:84% when the sales are driven by people 00:20:30.110 --> 00:20:37.630 align:middle line:84% who want to have the social signaling and not the context. 00:20:37.630 --> 00:20:40.880 align:middle line:84% NELSON REPENNING: OK, that sounds great. 00:20:40.880 --> 00:20:42.630 align:middle line:84% STUDENT: So then an alternative hypothesis 00:20:42.630 --> 00:20:44.290 align:middle line:90% is that there's a heterogeneity. 00:20:44.290 --> 00:20:49.590 align:middle line:84% So that people have kind of diverged into the readers 00:20:49.590 --> 00:20:50.770 align:middle line:90% and the super readers. 00:20:50.770 --> 00:20:55.270 align:middle line:84% The super readers are really, really high volume now more 00:20:55.270 --> 00:20:57.830 align:middle line:90% than they used to be. 00:20:57.830 --> 00:21:00.230 align:middle line:84% And so there are people who are reading, like, 00:21:00.230 --> 00:21:02.050 align:middle line:90% a hundred or 150 books a year. 00:21:02.050 --> 00:21:06.140 align:middle line:84% Largely, they're reading ebooks and audiobooks 00:21:06.140 --> 00:21:08.770 align:middle line:84% just because it's cheaper and easier to do it that way. 00:21:08.770 --> 00:21:13.670 align:middle line:84% So the people who are super readers are classified as such, 00:21:13.670 --> 00:21:17.920 align:middle line:90% are tipping the scales of sales. 00:21:17.920 --> 00:21:21.260 align:middle line:84% But then the kind of middle is being hollowed out. 00:21:21.260 --> 00:21:24.680 align:middle line:84% And a lot more people are on readers or light readers 00:21:24.680 --> 00:21:25.400 align:middle line:90% as well. 00:21:25.400 --> 00:21:27.137 align:middle line:90% NELSON REPENNING: OK, great. 00:21:27.137 --> 00:21:27.720 align:middle line:90% STUDENT: Yeah. 00:21:27.720 --> 00:21:28.540 align:middle line:84% NELSON REPENNING: Please, go ahead. 00:21:28.540 --> 00:21:29.600 align:middle line:84% STUDENT: What I don't understand, 00:21:29.600 --> 00:21:31.338 align:middle line:84% and this is a model about signaling, 00:21:31.338 --> 00:21:33.880 align:middle line:84% is why do you really have to buy the book if you don't really 00:21:33.880 --> 00:21:35.040 align:middle line:90% need to read it? 00:21:35.040 --> 00:21:38.380 align:middle line:84% You don't really need to buy the book for the signaling. 00:21:38.380 --> 00:21:42.200 align:middle line:84% You can just participate and you don't need to buy. 00:21:42.200 --> 00:21:43.680 align:middle line:90% STUDENT: Yeah, I think-- 00:21:43.680 --> 00:21:46.640 align:middle line:90% that's a good point. 00:21:46.640 --> 00:21:49.840 align:middle line:84% On the one hand, you can just totally 00:21:49.840 --> 00:21:51.520 align:middle line:84% put a picture of the book and say, 00:21:51.520 --> 00:21:53.160 align:middle line:90% oh my gosh, I really like this. 00:21:53.160 --> 00:21:56.400 align:middle line:84% On the other hand, I see a lot of content 00:21:56.400 --> 00:21:58.520 align:middle line:84% where the person's like, listening. 00:21:58.520 --> 00:21:59.680 align:middle line:90% STUDENT: OK, OK. 00:21:59.680 --> 00:22:03.480 align:middle line:90% STUDENT: So they have to-- 00:22:03.480 --> 00:22:04.020 align:middle line:90% but yeah. 00:22:04.020 --> 00:22:06.200 align:middle line:84% I think that's a fair point and leans 00:22:06.200 --> 00:22:09.880 align:middle line:84% more into the heterogeneity diverging populations 00:22:09.880 --> 00:22:14.920 align:middle line:84% theory rather than the kind of social signaling theory. 00:22:14.920 --> 00:22:18.650 align:middle line:84% NELSON REPENNING: OK, and so then, where did you get stuck? 00:22:18.650 --> 00:22:23.130 align:middle line:84% STUDENT: I think kind of, how do I model these, like, 00:22:23.130 --> 00:22:25.170 align:middle line:90% sources of capital? 00:22:25.170 --> 00:22:27.010 align:middle line:84% Social capital and human capital, 00:22:27.010 --> 00:22:35.890 align:middle line:84% how they change over time, whether people, I don't know. 00:22:35.890 --> 00:22:37.670 align:middle line:90% How are people perceiving that? 00:22:37.670 --> 00:22:39.587 align:middle line:84% And how are they making a decision really 00:22:39.587 --> 00:22:41.170 align:middle line:84% based on what they're going to get out 00:22:41.170 --> 00:22:45.090 align:middle line:84% of the book and these components? 00:22:45.090 --> 00:22:51.655 align:middle line:84% And yeah, I just, I got a little stuck on that thought, 00:22:51.655 --> 00:22:52.530 align:middle line:90% NELSON REPENNING: OK. 00:22:52.530 --> 00:22:53.850 align:middle line:84% So I think here's the trick, and I'm 00:22:53.850 --> 00:22:55.767 align:middle line:84% going to ask everyone to help you in a second, 00:22:55.767 --> 00:22:59.050 align:middle line:84% is like, we could probably cover the board 00:22:59.050 --> 00:23:01.630 align:middle line:90% with different causal diagrams. 00:23:01.630 --> 00:23:04.050 align:middle line:84% So kind of a little bit of the art of this 00:23:04.050 --> 00:23:10.050 align:middle line:84% is to consolidate this down to the most bare-bones formulation 00:23:10.050 --> 00:23:12.830 align:middle line:90% to kind of get started. 00:23:12.830 --> 00:23:16.500 align:middle line:84% And then you can add the complexity as you go. 00:23:16.500 --> 00:23:20.660 align:middle line:84% And so I think often, you see this-- 00:23:20.660 --> 00:23:23.200 align:middle line:84% it's rare that I give positive feedback to economists. 00:23:23.200 --> 00:23:24.500 align:middle line:84% But economists are very good at this part 00:23:24.500 --> 00:23:25.917 align:middle line:84% actually because they rely on sort 00:23:25.917 --> 00:23:28.580 align:middle line:84% of closed-form models, which is they'll strip a situation down 00:23:28.580 --> 00:23:31.527 align:middle line:90% to its absolute essence. 00:23:31.527 --> 00:23:33.860 align:middle line:84% But then I think, you can get a really complete analysis 00:23:33.860 --> 00:23:34.440 align:middle line:90% of that. 00:23:34.440 --> 00:23:37.925 align:middle line:84% So if you were going to, please help Cathy out. 00:23:37.925 --> 00:23:39.300 align:middle line:84% If we're going to strip this down 00:23:39.300 --> 00:23:41.540 align:middle line:84% to its really essential elements to get 00:23:41.540 --> 00:23:43.377 align:middle line:84% sort of a first formulation, I think 00:23:43.377 --> 00:23:45.960 align:middle line:84% you've actually given us most of the breadcrumbs that we need. 00:23:45.960 --> 00:23:49.563 align:middle line:84% But well, I'll ask you first, but others can chime in. 00:23:49.563 --> 00:23:50.480 align:middle line:90% Where would you start? 00:23:50.480 --> 00:23:54.100 align:middle line:90% 00:23:54.100 --> 00:23:57.580 align:middle line:84% STUDENT: Well, I think I was starting with, 00:23:57.580 --> 00:23:58.600 align:middle line:90% what are the stocks? 00:23:58.600 --> 00:24:04.420 align:middle line:84% The stocks are different kinds of populations of readers. 00:24:04.420 --> 00:24:08.220 align:middle line:84% So I was just dividing like non-readers, regular readers, 00:24:08.220 --> 00:24:11.500 align:middle line:90% and super readers. 00:24:11.500 --> 00:24:17.830 align:middle line:84% And then laying down how many books each of those populations 00:24:17.830 --> 00:24:20.430 align:middle line:84% buy, how many books each of those populations 00:24:20.430 --> 00:24:28.120 align:middle line:90% read, and then kind of the-- 00:24:28.120 --> 00:24:30.190 align:middle line:84% I don't know, the marginal value. 00:24:30.190 --> 00:24:38.390 align:middle line:84% I guess I haven't thought about it-- of reading versus buying, 00:24:38.390 --> 00:24:40.990 align:middle line:90% and then kind of divergent-- 00:24:40.990 --> 00:24:42.865 align:middle line:90% the divergent ability of this. 00:24:42.865 --> 00:24:43.990 align:middle line:90% NELSON REPENNING: Yeah, OK. 00:24:43.990 --> 00:24:46.338 align:middle line:84% I would say as a piece of advice, 00:24:46.338 --> 00:24:48.130 align:middle line:84% and I noticed in yours too, Jake, actually, 00:24:48.130 --> 00:24:51.672 align:middle line:84% is that when you are thinking of a model that you might index, 00:24:51.672 --> 00:24:53.130 align:middle line:84% we're going to have multiple types. 00:24:53.130 --> 00:24:55.230 align:middle line:84% So I think in yours, you had multiple cities 00:24:55.230 --> 00:24:59.510 align:middle line:84% that might become the epicenter of Los Angeles, and Vancouver, 00:24:59.510 --> 00:25:03.170 align:middle line:84% and Atlanta, I often would recommend at the beginning, 00:25:03.170 --> 00:25:04.710 align:middle line:90% just go with two. 00:25:04.710 --> 00:25:06.870 align:middle line:84% Just build the model with two types, 00:25:06.870 --> 00:25:09.555 align:middle line:84% and don't use the indexing function in Vensim 00:25:09.555 --> 00:25:11.722 align:middle line:84% because I think it's easy to get lost in what's sort 00:25:11.722 --> 00:25:13.642 align:middle line:90% of happening behind the screen. 00:25:13.642 --> 00:25:15.850 align:middle line:84% And once you've worked out with two, so in your case, 00:25:15.850 --> 00:25:18.200 align:middle line:84% you could be like, we have LA and the rest of the world, 00:25:18.200 --> 00:25:19.950 align:middle line:84% and you're lumping all these cities, which 00:25:19.950 --> 00:25:22.230 align:middle line:84% of course, has some nuance, and you're aggregating. 00:25:22.230 --> 00:25:24.110 align:middle line:84% But that way, I think you can kind of fully 00:25:24.110 --> 00:25:26.330 align:middle line:84% understand the dynamics with the simpler formulation, 00:25:26.330 --> 00:25:28.270 align:middle line:84% and then going back and indexing to multiple 00:25:28.270 --> 00:25:29.450 align:middle line:90% would be much, much easier. 00:25:29.450 --> 00:25:32.310 align:middle line:84% So in your case, Cathy, I think if you're going 00:25:32.310 --> 00:25:36.410 align:middle line:90% to go with types, just two. 00:25:36.410 --> 00:25:38.857 align:middle line:84% But actually, I'm not sure I would start with the stocks. 00:25:38.857 --> 00:25:40.690 align:middle line:84% I mean, you would need to get there quickly. 00:25:40.690 --> 00:25:44.430 align:middle line:84% But do others have thoughts on how do we strip this down 00:25:44.430 --> 00:25:46.583 align:middle line:90% to the essential formulation? 00:25:46.583 --> 00:25:48.750 align:middle line:84% Because I think there is a standard formulation that 00:25:48.750 --> 00:25:50.430 align:middle line:84% jumps out of this, or at least part of one, 00:25:50.430 --> 00:25:51.570 align:middle line:90% that could help you get started. 00:25:51.570 --> 00:25:52.737 align:middle line:90% But I don't know if others-- 00:25:52.737 --> 00:25:55.735 align:middle line:90% 00:25:55.735 --> 00:25:57.110 align:middle line:84% were you going to jump in, Jason? 00:25:57.110 --> 00:26:01.255 align:middle line:90% 00:26:01.255 --> 00:26:03.630 align:middle line:84% STUDENT: It was more of a question, how committed are you 00:26:03.630 --> 00:26:07.230 align:middle line:84% to the role of social capital versus human capital? 00:26:07.230 --> 00:26:09.660 align:middle line:84% STUDENT: I just think that that's all, my hypothesis. 00:26:09.660 --> 00:26:10.160 align:middle line:90% But-- 00:26:10.160 --> 00:26:17.920 align:middle line:90% 00:26:17.920 --> 00:26:21.635 align:middle line:84% STUDENT: I don't think [INAUDIBLE]. 00:26:21.635 --> 00:26:23.760 align:middle line:84% NELSON REPENNING: So I think the place that I would 00:26:23.760 --> 00:26:26.480 align:middle line:84% be very tempted to start is just try 00:26:26.480 --> 00:26:29.772 align:middle line:84% to formulate sort of the demand for books 00:26:29.772 --> 00:26:31.480 align:middle line:84% with the sort of-- because what jumps out 00:26:31.480 --> 00:26:34.280 align:middle line:84% to me is that the attractiveness, or the market 00:26:34.280 --> 00:26:37.920 align:middle line:84% share of reading, or the appeal of a particular book, 00:26:37.920 --> 00:26:42.400 align:middle line:84% is changing because in some previous epoch 00:26:42.400 --> 00:26:46.000 align:middle line:84% prior to social media, you don't really need to model this. 00:26:46.000 --> 00:26:47.880 align:middle line:84% You would just have some baseline demand 00:26:47.880 --> 00:26:52.958 align:middle line:84% that any given book would have a certain appeal to me. 00:26:52.958 --> 00:26:55.000 align:middle line:84% But then you've added this sort of second element 00:26:55.000 --> 00:26:56.680 align:middle line:84% to the attractiveness function of, 00:26:56.680 --> 00:26:58.888 align:middle line:84% oh, this is what my friends are reading, or this is-- 00:26:58.888 --> 00:27:02.340 align:middle line:84% and then you could formulate, I think, 00:27:02.340 --> 00:27:04.720 align:middle line:84% your kind of first reinforcing feedback loop that 00:27:04.720 --> 00:27:08.053 align:middle line:90% would be of the flavor. 00:27:08.053 --> 00:27:09.970 align:middle line:84% And there's a sort of tricky aggregation thing 00:27:09.970 --> 00:27:14.250 align:middle line:84% here that I have to think about a little bit in terms of you're 00:27:14.250 --> 00:27:19.770 align:middle line:84% obviously not going to model the whole universe of books. 00:27:19.770 --> 00:27:23.516 align:middle line:84% I have to think a little bit about how to handle that. 00:27:23.516 --> 00:27:27.200 align:middle line:84% All right, because what you're saying is that some books are-- 00:27:27.200 --> 00:27:28.950 align:middle line:84% well, actually, I guess you don't have to. 00:27:28.950 --> 00:27:31.530 align:middle line:84% You can just model the aggregate demand for books. 00:27:31.530 --> 00:27:33.639 align:middle line:84% So you don't have to worry about-- 00:27:33.639 --> 00:27:34.222 align:middle line:90% STUDENT: Good. 00:27:34.222 --> 00:27:35.040 align:middle line:90% Yeah. 00:27:35.040 --> 00:27:38.770 align:middle line:84% I was thinking in my original problem definition I had, 00:27:38.770 --> 00:27:42.710 align:middle line:84% and I think this exists in the social media hypothesis, 00:27:42.710 --> 00:27:46.970 align:middle line:84% there is a demand for a certain type 00:27:46.970 --> 00:27:51.050 align:middle line:84% of book, a very important [? wave ?] type of book. 00:27:51.050 --> 00:27:55.510 align:middle line:84% And I think that authors can produce to that demand. 00:27:55.510 --> 00:28:00.690 align:middle line:84% And I think self-publishing has boomed, like alternative, 00:28:00.690 --> 00:28:03.490 align:middle line:84% outside of traditional publishing avenues and offer 00:28:03.490 --> 00:28:08.300 align:middle line:84% this, just like online, [INAUDIBLE] stuff, and then 00:28:08.300 --> 00:28:12.620 align:middle line:84% also, self-publishing, and selling directly through Amazon, 00:28:12.620 --> 00:28:14.140 align:middle line:90% or something like that. 00:28:14.140 --> 00:28:18.700 align:middle line:84% And so the quality and variety of what's produced 00:28:18.700 --> 00:28:20.800 align:middle line:90% is changing, as well. 00:28:20.800 --> 00:28:25.260 align:middle line:84% But I think [INAUDIBLE] probably too 00:28:25.260 --> 00:28:27.388 align:middle line:84% much to start with the first thing. 00:28:27.388 --> 00:28:28.680 align:middle line:90% NELSON REPENNING: Well-- sorry. 00:28:28.680 --> 00:28:29.180 align:middle line:90% Go ahead. 00:28:29.180 --> 00:28:30.560 align:middle line:90% STUDENT: [INAUDIBLE] 00:28:30.560 --> 00:28:33.420 align:middle line:84% STUDENT: I'm just thinking also because if the Super users are 00:28:33.420 --> 00:28:36.620 align:middle line:84% really powerful there, it's quite quick 00:28:36.620 --> 00:28:40.340 align:middle line:90% to calculate their effect. 00:28:40.340 --> 00:28:45.260 align:middle line:84% So what market share do the super users actually buy? 00:28:45.260 --> 00:28:47.080 align:middle line:84% So you can see, if they've been growing, 00:28:47.080 --> 00:28:49.820 align:middle line:84% and if you only look at the super users and the market 00:28:49.820 --> 00:28:51.460 align:middle line:84% for the past 10 years, you can see 00:28:51.460 --> 00:28:54.700 align:middle line:84% if that line is kind of growing faster than the-- 00:28:54.700 --> 00:28:56.580 align:middle line:90% So and because yeah. 00:28:56.580 --> 00:29:00.420 align:middle line:84% And see if that is really, like, is it a big part of the market 00:29:00.420 --> 00:29:03.850 align:middle line:90% share or is not. 00:29:03.850 --> 00:29:05.350 align:middle line:84% NELSON REPENNING: Yeah, so I think-- 00:29:05.350 --> 00:29:06.528 align:middle line:90% I mean, if you have-- 00:29:06.528 --> 00:29:08.070 align:middle line:84% and again, I can already see you sort 00:29:08.070 --> 00:29:09.610 align:middle line:84% of wandering into other hypotheses. 00:29:09.610 --> 00:29:11.568 align:middle line:84% So you're going to have to be very disciplined. 00:29:11.568 --> 00:29:15.190 align:middle line:84% But I think you have two that I think sound pretty interesting, 00:29:15.190 --> 00:29:16.330 align:middle line:90% and good places to start. 00:29:16.330 --> 00:29:18.470 align:middle line:84% One is essentially just a technology shift 00:29:18.470 --> 00:29:20.370 align:middle line:84% that's made it easier to consume books. 00:29:20.370 --> 00:29:23.870 align:middle line:84% So now, I listen to a book when I drive into work in the morning 00:29:23.870 --> 00:29:24.570 align:middle line:90% or whatever. 00:29:24.570 --> 00:29:31.280 align:middle line:84% And I think that would be relatively easy to model. 00:29:31.280 --> 00:29:33.530 align:middle line:84% That part's not a particularly dynamic problem, right, 00:29:33.530 --> 00:29:37.250 align:middle line:84% which is just the cost of consumption has gone down. 00:29:37.250 --> 00:29:38.810 align:middle line:90% And so demand is going to go up. 00:29:38.810 --> 00:29:43.270 align:middle line:84% But there probably is some pretty simple dynamics 00:29:43.270 --> 00:29:47.990 align:middle line:84% as Audible has been introduced, or as there's online books, 00:29:47.990 --> 00:29:52.830 align:middle line:84% or different sort of transcription technologies, that 00:29:52.830 --> 00:29:54.287 align:middle line:90% should cause demand to go up. 00:29:54.287 --> 00:29:55.870 align:middle line:84% And then I think you have a little bit 00:29:55.870 --> 00:29:57.990 align:middle line:84% more feedback-rich version, which is basically 00:29:57.990 --> 00:30:02.880 align:middle line:84% the kind of positive feedback around Instagram, 00:30:02.880 --> 00:30:04.580 align:middle line:84% and social users, and stuff like that. 00:30:04.580 --> 00:30:07.480 align:middle line:84% I think each of those could be modeled in a really 00:30:07.480 --> 00:30:10.800 align:middle line:90% spare, stylized way. 00:30:10.800 --> 00:30:14.120 align:middle line:84% And you might well find, this is the kind of places where i think 00:30:14.120 --> 00:30:16.400 align:middle line:84% system dynamics is good, is if we 00:30:16.400 --> 00:30:18.560 align:middle line:84% were to go with the first hypothesis, which is just 00:30:18.560 --> 00:30:20.080 align:middle line:84% technology, right, you probably have 00:30:20.080 --> 00:30:22.520 align:middle line:84% to make some very extreme assumptions about the demand 00:30:22.520 --> 00:30:26.120 align:middle line:84% curve, that the marginal benefit of being 00:30:26.120 --> 00:30:29.440 align:middle line:84% able to listen over read has got to probably have gone up 00:30:29.440 --> 00:30:32.000 align:middle line:84% quite dramatically in order to account 00:30:32.000 --> 00:30:35.240 align:middle line:84% for the increase in demand, whereas the reinforcing feedback 00:30:35.240 --> 00:30:37.540 align:middle line:90% loop might offer more pieces. 00:30:37.540 --> 00:30:39.460 align:middle line:84% But I would focus on just those two. 00:30:39.460 --> 00:30:40.627 align:middle line:90% Do we have a model from you? 00:30:40.627 --> 00:30:41.127 align:middle line:90% You? 00:30:41.127 --> 00:30:41.680 align:middle line:90% No? 00:30:41.680 --> 00:30:42.360 align:middle line:90% Nothing? 00:30:42.360 --> 00:30:43.120 align:middle line:90% OK. 00:30:43.120 --> 00:30:44.120 align:middle line:90% So you're still in the-- 00:30:44.120 --> 00:30:45.800 align:middle line:84% STUDENT: I have instructure-- instructed-- 00:30:45.800 --> 00:30:46.675 align:middle line:90% NELSON REPENNING: OK. 00:30:46.675 --> 00:30:47.415 align:middle line:90% STUDENT: OK. 00:30:47.415 --> 00:30:49.040 align:middle line:84% NELSON REPENNING: So I think the advice 00:30:49.040 --> 00:30:54.580 align:middle line:84% I would give for you is stick with just those two hypotheses, 00:30:54.580 --> 00:30:59.490 align:middle line:84% which is basically one feedback loop apiece, plus or minus. 00:30:59.490 --> 00:31:01.608 align:middle line:84% Aggregate all books together for now. 00:31:01.608 --> 00:31:03.650 align:middle line:84% I know there's these different modalities of self 00:31:03.650 --> 00:31:05.070 align:middle line:90% published and stuff like that. 00:31:05.070 --> 00:31:07.033 align:middle line:84% And I'm happy to share with all of you, 00:31:07.033 --> 00:31:09.450 align:middle line:84% for someone who's recently written a book, all the details 00:31:09.450 --> 00:31:10.575 align:middle line:90% of the publishing industry. 00:31:10.575 --> 00:31:13.170 align:middle line:90% But that's for another time. 00:31:13.170 --> 00:31:15.072 align:middle line:84% And then I think in this case-- and so 00:31:15.072 --> 00:31:16.530 align:middle line:84% this is an interesting thing, which 00:31:16.530 --> 00:31:20.150 align:middle line:84% I don't think is, again, a theorem or an algorithm. 00:31:20.150 --> 00:31:25.250 align:middle line:84% But this is a case where I think the feedback processes jump out 00:31:25.250 --> 00:31:27.270 align:middle line:84% to me more than the stock flow structure. 00:31:27.270 --> 00:31:29.312 align:middle line:84% There are some cases where I think the stock flow 00:31:29.312 --> 00:31:30.530 align:middle line:90% structure is really clear. 00:31:30.530 --> 00:31:33.870 align:middle line:84% We have patients, and then they become whatever, right? 00:31:33.870 --> 00:31:37.430 align:middle line:84% In this case, we have readers and non-readers. 00:31:37.430 --> 00:31:41.930 align:middle line:84% But I think the kernel of the dynamic hypothesis 00:31:41.930 --> 00:31:44.823 align:middle line:84% you have is more feedback rich than it is stock and flow rich. 00:31:44.823 --> 00:31:46.490 align:middle line:84% And that's why I think this debate about 00:31:46.490 --> 00:31:47.850 align:middle line:84% should you start with stocks and flows 00:31:47.850 --> 00:31:49.690 align:middle line:84% or feedback loops is just utterly stupid 00:31:49.690 --> 00:31:52.813 align:middle line:84% because it doesn't matter where you start. 00:31:52.813 --> 00:31:54.730 align:middle line:84% Hopefully, you're going to end up with a model 00:31:54.730 --> 00:31:55.610 align:middle line:90% when you're done. 00:31:55.610 --> 00:31:57.120 align:middle line:84% And so you should pick the doorway 00:31:57.120 --> 00:31:59.258 align:middle line:84% that is kind of the easiest to get going. 00:31:59.258 --> 00:32:01.300 align:middle line:84% I think in your case, the feedback loops probably 00:32:01.300 --> 00:32:03.242 align:middle line:84% jump out a little bit more than the underlying 00:32:03.242 --> 00:32:05.700 align:middle line:84% sort of stock flow structure because, yeah, you're probably 00:32:05.700 --> 00:32:09.660 align:middle line:84% going to have a flow of kids that learn how to read, 00:32:09.660 --> 00:32:13.660 align:middle line:84% and then they either become regular readers, or non-readers. 00:32:13.660 --> 00:32:18.760 align:middle line:84% And then the regular readers are going 00:32:18.760 --> 00:32:20.260 align:middle line:84% to have some certain number of books 00:32:20.260 --> 00:32:23.040 align:middle line:84% that they consume per year or something like that. 00:32:23.040 --> 00:32:25.960 align:middle line:84% So that's probably relatively straightforward. 00:32:25.960 --> 00:32:26.800 align:middle line:90% So-- 00:32:26.800 --> 00:32:27.300 align:middle line:90% STUDENT: OK. 00:32:27.300 --> 00:32:28.060 align:middle line:90% That's helpful. 00:32:28.060 --> 00:32:29.060 align:middle line:90% NELSON REPENNING: OK. 00:32:29.060 --> 00:32:29.680 align:middle line:90% That's good. 00:32:29.680 --> 00:32:30.740 align:middle line:90% Does that help? 00:32:30.740 --> 00:32:31.948 align:middle line:90% STUDENT: Yeah, it's succinct. 00:32:31.948 --> 00:32:33.865 align:middle line:84% NELSON REPENNING: I think it's a good problem. 00:32:33.865 --> 00:32:36.060 align:middle line:84% And I think there's a corollary, which this is just 00:32:36.060 --> 00:32:39.120 align:middle line:84% an editorializing that I have long been interested in, 00:32:39.120 --> 00:32:43.020 align:middle line:84% which is if you look at the sort of different forms of media 00:32:43.020 --> 00:32:51.120 align:middle line:84% we consume, so say, just books, movies, and music, to me, 00:32:51.120 --> 00:32:52.540 align:middle line:84% it's always looked like the movie 00:32:52.540 --> 00:32:54.340 align:middle line:90% industry is the most top heavy. 00:32:54.340 --> 00:32:56.870 align:middle line:84% If you were to plot demand, there's 00:32:56.870 --> 00:32:58.970 align:middle line:84% a few movies that get zillions of views 00:32:58.970 --> 00:33:00.490 align:middle line:84% and then it falls off very rapidly. 00:33:00.490 --> 00:33:04.412 align:middle line:84% And my impression is that books are much more egalitarian. 00:33:04.412 --> 00:33:05.870 align:middle line:84% I mean, you still have bestsellers, 00:33:05.870 --> 00:33:07.170 align:middle line:90% and it's probably a power law. 00:33:07.170 --> 00:33:11.190 align:middle line:84% But I've never seen a good story as to why 00:33:11.190 --> 00:33:15.670 align:middle line:84% what sort of determines that distribution. 00:33:15.670 --> 00:33:19.710 align:middle line:84% And why is it that maybe there's 10,000 book titles that are 00:33:19.710 --> 00:33:22.110 align:middle line:84% viable, and there's only a thousand movie titles that are 00:33:22.110 --> 00:33:23.517 align:middle line:90% viable, or something like that? 00:33:23.517 --> 00:33:25.350 align:middle line:84% And then but then there's been a huge change 00:33:25.350 --> 00:33:28.377 align:middle line:84% in technology, Netflix, and streaming, and ebooks, 00:33:28.377 --> 00:33:29.210 align:middle line:90% and stuff like that. 00:33:29.210 --> 00:33:31.270 align:middle line:84% And so has that sort of changed that kind 00:33:31.270 --> 00:33:34.270 align:middle line:84% of power-law-esque distribution of the winner 00:33:34.270 --> 00:33:36.320 align:middle line:90% sort of take all pieces. 00:33:36.320 --> 00:33:38.070 align:middle line:84% And what's happened in the music industry, 00:33:38.070 --> 00:33:40.710 align:middle line:84% where now no one seems to make any money except for Taylor 00:33:40.710 --> 00:33:42.470 align:middle line:90% Swift or whatever? 00:33:42.470 --> 00:33:45.110 align:middle line:84% But it does seem like there's a huge diversity of titles 00:33:45.110 --> 00:33:46.792 align:middle line:90% that are out there available. 00:33:46.792 --> 00:33:48.250 align:middle line:84% And you could probably think of it. 00:33:48.250 --> 00:33:50.310 align:middle line:84% I'm not proposing that you do that. 00:33:50.310 --> 00:33:51.952 align:middle line:84% But I think if you get this sorted, 00:33:51.952 --> 00:33:54.160 align:middle line:84% it could lead to all sorts of really interesting kind 00:33:54.160 --> 00:33:57.880 align:middle line:84% of industry dynamics kind of questions. 00:33:57.880 --> 00:34:01.320 align:middle line:84% STUDENT: There's the book by Anita Elberse at Harvard 00:34:01.320 --> 00:34:03.960 align:middle line:84% called Blockbuster that looks specifically 00:34:03.960 --> 00:34:07.700 align:middle line:84% at that question, like why, in [? US ?] entertainment, 00:34:07.700 --> 00:34:10.800 align:middle line:84% are there some, usually popular ones, a lot. 00:34:10.800 --> 00:34:13.062 align:middle line:90% That's [? important. ?] 00:34:13.062 --> 00:34:14.520 align:middle line:84% NELSON REPENNING: And I assume it's 00:34:14.520 --> 00:34:18.538 align:middle line:84% a sort of positive-feedback-ish argument. 00:34:18.538 --> 00:34:20.080 align:middle line:84% Is there a form of modeling in there? 00:34:20.080 --> 00:34:20.340 align:middle line:90% Or-- 00:34:20.340 --> 00:34:21.820 align:middle line:84% STUDENT: I don't think there's any formal modeling. 00:34:21.820 --> 00:34:21.920 align:middle line:90% No. 00:34:21.920 --> 00:34:23.639 align:middle line:84% But I mean, it's basically the idea 00:34:23.639 --> 00:34:29.480 align:middle line:84% of if you have a huge success, it's initially, 00:34:29.480 --> 00:34:31.653 align:middle line:84% it's because of word of mouth and people 00:34:31.653 --> 00:34:33.820 align:middle line:84% telling their friends like, oh, this really is good. 00:34:33.820 --> 00:34:35.580 align:middle line:84% And once you've had one huge success, 00:34:35.580 --> 00:34:38.000 align:middle line:84% you're already in the blockbuster category. 00:34:38.000 --> 00:34:38.820 align:middle line:90% [INAUDIBLE] 00:34:38.820 --> 00:34:43.920 align:middle line:84% So you're able to go around and get deals, the best marketing 00:34:43.920 --> 00:34:45.000 align:middle line:90% in the world, basically. 00:34:45.000 --> 00:34:46.500 align:middle line:84% And everyone already knows, oh yeah. 00:34:46.500 --> 00:34:47.600 align:middle line:90% I heard of that movie. 00:34:47.600 --> 00:34:49.880 align:middle line:90% And [INAUDIBLE] hits. 00:34:49.880 --> 00:34:50.880 align:middle line:90% NELSON REPENNING: Right. 00:34:50.880 --> 00:34:54.850 align:middle line:84% But I do wonder, and we'll get to the next one, 00:34:54.850 --> 00:34:56.100 align:middle line:90% this is an empirical question. 00:34:56.100 --> 00:34:59.680 align:middle line:84% Is the top heaviness in the movie industry and the book 00:34:59.680 --> 00:35:02.100 align:middle line:90% industry of the same degree? 00:35:02.100 --> 00:35:04.440 align:middle line:84% And it might be as simple as movies are just 00:35:04.440 --> 00:35:07.500 align:middle line:84% way more expensive to create than books. 00:35:07.500 --> 00:35:09.225 align:middle line:84% Video games, probably the same thing. 00:35:09.225 --> 00:35:11.600 align:middle line:84% But it's an interesting question of what kind of allows-- 00:35:11.600 --> 00:35:14.257 align:middle line:84% and who's working on the luxury goods? 00:35:14.257 --> 00:35:14.840 align:middle line:90% STUDENT: I am. 00:35:14.840 --> 00:35:15.798 align:middle line:90% NELSON REPENNING: Yeah. 00:35:15.798 --> 00:35:20.927 align:middle line:84% So you have a really interesting flavor of this because-- 00:35:20.927 --> 00:35:22.760 align:middle line:84% STUDENT: [INAUDIBLE] There's two of us here. 00:35:22.760 --> 00:35:23.802 align:middle line:90% NELSON REPENNING: Oh, OK. 00:35:23.802 --> 00:35:24.620 align:middle line:90% STUDENT: Which one? 00:35:24.620 --> 00:35:25.060 align:middle line:90% NELSON REPENNING: Oh, OK. 00:35:25.060 --> 00:35:26.460 align:middle line:90% Is someone else working on this? 00:35:26.460 --> 00:35:27.335 align:middle line:90% STUDENT: [INAUDIBLE]. 00:35:27.335 --> 00:35:29.023 align:middle line:90% NELSON REPENNING: Oh, OK. 00:35:29.023 --> 00:35:30.940 align:middle line:84% So I think this is also a fascinating problem. 00:35:30.940 --> 00:35:34.720 align:middle line:84% Maybe we'll come to yours next because there's 00:35:34.720 --> 00:35:36.920 align:middle line:84% a simple version of increasing returns, which 00:35:36.920 --> 00:35:40.120 align:middle line:84% is the more people that read this book or watch this movie, 00:35:40.120 --> 00:35:42.235 align:middle line:84% then that just leads to more demand. 00:35:42.235 --> 00:35:44.860 align:middle line:84% But then there's the flavor that I think you're thinking about, 00:35:44.860 --> 00:35:46.507 align:middle line:84% which is that's true up to a degree. 00:35:46.507 --> 00:35:49.090 align:middle line:84% And then suddenly, I don't want to listen to that band anymore 00:35:49.090 --> 00:35:51.730 align:middle line:84% because you all listen to it, and so it's no longer cool, 00:35:51.730 --> 00:35:54.450 align:middle line:84% which adds, I think this really interesting kind 00:35:54.450 --> 00:35:58.520 align:middle line:90% of balancing feedback loop. 00:35:58.520 --> 00:36:01.330 align:middle line:84% And I don't know if anyone maybe there's someone's worked out 00:36:01.330 --> 00:36:03.710 align:middle line:84% that if I want to be in that niche market, 00:36:03.710 --> 00:36:06.170 align:middle line:84% how do you sort of manage the scarcity, and the hype, 00:36:06.170 --> 00:36:10.170 align:middle line:84% and sort of all that stuff so that you don't-- because I was 00:36:10.170 --> 00:36:13.290 align:middle line:84% just listening to some podcast about how for many years, 00:36:13.290 --> 00:36:15.150 align:middle line:84% there were the same luxury hotels. 00:36:15.150 --> 00:36:17.070 align:middle line:84% There was like The Four Seasons, and The Ritz, 00:36:17.070 --> 00:36:19.410 align:middle line:84% and then suddenly, there's been a whole new group that 00:36:19.410 --> 00:36:23.410 align:middle line:84% has come in on top that charges even more than The Four Seasons. 00:36:23.410 --> 00:36:25.170 align:middle line:84% And like, it just seems like there's 00:36:25.170 --> 00:36:28.530 align:middle line:84% this dynamic where every so often, another brand will show 00:36:28.530 --> 00:36:32.370 align:middle line:84% up that is even more premium and more scarce than what we thought 00:36:32.370 --> 00:36:34.590 align:middle line:90% was the logical conclusion. 00:36:34.590 --> 00:36:36.090 align:middle line:84% And I don't know if anyone's sort of 00:36:36.090 --> 00:36:37.530 align:middle line:90% worked out the dynamics of it. 00:36:37.530 --> 00:36:40.010 align:middle line:84% Seems like a really interesting question. 00:36:40.010 --> 00:36:40.510 align:middle line:90% Yeah. 00:36:40.510 --> 00:36:42.670 align:middle line:84% STUDENT: There's so many ideas that I could share. 00:36:42.670 --> 00:36:48.540 align:middle line:84% But one thing that, before I forget, since in your model, 00:36:48.540 --> 00:36:52.660 align:middle line:84% one thing I'm realizing that might be helpful to think about 00:36:52.660 --> 00:37:01.180 align:middle line:84% separately is kind of the weight that readers might put on, like, 00:37:01.180 --> 00:37:04.100 align:middle line:84% the intrinsic value of the media produce 00:37:04.100 --> 00:37:07.260 align:middle line:84% versus the weight they would put on what 00:37:07.260 --> 00:37:10.200 align:middle line:90% it does for them socially. 00:37:10.200 --> 00:37:16.747 align:middle line:84% So I wouldn't-- if this book was just here, I wouldn't read it. 00:37:16.747 --> 00:37:19.080 align:middle line:84% But because other people are reading it, I'm interested. 00:37:19.080 --> 00:37:21.380 align:middle line:84% So there's a weighting function, and then there's 00:37:21.380 --> 00:37:24.900 align:middle line:84% also the actual quality of what gets produced. 00:37:24.900 --> 00:37:27.220 align:middle line:90% They're different. 00:37:27.220 --> 00:37:29.540 align:middle line:84% STUDENT: And the quality of what is produced 00:37:29.540 --> 00:37:33.480 align:middle line:84% is what was suggested that I got as dynamic. 00:37:33.480 --> 00:37:34.420 align:middle line:90% Yeah. 00:37:34.420 --> 00:37:36.380 align:middle line:90% STUDENT: So hold that constant? 00:37:36.380 --> 00:37:42.552 align:middle line:84% STUDENT: Yeah, just not worry first about the supply side. 00:37:42.552 --> 00:37:44.260 align:middle line:84% NELSON REPENNING: Well, so again, I mean, 00:37:44.260 --> 00:37:45.450 align:middle line:90% I agree with that. 00:37:45.450 --> 00:37:47.870 align:middle line:84% But I want you to know why you're 00:37:47.870 --> 00:37:51.830 align:middle line:84% doing that, which is you have a dynamic hypothesis about what's 00:37:51.830 --> 00:37:52.610 align:middle line:90% important here. 00:37:52.610 --> 00:37:54.390 align:middle line:84% It could be right, it could be wrong. 00:37:54.390 --> 00:37:56.290 align:middle line:84% And I have no particular view of that. 00:37:56.290 --> 00:37:59.390 align:middle line:84% But I think, don't let any of yourselves 00:37:59.390 --> 00:38:00.930 align:middle line:84% kind of get wrapped around the axle. 00:38:00.930 --> 00:38:02.680 align:middle line:84% Well, geez, I know all six of these things 00:38:02.680 --> 00:38:05.310 align:middle line:84% are important, so I'm going to put them all in at once. 00:38:05.310 --> 00:38:06.830 align:middle line:84% Then you're just creating a big mess 00:38:06.830 --> 00:38:08.655 align:middle line:84% that's going to be very hard to sort out. 00:38:08.655 --> 00:38:10.030 align:middle line:84% So it kind of goes back to Jason, 00:38:10.030 --> 00:38:11.697 align:middle line:84% where it's hard to make those decisions, 00:38:11.697 --> 00:38:14.510 align:middle line:84% particularly when you have a lot of instincts 00:38:14.510 --> 00:38:16.630 align:middle line:90% about what's going on. 00:38:16.630 --> 00:38:19.670 align:middle line:84% But that typically is trouble to try 00:38:19.670 --> 00:38:22.190 align:middle line:84% to get too far away from a model that you can simulate 00:38:22.190 --> 00:38:24.030 align:middle line:84% and that you can understand and analyze. 00:38:24.030 --> 00:38:27.070 align:middle line:84% STUDENT: Yeah, but what I don't really understand is if Cathy, 00:38:27.070 --> 00:38:31.230 align:middle line:84% and I agree with her, thinks that's an important contributor, 00:38:31.230 --> 00:38:32.690 align:middle line:90% how do you-- 00:38:32.690 --> 00:38:36.230 align:middle line:84% and that's where you want to land. 00:38:36.230 --> 00:38:41.513 align:middle line:84% It seems to me, irreducible, and you cannot leave it out. 00:38:41.513 --> 00:38:42.930 align:middle line:84% NELSON REPENNING: In the long run. 00:38:42.930 --> 00:38:45.620 align:middle line:90% 00:38:45.620 --> 00:38:50.600 align:middle line:84% The key is you don't want to get too far away from a model 00:38:50.600 --> 00:38:52.820 align:middle line:84% that you understand why it does what it does. 00:38:52.820 --> 00:38:55.480 align:middle line:84% So here's another, like, editorial comment. 00:38:55.480 --> 00:38:57.600 align:middle line:84% Like, this is one of the many reasons 00:38:57.600 --> 00:39:00.880 align:middle line:84% that I don't go to the conferences often anymore. 00:39:00.880 --> 00:39:03.020 align:middle line:84% This is my least-favorite presentation, 00:39:03.020 --> 00:39:07.097 align:middle line:84% which is I built this big model, I threw in every feedback loop 00:39:07.097 --> 00:39:08.680 align:middle line:84% I could possibly think of, and a few I 00:39:08.680 --> 00:39:12.000 align:middle line:84% got from my mom, and my dog, and so on, and so forth. 00:39:12.000 --> 00:39:14.630 align:middle line:84% And look, the behavior is really complicated. 00:39:14.630 --> 00:39:16.223 align:middle line:90% [LAUGHTER] 00:39:16.223 --> 00:39:18.640 align:middle line:84% NELSON REPENNING: It just sends me over the edge because-- 00:39:18.640 --> 00:39:22.720 align:middle line:84% or nowadays, because the computational is so cheap, 00:39:22.720 --> 00:39:25.600 align:middle line:84% and look, it fits the data really well. 00:39:25.600 --> 00:39:27.180 align:middle line:90% OK, great. 00:39:27.180 --> 00:39:29.020 align:middle line:90% What have I learned from this? 00:39:29.020 --> 00:39:30.780 align:middle line:90% Right? 00:39:30.780 --> 00:39:35.160 align:middle line:84% Again, it goes back to, has it either helped me out the puzzle 00:39:35.160 --> 00:39:36.840 align:middle line:90% that we were targeting? 00:39:36.840 --> 00:39:40.440 align:middle line:84% Or have I identified some policy where 00:39:40.440 --> 00:39:43.010 align:middle line:84% I could intervene if I was going to give advice 00:39:43.010 --> 00:39:45.770 align:middle line:84% to the next high-end handbag maker that's going to come over, 00:39:45.770 --> 00:39:47.090 align:middle line:90% Birkins or whatever? 00:39:47.090 --> 00:39:48.870 align:middle line:84% Do I have some policy conclusions? 00:39:48.870 --> 00:39:50.828 align:middle line:84% And in some cases, they have policy conclusions 00:39:50.828 --> 00:39:52.810 align:middle line:84% that have been computationally derived. 00:39:52.810 --> 00:39:56.690 align:middle line:90% But I don't have any faith. 00:39:56.690 --> 00:39:58.590 align:middle line:90% I don't know why they work. 00:39:58.590 --> 00:40:02.510 align:middle line:90% 00:40:02.510 --> 00:40:05.010 align:middle line:84% And frankly, if you were going to build a model where you're 00:40:05.010 --> 00:40:07.450 align:middle line:84% not going to worry about the why it works and you're just going 00:40:07.450 --> 00:40:09.490 align:middle line:84% to take the model on faith, you might as well 00:40:09.490 --> 00:40:11.650 align:middle line:84% just use the AI models because they're bigger 00:40:11.650 --> 00:40:13.610 align:middle line:84% and they're more accurate, right? 00:40:13.610 --> 00:40:14.370 align:middle line:84% STUDENT: Yeah, they fit data better. 00:40:14.370 --> 00:40:16.245 align:middle line:84% NELSON REPENNING: Yeah, they fit data better. 00:40:16.245 --> 00:40:17.890 align:middle line:84% And so if you're going to give up 00:40:17.890 --> 00:40:20.190 align:middle line:84% on understanding the underlying structure of why, 00:40:20.190 --> 00:40:23.040 align:middle line:84% then you might as well go with the more powerful sort 00:40:23.040 --> 00:40:24.113 align:middle line:90% of technology. 00:40:24.113 --> 00:40:26.030 align:middle line:84% So yeah, I think it's proceeded incrementally. 00:40:26.030 --> 00:40:28.210 align:middle line:84% And again, I will place this as John 00:40:28.210 --> 00:40:30.710 align:middle line:84% and I have had big fights about this over the years, right? 00:40:30.710 --> 00:40:34.490 align:middle line:84% I think he's more comfortable putting a lot of stuff 00:40:34.490 --> 00:40:36.250 align:middle line:84% in, and often because well, because we 00:40:36.250 --> 00:40:38.290 align:middle line:90% know it has to be there. 00:40:38.290 --> 00:40:42.140 align:middle line:84% And I think, again, you're just creating risk for yourself 00:40:42.140 --> 00:40:45.540 align:middle line:84% in terms of your ability to understand what's going on. 00:40:45.540 --> 00:40:48.780 align:middle line:84% STUDENT: As a practical matter, like when you're planning ahead, 00:40:48.780 --> 00:40:50.040 align:middle line:90% OK, here's my vision. 00:40:50.040 --> 00:40:51.800 align:middle line:84% So I want to build this structure 00:40:51.800 --> 00:40:56.120 align:middle line:84% so that it can accommodate a changing quality of content. 00:40:56.120 --> 00:40:59.300 align:middle line:90% 00:40:59.300 --> 00:41:06.220 align:middle line:84% What I'm hearing from you is, kind of build it 00:41:06.220 --> 00:41:11.120 align:middle line:84% in airtight kind of compartments, 00:41:11.120 --> 00:41:14.940 align:middle line:90% not necessarily planning ahead. 00:41:14.940 --> 00:41:15.482 align:middle line:90% I don't know. 00:41:15.482 --> 00:41:17.773 align:middle line:84% NELSON REPENNING: I'm not so worried about the planning 00:41:17.773 --> 00:41:18.433 align:middle line:90% ahead part. 00:41:18.433 --> 00:41:19.600 align:middle line:90% I mean, I think that's fine. 00:41:19.600 --> 00:41:21.140 align:middle line:84% I think another way of saying what 00:41:21.140 --> 00:41:24.380 align:middle line:84% my advice is I think the model size, and your understanding 00:41:24.380 --> 00:41:25.895 align:middle line:90% of it, should go together. 00:41:25.895 --> 00:41:26.520 align:middle line:90% STUDENT: I see. 00:41:26.520 --> 00:41:27.520 align:middle line:90% NELSON REPENNING: Right? 00:41:27.520 --> 00:41:29.998 align:middle line:84% Rather than, I think there's an instinct of, 00:41:29.998 --> 00:41:31.540 align:middle line:84% I'm going to put in everything that I 00:41:31.540 --> 00:41:34.795 align:middle line:84% think I need to put in, formulate it properly, use 00:41:34.795 --> 00:41:36.920 align:middle line:84% all standard formulations, and so on, and so forth. 00:41:36.920 --> 00:41:40.110 align:middle line:84% And then when it's done, then I'm going to analyze it. 00:41:40.110 --> 00:41:41.932 align:middle line:84% And I think that, for all the reasons 00:41:41.932 --> 00:41:44.390 align:middle line:84% you cited, because it's dynamic, and non-linear, and so on, 00:41:44.390 --> 00:41:46.350 align:middle line:84% and so forth, it's going to be very 00:41:46.350 --> 00:41:49.830 align:middle line:84% hard to derive any meaningful insight from that 00:41:49.830 --> 00:41:51.530 align:middle line:90% that you actually understand. 00:41:51.530 --> 00:41:53.330 align:middle line:84% There may be insights embodied in it, 00:41:53.330 --> 00:41:54.770 align:middle line:90% but they're very hard to find. 00:41:54.770 --> 00:41:56.670 align:middle line:84% And I promise if you go that route, you 00:41:56.670 --> 00:41:59.710 align:middle line:84% will end up simplifying it after the fact 00:41:59.710 --> 00:42:03.533 align:middle line:84% to try to test your instincts about what's going on. 00:42:03.533 --> 00:42:04.950 align:middle line:84% And then this is a little bit more 00:42:04.950 --> 00:42:08.010 align:middle line:84% for just people who have sort of academic ambitions. 00:42:08.010 --> 00:42:11.630 align:middle line:84% At least in the organizations strategy world, 00:42:11.630 --> 00:42:14.730 align:middle line:84% right, model size is, at least in my experience, 00:42:14.730 --> 00:42:17.410 align:middle line:84% not dictated by the complexity of the phenomenon. 00:42:17.410 --> 00:42:19.750 align:middle line:84% It is dictated by the complexity of the existing 00:42:19.750 --> 00:42:21.990 align:middle line:90% theories in the literature. 00:42:21.990 --> 00:42:26.350 align:middle line:84% And what I mean by that is if a theory in the strategy world 00:42:26.350 --> 00:42:28.850 align:middle line:84% has sort of implicitly three feedback loops, 00:42:28.850 --> 00:42:32.130 align:middle line:84% and you bring in a model that has 50 of them, 00:42:32.130 --> 00:42:34.150 align:middle line:84% your paper is going to get rejected as, 00:42:34.150 --> 00:42:36.060 align:middle line:90% like, I don't understand this. 00:42:36.060 --> 00:42:37.840 align:middle line:90% This is way too complicated. 00:42:37.840 --> 00:42:40.200 align:middle line:84% And I think this is an era that system dynamics has 00:42:40.200 --> 00:42:42.280 align:middle line:90% made for decades. 00:42:42.280 --> 00:42:45.940 align:middle line:84% Now if it's a policy model, that's a different question. 00:42:45.940 --> 00:42:47.940 align:middle line:84% But if it's something you're going to publish, 00:42:47.940 --> 00:42:51.023 align:middle line:84% you can't get too far ahead of the sort of state of the art 00:42:51.023 --> 00:42:52.440 align:middle line:84% because referees are going to have 00:42:52.440 --> 00:42:55.740 align:middle line:84% no idea how to kind of process that or make sense of it. 00:42:55.740 --> 00:42:58.640 align:middle line:84% So the doctoral students already hold this. 00:42:58.640 --> 00:43:01.280 align:middle line:84% So many of the things I publish are these very simple models. 00:43:01.280 --> 00:43:03.260 align:middle line:84% And many cases, one of them in particular, 00:43:03.260 --> 00:43:04.600 align:middle line:90% really started out as a joke. 00:43:04.600 --> 00:43:07.240 align:middle line:84% Like, let's see what is the simplest thing that I 00:43:07.240 --> 00:43:09.160 align:middle line:90% could send to the journal. 00:43:09.160 --> 00:43:10.940 align:middle line:84% And it turned out it did very well, 00:43:10.940 --> 00:43:12.200 align:middle line:90% and it was a popular paper. 00:43:12.200 --> 00:43:14.240 align:middle line:84% It was the disaster dynamics model. 00:43:14.240 --> 00:43:16.560 align:middle line:84% But that was just because in that particular thread 00:43:16.560 --> 00:43:19.180 align:middle line:84% of the literature, there were no models at all. 00:43:19.180 --> 00:43:25.360 align:middle line:84% So one stock was actually an advance over the discussion 00:43:25.360 --> 00:43:26.760 align:middle line:90% that had happened before. 00:43:26.760 --> 00:43:29.400 align:middle line:84% Different facets of the literature 00:43:29.400 --> 00:43:32.560 align:middle line:90% are in different places. 00:43:32.560 --> 00:43:34.730 align:middle line:90% OK, who'd like to go next? 00:43:34.730 --> 00:43:35.881 align:middle line:90% STUDENT: I can do-- 00:43:35.881 --> 00:43:36.714 align:middle line:90% STUDENT: [INAUDIBLE] 00:43:36.714 --> 00:43:37.210 align:middle line:90% STUDENT: I can do mine. 00:43:37.210 --> 00:43:38.530 align:middle line:90% NELSON REPENNING: OK. 00:43:38.530 --> 00:43:39.750 align:middle line:90% Do I have yours in my-- 00:43:39.750 --> 00:43:43.570 align:middle line:84% STUDENT: Yeah, but my diagram, it's not readable 00:43:43.570 --> 00:43:46.930 align:middle line:84% because I started doing it, and then I 00:43:46.930 --> 00:43:48.985 align:middle line:84% kind of like lost sight of what the boundary was. 00:43:48.985 --> 00:43:50.610 align:middle line:84% It should be under Sampson [INAUDIBLE]. 00:43:50.610 --> 00:43:51.670 align:middle line:90% NELSON REPENNING: OK. 00:43:51.670 --> 00:43:52.890 align:middle line:90% Yeah. 00:43:52.890 --> 00:43:54.665 align:middle line:84% STUDENT: And so it doesn't make any sense. 00:43:54.665 --> 00:43:55.790 align:middle line:90% NELSON REPENNING: Oh, shit. 00:43:55.790 --> 00:43:56.665 align:middle line:90% I did it wrong again. 00:43:56.665 --> 00:43:57.490 align:middle line:90% Sorry. 00:43:57.490 --> 00:44:00.608 align:middle line:84% So tell us about the problem you're solving, or the puzzle, 00:44:00.608 --> 00:44:02.150 align:middle line:84% and then we'll come back to the model 00:44:02.150 --> 00:44:04.050 align:middle line:84% because I opened up in the wrong thing. 00:44:04.050 --> 00:44:07.630 align:middle line:84% STUDENT: Yeah, so what he was saying, 00:44:07.630 --> 00:44:10.210 align:middle line:84% one of the main ideas I was trying to look at 00:44:10.210 --> 00:44:16.250 align:middle line:84% was how you maintain the coolness of a brand when 00:44:16.250 --> 00:44:19.450 align:middle line:84% people from the out-group start buying more of it. 00:44:19.450 --> 00:44:25.290 align:middle line:84% Which was the problem started kind of because I initially 00:44:25.290 --> 00:44:27.170 align:middle line:84% had been thinking a lot about how resellers 00:44:27.170 --> 00:44:30.810 align:middle line:84% impact the amount of profit that the actual brand can get. 00:44:30.810 --> 00:44:34.725 align:middle line:84% But then I started looking at the in-group, out-group dynamic. 00:44:34.725 --> 00:44:36.350 align:middle line:84% And I thought that was more interesting 00:44:36.350 --> 00:44:37.308 align:middle line:90% so that I didn't have-- 00:44:37.308 --> 00:44:40.330 align:middle line:84% I don't have any of really the reseller or profit 00:44:40.330 --> 00:44:41.810 align:middle line:90% dynamic in here at all. 00:44:41.810 --> 00:44:44.970 align:middle line:84% That might need to get put in later. 00:44:44.970 --> 00:44:46.770 align:middle line:84% NELSON REPENNING: OK, so let's back up. 00:44:46.770 --> 00:44:50.337 align:middle line:84% What is the puzzle you want to answer or the policy-- 00:44:50.337 --> 00:44:52.170 align:middle line:84% STUDENT: So the puzzle is like you imagine-- 00:44:52.170 --> 00:44:54.220 align:middle line:84% NELSON REPENNING: You want to figure it-- 00:44:54.220 --> 00:44:55.470 align:middle line:90% STUDENT: --the brand, Supreme. 00:44:55.470 --> 00:44:57.470 align:middle line:84% I kind of already said this in a prior class, 00:44:57.470 --> 00:45:01.030 align:middle line:84% but they became really popular because it was like, cool, 00:45:01.030 --> 00:45:03.848 align:middle line:84% all the skaters were wearing it like any other street brand. 00:45:03.848 --> 00:45:05.890 align:middle line:84% Then people from the out-group who aren't skaters 00:45:05.890 --> 00:45:07.515 align:middle line:84% start buying it because you want to buy 00:45:07.515 --> 00:45:11.810 align:middle line:84% into this culture of being cool streetwear, or skater person. 00:45:11.810 --> 00:45:14.770 align:middle line:84% But then people from the in-group as the out-group 00:45:14.770 --> 00:45:16.770 align:middle line:84% starts to claim a bigger share of the people who 00:45:16.770 --> 00:45:17.395 align:middle line:90% are wearing it. 00:45:17.395 --> 00:45:20.470 align:middle line:84% It's no longer cool to wear it, and they go to alternatives, 00:45:20.470 --> 00:45:22.170 align:middle line:84% or they stop buying it altogether. 00:45:22.170 --> 00:45:24.610 align:middle line:84% And additionally, the more out-group people 00:45:24.610 --> 00:45:27.228 align:middle line:84% come in, particularly in streetwear culture, 00:45:27.228 --> 00:45:28.770 align:middle line:84% and this would be different depending 00:45:28.770 --> 00:45:31.100 align:middle line:84% on what brand you were looking at, 00:45:31.100 --> 00:45:32.980 align:middle line:84% the price starts to get jacked up, 00:45:32.980 --> 00:45:34.965 align:middle line:84% which makes it cooler for the out-group, 00:45:34.965 --> 00:45:36.840 align:middle line:84% but makes it even less cool for the in-group. 00:45:36.840 --> 00:45:38.860 align:middle line:84% And they're usually more price sensitive, 00:45:38.860 --> 00:45:40.540 align:middle line:84% or care less about that because they're 00:45:40.540 --> 00:45:42.820 align:middle line:90% the ones making the culture. 00:45:42.820 --> 00:45:44.740 align:middle line:90% And so they leave. 00:45:44.740 --> 00:45:47.460 align:middle line:84% And then with some delay, eventually, out-group people 00:45:47.460 --> 00:45:50.060 align:middle line:84% realize that this brand is no longer cool. 00:45:50.060 --> 00:45:54.390 align:middle line:90% And then their demand also ends. 00:45:54.390 --> 00:45:56.140 align:middle line:84% NELSON REPENNING: And then the brand dies. 00:45:56.140 --> 00:45:58.980 align:middle line:84% STUDENT: And then the brand dies, depending on the brand. 00:45:58.980 --> 00:46:04.540 align:middle line:84% There could be an argument that maybe eventually, the in-group 00:46:04.540 --> 00:46:06.600 align:middle line:84% comes back once prices have cooled, 00:46:06.600 --> 00:46:12.785 align:middle line:84% like depending on [INAUDIBLE] how good the alternatives are. 00:46:12.785 --> 00:46:14.160 align:middle line:84% I was thinking, it would be cool, 00:46:14.160 --> 00:46:16.640 align:middle line:84% like this would be an extreme generalization of it, 00:46:16.640 --> 00:46:21.040 align:middle line:84% but it could also maybe eventually pertain to language 00:46:21.040 --> 00:46:22.040 align:middle line:90% as well. 00:46:22.040 --> 00:46:25.220 align:middle line:84% Like, there's specific ways that for example, like, gay people 00:46:25.220 --> 00:46:28.710 align:middle line:84% talk or frat people talk that signal through their voice, 00:46:28.710 --> 00:46:31.430 align:middle line:84% like, I'm part of this in-group and you're not 00:46:31.430 --> 00:46:32.610 align:middle line:90% part of that in-group. 00:46:32.610 --> 00:46:34.270 align:middle line:84% But then once mainstream hits it, 00:46:34.270 --> 00:46:35.990 align:middle line:84% and everyone starts using that, that's 00:46:35.990 --> 00:46:38.310 align:middle line:84% no longer signaling that you're part of the in-group. 00:46:38.310 --> 00:46:42.150 align:middle line:84% So that in-group itself has to invent new ways of language 00:46:42.150 --> 00:46:45.170 align:middle line:84% to show each other that you're within that in-group. 00:46:45.170 --> 00:46:47.910 align:middle line:84% And so that's an extreme generalization of this. 00:46:47.910 --> 00:46:49.590 align:middle line:84% But it's kind of why I started thinking 00:46:49.590 --> 00:46:51.830 align:middle line:84% the in-group, out-group dynamic was 00:46:51.830 --> 00:46:53.997 align:middle line:84% more interesting than like reseller profit dynamics. 00:46:53.997 --> 00:46:56.163 align:middle line:84% NELSON REPENNING: Yeah, I think this is fascinating. 00:46:56.163 --> 00:46:59.090 align:middle line:84% And I have children that are probably close to your age. 00:46:59.090 --> 00:47:00.590 align:middle line:84% And they get so mad at me when I try 00:47:00.590 --> 00:47:02.393 align:middle line:90% to appropriate their language. 00:47:02.393 --> 00:47:04.810 align:middle line:84% They're like, Dad, you're not using it right, or whatever. 00:47:04.810 --> 00:47:05.150 align:middle line:90% So-- 00:47:05.150 --> 00:47:06.170 align:middle line:84% STUDENT: What are you trying to say? 00:47:06.170 --> 00:47:07.503 align:middle line:90% NELSON REPENNING: Yeah, exactly. 00:47:07.503 --> 00:47:08.970 align:middle line:90% [LAUGHS] 00:47:08.970 --> 00:47:09.470 align:middle line:90% 00:47:09.470 --> 00:47:11.370 align:middle line:90% OK, so I get the phenomenon. 00:47:11.370 --> 00:47:16.430 align:middle line:84% But again, what's the question you're trying to answer here? 00:47:16.430 --> 00:47:18.870 align:middle line:84% And this is a hard question actually. 00:47:18.870 --> 00:47:20.950 align:middle line:84% STUDENT: Yeah, so initially, the idea 00:47:20.950 --> 00:47:25.720 align:middle line:84% was looking at it from brand manager perspective, 00:47:25.720 --> 00:47:30.940 align:middle line:84% how do you maintain your brand image? 00:47:30.940 --> 00:47:34.200 align:middle line:84% How do you make sure only the in-group or more in-group 00:47:34.200 --> 00:47:36.360 align:middle line:84% is getting your products, rather than allowing 00:47:36.360 --> 00:47:39.880 align:middle line:90% out-group people to get it? 00:47:39.880 --> 00:47:42.200 align:middle line:84% So there's one idea, which is just 00:47:42.200 --> 00:47:46.080 align:middle line:84% make sure out-group people never know that this brand exists. 00:47:46.080 --> 00:47:49.500 align:middle line:84% Like, that's true of a lot of super high luxury brands, 00:47:49.500 --> 00:47:52.200 align:middle line:84% like Brunello Cucinelli or Hermes 00:47:52.200 --> 00:47:53.900 align:middle line:84% before the whole Birkin thing blew up. 00:47:53.900 --> 00:47:57.740 align:middle line:84% They did not want average people to know that they existed, 00:47:57.740 --> 00:48:00.640 align:middle line:84% and that's how they maintained their in-group. 00:48:00.640 --> 00:48:02.260 align:middle line:84% But then with brands like Supreme, 00:48:02.260 --> 00:48:06.580 align:middle line:84% which like part of the allure is also that it's very scarce. 00:48:06.580 --> 00:48:08.330 align:middle line:84% And it's not just on price it's scarce, 00:48:08.330 --> 00:48:10.080 align:middle line:84% it's just that no one buys it because they 00:48:10.080 --> 00:48:12.440 align:middle line:90% make such huge quantities. 00:48:12.440 --> 00:48:15.760 align:middle line:84% Like, is it even possible to balance that hype 00:48:15.760 --> 00:48:17.708 align:middle line:90% with maintaining a few? 00:48:17.708 --> 00:48:20.000 align:middle line:84% NELSON REPENNING: OK, so let me just repeat back to you 00:48:20.000 --> 00:48:21.640 align:middle line:84% what I thought you said, which is 00:48:21.640 --> 00:48:24.850 align:middle line:84% that one question to which modeling could be directed 00:48:24.850 --> 00:48:29.070 align:middle line:84% is, I am the brand manager, or the CEO, of a small company. 00:48:29.070 --> 00:48:31.427 align:middle line:90% Like, how do I manage this? 00:48:31.427 --> 00:48:33.510 align:middle line:84% If I'm in a market that might have these dynamics, 00:48:33.510 --> 00:48:34.885 align:middle line:84% how do I manage it sustainably so 00:48:34.885 --> 00:48:38.247 align:middle line:84% that I don't get this crazy overshoot and collapse? 00:48:38.247 --> 00:48:38.830 align:middle line:90% STUDENT: Yeah. 00:48:38.830 --> 00:48:41.080 align:middle line:84% NELSON REPENNING: OK, I think that's a great question. 00:48:41.080 --> 00:48:43.210 align:middle line:84% And it requires a diagnosis, which 00:48:43.210 --> 00:48:46.970 align:middle line:84% is you need a theory, which it sounds like you already have, 00:48:46.970 --> 00:48:51.210 align:middle line:84% of this kind of life cycle of these hype brands 00:48:51.210 --> 00:48:53.910 align:middle line:84% where cool kids do it, and then the nerds start doing it, 00:48:53.910 --> 00:48:55.050 align:middle line:84% and then it ruins it for the cool kids, 00:48:55.050 --> 00:48:56.070 align:middle line:90% and so on, and so forth. 00:48:56.070 --> 00:48:58.070 align:middle line:84% But then I think there's a really clear question 00:48:58.070 --> 00:49:00.090 align:middle line:84% at the end, which is if I'm in that space, 00:49:00.090 --> 00:49:02.717 align:middle line:84% what combo of pricing, exclusivity, 00:49:02.717 --> 00:49:03.550 align:middle line:90% and stuff like that? 00:49:03.550 --> 00:49:04.425 align:middle line:90% I think that's great. 00:49:04.425 --> 00:49:08.850 align:middle line:84% And by the way, this is an age-old problem. 00:49:08.850 --> 00:49:10.870 align:middle line:84% When I was of sort of where you all are sitting, 00:49:10.870 --> 00:49:12.745 align:middle line:84% one of the companies that we worked with here 00:49:12.745 --> 00:49:15.090 align:middle line:84% was Harley-Davidson, the big motorcycle manufacturer. 00:49:15.090 --> 00:49:19.970 align:middle line:84% And through the 1970s, they had had a core user base that was 00:49:19.970 --> 00:49:24.438 align:middle line:84% largely blue-collar, relatively middle-income people. 00:49:24.438 --> 00:49:26.980 align:middle line:84% And during that time period, the quality of their motorcycles 00:49:26.980 --> 00:49:28.120 align:middle line:90% had been crap. 00:49:28.120 --> 00:49:32.160 align:middle line:84% And but their kind of core, sort of blue-collar user base 00:49:32.160 --> 00:49:33.660 align:middle line:84% had stuck with them, and the company 00:49:33.660 --> 00:49:34.880 align:middle line:90% had survived largely on them. 00:49:34.880 --> 00:49:38.540 align:middle line:84% And then in the '80s and sort of early '90s, suddenly, 00:49:38.540 --> 00:49:42.340 align:middle line:84% Harleys became the hot thing to have if you were like a dentist 00:49:42.340 --> 00:49:45.120 align:middle line:84% or a plastic surgeon in Southern California or whatever. 00:49:45.120 --> 00:49:46.580 align:middle line:90% And so they became super hot. 00:49:46.580 --> 00:49:49.740 align:middle line:84% So their demand just goes through the roof. 00:49:49.740 --> 00:49:51.900 align:middle line:84% But to their credit, Harley was really worried 00:49:51.900 --> 00:49:55.580 align:middle line:84% that they were basically pricing their original user base out 00:49:55.580 --> 00:49:56.640 align:middle line:90% of the market. 00:49:56.640 --> 00:49:58.540 align:middle line:84% And at some point, the plastic surgeons 00:49:58.540 --> 00:50:01.318 align:middle line:84% were going to switch to Ducatis, or sports cars, or whatever. 00:50:01.318 --> 00:50:02.860 align:middle line:84% And they were worried there was going 00:50:02.860 --> 00:50:06.500 align:middle line:90% to be no kind of core left. 00:50:06.500 --> 00:50:11.460 align:middle line:84% And so that actually was a super hard problem to manage. 00:50:11.460 --> 00:50:14.100 align:middle line:90% So that sounds great. 00:50:14.100 --> 00:50:16.637 align:middle line:84% So are there comments for-- is it Oriana? 00:50:16.637 --> 00:50:17.220 align:middle line:90% STUDENT: Yeah. 00:50:17.220 --> 00:50:18.540 align:middle line:84% NELSON REPENNING: Yeah, about the problem formulation 00:50:18.540 --> 00:50:20.498 align:middle line:84% and the setup before we go look at the-- sorry, 00:50:20.498 --> 00:50:23.190 align:middle line:90% my computer keeps turning off. 00:50:23.190 --> 00:50:24.810 align:middle line:84% Or suggestions for what she just said? 00:50:24.810 --> 00:50:27.810 align:middle line:84% STUDENT: The how question might be very hard. 00:50:27.810 --> 00:50:29.970 align:middle line:90% It might be a later-stage thing. 00:50:29.970 --> 00:50:34.710 align:middle line:84% I think it is impossible to not collapse. 00:50:34.710 --> 00:50:37.950 align:middle line:84% Like, is it possible to have a takeoff without a collapse 00:50:37.950 --> 00:50:40.270 align:middle line:90% at some point? 00:50:40.270 --> 00:50:41.230 align:middle line:90% STUDENT: Yeah. 00:50:41.230 --> 00:50:41.950 align:middle line:90% STUDENT: That's-- 00:50:41.950 --> 00:50:44.590 align:middle line:84% STUDENT: Yeah, because I guess Harley-Davidson still exists. 00:50:44.590 --> 00:50:45.858 align:middle line:90% I think Gucci managed-- 00:50:45.858 --> 00:50:48.150 align:middle line:84% NELSON REPENNING: They're a fraction of what they were. 00:50:48.150 --> 00:50:48.650 align:middle line:90% Yeah. 00:50:48.650 --> 00:50:50.990 align:middle line:84% STUDENT: Gucci managed to still survive 00:50:50.990 --> 00:50:55.970 align:middle line:84% after-- they had their own hype cycle, I don't know, 00:50:55.970 --> 00:50:58.250 align:middle line:90% six to eight years ago. 00:50:58.250 --> 00:50:59.090 align:middle line:90% They still exist. 00:50:59.090 --> 00:51:01.790 align:middle line:84% I guess I should look at what fraction they're at now compared 00:51:01.790 --> 00:51:06.290 align:middle line:84% to [INAUDIBLE] maybe cycle of this to see if is it possible 00:51:06.290 --> 00:51:06.790 align:middle line:90% to-- 00:51:06.790 --> 00:51:09.040 align:middle line:84% STUDENT: There might be something structural with this 00:51:09.040 --> 00:51:13.280 align:middle line:84% thing that you actually cannot become super popular-- 00:51:13.280 --> 00:51:15.030 align:middle line:84% STUDENT: Yeah, but the in-group will never 00:51:15.030 --> 00:51:17.230 align:middle line:90% fully return once it's lost. 00:51:17.230 --> 00:51:20.360 align:middle line:84% STUDENT: So that's just one version of the question. 00:51:20.360 --> 00:51:23.800 align:middle line:84% STUDENT: I don't know if there's something because they 00:51:23.800 --> 00:51:27.520 align:middle line:84% have a lot of back and forth between going public, and being 00:51:27.520 --> 00:51:31.000 align:middle line:84% private, and should be the shareholders or what. 00:51:31.000 --> 00:51:31.920 align:middle line:90% And I wonder if-- 00:51:31.920 --> 00:51:34.820 align:middle line:84% I mean, it's not a feedback story really, 00:51:34.820 --> 00:51:39.800 align:middle line:84% but I wonder if the ability to be disciplined on how much I'm 00:51:39.800 --> 00:51:42.920 align:middle line:84% buying into the hype that exists, 00:51:42.920 --> 00:51:46.960 align:middle line:84% how much I'm capitalizing on the hype that exists 00:51:46.960 --> 00:51:49.040 align:middle line:84% is a major factor in whether these companies are 00:51:49.040 --> 00:51:52.240 align:middle line:84% able to sustain their core base versus kind 00:51:52.240 --> 00:51:58.080 align:middle line:84% of being forced by market demands to alienate their base. 00:51:58.080 --> 00:51:59.880 align:middle line:90% STUDENT: Yeah. 00:51:59.880 --> 00:52:01.920 align:middle line:84% Yeah, there's a lot of interesting strategies 00:52:01.920 --> 00:52:06.520 align:middle line:84% that I think brands have developed to try and get 00:52:06.520 --> 00:52:09.580 align:middle line:84% as much profit out of people without alienating their base. 00:52:09.580 --> 00:52:12.240 align:middle line:84% I think Dior and some other brands 00:52:12.240 --> 00:52:15.690 align:middle line:84% have interesting dynamics, which is the really rich people who 00:52:15.690 --> 00:52:19.370 align:middle line:84% are the in-group that people are fantasizing that they are buying 00:52:19.370 --> 00:52:21.170 align:middle line:90% ready-to-wear and couture. 00:52:21.170 --> 00:52:23.430 align:middle line:84% And then if you can't afford that, 00:52:23.430 --> 00:52:25.870 align:middle line:84% you can buy their canvas tote bags. 00:52:25.870 --> 00:52:28.395 align:middle line:84% And if you can't afford that, you buy their fragrance line. 00:52:28.395 --> 00:52:29.770 align:middle line:84% And if you can't afford that, you 00:52:29.770 --> 00:52:31.253 align:middle line:90% buy their blushes [INAUDIBLE]. 00:52:31.253 --> 00:52:33.170 align:middle line:84% So like, I feel like that's kind of a good way 00:52:33.170 --> 00:52:36.530 align:middle line:84% that they like segment everyone and they still 00:52:36.530 --> 00:52:39.010 align:middle line:84% manage to keep the allure of being very classy, very 00:52:39.010 --> 00:52:41.690 align:middle line:84% well up brand, while if you can only 00:52:41.690 --> 00:52:44.550 align:middle line:84% afford to save up $50 for their blush, 00:52:44.550 --> 00:52:48.192 align:middle line:84% then you can still imagine that you're part of their consumer 00:52:48.192 --> 00:52:49.650 align:middle line:84% base if you can imagine that you're 00:52:49.650 --> 00:52:52.030 align:middle line:84% one of those people sitting in New York Fashion Week, 00:52:52.030 --> 00:52:54.550 align:middle line:90% even though you never will be. 00:52:54.550 --> 00:52:58.470 align:middle line:84% STUDENT: So diffusion length are a good story there. 00:52:58.470 --> 00:52:58.970 align:middle line:90% Like-- 00:52:58.970 --> 00:52:59.710 align:middle line:90% STUDENT: Yeah. 00:52:59.710 --> 00:53:02.450 align:middle line:84% But I think that would almost be an entirely different-- 00:53:02.450 --> 00:53:04.410 align:middle line:84% STUDENT: Well, that's a product category, 00:53:04.410 --> 00:53:07.050 align:middle line:84% but these brands also create diffusion lines like Marc 00:53:07.050 --> 00:53:09.450 align:middle line:90% by Marc Jacobs is everywhere. 00:53:09.450 --> 00:53:11.730 align:middle line:84% Like, yeah, those are, yeah, targeted 00:53:11.730 --> 00:53:16.160 align:middle line:84% at people who want the feeling but don't have the money. 00:53:16.160 --> 00:53:17.060 align:middle line:90% STUDENT: Yeah. 00:53:17.060 --> 00:53:18.020 align:middle line:90% So it's kind of-- 00:53:18.020 --> 00:53:19.980 align:middle line:84% STUDENT: And then [INAUDIBLE] people 00:53:19.980 --> 00:53:22.560 align:middle line:84% who are purchasing the kind of top thing. 00:53:22.560 --> 00:53:24.560 align:middle line:84% It's like, oh, that's the diffusion line. 00:53:24.560 --> 00:53:26.497 align:middle line:90% STUDENT: It doesn't affect the-- 00:53:26.497 --> 00:53:27.580 align:middle line:90% STUDENT: Yeah, it's just-- 00:53:27.580 --> 00:53:29.580 align:middle line:84% STUDENT: The caliber of the top ones. 00:53:29.580 --> 00:53:31.180 align:middle line:84% STUDENT: Yeah, so I guess then, that's 00:53:31.180 --> 00:53:36.057 align:middle line:84% the way that the managers are reacting to pressure like such-- 00:53:36.057 --> 00:53:37.140 align:middle line:90% STUDENT: Market pressures. 00:53:37.140 --> 00:53:40.380 align:middle line:84% STUDENT: Market pressures to sell without hopefully, 00:53:40.380 --> 00:53:42.200 align:middle line:84% like, destroying their core base. 00:53:42.200 --> 00:53:44.540 align:middle line:90% STUDENT: [INAUDIBLE] 00:53:44.540 --> 00:53:47.278 align:middle line:84% NELSON REPENNING: Yeah, and I mean, 00:53:47.278 --> 00:53:49.320 align:middle line:84% don't take this as contrary to my earlier advice. 00:53:49.320 --> 00:53:51.540 align:middle line:84% I think there's a bunch of other feedbacks 00:53:51.540 --> 00:53:53.660 align:middle line:90% you could conceivably add. 00:53:53.660 --> 00:53:56.780 align:middle line:84% I think this is a place where investors, and particularly 00:53:56.780 --> 00:54:00.920 align:middle line:84% Wall Street, can be a huge enemy to doing it correctly, 00:54:00.920 --> 00:54:04.700 align:middle line:84% because you often see these dramatic growth patterns. 00:54:04.700 --> 00:54:07.400 align:middle line:84% So I think we talk about this in the core class. 00:54:07.400 --> 00:54:08.920 align:middle line:90% But think about Peloton. 00:54:08.920 --> 00:54:11.400 align:middle line:84% So it's growing quickly, and then we have a pandemic, 00:54:11.400 --> 00:54:13.100 align:middle line:84% and it was like the go-to Christmas 00:54:13.100 --> 00:54:16.400 align:middle line:84% gift for every wealthy family in America, 00:54:16.400 --> 00:54:20.820 align:middle line:84% which then, just their share price has gone through the roof. 00:54:20.820 --> 00:54:23.220 align:middle line:84% But it was the classic boom and bust from John 00:54:23.220 --> 00:54:24.900 align:middle line:84% and Mark [INAUDIBLE] work from long ago. 00:54:24.900 --> 00:54:26.300 align:middle line:84% There just was no way that you could 00:54:26.300 --> 00:54:27.520 align:middle line:90% sustain that level of growth. 00:54:27.520 --> 00:54:32.720 align:middle line:84% But somehow, their expectations don't sort of account for that. 00:54:32.720 --> 00:54:35.020 align:middle line:84% So you can imagine owners or investors, 00:54:35.020 --> 00:54:37.380 align:middle line:84% like, see that sort of short-term payoff, which just 00:54:37.380 --> 00:54:39.798 align:middle line:90% makes all these dynamics worth. 00:54:39.798 --> 00:54:40.840 align:middle line:90% So I think this is great. 00:54:40.840 --> 00:54:42.923 align:middle line:84% So tell us a little bit about your model structure 00:54:42.923 --> 00:54:44.980 align:middle line:90% that we have here. 00:54:44.980 --> 00:54:46.620 align:middle line:84% STUDENT: There's no model structure. 00:54:46.620 --> 00:54:54.020 align:middle line:84% So the attempt was to, here are our out-group people, 00:54:54.020 --> 00:54:58.780 align:middle line:84% which these are like anyone who's in the out-group is 00:54:58.780 --> 00:55:00.520 align:middle line:90% these people. 00:55:00.520 --> 00:55:02.020 align:middle line:84% These are people who are interested. 00:55:02.020 --> 00:55:04.220 align:middle line:90% And the way I-- 00:55:04.220 --> 00:55:07.240 align:middle line:84% my variables are kind of vague, so it's not good. 00:55:07.240 --> 00:55:09.100 align:middle line:84% But interested kind of started doubling 00:55:09.100 --> 00:55:12.310 align:middle line:84% for people who would be willing to purchase, which 00:55:12.310 --> 00:55:14.910 align:middle line:84% gets complicated because there's two purchasing levels, which 00:55:14.910 --> 00:55:18.950 align:middle line:84% is retail purchase price and resale purchase price, which you 00:55:18.950 --> 00:55:21.790 align:middle line:84% may or may not be able to access depending on how in demand it 00:55:21.790 --> 00:55:22.290 align:middle line:90% is. 00:55:22.290 --> 00:55:23.972 align:middle line:84% So that was kind of its own issue. 00:55:23.972 --> 00:55:25.430 align:middle line:84% And then hypebeasts are here, which 00:55:25.430 --> 00:55:27.150 align:middle line:90% I don't know if I really need. 00:55:27.150 --> 00:55:29.790 align:middle line:84% But these are people who are just do not care about price, 00:55:29.790 --> 00:55:33.670 align:middle line:84% and will buy at any level, and are the reason that resellers 00:55:33.670 --> 00:55:35.948 align:middle line:84% can start jacking up prices to 10x retail. 00:55:35.948 --> 00:55:38.490 align:middle line:84% NELSON REPENNING: I think that's the best variable name ever. 00:55:38.490 --> 00:55:39.810 align:middle line:90% But-- [LAUGHS] 00:55:39.810 --> 00:55:39.977 align:middle line:90% 00:55:39.977 --> 00:55:40.810 align:middle line:90% STUDENT: Well, yeah. 00:55:40.810 --> 00:55:42.290 align:middle line:90% That's [INAUDIBLE]. 00:55:42.290 --> 00:55:43.565 align:middle line:90% And these are-- 00:55:43.565 --> 00:55:45.190 align:middle line:84% STUDENT: It's really called hypebeasts? 00:55:45.190 --> 00:55:46.898 align:middle line:84% STUDENT: Yeah, they're called hypebeasts. 00:55:46.898 --> 00:55:51.530 align:middle line:84% Like, people who buy brands just to be cool, but don't get it. 00:55:51.530 --> 00:55:53.230 align:middle line:90% They're called this. 00:55:53.230 --> 00:55:55.585 align:middle line:84% And then these are our in-group people, 00:55:55.585 --> 00:55:57.210 align:middle line:84% which is anyone who's in that in-group. 00:55:57.210 --> 00:56:00.270 align:middle line:84% So for Supreme, you could imagine skaters. 00:56:00.270 --> 00:56:03.070 align:middle line:84% And then these are people who are willing to purchase 00:56:03.070 --> 00:56:06.550 align:middle line:90% at whatever retail prices. 00:56:06.550 --> 00:56:08.600 align:middle line:84% And so they lose interest depending 00:56:08.600 --> 00:56:10.277 align:middle line:84% on how many alternatives are there. 00:56:10.277 --> 00:56:12.360 align:middle line:84% Are there other brands that can signal that you're 00:56:12.360 --> 00:56:14.380 align:middle line:90% a skater pretty easily? 00:56:14.380 --> 00:56:16.000 align:middle line:90% Or is Supreme the only one? 00:56:16.000 --> 00:56:18.380 align:middle line:84% And then also, perceived coolness. 00:56:18.380 --> 00:56:21.560 align:middle line:84% So for in-group, I think the perceived coolness drops really 00:56:21.560 --> 00:56:24.520 align:middle line:84% quickly the second a lot of these people start entering. 00:56:24.520 --> 00:56:26.840 align:middle line:84% And it's gained, like, dramatically slow. 00:56:26.840 --> 00:56:29.960 align:middle line:84% While on the other hand, perceived coolness 00:56:29.960 --> 00:56:31.960 align:middle line:84% by the out-group is immediately really high 00:56:31.960 --> 00:56:35.480 align:middle line:84% because it's at first essentially only 00:56:35.480 --> 00:56:37.960 align:middle line:84% in-group people, and then it cools down, 00:56:37.960 --> 00:56:40.600 align:middle line:84% or it loses coolness very, very, very slowly 00:56:40.600 --> 00:56:43.920 align:middle line:84% because they actually don't understand these people at all. 00:56:43.920 --> 00:56:48.320 align:middle line:84% And then this part was kind of attempting 00:56:48.320 --> 00:56:54.680 align:middle line:84% to start to understand how this demand and sellout happens. 00:56:54.680 --> 00:56:56.680 align:middle line:84% So this is, like, total interest. 00:56:56.680 --> 00:56:57.560 align:middle line:90% Oh. 00:56:57.560 --> 00:57:01.760 align:middle line:84% Then there's the exogenous quantity 00:57:01.760 --> 00:57:03.600 align:middle line:84% that the brand is willing to produce. 00:57:03.600 --> 00:57:06.470 align:middle line:84% Like, Supreme could start making a million shirts, 00:57:06.470 --> 00:57:09.850 align:middle line:84% but they choose to keep very limited drops. 00:57:09.850 --> 00:57:14.970 align:middle line:84% And so then surplus demand means that the resale price 00:57:14.970 --> 00:57:17.210 align:middle line:90% can get jacked up. 00:57:17.210 --> 00:57:21.450 align:middle line:84% And then as that gap increases, my thought process was, 00:57:21.450 --> 00:57:22.930 align:middle line:84% less and less in-group people are 00:57:22.930 --> 00:57:25.490 align:middle line:84% going to be able to buy it because they're either 00:57:25.490 --> 00:57:26.930 align:middle line:84% more price sensitive or they just 00:57:26.930 --> 00:57:31.530 align:middle line:84% have other things they'd rather be buying. 00:57:31.530 --> 00:57:36.210 align:middle line:84% And so that fuels a loop through perceived coolness eventually. 00:57:36.210 --> 00:57:39.630 align:middle line:84% And then I had, down in this corner, if you scroll, 00:57:39.630 --> 00:57:44.050 align:middle line:84% there's the case study kind of of my hypothesis of how-- 00:57:44.050 --> 00:57:47.427 align:middle line:84% [INAUDIBLE] Oh, why does it look like that? 00:57:47.427 --> 00:57:49.010 align:middle line:84% Well, it used to look better, but this 00:57:49.010 --> 00:57:54.850 align:middle line:84% is kind of like my hypothesis of how the Supreme case should go. 00:57:54.850 --> 00:57:57.290 align:middle line:84% NELSON REPENNING: Do you have data from Supreme? 00:57:57.290 --> 00:57:57.930 align:middle line:90% STUDENT: No. 00:57:57.930 --> 00:57:58.850 align:middle line:90% NELSON REPENNING: OK. 00:57:58.850 --> 00:58:00.550 align:middle line:84% STUDENT: But so I need to get that. 00:58:00.550 --> 00:58:03.070 align:middle line:90% But this is the general idea. 00:58:03.070 --> 00:58:09.260 align:middle line:84% I think maybe a way I was thinking about modeling interest 00:58:09.260 --> 00:58:11.900 align:middle line:84% maybe could be just like, how often does like the name, 00:58:11.900 --> 00:58:15.500 align:middle line:84% "Supreme" come up on TikTok reels, articles, 00:58:15.500 --> 00:58:18.620 align:middle line:90% might be a good place to start. 00:58:18.620 --> 00:58:19.620 align:middle line:90% NELSON REPENNING: Yeah. 00:58:19.620 --> 00:58:21.820 align:middle line:84% STUDENT: But that's the general idea. 00:58:21.820 --> 00:58:24.140 align:middle line:90% NELSON REPENNING: All right. 00:58:24.140 --> 00:58:24.700 align:middle line:90% Suggestions? 00:58:24.700 --> 00:58:27.355 align:middle line:90% Comments? 00:58:27.355 --> 00:58:28.980 align:middle line:84% STUDENT: Yeah, I don't really know what 00:58:28.980 --> 00:58:31.580 align:middle line:90% my stock and flow should be. 00:58:31.580 --> 00:58:36.000 align:middle line:84% Or the whole pricing thing is a bit confusing. 00:58:36.000 --> 00:58:40.100 align:middle line:90% 00:58:40.100 --> 00:58:43.460 align:middle line:84% STUDENT: So in-group relevance, or getting cooler 00:58:43.460 --> 00:58:49.500 align:middle line:84% for the in-group, and sort of in-group relevance, which is a-- 00:58:49.500 --> 00:58:52.922 align:middle line:84% STUDENT: In-group relevance is essentially just like, 00:58:52.922 --> 00:58:55.380 align:middle line:84% out of everyone who owns the brand, what percentage of them 00:58:55.380 --> 00:58:56.980 align:middle line:90% are part of it? 00:58:56.980 --> 00:59:00.260 align:middle line:84% If it's like 100%, then it's super cool. 00:59:00.260 --> 00:59:06.073 align:middle line:84% If it's 50%, it's in the not-cool [INAUDIBLE]. 00:59:06.073 --> 00:59:10.070 align:middle line:90% 00:59:10.070 --> 00:59:13.070 align:middle line:84% NELSON REPENNING: So this is for all of you. 00:59:13.070 --> 00:59:16.910 align:middle line:84% What do you think is the biggest sort of crossroads here in terms 00:59:16.910 --> 00:59:19.190 align:middle line:84% of modeling assumptions when getting started 00:59:19.190 --> 00:59:21.990 align:middle line:90% with this formulation? 00:59:21.990 --> 00:59:23.112 align:middle line:90% STUDENT: [INAUDIBLE] 00:59:23.112 --> 00:59:24.070 align:middle line:90% NELSON REPENNING: What? 00:59:24.070 --> 00:59:25.690 align:middle line:84% STUDENT: I can't tell if this is [INAUDIBLE]. 00:59:25.690 --> 00:59:26.523 align:middle line:90% It's a huge product. 00:59:26.523 --> 00:59:29.927 align:middle line:90% 00:59:29.927 --> 00:59:30.510 align:middle line:90% STUDENT: Yeah. 00:59:30.510 --> 00:59:31.730 align:middle line:90% STUDENT: Wait, am I not-- 00:59:31.730 --> 00:59:32.770 align:middle line:90% there's no formulas. 00:59:32.770 --> 00:59:33.590 align:middle line:90% So I don't know [INAUDIBLE]. 00:59:33.590 --> 00:59:35.423 align:middle line:84% STUDENT: Yeah, but just your stock and flow. 00:59:35.423 --> 00:59:39.030 align:middle line:84% Like, you're interested in your buyers' potential buyers, 00:59:39.030 --> 00:59:41.190 align:middle line:90% seem separate from hypebeasts. 00:59:41.190 --> 00:59:41.810 align:middle line:90% STUDENT: Yeah. 00:59:41.810 --> 00:59:43.310 align:middle line:84% STUDENT: Hypebeasts are [INAUDIBLE]. 00:59:43.310 --> 00:59:45.245 align:middle line:84% STUDENT: These are entirely out-group. 00:59:45.245 --> 00:59:46.370 align:middle line:90% STUDENT: They're out-group? 00:59:46.370 --> 00:59:47.038 align:middle line:90% STUDENT: Yeah. 00:59:47.038 --> 00:59:48.330 align:middle line:90% STUDENT: So I don't understand. 00:59:48.330 --> 00:59:49.788 align:middle line:84% STUDENT: Hypebeasts are just people 00:59:49.788 --> 00:59:54.790 align:middle line:84% who are super obsessed with the brand kind of just 00:59:54.790 --> 00:59:57.852 align:middle line:84% to buy the brand because they can buy the brand and it's cool. 00:59:57.852 --> 00:59:59.810 align:middle line:84% And they, like, will spend any amount of money. 00:59:59.810 --> 01:00:01.685 align:middle line:84% These are the people who spend, I don't know, 01:00:01.685 --> 01:00:05.320 align:middle line:84% $1,000 for a crowbar with the word, "Supreme", on it. 01:00:05.320 --> 01:00:06.720 align:middle line:90% Those are hypebeasts. 01:00:06.720 --> 01:00:08.175 align:middle line:90% STUDENT: OK. 01:00:08.175 --> 01:00:09.300 align:middle line:90% STUDENT: I have a question. 01:00:09.300 --> 01:00:14.600 align:middle line:84% So this is about Supreme, and then the in-group 01:00:14.600 --> 01:00:17.780 align:middle line:84% is defined by participating in like skater culture. 01:00:17.780 --> 01:00:24.900 align:middle line:84% So their ability to pay the higher resale prices is limited. 01:00:24.900 --> 01:00:30.440 align:middle line:90% 01:00:30.440 --> 01:00:33.440 align:middle line:90% And so it's driven by coolness. 01:00:33.440 --> 01:00:35.985 align:middle line:84% They don't want to buy the thing because it's uncool. 01:00:35.985 --> 01:00:38.360 align:middle line:84% But also, they can't buy the thing if it's too expensive, 01:00:38.360 --> 01:00:43.600 align:middle line:84% and it sells out, and they can get it-- the next price later. 01:00:43.600 --> 01:00:47.520 align:middle line:84% But for luxury brands, if we're talking about buzz, 01:00:47.520 --> 01:00:50.560 align:middle line:84% the in-group is everybody who's kind 01:00:50.560 --> 01:00:54.692 align:middle line:84% in a social group with a ton of money anyway. 01:00:54.692 --> 01:00:56.900 align:middle line:84% So they're not going to get priced out of the market. 01:00:56.900 --> 01:00:59.226 align:middle line:84% It's only going to be the coolest anyway. 01:00:59.226 --> 01:01:00.758 align:middle line:90% STUDENT: The price there is-- 01:01:00.758 --> 01:01:03.050 align:middle line:84% yeah, I was thinking about that, which is it's probably 01:01:03.050 --> 01:01:04.390 align:middle line:90% more easy alternatives. 01:01:04.390 --> 01:01:06.730 align:middle line:84% And how much is it signaling that you're 01:01:06.730 --> 01:01:07.970 align:middle line:90% part of the in-group? 01:01:07.970 --> 01:01:12.070 align:middle line:84% I do think depending on if this model was made correctly, 01:01:12.070 --> 01:01:16.290 align:middle line:84% I do think changing price sensitivity 01:01:16.290 --> 01:01:18.750 align:middle line:90% for both should like change. 01:01:18.750 --> 01:01:23.310 align:middle line:84% Like for example, for Birkins, it's just also a $10,000, 01:01:23.310 --> 01:01:25.050 align:middle line:90% $50,000, $100,000 bag. 01:01:25.050 --> 01:01:27.550 align:middle line:84% You just have less people who could even buy that. 01:01:27.550 --> 01:01:30.970 align:middle line:84% And so then the boom-and-bust cycle has to be less 01:01:30.970 --> 01:01:33.410 align:middle line:84% because rather than a, I don't know, 01:01:33.410 --> 01:01:38.530 align:middle line:84% $70 t-shirt that a lot of people could reasonably save up to buy, 01:01:38.530 --> 01:01:41.250 align:middle line:84% you're not saving up for a $100,000 bag. 01:01:41.250 --> 01:01:42.930 align:middle line:90% And so I think that would-- 01:01:42.930 --> 01:01:45.850 align:middle line:84% the difference would just be if the price sensitivity for both 01:01:45.850 --> 01:01:48.710 align:middle line:84% was pretty much 0, like they'll pay whatever, 01:01:48.710 --> 01:01:53.530 align:middle line:84% then I think in a good model, that cycle should just 01:01:53.530 --> 01:01:55.282 align:middle line:90% be temporary there. 01:01:55.282 --> 01:01:55.990 align:middle line:90% STUDENT: I agree. 01:01:55.990 --> 01:02:00.220 align:middle line:84% I think in the spirit of minimum, 01:02:00.220 --> 01:02:02.980 align:middle line:84% maybe what about taking out the price dynamic 01:02:02.980 --> 01:02:07.472 align:middle line:84% and just thinking about in-group, out-group coolness? 01:02:07.472 --> 01:02:12.300 align:middle line:84% Rather than these kind of slow-to-learn out-group people 01:02:12.300 --> 01:02:18.060 align:middle line:84% can keep afloat an artificial sales cycle 01:02:18.060 --> 01:02:20.100 align:middle line:84% long after the in-group has [INAUDIBLE]. 01:02:20.100 --> 01:02:22.180 align:middle line:90% STUDENT: I did not make sense. 01:02:22.180 --> 01:02:25.700 align:middle line:84% STUDENT: And then maybe the price comes back in-- 01:02:25.700 --> 01:02:27.880 align:middle line:84% STUDENT: After this is worked out. 01:02:27.880 --> 01:02:29.300 align:middle line:90% Yeah. 01:02:29.300 --> 01:02:31.540 align:middle line:84% NELSON REPENNING: So one thing to be very careful of, 01:02:31.540 --> 01:02:33.915 align:middle line:84% and this is an easy mistake to make that people make over 01:02:33.915 --> 01:02:38.340 align:middle line:84% and over again, is if in that stock flow 01:02:38.340 --> 01:02:41.460 align:middle line:84% chain of the out-group interested in hypebeasts, 01:02:41.460 --> 01:02:43.880 align:middle line:84% if they're in the same stock flow chain, 01:02:43.880 --> 01:02:45.700 align:middle line:84% you have to assume that they all have 01:02:45.700 --> 01:02:48.460 align:middle line:90% the same average attributes. 01:02:48.460 --> 01:02:52.460 align:middle line:84% So you can't mix people that have more or less price 01:02:52.460 --> 01:02:54.160 align:middle line:90% sensitivity in there. 01:02:54.160 --> 01:02:57.598 align:middle line:84% And this is a mistake people make all the time, 01:02:57.598 --> 01:02:59.890 align:middle line:84% and they try to work their way around it with co-flows. 01:02:59.890 --> 01:03:03.530 align:middle line:90% But it never works. 01:03:03.530 --> 01:03:07.470 align:middle line:84% STUDENT: [INAUDIBLE] It's like average out-group people, 01:03:07.470 --> 01:03:08.147 align:middle line:90% wealthy people. 01:03:08.147 --> 01:03:10.230 align:middle line:84% NELSON REPENNING: Well, so where I was going with, 01:03:10.230 --> 01:03:11.470 align:middle line:84% I think the biggest assumption you've 01:03:11.470 --> 01:03:14.137 align:middle line:84% made, which I'm fine with, it's just I want you to be clear what 01:03:14.137 --> 01:03:16.570 align:middle line:84% you're doing, is you have now, to me, 01:03:16.570 --> 01:03:17.950 align:middle line:84% the first kind of analytic divide 01:03:17.950 --> 01:03:21.735 align:middle line:84% is you have now differentiated between sort of cool kids 01:03:21.735 --> 01:03:22.610 align:middle line:90% and out-group, right? 01:03:22.610 --> 01:03:23.930 align:middle line:90% And that's fine. 01:03:23.930 --> 01:03:25.590 align:middle line:84% But I actually think there's a way 01:03:25.590 --> 01:03:28.990 align:middle line:84% to build this model where you have just one chain, 01:03:28.990 --> 01:03:34.700 align:middle line:84% and the in-group out-group actually emerges endogenously. 01:03:34.700 --> 01:03:36.450 align:middle line:84% That may or may not be a good thing to do, 01:03:36.450 --> 01:03:38.250 align:middle line:84% but I just, again, I want you to be clear. 01:03:38.250 --> 01:03:40.550 align:middle line:84% So for now, I think it's not a bad assumption 01:03:40.550 --> 01:03:42.650 align:middle line:84% to say, look, simplest formulation. 01:03:42.650 --> 01:03:44.590 align:middle line:84% We're going to start with two kinds of people. 01:03:44.590 --> 01:03:49.350 align:middle line:84% We've got cool kids and we've got everybody else. 01:03:49.350 --> 01:03:52.630 align:middle line:84% And then I think in terms of standard formulations, what 01:03:52.630 --> 01:03:54.990 align:middle line:84% you need to then sort of formulate for these two 01:03:54.990 --> 01:03:56.910 align:middle line:84% different groups, and it kind of goes 01:03:56.910 --> 01:04:00.270 align:middle line:84% back to where I was headed with Cathy's idea is, 01:04:00.270 --> 01:04:03.550 align:middle line:84% you've got to sort of formulate a model of their demand 01:04:03.550 --> 01:04:05.630 align:middle line:90% for these items. 01:04:05.630 --> 01:04:06.282 align:middle line:90% Goddamn this. 01:04:06.282 --> 01:04:07.490 align:middle line:90% I got to turn this thing off. 01:04:07.490 --> 01:04:07.990 align:middle line:90% Sorry. 01:04:07.990 --> 01:04:12.140 align:middle line:90% 01:04:12.140 --> 01:04:14.390 align:middle line:84% And the reason you're distinguishing between these two 01:04:14.390 --> 01:04:16.590 align:middle line:84% groups is you have a hypothesis that they 01:04:16.590 --> 01:04:18.885 align:middle line:84% have very different demand sort of functions 01:04:18.885 --> 01:04:20.010 align:middle line:90% in terms of what they want. 01:04:20.010 --> 01:04:23.830 align:middle line:84% Basically, the nerds, or the hypebeasts 01:04:23.830 --> 01:04:25.667 align:middle line:84% are looking at what the cool kids are doing. 01:04:25.667 --> 01:04:27.250 align:middle line:84% And the cool kids are, to some degree, 01:04:27.250 --> 01:04:30.250 align:middle line:84% looking at what the nerds are not doing in some sense. 01:04:30.250 --> 01:04:33.050 align:middle line:84% So I think, again, back to what they were doing before, 01:04:33.050 --> 01:04:35.467 align:middle line:84% what we were talking with her, like you need to figure out 01:04:35.467 --> 01:04:39.430 align:middle line:84% in a very simple way, what is the attractiveness or the appeal 01:04:39.430 --> 01:04:43.190 align:middle line:90% of this particular-- 01:04:43.190 --> 01:04:46.270 align:middle line:84% I assume this is a generic product from Supreme 01:04:46.270 --> 01:04:47.948 align:middle line:90% or whatever it is. 01:04:47.948 --> 01:04:50.240 align:middle line:84% And I think you could get, using standard formulations, 01:04:50.240 --> 01:04:53.518 align:middle line:84% a very sort of stripped-down model. 01:04:53.518 --> 01:04:55.560 align:middle line:84% The other thing is a kind of standard formulation 01:04:55.560 --> 01:04:58.840 align:middle line:84% is I do imagine you need some kind of diffusion style 01:04:58.840 --> 01:04:59.800 align:middle line:90% introduction. 01:04:59.800 --> 01:05:04.100 align:middle line:84% So when they offer a t-shirt or whatever it is, 01:05:04.100 --> 01:05:05.920 align:middle line:84% it's going to go through some Bass style 01:05:05.920 --> 01:05:07.003 align:middle line:90% kind of diffusion process. 01:05:07.003 --> 01:05:09.670 align:middle line:84% But then I think this is really interesting because there's sort 01:05:09.670 --> 01:05:11.780 align:middle line:90% of two diffusions going on. 01:05:11.780 --> 01:05:13.720 align:middle line:84% And you have to think about how they interact 01:05:13.720 --> 01:05:15.262 align:middle line:84% with each other, which would actually 01:05:15.262 --> 01:05:19.060 align:middle line:84% be an interesting little problem I have to think about. 01:05:19.060 --> 01:05:21.720 align:middle line:84% But I think where I would start is I think the stock flow 01:05:21.720 --> 01:05:23.160 align:middle line:90% chains are pretty-- 01:05:23.160 --> 01:05:26.375 align:middle line:84% I don't even know if you need three stocks for the-- 01:05:26.375 --> 01:05:29.000 align:middle line:84% you could probably start with, I just have two types of people, 01:05:29.000 --> 01:05:31.000 align:middle line:84% and they're either buyers, they're 01:05:31.000 --> 01:05:32.540 align:middle line:90% either owners, or not owners. 01:05:32.540 --> 01:05:34.302 align:middle line:90% So just two on both sides. 01:05:34.302 --> 01:05:35.760 align:middle line:84% And then I think the tricky part is 01:05:35.760 --> 01:05:37.520 align:middle line:84% to formulate-- the interesting part is 01:05:37.520 --> 01:05:41.600 align:middle line:84% to formulate the two demand functions for the two groups. 01:05:41.600 --> 01:05:43.640 align:middle line:84% And that's really where the two groups 01:05:43.640 --> 01:05:47.240 align:middle line:84% will differ in this context because your cool kids will 01:05:47.240 --> 01:05:49.490 align:middle line:84% basically be looking at some attribute 01:05:49.490 --> 01:05:52.630 align:middle line:84% and making sure that the non-cool kids don't have it. 01:05:52.630 --> 01:05:55.590 align:middle line:84% And then the non-cool kids will sort of be doing the opposite. 01:05:55.590 --> 01:05:58.310 align:middle line:84% And I think from there, you can get sort of the core dynamic. 01:05:58.310 --> 01:05:59.935 align:middle line:84% And then you can start to layer in some 01:05:59.935 --> 01:06:02.590 align:middle line:90% of these other feedbacks. 01:06:02.590 --> 01:06:05.612 align:middle line:84% But again, kind of on what is becoming the theme for the day, 01:06:05.612 --> 01:06:07.570 align:middle line:84% I would do incrementally, sort of feedback loop 01:06:07.570 --> 01:06:08.278 align:middle line:90% by feedback loop. 01:06:08.278 --> 01:06:13.610 align:middle line:84% STUDENT: So look for stocks just like [INAUDIBLE]? 01:06:13.610 --> 01:06:16.290 align:middle line:84% NELSON REPENNING: I think that's the sparest sort of formulation 01:06:16.290 --> 01:06:18.917 align:middle line:90% you could come up with. 01:06:18.917 --> 01:06:20.250 align:middle line:90% Again, I have to think about it. 01:06:20.250 --> 01:06:21.875 align:middle line:84% I don't know if anybody's in instincts. 01:06:21.875 --> 01:06:23.890 align:middle line:84% There is a sort of diffusion process 01:06:23.890 --> 01:06:25.770 align:middle line:84% that's going, where both of these groups 01:06:25.770 --> 01:06:27.650 align:middle line:90% are learning about it. 01:06:27.650 --> 01:06:29.730 align:middle line:84% And you have to think a little bit 01:06:29.730 --> 01:06:32.430 align:middle line:84% about who's looking at whom in terms of this. 01:06:32.430 --> 01:06:34.338 align:middle line:90% The cool kids are probably-- 01:06:34.338 --> 01:06:36.130 align:middle line:84% if you think of the standard Bass diffusion 01:06:36.130 --> 01:06:41.170 align:middle line:84% model or simple epidemic model, first, I have to learn of it. 01:06:41.170 --> 01:06:43.890 align:middle line:84% OK, Supreme's got a new t-shirt or something like that. 01:06:43.890 --> 01:06:47.140 align:middle line:84% And then I have to form, what's the appeal to me? 01:06:47.140 --> 01:06:51.440 align:middle line:84% And so I guess for the cool kids, well, you tell me. 01:06:51.440 --> 01:06:54.520 align:middle line:84% But I imagine that they look at the other cool kids. 01:06:54.520 --> 01:06:57.480 align:middle line:84% And the more the cool kids have it, that increases their demand. 01:06:57.480 --> 01:06:59.637 align:middle line:84% But then when I see middle-aged people 01:06:59.637 --> 01:07:01.720 align:middle line:84% like me walking down the street with that t-shirt, 01:07:01.720 --> 01:07:03.645 align:middle line:90% that's probably a negative. 01:07:03.645 --> 01:07:05.020 align:middle line:84% So you'll have to think carefully 01:07:05.020 --> 01:07:08.380 align:middle line:84% about how to formulate the relative attractiveness of it 01:07:08.380 --> 01:07:11.240 align:middle line:84% because I think that's going to be where the action is for this. 01:07:11.240 --> 01:07:14.460 align:middle line:84% I think it's a great problem actually. 01:07:14.460 --> 01:07:17.620 align:middle line:84% And I think this is a case where actually, 01:07:17.620 --> 01:07:21.310 align:middle line:84% to the extent that you can start with a fairly specific example, 01:07:21.310 --> 01:07:23.060 align:middle line:84% so I think it's good you picked one brand, 01:07:23.060 --> 01:07:25.320 align:middle line:84% and even with one product in the brand. 01:07:25.320 --> 01:07:29.003 align:middle line:84% And you may want to make this a more generic treatment 01:07:29.003 --> 01:07:30.420 align:middle line:84% at some point, but I think it will 01:07:30.420 --> 01:07:35.420 align:middle line:84% be easier to sort of kind of model formulations 01:07:35.420 --> 01:07:39.393 align:middle line:84% if you're focused on, OK, I want to look at t-shirts 01:07:39.393 --> 01:07:40.560 align:middle line:90% from this brand or whatever. 01:07:40.560 --> 01:07:42.060 align:middle line:84% And back to Cathy, I think you're 01:07:42.060 --> 01:07:46.510 align:middle line:84% smart to focus on just books rather than-- you may extend it 01:07:46.510 --> 01:07:49.170 align:middle line:84% someday to music, and movies, or whatever. 01:07:49.170 --> 01:07:50.830 align:middle line:84% But I think for now, it's actually 01:07:50.830 --> 01:07:52.710 align:middle line:84% having a specific case will be more helpful 01:07:52.710 --> 01:07:56.500 align:middle line:84% rather than trying to kind of build the grand, perfect sort 01:07:56.500 --> 01:07:59.630 align:middle line:90% of universal model. 01:07:59.630 --> 01:08:00.930 align:middle line:90% Other suggestions for this? 01:08:00.930 --> 01:08:02.530 align:middle line:90% I think this is a great choice. 01:08:02.530 --> 01:08:06.328 align:middle line:90% 01:08:06.328 --> 01:08:08.370 align:middle line:84% And you haven't written any equations yet, right? 01:08:08.370 --> 01:08:13.630 align:middle line:84% So it's just a stock flow structure. 01:08:13.630 --> 01:08:17.229 align:middle line:84% Yeah, and I would say too, I think the resale market is also 01:08:17.229 --> 01:08:18.670 align:middle line:90% super interesting. 01:08:18.670 --> 01:08:20.390 align:middle line:84% Again, as a matter of strategy, I 01:08:20.390 --> 01:08:22.870 align:middle line:84% would get sort of the core dynamic with just these two 01:08:22.870 --> 01:08:25.149 align:middle line:84% groups and sort of the life cycle 01:08:25.149 --> 01:08:31.380 align:middle line:84% first, and then probably add that resale 01:08:31.380 --> 01:08:33.782 align:middle line:84% sort of feedback or stock flow structure in later. 01:08:33.782 --> 01:08:35.990 align:middle line:84% I mean, doesn't this happen in sneakers all the time, 01:08:35.990 --> 01:08:40.350 align:middle line:84% that there's this sort of secondary market for hundreds 01:08:40.350 --> 01:08:43.439 align:middle line:84% of dollars, and multiple markups, and stuff like that? 01:08:43.439 --> 01:08:46.680 align:middle line:90% I think it's a fascinating area. 01:08:46.680 --> 01:08:52.520 align:middle line:84% STUDENT: Would you be able to have a simple structure that 01:08:52.520 --> 01:08:59.819 align:middle line:84% can showcase both behaviors of, [INAUDIBLE] screen thrasher? 01:08:59.819 --> 01:09:02.680 align:middle line:84% Like thrasher feels like it hasn't really hit 01:09:02.680 --> 01:09:03.460 align:middle line:90% people as much. 01:09:03.460 --> 01:09:08.040 align:middle line:84% It's still sort of [INAUDIBLE] constant demand on group, 01:09:08.040 --> 01:09:10.580 align:middle line:84% whereas Supreme does hit out-group, 01:09:10.580 --> 01:09:12.859 align:middle line:84% and then suddenly, it's [INAUDIBLE]. 01:09:12.859 --> 01:09:15.680 align:middle line:84% And then, you can just showcase it. 01:09:15.680 --> 01:09:18.460 align:middle line:84% What happens to have it be sustainable? 01:09:18.460 --> 01:09:21.200 align:middle line:84% And what happens to have it definitely crash? 01:09:21.200 --> 01:09:26.080 align:middle line:84% And then all the management group goes away. 01:09:26.080 --> 01:09:29.941 align:middle line:84% Then it's already pretty cool on [INAUDIBLE]. 01:09:29.941 --> 01:09:31.399 align:middle line:84% NELSON REPENNING: Yeah, and I think 01:09:31.399 --> 01:09:36.600 align:middle line:84% there's some really interesting policy questions, which is if, 01:09:36.600 --> 01:09:40.562 align:middle line:84% under a certain set of conditions, if basically 01:09:40.562 --> 01:09:42.729 align:middle line:84% the brand is going to kill itself because it's going 01:09:42.729 --> 01:09:45.396 align:middle line:84% to become too popular, and then the cool kids don't want it, OK, 01:09:45.396 --> 01:09:48.330 align:middle line:90% so I need to restrict It. 01:09:48.330 --> 01:09:51.450 align:middle line:84% Well, pricing will have some desirable properties 01:09:51.450 --> 01:09:53.689 align:middle line:84% and probably some negative properties basically 01:09:53.689 --> 01:09:56.470 align:middle line:84% depending on the in-group versus the out-groups. 01:09:56.470 --> 01:09:59.330 align:middle line:84% I mean, you could end up with a really interesting story. 01:09:59.330 --> 01:10:02.560 align:middle line:84% If the in-group are also the wealthy people, 01:10:02.560 --> 01:10:05.070 align:middle line:84% like Kardashians or something like that, 01:10:05.070 --> 01:10:07.810 align:middle line:84% that probably leads to a very specific set of strategies that 01:10:07.810 --> 01:10:09.118 align:middle line:90% goes with the Birkin thing. 01:10:09.118 --> 01:10:11.410 align:middle line:84% Basically, we're just going to price everybody else out 01:10:11.410 --> 01:10:13.535 align:middle line:84% of the market, and we have this super luxury brand. 01:10:13.535 --> 01:10:16.610 align:middle line:84% And then maybe, you have the kind of multi-level thing 01:10:16.610 --> 01:10:20.330 align:middle line:84% where we then create other versions. 01:10:20.330 --> 01:10:22.690 align:middle line:84% But if the cool kids are actually on the lower income 01:10:22.690 --> 01:10:27.930 align:middle line:84% side, then it probably becomes a much thornier sort of problem. 01:10:27.930 --> 01:10:30.490 align:middle line:84% So then how do you restrict demand in other ways 01:10:30.490 --> 01:10:31.970 align:middle line:90% besides pricing is-- 01:10:31.970 --> 01:10:35.050 align:middle line:84% STUDENT: I know Supreme kind of-- what-- kind of 01:10:35.050 --> 01:10:38.090 align:middle line:84% failed by just trying to make itself like an extreme luxury 01:10:38.090 --> 01:10:38.590 align:middle line:90% brand. 01:10:38.590 --> 01:10:43.740 align:middle line:84% It [INAUDIBLE] like $10,000 bags or whatnot. 01:10:43.740 --> 01:10:45.420 align:middle line:84% So I don't think we're [INAUDIBLE]. 01:10:45.420 --> 01:10:47.003 align:middle line:84% NELSON REPENNING: Well, and I imagine, 01:10:47.003 --> 01:10:51.060 align:middle line:84% and you could even sort of show this in your analysis. 01:10:51.060 --> 01:10:54.040 align:middle line:84% It's got to be enormously seductive to do the wrong thing. 01:10:54.040 --> 01:10:56.080 align:middle line:84% So if I start this low-end brand, 01:10:56.080 --> 01:10:59.380 align:middle line:84% suddenly, all of Hollywood is wearing my sneakers 01:10:59.380 --> 01:11:01.160 align:middle line:84% or they're in the Academy Awards, 01:11:01.160 --> 01:11:05.300 align:middle line:84% like swag bag or whatever, there is an enormous payday that you 01:11:05.300 --> 01:11:07.238 align:middle line:90% could probably capitalize on. 01:11:07.238 --> 01:11:08.780 align:middle line:84% But then you could probably also kill 01:11:08.780 --> 01:11:14.380 align:middle line:84% the goose that laying the golden egg pretty quickly. 01:11:14.380 --> 01:11:18.020 align:middle line:84% It's ancient at this point, but if you read the original version 01:11:18.020 --> 01:11:21.260 align:middle line:84% of Malcolm Gladwell's Tipping Point, 01:11:21.260 --> 01:11:23.780 align:middle line:84% the book is probably older than you are, 01:11:23.780 --> 01:11:25.760 align:middle line:84% there's a whole chapter on hush puppies, 01:11:25.760 --> 01:11:28.945 align:middle line:84% which were these dorky shoes that suddenly became cool again. 01:11:28.945 --> 01:11:30.820 align:middle line:84% And like, so there's a nice little case study 01:11:30.820 --> 01:11:34.100 align:middle line:84% that you could sort of grab there about-- 01:11:34.100 --> 01:11:36.140 align:middle line:84% and the same thing happened with Vans. 01:11:36.140 --> 01:11:39.190 align:middle line:84% Vans sneakers were, like, orthopedic sneakers and then 01:11:39.190 --> 01:11:40.490 align:middle line:90% became popular with skaters. 01:11:40.490 --> 01:11:41.830 align:middle line:90% And I don't know. 01:11:41.830 --> 01:11:43.290 align:middle line:90% They seem to still survive. 01:11:43.290 --> 01:11:45.458 align:middle line:84% So they may have figured out a way to do it. 01:11:45.458 --> 01:11:47.250 align:middle line:84% But I would focus on your Supreme case now. 01:11:47.250 --> 01:11:49.550 align:middle line:84% But once you've kind of got your story, 01:11:49.550 --> 01:11:52.690 align:middle line:84% then if you were going to kind of continue with this, 01:11:52.690 --> 01:11:55.630 align:middle line:84% you could then use your model with different parameters 01:11:55.630 --> 01:11:57.650 align:middle line:84% to explain, this is what happened to Vans, 01:11:57.650 --> 01:11:59.950 align:middle line:84% this is what happened with this and that. 01:11:59.950 --> 01:12:04.630 align:middle line:84% And it could be a really interesting piece. 01:12:04.630 --> 01:12:05.768 align:middle line:90% OK, should we take a break? 01:12:05.768 --> 01:12:07.810 align:middle line:84% Do you guys normally take a break for this class? 01:12:07.810 --> 01:12:09.430 align:middle line:90% OK, what's the normal-- 01:12:09.430 --> 01:12:10.270 align:middle line:90% STUDENT: 10 minutes. 01:12:10.270 --> 01:12:10.850 align:middle line:90% NELSON REPENNING: 10 minutes? 01:12:10.850 --> 01:12:11.190 align:middle line:90% OK. 01:12:11.190 --> 01:12:11.950 align:middle line:90% STUDENT: 40 maybe 01:12:11.950 --> 01:12:15.390 align:middle line:84% NELSON REPENNING: Yeah, 10:40 and then plus the usual MIT 01:12:15.390 --> 01:12:16.170 align:middle line:90% discount. 01:12:16.170 --> 01:12:19.150 align:middle line:90% 01:12:19.150 --> 01:12:22.110 align:middle line:84% So we're probably not to get to everyone, given that not 01:12:22.110 --> 01:12:23.330 align:middle line:90% everyone posted their model. 01:12:23.330 --> 01:12:25.997 align:middle line:84% I'm guessing that's not going to be a heartbreaker for everyone. 01:12:25.997 --> 01:12:28.390 align:middle line:84% But if you want to think about whether you 01:12:28.390 --> 01:12:31.570 align:middle line:84% want to get in the queue before we get to 12 o'clock. 01:12:31.570 --> 01:12:35.187 align:middle line:84% But I would like to talk about yours. 01:12:35.187 --> 01:12:36.020 align:middle line:90% STUDENT: Yeah, yeah. 01:12:36.020 --> 01:12:37.060 align:middle line:84% NELSON REPENNING: I think it's interesting too. 01:12:37.060 --> 01:12:37.935 align:middle line:90% STUDENT: I need help. 01:12:37.935 --> 01:12:39.005 align:middle line:90% NELSON REPENNING: OK. 01:12:39.005 --> 01:12:40.380 align:middle line:84% OK, well, why don't we get going? 01:12:40.380 --> 01:12:43.020 align:middle line:84% Hopefully Oriana will-- is she the only one that's missing? 01:12:43.020 --> 01:12:43.520 align:middle line:90% I guess. 01:12:43.520 --> 01:12:45.078 align:middle line:90% OK. 01:12:45.078 --> 01:12:47.620 align:middle line:84% Yeah, so we don't need to worry about the model for a second. 01:12:47.620 --> 01:12:51.580 align:middle line:84% What's the problem or puzzle you're trying to sort here? 01:12:51.580 --> 01:12:58.000 align:middle line:90% 01:12:58.000 --> 01:13:01.420 align:middle line:84% STUDENT: So in the current state, 01:13:01.420 --> 01:13:03.993 align:middle line:84% the problem is with one of these, 01:13:03.993 --> 01:13:06.160 align:middle line:84% what are the consequences of having a structure that 01:13:06.160 --> 01:13:09.510 align:middle line:90% looks like this? 01:13:09.510 --> 01:13:11.880 align:middle line:90% What are the consequences? 01:13:11.880 --> 01:13:19.160 align:middle line:84% So the idea is building a simple model 01:13:19.160 --> 01:13:25.680 align:middle line:84% of how a flight books from the day 01:13:25.680 --> 01:13:28.360 align:middle line:84% it's born to the day it takes off 01:13:28.360 --> 01:13:35.140 align:middle line:84% with realistic kind of booking behavior on both ends. 01:13:35.140 --> 01:13:38.680 align:middle line:84% So customers show up, they're willing to pay 01:13:38.680 --> 01:13:41.420 align:middle line:90% if the price is low enough. 01:13:41.420 --> 01:13:43.662 align:middle line:84% But when booking, they might wait. 01:13:43.662 --> 01:13:45.120 align:middle line:84% And then on the other side, there's 01:13:45.120 --> 01:13:47.160 align:middle line:84% a revenue manager kind of adjusting prices 01:13:47.160 --> 01:13:49.680 align:middle line:84% based on where they are relative to that point. 01:13:49.680 --> 01:13:57.960 align:middle line:84% The thought was, can we build first just 01:13:57.960 --> 01:14:01.680 align:middle line:84% one flight on one departure day, kind 01:14:01.680 --> 01:14:10.900 align:middle line:84% of booking until it takes off, and then layer in a competitor? 01:14:10.900 --> 01:14:15.120 align:middle line:90% OK, why bother with any of this? 01:14:15.120 --> 01:14:22.800 align:middle line:84% The hypothesis kind of idea is that because demand is not 01:14:22.800 --> 01:14:29.000 align:middle line:84% fully observable, meaning if you're priced too high, 01:14:29.000 --> 01:14:31.730 align:middle line:84% you cut off some demand at certain times, 01:14:31.730 --> 01:14:36.130 align:middle line:84% and then you can lower it later, get some of that. 01:14:36.130 --> 01:14:43.450 align:middle line:84% There's a bit of a challenge in attributing 01:14:43.450 --> 01:14:51.690 align:middle line:84% the perspective of the airline, like whether your forecast was 01:14:51.690 --> 01:14:56.370 align:middle line:84% off, if the competitor priced too aggressively. 01:14:56.370 --> 01:14:58.630 align:middle line:84% And if you play this game over and over again, 01:14:58.630 --> 01:15:02.490 align:middle line:84% you can wind up perhaps learning the wrong lesson 01:15:02.490 --> 01:15:11.770 align:middle line:84% and getting stuck in these very sort of misaligned beliefs 01:15:11.770 --> 01:15:14.770 align:middle line:84% about what the demand environment actually is. 01:15:14.770 --> 01:15:18.172 align:middle line:84% So I didn't really give you a problem statement. 01:15:18.172 --> 01:15:19.130 align:middle line:90% NELSON REPENNING: Good. 01:15:19.130 --> 01:15:21.530 align:middle line:90% You didn't. 01:15:21.530 --> 01:15:22.816 align:middle line:90% Who can help Jason? 01:15:22.816 --> 01:15:24.983 align:middle line:84% Or what question are we about to ask next, which you 01:15:24.983 --> 01:15:26.330 align:middle line:90% kind of already are on the-- 01:15:26.330 --> 01:15:31.898 align:middle line:90% 01:15:31.898 --> 01:15:33.440 align:middle line:84% STUDENT: What exactly is the problem? 01:15:33.440 --> 01:15:34.640 align:middle line:90% NELSON REPENNING: Yeah, exactly. 01:15:34.640 --> 01:15:35.860 align:middle line:84% STUDENT: Yeah, what's the problem we're trying-- 01:15:35.860 --> 01:15:37.780 align:middle line:84% NELSON REPENNING: Or why should we care? 01:15:37.780 --> 01:15:40.588 align:middle line:84% So I mean, I know you know this, but you're 01:15:40.588 --> 01:15:42.380 align:middle line:84% getting the cart and the horse a little bit 01:15:42.380 --> 01:15:43.600 align:middle line:90% in the wrong order here. 01:15:43.600 --> 01:15:46.620 align:middle line:84% You have a sort of structural, I want 01:15:46.620 --> 01:15:49.080 align:middle line:84% to model this, the story of what happens. 01:15:49.080 --> 01:15:50.080 align:middle line:90% But to what end? 01:15:50.080 --> 01:15:51.622 align:middle line:84% What are we going to get out of this? 01:15:51.622 --> 01:15:52.913 align:middle line:90% STUDENT: I need help with that. 01:15:52.913 --> 01:15:54.360 align:middle line:90% NELSON REPENNING: OK. 01:15:54.360 --> 01:16:00.740 align:middle line:90% 01:16:00.740 --> 01:16:07.380 align:middle line:84% STUDENT: It's accurately finding demand 01:16:07.380 --> 01:16:11.840 align:middle line:84% from the sale percent [INAUDIBLE]. 01:16:11.840 --> 01:16:16.580 align:middle line:90% 01:16:16.580 --> 01:16:18.040 align:middle line:90% It's not-- it's-- 01:16:18.040 --> 01:16:23.840 align:middle line:90% 01:16:23.840 --> 01:16:31.710 align:middle line:84% STUDENT: I mean, one question you could ask is like, 01:16:31.710 --> 01:16:33.190 align:middle line:84% is it important to have an accurate 01:16:33.190 --> 01:16:36.130 align:middle line:90% forecast of remaining demand? 01:16:36.130 --> 01:16:39.830 align:middle line:90% Or when is it important? 01:16:39.830 --> 01:16:43.810 align:middle line:84% It's one question, but it's not really-- 01:16:43.810 --> 01:16:45.310 align:middle line:84% STUDENT: Let me look at my problems. 01:16:45.310 --> 01:16:48.210 align:middle line:84% NELSON REPENNING: So you need a puzzle or a policy question. 01:16:48.210 --> 01:16:49.710 align:middle line:90% Yeah. 01:16:49.710 --> 01:16:51.030 align:middle line:90% STUDENT: I need a puzzle. 01:16:51.030 --> 01:16:51.790 align:middle line:84% NELSON REPENNING: I mean, I think the one 01:16:51.790 --> 01:16:53.790 align:middle line:84% that we were talking about in the hallway, which 01:16:53.790 --> 01:16:56.810 align:middle line:84% is how is this industry that seems to be very lucrative, 01:16:56.810 --> 01:17:00.710 align:middle line:84% in fact, has cumulatively lost money over its entire existence? 01:17:00.710 --> 01:17:02.910 align:middle line:84% It's a very interesting question. 01:17:02.910 --> 01:17:07.150 align:middle line:84% That may not be the one that you're interested in. 01:17:07.150 --> 01:17:09.010 align:middle line:84% But just so, free associate a little bit. 01:17:09.010 --> 01:17:10.635 align:middle line:84% I mean, why is this interesting to you? 01:17:10.635 --> 01:17:14.270 align:middle line:84% I know you're from the industry, but I'll pin you down 01:17:14.270 --> 01:17:16.110 align:middle line:84% on a problem statement later, but just 01:17:16.110 --> 01:17:18.010 align:middle line:90% what grabs you about this? 01:17:18.010 --> 01:17:19.690 align:middle line:90% Why is it appealing to you? 01:17:19.690 --> 01:17:25.280 align:middle line:90% 01:17:25.280 --> 01:17:28.720 align:middle line:84% STUDENT: Because it doesn't seem like it should work. 01:17:28.720 --> 01:17:30.240 align:middle line:90% So OK. 01:17:30.240 --> 01:17:33.680 align:middle line:84% NELSON REPENNING: What doesn't seem like it should work? 01:17:33.680 --> 01:17:39.320 align:middle line:84% STUDENT: The claim is that the revenue manager, so the person 01:17:39.320 --> 01:17:43.600 align:middle line:84% who's managing the flight and deciding what the fare should 01:17:43.600 --> 01:17:52.840 align:middle line:84% be at any point in time, has some expectation about what 01:17:52.840 --> 01:17:56.520 align:middle line:84% kind of booking will materialize and when. 01:17:56.520 --> 01:18:03.000 align:middle line:84% And they wind up kind of claiming-- the whole system is 01:18:03.000 --> 01:18:06.200 align:middle line:84% set up to match the right passenger with the right seat 01:18:06.200 --> 01:18:09.480 align:middle line:90% at the right time. 01:18:09.480 --> 01:18:11.760 align:middle line:90% So that's the claim. 01:18:11.760 --> 01:18:16.720 align:middle line:84% But in reality, there's a lots of places where it breaks down. 01:18:16.720 --> 01:18:18.712 align:middle line:90% And I don't know. 01:18:18.712 --> 01:18:21.240 align:middle line:84% I just want to know how it works. 01:18:21.240 --> 01:18:26.890 align:middle line:84% I want to see what the consequences are. 01:18:26.890 --> 01:18:28.530 align:middle line:90% Why is it interesting? 01:18:28.530 --> 01:18:35.583 align:middle line:90% 01:18:35.583 --> 01:18:36.750 align:middle line:90% Here's why it's interesting. 01:18:36.750 --> 01:18:38.190 align:middle line:90% It's incredibly complex. 01:18:38.190 --> 01:18:40.890 align:middle line:84% The true dynamics are incredibly complex. 01:18:40.890 --> 01:18:44.530 align:middle line:84% And you have to design kind of roles 01:18:44.530 --> 01:18:50.270 align:middle line:84% in the running of the organization that are reducibly 01:18:50.270 --> 01:18:51.230 align:middle line:90% kind of simple. 01:18:51.230 --> 01:18:54.730 align:middle line:84% So a lot of that complexity has to be like confined 01:18:54.730 --> 01:18:56.650 align:middle line:90% to the background. 01:18:56.650 --> 01:19:02.410 align:middle line:84% And my confusion, or my kind of instinct 01:19:02.410 --> 01:19:06.410 align:middle line:84% is that too much is shifted to the background, 01:19:06.410 --> 01:19:10.530 align:middle line:84% and forgotten about, and then the actual simplicity 01:19:10.530 --> 01:19:16.010 align:middle line:84% of kind of the roles that get enacted 01:19:16.010 --> 01:19:18.790 align:middle line:84% don't account or appreciate for what's actually going on. 01:19:18.790 --> 01:19:22.940 align:middle line:84% And that lack of accounting and lack of appreciation 01:19:22.940 --> 01:19:27.220 align:middle line:84% gets embedded into the system and feeds back 01:19:27.220 --> 01:19:32.508 align:middle line:84% in horrible ways that can lead to divergences. 01:19:32.508 --> 01:19:34.800 align:middle line:84% NELSON REPENNING: Give me an example of a horrible way. 01:19:34.800 --> 01:19:38.460 align:middle line:90% 01:19:38.460 --> 01:19:40.826 align:middle line:90% STUDENT: OK. 01:19:40.826 --> 01:19:44.060 align:middle line:84% It's been a while since I've thought through this. 01:19:44.060 --> 01:19:50.966 align:middle line:84% One example is, like, a, self-fulfilling prophecy sort 01:19:50.966 --> 01:19:52.380 align:middle line:90% of story, where-- 01:19:52.380 --> 01:19:56.060 align:middle line:90% 01:19:56.060 --> 01:19:58.960 align:middle line:84% I know that Easter is a week earlier this year. 01:19:58.960 --> 01:20:05.140 align:middle line:84% So the demand for my flight is going 01:20:05.140 --> 01:20:07.240 align:middle line:84% to be a lot higher than it was last year. 01:20:07.240 --> 01:20:14.060 align:middle line:84% So I set the fare level 10% higher across the whole curve. 01:20:14.060 --> 01:20:21.450 align:middle line:84% And then that demand, suppose it does not materialize. 01:20:21.450 --> 01:20:24.630 align:middle line:90% 01:20:24.630 --> 01:20:28.350 align:middle line:84% Your new booking history is going 01:20:28.350 --> 01:20:35.670 align:middle line:84% to be lower than what it would have been if you 01:20:35.670 --> 01:20:38.810 align:middle line:90% had priced more accurately. 01:20:38.810 --> 01:20:41.690 align:middle line:84% So now, you have a biased forecast for next time. 01:20:41.690 --> 01:20:45.470 align:middle line:90% 01:20:45.470 --> 01:20:48.910 align:middle line:84% And no one's accounting for that unconstraining process of-- 01:20:48.910 --> 01:20:52.330 align:middle line:84% I mean, there is nominal ways they account for it, but-- 01:20:52.330 --> 01:20:54.910 align:middle line:84% NELSON REPENNING: OK, so let me repeat back 01:20:54.910 --> 01:20:56.670 align:middle line:84% to what I think I'm hearing, is you 01:20:56.670 --> 01:21:00.030 align:middle line:84% have an instinct that this process doesn't work as well 01:21:00.030 --> 01:21:02.110 align:middle line:90% as it could have, it could. 01:21:02.110 --> 01:21:05.390 align:middle line:84% And so there is a pathology, or a problem here, in the sense 01:21:05.390 --> 01:21:09.370 align:middle line:84% that this is somehow suboptimized. 01:21:09.370 --> 01:21:11.890 align:middle line:84% OK, so let me give you the alternative. 01:21:11.890 --> 01:21:14.430 align:middle line:84% So I know nothing about this industry, right? 01:21:14.430 --> 01:21:17.280 align:middle line:84% When I was a student, this kind of dynamic pricing 01:21:17.280 --> 01:21:18.500 align:middle line:90% was a very hot topic. 01:21:18.500 --> 01:21:20.763 align:middle line:84% And so my naive take on this would have been, 01:21:20.763 --> 01:21:22.680 align:middle line:84% actually, it works pretty well because there's 01:21:22.680 --> 01:21:24.763 align:middle line:84% a feedback loop where they essentially auction off 01:21:24.763 --> 01:21:26.037 align:middle line:90% each seat in the airplane. 01:21:26.037 --> 01:21:28.620 align:middle line:84% And every time I fly, it seems like the seats are pretty full. 01:21:28.620 --> 01:21:29.080 align:middle line:90% STUDENT: Yes. 01:21:29.080 --> 01:21:30.440 align:middle line:84% NELSON REPENNING: So I don't know. 01:21:30.440 --> 01:21:31.898 align:middle line:84% It seems like it works pretty well. 01:21:31.898 --> 01:21:34.097 align:middle line:84% But it sounds like your instinct is, 01:21:34.097 --> 01:21:36.180 align:middle line:84% there's something in there that doesn't work well, 01:21:36.180 --> 01:21:40.200 align:middle line:84% or there's a potential pathology that can recur 01:21:40.200 --> 01:21:43.120 align:middle line:90% or something like that? 01:21:43.120 --> 01:21:46.360 align:middle line:84% STUDENT: I think there's a lack of accounting for the feedback 01:21:46.360 --> 01:21:52.720 align:middle line:84% effects in the forecasting process, proper accounting. 01:21:52.720 --> 01:21:57.440 align:middle line:84% And I'm curious to note that is consequential. 01:21:57.440 --> 01:21:59.480 align:middle line:84% I'm not sure, but I have a feeling it is. 01:21:59.480 --> 01:22:02.058 align:middle line:90% 01:22:02.058 --> 01:22:03.600 align:middle line:84% NELSON REPENNING: OK, so I think this 01:22:03.600 --> 01:22:05.420 align:middle line:84% is an interesting challenge, right? 01:22:05.420 --> 01:22:08.200 align:middle line:84% Because you have an instinct that maybe 01:22:08.200 --> 01:22:10.400 align:middle line:90% there's something going wrong. 01:22:10.400 --> 01:22:13.440 align:middle line:84% But unless-- God damn this thing. 01:22:13.440 --> 01:22:17.555 align:middle line:84% Unless I am misinterpreting you, you don't have a lot of evidence 01:22:17.555 --> 01:22:19.430 align:middle line:84% that this thing is going wrong at this point. 01:22:19.430 --> 01:22:20.550 align:middle line:90% It's more just-- 01:22:20.550 --> 01:22:21.010 align:middle line:90% STUDENT: I have no evidence. 01:22:21.010 --> 01:22:22.510 align:middle line:84% NELSON REPENNING: Yeah, no evidence. 01:22:22.510 --> 01:22:28.450 align:middle line:90% 01:22:28.450 --> 01:22:32.130 align:middle line:84% So I would suggest you go in two directions at once, which 01:22:32.130 --> 01:22:36.652 align:middle line:84% is it'd be interesting to look at the data 01:22:36.652 --> 01:22:38.610 align:middle line:84% and see if there's a kind of thought experiment 01:22:38.610 --> 01:22:41.010 align:middle line:84% that you could run that would give you a better sense 01:22:41.010 --> 01:22:43.150 align:middle line:84% that the fact there is something going wrong. 01:22:43.150 --> 01:22:46.650 align:middle line:84% I mean, I would imagine, for example, there's 01:22:46.650 --> 01:22:49.790 align:middle line:84% an upper bound of I have x number of flights, 01:22:49.790 --> 01:22:52.050 align:middle line:90% I have y number of seats. 01:22:52.050 --> 01:22:54.410 align:middle line:84% That should add up to this much money. 01:22:54.410 --> 01:22:56.830 align:middle line:84% And what they get is somehow less than that. 01:22:56.830 --> 01:22:59.163 align:middle line:84% I mean, my casual empiricism is it 01:22:59.163 --> 01:23:01.830 align:middle line:84% doesn't seem like there's a lot of empty seats on these flights. 01:23:01.830 --> 01:23:04.390 align:middle line:84% But maybe they're not filling them in some optimal way. 01:23:04.390 --> 01:23:06.870 align:middle line:84% But that casual empiricism could also just be wrong. 01:23:06.870 --> 01:23:09.330 align:middle line:84% STUDENT: Yeah, I mean, there's a trade off 01:23:09.330 --> 01:23:10.930 align:middle line:90% between load and yield. 01:23:10.930 --> 01:23:13.130 align:middle line:90% You can sell out a plane. 01:23:13.130 --> 01:23:18.350 align:middle line:84% I saw a plane tomorrow if I make all the tickets $50. 01:23:18.350 --> 01:23:20.130 align:middle line:90% NELSON REPENNING: Right. 01:23:20.130 --> 01:23:24.970 align:middle line:84% STUDENT: So the kind of challenge of the revenue manager 01:23:24.970 --> 01:23:29.890 align:middle line:84% is to decide kind of what kind of demand is remaining, 01:23:29.890 --> 01:23:33.610 align:middle line:90% and what value is of-- 01:23:33.610 --> 01:23:40.050 align:middle line:84% I think-- like, I could articulate 01:23:40.050 --> 01:23:43.560 align:middle line:90% a much clearer feedback story. 01:23:43.560 --> 01:23:45.810 align:middle line:84% But then, I'm committing myself to a story for which I 01:23:45.810 --> 01:23:47.610 align:middle line:90% don't really have evidence. 01:23:47.610 --> 01:23:48.610 align:middle line:90% NELSON REPENNING: Right. 01:23:48.610 --> 01:23:52.090 align:middle line:84% But I don't think you have a choice there because so here's-- 01:23:52.090 --> 01:23:55.930 align:middle line:84% I think one path is to see is, can I go collect some data 01:23:55.930 --> 01:24:02.090 align:middle line:84% or talk to people, and kind of give more detail to this what 01:24:02.090 --> 01:24:04.810 align:middle line:84% now sounds like a fairly vague instinct that there's something 01:24:04.810 --> 01:24:08.010 align:middle line:84% wrong and then there hills there that something's going wrong. 01:24:08.010 --> 01:24:11.760 align:middle line:84% And then I think, you can try to model it, right? 01:24:11.760 --> 01:24:14.820 align:middle line:84% Because you have the nuggets of a dynamic hypothesis 01:24:14.820 --> 01:24:16.940 align:middle line:84% that somehow the different kind of expectations 01:24:16.940 --> 01:24:23.620 align:middle line:84% processes in the players are not adding up to the right answer. 01:24:23.620 --> 01:24:25.120 align:middle line:90% STUDENT: Right. 01:24:25.120 --> 01:24:30.060 align:middle line:90% 01:24:30.060 --> 01:24:32.820 align:middle line:84% Here's what I'd do from my experience. 01:24:32.820 --> 01:24:37.140 align:middle line:84% There's a demand forecast that gets used as a tool every day 01:24:37.140 --> 01:24:39.380 align:middle line:84% as a revenue manager is managing. 01:24:39.380 --> 01:24:43.060 align:middle line:84% That forecast is produced as a result of an automated process. 01:24:43.060 --> 01:24:46.380 align:middle line:84% The revenue manager uses that forecast 01:24:46.380 --> 01:24:49.340 align:middle line:84% as an input in their decision on how to adjust fares. 01:24:49.340 --> 01:24:52.660 align:middle line:84% And as a consequence of the forecasts 01:24:52.660 --> 01:24:55.220 align:middle line:84% and the decision they make, it's going 01:24:55.220 --> 01:24:58.560 align:middle line:84% to feed back and create future forecasts. 01:24:58.560 --> 01:25:01.700 align:middle line:90% 01:25:01.700 --> 01:25:03.180 align:middle line:90% What's left out? 01:25:03.180 --> 01:25:05.520 align:middle line:84% There's no evaluation of the forecast accuracy. 01:25:05.520 --> 01:25:07.060 align:middle line:90% Zero. 01:25:07.060 --> 01:25:12.390 align:middle line:84% So this is a recipe for having very divergent bad forecasts, 01:25:12.390 --> 01:25:15.010 align:middle line:90% and no one trusts the forecast. 01:25:15.010 --> 01:25:15.510 align:middle line:90% So-- 01:25:15.510 --> 01:25:18.653 align:middle line:90% 01:25:18.653 --> 01:25:21.070 align:middle line:84% NELSON REPENNING: OK, so this is getting more interesting. 01:25:21.070 --> 01:25:22.950 align:middle line:84% I like that no one trusts the forecasts. 01:25:22.950 --> 01:25:28.190 align:middle line:84% That gets my spider senses tingling a little bit. 01:25:28.190 --> 01:25:33.190 align:middle line:84% I think the way to go after it is without worrying about Vensim 01:25:33.190 --> 01:25:36.210 align:middle line:84% equations yet is to really-- again, 01:25:36.210 --> 01:25:39.510 align:middle line:84% you have kind of a loose dynamic hypothesis 01:25:39.510 --> 01:25:41.370 align:middle line:90% about something going wrong. 01:25:41.370 --> 01:25:44.910 align:middle line:84% And I think the question here is, does it matter? 01:25:44.910 --> 01:25:46.990 align:middle line:84% Because this is a system that has 01:25:46.990 --> 01:25:50.530 align:middle line:84% quite a bit of negative feedback to it, which is, again, 01:25:50.530 --> 01:25:52.970 align:middle line:84% if I understand revenue management, 01:25:52.970 --> 01:25:55.710 align:middle line:84% if we have a hundred seats left on this flight with a week 01:25:55.710 --> 01:25:58.590 align:middle line:84% to go, we're going to start lowering the price until-- now 01:25:58.590 --> 01:26:00.190 align:middle line:90% whether that's optimal or not. 01:26:00.190 --> 01:26:04.910 align:middle line:84% But so I think you need to kind of think really 01:26:04.910 --> 01:26:06.958 align:middle line:84% carefully at a conceptual level of, 01:26:06.958 --> 01:26:09.000 align:middle line:84% and you are committing yourself to a story, which 01:26:09.000 --> 01:26:10.560 align:middle line:84% I think is the thing I'm getting the sense is 01:26:10.560 --> 01:26:11.840 align:middle line:90% making you uncomfortable. 01:26:11.840 --> 01:26:13.600 align:middle line:84% But you don't have a choice, right? 01:26:13.600 --> 01:26:15.840 align:middle line:84% Again, if you don't, you're going 01:26:15.840 --> 01:26:17.370 align:middle line:90% to end up modeling the system. 01:26:17.370 --> 01:26:19.120 align:middle line:84% You're going to just throw in every detail 01:26:19.120 --> 01:26:20.520 align:middle line:84% that you can possibly think of and you're 01:26:20.520 --> 01:26:22.103 align:middle line:84% going to end up with a Jackson Pollock 01:26:22.103 --> 01:26:23.817 align:middle line:84% painting of a model that's probably not 01:26:23.817 --> 01:26:24.900 align:middle line:90% going to be super helpful. 01:26:24.900 --> 01:26:28.600 align:middle line:84% So I think you need to think hard about what 01:26:28.600 --> 01:26:32.020 align:middle line:90% your story is and commit to it. 01:26:32.020 --> 01:26:34.380 align:middle line:84% And when I say you have to have a puzzle or policy, 01:26:34.380 --> 01:26:36.800 align:middle line:84% I think the real world is a little more iterative than I'm 01:26:36.800 --> 01:26:38.717 align:middle line:84% characterizing it, because I think, often, you 01:26:38.717 --> 01:26:40.140 align:middle line:90% will start building your model. 01:26:40.140 --> 01:26:41.200 align:middle line:84% And then you realize that the question 01:26:41.200 --> 01:26:42.560 align:middle line:84% that you were interested in is actually 01:26:42.560 --> 01:26:43.580 align:middle line:90% not the most interesting one. 01:26:43.580 --> 01:26:45.580 align:middle line:84% And then you come back, and that's totally fine. 01:26:45.580 --> 01:26:46.880 align:middle line:90% I have no problem with that. 01:26:46.880 --> 01:26:50.120 align:middle line:84% What I don't want you to do is kind of go down the kitchen sink 01:26:50.120 --> 01:26:54.380 align:middle line:84% strategy of, I think there's a problem somewhere in the city. 01:26:54.380 --> 01:26:55.760 align:middle line:84% And so I'm going to start drawing 01:26:55.760 --> 01:26:58.720 align:middle line:84% a comprehensive map of all of Boston 01:26:58.720 --> 01:27:00.480 align:middle line:84% to see if I luck into finding it. 01:27:00.480 --> 01:27:02.980 align:middle line:84% That I think is going to be a little bit of a fool's errand. 01:27:02.980 --> 01:27:04.073 align:middle line:90% STUDENT: Right. 01:27:04.073 --> 01:27:06.490 align:middle line:84% NELSON REPENNING: And again, I think the question, though, 01:27:06.490 --> 01:27:11.150 align:middle line:84% is if you identify a pathology around forecasts, 01:27:11.150 --> 01:27:14.030 align:middle line:84% and this is a question a dynamic model could answer, 01:27:14.030 --> 01:27:15.600 align:middle line:90% is, does it matter? 01:27:15.600 --> 01:27:17.850 align:middle line:84% Because there might be enough self-correcting negative 01:27:17.850 --> 01:27:19.610 align:middle line:84% feedback there that it really doesn't. 01:27:19.610 --> 01:27:22.050 align:middle line:84% Or it might matter a lot, in which case, you 01:27:22.050 --> 01:27:24.670 align:middle line:84% would have discovered sort of something new. 01:27:24.670 --> 01:27:27.730 align:middle line:84% But I think it's your intuition needs 01:27:27.730 --> 01:27:30.170 align:middle line:84% to be made quite a bit more kind of concrete 01:27:30.170 --> 01:27:33.430 align:middle line:84% about where you think something's going sideways. 01:27:33.430 --> 01:27:34.210 align:middle line:90% STUDENT: Got it. 01:27:34.210 --> 01:27:37.570 align:middle line:90% So [INAUDIBLE]. 01:27:37.570 --> 01:27:40.130 align:middle line:90% 01:27:40.130 --> 01:27:45.530 align:middle line:84% My sense is, and I can make it more concrete, 01:27:45.530 --> 01:27:49.610 align:middle line:84% just a little bit thinking, but my sense 01:27:49.610 --> 01:27:52.090 align:middle line:84% is there is a hidden kind of reinforcing factor 01:27:52.090 --> 01:27:54.010 align:middle line:90% that's not really the problem. 01:27:54.010 --> 01:27:56.210 align:middle line:84% NELSON REPENNING: Now we're getting closer. 01:27:56.210 --> 01:27:58.090 align:middle line:84% STUDENT: And it won't be too much effort 01:27:58.090 --> 01:28:00.770 align:middle line:84% for me to identify because that will give rise 01:28:00.770 --> 01:28:04.580 align:middle line:90% to some path dependency. 01:28:04.580 --> 01:28:08.948 align:middle line:84% And then the question is, are the balancing feedbacks 01:28:08.948 --> 01:28:10.740 align:middle line:84% that exists, that are built into the system 01:28:10.740 --> 01:28:16.202 align:middle line:84% and behavioral feedbacks enough to correct for that? 01:28:16.202 --> 01:28:17.160 align:middle line:90% NELSON REPENNING: Yeah. 01:28:17.160 --> 01:28:18.380 align:middle line:90% STUDENT: Yeah. 01:28:18.380 --> 01:28:22.717 align:middle line:84% NELSON REPENNING: And enough on a time scale that matters. 01:28:22.717 --> 01:28:23.300 align:middle line:90% STUDENT: Yeah. 01:28:23.300 --> 01:28:24.300 align:middle line:90% NELSON REPENNING: Right? 01:28:24.300 --> 01:28:26.675 align:middle line:84% And so and it is fairly complicated, right? 01:28:26.675 --> 01:28:28.800 align:middle line:84% Well, there's at least another layer of complexity, 01:28:28.800 --> 01:28:31.180 align:middle line:84% right, which is what little I know of the industry 01:28:31.180 --> 01:28:35.660 align:middle line:84% is, yes, I'm auctioning off seats on this flight. 01:28:35.660 --> 01:28:40.380 align:middle line:84% And I presume that if I sell out the flight months in advance 01:28:40.380 --> 01:28:42.282 align:middle line:84% at a high price, then oh, maybe we 01:28:42.282 --> 01:28:44.240 align:middle line:84% should add a couple more flights on that route. 01:28:44.240 --> 01:28:45.698 align:middle line:84% So then there's and the time delays 01:28:45.698 --> 01:28:48.180 align:middle line:84% between changing pricing and changing capacity probably are 01:28:48.180 --> 01:28:48.960 align:middle line:90% very different. 01:28:48.960 --> 01:28:51.340 align:middle line:84% STUDENT: Yeah, I'm not mentioning that yet. 01:28:51.340 --> 01:28:53.718 align:middle line:90% 01:28:53.718 --> 01:28:56.260 align:middle line:84% NELSON REPENNING: But I think you have an instinct that there 01:28:56.260 --> 01:28:58.020 align:middle line:84% is a kind of, I think path dependence is 01:28:58.020 --> 01:29:02.390 align:middle line:84% the right answer here, is that forecast kind of anchors us 01:29:02.390 --> 01:29:03.230 align:middle line:90% in a certain-- 01:29:03.230 --> 01:29:04.070 align:middle line:90% STUDENT: Realm. 01:29:04.070 --> 01:29:05.870 align:middle line:84% NELSON REPENNING: Quasi equilibrium 01:29:05.870 --> 01:29:08.590 align:middle line:84% that then, we don't search outside of it, 01:29:08.590 --> 01:29:10.470 align:middle line:84% that we may be missing opportunities, 01:29:10.470 --> 01:29:13.370 align:middle line:84% or it allows competitors to come in, or something like that. 01:29:13.370 --> 01:29:17.950 align:middle line:84% STUDENT: Yeah, so clarifying really, so what I need to do 01:29:17.950 --> 01:29:22.210 align:middle line:84% is kind of really articulate the dynamic hypothesis. 01:29:22.210 --> 01:29:25.070 align:middle line:90% 01:29:25.070 --> 01:29:30.590 align:middle line:90% And then why do I need a model? 01:29:30.590 --> 01:29:33.190 align:middle line:90% I need a model to test it. 01:29:33.190 --> 01:29:34.790 align:middle line:90% NELSON REPENNING: Yeah. 01:29:34.790 --> 01:29:35.290 align:middle line:90% Right. 01:29:35.290 --> 01:29:37.750 align:middle line:84% And I think, again, it goes back to what I was asking you, 01:29:37.750 --> 01:29:41.610 align:middle line:84% just to get clear on what dynamic models are good for. 01:29:41.610 --> 01:29:44.110 align:middle line:84% That balancing loop of I auction off seats, 01:29:44.110 --> 01:29:46.250 align:middle line:84% and that I see that this flight is full, 01:29:46.250 --> 01:29:49.990 align:middle line:84% and so I add new routes, that operates at a certain speed. 01:29:49.990 --> 01:29:52.990 align:middle line:84% And that may or may not be fast enough 01:29:52.990 --> 01:29:56.045 align:middle line:84% to offset the errors that you're identifying. 01:29:56.045 --> 01:29:57.670 align:middle line:84% That's going to be a very hard question 01:29:57.670 --> 01:30:00.580 align:middle line:84% to answer just through kind of casual thought, right? 01:30:00.580 --> 01:30:02.720 align:middle line:84% And so I think a dynamic model would 01:30:02.720 --> 01:30:07.800 align:middle line:84% be very helpful to figure out sort of which of those feedbacks 01:30:07.800 --> 01:30:09.940 align:middle line:84% is sort of dominant in that context. 01:30:09.940 --> 01:30:12.760 align:middle line:84% STUDENT: I think I'm also running into a trap, 01:30:12.760 --> 01:30:14.180 align:middle line:90% proximity to the thing. 01:30:14.180 --> 01:30:16.200 align:middle line:84% So I have this thinking complexity 01:30:16.200 --> 01:30:19.160 align:middle line:90% that I want to capture. 01:30:19.160 --> 01:30:20.000 align:middle line:90% And-- 01:30:20.000 --> 01:30:20.866 align:middle line:84% NELSON REPENNING: See, the problem is, is you 01:30:20.866 --> 01:30:23.020 align:middle line:84% know too much about how it actually works, right? 01:30:23.020 --> 01:30:24.880 align:middle line:84% So you want to leverage all that. 01:30:24.880 --> 01:30:27.880 align:middle line:90% STUDENT: It's not useful. 01:30:27.880 --> 01:30:31.223 align:middle line:84% I mean, so I had actually started 01:30:31.223 --> 01:30:33.140 align:middle line:84% with more at the level you were talking about, 01:30:33.140 --> 01:30:36.140 align:middle line:84% where there was multiple time horizons. 01:30:36.140 --> 01:30:39.400 align:middle line:84% And I was encouraged to scale it down 01:30:39.400 --> 01:30:44.280 align:middle line:84% to the most atomistic level, one flight on one departure day. 01:30:44.280 --> 01:30:48.200 align:middle line:90% Kind of how does that work? 01:30:48.200 --> 01:30:51.760 align:middle line:84% But I actually think there's some middle ground because where 01:30:51.760 --> 01:30:54.140 align:middle line:84% I started off is too vague, and that's probably where it 01:30:54.140 --> 01:30:54.640 align:middle line:90% [INAUDIBLE]. 01:30:54.640 --> 01:30:56.015 align:middle line:84% NELSON REPENNING: Well, but also, 01:30:56.015 --> 01:30:59.930 align:middle line:84% it doesn't sound like the one flight on one day 01:30:59.930 --> 01:31:01.993 align:middle line:84% captures your dynamic hypothesis. 01:31:01.993 --> 01:31:03.410 align:middle line:84% The thing that I've heard from you 01:31:03.410 --> 01:31:07.650 align:middle line:84% is that you think that somehow, a particular forecast gets kind 01:31:07.650 --> 01:31:12.530 align:middle line:84% of embedded in the system, and either limits search or is 01:31:12.530 --> 01:31:14.150 align:middle line:90% self-reinforcing in some way. 01:31:14.150 --> 01:31:15.567 align:middle line:90% One flight in one day is not-- 01:31:15.567 --> 01:31:16.650 align:middle line:90% STUDENT: Going to do that. 01:31:16.650 --> 01:31:18.490 align:middle line:84% NELSON REPENNING: That aggregation choice 01:31:18.490 --> 01:31:21.050 align:middle line:84% won't help you with a dynamic hypothesis. 01:31:21.050 --> 01:31:23.930 align:middle line:84% See, this is why that dynamic hypothesis is pretty-- 01:31:23.930 --> 01:31:24.763 align:middle line:90% STUDENT: Important. 01:31:24.763 --> 01:31:25.930 align:middle line:90% NELSON REPENNING: That's it. 01:31:25.930 --> 01:31:26.722 align:middle line:90% STUDENT: All right. 01:31:26.722 --> 01:31:27.310 align:middle line:90% [INAUDIBLE] 01:31:27.310 --> 01:31:28.185 align:middle line:90% NELSON REPENNING: OK. 01:31:28.185 --> 01:31:30.290 align:middle line:84% STUDENT: I'll go back to the dynamic hypothesis, 01:31:30.290 --> 01:31:33.330 align:middle line:90% rework the model. 01:31:33.330 --> 01:31:34.372 align:middle line:90% I had a question. 01:31:34.372 --> 01:31:35.330 align:middle line:90% NELSON REPENNING: Yeah. 01:31:35.330 --> 01:31:36.705 align:middle line:84% STUDENT: It might have gone away, 01:31:36.705 --> 01:31:38.550 align:middle line:90% but you can check it real fast. 01:31:38.550 --> 01:31:39.925 align:middle line:84% NELSON REPENNING: Yeah, go ahead. 01:31:39.925 --> 01:31:42.650 align:middle line:90% 01:31:42.650 --> 01:31:50.336 align:middle line:90% STUDENT: So if you-- 01:31:50.336 --> 01:31:53.810 align:middle line:84% well, it might be answered with your statement 01:31:53.810 --> 01:31:56.980 align:middle line:84% earlier that you had the stock flow chain, 01:31:56.980 --> 01:32:00.200 align:middle line:84% the average has to be the same for everyone in that chain. 01:32:00.200 --> 01:32:03.300 align:middle line:84% So if you want to model the distribution, 01:32:03.300 --> 01:32:05.460 align:middle line:84% you have to just have separate stocks for each 01:32:05.460 --> 01:32:06.220 align:middle line:90% [? individual distribution. ?] 01:32:06.220 --> 01:32:08.345 align:middle line:84% NELSON REPENNING: Yeah, so the technical assumption 01:32:08.345 --> 01:32:09.440 align:middle line:90% is perfect mixing, right? 01:32:09.440 --> 01:32:13.820 align:middle line:84% Which means that if we just take seats remaining-- 01:32:13.820 --> 01:32:14.700 align:middle line:90% STUDENT: Well-- 01:32:14.700 --> 01:32:16.360 align:middle line:84% NELSON REPENNING: I mean, it's maybe not a good example. 01:32:16.360 --> 01:32:17.443 align:middle line:90% STUDENT: Yeah, yeah, yeah. 01:32:17.443 --> 01:32:19.480 align:middle line:90% Yeah, scroll down a little bit. 01:32:19.480 --> 01:32:23.820 align:middle line:90% 01:32:23.820 --> 01:32:28.060 align:middle line:84% So this average reference price, people wouldn't book. 01:32:28.060 --> 01:32:29.720 align:middle line:90% It's the same for everyone. 01:32:29.720 --> 01:32:31.660 align:middle line:84% But really, that average represents kind of, 01:32:31.660 --> 01:32:34.640 align:middle line:84% it's a summary of some distribution. 01:32:34.640 --> 01:32:39.220 align:middle line:84% And the shape of that distribution could change. 01:32:39.220 --> 01:32:42.400 align:middle line:84% It's actually what we care about, not just the average. 01:32:42.400 --> 01:32:43.500 align:middle line:90% We care about the shape. 01:32:43.500 --> 01:32:44.500 align:middle line:90% NELSON REPENNING: Yeah. 01:32:44.500 --> 01:32:47.660 align:middle line:84% STUDENT: And to capture that, I need multiple stocks. 01:32:47.660 --> 01:32:49.260 align:middle line:90% I can't [INAUDIBLE]. 01:32:49.260 --> 01:32:52.940 align:middle line:84% And I was trying to get the shape through a parameter, 01:32:52.940 --> 01:32:54.480 align:middle line:90% but the parameter is fixed. 01:32:54.480 --> 01:32:57.080 align:middle line:84% So I'm just basically giving it the same shape every time. 01:32:57.080 --> 01:32:58.580 align:middle line:84% NELSON REPENNING: Yeah, I mean, I've 01:32:58.580 --> 01:33:03.180 align:middle line:84% often wondered if you think about how the co-flow works, 01:33:03.180 --> 01:33:05.660 align:middle line:84% you're basically just capturing the first moment 01:33:05.660 --> 01:33:07.380 align:middle line:90% of a distribution. 01:33:07.380 --> 01:33:10.300 align:middle line:84% So you're just capturing the average attribute. 01:33:10.300 --> 01:33:12.260 align:middle line:84% And the perfect mixing assumption 01:33:12.260 --> 01:33:16.220 align:middle line:84% means that anyone that comes out of that box, 01:33:16.220 --> 01:33:18.220 align:middle line:84% it's just a random draw and they have 01:33:18.220 --> 01:33:19.800 align:middle line:90% the same average propensity. 01:33:19.800 --> 01:33:24.800 align:middle line:84% So you can't take out of a stock the high-propensity demand, 01:33:24.800 --> 01:33:27.460 align:middle line:90% the tail. 01:33:27.460 --> 01:33:29.720 align:middle line:84% And I don't know enough probability to work this out. 01:33:29.720 --> 01:33:33.460 align:middle line:84% But if you had a co-flow that had higher moments, so 01:33:33.460 --> 01:33:37.427 align:middle line:84% if we had variance, and-- could you get around this? 01:33:37.427 --> 01:33:38.760 align:middle line:90% I don't know the answer to that. 01:33:38.760 --> 01:33:41.838 align:middle line:84% And again, I don't know enough probability to work on that. 01:33:41.838 --> 01:33:42.880 align:middle line:90% STUDENT: It's worthwhile. 01:33:42.880 --> 01:33:44.560 align:middle line:84% NELSON REPENNING: Yeah, but I think you could have two types. 01:33:44.560 --> 01:33:46.120 align:middle line:84% So you can have n types or, right? 01:33:46.120 --> 01:33:47.800 align:middle line:84% I mean, you could get a distribution. 01:33:47.800 --> 01:33:50.860 align:middle line:84% You can get around it by having-- 01:33:50.860 --> 01:33:55.350 align:middle line:84% we could have business customers and leisure travel customers. 01:33:55.350 --> 01:33:59.430 align:middle line:84% And then we could get two different propensities for-- 01:33:59.430 --> 01:34:00.770 align:middle line:90% You see this error all the time. 01:34:00.770 --> 01:34:03.062 align:middle line:84% People try to work their way around that perfect mixing 01:34:03.062 --> 01:34:05.670 align:middle line:84% assumption, and it always gets them in trouble 01:34:05.670 --> 01:34:07.030 align:middle line:90% because you're just violating. 01:34:07.030 --> 01:34:09.230 align:middle line:84% There's a whole kind of distributional assumptions 01:34:09.230 --> 01:34:10.550 align:middle line:90% underlying all of this that-- 01:34:10.550 --> 01:34:13.382 align:middle line:84% STUDENT: Yeah, I think the reason 01:34:13.382 --> 01:34:15.590 align:middle line:84% it's tempting to want to work your way around that is 01:34:15.590 --> 01:34:18.750 align:middle line:90% because if I introduce more-- 01:34:18.750 --> 01:34:21.250 align:middle line:84% more and more I disaggregate intuitively, 01:34:21.250 --> 01:34:25.190 align:middle line:84% the closer I am to something that's-- 01:34:25.190 --> 01:34:26.430 align:middle line:90% NELSON REPENNING: To God. 01:34:26.430 --> 01:34:27.710 align:middle line:90% STUDENT: Continues. 01:34:27.710 --> 01:34:30.030 align:middle line:90% Not to God. 01:34:30.030 --> 01:34:33.970 align:middle line:84% The more and more I disaggregate, like, OK, 01:34:33.970 --> 01:34:37.650 align:middle line:84% I could have a hundred different buckets 01:34:37.650 --> 01:34:41.510 align:middle line:84% and I still wouldn't cover the full demand curve distribution. 01:34:41.510 --> 01:34:43.330 align:middle line:84% Over 1,000 buckets still doesn't cover it. 01:34:43.330 --> 01:34:46.530 align:middle line:84% So I'm picturing a continuous thing, 01:34:46.530 --> 01:34:48.990 align:middle line:84% and then it's tempting for me to want to represent 01:34:48.990 --> 01:34:53.020 align:middle line:90% that with one stock. 01:34:53.020 --> 01:34:54.280 align:middle line:90% But I think that doesn't work. 01:34:54.280 --> 01:34:57.000 align:middle line:84% NELSON REPENNING: Yeah, and I don't-- 01:34:57.000 --> 01:34:59.000 align:middle line:84% Tom Friedman gave a paper at the last conference 01:34:59.000 --> 01:35:03.040 align:middle line:84% I went to, where the model would sort of rewrite stocks 01:35:03.040 --> 01:35:04.020 align:middle line:90% sort of as needed. 01:35:04.020 --> 01:35:05.480 align:middle line:84% I mean, I think there probably are 01:35:05.480 --> 01:35:07.320 align:middle line:84% other mathematical formulations that someone 01:35:07.320 --> 01:35:09.460 align:middle line:84% could invent to solve that problem, 01:35:09.460 --> 01:35:11.040 align:middle line:90% but I haven't seen one that-- 01:35:11.040 --> 01:35:14.880 align:middle line:84% PJ, I know, worked on this a little bit. 01:35:14.880 --> 01:35:17.680 align:middle line:84% But yeah, the moment you say, well, I'm 01:35:17.680 --> 01:35:21.608 align:middle line:84% going to take out the one tail of the distribution first-- 01:35:21.608 --> 01:35:23.400 align:middle line:84% STUDENT: You need to put it in [INAUDIBLE]. 01:35:23.400 --> 01:35:25.180 align:middle line:90% NELSON REPENNING: Exactly right. 01:35:25.180 --> 01:35:25.680 align:middle line:90% STUDENT: OK. 01:35:25.680 --> 01:35:26.420 align:middle line:90% This helps. 01:35:26.420 --> 01:35:29.800 align:middle line:84% I might not even need to deal with this part. 01:35:29.800 --> 01:35:31.160 align:middle line:90% OK. 01:35:31.160 --> 01:35:32.240 align:middle line:90% NELSON REPENNING: OK? 01:35:32.240 --> 01:35:32.840 align:middle line:90% All right. 01:35:32.840 --> 01:35:33.783 align:middle line:90% Awesome. 01:35:33.783 --> 01:35:34.700 align:middle line:90% All right, who's next? 01:35:34.700 --> 01:35:35.533 align:middle line:90% STUDENT: [INAUDIBLE] 01:35:35.533 --> 01:35:36.950 align:middle line:84% NELSON REPENNING: Yeah, I'm sorry. 01:35:36.950 --> 01:35:38.320 align:middle line:90% STUDENT: No, you should go. 01:35:38.320 --> 01:35:39.660 align:middle line:84% NELSON REPENNING: All right, so how many people do we have left? 01:35:39.660 --> 01:35:42.920 align:middle line:84% I'll actually be a little more disciplined than I normally am. 01:35:42.920 --> 01:35:43.980 align:middle line:90% So we have an hour left. 01:35:43.980 --> 01:35:45.440 align:middle line:90% And how many want to go? 01:35:45.440 --> 01:35:46.420 align:middle line:90% So we have four of you? 01:35:46.420 --> 01:35:48.690 align:middle line:84% STUDENT: I could have on little [INAUDIBLE]. 01:35:48.690 --> 01:35:51.190 align:middle line:84% NELSON REPENNING: OK, so we'll probably do about 10 minutes. 01:35:51.190 --> 01:35:55.810 align:middle line:84% So I'm going to have to actually not be myself and be 01:35:55.810 --> 01:35:56.550 align:middle line:90% more efficient. 01:35:56.550 --> 01:35:57.990 align:middle line:90% OK, so who wants to go first? 01:35:57.990 --> 01:35:58.570 align:middle line:90% Jake, do you want to go? 01:35:58.570 --> 01:35:59.090 align:middle line:90% STUDENT: OK. 01:35:59.090 --> 01:36:01.955 align:middle line:90% NELSON REPENNING: OK. 01:36:01.955 --> 01:36:05.130 align:middle line:90% So again, it's the same-- 01:36:05.130 --> 01:36:06.110 align:middle line:90% which one was yours? 01:36:06.110 --> 01:36:07.270 align:middle line:90% The runaway production? 01:36:07.270 --> 01:36:07.853 align:middle line:90% STUDENT: Yeah. 01:36:07.853 --> 01:36:10.050 align:middle line:90% 01:36:10.050 --> 01:36:13.210 align:middle line:84% NELSON REPENNING: So tell us your story, or what-- 01:36:13.210 --> 01:36:14.710 align:middle line:90% STUDENT: It didn't upload. 01:36:14.710 --> 01:36:16.335 align:middle line:84% NELSON REPENNING: Yeah, here, I got it. 01:36:16.335 --> 01:36:17.450 align:middle line:90% No, I think I have it. 01:36:17.450 --> 01:36:21.210 align:middle line:84% STUDENT: So the story is Los Angeles has basically always 01:36:21.210 --> 01:36:25.230 align:middle line:84% been where basically all film and TV production has happened, 01:36:25.230 --> 01:36:29.610 align:middle line:84% at least for major English-language blockbusters. 01:36:29.610 --> 01:36:35.290 align:middle line:84% And recently, it's been sort of a saving to other locations 01:36:35.290 --> 01:36:37.090 align:middle line:90% at an increasing rate. 01:36:37.090 --> 01:36:39.610 align:middle line:84% And there's a bunch of explanations for that that 01:36:39.610 --> 01:36:44.330 align:middle line:84% come from various, like, spokepeople 01:36:44.330 --> 01:36:49.060 align:middle line:84% in the industry, whether it's studio heads, whether it's 01:36:49.060 --> 01:36:52.680 align:middle line:84% other types of advocates, that boil down to labor cost 01:36:52.680 --> 01:36:55.860 align:middle line:84% is too much now because of all the strikes and the agreements 01:36:55.860 --> 01:36:59.740 align:middle line:84% that they came to, tax incentives elsewhere are just 01:36:59.740 --> 01:37:05.700 align:middle line:84% really way too enticing for us to ever do anything here, 01:37:05.700 --> 01:37:10.300 align:middle line:84% to maybe there is just some aspect of nowhere else 01:37:10.300 --> 01:37:12.920 align:middle line:84% had the ability to produce things before, and now they do, 01:37:12.920 --> 01:37:15.620 align:middle line:84% and it's a reinforcing feedback loop that makes it possible 01:37:15.620 --> 01:37:17.820 align:middle line:90% [INAUDIBLE]. 01:37:17.820 --> 01:37:22.700 align:middle line:84% So the problem question to me is like, 01:37:22.700 --> 01:37:27.360 align:middle line:84% if the story is a tax incentive thing, 01:37:27.360 --> 01:37:29.900 align:middle line:84% or can you work within the system 01:37:29.900 --> 01:37:33.340 align:middle line:84% to place a tax incentive as the city of Los Angeles 01:37:33.340 --> 01:37:36.800 align:middle line:84% to bring back a lot of this production, 01:37:36.800 --> 01:37:38.320 align:middle line:84% or at least stop it from leaving? 01:37:38.320 --> 01:37:40.070 align:middle line:84% Because right now, there's a lot of debate 01:37:40.070 --> 01:37:43.220 align:middle line:84% on what should that incentive level kind be? 01:37:43.220 --> 01:37:44.430 align:middle line:90% So that's sort of why. 01:37:44.430 --> 01:37:47.550 align:middle line:84% NELSON REPENNING: So it's a policy question. 01:37:47.550 --> 01:37:54.230 align:middle line:84% You want a policy answer to what suite of interventions, taxes, 01:37:54.230 --> 01:37:58.090 align:middle line:84% or otherwise, that would keep the business in Los Angeles. 01:37:58.090 --> 01:37:58.930 align:middle line:90% OK, great. 01:37:58.930 --> 01:38:00.405 align:middle line:90% Love that. 01:38:00.405 --> 01:38:02.030 align:middle line:84% STUDENT: So then here, in [INAUDIBLE]-- 01:38:02.030 --> 01:38:04.510 align:middle line:84% so I originally set it up exactly the way you're talking 01:38:04.510 --> 01:38:05.110 align:middle line:90% about it. 01:38:05.110 --> 01:38:06.930 align:middle line:84% Like it's LA, and it's rest of the world. 01:38:06.930 --> 01:38:08.490 align:middle line:84% Yeah, I did subscript everything, 01:38:08.490 --> 01:38:11.590 align:middle line:84% but it's just two locations I have [INAUDIBLE] the rest 01:38:11.590 --> 01:38:13.710 align:middle line:90% of the world at the moment. 01:38:13.710 --> 01:38:18.290 align:middle line:84% I originally was like, LA attractiveness, through 01:38:18.290 --> 01:38:21.210 align:middle line:90% [INAUDIBLE]. 01:38:21.210 --> 01:38:23.670 align:middle line:84% But it felt like everything would 01:38:23.670 --> 01:38:28.930 align:middle line:84% need like, you would need the LA town experience and [INAUDIBLE]. 01:38:28.930 --> 01:38:31.455 align:middle line:84% So I was like, it might just make more sense to [INAUDIBLE]. 01:38:31.455 --> 01:38:33.330 align:middle line:84% NELSON REPENNING: So I think in the long run, 01:38:33.330 --> 01:38:34.330 align:middle line:90% subscripting is fine. 01:38:34.330 --> 01:38:36.290 align:middle line:84% And this may just be my own foibles, 01:38:36.290 --> 01:38:39.350 align:middle line:84% but I find it just is much easier for me 01:38:39.350 --> 01:38:42.460 align:middle line:84% to visualize what's going on when I do the model of two 01:38:42.460 --> 01:38:44.560 align:middle line:84% and I have the same stock and flow for LA 01:38:44.560 --> 01:38:45.660 align:middle line:90% and rest of the world. 01:38:45.660 --> 01:38:47.470 align:middle line:84% Your brain's maybe bigger than mine, 01:38:47.470 --> 01:38:48.720 align:middle line:90% and you can sort of keep that. 01:38:48.720 --> 01:38:51.655 align:middle line:84% But I just think, I find it much easier 01:38:51.655 --> 01:38:54.280 align:middle line:84% to work out the details where I can see all the structure first 01:38:54.280 --> 01:38:55.840 align:middle line:84% before going to the subscripting that. 01:38:55.840 --> 01:38:57.382 align:middle line:84% STUDENT: If I was going to explain it 01:38:57.382 --> 01:38:59.060 align:middle line:84% to somebody who's setting policy, 01:38:59.060 --> 01:39:00.480 align:middle line:90% I would need [INAUDIBLE]. 01:39:00.480 --> 01:39:03.197 align:middle line:90% 01:39:03.197 --> 01:39:05.280 align:middle line:84% NELSON REPENNING: Before we jump into this, so why 01:39:05.280 --> 01:39:07.540 align:middle line:84% do you think this is a dynamic problem? 01:39:07.540 --> 01:39:09.560 align:middle line:84% So an economist would walk in here 01:39:09.560 --> 01:39:12.885 align:middle line:84% and say, yeah, you just need to calculate the optimal tax 01:39:12.885 --> 01:39:14.760 align:middle line:84% incentive so that the price of doing business 01:39:14.760 --> 01:39:17.960 align:middle line:84% is Los Angeles is cheaper than Vancouver. 01:39:17.960 --> 01:39:21.380 align:middle line:90% We don't need dynamics for that. 01:39:21.380 --> 01:39:23.940 align:middle line:84% We can just sort of estimate the relative demand curve. 01:39:23.940 --> 01:39:25.840 align:middle line:84% So again, it kind comes back to the question 01:39:25.840 --> 01:39:26.980 align:middle line:90% I've been asking everyone. 01:39:26.980 --> 01:39:28.640 align:middle line:84% My guess is you have some instinct that there are 01:39:28.640 --> 01:39:30.200 align:middle line:90% some interesting dynamics here. 01:39:30.200 --> 01:39:31.080 align:middle line:90% STUDENT: Yeah, I do. 01:39:31.080 --> 01:39:33.520 align:middle line:84% So I mean, I think that it is definitely 01:39:33.520 --> 01:39:38.560 align:middle line:84% like there's a sort of agglomeration dynamic piece 01:39:38.560 --> 01:39:40.590 align:middle line:84% of it for sure that I think this, 01:39:40.590 --> 01:39:44.910 align:middle line:84% if you have a really successful movie that's produced in the UK, 01:39:44.910 --> 01:39:47.930 align:middle line:84% suddenly you've proven that you can make a really 01:39:47.930 --> 01:39:49.310 align:middle line:90% successful movie in the UK. 01:39:49.310 --> 01:39:51.590 align:middle line:84% And people are more willing to sort of go and do that. 01:39:51.590 --> 01:39:52.530 align:middle line:90% NELSON REPENNING: OK. 01:39:52.530 --> 01:39:56.270 align:middle line:84% STUDENT: And I think even within the tax credit sort of system, 01:39:56.270 --> 01:40:00.490 align:middle line:84% there's obviously going to be some sort of response 01:40:00.490 --> 01:40:03.850 align:middle line:84% from the sort of rest of the world to the decisions 01:40:03.850 --> 01:40:05.410 align:middle line:90% that you're making. 01:40:05.410 --> 01:40:11.650 align:middle line:84% But really, I think it's the reason I find it dynamic 01:40:11.650 --> 01:40:16.130 align:middle line:84% is because there's a level at which if enough production goes 01:40:16.130 --> 01:40:19.810 align:middle line:84% away, suddenly, LA is no longer the spot that 01:40:19.810 --> 01:40:23.430 align:middle line:84% has the advantage of being the better spot 01:40:23.430 --> 01:40:26.090 align:middle line:84% because there's just been too much experience gained 01:40:26.090 --> 01:40:27.242 align:middle line:90% by local talent elsewhere. 01:40:27.242 --> 01:40:29.450 align:middle line:84% There's been too much local infrastructure investment 01:40:29.450 --> 01:40:31.530 align:middle line:90% that's happened elsewhere. 01:40:31.530 --> 01:40:33.230 align:middle line:84% So it's like, it's never coming back. 01:40:33.230 --> 01:40:37.450 align:middle line:84% So I thought that was maybe more feeds into itself. 01:40:37.450 --> 01:40:39.280 align:middle line:84% NELSON REPENNING: So I heard two dynamics. 01:40:39.280 --> 01:40:42.240 align:middle line:84% One is kind of like a proof dynamic, which is like, OK, 01:40:42.240 --> 01:40:45.680 align:middle line:84% I had a blockbuster that was made in Akron, Ohio. 01:40:45.680 --> 01:40:47.880 align:middle line:84% So now we can prove we can do it. 01:40:47.880 --> 01:40:50.962 align:middle line:84% That dynamic sounds relatively speedy. 01:40:50.962 --> 01:40:52.420 align:middle line:84% But then it sounds like you're also 01:40:52.420 --> 01:40:55.460 align:middle line:84% talking about a much more infrastructure-based story of, 01:40:55.460 --> 01:41:00.525 align:middle line:84% I assume there's skilled trades, and infrastructure, and people, 01:41:00.525 --> 01:41:01.900 align:middle line:84% and stuff you need in a location. 01:41:01.900 --> 01:41:02.700 align:middle line:90% STUDENT: [INAUDIBLE] 01:41:02.700 --> 01:41:03.680 align:middle line:90% NELSON REPENNING: Yeah, exactly. 01:41:03.680 --> 01:41:04.180 align:middle line:90% OK. 01:41:04.180 --> 01:41:06.580 align:middle line:90% 01:41:06.580 --> 01:41:08.540 align:middle line:90% OK, good. 01:41:08.540 --> 01:41:10.480 align:middle line:84% So you have a pretty clear dynamic hypothesis. 01:41:10.480 --> 01:41:11.897 align:middle line:84% And actually, what I think is kind 01:41:11.897 --> 01:41:14.500 align:middle line:84% of interesting about your dynamic hypothesis is, 01:41:14.500 --> 01:41:16.360 align:middle line:84% if I hear what you're saying correctly, 01:41:16.360 --> 01:41:17.960 align:middle line:90% is you're worried about a cliff. 01:41:17.960 --> 01:41:18.740 align:middle line:90% STUDENT: Yeah. 01:41:18.740 --> 01:41:20.360 align:middle line:84% NELSON REPENNING: Which is like, OK, it's fine. 01:41:20.360 --> 01:41:22.110 align:middle line:84% Production is going to these other places. 01:41:22.110 --> 01:41:26.060 align:middle line:84% But there is the fear of a reinforcing feedback 01:41:26.060 --> 01:41:29.860 align:middle line:84% that if it gets going too strongly in the wrong direction, 01:41:29.860 --> 01:41:31.080 align:middle line:90% LA could be kind of screwed. 01:41:31.080 --> 01:41:31.720 align:middle line:90% And then-- 01:41:31.720 --> 01:41:32.303 align:middle line:90% STUDENT: Yeah. 01:41:32.303 --> 01:41:33.980 align:middle line:90% NELSON REPENNING: Yeah. 01:41:33.980 --> 01:41:36.910 align:middle line:84% STUDENT: Film production in LA is like water, in [INAUDIBLE]. 01:41:36.910 --> 01:41:41.903 align:middle line:90% 01:41:41.903 --> 01:41:43.570 align:middle line:84% NELSON REPENNING: OK, that sounds great. 01:41:43.570 --> 01:41:46.710 align:middle line:84% So tell us a little bit about that model so far. 01:41:46.710 --> 01:41:51.510 align:middle line:84% STUDENT: So this is like the sort of most basic version 01:41:51.510 --> 01:41:56.950 align:middle line:84% I could make of where it has sort of a tax credit 01:41:56.950 --> 01:42:01.630 align:middle line:84% that starts somewhere else that starts to pull production away. 01:42:01.630 --> 01:42:05.190 align:middle line:84% In general, like, if you were to run it 01:42:05.190 --> 01:42:07.450 align:middle line:84% with no tax credit for elsewhere, 01:42:07.450 --> 01:42:12.870 align:middle line:84% then LA just builds up all of the local talent experience 01:42:12.870 --> 01:42:14.710 align:middle line:84% because they start with a little bit more 01:42:14.710 --> 01:42:19.030 align:middle line:84% of the experience, and infrastructure, and production 01:42:19.030 --> 01:42:19.530 align:middle line:90% scaling. 01:42:19.530 --> 01:42:22.870 align:middle line:84% So they're just going to continuously become 01:42:22.870 --> 01:42:27.150 align:middle line:84% stronger and stronger as a reason to go there. 01:42:27.150 --> 01:42:29.830 align:middle line:84% And then I made a switch so that you could set it 01:42:29.830 --> 01:42:34.050 align:middle line:84% so that, OK, now there is a 30% tax credit in Georgia, 01:42:34.050 --> 01:42:35.970 align:middle line:84% we start seeing some productions away, 01:42:35.970 --> 01:42:39.415 align:middle line:84% and over time, that sort of share of productions 01:42:39.415 --> 01:42:41.790 align:middle line:84% grows, and grows, and grows because there's a tax credit. 01:42:41.790 --> 01:42:45.750 align:middle line:84% And we like local infrastructure and talent 01:42:45.750 --> 01:42:50.030 align:middle line:84% that it sort of built up that pull everything away. 01:42:50.030 --> 01:42:55.150 align:middle line:84% And then I don't yet have things that feel important to me 01:42:55.150 --> 01:43:00.270 align:middle line:84% in this of, I think there's sort of a cost 01:43:00.270 --> 01:43:04.650 align:middle line:84% and a benefit piece of talent, for instance, 01:43:04.650 --> 01:43:08.050 align:middle line:84% whereas I only have sort of a benefit piece of that right now 01:43:08.050 --> 01:43:08.710 align:middle line:90% [INAUDIBLE]. 01:43:08.710 --> 01:43:13.710 align:middle line:84% I think there's a cost and a benefit 01:43:13.710 --> 01:43:15.450 align:middle line:90% to a lot of different pieces. 01:43:15.450 --> 01:43:18.750 align:middle line:84% Like, I might be more willing to shoot in an expensive location 01:43:18.750 --> 01:43:22.330 align:middle line:84% if I know that I can make a better movie, for instance. 01:43:22.330 --> 01:43:24.330 align:middle line:84% But if it's going to be cheap enough, 01:43:24.330 --> 01:43:28.590 align:middle line:84% I might be willing to do it in a worse-off place. 01:43:28.590 --> 01:43:30.170 align:middle line:84% So I don't have that baked in yet. 01:43:30.170 --> 01:43:33.600 align:middle line:84% But right now, I have sort of feedback loops 01:43:33.600 --> 01:43:37.560 align:middle line:84% for infrastructures or [INAUDIBLE] talents 01:43:37.560 --> 01:43:39.160 align:middle line:90% and [INAUDIBLE]. 01:43:39.160 --> 01:43:42.360 align:middle line:84% And then there's sort of a tax-credit piece of it too 01:43:42.360 --> 01:43:46.760 align:middle line:84% that goes into locations of active attractiveness. 01:43:46.760 --> 01:43:49.280 align:middle line:84% NELSON REPENNING: So have you have 01:43:49.280 --> 01:43:52.980 align:middle line:84% evidence, casual, anecdote, or otherwise, 01:43:52.980 --> 01:43:57.320 align:middle line:84% that the positive feedbacks have been working 01:43:57.320 --> 01:43:59.280 align:middle line:84% in these other locations in the sense 01:43:59.280 --> 01:44:03.120 align:middle line:84% that Georgia is starting to develop soundstages 01:44:03.120 --> 01:44:06.180 align:middle line:84% or I don't know what skilled trades you need for a movie, 01:44:06.180 --> 01:44:08.080 align:middle line:84% but electricians, or photographers, 01:44:08.080 --> 01:44:09.060 align:middle line:90% or something like that. 01:44:09.060 --> 01:44:09.400 align:middle line:90% STUDENT: Yeah. 01:44:09.400 --> 01:44:10.358 align:middle line:90% NELSON REPENNING: Yeah. 01:44:10.358 --> 01:44:11.400 align:middle line:90% OK. 01:44:11.400 --> 01:44:13.280 align:middle line:84% STUDENT: Canada and the UK both, like if you 01:44:13.280 --> 01:44:15.300 align:middle line:84% look at the UK's level of soundstages, 01:44:15.300 --> 01:44:18.520 align:middle line:84% they now have as many as Los Angeles does and sort of put 01:44:18.520 --> 01:44:20.880 align:middle line:84% a lot of investment in specifically 01:44:20.880 --> 01:44:22.980 align:middle line:84% making that infrastructure become available. 01:44:22.980 --> 01:44:25.063 align:middle line:84% And as they've done so, there's been more and more 01:44:25.063 --> 01:44:27.880 align:middle line:90% [? adherence ?] from the UK. 01:44:27.880 --> 01:44:30.610 align:middle line:84% There's smaller pieces of anecdotal evidence. 01:44:30.610 --> 01:44:34.990 align:middle line:84% Like right now, so The Lord of the Rings time, 01:44:34.990 --> 01:44:37.530 align:middle line:84% like they did a lot of the special effects and museum 01:44:37.530 --> 01:44:40.690 align:middle line:90% [INAUDIBLE] be easier. 01:44:40.690 --> 01:44:44.010 align:middle line:84% And as a result, the museum now has the most state-of-the-art 01:44:44.010 --> 01:44:46.410 align:middle line:84% visual effects in the entire world. 01:44:46.410 --> 01:44:49.850 align:middle line:84% So now, movies like The Avengers will even go to New Zealand 01:44:49.850 --> 01:44:52.000 align:middle line:84% to have them do the visual effects. 01:44:52.000 --> 01:44:53.250 align:middle line:90% NELSON REPENNING: Interesting. 01:44:53.250 --> 01:44:55.915 align:middle line:90% OK. 01:44:55.915 --> 01:44:57.790 align:middle line:84% Well, in the interest of getting to everyone, 01:44:57.790 --> 01:45:00.170 align:middle line:84% let me give you a couple thoughts. 01:45:00.170 --> 01:45:05.187 align:middle line:84% So I think the aggregation level looks pretty good for now. 01:45:05.187 --> 01:45:06.770 align:middle line:84% I mean, I can imagine a world in which 01:45:06.770 --> 01:45:09.210 align:middle line:84% you start to disaggregate between soundstages, 01:45:09.210 --> 01:45:11.150 align:middle line:84% and visual effects, and stuff like that. 01:45:11.150 --> 01:45:14.750 align:middle line:84% But I think basically, now, if I understand this correctly, 01:45:14.750 --> 01:45:17.455 align:middle line:84% you've kind of just disaggregated 01:45:17.455 --> 01:45:19.330 align:middle line:84% between kind of hardware and software, right? 01:45:19.330 --> 01:45:21.610 align:middle line:90% So we have people and stuff. 01:45:21.610 --> 01:45:24.290 align:middle line:90% I think that's fine. 01:45:24.290 --> 01:45:27.210 align:middle line:84% I think you'd be smart to have, even in just 01:45:27.210 --> 01:45:30.700 align:middle line:84% in your documentation, some examples of each of those 01:45:30.700 --> 01:45:33.500 align:middle line:84% because a lot of this is going to turn on relative delays. 01:45:33.500 --> 01:45:37.158 align:middle line:84% How long does it take to develop a soundstage or whatever. 01:45:37.158 --> 01:45:38.700 align:middle line:84% So I think you want examples in there 01:45:38.700 --> 01:45:39.780 align:middle line:90% just to help you think about. 01:45:39.780 --> 01:45:42.120 align:middle line:84% I have no idea how long it takes to build a soundstage, 01:45:42.120 --> 01:45:47.180 align:middle line:90% but that will matter. 01:45:47.180 --> 01:45:52.020 align:middle line:84% In this local attractiveness thing, 01:45:52.020 --> 01:45:53.980 align:middle line:84% this is probably the only detailed formulation 01:45:53.980 --> 01:45:59.300 align:middle line:84% feedback I will give, but I looked at this earlier. 01:45:59.300 --> 01:46:02.580 align:middle line:84% Almost certainly, you're going to want to use the-- 01:46:02.580 --> 01:46:04.580 align:middle line:84% and this goes generally for these attractiveness 01:46:04.580 --> 01:46:08.390 align:middle line:84% formulations, the multinomial logit formulation. 01:46:08.390 --> 01:46:09.640 align:middle line:90% It's discussed in John's book. 01:46:09.640 --> 01:46:11.900 align:middle line:90% You can find it on the web. 01:46:11.900 --> 01:46:14.220 align:middle line:84% It's just the version of this formulation that 01:46:14.220 --> 01:46:17.478 align:middle line:84% has the kind of most robust story underlying it 01:46:17.478 --> 01:46:20.020 align:middle line:84% because I think you had like just a fractional, like a convex 01:46:20.020 --> 01:46:22.640 align:middle line:90% combination, of the two things. 01:46:22.640 --> 01:46:26.000 align:middle line:84% So when doing these attractiveness, 01:46:26.000 --> 01:46:30.310 align:middle line:84% that should probably, in most cases, be the default option. 01:46:30.310 --> 01:46:32.010 align:middle line:84% And what I would suggest you do actually 01:46:32.010 --> 01:46:35.150 align:middle line:84% as a kind of analytical strategy is 01:46:35.150 --> 01:46:37.630 align:middle line:84% the equilibrium of this model, I think, 01:46:37.630 --> 01:46:41.710 align:middle line:84% I would suggest you start with most of the production is in LA, 01:46:41.710 --> 01:46:45.670 align:middle line:84% and then your intervention is the most of world 01:46:45.670 --> 01:46:48.030 align:middle line:90% gives this tax incentive. 01:46:48.030 --> 01:46:51.610 align:middle line:84% And then some small number of movies will go, 01:46:51.610 --> 01:46:55.030 align:middle line:84% but then you can see how these other feedbacks start 01:46:55.030 --> 01:46:58.630 align:middle line:84% to then grow stronger and maybe start driving 01:46:58.630 --> 01:47:01.330 align:middle line:90% some sort of more balance. 01:47:01.330 --> 01:47:03.570 align:middle line:84% And again, I'd do the two pieces. 01:47:03.570 --> 01:47:06.550 align:middle line:84% But I think this is a great question. 01:47:06.550 --> 01:47:09.750 align:middle line:84% And you may end up disaggregating 01:47:09.750 --> 01:47:12.678 align:middle line:84% across soundstages, visual effects, and stuff like that. 01:47:12.678 --> 01:47:14.470 align:middle line:84% But I would stick with kind of the hardware 01:47:14.470 --> 01:47:19.990 align:middle line:84% and the software for now, and sort of get this working 01:47:19.990 --> 01:47:23.353 align:middle line:90% so that you understand it. 01:47:23.353 --> 01:47:26.020 align:middle line:84% And it could lead to some really interesting policy conclusions, 01:47:26.020 --> 01:47:26.520 align:middle line:90% right? 01:47:26.520 --> 01:47:29.373 align:middle line:84% I mean, LA may have to work hard to keep-- 01:47:29.373 --> 01:47:31.540 align:middle line:84% because I don't know if they have cost disadvantage. 01:47:31.540 --> 01:47:32.720 align:middle line:84% I mean, it's probably more expensive to be 01:47:32.720 --> 01:47:33.678 align:middle line:90% in Southern California. 01:47:33.678 --> 01:47:37.900 align:middle line:84% Just so this may be a hard thing to do in order to-- 01:47:37.900 --> 01:47:40.480 align:middle line:90% 01:47:40.480 --> 01:47:42.600 align:middle line:90% it's a great question, though. 01:47:42.600 --> 01:47:45.660 align:middle line:84% So I mean, if you had a client for this, 01:47:45.660 --> 01:47:49.040 align:middle line:84% would it be the movie studios, or the Los Angeles 01:47:49.040 --> 01:47:49.960 align:middle line:90% Chamber of Commerce? 01:47:49.960 --> 01:47:50.280 align:middle line:90% Or-- 01:47:50.280 --> 01:47:51.238 align:middle line:90% STUDENT: Yeah, exactly. 01:47:51.238 --> 01:47:53.500 align:middle line:84% MPA, which is the [? parent ?] of all studios. 01:47:53.500 --> 01:47:56.533 align:middle line:90% And then I think, [INAUDIBLE]. 01:47:56.533 --> 01:47:58.200 align:middle line:84% NELSON REPENNING: They're going to think 01:47:58.200 --> 01:47:59.880 align:middle line:84% you're a Dr. Doom when you walk in and be, 01:47:59.880 --> 01:48:01.420 align:middle line:84% like, we're going to lose the whole thing if we 01:48:01.420 --> 01:48:02.080 align:middle line:90% don't change the-- 01:48:02.080 --> 01:48:04.760 align:middle line:84% STUDENT: Well, as you say, it's easy to sort of share the story, 01:48:04.760 --> 01:48:07.520 align:middle line:84% and like if other places have these tax incentives, 01:48:07.520 --> 01:48:11.160 align:middle line:84% and then what will be happening there is all these feedback 01:48:11.160 --> 01:48:12.628 align:middle line:84% loops, here's what necessarily will 01:48:12.628 --> 01:48:13.920 align:middle line:90% happen in terms of [INAUDIBLE]. 01:48:13.920 --> 01:48:14.920 align:middle line:90% NELSON REPENNING: Right. 01:48:14.920 --> 01:48:16.580 align:middle line:84% And tax incentives is only one lever. 01:48:16.580 --> 01:48:18.038 align:middle line:84% And I know we need to move forward. 01:48:18.038 --> 01:48:22.200 align:middle line:84% But you could imagine infrastructure investments. 01:48:22.200 --> 01:48:25.890 align:middle line:84% You could subsidize computing clusters 01:48:25.890 --> 01:48:28.270 align:middle line:84% for visual effects, which would require, 01:48:28.270 --> 01:48:32.588 align:middle line:84% I think, a pretty careful theory of what 01:48:32.588 --> 01:48:34.630 align:middle line:84% is the future of movie making going to look like, 01:48:34.630 --> 01:48:37.930 align:middle line:84% and what are going to be the scarce, hardest-to-replicate 01:48:37.930 --> 01:48:38.990 align:middle line:90% resources? 01:48:38.990 --> 01:48:40.510 align:middle line:90% Is it going to be soundstages? 01:48:40.510 --> 01:48:41.927 align:middle line:84% Is it going to be computing power? 01:48:41.927 --> 01:48:44.272 align:middle line:90% Is it going to be engineers? 01:48:44.272 --> 01:48:46.730 align:middle line:84% And also, I think there's some interesting questions about, 01:48:46.730 --> 01:48:52.710 align:middle line:84% as technology requires fewer and fewer physical assets, 01:48:52.710 --> 01:48:55.530 align:middle line:84% the terms of competition could change in some interesting ways 01:48:55.530 --> 01:48:57.050 align:middle line:90% as well. 01:48:57.050 --> 01:48:59.730 align:middle line:90% I think this is cool. 01:48:59.730 --> 01:49:01.410 align:middle line:90% OK, who's next? 01:49:01.410 --> 01:49:02.970 align:middle line:90% STUDENT: [INAUDIBLE] 01:49:02.970 --> 01:49:04.930 align:middle line:84% NELSON REPENNING: Yeah, did you-- 01:49:04.930 --> 01:49:06.310 align:middle line:90% I think I have yours, right? 01:49:06.310 --> 01:49:13.010 align:middle line:90% 01:49:13.010 --> 01:49:14.150 align:middle line:90% So is this one right here? 01:49:14.150 --> 01:49:15.510 align:middle line:90% STUDENT: [INAUDIBLE]. 01:49:15.510 --> 01:49:16.370 align:middle line:90% Yeah. 01:49:16.370 --> 01:49:18.050 align:middle line:90% NELSON REPENNING: OK. 01:49:18.050 --> 01:49:18.630 align:middle line:90% All right. 01:49:18.630 --> 01:49:20.820 align:middle line:90% Tell us about your problem. 01:49:20.820 --> 01:49:25.780 align:middle line:84% STUDENT: So it's about heat principles and the effects 01:49:25.780 --> 01:49:27.580 align:middle line:90% on health. 01:49:27.580 --> 01:49:28.860 align:middle line:90% NELSON REPENNING: Yeah. 01:49:28.860 --> 01:49:31.940 align:middle line:84% STUDENT: I'm starting with mortality. 01:49:31.940 --> 01:49:35.580 align:middle line:84% And basically, it should be a policy problem 01:49:35.580 --> 01:49:42.660 align:middle line:84% because we know that originally, the puzzle that I looked at 01:49:42.660 --> 01:49:46.800 align:middle line:84% was that basically, despite some measures we're taking, 01:49:46.800 --> 01:49:50.320 align:middle line:84% like early warning systems and awareness campaigns, 01:49:50.320 --> 01:49:53.780 align:middle line:84% if you look at the hot and hottest days, which 01:49:53.780 --> 01:49:58.460 align:middle line:84% are controlled in Celsius or [INAUDIBLE] Fahrenheit, 01:49:58.460 --> 01:50:01.780 align:middle line:84% there was still an increase in mortality. 01:50:01.780 --> 01:50:06.540 align:middle line:84% But I think the puzzle is not why because we know why. 01:50:06.540 --> 01:50:11.060 align:middle line:84% And we know that we have more hot days, and they're hotter. 01:50:11.060 --> 01:50:14.340 align:middle line:84% And we know that the population is getting older, 01:50:14.340 --> 01:50:22.030 align:middle line:84% and the housing is not made for this weather. 01:50:22.030 --> 01:50:23.750 align:middle line:84% So I think it's a [INAUDIBLE] question, 01:50:23.750 --> 01:50:27.970 align:middle line:84% and kind of seeing which policies 01:50:27.970 --> 01:50:31.350 align:middle line:90% could help to reduce this trend. 01:50:31.350 --> 01:50:36.470 align:middle line:84% And I think I probably have to take a step back and simplify 01:50:36.470 --> 01:50:37.190 align:middle line:90% this. 01:50:37.190 --> 01:50:40.230 align:middle line:84% But I think the question in my head 01:50:40.230 --> 01:50:44.430 align:middle line:84% was, what would this structure look like? 01:50:44.430 --> 01:50:48.390 align:middle line:84% Because I basically have the population 01:50:48.390 --> 01:50:52.470 align:middle line:90% separated in three cohorts. 01:50:52.470 --> 01:50:56.510 align:middle line:84% So because the most vulnerable are the young and the old, 01:50:56.510 --> 01:51:00.950 align:middle line:84% so I now have the population separated as young up to 15, 01:51:00.950 --> 01:51:05.990 align:middle line:84% and then adult from 15 to 65, and then 65 plus. 01:51:05.990 --> 01:51:09.910 align:middle line:84% And then I wanted to separate at least high and low income 01:51:09.910 --> 01:51:12.990 align:middle line:84% because that's another vulnerability factor. 01:51:12.990 --> 01:51:15.010 align:middle line:90% And then I wanted to separate-- 01:51:15.010 --> 01:51:17.880 align:middle line:90% 01:51:17.880 --> 01:51:18.828 align:middle line:90% [INAUDIBLE] 01:51:18.828 --> 01:51:22.750 align:middle line:90% 01:51:22.750 --> 01:51:24.400 align:middle line:90% Oh yeah, the housing. 01:51:24.400 --> 01:51:26.320 align:middle line:90% So [INAUDIBLE]. 01:51:26.320 --> 01:51:30.600 align:middle line:90% 01:51:30.600 --> 01:51:32.800 align:middle line:84% And yeah, I put some numbers, but they're just, 01:51:32.800 --> 01:51:36.700 align:middle line:84% like, fake numbers to see if this thing works. 01:51:36.700 --> 01:51:39.840 align:middle line:90% 01:51:39.840 --> 01:51:44.200 align:middle line:84% Yeah, this is just the very beginning 01:51:44.200 --> 01:51:46.360 align:middle line:90% of the population structure. 01:51:46.360 --> 01:51:49.560 align:middle line:84% NELSON REPENNING: OK, so the question 01:51:49.560 --> 01:51:52.440 align:middle line:84% you want to answer is, given finite government 01:51:52.440 --> 01:51:54.840 align:middle line:84% resources to manage-- or what's the best 01:51:54.840 --> 01:51:57.760 align:middle line:84% way to manage this problem, to minimize sort of climate 01:51:57.760 --> 01:52:00.540 align:middle line:90% change, heat-related mortality? 01:52:00.540 --> 01:52:01.900 align:middle line:90% OK, that sounds great. 01:52:01.900 --> 01:52:05.384 align:middle line:90% That seems pretty clear. 01:52:05.384 --> 01:52:08.140 align:middle line:84% OK, so what do y'all think about this structure? 01:52:08.140 --> 01:52:10.820 align:middle line:90% So you have four stocks, right? 01:52:10.820 --> 01:52:15.280 align:middle line:84% So there's population, and you have a low-income population, 01:52:15.280 --> 01:52:20.840 align:middle line:84% and adequate housing, low-income. 01:52:20.840 --> 01:52:24.557 align:middle line:84% So we have low-income people on adequate and inadequate housing, 01:52:24.557 --> 01:52:26.640 align:middle line:84% and they have different mortality rates presumably 01:52:26.640 --> 01:52:28.340 align:middle line:90% from the heat increase. 01:52:28.340 --> 01:52:30.720 align:middle line:84% And then the same thing for high income. 01:52:30.720 --> 01:52:33.520 align:middle line:84% And then you have this kind of maximally connected 01:52:33.520 --> 01:52:37.800 align:middle line:84% so that people are able to move between all 01:52:37.800 --> 01:52:39.160 align:middle line:90% the different states. 01:52:39.160 --> 01:52:42.040 align:middle line:84% STUDENT: This looks nice expected for the age cohorts. 01:52:42.040 --> 01:52:45.360 align:middle line:90% 01:52:45.360 --> 01:52:49.380 align:middle line:84% So first, I made this with 12 stocks. 01:52:49.380 --> 01:52:54.240 align:middle line:84% So each separate cohort with the-- just to understand, 01:52:54.240 --> 01:52:57.000 align:middle line:90% what do I have here [INAUDIBLE]. 01:52:57.000 --> 01:53:00.480 align:middle line:90% And then I simplified this way. 01:53:00.480 --> 01:53:02.800 align:middle line:84% In between this, I was thinking that maybe, I 01:53:02.800 --> 01:53:04.840 align:middle line:84% should have just one stock of population, 01:53:04.840 --> 01:53:06.800 align:middle line:90% and everything substituted. 01:53:06.800 --> 01:53:09.020 align:middle line:84% But then I couldn't really get my head around it, 01:53:09.020 --> 01:53:10.380 align:middle line:90% so I didn't say. 01:53:10.380 --> 01:53:12.520 align:middle line:90% 01:53:12.520 --> 01:53:14.270 align:middle line:84% NELSON REPENNING: OK, what do y'all think? 01:53:14.270 --> 01:53:17.130 align:middle line:90% 01:53:17.130 --> 01:53:19.210 align:middle line:90% Suggestions or critiques? 01:53:19.210 --> 01:53:22.570 align:middle line:84% STUDENT: I would stay away from the heat stratification 01:53:22.570 --> 01:53:25.170 align:middle line:90% for [INAUDIBLE]. 01:53:25.170 --> 01:53:27.970 align:middle line:84% I think if there was dynamics [INAUDIBLE]. 01:53:27.970 --> 01:53:32.008 align:middle line:90% 01:53:32.008 --> 01:53:32.800 align:middle line:90% STUDENT: Say again. 01:53:32.800 --> 01:53:33.300 align:middle line:90% Sorry. 01:53:33.300 --> 01:53:36.050 align:middle line:84% STUDENT: I would go smaller in terms of stratification. 01:53:36.050 --> 01:53:39.210 align:middle line:84% Instead of 12, I would stick with these four for now. 01:53:39.210 --> 01:53:40.370 align:middle line:90% STUDENT: Without that age? 01:53:40.370 --> 01:53:42.130 align:middle line:90% STUDENT: Yeah. 01:53:42.130 --> 01:53:42.930 align:middle line:90% For now. 01:53:42.930 --> 01:53:46.250 align:middle line:90% You can add more complexity. 01:53:46.250 --> 01:53:48.332 align:middle line:90% It's a lot of [? problem. ?] 01:53:48.332 --> 01:53:49.290 align:middle line:90% NELSON REPENNING: Yeah. 01:53:49.290 --> 01:53:52.170 align:middle line:84% STUDENT: Are there other feedbacks 01:53:52.170 --> 01:53:56.690 align:middle line:84% very important or linked to the flows between these stocks? 01:53:56.690 --> 01:54:01.170 align:middle line:84% Or are the dynamics driven by people transitioning 01:54:01.170 --> 01:54:03.190 align:middle line:90% from one state to another? 01:54:03.190 --> 01:54:08.730 align:middle line:84% Or just the dynamics are different for each group? 01:54:08.730 --> 01:54:11.845 align:middle line:90% STUDENT: I don't know exactly. 01:54:11.845 --> 01:54:12.720 align:middle line:90% It's a good question. 01:54:12.720 --> 01:54:17.780 align:middle line:84% I wouldn't imagine that they're driven by the transition, 01:54:17.780 --> 01:54:21.380 align:middle line:84% but I think the mechanisms are different. 01:54:21.380 --> 01:54:25.980 align:middle line:84% So that's why I wanted to separate them. 01:54:25.980 --> 01:54:27.540 align:middle line:90% But I'm not sure. 01:54:27.540 --> 01:54:32.100 align:middle line:90% I might have to change it. 01:54:32.100 --> 01:54:35.180 align:middle line:84% NELSON REPENNING: So what's your guesstimate 01:54:35.180 --> 01:54:38.600 align:middle line:84% over the appropriate time horizon for this analysis? 01:54:38.600 --> 01:54:39.100 align:middle line:90% Right? 01:54:39.100 --> 01:54:41.433 align:middle line:84% So how long are we going to have to look into the future 01:54:41.433 --> 01:54:45.040 align:middle line:84% to really think properly about this? 01:54:45.040 --> 01:54:49.633 align:middle line:84% So I'm imagining that it's not going to be like-- in one year, 01:54:49.633 --> 01:54:50.800 align:middle line:90% not much is going to change. 01:54:50.800 --> 01:54:52.910 align:middle line:90% And so this is a-- 01:54:52.910 --> 01:54:54.040 align:middle line:90% I think it's-- 01:54:54.040 --> 01:54:56.460 align:middle line:90% STUDENT: [INAUDIBLE] 01:54:56.460 --> 01:54:57.910 align:middle line:90% 01:54:57.910 --> 01:55:00.160 align:middle line:84% NELSON REPENNING: Yeah, I would guess probably longer. 01:55:00.160 --> 01:55:01.500 align:middle line:90% But so it's measured in decades. 01:55:01.500 --> 01:55:02.040 align:middle line:90% For sure. 01:55:02.040 --> 01:55:03.867 align:middle line:90% OK. 01:55:03.867 --> 01:55:05.700 align:middle line:84% STUDENT: Especially if you want to see that, 01:55:05.700 --> 01:55:10.470 align:middle line:84% because the population is aging, so if we want to capture that, 01:55:10.470 --> 01:55:14.810 align:middle line:84% since we expect more people to go to the elderly category. 01:55:14.810 --> 01:55:16.645 align:middle line:84% So more people will become older. 01:55:16.645 --> 01:55:17.770 align:middle line:90% NELSON REPENNING: Yeah, OK. 01:55:17.770 --> 01:55:19.850 align:middle line:84% So that makes a lot of sense to me. 01:55:19.850 --> 01:55:22.950 align:middle line:90% 01:55:22.950 --> 01:55:24.090 align:middle line:90% Yeah, so other thoughts? 01:55:24.090 --> 01:55:26.910 align:middle line:90% 01:55:26.910 --> 01:55:32.030 align:middle line:84% STUDENT: Yeah, I think I would probably separate it 01:55:32.030 --> 01:55:40.470 align:middle line:84% by age and then housing and that because income level 01:55:40.470 --> 01:55:45.270 align:middle line:84% and adequacy of housing are so [INAUDIBLE] here. 01:55:45.270 --> 01:55:49.550 align:middle line:84% But I think you could take out income level and just focus 01:55:49.550 --> 01:55:52.070 align:middle line:84% on housing and age because those are the things, 01:55:52.070 --> 01:55:55.710 align:middle line:84% like if you're an older vulnerable person, 01:55:55.710 --> 01:55:59.190 align:middle line:84% and inadequate housing-- that's the highest risk-- 01:55:59.190 --> 01:56:00.110 align:middle line:90% going to be. 01:56:00.110 --> 01:56:05.237 align:middle line:84% And the income's kind of in the background of that. 01:56:05.237 --> 01:56:07.320 align:middle line:84% STUDENT: I think you have a good point, especially 01:56:07.320 --> 01:56:11.560 align:middle line:84% because I noticed, actually, the income doesn't matter 01:56:11.560 --> 01:56:14.740 align:middle line:84% because you could be high income, 01:56:14.740 --> 01:56:19.640 align:middle line:84% but still the housing may be so old that it 01:56:19.640 --> 01:56:21.300 align:middle line:90% doesn't have any insulation. 01:56:21.300 --> 01:56:26.460 align:middle line:84% And it's kind of air con is not very present in Switzerland. 01:56:26.460 --> 01:56:29.080 align:middle line:84% Like, only the very modern buildings have it. 01:56:29.080 --> 01:56:30.215 align:middle line:90% So I think you're right. 01:56:30.215 --> 01:56:31.840 align:middle line:84% STUDENT: The money in your bank account 01:56:31.840 --> 01:56:33.540 align:middle line:90% doesn't save you from this. 01:56:33.540 --> 01:56:36.500 align:middle line:84% It's like, it's housing risk factor. 01:56:36.500 --> 01:56:38.280 align:middle line:84% STUDENT: I think that's a good point. 01:56:38.280 --> 01:56:39.920 align:middle line:90% STUDENT: But people tend to-- 01:56:39.920 --> 01:56:43.495 align:middle line:84% I mean, all the people live with children, obviously. 01:56:43.495 --> 01:56:44.120 align:middle line:90% STUDENT: Sorry? 01:56:44.120 --> 01:56:47.400 align:middle line:90% STUDENT: Muslim house. 01:56:47.400 --> 01:56:50.220 align:middle line:84% I mean, if you have 15-year-old, I mean, 01:56:50.220 --> 01:56:52.537 align:middle line:84% they of course live with the older people. 01:56:52.537 --> 01:56:53.120 align:middle line:90% STUDENT: Yeah. 01:56:53.120 --> 01:56:55.080 align:middle line:90% STUDENT: Like, in the other. 01:56:55.080 --> 01:56:57.460 align:middle line:84% So I'd rather aggregate it in two. 01:56:57.460 --> 01:56:59.163 align:middle line:90% I mean, the old and the-- 01:56:59.163 --> 01:57:01.080 align:middle line:84% because you have the children living with the, 01:57:01.080 --> 01:57:04.800 align:middle line:90% I don't know, 50 minors. 01:57:04.800 --> 01:57:06.790 align:middle line:84% STUDENT: I mean, if you have it by age, 01:57:06.790 --> 01:57:09.170 align:middle line:84% then there might be a person in the young person stock 01:57:09.170 --> 01:57:11.890 align:middle line:84% and a person from the old person stock who are both living-- 01:57:11.890 --> 01:57:13.350 align:middle line:90% STUDENT: In the same house. 01:57:13.350 --> 01:57:15.690 align:middle line:90% STUDENT: [INAUDIBLE] 01:57:15.690 --> 01:57:18.310 align:middle line:90% 01:57:18.310 --> 01:57:20.890 align:middle line:84% STUDENT: Yeah, I think the income kind of makes 01:57:20.890 --> 01:57:25.890 align:middle line:84% sense to me because if you're designing a policy 01:57:25.890 --> 01:57:27.870 align:middle line:84% and you want to see how it changes, 01:57:27.870 --> 01:57:31.090 align:middle line:84% moving somebody from low income to high income 01:57:31.090 --> 01:57:34.610 align:middle line:84% feels like less important than a government's 01:57:34.610 --> 01:57:39.430 align:middle line:84% perspective than moving them from the ability to being 01:57:39.430 --> 01:57:43.530 align:middle line:84% in adequate housing to [INAUDIBLE]. 01:57:43.530 --> 01:57:45.570 align:middle line:84% So it feels like it distracts you 01:57:45.570 --> 01:57:47.810 align:middle line:84% from the focus of the [INAUDIBLE]. 01:57:47.810 --> 01:57:53.330 align:middle line:84% STUDENT: So you're also going to get rid of the income? 01:57:53.330 --> 01:57:58.850 align:middle line:84% STUDENT: Just in terms of modeling actual sort of 01:57:58.850 --> 01:57:59.890 align:middle line:90% [INAUDIBLE]. 01:57:59.890 --> 01:58:03.460 align:middle line:84% I think it's harder to make that sort of leap 01:58:03.460 --> 01:58:05.700 align:middle line:84% from a policy persepective [INAUDIBLE]. 01:58:05.700 --> 01:58:10.005 align:middle line:90% 01:58:10.005 --> 01:58:11.380 align:middle line:84% NELSON REPENNING: So do you think 01:58:11.380 --> 01:58:14.940 align:middle line:84% you'll have wealthy people who might still 01:58:14.940 --> 01:58:16.860 align:middle line:90% be in inadequate housing? 01:58:16.860 --> 01:58:17.560 align:middle line:90% You do? 01:58:17.560 --> 01:58:18.580 align:middle line:90% STUDENT: Yeah. 01:58:18.580 --> 01:58:20.980 align:middle line:84% STUDENT: It's illegal to have air conditioning, at least 01:58:20.980 --> 01:58:22.880 align:middle line:90% in Geneva, in your house. 01:58:22.880 --> 01:58:24.460 align:middle line:90% I mean, it's not legal. 01:58:24.460 --> 01:58:28.300 align:middle line:84% So even if you're rich, I mean, it's not enough. 01:58:28.300 --> 01:58:30.980 align:middle line:90% STUDENT: Yeah. 01:58:30.980 --> 01:58:33.860 align:middle line:84% STUDENT: So but the policy question 01:58:33.860 --> 01:58:36.020 align:middle line:84% is, should you allow for air conditioning? 01:58:36.020 --> 01:58:37.040 align:middle line:90% STUDENT: In houses. 01:58:37.040 --> 01:58:41.700 align:middle line:90% [INTERPOSING VOICES] 01:58:41.700 --> 01:58:43.760 align:middle line:84% STUDENT: I mean, you sleep in the basement. 01:58:43.760 --> 01:58:47.090 align:middle line:84% That's what-- yeah-- if you have a basement. 01:58:47.090 --> 01:58:50.220 align:middle line:90% [LAUGHTER] 01:58:50.220 --> 01:58:51.660 align:middle line:90% 01:58:51.660 --> 01:58:53.353 align:middle line:90% STUDENT: I do hope [INAUDIBLE]. 01:58:53.353 --> 01:58:54.520 align:middle line:90% STUDENT: At least in Geneva. 01:58:54.520 --> 01:58:55.160 align:middle line:90% It's not a lot. 01:58:55.160 --> 01:58:57.620 align:middle line:84% STUDENT: Is it for all buildings? 01:58:57.620 --> 01:59:00.380 align:middle line:84% STUDENT: It's at least for where you live. 01:59:00.380 --> 01:59:02.830 align:middle line:90% It's allowed for offices. 01:59:02.830 --> 01:59:06.110 align:middle line:90% And if you use-- yeah. 01:59:06.110 --> 01:59:10.470 align:middle line:84% STUDENT: But does it depend on the age of the house? 01:59:10.470 --> 01:59:15.690 align:middle line:84% Maybe some of the very new buildings, they have it already? 01:59:15.690 --> 01:59:18.110 align:middle line:84% But if it's like an old property, 01:59:18.110 --> 01:59:19.330 align:middle line:90% you're not allowed to make-- 01:59:19.330 --> 01:59:20.790 align:middle line:84% STUDENT: Well, you have to check. 01:59:20.790 --> 01:59:24.410 align:middle line:84% I lived there, and there was no aircon in the houses. 01:59:24.410 --> 01:59:28.350 align:middle line:84% And you're not allowed to install the aircon. 01:59:28.350 --> 01:59:30.990 align:middle line:84% But it's different with office buildings 01:59:30.990 --> 01:59:34.810 align:middle line:84% and maybe if it's a certain size or certain efficiency. 01:59:34.810 --> 01:59:46.270 align:middle line:84% And they have a system coming from the lake of Geneva. 01:59:46.270 --> 01:59:49.690 align:middle line:84% So if you use that, so you get the [INAUDIBLE]. 01:59:49.690 --> 01:59:53.210 align:middle line:84% So it's also a regulatory thing that you're [INAUDIBLE]. 01:59:53.210 --> 01:59:56.110 align:middle line:84% NELSON REPENNING: So what's your dynamic hypothesis, right? 01:59:56.110 --> 01:59:59.280 align:middle line:90% 01:59:59.280 --> 02:00:00.620 align:middle line:90% This is a guess, right? 02:00:00.620 --> 02:00:01.680 align:middle line:84% It might make you uncomfortable the way 02:00:01.680 --> 02:00:02.860 align:middle line:90% it made Jason uncomfortable. 02:00:02.860 --> 02:00:07.720 align:middle line:84% But what do you think is the most important dynamic that's 02:00:07.720 --> 02:00:10.840 align:middle line:90% going to drive this? 02:00:10.840 --> 02:00:14.520 align:middle line:84% Or do you have a guess at which policy you think 02:00:14.520 --> 02:00:16.680 align:middle line:90% has the most leverage here? 02:00:16.680 --> 02:00:19.600 align:middle line:84% STUDENT: I think it would be some update of the building's 02:00:19.600 --> 02:00:27.240 align:middle line:84% insulation, some type of passive cooling, 02:00:27.240 --> 02:00:32.580 align:middle line:84% and probably more like green areas on the streets and so on. 02:00:32.580 --> 02:00:33.723 align:middle line:90% [INAUDIBLE] 02:00:33.723 --> 02:00:35.640 align:middle line:84% NELSON REPENNING: So what I would suggest here 02:00:35.640 --> 02:00:38.600 align:middle line:84% is I think a lot of people are kind 02:00:38.600 --> 02:00:41.080 align:middle line:84% of saying in different ways, I think 02:00:41.080 --> 02:00:42.920 align:middle line:84% your stock flow structure is probably 02:00:42.920 --> 02:00:46.600 align:middle line:90% too complicated for now. 02:00:46.600 --> 02:00:49.200 align:middle line:84% For example, right now, you allow 02:00:49.200 --> 02:00:53.280 align:middle line:84% people to move from low income to high income and back again. 02:00:53.280 --> 02:00:57.770 align:middle line:84% I'm sure that there is some transition between those states, 02:00:57.770 --> 02:01:00.618 align:middle line:84% but you would have to work pretty hard to convince me 02:01:00.618 --> 02:01:02.410 align:middle line:84% that's going to be relevant for the problem 02:01:02.410 --> 02:01:04.035 align:middle line:84% that you're trying to solve because I'm 02:01:04.035 --> 02:01:07.190 align:middle line:90% guessing that's a fairly tiny-- 02:01:07.190 --> 02:01:09.110 align:middle line:84% relative to the dynamics you're talking about, 02:01:09.110 --> 02:01:11.690 align:middle line:84% that's a relatively small fraction. 02:01:11.690 --> 02:01:15.230 align:middle line:84% So you could imagine just a stock flow chain, right? 02:01:15.230 --> 02:01:18.210 align:middle line:84% I think one dynamic that you said that to me sounds important 02:01:18.210 --> 02:01:24.570 align:middle line:84% is just the demographics, which is that the balance of the age 02:01:24.570 --> 02:01:27.068 align:middle line:90% cohorts is going to skew older. 02:01:27.068 --> 02:01:29.610 align:middle line:84% So which could make this problem worse because old people are 02:01:29.610 --> 02:01:31.930 align:middle line:90% more susceptible. 02:01:31.930 --> 02:01:33.690 align:middle line:84% And I'm perfectly comfortable for now 02:01:33.690 --> 02:01:39.110 align:middle line:84% with your kind of, we have zero to 15, 15 to 65, and 65 plus. 02:01:39.110 --> 02:01:40.770 align:middle line:90% That seems totally fine. 02:01:40.770 --> 02:01:44.747 align:middle line:84% And you could make that down to one-year age increments 02:01:44.747 --> 02:01:46.830 align:middle line:84% at some point, if you wanted to be really precise. 02:01:46.830 --> 02:01:48.830 align:middle line:84% But I don't think you need that as a first step. 02:01:48.830 --> 02:01:53.630 align:middle line:84% So I would maybe start with just that aging chain of 15, 02:01:53.630 --> 02:01:56.150 align:middle line:90% 15 to 65, and 65 plus. 02:01:56.150 --> 02:01:58.750 align:middle line:84% And then presumably, those data are available. 02:01:58.750 --> 02:02:01.290 align:middle line:84% So you could get that to match the existing 02:02:01.290 --> 02:02:02.535 align:middle line:90% demographics roughly. 02:02:02.535 --> 02:02:04.410 align:middle line:84% And there's probably good forecasts out there 02:02:04.410 --> 02:02:06.410 align:middle line:84% because these are the kinds of things that are-- 02:02:06.410 --> 02:02:10.310 align:middle line:84% I don't know how the retirement funding system works here. 02:02:10.310 --> 02:02:12.867 align:middle line:84% But if it was the US, our Social Security Administration 02:02:12.867 --> 02:02:14.450 align:middle line:84% is forecasting this stuff all the time 02:02:14.450 --> 02:02:16.890 align:middle line:90% in terms of what happens. 02:02:16.890 --> 02:02:19.730 align:middle line:84% And you could do that either for low income 02:02:19.730 --> 02:02:22.650 align:middle line:84% only, or for low income and high income. 02:02:22.650 --> 02:02:27.010 align:middle line:90% I'm kind of agnostic about that. 02:02:27.010 --> 02:02:29.370 align:middle line:84% Given that you think that this may well 02:02:29.370 --> 02:02:31.230 align:middle line:84% be a problem for high-income people, 02:02:31.230 --> 02:02:34.810 align:middle line:84% I might just do one aging chain for now. 02:02:34.810 --> 02:02:37.190 align:middle line:84% And then I think the other thing I would do would be, 02:02:37.190 --> 02:02:39.825 align:middle line:84% and this is kind of a system dynamics classic 02:02:39.825 --> 02:02:41.450 align:middle line:84% that would make John Sterman very, very 02:02:41.450 --> 02:02:45.650 align:middle line:84% happy is to then have a similar aging chain for the housing 02:02:45.650 --> 02:02:52.410 align:middle line:84% stock because presumably, the newer buildings are, 02:02:52.410 --> 02:02:54.900 align:middle line:84% even if they don't have air conditioning, 02:02:54.900 --> 02:02:57.700 align:middle line:84% or ventilated better, or have other cooling mechanisms, 02:02:57.700 --> 02:03:01.500 align:middle line:84% or whatever, I don't really have a good instinct 02:03:01.500 --> 02:03:04.020 align:middle line:84% off the top of my head about how many stocks 02:03:04.020 --> 02:03:05.622 align:middle line:90% you would want in that chain. 02:03:05.622 --> 02:03:07.580 align:middle line:84% You might have to do a little research in terms 02:03:07.580 --> 02:03:12.060 align:middle line:84% of the innovations, but you might do it in 20-year age 02:03:12.060 --> 02:03:13.440 align:middle line:90% cohorts or something like that. 02:03:13.440 --> 02:03:17.660 align:middle line:84% And then I think if you know the rough demographics 02:03:17.660 --> 02:03:21.360 align:middle line:84% and you have some sense of the age of the housing stock, 02:03:21.360 --> 02:03:23.940 align:middle line:84% then you can use the perfect mixing assumption 02:03:23.940 --> 02:03:26.860 align:middle line:84% to kind of calculate just some averages of, 02:03:26.860 --> 02:03:29.300 align:middle line:84% given the number of people and the number of houses, 02:03:29.300 --> 02:03:30.920 align:middle line:84% assuming some kind of perfect mixing, 02:03:30.920 --> 02:03:36.300 align:middle line:84% we are going to have x number of vulnerable people living 02:03:36.300 --> 02:03:38.640 align:middle line:90% in outdated housing. 02:03:38.640 --> 02:03:41.820 align:middle line:84% And then from there, you could sort of forecast the mortality 02:03:41.820 --> 02:03:43.440 align:middle line:90% rates that go with that. 02:03:43.440 --> 02:03:47.245 align:middle line:84% And I don't know what kind of data are out there. 02:03:47.245 --> 02:03:49.620 align:middle line:84% So I would say if you could get that to work, and to kind 02:03:49.620 --> 02:03:53.933 align:middle line:84% of at least roughly match the data that you have. 02:03:53.933 --> 02:03:55.350 align:middle line:84% And the reason I'm chuckling about 02:03:55.350 --> 02:03:58.550 align:middle line:84% this is if you really want to go deep tracks, 02:03:58.550 --> 02:04:00.550 align:middle line:84% this was essentially the question that John 02:04:00.550 --> 02:04:05.750 align:middle line:84% Sterman answered in his general exam, which was kind of, how 02:04:05.750 --> 02:04:11.070 align:middle line:84% quickly would the housing stocks energy consumption change based 02:04:11.070 --> 02:04:14.250 align:middle line:84% on new housing, and retrofits, and stuff like that? 02:04:14.250 --> 02:04:17.890 align:middle line:84% So I remember he handed this to me when I was a PhD student. 02:04:17.890 --> 02:04:19.490 align:middle line:84% So documentation of this must exist. 02:04:19.490 --> 02:04:20.490 align:middle line:90% It was in Dynamo, right? 02:04:20.490 --> 02:04:25.710 align:middle line:84% It was in before Stella, or Vensim, or whatever. 02:04:25.710 --> 02:04:28.630 align:middle line:84% But then I think it would allow you, even that would allow, 02:04:28.630 --> 02:04:31.290 align:middle line:84% some interesting kind of policy explorations basically, 02:04:31.290 --> 02:04:34.990 align:middle line:84% which would boil down to, given your hypothesis is, 02:04:34.990 --> 02:04:38.390 align:middle line:90% should we focus on new houses? 02:04:38.390 --> 02:04:41.030 align:middle line:90% Or should we focus on retrofits? 02:04:41.030 --> 02:04:43.170 align:middle line:84% And I don't know anything about this world. 02:04:43.170 --> 02:04:45.310 align:middle line:84% So I don't even know if retrofitting is even 02:04:45.310 --> 02:04:48.070 align:middle line:90% remotely possible. 02:04:48.070 --> 02:04:50.960 align:middle line:84% But there might also be a little bit of a shock-and-awe thing, 02:04:50.960 --> 02:04:51.460 align:middle line:90% right? 02:04:51.460 --> 02:04:54.120 align:middle line:84% If you take the kind of increase in mortality from temperature 02:04:54.120 --> 02:04:57.080 align:middle line:84% and mesh that with the changing demographics, 02:04:57.080 --> 02:05:00.240 align:middle line:84% I could imagine that 20 years from now, the mortality 02:05:00.240 --> 02:05:02.120 align:middle line:84% rates, the raw numbers could look 02:05:02.120 --> 02:05:05.575 align:middle line:84% much higher because of the confluence of those two things. 02:05:05.575 --> 02:05:07.200 align:middle line:84% And who knows whether that's showing up 02:05:07.200 --> 02:05:10.460 align:middle line:90% in anyone else's forecast? 02:05:10.460 --> 02:05:12.840 align:middle line:84% But I think that would be a simpler formulation than what 02:05:12.840 --> 02:05:15.160 align:middle line:84% you have because right now, you have, essentially, 02:05:15.160 --> 02:05:16.960 align:middle line:84% a lot of transition probabilities from one 02:05:16.960 --> 02:05:21.040 align:middle line:84% state to the next that I think, to first order, you can assume 02:05:21.040 --> 02:05:24.260 align:middle line:84% are not relevant to this problem. 02:05:24.260 --> 02:05:29.308 align:middle line:84% I mean, I suppose you could imagine another policy, 02:05:29.308 --> 02:05:31.100 align:middle line:84% which would be rather than building houses, 02:05:31.100 --> 02:05:32.642 align:middle line:84% we could just give people more money. 02:05:32.642 --> 02:05:36.100 align:middle line:84% So we could move more people from low income to high income. 02:05:36.100 --> 02:05:40.680 align:middle line:84% But it didn't sound like you thought that the income 02:05:40.680 --> 02:05:43.103 align:middle line:84% demographic was going to-- or the income stratification 02:05:43.103 --> 02:05:44.520 align:middle line:84% was going to make a big difference 02:05:44.520 --> 02:05:45.625 align:middle line:90% in terms of mortality. 02:05:45.625 --> 02:05:48.250 align:middle line:84% So you're going to have to make some bets on what you think are 02:05:48.250 --> 02:05:50.950 align:middle line:90% the higher leverage policies. 02:05:50.950 --> 02:05:52.120 align:middle line:90% But-- 02:05:52.120 --> 02:05:55.630 align:middle line:84% STUDENT: And just to make sure I understand correctly, 02:05:55.630 --> 02:06:00.290 align:middle line:84% so you're saying half the stock and flow aging structure 02:06:00.290 --> 02:06:04.290 align:middle line:84% just for the age, just for the-- so half 02:06:04.290 --> 02:06:08.090 align:middle line:84% is subscripted by cohort, by age cohort, 02:06:08.090 --> 02:06:14.450 align:middle line:84% remove from the people stock and flow the housing 02:06:14.450 --> 02:06:20.050 align:middle line:90% and the socioeconomic status? 02:06:20.050 --> 02:06:22.610 align:middle line:84% NELSON REPENNING: So I'm not sure I would even subscript it, 02:06:22.610 --> 02:06:25.513 align:middle line:84% because again, I tend to get confused with the subscripts 02:06:25.513 --> 02:06:27.430 align:middle line:84% until I really understand what the model does. 02:06:27.430 --> 02:06:29.180 align:middle line:84% But that's sort of neither here nor there. 02:06:29.180 --> 02:06:32.090 align:middle line:84% I think I would start with just a 02:06:32.090 --> 02:06:37.130 align:middle line:84% exactly as you delineated a three stock chain, 0 to 15, 15 02:06:37.130 --> 02:06:40.290 align:middle line:84% to 65, 65 plus, you probably want third-order delays 02:06:40.290 --> 02:06:42.650 align:middle line:90% in between those. 02:06:42.650 --> 02:06:46.580 align:middle line:84% And again, you could go more detailed if you want, 02:06:46.580 --> 02:06:50.180 align:middle line:84% but I would only do that if you had some data suggesting 02:06:50.180 --> 02:06:53.180 align:middle line:84% that the mortality rates, you would have 02:06:53.180 --> 02:06:54.720 align:middle line:90% to have pretty granular data. 02:06:54.720 --> 02:06:58.940 align:middle line:84% If I asked you, what's the mortality rate for 50-year-olds 02:06:58.940 --> 02:07:00.320 align:middle line:90% versus 55-year-olds? 02:07:00.320 --> 02:07:03.880 align:middle line:84% I'm guessing that you don't have that level of precision of-- 02:07:03.880 --> 02:07:05.550 align:middle line:90% STUDENT: I [INAUDIBLE]. 02:07:05.550 --> 02:07:07.800 align:middle line:84% NELSON REPENNING: Yeah, so you could go look for that. 02:07:07.800 --> 02:07:10.060 align:middle line:84% But I would do that, and then you 02:07:10.060 --> 02:07:12.780 align:middle line:84% have mortality rates out of that. 02:07:12.780 --> 02:07:19.360 align:middle line:84% And then I think, as a general rule, or guideline, 02:07:19.360 --> 02:07:20.980 align:middle line:84% you want the stocks in your model 02:07:20.980 --> 02:07:23.560 align:middle line:84% to match the physical stocks in the system. 02:07:23.560 --> 02:07:25.420 align:middle line:84% And if you think about what you're doing, 02:07:25.420 --> 02:07:29.100 align:middle line:84% you have a stock of people, and the housing 02:07:29.100 --> 02:07:30.680 align:middle line:90% is a very physical stock. 02:07:30.680 --> 02:07:33.200 align:middle line:84% And so I think I would capture the housing. 02:07:33.200 --> 02:07:35.220 align:middle line:84% And I think the reason capturing The stock flow 02:07:35.220 --> 02:07:37.220 align:middle line:84% chain of the housing might be helpful is then, 02:07:37.220 --> 02:07:41.040 align:middle line:84% it would give you a very clear way of thinking about, 02:07:41.040 --> 02:07:42.640 align:middle line:90% do we build new houses? 02:07:42.640 --> 02:07:44.450 align:middle line:90% Do we retrofit old ones? 02:07:44.450 --> 02:07:45.990 align:middle line:90% Do we destroy old ones? 02:07:45.990 --> 02:07:48.655 align:middle line:84% Because now, the policy choices are 02:07:48.655 --> 02:07:50.030 align:middle line:84% going to be basically represented 02:07:50.030 --> 02:07:52.490 align:middle line:90% by the flows in that. 02:07:52.490 --> 02:07:54.790 align:middle line:84% And then I think, to calculate the mortality, which 02:07:54.790 --> 02:07:57.110 align:middle line:84% is the thing that you're after, now you 02:07:57.110 --> 02:07:59.230 align:middle line:84% can do some sort of simple algebra of, 02:07:59.230 --> 02:08:02.730 align:middle line:84% we have this many housing units in these states. 02:08:02.730 --> 02:08:05.030 align:middle line:84% We could then calculate kind of on average, 02:08:05.030 --> 02:08:07.110 align:middle line:84% how many people in the 0 to 15 cohort 02:08:07.110 --> 02:08:11.150 align:middle line:84% are in inadequate housing, for example? 02:08:11.150 --> 02:08:14.130 align:middle line:84% And then you could enrich that a million different ways. 02:08:14.130 --> 02:08:16.250 align:middle line:84% But I would get that kind of, again, 02:08:16.250 --> 02:08:18.590 align:middle line:84% sort of bare bones formulation sort of working. 02:08:18.590 --> 02:08:21.070 align:middle line:84% And I think in this case, again, you 02:08:21.070 --> 02:08:24.775 align:middle line:84% could probably have it loosely fit the available data. 02:08:24.775 --> 02:08:27.150 align:middle line:84% I mean, with that simple model, you're probably not going 02:08:27.150 --> 02:08:28.470 align:middle line:84% to get super high fidelity, but at least you 02:08:28.470 --> 02:08:29.970 align:middle line:84% can get the order of magnitudes kind 02:08:29.970 --> 02:08:32.230 align:middle line:84% of in the right ballpark, which I think 02:08:32.230 --> 02:08:34.910 align:middle line:90% would also be a good start. 02:08:34.910 --> 02:08:35.850 align:middle line:90% Does that make sense? 02:08:35.850 --> 02:08:36.150 align:middle line:90% STUDENT: Yes. 02:08:36.150 --> 02:08:36.970 align:middle line:90% NELSON REPENNING: OK, that's it. 02:08:36.970 --> 02:08:37.770 align:middle line:90% Another good one. 02:08:37.770 --> 02:08:41.595 align:middle line:90% OK, so who do we have left? 02:08:41.595 --> 02:08:43.080 align:middle line:90% STUDENT: I can also go next. 02:08:43.080 --> 02:08:45.320 align:middle line:90% NELSON REPENNING: OK, great. 02:08:45.320 --> 02:08:46.620 align:middle line:90% Do I have yours in my-- 02:08:46.620 --> 02:08:47.280 align:middle line:90% STUDENT: Yes. 02:08:47.280 --> 02:08:48.405 align:middle line:90% NELSON REPENNING: Yeah, OK. 02:08:48.405 --> 02:08:51.640 align:middle line:90% 02:08:51.640 --> 02:08:52.480 align:middle line:90% Which one? 02:08:52.480 --> 02:08:53.380 align:middle line:90% This one? 02:08:53.380 --> 02:08:53.880 align:middle line:90% OK. 02:08:53.880 --> 02:08:58.600 align:middle line:90% 02:08:58.600 --> 02:09:01.640 align:middle line:84% STUDENT: So first of all, the model is halfway there. 02:09:01.640 --> 02:09:04.200 align:middle line:84% It's a combination of a stock management 02:09:04.200 --> 02:09:06.880 align:middle line:84% problem for clinical resources problem 02:09:06.880 --> 02:09:13.160 align:middle line:84% and the self-governance mechanisms 02:09:13.160 --> 02:09:15.760 align:middle line:90% we have in econ division. 02:09:15.760 --> 02:09:18.680 align:middle line:90% 02:09:18.680 --> 02:09:21.800 align:middle line:84% And they're [INAUDIBLE] 1980s and 1990s. 02:09:21.800 --> 02:09:24.400 align:middle line:84% They did a bunch of experiments showing 02:09:24.400 --> 02:09:29.380 align:middle line:84% that actually, communities can self-sustain their resources. 02:09:29.380 --> 02:09:32.560 align:middle line:84% And they introduced a controlling mechanism, 02:09:32.560 --> 02:09:35.040 align:middle line:84% like a punishment mechanism, so that if I 02:09:35.040 --> 02:09:39.650 align:middle line:84% see that Jason is using that resource more 02:09:39.650 --> 02:09:41.650 align:middle line:84% than he should have used, so I'm going 02:09:41.650 --> 02:09:45.370 align:middle line:84% to impose a punishment like this controlling mechanism 02:09:45.370 --> 02:09:49.370 align:middle line:84% to control his vehicle, break his speed. 02:09:49.370 --> 02:09:53.950 align:middle line:84% And I have shown that this have worked for a lot of situations. 02:09:53.950 --> 02:09:57.970 align:middle line:84% But there have been a grey area in the empirical data that is 02:09:57.970 --> 02:10:00.810 align:middle line:84% still some of the comments, despite these mechanisms, 02:10:00.810 --> 02:10:02.330 align:middle line:90% have failed. 02:10:02.330 --> 02:10:05.210 align:middle line:84% And during my PhD, I realized that one 02:10:05.210 --> 02:10:08.410 align:middle line:84% of the things that we haven't recognized 02:10:08.410 --> 02:10:12.650 align:middle line:84% is that all this decision making and capital 02:10:12.650 --> 02:10:14.570 align:middle line:84% investment, and resource regeneration, 02:10:14.570 --> 02:10:17.090 align:middle line:90% all involve delays. 02:10:17.090 --> 02:10:20.330 align:middle line:84% And we actually replicated that if the delay 02:10:20.330 --> 02:10:24.250 align:middle line:84% is short in this context, so these punishment mechanisms 02:10:24.250 --> 02:10:27.450 align:middle line:90% of governance mechanism will go. 02:10:27.450 --> 02:10:32.970 align:middle line:84% But if the delay is longer because the self stock 02:10:32.970 --> 02:10:36.010 align:middle line:84% management problem becomes harder, 02:10:36.010 --> 02:10:38.540 align:middle line:84% then those self-governance mechanism 02:10:38.540 --> 02:10:43.140 align:middle line:84% to actually make the situation even worse because people start 02:10:43.140 --> 02:10:47.240 align:middle line:84% blaming each other, that in that case, for example, 02:10:47.240 --> 02:10:49.680 align:middle line:84% Jason has corrected himself, his behavior. 02:10:49.680 --> 02:10:52.940 align:middle line:84% But because I'm not seeing the immediate outcome 02:10:52.940 --> 02:10:56.940 align:middle line:84% of his corrective actions, so I keep blaming him 02:10:56.940 --> 02:11:00.605 align:middle line:84% and I feel that-- sorry, that's [INAUDIBLE]. 02:11:00.605 --> 02:11:01.480 align:middle line:90% STUDENT: No, it's OK. 02:11:01.480 --> 02:11:04.560 align:middle line:90% I think this is awesome. 02:11:04.560 --> 02:11:08.560 align:middle line:84% STUDENT: That is the general thing that I wanted. 02:11:08.560 --> 02:11:11.140 align:middle line:84% So we have done experiments, actually, on this. 02:11:11.140 --> 02:11:17.100 align:middle line:84% It was a open model, not a closed feedback one. 02:11:17.100 --> 02:11:20.980 align:middle line:84% So right now, I'm trying to close all the feedback loops 02:11:20.980 --> 02:11:23.660 align:middle line:90% so that I can run experiments. 02:11:23.660 --> 02:11:26.080 align:middle line:84% NELSON REPENNING: So you probably know this, 02:11:26.080 --> 02:11:28.260 align:middle line:84% but there's a fairly substantial literature 02:11:28.260 --> 02:11:32.607 align:middle line:84% in system dynamics on this and the famous fish banks game, 02:11:32.607 --> 02:11:33.440 align:middle line:90% and stuff like that. 02:11:33.440 --> 02:11:37.060 align:middle line:84% And actually, I don't know how much experimental evidence 02:11:37.060 --> 02:11:37.720 align:middle line:90% is out there. 02:11:37.720 --> 02:11:41.500 align:middle line:84% But so given that this is a place where there's 02:11:41.500 --> 02:11:45.740 align:middle line:84% existing work, what's the puzzle kind of in the existing 02:11:45.740 --> 02:11:48.260 align:middle line:84% literature that you're trying to sort? 02:11:48.260 --> 02:11:50.625 align:middle line:84% I just want to make sure I'm clear on-- 02:11:50.625 --> 02:11:52.000 align:middle line:84% STUDENT: You're absolutely right. 02:11:52.000 --> 02:11:55.140 align:middle line:84% The puzzle is that, first of all, 02:11:55.140 --> 02:11:58.460 align:middle line:84% these self-governance mechanisms haven't really 02:11:58.460 --> 02:12:02.520 align:middle line:84% been captured thoroughly, at least, and understand. 02:12:02.520 --> 02:12:07.580 align:middle line:84% But the puzzle is, why is this contradictory outcomes 02:12:07.580 --> 02:12:10.500 align:middle line:84% or observations that some studies suggest 02:12:10.500 --> 02:12:14.740 align:middle line:84% that these mechanisms are working, but on the other hand, 02:12:14.740 --> 02:12:16.780 align:middle line:84% the data shows that in some situations, which 02:12:16.780 --> 02:12:19.620 align:middle line:84% is a little bit also unclear in the literature, that they're 02:12:19.620 --> 02:12:20.760 align:middle line:90% failing as well? 02:12:20.760 --> 02:12:23.400 align:middle line:84% NELSON REPENNING: OK, so at a very 50,000 feet, 02:12:23.400 --> 02:12:25.500 align:middle line:84% the empirical puzzle is, in some cases, they work, 02:12:25.500 --> 02:12:28.340 align:middle line:84% in some cases-- and just I'm not super familiar with this 02:12:28.340 --> 02:12:28.840 align:middle line:90% literature. 02:12:28.840 --> 02:12:31.940 align:middle line:84% What's an example of a self-governance? 02:12:31.940 --> 02:12:35.710 align:middle line:84% STUDENT: So in the case of fisheries, people sit together 02:12:35.710 --> 02:12:40.190 align:middle line:90% and they created an NGO. 02:12:40.190 --> 02:12:42.390 align:middle line:90% And each of us have a license. 02:12:42.390 --> 02:12:46.770 align:middle line:84% So if somebody deviates from agreed-upon behavior, 02:12:46.770 --> 02:12:48.510 align:middle line:84% their license is being revoked, they're 02:12:48.510 --> 02:12:51.590 align:middle line:84% not going to have access to that system. 02:12:51.590 --> 02:12:53.390 align:middle line:90% NELSON REPENNING: OK. 02:12:53.390 --> 02:12:56.867 align:middle line:84% And there's some empirical evidence that sometimes, 02:12:56.867 --> 02:12:58.450 align:middle line:84% this works, and sometimes, it doesn't. 02:12:58.450 --> 02:13:00.270 align:middle line:90% OK, great. 02:13:00.270 --> 02:13:02.570 align:middle line:84% And so I think I heard you say this, 02:13:02.570 --> 02:13:07.670 align:middle line:84% but your dynamic hypothesis as to the source of this puzzle 02:13:07.670 --> 02:13:08.950 align:middle line:90% is what? 02:13:08.950 --> 02:13:13.070 align:middle line:84% STUDENT: So what we found so far is 02:13:13.070 --> 02:13:17.630 align:middle line:84% that if the delay of capital growth and resource 02:13:17.630 --> 02:13:21.950 align:middle line:84% regeneration, or maybe just let's focus on capital growth 02:13:21.950 --> 02:13:26.990 align:middle line:84% is shorter, usually managing even absent all 02:13:26.990 --> 02:13:30.390 align:middle line:84% these mechanisms, managing the resources easier. 02:13:30.390 --> 02:13:32.360 align:middle line:84% But these mechanisms actually help 02:13:32.360 --> 02:13:37.680 align:middle line:84% for managing these resources to be improved that are 02:13:37.680 --> 02:13:39.120 align:middle line:90% having better management. 02:13:39.120 --> 02:13:41.480 align:middle line:84% But if the delay is longer, not only 02:13:41.480 --> 02:13:45.520 align:middle line:84% the management of the resource becomes challenging, 02:13:45.520 --> 02:13:49.120 align:middle line:84% more challenging, but if you add these self-governance 02:13:49.120 --> 02:13:52.365 align:middle line:84% mechanisms, they actually make the situation worse. 02:13:52.365 --> 02:13:53.240 align:middle line:90% NELSON REPENNING: OK. 02:13:53.240 --> 02:13:55.615 align:middle line:84% So let me just repeat back to you what I thought I heard, 02:13:55.615 --> 02:13:58.680 align:middle line:84% which is your hypothesis is that these governance mechanisms work 02:13:58.680 --> 02:14:02.740 align:middle line:84% well as long as the delays are short enough, basically, 02:14:02.740 --> 02:14:05.240 align:middle line:84% and that there is some threshold beyond which the delays are 02:14:05.240 --> 02:14:08.280 align:middle line:84% so long, not only does it not help, 02:14:08.280 --> 02:14:10.240 align:middle line:90% but maybe even makes it worse. 02:14:10.240 --> 02:14:12.380 align:middle line:90% OK, so that sounds great. 02:14:12.380 --> 02:14:15.940 align:middle line:84% I mean, and that hasn't been, as far as you know, 02:14:15.940 --> 02:14:20.020 align:middle line:84% studied, or experimented upon in the literature? 02:14:20.020 --> 02:14:23.160 align:middle line:84% STUDENT: People are starting experimenting or exploring 02:14:23.160 --> 02:14:27.700 align:middle line:84% the impact of delay, but different parts of it, 02:14:27.700 --> 02:14:29.140 align:middle line:90% like the whole thing. 02:14:29.140 --> 02:14:32.710 align:middle line:84% Some people will say that especially economists. 02:14:32.710 --> 02:14:35.850 align:middle line:84% So they say, OK, let's do the impact of delay 02:14:35.850 --> 02:14:37.690 align:middle line:90% on the punishment [INAUDIBLE]. 02:14:37.690 --> 02:14:40.770 align:middle line:84% I imposed this sanction mechanism, 02:14:40.770 --> 02:14:44.370 align:middle line:84% but it will be in place with some sort of delay. 02:14:44.370 --> 02:14:48.470 align:middle line:84% NELSON REPENNING: OK, so you may or may not be aware of this. 02:14:48.470 --> 02:14:52.662 align:middle line:84% Not exactly in this context, but the sort 02:14:52.662 --> 02:14:56.750 align:middle line:84% of generation of John's PhD students that were before me, 02:14:56.750 --> 02:14:59.290 align:middle line:84% almost all of them did sort of experimental studies 02:14:59.290 --> 02:15:03.010 align:middle line:84% in different systems that showed basically what you're after, 02:15:03.010 --> 02:15:06.590 align:middle line:84% is that as the delay went up, performance went down. 02:15:06.590 --> 02:15:11.870 align:middle line:84% So Mark did it in the boom-and-bust context, 02:15:11.870 --> 02:15:14.670 align:middle line:84% Christian Kampman was on markets and so on. 02:15:14.670 --> 02:15:18.750 align:middle line:84% So there's a long, pretty old literature, at this point. 02:15:18.750 --> 02:15:22.630 align:middle line:84% OK, so if you haven't, I would just double 02:15:22.630 --> 02:15:24.130 align:middle line:84% check to make sure that no one's run 02:15:24.130 --> 02:15:26.630 align:middle line:84% exactly this experiment before if that's where you're going. 02:15:26.630 --> 02:15:29.540 align:middle line:84% But it sounds like a great problem. 02:15:29.540 --> 02:15:32.020 align:middle line:84% And I'm imagining where we're getting to is 02:15:32.020 --> 02:15:34.580 align:middle line:84% that the hard modeling problem is 02:15:34.580 --> 02:15:38.180 align:middle line:84% to model the self-governance, the kind of behavioral mechanics 02:15:38.180 --> 02:15:39.160 align:middle line:90% of how that works. 02:15:39.160 --> 02:15:39.820 align:middle line:90% OK. 02:15:39.820 --> 02:15:40.320 align:middle line:90% All right. 02:15:40.320 --> 02:15:41.920 align:middle line:84% So why don't you show us what we've got? 02:15:41.920 --> 02:15:44.040 align:middle line:84% STUDENT: So I think that stock management problem, 02:15:44.040 --> 02:15:47.660 align:middle line:84% I really don't have any issues, as you pointed out. 02:15:47.660 --> 02:15:51.380 align:middle line:90% It's a classical stock and flow. 02:15:51.380 --> 02:15:53.920 align:middle line:84% But if you scroll up a little bit, 02:15:53.920 --> 02:16:00.540 align:middle line:84% so this is just for an example of a CPR resource problem. 02:16:00.540 --> 02:16:03.120 align:middle line:84% The shared resources here is water. 02:16:03.120 --> 02:16:06.060 align:middle line:84% And these are, you can imagine, a bunch 02:16:06.060 --> 02:16:09.462 align:middle line:84% of farmers that have some land, and they all rely on that water. 02:16:09.462 --> 02:16:10.420 align:middle line:90% NELSON REPENNING: Yeah. 02:16:10.420 --> 02:16:13.020 align:middle line:84% STUDENT: So as the water becomes scarce, 02:16:13.020 --> 02:16:15.060 align:middle line:84% it is also the case for other resources, 02:16:15.060 --> 02:16:19.340 align:middle line:84% and the costs of obtaining that resource goes higher. 02:16:19.340 --> 02:16:22.080 align:middle line:84% And that actually contributes to lower profits. 02:16:22.080 --> 02:16:26.230 align:middle line:84% So that is how the whole thing is being regulated. 02:16:26.230 --> 02:16:31.070 align:middle line:84% What I haven't added yet is the punishment mechanism so 02:16:31.070 --> 02:16:33.870 align:middle line:90% that when people see-- 02:16:33.870 --> 02:16:35.830 align:middle line:84% first of all, one of the things that I 02:16:35.830 --> 02:16:39.549 align:middle line:84% need feedback is usually these people 02:16:39.549 --> 02:16:42.910 align:middle line:84% need a signal to impose that action. 02:16:42.910 --> 02:16:45.230 align:middle line:84% In the experiments that we did, we 02:16:45.230 --> 02:16:47.809 align:middle line:84% relied on the spark of-- they're capitalists. 02:16:47.809 --> 02:16:49.790 align:middle line:84% For example, again, an example, I 02:16:49.790 --> 02:16:53.090 align:middle line:84% see that Jason has a larger fleet. 02:16:53.090 --> 02:16:57.830 align:middle line:84% So I would relay him that, you're catching the most fishes. 02:16:57.830 --> 02:17:01.389 align:middle line:90% So you are [INAUDIBLE]. 02:17:01.389 --> 02:17:04.070 align:middle line:84% So that is one of the things that I'm 02:17:04.070 --> 02:17:05.846 align:middle line:90% right now sorting at the time. 02:17:05.846 --> 02:17:08.165 align:middle line:90% 02:17:08.165 --> 02:17:09.790 align:middle line:84% NELSON REPENNING: It's a great problem. 02:17:09.790 --> 02:17:11.650 align:middle line:90% So I think this is hard, right? 02:17:11.650 --> 02:17:13.150 align:middle line:84% But I think you're thinking about it 02:17:13.150 --> 02:17:16.590 align:middle line:84% in the right way, which is I think this is a case where 02:17:16.590 --> 02:17:18.309 align:middle line:84% when we don't have a standard kind 02:17:18.309 --> 02:17:22.389 align:middle line:84% of stylized, canonical formulation, 02:17:22.389 --> 02:17:26.780 align:middle line:84% it really pays to think through the micro mechanics. 02:17:26.780 --> 02:17:28.799 align:middle line:84% And so it sounds like what you're-- 02:17:28.799 --> 02:17:31.507 align:middle line:84% and I know this is sort of where you were headed in your problem, 02:17:31.507 --> 02:17:35.240 align:middle line:90% but so what's the cover story? 02:17:35.240 --> 02:17:36.719 align:middle line:90% The water is the-- 02:17:36.719 --> 02:17:37.980 align:middle line:90% STUDENT: Yeah, the water. 02:17:37.980 --> 02:17:40.760 align:middle line:90% Yeah. 02:17:40.760 --> 02:17:48.799 align:middle line:84% So a bunch of farmers, and they rely on irrigating their lands 02:17:48.799 --> 02:17:51.559 align:middle line:90% based on the water. 02:17:51.559 --> 02:17:54.760 align:middle line:84% But then water becomes a sparse, so their profits 02:17:54.760 --> 02:17:56.200 align:middle line:90% start to decline. 02:17:56.200 --> 02:18:00.180 align:middle line:84% But then they would look to who has the most land. 02:18:00.180 --> 02:18:03.040 align:middle line:84% [INAUDIBLE] So that's the person who's [INAUDIBLE]. 02:18:03.040 --> 02:18:05.400 align:middle line:84% NELSON REPENNING: OK, so how do they know that water 02:18:05.400 --> 02:18:07.879 align:middle line:90% is becoming more scarce? 02:18:07.879 --> 02:18:10.740 align:middle line:84% The price goes up from the utility or just less comes out-- 02:18:10.740 --> 02:18:11.640 align:middle line:90% STUDENT: [INAUDIBLE] 02:18:11.640 --> 02:18:12.515 align:middle line:90% NELSON REPENNING: OK. 02:18:12.515 --> 02:18:16.520 align:middle line:84% So I think the first thing is to model the kind of perceptual-- 02:18:16.520 --> 02:18:18.320 align:middle line:90% you have a flow basically. 02:18:18.320 --> 02:18:19.559 align:middle line:90% It's water. 02:18:19.559 --> 02:18:21.799 align:middle line:90% And it's going down. 02:18:21.799 --> 02:18:25.987 align:middle line:84% And then there's some perceptual dynamic. 02:18:25.987 --> 02:18:27.570 align:middle line:84% It's got to get to a certain threshold 02:18:27.570 --> 02:18:28.629 align:middle line:90% before I can notice it. 02:18:28.629 --> 02:18:30.530 align:middle line:84% So is it the actual flow of water that I see, 02:18:30.530 --> 02:18:33.250 align:middle line:90% or the pricing from the utility? 02:18:33.250 --> 02:18:38.330 align:middle line:84% STUDENT: I think you actually see the stock of water. 02:18:38.330 --> 02:18:41.028 align:middle line:84% And you also see that it also [INAUDIBLE]. 02:18:41.028 --> 02:18:42.570 align:middle line:84% NELSON REPENNING: OK, so you're going 02:18:42.570 --> 02:18:47.330 align:middle line:84% to need some simple underlying model of essentially 02:18:47.330 --> 02:18:49.610 align:middle line:84% the aquifer, which could just be a stock 02:18:49.610 --> 02:18:51.090 align:middle line:90% and people taking out of it. 02:18:51.090 --> 02:18:53.850 align:middle line:84% OK, so then I notice that there's a problem. 02:18:53.850 --> 02:18:56.129 align:middle line:84% And then the next step is now, there 02:18:56.129 --> 02:18:59.174 align:middle line:84% has to be some attribution of blame. 02:18:59.174 --> 02:18:59.799 align:middle line:90% STUDENT: Right. 02:18:59.799 --> 02:19:00.799 align:middle line:90% NELSON REPENNING: Right? 02:19:00.799 --> 02:19:03.129 align:middle line:84% And so what's-- so how do you think that works? 02:19:03.129 --> 02:19:04.049 align:middle line:90% This is a great-- 02:19:04.049 --> 02:19:05.070 align:middle line:90% STUDENT: [LAUGHTER] 02:19:05.070 --> 02:19:06.362 align:middle line:90% NELSON REPENNING: This is hard. 02:19:06.362 --> 02:19:07.290 align:middle line:90% I mean, this is not-- 02:19:07.290 --> 02:19:08.665 align:middle line:84% maybe someone's figured this out. 02:19:08.665 --> 02:19:14.490 align:middle line:84% But I don't know of a solid formulation for this. 02:19:14.490 --> 02:19:17.010 align:middle line:84% STUDENT: So can I also share this? 02:19:17.010 --> 02:19:19.980 align:middle line:84% In the experiments, we provided all this information 02:19:19.980 --> 02:19:23.940 align:middle line:84% and I never thought that how if I 02:19:23.940 --> 02:19:29.139 align:middle line:84% wanted to do a feedback-rich model, how should I do this? 02:19:29.139 --> 02:19:32.440 align:middle line:84% That part is actually a little bit unclear. 02:19:32.440 --> 02:19:35.420 align:middle line:84% But what I'm hypothesizing is that probably, people 02:19:35.420 --> 02:19:40.219 align:middle line:84% would see the level of water like aquifer compared 02:19:40.219 --> 02:19:47.059 align:middle line:84% to the thresholds is that those thresholds would start letting 02:19:47.059 --> 02:19:50.600 align:middle line:84% others see that you have to stay in the rules, 02:19:50.600 --> 02:19:54.290 align:middle line:84% you don't have to grow your orchards or lands. 02:19:54.290 --> 02:19:57.180 align:middle line:90% [INAUDIBLE] 02:19:57.180 --> 02:19:58.660 align:middle line:84% NELSON REPENNING: Yeah, so I think 02:19:58.660 --> 02:20:00.220 align:middle line:84% what you're going to have to sort out 02:20:00.220 --> 02:20:02.740 align:middle line:84% is there's going to be different cases that 02:20:02.740 --> 02:20:07.360 align:middle line:84% are going to sort of turn on the observability of the usage. 02:20:07.360 --> 02:20:12.220 align:middle line:84% So I mean, so you've, I think, advanced a couple of times 02:20:12.220 --> 02:20:13.860 align:middle line:84% a sort of hypothesis that I'm just 02:20:13.860 --> 02:20:15.540 align:middle line:84% going to yell at the person with the biggest orchard, 02:20:15.540 --> 02:20:17.165 align:middle line:84% because they're the one that's probably 02:20:17.165 --> 02:20:18.910 align:middle line:90% taking-- so that's one rule. 02:20:18.910 --> 02:20:23.870 align:middle line:84% We think about the fish, I don't know. 02:20:23.870 --> 02:20:25.670 align:middle line:84% In these fisheries, are there records of-- 02:20:25.670 --> 02:20:27.530 align:middle line:84% I mean, do I have to log in the number of fish? 02:20:27.530 --> 02:20:29.405 align:middle line:84% Or do I have the number of fleet or whatever? 02:20:29.405 --> 02:20:32.642 align:middle line:84% So these settings are going to differ in terms of kind 02:20:32.642 --> 02:20:34.170 align:middle line:90% of how observable it is. 02:20:34.170 --> 02:20:35.670 align:middle line:84% Like, if I just look at your orchard 02:20:35.670 --> 02:20:38.550 align:middle line:84% and it looks greener than mine, maybe that I-- so you 02:20:38.550 --> 02:20:42.150 align:middle line:84% could have an underlying model of how much each person is 02:20:42.150 --> 02:20:42.650 align:middle line:90% taking. 02:20:42.650 --> 02:20:44.970 align:middle line:84% I mean, you might, again, to keep it simple, 02:20:44.970 --> 02:20:47.270 align:middle line:90% just start with two. 02:20:47.270 --> 02:20:49.572 align:middle line:84% I could look at the productivity, 02:20:49.572 --> 02:20:51.030 align:middle line:84% but I think the interesting part is 02:20:51.030 --> 02:20:54.290 align:middle line:84% there's both delays and uncertainty in that. 02:20:54.290 --> 02:21:00.070 align:middle line:84% And then so you'll probably have a stock 02:21:00.070 --> 02:21:03.830 align:middle line:84% that's going to be kind of my mental estimate 02:21:03.830 --> 02:21:05.510 align:middle line:84% of how much water you're using, which 02:21:05.510 --> 02:21:07.570 align:middle line:84% is going to be updated through these processes. 02:21:07.570 --> 02:21:10.410 align:middle line:84% And then if that exceeds my expectations, 02:21:10.410 --> 02:21:12.530 align:middle line:90% then there's going to be some-- 02:21:12.530 --> 02:21:14.030 align:middle line:84% I read Hasier's comments, and he was 02:21:14.030 --> 02:21:16.600 align:middle line:84% talking about how this is a sort of game theoretic thing. 02:21:16.600 --> 02:21:19.610 align:middle line:84% So I mean, you could solve for some kind of equilibrium, 02:21:19.610 --> 02:21:23.012 align:middle line:84% but it's the inputs that go into it that are interesting. 02:21:23.012 --> 02:21:24.470 align:middle line:84% But I guess I would urge you, and I 02:21:24.470 --> 02:21:27.310 align:middle line:84% know we're running low on time, is, 02:21:27.310 --> 02:21:30.590 align:middle line:84% this is a case where I think having a more specific setting 02:21:30.590 --> 02:21:35.550 align:middle line:84% you're modeling, fish, water, I don't care, 02:21:35.550 --> 02:21:37.310 align:middle line:84% will really be helpful because I think 02:21:37.310 --> 02:21:39.710 align:middle line:84% that will help you kind of sort the actual kind 02:21:39.710 --> 02:21:45.410 align:middle line:84% of micromechanics of I see that I'm running out of water, 02:21:45.410 --> 02:21:47.050 align:middle line:84% I see that your field is greener, 02:21:47.050 --> 02:21:50.030 align:middle line:84% or you're getting more apples out of the orchard or whatever, 02:21:50.030 --> 02:21:52.510 align:middle line:84% we have our monthly meeting, I come in 02:21:52.510 --> 02:21:56.650 align:middle line:84% and I negotiate, and maybe you have some simple Nash bargaining 02:21:56.650 --> 02:21:58.150 align:middle line:84% solution or something like that that 02:21:58.150 --> 02:22:01.870 align:middle line:84% would be a real simple model of those negotiations. 02:22:01.870 --> 02:22:04.310 align:middle line:84% And then you could kind of see, there'd 02:22:04.310 --> 02:22:05.890 align:middle line:90% be some feedback adjustment. 02:22:05.890 --> 02:22:10.150 align:middle line:84% And then I think, kind of having had hypotheses and all 02:22:10.150 --> 02:22:12.630 align:middle line:84% those delays, you could see whether essentially, 02:22:12.630 --> 02:22:19.440 align:middle line:84% that response is fast enough to, I could imagine either it 02:22:19.440 --> 02:22:25.640 align:middle line:84% could die, or you could get some kind of cyclicality to that. 02:22:25.640 --> 02:22:27.000 align:middle line:90% Yeah, that's good. 02:22:27.000 --> 02:22:31.875 align:middle line:84% I don't know if this is helpful, but I think it's a good problem. 02:22:31.875 --> 02:22:32.500 align:middle line:90% And it's funny. 02:22:32.500 --> 02:22:34.200 align:middle line:84% John has a paper too on, I think it 02:22:34.200 --> 02:22:36.885 align:middle line:84% was called Behavioral Game Theory or something like that, 02:22:36.885 --> 02:22:37.760 align:middle line:90% which you've seen it? 02:22:37.760 --> 02:22:38.060 align:middle line:90% STUDENT: Yeah. 02:22:38.060 --> 02:22:39.143 align:middle line:90% NELSON REPENNING: So yeah. 02:22:39.143 --> 02:22:41.140 align:middle line:84% OK, so I know he's worked on this a little bit. 02:22:41.140 --> 02:22:43.520 align:middle line:84% But this is an interesting kind of convergence 02:22:43.520 --> 02:22:47.400 align:middle line:84% of sort of an SD classic with some econ that I'm not-- 02:22:47.400 --> 02:22:50.097 align:middle line:84% I don't always know the literature, every paper. 02:22:50.097 --> 02:22:52.180 align:middle line:84% But I'm not sure I've seen those two put together. 02:22:52.180 --> 02:22:53.060 align:middle line:90% So it's kind of cool. 02:22:53.060 --> 02:22:55.720 align:middle line:90% So-- 02:22:55.720 --> 02:22:58.700 align:middle line:84% STUDENT: The nice thing about it was through the experience, 02:22:58.700 --> 02:23:02.500 align:middle line:84% he actually could replicate findings in the econ literature. 02:23:02.500 --> 02:23:04.482 align:middle line:90% So it's [INAUDIBLE] short. 02:23:04.482 --> 02:23:06.940 align:middle line:84% Actually, the situation is going to be better for everyone. 02:23:06.940 --> 02:23:07.898 align:middle line:90% NELSON REPENNING: Yeah. 02:23:07.898 --> 02:23:12.530 align:middle line:84% STUDENT: But the other end of the spectrum, delay is longer. 02:23:12.530 --> 02:23:13.857 align:middle line:90% No one has looked at it. 02:23:13.857 --> 02:23:15.690 align:middle line:84% And we actually showed that the situation is 02:23:15.690 --> 02:23:17.370 align:middle line:90% going to get worse and worse. 02:23:17.370 --> 02:23:18.870 align:middle line:84% NELSON REPENNING: In the experiment. 02:23:18.870 --> 02:23:21.850 align:middle line:84% OK, so now, you're kind of building a theory of why 02:23:21.850 --> 02:23:24.490 align:middle line:90% that happens in some sense. 02:23:24.490 --> 02:23:25.350 align:middle line:90% Yeah, OK. 02:23:25.350 --> 02:23:25.850 align:middle line:90% Awesome. 02:23:25.850 --> 02:23:26.510 align:middle line:90% That's great. 02:23:26.510 --> 02:23:26.790 align:middle line:90% All right. 02:23:26.790 --> 02:23:27.290 align:middle line:90% Who's left? 02:23:27.290 --> 02:23:28.230 align:middle line:90% We we have two more? 02:23:28.230 --> 02:23:28.730 align:middle line:90% Yep. 02:23:28.730 --> 02:23:30.210 align:middle line:90% OK. 02:23:30.210 --> 02:23:31.110 align:middle line:90% STUDENT: So I could. 02:23:31.110 --> 02:23:31.610 align:middle line:90% Yeah. 02:23:31.610 --> 02:23:32.727 align:middle line:90% NELSON REPENNING: OK. 02:23:32.727 --> 02:23:34.310 align:middle line:84% STUDENT: I have [INAUDIBLE] something, 02:23:34.310 --> 02:23:35.550 align:middle line:90% but I'm not very happy with it. 02:23:35.550 --> 02:23:36.425 align:middle line:90% NELSON REPENNING: OK. 02:23:36.425 --> 02:23:39.450 align:middle line:84% STUDENT: So I'd rather maybe just get sort of a feedback 02:23:39.450 --> 02:23:40.370 align:middle line:90% on how it should go. 02:23:40.370 --> 02:23:41.370 align:middle line:90% NELSON REPENNING: OK. 02:23:41.370 --> 02:23:43.690 align:middle line:84% STUDENT: My research is on the lung cancer screening. 02:23:43.690 --> 02:23:46.806 align:middle line:90% NELSON REPENNING: OK. 02:23:46.806 --> 02:23:49.690 align:middle line:84% STUDENT: The problem there is we're getting better 02:23:49.690 --> 02:23:52.950 align:middle line:84% at identifying the population we need to target this group. 02:23:52.950 --> 02:23:55.170 align:middle line:90% [INAUDIBLE] not sure. 02:23:55.170 --> 02:23:57.210 align:middle line:84% We're getting somewhere around 15% 02:23:57.210 --> 02:23:59.310 align:middle line:90% of the [INAUDIBLE] screening. 02:23:59.310 --> 02:24:01.490 align:middle line:84% There's always this hypothesis that, especially 02:24:01.490 --> 02:24:03.510 align:middle line:84% for lung cancer, these are people who smoke, 02:24:03.510 --> 02:24:05.450 align:middle line:90% who [INAUDIBLE]. 02:24:05.450 --> 02:24:09.083 align:middle line:84% So they may not be very caring about their health care. 02:24:09.083 --> 02:24:10.250 align:middle line:90% That's why they're not sure. 02:24:10.250 --> 02:24:13.260 align:middle line:84% The very same studies show that the same people who 02:24:13.260 --> 02:24:16.860 align:middle line:84% are eligible for lung cancer screening, 15% of them 02:24:16.860 --> 02:24:18.700 align:middle line:84% show up for lung cancer screening. 02:24:18.700 --> 02:24:20.240 align:middle line:84% If they're eligible for, let's say, 02:24:20.240 --> 02:24:23.700 align:middle line:84% breast cancer screening, colorectal, 65%. 02:24:23.700 --> 02:24:25.420 align:middle line:90% STUDENT: Show up to this-- 02:24:25.420 --> 02:24:29.568 align:middle line:84% There's a 50% gap between showing up for lung cancer 02:24:29.568 --> 02:24:30.860 align:middle line:90% screening and with the cancer-- 02:24:30.860 --> 02:24:32.420 align:middle line:90% the same group of people. 02:24:32.420 --> 02:24:34.620 align:middle line:84% So it's not an attribute of the people. 02:24:34.620 --> 02:24:36.960 align:middle line:84% The test, it's just a simple [? step. ?] 02:24:36.960 --> 02:24:38.380 align:middle line:90% It's not very expensive. 02:24:38.380 --> 02:24:40.740 align:middle line:84% Insurance covers it because they're eligible. 02:24:40.740 --> 02:24:44.140 align:middle line:84% It's not intrusive, less intrusive than the other tests. 02:24:44.140 --> 02:24:46.140 align:middle line:84% So there's this puzzle, why they're not 02:24:46.140 --> 02:24:47.740 align:middle line:90% showing up to the screen. 02:24:47.740 --> 02:24:49.620 align:middle line:90% [INAUDIBLE] 02:24:49.620 --> 02:24:51.540 align:middle line:84% I have this hypothesis that there's 02:24:51.540 --> 02:24:55.880 align:middle line:84% sort of a cost-benefit analysis for people. 02:24:55.880 --> 02:25:05.380 align:middle line:84% And there's lots of false positives in the first layer 02:25:05.380 --> 02:25:06.780 align:middle line:90% of screening. 02:25:06.780 --> 02:25:09.190 align:middle line:84% And if there's something seen there called 02:25:09.190 --> 02:25:12.190 align:middle line:84% for for a second follow up among those 02:25:12.190 --> 02:25:16.590 align:middle line:84% who get any positive tests in the first screening, around 40% 02:25:16.590 --> 02:25:19.775 align:middle line:84% of them show up for follow up, which is still very low. 02:25:19.775 --> 02:25:22.150 align:middle line:84% NELSON REPENNING: So even if you have a positive on round 02:25:22.150 --> 02:25:26.070 align:middle line:90% one, 60%-- 02:25:26.070 --> 02:25:27.590 align:middle line:90% STUDENT: Absolutely positive. 02:25:27.590 --> 02:25:29.110 align:middle line:90% So there are different layers. 02:25:29.110 --> 02:25:32.870 align:middle line:84% So if it's something very bad in there, so they go to treatment. 02:25:32.870 --> 02:25:36.790 align:middle line:90% But if something [INAUDIBLE]. 02:25:36.790 --> 02:25:39.190 align:middle line:84% It's usually-- the decision is, you're OK. 02:25:39.190 --> 02:25:40.350 align:middle line:90% There's something there. 02:25:40.350 --> 02:25:43.210 align:middle line:84% Come back in six months, and we'll see again. 02:25:43.210 --> 02:25:45.890 align:middle line:84% And then that come back doesn't happen all the time. 02:25:45.890 --> 02:25:48.990 align:middle line:90% So you only have 40%, 50%. 02:25:48.990 --> 02:25:51.435 align:middle line:84% So there is something there, and I'm trying to-- 02:25:51.435 --> 02:25:53.310 align:middle line:84% NELSON REPENNING: OK, so it's a great puzzle. 02:25:53.310 --> 02:25:56.550 align:middle line:84% Why do you think a dynamic model is the way 02:25:56.550 --> 02:25:57.950 align:middle line:90% to answer that puzzle? 02:25:57.950 --> 02:26:00.110 align:middle line:84% STUDENT: I have a sense that there's 02:26:00.110 --> 02:26:05.030 align:middle line:84% this dread of the worst-case scenario and a blame game 02:26:05.030 --> 02:26:07.680 align:middle line:84% that gets this reinforcing dynamics, which 02:26:07.680 --> 02:26:11.340 align:middle line:84% is if I'm a smoker and I've done this to myself, 02:26:11.340 --> 02:26:12.480 align:middle line:90% I don't want to know. 02:26:12.480 --> 02:26:15.560 align:middle line:84% There's somewhat of a willful ignorance. 02:26:15.560 --> 02:26:19.680 align:middle line:84% And as time goes on, my situation is getting worse. 02:26:19.680 --> 02:26:22.700 align:middle line:90% So I don't want to know more. 02:26:22.700 --> 02:26:25.760 align:middle line:84% So I feel like there's some dynamics that-- 02:26:25.760 --> 02:26:26.460 align:middle line:90% I'm not sure. 02:26:26.460 --> 02:26:28.120 align:middle line:90% I don't think anyone that-- 02:26:28.120 --> 02:26:29.540 align:middle line:90% has any idea of why. 02:26:29.540 --> 02:26:32.578 align:middle line:90% 02:26:32.578 --> 02:26:34.120 align:middle line:84% NELSON REPENNING: It's a fascinating, 02:26:34.120 --> 02:26:36.220 align:middle line:84% and super interesting, and important problem. 02:26:36.220 --> 02:26:38.960 align:middle line:84% It doesn't immediately scream to me 02:26:38.960 --> 02:26:43.380 align:middle line:84% that a kind of feedback-rich model is the way to go after it. 02:26:43.380 --> 02:26:46.600 align:middle line:84% I mean, so I think there's another story you could tell 02:26:46.600 --> 02:26:52.440 align:middle line:84% that there's some kind of biased risk calculus 02:26:52.440 --> 02:26:57.080 align:middle line:84% that's going on in people's head of, or some dread, 02:26:57.080 --> 02:26:58.400 align:middle line:90% or something like that. 02:26:58.400 --> 02:27:00.800 align:middle line:90% That's a very different class. 02:27:00.800 --> 02:27:03.400 align:middle line:84% So where I thought you were going 02:27:03.400 --> 02:27:07.410 align:middle line:84% is that the treatment had developed 02:27:07.410 --> 02:27:10.990 align:middle line:84% a bad reputation, basically, because of the false positives. 02:27:10.990 --> 02:27:13.910 align:middle line:84% And so that was driving people away, or something like that. 02:27:13.910 --> 02:27:16.370 align:middle line:84% STUDENT: There is dynamics there for sure too, that because 02:27:16.370 --> 02:27:19.410 align:middle line:84% of false positives, people don't think that there's 02:27:19.410 --> 02:27:21.810 align:middle line:90% much benefit to this screening. 02:27:21.810 --> 02:27:24.910 align:middle line:84% As a slower dynamics in testing, I'm not sure. 02:27:24.910 --> 02:27:26.850 align:middle line:90% But there is some data. 02:27:26.850 --> 02:27:28.988 align:middle line:84% I still don't know what to go after. 02:27:28.988 --> 02:27:30.530 align:middle line:84% So what I've done so far, and I think 02:27:30.530 --> 02:27:33.370 align:middle line:84% I got my first round of the [INAUDIBLE] 02:27:33.370 --> 02:27:37.950 align:middle line:84% was to focus on building the structure, 02:27:37.950 --> 02:27:39.890 align:middle line:90% this actual structure. 02:27:39.890 --> 02:27:42.890 align:middle line:84% And I did go through that, lots of stocks, 02:27:42.890 --> 02:27:45.390 align:middle line:84% and lots of ideas for subscription. 02:27:45.390 --> 02:27:47.870 align:middle line:84% But then I felt like whatever I'm doing there, 02:27:47.870 --> 02:27:52.330 align:middle line:84% I already have as math models coming from ACI. 02:27:52.330 --> 02:27:57.530 align:middle line:84% All the parameters are calibrated by ACI [INAUDIBLE]. 02:27:57.530 --> 02:27:58.690 align:middle line:90% I can recreate. 02:27:58.690 --> 02:28:00.490 align:middle line:90% I think I managed to do that. 02:28:00.490 --> 02:28:04.620 align:middle line:84% But I don't think it got me any closer to the dynamic 02:28:04.620 --> 02:28:06.220 align:middle line:90% that I would want to. 02:28:06.220 --> 02:28:07.970 align:middle line:84% NELSON REPENNING: Yeah, so it sounds to me 02:28:07.970 --> 02:28:10.220 align:middle line:84% like what you're missing here, and this 02:28:10.220 --> 02:28:13.080 align:middle line:84% is kind of a fundamentally creative act, 02:28:13.080 --> 02:28:16.743 align:middle line:84% is the sort of core dynamic hypothesis. 02:28:16.743 --> 02:28:18.660 align:middle line:84% And I'm certainly willing to accept that there 02:28:18.660 --> 02:28:19.880 align:middle line:90% might be some dynamics there. 02:28:19.880 --> 02:28:22.040 align:middle line:90% But there may not be, right? 02:28:22.040 --> 02:28:26.000 align:middle line:84% I mean, it may just be a kind set of perceptual biases. 02:28:26.000 --> 02:28:28.140 align:middle line:90% I mean, it's a fascinating-- 02:28:28.140 --> 02:28:31.360 align:middle line:84% Why people would sign up for this less than a colonoscopy, 02:28:31.360 --> 02:28:39.060 align:middle line:84% which is invasive and enormously disruptive, 02:28:39.060 --> 02:28:43.120 align:middle line:84% is really an interesting paradox that makes no sense. 02:28:43.120 --> 02:28:46.512 align:middle line:84% But I mean, the only place where I 02:28:46.512 --> 02:28:48.220 align:middle line:84% can see that system dynamics would really 02:28:48.220 --> 02:28:51.580 align:middle line:84% bear on this, if there is a kind of social learning 02:28:51.580 --> 02:28:56.340 align:middle line:84% set of dynamics that goes along, that somehow, by reading 02:28:56.340 --> 02:28:58.200 align:middle line:84% the paper, and talking to their friends, 02:28:58.200 --> 02:29:02.490 align:middle line:84% people have learned that the colonoscopies are helpful, 02:29:02.490 --> 02:29:04.110 align:middle line:84% and they catch this early, and you 02:29:04.110 --> 02:29:06.110 align:middle line:84% can live a long, happy life, where somehow, they 02:29:06.110 --> 02:29:09.750 align:middle line:84% have informed the same perception of the lung cancer 02:29:09.750 --> 02:29:10.270 align:middle line:90% screening. 02:29:10.270 --> 02:29:12.830 align:middle line:84% So it's a case where it seems like the learning mechanism has 02:29:12.830 --> 02:29:16.050 align:middle line:84% really gone sideways in some weird-- 02:29:16.050 --> 02:29:17.150 align:middle line:90% STUDENT: Like [INAUDIBLE]. 02:29:17.150 --> 02:29:18.483 align:middle line:90% NELSON REPENNING: Yeah, exactly. 02:29:18.483 --> 02:29:22.230 align:middle line:84% So I actually think this is a case where you might 02:29:22.230 --> 02:29:24.550 align:middle line:90% want to read the literature. 02:29:24.550 --> 02:29:27.390 align:middle line:84% If there's first-person accounts of, 02:29:27.390 --> 02:29:30.150 align:middle line:84% have they interviewed people that don't show up or whatever? 02:29:30.150 --> 02:29:32.690 align:middle line:84% And start scribbling causal loop diagrams, 02:29:32.690 --> 02:29:34.510 align:middle line:84% actually, I think, because I'm not 02:29:34.510 --> 02:29:37.210 align:middle line:84% sure you have anything to model formally yet. 02:29:37.210 --> 02:29:39.442 align:middle line:84% I mean, you could model kind of how people 02:29:39.442 --> 02:29:40.650 align:middle line:90% move through the aging chain. 02:29:40.650 --> 02:29:42.550 align:middle line:84% But like you said, that's been done, 02:29:42.550 --> 02:29:44.530 align:middle line:84% and that alone hasn't solved the problem. 02:29:44.530 --> 02:29:48.030 align:middle line:84% I think you need a hypothesis basically. 02:29:48.030 --> 02:29:50.910 align:middle line:90% But it's a great-- 02:29:50.910 --> 02:29:52.470 align:middle line:90% I don't know. 02:29:52.470 --> 02:29:54.798 align:middle line:84% I have trouble wrapping my head around that one. 02:29:54.798 --> 02:29:55.590 align:middle line:90% STUDENT: Thank you. 02:29:55.590 --> 02:29:56.643 align:middle line:90% NELSON REPENNING: OK. 02:29:56.643 --> 02:29:57.310 align:middle line:90% Do you want to-- 02:29:57.310 --> 02:30:02.400 align:middle line:84% STUDENT: Yeah, just discuss it a little bit because I haven't 02:30:02.400 --> 02:30:06.480 align:middle line:84% done much modeling around this, but I'm 02:30:06.480 --> 02:30:10.600 align:middle line:84% interviewing businesses in a ecosystem, 02:30:10.600 --> 02:30:17.720 align:middle line:84% and trying to find the circular economy KPIs, so Key Performance 02:30:17.720 --> 02:30:20.400 align:middle line:84% Indicators, for where they actually 02:30:20.400 --> 02:30:22.560 align:middle line:84% are [INAUDIBLE] where you can see 02:30:22.560 --> 02:30:27.080 align:middle line:90% the advantages of collaboration. 02:30:27.080 --> 02:30:33.720 align:middle line:84% So if they are sharing resources, or it's a gov-- 02:30:33.720 --> 02:30:37.760 align:middle line:84% so to try to have a governance system for the whole ecosystem 02:30:37.760 --> 02:30:40.820 align:middle line:90% system to show the benefits. 02:30:40.820 --> 02:30:44.700 align:middle line:84% And I have done some interviews, four or five interviews, 02:30:44.700 --> 02:30:48.960 align:middle line:84% and I see that there are a lot of shared needs 02:30:48.960 --> 02:30:52.360 align:middle line:84% that they are satisfying within that ecosystem. 02:30:52.360 --> 02:30:58.920 align:middle line:84% And they all share those needs, and this is not really 02:30:58.920 --> 02:31:02.000 align:middle line:84% captured by the ecosystem itself because it's 02:31:02.000 --> 02:31:05.240 align:middle line:84% more like a legitimacy or reputation. 02:31:05.240 --> 02:31:08.360 align:middle line:84% It's a motivation of the employees. 02:31:08.360 --> 02:31:10.540 align:middle line:84% It's very important for the ecosystem, 02:31:10.540 --> 02:31:13.320 align:middle line:84% but it's not being captured in any performance. 02:31:13.320 --> 02:31:16.320 align:middle line:84% NELSON REPENNING: So this feels a little bit 02:31:16.320 --> 02:31:19.040 align:middle line:84% like a solution chasing a problem. 02:31:19.040 --> 02:31:24.560 align:middle line:84% I mean, so if you're able to solve this problem, 02:31:24.560 --> 02:31:25.580 align:middle line:90% what will get better? 02:31:25.580 --> 02:31:28.040 align:middle line:84% What's the thing that you care about at the end of the day? 02:31:28.040 --> 02:31:30.720 align:middle line:84% STUDENT: Yeah, it's really helpful for the ecosystem 02:31:30.720 --> 02:31:32.180 align:middle line:90% to organize itself. 02:31:32.180 --> 02:31:35.240 align:middle line:90% 02:31:35.240 --> 02:31:39.160 align:middle line:84% I mean, to capture the benefits of what they're doing. 02:31:39.160 --> 02:31:42.060 align:middle line:84% So today, they are, for example, they are doing some stuff, 02:31:42.060 --> 02:31:44.700 align:middle line:84% but they don't really know the benefits from it. 02:31:44.700 --> 02:31:46.700 align:middle line:84% But they believe they're doing the right things. 02:31:46.700 --> 02:31:49.075 align:middle line:84% NELSON REPENNING: Because I'm missing a piece of context. 02:31:49.075 --> 02:31:50.540 align:middle line:90% I mean, doing what? 02:31:50.540 --> 02:31:51.385 align:middle line:90% Trying to-- 02:31:51.385 --> 02:31:52.760 align:middle line:84% STUDENT: So the ecosystem is kind 02:31:52.760 --> 02:31:57.530 align:middle line:84% of helping-- there is one party within that system that 02:31:57.530 --> 02:32:01.690 align:middle line:90% is supporting them. 02:32:01.690 --> 02:32:08.010 align:middle line:84% So they do events, they invite people to come over, and visit. 02:32:08.010 --> 02:32:13.630 align:middle line:84% And there's-- so they organize events. 02:32:13.630 --> 02:32:17.090 align:middle line:84% They also organize the travel and visit 02:32:17.090 --> 02:32:20.370 align:middle line:90% other potential customers. 02:32:20.370 --> 02:32:24.290 align:middle line:84% And they also advise on, or giving market advice, 02:32:24.290 --> 02:32:25.730 align:middle line:90% negotiating advice. 02:32:25.730 --> 02:32:30.210 align:middle line:84% So there's a lot of support for that system. 02:32:30.210 --> 02:32:31.633 align:middle line:84% NELSON REPENNING: But to what end? 02:32:31.633 --> 02:32:33.050 align:middle line:84% Another way of saying this, if you 02:32:33.050 --> 02:32:35.383 align:middle line:84% build a model, what's the performance metric that you're 02:32:35.383 --> 02:32:36.110 align:middle line:90% going to look at? 02:32:36.110 --> 02:32:39.450 align:middle line:84% STUDENT: I'm trying to capture the KPIs that they could kind 02:32:39.450 --> 02:32:41.770 align:middle line:90% of measure their performance. 02:32:41.770 --> 02:32:44.610 align:middle line:90% So by looking at that-- 02:32:44.610 --> 02:32:47.690 align:middle line:90% 02:32:47.690 --> 02:32:49.030 align:middle line:90% I'm not explaining this-- 02:32:49.030 --> 02:32:49.850 align:middle line:84% NELSON REPENNING: Well, I know we're rushing 02:32:49.850 --> 02:32:50.950 align:middle line:90% you, which is not very helpful. 02:32:50.950 --> 02:32:53.040 align:middle line:84% But I mean, and I'm happy to talk to you about it. 02:32:53.040 --> 02:32:55.700 align:middle line:84% STUDENT: It's like you are a party, 02:32:55.700 --> 02:32:57.560 align:middle line:90% you are developing a product. 02:32:57.560 --> 02:33:00.020 align:middle line:84% And within the ecosystem, you find somebody 02:33:00.020 --> 02:33:02.900 align:middle line:84% within the ecosystem who actually is using that product. 02:33:02.900 --> 02:33:05.940 align:middle line:84% So you can work together and you develop the product together. 02:33:05.940 --> 02:33:08.700 align:middle line:84% So you have the benefits of having access 02:33:08.700 --> 02:33:12.940 align:middle line:84% to a product development partner, which 02:33:12.940 --> 02:33:17.908 align:middle line:84% is also benefiting from you because you provide the service. 02:33:17.908 --> 02:33:19.700 align:middle line:84% And then they develop the product together, 02:33:19.700 --> 02:33:21.860 align:middle line:84% and they get the benefits because they 02:33:21.860 --> 02:33:23.205 align:middle line:90% are both in the ecosystem. 02:33:23.205 --> 02:33:24.580 align:middle line:84% NELSON REPENNING: OK, so you have 02:33:24.580 --> 02:33:27.660 align:middle line:84% a hypothesis that if people collaborate more, 02:33:27.660 --> 02:33:28.820 align:middle line:90% something will get better. 02:33:28.820 --> 02:33:29.700 align:middle line:90% STUDENT: Yes. 02:33:29.700 --> 02:33:32.060 align:middle line:84% And it's in the concept of circular economy 02:33:32.060 --> 02:33:33.700 align:middle line:84% because circular economy, you need 02:33:33.700 --> 02:33:37.140 align:middle line:84% to collaborate more in order to, I mean [INAUDIBLE]. 02:33:37.140 --> 02:33:38.140 align:middle line:90% NELSON REPENNING: Right. 02:33:38.140 --> 02:33:40.497 align:middle line:84% But again, if your hypothesis is true, 02:33:40.497 --> 02:33:42.080 align:middle line:84% what's the thing that will get better? 02:33:42.080 --> 02:33:44.740 align:middle line:84% Is it GDP is going to go up, or environmental impact is 02:33:44.740 --> 02:33:45.720 align:middle line:90% going to go down? 02:33:45.720 --> 02:33:48.620 align:middle line:84% Or what's the final performance metric? 02:33:48.620 --> 02:33:51.390 align:middle line:84% Or is that the thing you're trying to figure out? 02:33:51.390 --> 02:33:53.230 align:middle line:90% STUDENT: Yes. 02:33:53.230 --> 02:33:55.830 align:middle line:90% I mean, is it happening? 02:33:55.830 --> 02:33:57.310 align:middle line:84% So that was kind of, is it really 02:33:57.310 --> 02:33:59.270 align:middle line:90% happening within that system? 02:33:59.270 --> 02:34:01.790 align:middle line:84% Which I have found out in the interview. 02:34:01.790 --> 02:34:03.670 align:middle line:90% But it's actually happening. 02:34:03.670 --> 02:34:09.070 align:middle line:84% And then also, what is going up or down, and how-- 02:34:09.070 --> 02:34:09.790 align:middle line:90% so-- 02:34:09.790 --> 02:34:11.748 align:middle line:84% NELSON REPENNING: Yeah, so I don't think that's 02:34:11.748 --> 02:34:13.570 align:middle line:84% a system dynamics problem per se. 02:34:13.570 --> 02:34:15.570 align:middle line:84% I mean, there might be some modeling to be done. 02:34:15.570 --> 02:34:20.090 align:middle line:84% I mean, I think that's much more of a kind of, 02:34:20.090 --> 02:34:23.098 align:middle line:84% not to quote my own papers, but you 02:34:23.098 --> 02:34:25.390 align:middle line:84% have to kind of figure out what problem you want to try 02:34:25.390 --> 02:34:27.150 align:middle line:90% to solve in this context. 02:34:27.150 --> 02:34:29.715 align:middle line:84% And again, I'm totally fine with reverse engineering 02:34:29.715 --> 02:34:31.090 align:middle line:84% in the following sense, which is, 02:34:31.090 --> 02:34:33.950 align:middle line:84% again, I think you have a hypothesis that collaboration is 02:34:33.950 --> 02:34:35.630 align:middle line:90% good and more of it would help. 02:34:35.630 --> 02:34:37.670 align:middle line:84% But you have to sort of reverse engineer 02:34:37.670 --> 02:34:40.770 align:middle line:84% to what end would that be directed? 02:34:40.770 --> 02:34:42.523 align:middle line:84% Will we have a more sustainable economy? 02:34:42.523 --> 02:34:45.190 align:middle line:84% Or we will have a more circular economy, or something like that? 02:34:45.190 --> 02:34:48.213 align:middle line:84% Because you need that final performance measure, 02:34:48.213 --> 02:34:50.380 align:middle line:84% because otherwise, you're going to build your model. 02:34:50.380 --> 02:34:51.755 align:middle line:84% You're going to have no yardstick 02:34:51.755 --> 02:34:54.700 align:middle line:84% to tell you what to put in, and what to put out. 02:34:54.700 --> 02:34:56.422 align:middle line:84% So you're going to have to-- this 02:34:56.422 --> 02:34:58.380 align:middle line:84% is what was making Jason uncomfortable earlier, 02:34:58.380 --> 02:34:59.880 align:middle line:84% right, which is you're going to have 02:34:59.880 --> 02:35:02.763 align:middle line:84% to make some choices about what to really focus on. 02:35:02.763 --> 02:35:04.680 align:middle line:84% And that's ultimately a value judgment, right? 02:35:04.680 --> 02:35:06.320 align:middle line:84% It's about what are the things that you really 02:35:06.320 --> 02:35:07.360 align:middle line:90% care about and that. 02:35:07.360 --> 02:35:09.280 align:middle line:84% STUDENT: Yes, because it really helps 02:35:09.280 --> 02:35:10.920 align:middle line:90% to do a causal loop diagram. 02:35:10.920 --> 02:35:14.180 align:middle line:84% I mean, and maybe that's enough because I mean, 02:35:14.180 --> 02:35:17.180 align:middle line:84% that really helps, and just count the loops in a way. 02:35:17.180 --> 02:35:19.160 align:middle line:84% I mean, I could see that the people who 02:35:19.160 --> 02:35:22.100 align:middle line:84% are benefiting the most from the ecosystem 02:35:22.100 --> 02:35:26.520 align:middle line:84% have the most loops because that makes the most contacts. 02:35:26.520 --> 02:35:29.085 align:middle line:90% But I don't really know if-- 02:35:29.085 --> 02:35:29.960 align:middle line:90% NELSON REPENNING: OK. 02:35:29.960 --> 02:35:31.620 align:middle line:84% Well, so I think they're waiting outside. 02:35:31.620 --> 02:35:32.700 align:middle line:90% But I'm happy to chat with you. 02:35:32.700 --> 02:35:34.908 align:middle line:84% And I apologize that we ran out of time a little bit. 02:35:34.908 --> 02:35:36.320 align:middle line:90% But we did it. 02:35:36.320 --> 02:35:36.820 align:middle line:90% All right. 02:35:36.820 --> 02:35:37.380 align:middle line:90% Thanks, everyone. 02:35:37.380 --> 02:35:38.260 align:middle line:90% Thanks for your patience. 02:35:38.260 --> 02:35:39.620 align:middle line:84% This was maybe a little bit disorganized, 02:35:39.620 --> 02:35:41.540 align:middle line:84% but hopefully, there was some useful nuggets along the way. 02:35:41.540 --> 02:35:41.960 align:middle line:90% So-- 02:35:41.960 --> 02:35:42.752 align:middle line:90% STUDENT: Thank you. 02:35:42.752 --> 02:35:44.270 align:middle line:90% NELSON REPENNING: OK. 02:35:44.270 --> 02:35:48.000 align:middle line:90%