WEBVTT 00:00:00.000 --> 00:00:00.974 [SQUEAKING] 00:00:02.435 --> 00:00:04.383 [RUSTLING] 00:00:04.383 --> 00:00:09.750 [CLICKING] 00:00:09.750 --> 00:00:13.060 FRANK SCHILLBACH Let me get started on lecture 20. 00:00:13.060 --> 00:00:15.060 So I'm going to finish up with what we discussed 00:00:15.060 --> 00:00:16.653 last time, which is lecture 19. 00:00:16.653 --> 00:00:18.570 And then we're going to talk about lecture 20, 00:00:18.570 --> 00:00:23.687 which is about malleability and inaccessibility of preferences. 00:00:23.687 --> 00:00:25.770 For now, we're going to talk for a little bit more 00:00:25.770 --> 00:00:30.290 about defaults and frames, and nudges in particular. 00:00:30.290 --> 00:00:33.030 So what we left it off last time was we 00:00:33.030 --> 00:00:34.673 talked about default effects. 00:00:34.673 --> 00:00:36.090 We talked about that in particular 00:00:36.090 --> 00:00:37.260 in retirement savings. 00:00:37.260 --> 00:00:39.860 Defaults can be sort of setting essentially 00:00:39.860 --> 00:00:41.790 in one default option. 00:00:41.790 --> 00:00:45.540 That is like, what happens if you do nothing in a retirement 00:00:45.540 --> 00:00:47.190 savings account. 00:00:47.190 --> 00:00:49.050 Setting a certain default option can 00:00:49.050 --> 00:00:53.430 have very powerful effects in affecting people's behavior 00:00:53.430 --> 00:00:55.920 in a domain that's really important in this case, 00:00:55.920 --> 00:01:01.020 savings, and in a domain where traditional economic tools have 00:01:01.020 --> 00:01:05.519 not really gone very far, as in matching contributions, 00:01:05.519 --> 00:01:09.420 or any sort of like types of financial education, et cetera. 00:01:09.420 --> 00:01:10.920 It's just not doing very much. 00:01:10.920 --> 00:01:12.000 It's expensive to do. 00:01:12.000 --> 00:01:16.110 And it's hard to change behavior, while using defaults 00:01:16.110 --> 00:01:20.070 is really powerful, and can cause large behavior change 00:01:20.070 --> 00:01:25.550 or large changes in outcomes even years later. 00:01:25.550 --> 00:01:28.430 Let me talk a little bit about what is the optimal default 00:01:28.430 --> 00:01:29.750 decision regime. 00:01:29.750 --> 00:01:34.040 And that's getting into tricky territory. 00:01:34.040 --> 00:01:36.380 We talked about this last time a little bit. 00:01:36.380 --> 00:01:38.810 We'll talk about this again in the last lecture, 00:01:38.810 --> 00:01:41.880 on lecture 23, when we talk about policy. 00:01:41.880 --> 00:01:44.430 So when do you think about active choice versus default, 00:01:44.430 --> 00:01:45.680 so what is active choice? 00:01:45.680 --> 00:01:48.440 Active choice to remind you is like somebody has asked 00:01:48.440 --> 00:01:50.600 at the beginning, or directly asked, 00:01:50.600 --> 00:01:51.770 you have to make a choice. 00:01:51.770 --> 00:01:55.410 And I'm just sort of forcing you to do it, more or less. 00:01:55.410 --> 00:01:58.040 [INAUDIBLE] the company hires somebody and in the hiring 00:01:58.040 --> 00:02:00.860 package, there is just a form that you need to fill out. 00:02:00.860 --> 00:02:03.483 And otherwise you cannot really start [INAUDIBLE] company. 00:02:03.483 --> 00:02:05.150 Sometimes that's not really enforceable, 00:02:05.150 --> 00:02:06.942 but effectively, essentially just everybody 00:02:06.942 --> 00:02:08.206 makes some choice. 00:02:08.206 --> 00:02:10.789 That's essentially active choice versus default, which is just 00:02:10.789 --> 00:02:12.247 like, if you do nothing, you either 00:02:12.247 --> 00:02:15.870 get zero retirement savings, or some positive amounts. 00:02:15.870 --> 00:02:19.008 Now, if people are really different, then essentially, 00:02:19.008 --> 00:02:20.300 active choice is a great thing. 00:02:20.300 --> 00:02:22.525 Because if people really have different preferences, 00:02:22.525 --> 00:02:24.650 then we shouldn't just push them into one direction 00:02:24.650 --> 00:02:25.250 or the other. 00:02:25.250 --> 00:02:26.090 They should choose. 00:02:26.090 --> 00:02:29.090 Everybody chooses on their own what's best for them. 00:02:29.090 --> 00:02:32.390 In contrast, if you think most people are the same anyway, 00:02:32.390 --> 00:02:33.890 and in addition, people don't really 00:02:33.890 --> 00:02:35.720 know what's best for them, well then 00:02:35.720 --> 00:02:39.050 it might be better to just provide one default [INAUDIBLE] 00:02:39.050 --> 00:02:40.930 most people that's best for most people. 00:02:40.930 --> 00:02:42.680 And then most people will stick to anyway, 00:02:42.680 --> 00:02:44.990 because these defaults tend to stick. 00:02:44.990 --> 00:02:46.880 And if people don't know what they want 00:02:46.880 --> 00:02:48.020 or what's good for them, then sort 00:02:48.020 --> 00:02:50.062 of letting them choose actively might make things 00:02:50.062 --> 00:02:53.240 potentially worse anyway. 00:02:53.240 --> 00:02:55.422 And so that's sort of the fundamental trade 00:02:55.422 --> 00:02:57.908 off when people have come down on different sides of this. 00:02:57.908 --> 00:02:59.450 And it depends on your view, kind of, 00:02:59.450 --> 00:03:01.880 about sophistication and so on of customers 00:03:01.880 --> 00:03:04.010 who are [INAUDIBLE]. 00:03:04.010 --> 00:03:07.460 And one big and important issue with defaults 00:03:07.460 --> 00:03:09.460 and with the nudges, in general, which 00:03:09.460 --> 00:03:11.630 I'm going to talk about to you in a second 00:03:11.630 --> 00:03:14.580 is, we kind of want to make sure that we don't make people 00:03:14.580 --> 00:03:16.860 worse off in some ways by pushing people 00:03:16.860 --> 00:03:17.970 in certain directions. 00:03:17.970 --> 00:03:19.470 For example, one thing you might say 00:03:19.470 --> 00:03:22.380 is, well, why not set the default to like 5% or 10% 00:03:22.380 --> 00:03:24.060 of retirement savings? 00:03:24.060 --> 00:03:27.090 But then you worry about people potentially over-saving 00:03:27.090 --> 00:03:28.777 in their retirement savings accounts. 00:03:28.777 --> 00:03:31.110 And then essentially they have lots of credit card debt, 00:03:31.110 --> 00:03:33.960 potentially, which might be even larger. 00:03:33.960 --> 00:03:36.163 Sorry, the interest rate on the credit card debt 00:03:36.163 --> 00:03:37.080 might be really large. 00:03:37.080 --> 00:03:38.163 And people get in trouble. 00:03:38.163 --> 00:03:40.740 Or people then withdraw prematurely, and so on, 00:03:40.740 --> 00:03:43.830 which is [INAUDIBLE] potentially worse for them. 00:03:43.830 --> 00:03:46.050 We talked very briefly about the smart plan, 00:03:46.050 --> 00:03:50.240 which again, I'm going to get back to in lecture 23 as well. 00:03:50.240 --> 00:03:57.170 Now I also talked briefly about sort of default effects 00:03:57.170 --> 00:03:59.210 also are quite powerful in other settings. 00:03:59.210 --> 00:04:01.940 For example, in organ donations, in some places, 00:04:01.940 --> 00:04:04.830 organ donations are explicit consent, 00:04:04.830 --> 00:04:06.500 which is like you have to opt in, 00:04:06.500 --> 00:04:08.660 and otherwise you will not be an organ donor. 00:04:08.660 --> 00:04:11.060 In other countries-- and this is from a while ago-- 00:04:11.060 --> 00:04:13.813 in other countries, things are different, 00:04:13.813 --> 00:04:15.980 where there is what's called presumed consent, which 00:04:15.980 --> 00:04:17.360 is like you have to opt out. 00:04:17.360 --> 00:04:19.160 And otherwise, if you do not opt out, 00:04:19.160 --> 00:04:21.800 you're presumed to consenting to organ [INAUDIBLE].. 00:04:21.800 --> 00:04:24.050 We see essentially huge differences 00:04:24.050 --> 00:04:27.020 in otherwise quite similar countries, 00:04:27.020 --> 00:04:29.060 which sort of shows you that the default here 00:04:29.060 --> 00:04:31.370 can have really large effects. 00:04:31.370 --> 00:04:34.070 There's some other examples of default effects in particular 00:04:34.070 --> 00:04:35.840 when it comes to voter registration, 00:04:35.840 --> 00:04:37.760 or green energy choices. 00:04:37.760 --> 00:04:39.350 In some cases, in particular when 00:04:39.350 --> 00:04:41.930 it comes to voter registration, it's actually quite hard 00:04:41.930 --> 00:04:43.700 to come with arguments against setting 00:04:43.700 --> 00:04:45.050 a default in a certain way. 00:04:45.050 --> 00:04:47.360 Like in some sense, getting voters 00:04:47.360 --> 00:04:49.840 to be registered when you get say 00:04:49.840 --> 00:04:52.550 a driver's license or the like seems just very 00:04:52.550 --> 00:04:54.400 straightforward and-- 00:04:54.400 --> 00:04:57.020 or when they turn 18, that seems like a very straightforward 00:04:57.020 --> 00:04:58.070 thing to do. 00:04:58.070 --> 00:05:00.950 And there, in some ways, the defaults 00:05:00.950 --> 00:05:05.210 can really sort of remove some barriers and inefficiencies 00:05:05.210 --> 00:05:09.950 where we think we should agree that everybody should 00:05:09.950 --> 00:05:14.270 be able to vote if they would like to, with some exceptions, 00:05:14.270 --> 00:05:14.990 perhaps. 00:05:14.990 --> 00:05:17.090 And so, if you can default people 00:05:17.090 --> 00:05:19.730 into getting registered to vote, that seems like a good thing, 00:05:19.730 --> 00:05:20.690 unambiguously. 00:05:20.690 --> 00:05:24.680 So there sort of default effects can be really, really powerful. 00:05:24.680 --> 00:05:28.180 Now, let's talk a little bit about-- not just again, 00:05:28.180 --> 00:05:29.680 we're going to talk about this again 00:05:29.680 --> 00:05:31.702 in lecture 23, when it comes, about policy, 00:05:31.702 --> 00:05:33.160 but let me sort of define the issue 00:05:33.160 --> 00:05:35.210 and sort of be clear about that. 00:05:35.210 --> 00:05:37.960 So you might have sort of come across the very famous book 00:05:37.960 --> 00:05:40.720 by Richard Thaler, who recently won 00:05:40.720 --> 00:05:44.350 the Nobel Prize in economics, and Cass Sunstein, 00:05:44.350 --> 00:05:46.380 called Nudge. 00:05:46.380 --> 00:05:51.030 I'll give you the definition by Cass Sunstein, who 00:05:51.030 --> 00:05:54.720 says a nudge is a feature of the social environment that affects 00:05:54.720 --> 00:05:57.270 people's choices without imposing coercion 00:05:57.270 --> 00:06:01.200 of any kind of material incentive. 00:06:01.200 --> 00:06:03.912 So that-- default is sort of like a clear case 00:06:03.912 --> 00:06:05.370 of that, where essentially you just 00:06:05.370 --> 00:06:07.260 say, OK, if you don't do anything, 00:06:07.260 --> 00:06:10.020 I'm going to just pick a choice for you. 00:06:10.020 --> 00:06:13.050 But you can choose, however, in whichever way you want. 00:06:13.050 --> 00:06:16.540 And there is essentially no material incentive. 00:06:16.540 --> 00:06:19.272 But I'm changing essentially your choice environment 00:06:19.272 --> 00:06:20.730 or what people would call sometimes 00:06:20.730 --> 00:06:22.980 the choice architecture. 00:06:22.980 --> 00:06:24.655 And there's other versions of that. 00:06:24.655 --> 00:06:28.620 There's simplification, information disclosure. 00:06:28.620 --> 00:06:31.110 Again, these are all things that are usually available. 00:06:31.110 --> 00:06:33.360 People can figure it out on their own. 00:06:33.360 --> 00:06:36.250 But we make it easier for people. 00:06:36.250 --> 00:06:39.930 And one of the big mantras [INAUDIBLE] 00:06:39.930 --> 00:06:43.290 about like, what's the main thing we should do 00:06:43.290 --> 00:06:45.600 to improve people's behavior? 00:06:45.600 --> 00:06:47.850 Dick Thaler would often say, make it easy for people. 00:06:47.850 --> 00:06:50.100 Essentially, people are often-- 00:06:50.100 --> 00:06:52.770 and he said that about himself-- people are lazy. 00:06:52.770 --> 00:06:54.600 And they make simple choices. 00:06:54.600 --> 00:06:57.550 And if you make it easy for them to make the right choice, 00:06:57.550 --> 00:07:01.620 they will be more likely to make good choices for them. 00:07:01.620 --> 00:07:05.040 There's some other things like warnings, reminders. 00:07:05.040 --> 00:07:07.370 Notice for reminders, for example, again, you already 00:07:07.370 --> 00:07:09.120 have the information that somebody gave it 00:07:09.120 --> 00:07:10.480 to you previously. 00:07:10.480 --> 00:07:15.270 So really it's only like if you have memory issues, 00:07:15.270 --> 00:07:16.680 that might be useful for you. 00:07:16.680 --> 00:07:19.590 There's no material incentives or [INAUDIBLE].. 00:07:19.590 --> 00:07:20.910 Use of social norms. 00:07:20.910 --> 00:07:24.480 That would be things like I'm telling you all of your friends 00:07:24.480 --> 00:07:25.110 did x. 00:07:25.110 --> 00:07:27.450 Now would you like to do x or y? 00:07:27.450 --> 00:07:31.110 Or I can send you letters and say your neighbors used 00:07:31.110 --> 00:07:34.140 so much energy, and you use more energy than your neighbor. 00:07:34.140 --> 00:07:37.350 And then I put like a sad face next to it, or something. 00:07:37.350 --> 00:07:40.860 And then you might feel bad about it because 00:07:40.860 --> 00:07:44.950 of social norms, and sort of reduce your energy usage, 00:07:44.950 --> 00:07:47.860 which people have in fact done. 00:07:47.860 --> 00:07:50.190 You can also like increase ease and convenience. 00:07:50.190 --> 00:07:53.550 This is exactly what like [INAUDIBLE] convenience. 00:07:53.550 --> 00:07:55.740 That's exactly what Dick Thaler would say. 00:07:55.740 --> 00:07:58.560 Those are things like, for example, I 00:07:58.560 --> 00:08:04.560 can put Snickers bars next to the counter at the cashier. 00:08:04.560 --> 00:08:07.300 Or I can put apples next to it. 00:08:07.300 --> 00:08:09.853 And people are more likely to buy apples 00:08:09.853 --> 00:08:11.520 when it's easy or simple or when they're 00:08:11.520 --> 00:08:13.320 reminded of eating apples. 00:08:13.320 --> 00:08:16.410 And they're more likely to eat Snickers bars essentially 00:08:16.410 --> 00:08:18.480 if [INAUDIBLE] to them. 00:08:18.480 --> 00:08:20.160 To the extent that you want people 00:08:20.160 --> 00:08:24.270 to make what one might think of as good choices, 00:08:24.270 --> 00:08:27.060 making it easy for them to make the good choice without sort 00:08:27.060 --> 00:08:29.700 of restricting anybody's choice sets 00:08:29.700 --> 00:08:33.750 or coercing them in any way is considered a nudge. 00:08:33.750 --> 00:08:37.320 And then there's also some things about framing of choices 00:08:37.320 --> 00:08:42.260 as in gains versus losses, in terms of their contracts. 00:08:42.260 --> 00:08:43.400 Any questions on this? 00:08:49.920 --> 00:08:50.420 OK. 00:08:50.420 --> 00:08:52.513 So then let me give you, briefly, sort of-- 00:08:52.513 --> 00:08:53.930 and again, we're going to get back 00:08:53.930 --> 00:08:55.202 to all of this in lecture 23. 00:08:55.202 --> 00:08:56.660 But let me briefly sort of give you 00:08:56.660 --> 00:08:59.140 a couple of examples of nudges in some cases 00:08:59.140 --> 00:09:03.820 where it's pretty clear that some nudges are a good idea. 00:09:03.820 --> 00:09:05.820 So one thing that we talked about quite 00:09:05.820 --> 00:09:07.800 a bit already when it comes to procrastination 00:09:07.800 --> 00:09:12.990 or self-control problems present bias, or [? losing focus ?] 00:09:12.990 --> 00:09:17.010 is the health domain, and what people and in particular, 00:09:17.010 --> 00:09:20.200 individuals and society often have aligned goals, 00:09:20.200 --> 00:09:23.015 where individuals often want behavioral change. 00:09:23.015 --> 00:09:24.390 They want to improve their diets, 00:09:24.390 --> 00:09:27.480 increase physical activity, stop smoking, get vaccinated, 00:09:27.480 --> 00:09:29.070 use less energy, and so on. 00:09:29.070 --> 00:09:33.190 And often, there are societal costs of these behaviors, 00:09:33.190 --> 00:09:35.430 or lack of those behaviors. 00:09:35.430 --> 00:09:39.870 Sometimes there's externalities from smoking or getting or not 00:09:39.870 --> 00:09:41.250 getting vaccinated. 00:09:41.250 --> 00:09:43.590 Sometimes [INAUDIBLE] health care costs and the 00:09:43.590 --> 00:09:47.640 like from obesity and so on, where it's bad for everybody 00:09:47.640 --> 00:09:51.120 if the population is sick. 00:09:51.120 --> 00:09:53.340 So these goals are now aligned. 00:09:53.340 --> 00:09:54.840 And the social planner might want 00:09:54.840 --> 00:09:59.100 people to improve their diet, or the government might want that, 00:09:59.100 --> 00:10:02.040 or might want to reduce smoking and so on. 00:10:02.040 --> 00:10:04.680 But individuals, as we saw previously, 00:10:04.680 --> 00:10:06.750 often fail to follow through. 00:10:06.750 --> 00:10:10.200 And sort of often, education and information interventions 00:10:10.200 --> 00:10:12.820 are often ineffective. 00:10:12.820 --> 00:10:15.030 Even price interventions are often ineffective, 00:10:15.030 --> 00:10:17.900 or in the sense of like increasing prices, 00:10:17.900 --> 00:10:19.900 or incentivizing people, it doesn't always work. 00:10:19.900 --> 00:10:21.555 Plus, it's quite expensive to do. 00:10:21.555 --> 00:10:23.430 So one natural question you might want to ask 00:10:23.430 --> 00:10:29.070 is, can we use some nudges to align intentions and actions? 00:10:29.070 --> 00:10:31.750 And the reason why this is really popular, in particular, 00:10:31.750 --> 00:10:35.220 among governments, is that it's really cheap to do this. 00:10:35.220 --> 00:10:37.710 So it's actually sending some reminders or the like 00:10:37.710 --> 00:10:40.050 or providing information in one way or the other, 00:10:40.050 --> 00:10:41.610 is essentially free. 00:10:41.610 --> 00:10:45.240 And if governments don't have a lot of money, 00:10:45.240 --> 00:10:46.860 that's an easy thing to do as opposed 00:10:46.860 --> 00:10:51.850 to paying people, which is quite expensive in many cases. 00:10:51.850 --> 00:10:54.480 So here's an example of a free intervention which is 00:10:54.480 --> 00:10:56.340 about flu shot communications. 00:10:56.340 --> 00:11:00.270 This is a study by Katy Milkman and co-authors from 2011. 00:11:00.270 --> 00:11:03.870 So here, the goal of the intervention 00:11:03.870 --> 00:11:08.220 is to get people to sign up, or to actually get flu shots. 00:11:08.220 --> 00:11:10.740 So the control group got an informational mailing. 00:11:10.740 --> 00:11:12.090 I'll show you this in a second. 00:11:12.090 --> 00:11:17.000 The treatment group got the same, plus an encouragement 00:11:17.000 --> 00:11:18.840 to make a date plan. 00:11:18.840 --> 00:11:23.550 And the treatment group got the same plus make a date plan, 00:11:23.550 --> 00:11:24.420 and a time plan. 00:11:24.420 --> 00:11:25.987 Let me show you what this looks like. 00:11:25.987 --> 00:11:27.320 So here's the control condition. 00:11:27.320 --> 00:11:30.368 This is just an informational mailing. 00:11:30.368 --> 00:11:32.448 It's saying here, the company-- 00:11:32.448 --> 00:11:33.990 this is a company trying to do this-- 00:11:33.990 --> 00:11:37.350 is holding a free flu shot clinic. 00:11:37.350 --> 00:11:40.290 And companies very much want their workers to get flu shots, 00:11:40.290 --> 00:11:44.070 because that reduces essentially sort of sick days. 00:11:44.070 --> 00:11:45.690 If everybody gets their flu shot, 00:11:45.690 --> 00:11:48.190 not only are they healthier and it's probably good for them. 00:11:48.190 --> 00:11:51.240 They're like less at risk of dying and infecting each other, 00:11:51.240 --> 00:11:52.692 or just being sick. 00:11:52.692 --> 00:11:54.150 But it's also good for the company, 00:11:54.150 --> 00:12:00.150 because the company now gets workers to actually show up 00:12:00.150 --> 00:12:02.533 more, and have fewer sick days. 00:12:02.533 --> 00:12:04.200 So here's sort of the control condition. 00:12:04.200 --> 00:12:08.510 It looks like a perfectly reasonable letter, 00:12:08.510 --> 00:12:12.090 where you say, here is these days, on which you can sign up. 00:12:12.090 --> 00:12:14.820 And you are informed about the dates and times 00:12:14.820 --> 00:12:17.520 of the workplace flu clinic. 00:12:17.520 --> 00:12:21.210 Now, the treatment condition is the date plan condition, 00:12:21.210 --> 00:12:24.360 which essentially is asking people to-- 00:12:24.360 --> 00:12:27.660 or invite people to choose a concrete date 00:12:27.660 --> 00:12:30.000 for getting a flu vaccine. 00:12:30.000 --> 00:12:33.510 And the rest of the information [? mailer ?] 00:12:33.510 --> 00:12:34.840 is exactly the same. 00:12:34.840 --> 00:12:37.500 So essentially, that's just asking people, 00:12:37.500 --> 00:12:43.200 pick a day of the week and a month, or a month and a day, 00:12:43.200 --> 00:12:45.100 and a day of week, and just write it down. 00:12:45.100 --> 00:12:47.100 And presumably, that also then encourages people 00:12:47.100 --> 00:12:50.010 to put it into their calendar or the like 00:12:50.010 --> 00:12:53.070 and sort of acts as some form of a reminder. 00:12:53.070 --> 00:12:54.527 Notice, there's no incentive here. 00:12:54.527 --> 00:12:55.860 There's no financial incentives. 00:12:55.860 --> 00:12:56.910 There's no coercion and so on. 00:12:56.910 --> 00:12:59.430 It's just saying why don't you just pick a day right now. 00:12:59.430 --> 00:13:02.220 Just have a look, check your calendar, and just figure out. 00:13:02.220 --> 00:13:04.250 Make a plan for this. 00:13:04.250 --> 00:13:08.872 And the second condition is then make a date and a time plan 00:13:08.872 --> 00:13:10.580 condition, which is to say, why don't you 00:13:10.580 --> 00:13:15.030 also pick a time, which is even more concrete. 00:13:15.030 --> 00:13:19.250 That would be Monday, October 26, at like 9:00 AM 00:13:19.250 --> 00:13:20.590 or whenever I'm going to go. 00:13:20.590 --> 00:13:22.215 Presumably, that also encourages people 00:13:22.215 --> 00:13:24.240 to put it into their calendar. 00:13:24.240 --> 00:13:26.570 Now what you then get in this experiment-- 00:13:26.570 --> 00:13:28.430 it is a very simple experiment. 00:13:28.430 --> 00:13:30.230 You get, when you just send a letter, 00:13:30.230 --> 00:13:32.630 33% actually get the flu shot. 00:13:32.630 --> 00:13:36.080 When you also send a date plan, 34%. 00:13:36.080 --> 00:13:39.122 And when you send a date plan and a time plan, it's 37%. 00:13:39.122 --> 00:13:40.580 Now, these effects are a little bit 00:13:40.580 --> 00:13:42.950 underwhelming in the sense they're kind of small. 00:13:42.950 --> 00:13:48.470 It's like 1.6 percentage points at 4.2 percentage points. 00:13:48.470 --> 00:13:51.800 And in relative terms that's not huge effect. 00:13:51.800 --> 00:13:54.050 But notice what's beautiful about here 00:13:54.050 --> 00:13:57.350 is that the costs are essentially zero. 00:13:57.350 --> 00:14:00.580 So in some sense, if you value additional flu shot adherence, 00:14:00.580 --> 00:14:06.770 if you value more people getting flu shots, then a 4.2 00:14:06.770 --> 00:14:09.230 percentage points increase, which is about like a 10% 00:14:09.230 --> 00:14:11.450 or whatever, 11%, 12% or something 00:14:11.450 --> 00:14:13.880 increase relative to the control group, that's 00:14:13.880 --> 00:14:16.460 quite a bit of an increase for something 00:14:16.460 --> 00:14:18.000 that's essentially free. 00:14:18.000 --> 00:14:20.310 So the cost effectiveness of this is, of course, very, 00:14:20.310 --> 00:14:20.900 very high. 00:14:20.900 --> 00:14:22.525 Because it's essentially free, or maybe 00:14:22.525 --> 00:14:25.190 it cost somebody a few minutes to sort of write this 00:14:25.190 --> 00:14:28.790 down, or design the sheet. 00:14:28.790 --> 00:14:31.220 But really it costs exactly the same 00:14:31.220 --> 00:14:33.500 to send the sheet here on the left 00:14:33.500 --> 00:14:35.580 versus the sheet here on the right. 00:14:35.580 --> 00:14:39.410 So that's kind of an example of a win-win kind of situation 00:14:39.410 --> 00:14:44.330 where we make things essentially more effective. 00:14:44.330 --> 00:14:47.580 It's not really hurting anybody, and it's cheap and free to do. 00:14:47.580 --> 00:14:48.980 And so everybody is happy. 00:14:48.980 --> 00:14:52.190 And it's sort of uncontroversial in the sense of nobody 00:14:52.190 --> 00:14:53.840 would argue that this is a bad thing. 00:14:53.840 --> 00:14:55.400 Of course, some people might sort of 00:14:55.400 --> 00:14:57.990 have some concerns about the flu shots themselves, 00:14:57.990 --> 00:15:00.710 which I don't think there's any actual evidence against that 00:15:00.710 --> 00:15:01.548 as far as I know. 00:15:01.548 --> 00:15:03.590 But that's sort of an example where we say, look, 00:15:03.590 --> 00:15:04.760 this is very cheap to do. 00:15:04.760 --> 00:15:08.457 So presumably you should sort of pick the right mailing, 00:15:08.457 --> 00:15:10.790 or the right design, if I'm mailing, to encourage people 00:15:10.790 --> 00:15:14.390 to do desirable things, at least from the company's perspective, 00:15:14.390 --> 00:15:17.240 while also preserving people's freedom, 00:15:17.240 --> 00:15:21.530 not spending any money, but also not encouraging anybody. 00:15:21.530 --> 00:15:22.640 Any questions on this? 00:15:28.220 --> 00:15:28.900 OK. 00:15:28.900 --> 00:15:31.690 So here's another example of a similar intervention, 00:15:31.690 --> 00:15:34.780 which is signing up for FAFSA. 00:15:34.780 --> 00:15:36.250 This is a paper by Bettinger et al. 00:15:36.250 --> 00:15:41.950 in 2009, where essentially, people were provided 00:15:41.950 --> 00:15:44.950 free additional assistance in completing and filing 00:15:44.950 --> 00:15:47.680 applications for college financial aid. 00:15:47.680 --> 00:15:51.220 This is FAFSA, the financial aid for college. 00:15:51.220 --> 00:15:55.660 And it really increased college enrollment-- 00:15:55.660 --> 00:15:59.020 not only FAFSA completion, but also college enrollment. 00:15:59.020 --> 00:16:01.940 And there's sort of different conditions. 00:16:01.940 --> 00:16:04.490 One is control versus just providing people information 00:16:04.490 --> 00:16:04.990 only. 00:16:04.990 --> 00:16:06.990 Here's [INAUDIBLE] information of like you could 00:16:06.990 --> 00:16:08.703 get some FAFSA help and so on. 00:16:08.703 --> 00:16:10.120 But it seems to be really what you 00:16:10.120 --> 00:16:15.850 need is some additional assistance of completing 00:16:15.850 --> 00:16:17.440 and filing this application. 00:16:17.440 --> 00:16:22.240 Just somebody essentially walking you through the filing 00:16:22.240 --> 00:16:23.730 has a huge impact. 00:16:23.730 --> 00:16:26.230 There's a bit of a question of, is that really just a nudge, 00:16:26.230 --> 00:16:28.060 or is it like a more powerful intervention. 00:16:28.060 --> 00:16:29.650 But it's a really minor intervention 00:16:29.650 --> 00:16:31.570 where some people just need some level of help 00:16:31.570 --> 00:16:33.670 in filling out this form. 00:16:33.670 --> 00:16:35.530 And it has pretty large effects, if you 00:16:35.530 --> 00:16:37.690 look at the FAFSA [INAUDIBLE] completion 00:16:37.690 --> 00:16:41.800 of these forms, reasonably large effect on college enrollment. 00:16:41.800 --> 00:16:44.470 And you might sort of think this is a hugely important choice, 00:16:44.470 --> 00:16:49.840 and sort of if you get people to make that really big change, 00:16:49.840 --> 00:16:53.420 that's pretty remarkable for doing some very simple things. 00:16:53.420 --> 00:16:57.340 [INAUDIBLE] help me complete and fill out a simple form, 00:16:57.340 --> 00:17:00.190 something that you could have done anyway. 00:17:00.190 --> 00:17:04.750 So, and here again, so this is a relatively cheap thing to do. 00:17:04.750 --> 00:17:07.089 And it's essentially equivalent to the impact 00:17:07.089 --> 00:17:10.569 of several thousands of dollars of education subsidy. 00:17:10.569 --> 00:17:14.180 So if you wanted to essentially pay people to go to college, 00:17:14.180 --> 00:17:15.790 [INAUDIBLE] subsidize that, which 00:17:15.790 --> 00:17:18.730 you might want to do for other reasons anyway, 00:17:18.730 --> 00:17:22.780 one key issue with that is that not only is it perhaps not 00:17:22.780 --> 00:17:24.970 particularly effective, but also you're 00:17:24.970 --> 00:17:26.800 going to pay a lot of for marginal people. 00:17:26.800 --> 00:17:28.270 You're going to pay a lot of people who would 00:17:28.270 --> 00:17:29.530 have gone to college anyway. 00:17:29.530 --> 00:17:30.890 You're going to pay the subsidy. 00:17:30.890 --> 00:17:34.810 So if you have a low budget, that's really tough to do. 00:17:34.810 --> 00:17:38.410 And sort of then there's not a lot of-- 00:17:38.410 --> 00:17:40.630 the cost effectiveness tends to be very low. 00:17:40.630 --> 00:17:43.180 Because you have essentially a low impact, 00:17:43.180 --> 00:17:45.580 but like really high costs. 00:17:45.580 --> 00:17:48.220 And you can read about this more in the link that I had. 00:17:48.220 --> 00:17:49.720 But essentially the bottom line here 00:17:49.720 --> 00:17:52.210 is, again, you can do very small changes that 00:17:52.210 --> 00:17:55.540 can have pretty large effects. 00:17:55.540 --> 00:17:57.970 There's other types of examples that government use, 00:17:57.970 --> 00:17:59.440 or where governments use-- 00:17:59.440 --> 00:18:02.380 not just-- another example would be 00:18:02.380 --> 00:18:05.440 sending out letters for people to pay their taxes, which 00:18:05.440 --> 00:18:07.960 essentially are like reminders, or often appealing 00:18:07.960 --> 00:18:12.460 to people's sort of prosociality in various ways. 00:18:12.460 --> 00:18:14.080 But the government sends you letters. 00:18:14.080 --> 00:18:15.670 And there's different ways in which 00:18:15.670 --> 00:18:17.290 you can write these letters. 00:18:17.290 --> 00:18:18.970 You could say you should pay your taxes, 00:18:18.970 --> 00:18:20.980 or everybody's paying their taxes, 00:18:20.980 --> 00:18:23.420 or good citizens pay their taxes, and so on. 00:18:23.420 --> 00:18:25.900 And people-- there's a whole industry of nudged units 00:18:25.900 --> 00:18:27.760 where people then try to figure out what's 00:18:27.760 --> 00:18:29.810 the best way of doing that. 00:18:29.810 --> 00:18:32.170 And by doing that in different ways, 00:18:32.170 --> 00:18:37.160 that can have a pretty large effect on people's behaviors. 00:18:37.160 --> 00:18:39.650 And the governments sort of love this, because it's 00:18:39.650 --> 00:18:40.940 a very cheap thing to do. 00:18:40.940 --> 00:18:43.910 And it can have large impact and change behavior quite a bit. 00:18:43.910 --> 00:18:46.730 In this case, in the tax case, increase revenue 00:18:46.730 --> 00:18:50.420 by quite a bit without costing the government very much. 00:18:50.420 --> 00:18:52.760 Because often, [INAUDIBLE] send some letters anyway. 00:18:52.760 --> 00:18:55.430 And they send some more in some less effective ways 00:18:55.430 --> 00:18:56.760 of doing that. 00:18:56.760 --> 00:18:58.700 When you think about the intervention 00:18:58.700 --> 00:19:00.710 that I have here on the slide, the FAFSA one 00:19:00.710 --> 00:19:02.240 is essentially just-- 00:19:02.240 --> 00:19:03.860 the effects almost surely will be 00:19:03.860 --> 00:19:07.250 persistent in the sense of like, this is a one shot choice, 00:19:07.250 --> 00:19:09.720 whether you send your kid to college or not. 00:19:09.720 --> 00:19:13.625 And once you increase the college enrollment, 00:19:13.625 --> 00:19:17.865 which might have some other issues, but once you do that, 00:19:17.865 --> 00:19:19.490 there's going to be persistent effects. 00:19:19.490 --> 00:19:21.770 Because once kids go to college, that 00:19:21.770 --> 00:19:23.850 has effects on their behaviors and so on. 00:19:23.850 --> 00:19:25.040 Now when you're saying something else, which 00:19:25.040 --> 00:19:26.060 is some other decisions, [INAUDIBLE] 00:19:26.060 --> 00:19:28.330 just like about paying taxes where every year 00:19:28.330 --> 00:19:30.770 you get this letter from the government that sort of tells 00:19:30.770 --> 00:19:33.590 you you should pay your taxes, maybe the first time it works. 00:19:33.590 --> 00:19:34.840 Maybe the second time as well. 00:19:34.840 --> 00:19:37.130 But at some point, you're like, whatever. 00:19:37.130 --> 00:19:38.610 I don't care. 00:19:38.610 --> 00:19:40.070 That's a great question. 00:19:40.070 --> 00:19:44.000 And I don't know whether we have lots of evidence or not. 00:19:44.000 --> 00:19:46.568 I think the governments would sort of-- 00:19:46.568 --> 00:19:48.110 I think there's some research on this 00:19:48.110 --> 00:19:51.610 that I just am not sure I'm aware of. 00:19:51.610 --> 00:19:54.830 I think governments or people would say, well, in some sense, 00:19:54.830 --> 00:19:57.150 we don't care that much about it. 00:19:57.150 --> 00:19:58.300 Of course we care. 00:19:58.300 --> 00:20:01.317 But it doesn't change the fact that in the first year-- 00:20:01.317 --> 00:20:03.650 so we know that in the first year, it works pretty well. 00:20:03.650 --> 00:20:05.858 And there's some things that work better than others. 00:20:05.858 --> 00:20:08.470 Many governments send some letters or some forms 00:20:08.470 --> 00:20:09.757 of mailers anyway. 00:20:09.757 --> 00:20:12.340 So what you want to do is kind of like make sure you optimize, 00:20:12.340 --> 00:20:15.670 at least the first or second time you contact people. 00:20:15.670 --> 00:20:18.100 And that might have some persistent effects 00:20:18.100 --> 00:20:20.500 by themselves, even if you don't re-contact them. 00:20:20.500 --> 00:20:22.300 Because once you start paying your taxes, 00:20:22.300 --> 00:20:25.120 maybe then it becomes a habit, and people do it anyway. 00:20:25.120 --> 00:20:26.950 But even if that's not persistent, 00:20:26.950 --> 00:20:29.650 governments would say, well, if that gets people to pay taxes 00:20:29.650 --> 00:20:32.890 more at least once or twice, that's worth 00:20:32.890 --> 00:20:34.450 sort of the money spent. 00:20:34.450 --> 00:20:38.770 [INAUDIBLE] somebody essentially needs to design this letters. 00:20:38.770 --> 00:20:40.810 There's like, templates of what you can use. 00:20:40.810 --> 00:20:48.060 And that's essentially very cheap to do 00:20:48.060 --> 00:20:53.400 and has pretty large effects at least in the short run. 00:20:53.400 --> 00:20:56.490 I think [INAUDIBLE] correct, which is sometimes it's like, 00:20:56.490 --> 00:20:57.450 some effects-- 00:20:57.450 --> 00:21:00.120 it's just information versus some things about making 00:21:00.120 --> 00:21:02.043 you feel bad, or social norms. 00:21:02.043 --> 00:21:04.710 I think the information things-- if you think people just forget 00:21:04.710 --> 00:21:07.200 stuff-- for example, if people forget to get their flu shot 00:21:07.200 --> 00:21:09.492 and so on, you should probably send reminders and so on 00:21:09.492 --> 00:21:11.670 and get them to do that. 00:21:11.670 --> 00:21:14.840 If there's sort of these social norms, or sort of [INAUDIBLE] 00:21:14.840 --> 00:21:17.070 makes you feel bad in some ways, maybe those things 00:21:17.070 --> 00:21:20.280 are maybe more likely to potentially at least go away 00:21:20.280 --> 00:21:22.527 once you do it like 17 times, because at some point, 00:21:22.527 --> 00:21:25.110 people are like, yeah, whatever, the same letter I got always. 00:21:25.110 --> 00:21:26.700 And I haven't paid my taxes last time. 00:21:26.700 --> 00:21:29.090 I'm not going to pay them now either. 00:21:29.090 --> 00:21:32.100 It's additional assistance in completing and filing 00:21:32.100 --> 00:21:33.480 applications. 00:21:33.480 --> 00:21:34.740 So essentially, it's not-- 00:21:34.740 --> 00:21:36.750 so the financial aid is always available. 00:21:36.750 --> 00:21:39.420 It's just who takes advantage of financial aid 00:21:39.420 --> 00:21:41.940 depends on how easy you make it to people, 00:21:41.940 --> 00:21:43.710 and whether you support them, provide 00:21:43.710 --> 00:21:47.310 some support in completing and filing applications. 00:21:47.310 --> 00:21:50.550 There's also [INAUDIBLE] with that where essentially, once 00:21:50.550 --> 00:21:52.440 you do your taxes, essentially, it already 00:21:52.440 --> 00:21:56.620 prefills you the forms, which is even easier to do. 00:21:56.620 --> 00:21:59.070 But essentially, this falls under the category of making 00:21:59.070 --> 00:22:00.930 things easy for people. 00:22:00.930 --> 00:22:05.140 And that can sort of change behaviors quite a bit. 00:22:05.140 --> 00:22:07.980 But the key part in the nudges is-- 00:22:07.980 --> 00:22:10.500 and let me get back again to the definition-- 00:22:10.500 --> 00:22:16.200 is that there's no financial incentives. 00:22:16.200 --> 00:22:20.340 So this part here is without imposing any coercion, 00:22:20.340 --> 00:22:24.210 or without paying any kind of material incentives, 00:22:24.210 --> 00:22:26.242 or at least, only very, very small incentives. 00:22:26.242 --> 00:22:28.200 There's a bit of a question how to define that. 00:22:28.200 --> 00:22:31.800 But once I pay you $1,000 to go to college or whatever, 00:22:31.800 --> 00:22:34.430 subsidize you, that's not a nudge anymore. 00:22:34.430 --> 00:22:38.340 Or once I'm taking away certain options from you, [INAUDIBLE] 00:22:38.340 --> 00:22:40.470 eliminating choices from your choice set, 00:22:40.470 --> 00:22:41.743 again, that's not a nudge. 00:22:41.743 --> 00:22:42.660 That's something else. 00:22:42.660 --> 00:22:43.993 That's changing your choice set. 00:22:43.993 --> 00:22:45.720 This one is-- nudges are explicitly 00:22:45.720 --> 00:22:51.600 about keeping people's choices fully free and available, 00:22:51.600 --> 00:22:56.060 but making it easier for them to make a certain choice. 00:22:56.060 --> 00:22:57.770 Presumably, a choice that's desired 00:22:57.770 --> 00:22:59.420 in some ways by the social planner, 00:22:59.420 --> 00:23:03.030 or about people's plans for their future. 00:23:03.030 --> 00:23:07.640 So anyway, there's a large sort of set of examples 00:23:07.640 --> 00:23:09.180 of different nudges and so on. 00:23:09.180 --> 00:23:11.000 And they can be fairly effective. 00:23:11.000 --> 00:23:12.500 Of course, not for everybody and not 00:23:12.500 --> 00:23:15.830 for the fraction of people who are swayed by these nudges 00:23:15.830 --> 00:23:19.020 is not huge, in part because if you think about it, 00:23:19.020 --> 00:23:21.542 essentially people need to be marginal in some ways. 00:23:21.542 --> 00:23:23.000 If people just really, really don't 00:23:23.000 --> 00:23:25.417 want to go to college because it's too expensive for them, 00:23:25.417 --> 00:23:27.980 or whatever for other reasons, or if people really 00:23:27.980 --> 00:23:30.240 want to go to college anyway, the nudge 00:23:30.240 --> 00:23:31.580 will not change that behavior. 00:23:31.580 --> 00:23:33.330 Whether somebody fills out the form or not 00:23:33.330 --> 00:23:35.000 doesn't really matter. 00:23:35.000 --> 00:23:36.500 But if somebody is sort of like, ah, 00:23:36.500 --> 00:23:38.030 maybe I should send the kid to college or not, 00:23:38.030 --> 00:23:39.500 but now [INAUDIBLE] form and then 00:23:39.500 --> 00:23:41.360 they forget about it and so on and so forth, 00:23:41.360 --> 00:23:42.818 for that kind of people, there will 00:23:42.818 --> 00:23:45.530 be potentially large effects. 00:23:45.530 --> 00:23:47.420 But sort of then by definition, in some ways, 00:23:47.420 --> 00:23:49.420 the fraction of people who are actually marginal 00:23:49.420 --> 00:23:50.570 will not be huge. 00:23:50.570 --> 00:23:53.930 But again, since the costs tend to be extremely low, 00:23:53.930 --> 00:23:58.370 it tends to be very much cost effective to do so. 00:23:58.370 --> 00:24:01.730 Now again, sort of previewing a little bit what 00:24:01.730 --> 00:24:05.180 comes in the policy lecture is, so minor interventions 00:24:05.180 --> 00:24:06.960 can have large effects. 00:24:06.960 --> 00:24:10.310 But nudges can often achieve sort of-- 00:24:10.310 --> 00:24:11.810 in some cases, nudges can achieve 00:24:11.810 --> 00:24:13.057 unambiguous improvements. 00:24:13.057 --> 00:24:15.140 Like, if you send people reminders and they forget 00:24:15.140 --> 00:24:17.780 to get the flu shot, and now they do the flu shot, 00:24:17.780 --> 00:24:21.080 or if people don't take their medication, 00:24:21.080 --> 00:24:23.360 and you remind them, or provide information, 00:24:23.360 --> 00:24:25.730 or some other forms [INAUDIBLE] [? tell ?] 00:24:25.730 --> 00:24:27.890 people to take their medication, it's 00:24:27.890 --> 00:24:30.970 pretty clear that that's making people better off. 00:24:30.970 --> 00:24:33.530 But there's also a bunch of challenges and other situations 00:24:33.530 --> 00:24:35.860 which is kind of like-- 00:24:35.860 --> 00:24:37.600 in some cases, it's not at all obvious. 00:24:37.600 --> 00:24:39.310 Like, which nudge to choose-- 00:24:39.310 --> 00:24:41.290 are we making everybody better off? 00:24:41.290 --> 00:24:43.047 Are some people made worse off? 00:24:43.047 --> 00:24:45.130 For example, should everybody save for retirement? 00:24:45.130 --> 00:24:48.370 Are we pushing people to save too much? 00:24:48.370 --> 00:24:49.750 Should everybody go to college? 00:24:49.750 --> 00:24:51.100 Maybe in some ways, for some people, 00:24:51.100 --> 00:24:53.267 it's more suitable than for others to go to college. 00:24:53.267 --> 00:24:56.210 Often it's a huge financial expense for the family. 00:24:56.210 --> 00:24:59.560 And if then the job prospects are not necessarily better 00:24:59.560 --> 00:25:00.950 compared to not going to college, 00:25:00.950 --> 00:25:03.257 then it's not clear that one should do that. 00:25:03.257 --> 00:25:05.590 There's also some evidence or some concerns about nudges 00:25:05.590 --> 00:25:07.503 making people feel bad. 00:25:07.503 --> 00:25:08.920 So you get these letters that says 00:25:08.920 --> 00:25:11.200 you're destroying the environment, 00:25:11.200 --> 00:25:13.630 and [INAUDIBLE] all these sad faces, 00:25:13.630 --> 00:25:15.250 and you're worse than your neighbors, 00:25:15.250 --> 00:25:16.840 people might just feel bad about it. 00:25:16.840 --> 00:25:22.450 And we should put some weight on that. 00:25:22.450 --> 00:25:26.560 If you send these letters to 1,000 households, 00:25:26.560 --> 00:25:28.640 and 50 households or 20 households 00:25:28.640 --> 00:25:30.640 change their behavior in certain ways, 00:25:30.640 --> 00:25:32.740 but 300 households feel bad about it 00:25:32.740 --> 00:25:35.500 because they feel uncomfortable getting these nudges, 00:25:35.500 --> 00:25:40.130 or you get all these phone calls and spam about the environment 00:25:40.130 --> 00:25:42.550 from sending too many letters, we 00:25:42.550 --> 00:25:43.940 should put some weight on that. 00:25:43.940 --> 00:25:46.280 And then it's like, then there's some trade offs. 00:25:46.280 --> 00:25:47.560 It's not like entirely free. 00:25:47.560 --> 00:25:50.257 And then there are some costs and benefits of that. 00:25:50.257 --> 00:25:52.090 There's also some questions about which self 00:25:52.090 --> 00:25:53.210 should we respect. 00:25:53.210 --> 00:25:54.770 We talked about this a lot before, 00:25:54.770 --> 00:25:59.070 which is if one self wants to be very virtuous 00:25:59.070 --> 00:26:01.080 and exercise a lot and so on, and the other one 00:26:01.080 --> 00:26:03.210 wants to sit on the couch and watch TV, 00:26:03.210 --> 00:26:06.370 or if one self wants to smoke and the other one does not, 00:26:06.370 --> 00:26:09.210 it's not obvious that we should respect the long run 00:26:09.210 --> 00:26:14.610 self compared to the short run self that wants to just enjoy 00:26:14.610 --> 00:26:15.600 themselves, and so on. 00:26:15.600 --> 00:26:18.360 And sort of now there's tricky issues [INAUDIBLE] 00:26:18.360 --> 00:26:21.178 are we actually making people better off or not. 00:26:21.178 --> 00:26:23.470 And we'll get back to these issues in the last lecture, 00:26:23.470 --> 00:26:26.400 which talks about policy. 00:26:26.400 --> 00:26:28.850 Any questions on this lecture? 00:26:35.030 --> 00:26:38.480 This is actually lecture 20 about malleability 00:26:38.480 --> 00:26:41.930 and inaccessibility of preferences. 00:26:41.930 --> 00:26:46.430 This is, in a way, a more radical deviation 00:26:46.430 --> 00:26:48.890 from neoclassical economics. 00:26:48.890 --> 00:26:50.750 Because so far, we have always said, 00:26:50.750 --> 00:26:52.490 people know their preferences. 00:26:52.490 --> 00:26:53.660 People have certain beliefs. 00:26:53.660 --> 00:26:55.610 Their beliefs might be wrong, and so on. 00:26:55.610 --> 00:26:59.180 But crucially, people knew what they were doing, 00:26:59.180 --> 00:27:01.730 and they were deliberately making certain choices. 00:27:01.730 --> 00:27:06.740 And whether their preferences were perhaps 00:27:06.740 --> 00:27:09.530 including present bias, or social preferences, a reference 00:27:09.530 --> 00:27:11.240 dependence or the like-- 00:27:11.240 --> 00:27:13.370 but these preferences were given and fixed. 00:27:13.370 --> 00:27:15.623 And people knew what those preferences are. 00:27:15.623 --> 00:27:17.790 Now we're going to deviate from that and think about 00:27:17.790 --> 00:27:22.970 [INAUDIBLE] evidence that people might not actually A-- 00:27:22.970 --> 00:27:24.740 these preferences might be malleable. 00:27:24.740 --> 00:27:26.990 And B, people might not even understand 00:27:26.990 --> 00:27:28.640 why they want what they want. 00:27:28.640 --> 00:27:30.230 And sort of the preferences are sort 00:27:30.230 --> 00:27:32.120 of quite mysterious to people. 00:27:32.120 --> 00:27:35.540 And it's quite easy in some ways to manipulate people 00:27:35.540 --> 00:27:38.660 in some ways, without them even understanding it, 00:27:38.660 --> 00:27:40.340 knowing about it. 00:27:40.340 --> 00:27:44.142 So I tell you about some very fascinating psychology work, 00:27:44.142 --> 00:27:46.600 which is called-- the paper is called "Telling more than we 00:27:46.600 --> 00:27:48.267 can know" from Nisbett and Wilson, which 00:27:48.267 --> 00:27:50.172 is a classic paper in social psychology. 00:27:50.172 --> 00:27:52.630 And then we're going to talk a little bit about willingness 00:27:52.630 --> 00:27:54.460 to pay, and then the paper by Ariely 00:27:54.460 --> 00:27:58.540 et al on coherent arbitrariness, which you were supposed 00:27:58.540 --> 00:28:00.970 to read for today. 00:28:00.970 --> 00:28:01.660 OK. 00:28:01.660 --> 00:28:04.630 So let's step back a little bit, and talk about the Nisbett 00:28:04.630 --> 00:28:07.220 and Wilson paper. 00:28:07.220 --> 00:28:10.540 So this is-- so many questions-- you 00:28:10.540 --> 00:28:14.560 might have any questions about cognitive processes underlying 00:28:14.560 --> 00:28:17.595 our choices, evaluations, judgments, and behaviors. 00:28:17.595 --> 00:28:18.970 You might wonder, why do you like 00:28:18.970 --> 00:28:20.143 a certain person versus not? 00:28:20.143 --> 00:28:21.310 What do you like about them? 00:28:21.310 --> 00:28:23.602 And why do you like this person and not another person? 00:28:23.602 --> 00:28:26.150 And why you're friends with this person versus somebody else? 00:28:26.150 --> 00:28:27.817 And why do you like Math versus English, 00:28:27.817 --> 00:28:29.770 and so on and so forth? 00:28:29.770 --> 00:28:31.395 How did you solve a certain problem 00:28:31.395 --> 00:28:33.520 once you were sort of asked to solve some problems? 00:28:33.520 --> 00:28:36.162 How did you come up with a solution? 00:28:36.162 --> 00:28:37.870 Why did you take this job, or why did you 00:28:37.870 --> 00:28:39.633 take this class, versus another class? 00:28:39.633 --> 00:28:41.050 And you might sort of say, well, I 00:28:41.050 --> 00:28:42.460 like this one and not that one. 00:28:42.460 --> 00:28:44.660 And I really like sort of like computer science, 00:28:44.660 --> 00:28:47.350 versus math versus economics. 00:28:47.350 --> 00:28:49.950 But when you then ask [INAUDIBLE] yourself 00:28:49.950 --> 00:28:52.810 for others, why exactly do you like that one, versus something 00:28:52.810 --> 00:28:53.440 else? 00:28:53.440 --> 00:28:57.070 You'll realize quickly that people often don't quite 00:28:57.070 --> 00:28:59.330 have a good answer to those questions. 00:28:59.330 --> 00:29:03.100 And Nisbett and Wilson's fairly provocative paper 00:29:03.100 --> 00:29:05.980 says, essentially, we have no idea where these preferences 00:29:05.980 --> 00:29:07.070 are coming from. 00:29:07.070 --> 00:29:08.830 And we're just making stuff up. 00:29:08.830 --> 00:29:11.920 And so let me give you some examples of that. 00:29:11.920 --> 00:29:15.060 The first example is what's called Maier's 00:29:15.060 --> 00:29:17.640 two-string problem from 1931. 00:29:17.640 --> 00:29:21.180 This is [INAUDIBLE] one of these classic psychology experiments. 00:29:21.180 --> 00:29:23.160 The experiments works as follows. 00:29:23.160 --> 00:29:26.280 There's two cords hung from the ceiling 00:29:26.280 --> 00:29:30.990 of a lab with many objects, such as poles, ring stands, clamps, 00:29:30.990 --> 00:29:33.150 pliers, and extension cords. 00:29:33.150 --> 00:29:35.640 Subjects are now asked-- 00:29:35.640 --> 00:29:38.550 are told that a task is to tie the two ends of the cords 00:29:38.550 --> 00:29:39.100 together. 00:29:39.100 --> 00:29:41.400 So there's two cords hanging on the side 00:29:41.400 --> 00:29:44.370 of each side of the room. 00:29:44.370 --> 00:29:48.700 And subjects are supposed to tie these two cords together. 00:29:48.700 --> 00:29:52.290 The problem is that the cords are placed far apart 00:29:52.290 --> 00:29:55.020 from each other, such that subjects 00:29:55.020 --> 00:29:57.943 can't, while holding onto one cord, reach the other. 00:29:57.943 --> 00:29:59.610 So you can't just take one cord and then 00:29:59.610 --> 00:30:01.110 sort of like try to reach the other, 00:30:01.110 --> 00:30:03.300 because it's too far away. 00:30:03.300 --> 00:30:07.080 And so there's all these different objects in the room. 00:30:07.080 --> 00:30:10.210 And subjects are sort of trying to figure out how to do this. 00:30:10.210 --> 00:30:14.547 And so then they usually come up with one or two solutions 00:30:14.547 --> 00:30:17.130 that are not really solutions, because it doesn't really work. 00:30:17.130 --> 00:30:19.590 So they can get [INAUDIBLE] like extension cord, 00:30:19.590 --> 00:30:22.350 they use the pliers, and so on, and the clamp, and some ring 00:30:22.350 --> 00:30:22.900 or whatever. 00:30:22.900 --> 00:30:24.710 And it just doesn't work. 00:30:24.710 --> 00:30:28.880 And then they sort of try it and keep trying, 00:30:28.880 --> 00:30:30.570 but it doesn't really work out. 00:30:30.570 --> 00:30:33.990 And then they're told to do it another way. 00:30:33.990 --> 00:30:36.110 And then Maier, the experimenter, 00:30:36.110 --> 00:30:38.600 walks into the room at some point, or walks around, 00:30:38.600 --> 00:30:40.617 and essentially, accidentally-- this 00:30:40.617 --> 00:30:42.950 is sort of like on purpose of course-- accidentally puts 00:30:42.950 --> 00:30:44.900 some of the cords in motion. 00:30:44.900 --> 00:30:48.230 And subjects then very quickly afterwards figure out 00:30:48.230 --> 00:30:51.050 the solution within the next 45 seconds. 00:30:51.050 --> 00:30:52.490 Of course, the reasoning here is, 00:30:52.490 --> 00:30:54.890 why they figure it out is because now the cords are 00:30:54.890 --> 00:30:55.730 put in motion. 00:30:55.730 --> 00:30:58.535 And subjects sort of figure out how to do this. 00:30:58.535 --> 00:31:00.410 And what you're supposed to do is essentially 00:31:00.410 --> 00:31:01.370 put the cords in motion. 00:31:01.370 --> 00:31:02.870 And once you put them in motion, you 00:31:02.870 --> 00:31:07.380 can essentially reach them both and then put it together. 00:31:07.380 --> 00:31:08.360 I have a video here. 00:31:08.360 --> 00:31:10.160 I'm a little traumatized the last time 00:31:10.160 --> 00:31:11.280 where it didn't work out. 00:31:11.280 --> 00:31:13.730 So I'm not sure [INAUDIBLE]. 00:31:13.730 --> 00:31:14.730 Now I don't even see it. 00:31:14.730 --> 00:31:15.960 So maybe let's just skip it. 00:31:15.960 --> 00:31:17.460 But you can watch the video of that. 00:31:17.460 --> 00:31:20.120 It's the first two and a half minutes video. 00:31:20.120 --> 00:31:21.390 [INAUDIBLE] see anything. 00:31:21.390 --> 00:31:25.410 But there's a video that sort of demonstrates this task that you 00:31:25.410 --> 00:31:28.030 can sort of just skip this. 00:31:28.030 --> 00:31:31.920 So now then afterwards, people are asked, well, 00:31:31.920 --> 00:31:36.780 how did you come up with the idea of using a pendulum? 00:31:36.780 --> 00:31:39.540 And people then have all these like explanations. 00:31:39.540 --> 00:31:40.980 Well, it just dawned on me. 00:31:40.980 --> 00:31:42.738 It was the only thing left. 00:31:42.738 --> 00:31:44.530 I thought about all these different things. 00:31:44.530 --> 00:31:46.655 And at the end, I came up with this other solution, 00:31:46.655 --> 00:31:48.570 which is using these strings as a pendulum. 00:31:48.570 --> 00:31:51.112 Or I just realized the cord would swing 00:31:51.112 --> 00:31:52.320 [INAUDIBLE] the weight on it. 00:31:52.320 --> 00:31:54.067 So what the solution at the end is like, 00:31:54.067 --> 00:31:56.400 you put the plier on one of the cords and then swing it. 00:31:56.400 --> 00:31:57.630 And then you sort of use the other one. 00:31:57.630 --> 00:31:59.430 And then you could put them together. 00:31:59.430 --> 00:32:02.520 There's one Harvard psychology faculty subject, in fact, 00:32:02.520 --> 00:32:04.020 had the following explanation, which 00:32:04.020 --> 00:32:06.320 is, having exhausted everything else, 00:32:06.320 --> 00:32:07.650 the next thing was to swing it. 00:32:07.650 --> 00:32:10.670 I thought of the situation of swinging across a river. 00:32:10.670 --> 00:32:13.560 I had an imagery of monkeys swinging from trees. 00:32:13.560 --> 00:32:17.410 This imagery appeared simultaneously with a solution. 00:32:17.410 --> 00:32:20.100 The idea appeared complete. 00:32:20.100 --> 00:32:22.500 Now of course, the problem here is that-- 00:32:25.050 --> 00:32:27.510 the problem here is that's all bullshit. 00:32:27.510 --> 00:32:30.720 Because the reason why people come up with a solution 00:32:30.720 --> 00:32:35.790 is because, as I told you, Maier was walking in and accidentally 00:32:35.790 --> 00:32:37.110 putting the cords in motion. 00:32:37.110 --> 00:32:39.990 And there's differences in timing. 00:32:39.990 --> 00:32:42.390 Sometimes they would come in earlier, sometimes later. 00:32:42.390 --> 00:32:45.360 And every time the experimenter walks in, 00:32:45.360 --> 00:32:46.900 people would sort of accidentally 00:32:46.900 --> 00:32:48.840 then figure it out, and then come up 00:32:48.840 --> 00:32:51.360 with these elaborate explanations. 00:32:51.360 --> 00:32:53.160 And the reason why we know that this 00:32:53.160 --> 00:32:56.820 is because of the experimenter are coming in 00:32:56.820 --> 00:32:59.790 is because if the experimenter doesn't accidentally come in, 00:32:59.790 --> 00:33:02.880 and puts the cords in motion, then essentially, 00:33:02.880 --> 00:33:04.580 people are not figuring out. 00:33:04.580 --> 00:33:07.290 So we know the true causal effect here 00:33:07.290 --> 00:33:09.840 is the experimenter coming in and suggesting 00:33:09.840 --> 00:33:11.340 a solution to people. 00:33:11.340 --> 00:33:14.940 But people then make up all these stories, 00:33:14.940 --> 00:33:16.410 including like the monkeys swinging 00:33:16.410 --> 00:33:18.990 from trees, how they came up with a solution, which 00:33:18.990 --> 00:33:21.340 is clearly not how they came up with the solution. 00:33:21.340 --> 00:33:23.215 The reason why they came up with the solution 00:33:23.215 --> 00:33:26.980 is because Maier touched the strings. 00:33:26.980 --> 00:33:27.790 OK. 00:33:27.790 --> 00:33:30.010 Now there's a second example here, 00:33:30.010 --> 00:33:31.760 or three examples in total. 00:33:31.760 --> 00:33:33.490 So let me tell you about example number-- 00:33:33.490 --> 00:33:38.910 or let me ask first, is there any questions about this study? 00:33:38.910 --> 00:33:41.990 By the way, you should lower your hands, 00:33:41.990 --> 00:33:43.670 if you asked [INAUDIBLE] previously, 00:33:43.670 --> 00:33:46.770 like [? Brian ?] and [INAUDIBLE].. 00:33:46.770 --> 00:33:48.610 Unless you have have additional questions. 00:33:53.510 --> 00:33:57.380 Let me tell you the second study now. 00:33:57.380 --> 00:34:01.460 This is a study by Latane and Darley, 00:34:01.460 --> 00:34:04.400 which is, I told you this a little bit 00:34:04.400 --> 00:34:08.150 about [INAUDIBLE] these are the impacts of bystanders 00:34:08.150 --> 00:34:10.460 and witnesses on helping behaviors. 00:34:10.460 --> 00:34:12.989 I told you about the story of the Good Samaritan before. 00:34:12.989 --> 00:34:15.670 This is sort of a similar study. 00:34:15.670 --> 00:34:17.980 And this is essentially a situation 00:34:17.980 --> 00:34:20.260 that's, again, created by social psychologists. 00:34:20.260 --> 00:34:22.239 It's not a real situation, but people are 00:34:22.239 --> 00:34:24.409 meant to think that it's real. 00:34:24.409 --> 00:34:27.550 And so here, the more people over here, someone 00:34:27.550 --> 00:34:29.139 in another room, having what sounds 00:34:29.139 --> 00:34:32.440 like an epileptic seizure, the lower the probability 00:34:32.440 --> 00:34:35.280 that any given individual will rush to help. 00:34:35.280 --> 00:34:38.699 And you get similar results for individuals' reaction 00:34:38.699 --> 00:34:40.358 to dangerous-looking smoke coming out 00:34:40.358 --> 00:34:41.400 of the ceiling of a room. 00:34:41.400 --> 00:34:43.108 So like there's people in the experiment. 00:34:43.108 --> 00:34:45.630 And they have smoke coming out in a room. 00:34:45.630 --> 00:34:48.060 And the more people that are there, the less likely 00:34:48.060 --> 00:34:50.520 is any given person, perhaps because they're freeriding 00:34:50.520 --> 00:34:53.620 is to actually do something about it. 00:34:53.620 --> 00:34:55.425 And this looks kind of very dangerous. 00:34:55.425 --> 00:34:57.300 But nobody does anything, because essentially 00:34:57.300 --> 00:34:59.550 other people are also not doing anything. 00:34:59.550 --> 00:35:01.470 Now when you then ask people afterwards, 00:35:01.470 --> 00:35:05.430 you know, what influenced your choices? 00:35:05.430 --> 00:35:10.290 And essentially, when you ask them directly and tactfully 00:35:10.290 --> 00:35:13.540 and bluntly, you always get the same answer, that essentially, 00:35:13.540 --> 00:35:15.900 they think their behavior had not been influenced 00:35:15.900 --> 00:35:17.355 by the other people present. 00:35:17.355 --> 00:35:19.230 They think you're like-- for whatever reason, 00:35:19.230 --> 00:35:21.157 they didn't do anything, or maybe they 00:35:21.157 --> 00:35:23.490 thought it wasn't so bad, or there wasn't really danger, 00:35:23.490 --> 00:35:26.050 or whatever. 00:35:26.050 --> 00:35:28.280 But of course, these are randomized experiments. 00:35:28.280 --> 00:35:30.030 We know that some subjects were influenced 00:35:30.030 --> 00:35:32.155 by the presence of other people, because precisely, 00:35:32.155 --> 00:35:33.960 the experiment was creating that variation, 00:35:33.960 --> 00:35:36.090 and it affected people's behavior. 00:35:36.090 --> 00:35:39.060 [INAUDIBLE] people essentially just don't understand that, 00:35:39.060 --> 00:35:41.610 and sort of then come up with other explanations 00:35:41.610 --> 00:35:46.990 of why they do or did what they, in fact, did. 00:35:46.990 --> 00:35:49.690 And number three is sort of very similar. 00:35:49.690 --> 00:35:54.890 These are called erroneous reports about position effects. 00:35:54.890 --> 00:35:58.840 These are studies where certain items 00:35:58.840 --> 00:36:04.330 are positioned in different ways, to get people to-- 00:36:04.330 --> 00:36:06.680 when people were evaluating them. 00:36:06.680 --> 00:36:12.310 So passerbys were asked to evaluate some clothing. 00:36:12.310 --> 00:36:16.630 And they were asked about the quality and the preferences-- 00:36:16.630 --> 00:36:18.460 the quality of certain goods and people's 00:36:18.460 --> 00:36:20.760 preferences between different goods. 00:36:20.760 --> 00:36:23.042 And the way this experiment was set up-- 00:36:23.042 --> 00:36:25.500 and this is sort of essentially like a marketing experiment 00:36:25.500 --> 00:36:26.280 if you want-- 00:36:26.280 --> 00:36:29.910 essentially, there's a pronounced left-to-right 00:36:29.910 --> 00:36:30.990 position effect. 00:36:30.990 --> 00:36:33.510 The rightmost object was heavily over-chosen 00:36:33.510 --> 00:36:34.600 in these experiments. 00:36:34.600 --> 00:36:36.990 And that's in some sense irrelevant why that's the case. 00:36:36.990 --> 00:36:39.540 But essentially, from previous experiments, 00:36:39.540 --> 00:36:41.790 we know that essentially the rightmost object 00:36:41.790 --> 00:36:45.300 is most likely to be chosen, perhaps because people 00:36:45.300 --> 00:36:46.320 look at it first. 00:36:46.320 --> 00:36:49.305 [INAUDIBLE] look at it last and so on, it doesn't matter. 00:36:49.305 --> 00:36:50.680 What's important for our purposes 00:36:50.680 --> 00:36:53.700 is that the rightmost object is most heavily chosen. 00:36:53.700 --> 00:36:56.880 People like that object the best in these experiments. 00:36:56.880 --> 00:36:59.190 And then there's going to be randomization 00:36:59.190 --> 00:37:02.070 in the positioning of experiments. 00:37:02.070 --> 00:37:03.630 And people are then asked, well, why 00:37:03.630 --> 00:37:06.480 did you choose this sweater versus another? 00:37:06.480 --> 00:37:09.810 And no subject whatsoever ever mentions 00:37:09.810 --> 00:37:14.040 the position of the article in the room, 00:37:14.040 --> 00:37:15.720 and virtually all subjects denied 00:37:15.720 --> 00:37:18.270 it was, when asked directly about the possible effect 00:37:18.270 --> 00:37:19.802 of the position of the article. 00:37:19.802 --> 00:37:21.760 And people come up with all these explanations. 00:37:21.760 --> 00:37:24.030 It's like, green is really always my [INAUDIBLE] favorite 00:37:24.030 --> 00:37:24.540 color. 00:37:24.540 --> 00:37:26.250 And this sweater is really fluffy. 00:37:26.250 --> 00:37:27.000 And this and that. 00:37:27.000 --> 00:37:30.580 And I really like it for reason x and y. 00:37:30.580 --> 00:37:33.090 Of course that's true for some people, 00:37:33.090 --> 00:37:34.860 but we know that at least some people must 00:37:34.860 --> 00:37:36.390 be swayed by these position effects, 00:37:36.390 --> 00:37:39.302 because we set up an experiment to precisely do that. 00:37:39.302 --> 00:37:41.010 And then that essentially swayed behavior 00:37:41.010 --> 00:37:42.880 in a very predictable way. 00:37:42.880 --> 00:37:44.790 So essentially, there are these determinants 00:37:44.790 --> 00:37:46.800 on people's preferences of behavior, 00:37:46.800 --> 00:37:51.500 that people do not seem to understand. 00:37:51.500 --> 00:37:52.010 OK. 00:37:52.010 --> 00:37:54.527 So let me summarize what did we learn. 00:37:54.527 --> 00:37:57.110 So there are many instances in which subjects have no idea why 00:37:57.110 --> 00:37:59.130 they choose what they choose. 00:37:59.130 --> 00:38:00.455 And then people-- 00:38:00.455 --> 00:38:02.520 [INAUDIBLE] can read this more in this paper, 00:38:02.520 --> 00:38:05.010 which I think is a beautiful paper-- 00:38:05.010 --> 00:38:07.940 people appear to make up stories that 00:38:07.940 --> 00:38:11.450 are based on their a priori, implicit causal theories. 00:38:11.450 --> 00:38:13.160 What I mean by that, essentially, they 00:38:13.160 --> 00:38:16.770 have some theory of why they do certain things. 00:38:16.770 --> 00:38:18.860 And when you ask them, then, OK, well, why 00:38:18.860 --> 00:38:20.420 did you do what you just did? 00:38:20.420 --> 00:38:22.190 People will sort of come up with some ways 00:38:22.190 --> 00:38:24.980 of justifying [INAUDIBLE] in some ways 00:38:24.980 --> 00:38:27.020 their behavior, based on some theories 00:38:27.020 --> 00:38:28.850 that they had about themselves. 00:38:28.850 --> 00:38:29.735 And sort of-- 00:38:29.735 --> 00:38:31.820 I always like fluffy sweaters, and therefore you 00:38:31.820 --> 00:38:32.820 choose the sweater. 00:38:32.820 --> 00:38:34.820 Of course, it happens to be the only [INAUDIBLE] 00:38:34.820 --> 00:38:38.270 sweater if it's [INAUDIBLE] on the right versus on the left. 00:38:38.270 --> 00:38:40.790 But people then sort of have their ways in which they then 00:38:40.790 --> 00:38:44.510 explain what they do based on essentially things that they 00:38:44.510 --> 00:38:47.950 sort of essentially make up. 00:38:47.950 --> 00:38:50.831 Any questions on this before I move on? 00:38:57.900 --> 00:39:02.010 And I encourage you to read the paper, more for fun 00:39:02.010 --> 00:39:03.265 than for anything. 00:39:03.265 --> 00:39:04.890 It makes you kind of wonder quite a lot 00:39:04.890 --> 00:39:07.680 in why you do what you do, and why you prefer 00:39:07.680 --> 00:39:11.540 certain things versus others. 00:39:11.540 --> 00:39:14.180 So let's briefly talk about two experimental design 00:39:14.180 --> 00:39:17.630 tools, which will be useful for the coherent arbitrariness 00:39:17.630 --> 00:39:18.642 paper. 00:39:18.642 --> 00:39:20.600 One is what's called the strategy method, which 00:39:20.600 --> 00:39:22.292 you already came across a little bit. 00:39:22.292 --> 00:39:23.750 And the second one is what's called 00:39:23.750 --> 00:39:26.750 the BDM, the Becker-DeGroot-Marschak 00:39:26.750 --> 00:39:28.850 procedure for eliciting valuations. 00:39:28.850 --> 00:39:31.010 I'll try to be quick. 00:39:31.010 --> 00:39:38.240 So we're often interested in behavior in rare contingencies. 00:39:38.240 --> 00:39:40.220 So often, we may ask the question, 00:39:40.220 --> 00:39:43.370 how would people behave in many different contingencies? 00:39:43.370 --> 00:39:44.840 Some of them are often quite rare, 00:39:44.840 --> 00:39:47.730 that don't happen very often. 00:39:47.730 --> 00:39:50.070 And why do we care about such contingencies? 00:39:50.070 --> 00:39:51.780 Well, sometimes we just care about them 00:39:51.780 --> 00:39:55.200 like [INAUDIBLE] it's inherently important 00:39:55.200 --> 00:39:59.640 how people behave or prevent certain contingencies, 00:39:59.640 --> 00:40:01.995 disasters, earthquakes, droughts, et cetera. 00:40:01.995 --> 00:40:03.370 We care a lot about these things, 00:40:03.370 --> 00:40:04.890 even if they're sort of rarely happening. 00:40:04.890 --> 00:40:06.432 So we kind of want to know how people 00:40:06.432 --> 00:40:09.510 behave in certain situations. 00:40:09.510 --> 00:40:10.980 But events and rare contingencies 00:40:10.980 --> 00:40:15.390 also can affect, on top of that, events in likely contingencies. 00:40:15.390 --> 00:40:16.660 Here's a simple example. 00:40:16.660 --> 00:40:19.173 If your roommates think you'll punch them in the face 00:40:19.173 --> 00:40:21.090 if they borrow your stuff without asking them, 00:40:21.090 --> 00:40:23.130 they will not do it. 00:40:23.130 --> 00:40:26.310 Of course, hopefully, you are not punching your roommates, 00:40:26.310 --> 00:40:30.010 and I very much do not want to encourage you to do so. 00:40:30.010 --> 00:40:33.270 But here, the key part here is that punching 00:40:33.270 --> 00:40:38.790 is rare but important, precisely because you might do so in case 00:40:38.790 --> 00:40:41.910 they misbehave, they will not borrow your stuff 00:40:41.910 --> 00:40:44.320 or steal your stuff in the first place. 00:40:44.320 --> 00:40:46.410 So essentially, sort of these [? equilibrium ?] 00:40:46.410 --> 00:40:49.512 or rare contingencies might be quite important, 00:40:49.512 --> 00:40:51.720 not because we think they actually happen very often, 00:40:51.720 --> 00:40:56.540 but they sort of discipline behavior in other cases. 00:40:56.540 --> 00:40:58.970 Now, what's [? known ?] as strategy method? 00:40:58.970 --> 00:41:02.480 Well, the strategy method helps you 00:41:02.480 --> 00:41:06.890 to elicit behavior in many potentially rare circumstances 00:41:06.890 --> 00:41:08.780 by asking subjects what they would 00:41:08.780 --> 00:41:13.290 do with a choice implemented if the circumstances arises. 00:41:13.290 --> 00:41:16.160 That is to say, I'm asking you-- for many different cases, 00:41:16.160 --> 00:41:18.990 I'm going to implement one of those cases. 00:41:18.990 --> 00:41:24.320 And that's going to allow me to be [INAUDIBLE] [? compatible ?] 00:41:24.320 --> 00:41:25.860 in various ways. 00:41:25.860 --> 00:41:30.870 So since the decision does count if the contingency occurs, 00:41:30.870 --> 00:41:33.180 subjects have an incentive to choose correctly 00:41:33.180 --> 00:41:35.100 for each contingency. 00:41:35.100 --> 00:41:39.660 That is to say, I could say, suppose this is what we're 00:41:39.660 --> 00:41:41.580 doing in class in some sense. 00:41:41.580 --> 00:41:43.140 We talked about social preferences. 00:41:43.140 --> 00:41:46.080 I said, you know, I'm asking you to make a choice. 00:41:46.080 --> 00:41:50.140 I'm going to only pick one of you to implement that choice. 00:41:50.140 --> 00:41:57.145 And you have the incentives now to answer truthfully, 00:41:57.145 --> 00:41:59.520 because it could always be the case that your choice will 00:41:59.520 --> 00:42:01.420 be implemented. 00:42:01.420 --> 00:42:05.190 And so that allows then the experimenter, me in that case, 00:42:05.190 --> 00:42:09.030 to generate a lot of data from one simple experiment, where 00:42:09.030 --> 00:42:10.800 essentially, you can give a subject 00:42:10.800 --> 00:42:12.900 many different decisions, and then say, 00:42:12.900 --> 00:42:14.760 only one of your choices will actually 00:42:14.760 --> 00:42:16.020 be implemented for sure. 00:42:16.020 --> 00:42:18.400 Or you can ask many people the same question and say, 00:42:18.400 --> 00:42:21.630 [INAUDIBLE] only one of you, or one of your choices 00:42:21.630 --> 00:42:23.140 will be implemented. 00:42:23.140 --> 00:42:27.480 And so then that's essentially incentive-compatible, 00:42:27.480 --> 00:42:30.420 because it can always be that your choice is the one, or one 00:42:30.420 --> 00:42:32.850 of the choices is the one that counts. 00:42:32.850 --> 00:42:35.680 And when you look at sort of experimental evidence, 00:42:35.680 --> 00:42:38.370 it seems to suggest that the strategy methods are asking, 00:42:38.370 --> 00:42:41.372 essentially, if I can ask you-- so there's experiments where 00:42:41.372 --> 00:42:43.830 I can ask you 100 different questions, and only one of them 00:42:43.830 --> 00:42:46.650 will be implemented, versus a different group would 00:42:46.650 --> 00:42:49.110 be asked-- a randomized group would be asked only one 00:42:49.110 --> 00:42:49.630 question. 00:42:49.630 --> 00:42:52.350 It turns out that people actually answer in experiments 00:42:52.350 --> 00:42:57.250 these questions pretty similarly in both of these cases. 00:42:57.250 --> 00:42:59.160 So that's to say the strategy method seems 00:42:59.160 --> 00:43:04.424 to elicit individuals' true preferences pretty well. 00:43:04.424 --> 00:43:05.103 Sorry. 00:43:05.103 --> 00:43:06.020 [INAUDIBLE] messed up. 00:43:06.020 --> 00:43:07.252 But I think that's fine. 00:43:07.252 --> 00:43:08.960 So what's now the Becker-DeGroot-Marschak 00:43:08.960 --> 00:43:09.620 procedure? 00:43:09.620 --> 00:43:11.910 It's essentially a version of that, 00:43:11.910 --> 00:43:17.825 which looks something like this, which is, people are asked-- 00:43:17.825 --> 00:43:19.200 [INAUDIBLE] so what the goal here 00:43:19.200 --> 00:43:22.523 is to understand people's willingness to pay for a good. 00:43:22.523 --> 00:43:24.440 So subjects are told that a price for the good 00:43:24.440 --> 00:43:25.730 will be randomly selected. 00:43:25.730 --> 00:43:29.000 And the price goes from like $0.50 to like, say, $10. 00:43:29.000 --> 00:43:31.400 For each of these prices, the person is asked, 00:43:31.400 --> 00:43:35.780 do you prefer buy versus not-buy in these options. 00:43:35.780 --> 00:43:38.943 And so then, I ask you to fill out the entire form. 00:43:38.943 --> 00:43:41.360 Tell me, for each of these prices, what do you want to do? 00:43:41.360 --> 00:43:43.070 If the price is $3, what do you want? 00:43:43.070 --> 00:43:44.690 If the price is $1, what do you want? 00:43:44.690 --> 00:43:47.030 If the price is $10, what do you want? 00:43:47.030 --> 00:43:50.870 And then I'm going to afterwards then pick a price and say, OK. 00:43:50.870 --> 00:43:52.378 Now the price is $3.50. 00:43:52.378 --> 00:43:54.920 And now I'm going to just look at like, what did you actually 00:43:54.920 --> 00:43:56.870 say, and whatever your choice was then 00:43:56.870 --> 00:44:03.440 for that specific price is going to be implemented. 00:44:03.440 --> 00:44:06.180 Is that clear? 00:44:06.180 --> 00:44:09.270 Now, just to be clear what the problem is that we're 00:44:09.270 --> 00:44:12.930 trying to overcome here-- 00:44:12.930 --> 00:44:17.210 suppose I'm trying to sell you a mug. 00:44:17.210 --> 00:44:18.758 Or suppose I have this mug here. 00:44:18.758 --> 00:44:20.550 I'm saying, would you like to buy this mug? 00:44:20.550 --> 00:44:23.220 It's a very beautiful mug with lots of trees on them. 00:44:23.220 --> 00:44:24.720 I'd like to sell it to you. 00:44:24.720 --> 00:44:27.950 And I'm eliciting your willingness to pay. 00:44:27.950 --> 00:44:30.470 What's the problem that arises from just 00:44:30.470 --> 00:44:33.240 asking you this directly, without the BDM procedure? 00:44:33.240 --> 00:44:37.548 What problem-- what is the BDM procedure sort of helping with? 00:44:37.548 --> 00:44:39.090 So if you go to a marketplace, or try 00:44:39.090 --> 00:44:41.820 to bargain with me on this mug, which again, 00:44:41.820 --> 00:44:45.210 it's a beautiful mug-- 00:44:45.210 --> 00:44:47.850 you might want to shade your valuation. 00:44:47.850 --> 00:44:50.640 You might be willing to pay $5, but you're 00:44:50.640 --> 00:44:54.030 going to say $2, because you are hoping to get a bargain 00:44:54.030 --> 00:44:56.340 and get [INAUDIBLE] a cheap price from me. 00:44:56.340 --> 00:44:57.090 Right? 00:44:57.090 --> 00:44:59.430 And the key part here is that you are essentially 00:44:59.430 --> 00:45:03.900 hoping that your willingness to pay whatever you are offering 00:45:03.900 --> 00:45:07.380 me will essentially change whatever price I'm 00:45:07.380 --> 00:45:09.210 offering the mug to you. 00:45:09.210 --> 00:45:12.240 Notice that that's not the case here in the BDM procedure. 00:45:12.240 --> 00:45:15.600 In the BDM procedure, whatever you say 00:45:15.600 --> 00:45:19.103 is independent of the actual price that's implemented. 00:45:19.103 --> 00:45:20.520 I'm saying, essentially, I'm going 00:45:20.520 --> 00:45:23.490 to randomly pick one of those 10 or 20 prices here. 00:45:23.490 --> 00:45:26.700 And you're going to tell me, for each of those prices, what you 00:45:26.700 --> 00:45:28.710 want, either buy or not buy. 00:45:28.710 --> 00:45:31.538 And your choices are-- since I'm randomizing afterwards, 00:45:31.538 --> 00:45:32.955 which price selection you're going 00:45:32.955 --> 00:45:38.810 to pick, your choices are irrelevant for which 00:45:38.810 --> 00:45:43.247 of the actual prices are selected. 00:45:43.247 --> 00:45:45.080 And so now essentially get around this issue 00:45:45.080 --> 00:45:47.480 about people shading or essentially underreporting 00:45:47.480 --> 00:45:49.220 their willingness to pay. 00:45:49.220 --> 00:45:52.130 Precisely that's what makes it incentive-compatible. 00:45:52.130 --> 00:45:54.770 And that's essentially what's called the BDM mechanism. 00:45:54.770 --> 00:45:59.155 That's what many economists use in many, many experiments. 00:45:59.155 --> 00:46:00.530 There's another version of what's 00:46:00.530 --> 00:46:07.102 called the BDM which is a more straightforward version of it. 00:46:07.102 --> 00:46:08.810 Notice here you have to ask 20 questions. 00:46:08.810 --> 00:46:10.977 Ask would you like to pay-- would you like to buy it 00:46:10.977 --> 00:46:14.910 for $0.50, $1.00, $1.50, $2.00, and so on and so forth. 00:46:14.910 --> 00:46:20.210 Another version of this is, I'm asking you straight up what's 00:46:20.210 --> 00:46:21.770 your willingness to pay. 00:46:21.770 --> 00:46:25.100 And the bid is then compared to a price determined 00:46:25.100 --> 00:46:26.750 by a random number generator. 00:46:26.750 --> 00:46:28.540 I'm just saying, like, tell me what your willingness to pay 00:46:28.540 --> 00:46:29.210 is. 00:46:29.210 --> 00:46:32.220 Then I'm going to do like a random number generator. 00:46:32.220 --> 00:46:35.030 If the subject's bid is greater than the price that's 00:46:35.030 --> 00:46:37.670 generated by the random number generator, 00:46:37.670 --> 00:46:40.430 he or she pays the price, not the announced willingness 00:46:40.430 --> 00:46:43.430 to pay, and receives the item being auctioned. 00:46:43.430 --> 00:46:45.470 If the subject's bid is lower than the price, 00:46:45.470 --> 00:46:48.740 he or she pays nothing and receives nothing. 00:46:48.740 --> 00:46:53.210 And so now here again, the final price the person must pay 00:46:53.210 --> 00:46:55.670 is independent of what the person indicated 00:46:55.670 --> 00:46:57.380 as your willingness to pay, which 00:46:57.380 --> 00:47:00.265 essentially solves the incentive compatibility issue. 00:47:00.265 --> 00:47:01.640 Think of like this version that I 00:47:01.640 --> 00:47:04.670 have on this line as just a more efficient way of eliciting 00:47:04.670 --> 00:47:06.720 the question that's on the previous slide. 00:47:06.720 --> 00:47:08.960 This is essentially exactly the same thing, 00:47:08.960 --> 00:47:11.370 except for that it's a more efficient thing to do. 00:47:11.370 --> 00:47:13.700 The problem with this version, often, is comprehension. 00:47:13.700 --> 00:47:14.533 People get confused. 00:47:14.533 --> 00:47:17.120 Or people tend to still try to bargain and give 00:47:17.120 --> 00:47:18.780 like lower willingness to pay. 00:47:18.780 --> 00:47:21.308 So sometimes, this version works better 00:47:21.308 --> 00:47:23.100 because it's much easier, much transparent, 00:47:23.100 --> 00:47:25.350 when we say, here's a price, do want to buy it or not, 00:47:25.350 --> 00:47:27.200 versus and so on. 00:47:27.200 --> 00:47:29.822 While here it's people are sort of [INAUDIBLE] because they're 00:47:29.822 --> 00:47:31.280 used to, essentially, bargain, they 00:47:31.280 --> 00:47:35.350 tend to sort of under-report at least sometimes. 00:47:35.350 --> 00:47:36.430 Any questions on this? 00:47:45.920 --> 00:47:49.940 So now I'm also going to skip this video. 00:47:49.940 --> 00:47:50.805 It's a fun video. 00:47:50.805 --> 00:47:52.805 Who knows the story of Tom Sawyer and the fence? 00:47:59.130 --> 00:48:02.250 I think it's in the readings, no? 00:48:02.250 --> 00:48:04.180 Tom Sawyer was misbehaving. 00:48:04.180 --> 00:48:06.880 He was essentially punished to paint the fence. 00:48:06.880 --> 00:48:08.750 He really didn't want to do it. 00:48:08.750 --> 00:48:10.980 But then his friend comes along, and essentially, 00:48:10.980 --> 00:48:12.730 he sort of tricks his friend into thinking 00:48:12.730 --> 00:48:15.115 this is really like such a fun task, 00:48:15.115 --> 00:48:16.870 and it ends up being essentially, 00:48:16.870 --> 00:48:21.190 the friend even paying him for painting the fence, 00:48:21.190 --> 00:48:23.380 and he not having to do it. 00:48:23.380 --> 00:48:25.040 And the point of the story, of course, 00:48:25.040 --> 00:48:30.080 is that people's willingness to pay is malleable in the sense, 00:48:30.080 --> 00:48:33.340 like, even to the extent that people don't even 00:48:33.340 --> 00:48:35.500 know whether they're willing to pay for something, 00:48:35.500 --> 00:48:39.140 or you have to pay them something to do it. 00:48:39.140 --> 00:48:41.140 And so essentially, people's preferences 00:48:41.140 --> 00:48:46.180 are inherently malleable by the way they're being marketed, 00:48:46.180 --> 00:48:49.732 or the way they appear to them, or being sold to them. 00:48:49.732 --> 00:48:53.520 You can watch the video in the slides if you download it. 00:48:53.520 --> 00:48:56.340 For whatever reason this is not working on Zoom, 00:48:56.340 --> 00:48:57.730 at least for now. 00:48:57.730 --> 00:49:01.940 So stepping back a little bit-- so overall, 00:49:01.940 --> 00:49:03.990 so far we talked about two key components 00:49:03.990 --> 00:49:05.250 of individual decision making. 00:49:05.250 --> 00:49:08.340 Utility functions-- what people want and what they care about, 00:49:08.340 --> 00:49:10.350 and beliefs-- how people perceive themselves 00:49:10.350 --> 00:49:12.250 and patterns in the world. 00:49:12.250 --> 00:49:13.960 And so understanding these, of course, 00:49:13.960 --> 00:49:17.580 is important because that determines people's choices. 00:49:17.580 --> 00:49:19.770 Now so far, we have always sort of pretended 00:49:19.770 --> 00:49:22.740 and said people's preferences and beliefs are always 00:49:22.740 --> 00:49:23.670 sharply-- 00:49:23.670 --> 00:49:26.400 people are always sharply aware of what they want and believe 00:49:26.400 --> 00:49:28.770 that costs [INAUDIBLE]. 00:49:28.770 --> 00:49:30.790 For example, a homeowner might have 00:49:30.790 --> 00:49:32.760 reference-dependent preferences, but they know 00:49:32.760 --> 00:49:34.907 what their preferences are. 00:49:34.907 --> 00:49:36.990 There might be some issues about how the reference 00:49:36.990 --> 00:49:39.600 point is determined, but the preference is the function. 00:49:39.600 --> 00:49:42.467 The functional form is always fixed, and it's known. 00:49:42.467 --> 00:49:44.550 A person might have the wrong theory of the world, 00:49:44.550 --> 00:49:46.860 but she always has some beliefs in mind, what 00:49:46.860 --> 00:49:49.020 she uses to make those choices. 00:49:49.020 --> 00:49:51.420 Or a smoker might act suboptimally, 00:49:51.420 --> 00:49:53.730 but he always has a fully specific strategy 00:49:53.730 --> 00:49:56.400 in mind for all his current and future decisions. 00:49:56.400 --> 00:49:58.020 So people-- the smoker might sort of 00:49:58.020 --> 00:49:59.940 have some issues with present bias, 00:49:59.940 --> 00:50:02.400 or some wrong beliefs in some ways or the other. 00:50:02.400 --> 00:50:04.440 But the utility function was always 00:50:04.440 --> 00:50:08.160 assumed to be fixed and known. 00:50:08.160 --> 00:50:10.510 Now we're going to deviate from that. 00:50:10.510 --> 00:50:14.610 If you think about it like, what are your preferences, 00:50:14.610 --> 00:50:16.980 and what do you like and what do you not like-- 00:50:16.980 --> 00:50:20.610 in many cases, we actually don't know. 00:50:20.610 --> 00:50:23.640 And in many cases, people sort of just 00:50:23.640 --> 00:50:26.560 make things up as they go along. 00:50:26.560 --> 00:50:29.090 So that's to say, like, people are essentially 00:50:29.090 --> 00:50:30.630 making some choices. 00:50:30.630 --> 00:50:32.795 And this one of the examples that I showed you 00:50:32.795 --> 00:50:34.170 from earlier-- the three examples 00:50:34.170 --> 00:50:36.780 that I showed you first, where people are just 00:50:36.780 --> 00:50:39.127 induced to make certain choices one way or the other, 00:50:39.127 --> 00:50:40.710 depending on their choice environment, 00:50:40.710 --> 00:50:42.210 depending on their social influence. 00:50:42.210 --> 00:50:45.480 In Tom Sawyer's case, what he tells his friend. 00:50:45.480 --> 00:50:48.420 His friend is now willing to pay to paint a fence, which 00:50:48.420 --> 00:50:49.840 actually is a punishment. 00:50:49.840 --> 00:50:54.730 So people's preferences seem inherently malleable. 00:50:54.730 --> 00:50:58.890 And so that's what this paper about coherent arbitrariness 00:50:58.890 --> 00:51:01.180 is very much about. 00:51:01.180 --> 00:51:07.050 And so if you think about it, in some ways, almost all 00:51:07.050 --> 00:51:11.730 of economics, and any class that you have taken, the professor-- 00:51:11.730 --> 00:51:15.540 or in any book, people will write down the utility function 00:51:15.540 --> 00:51:17.980 as if that's some truth that we know about the world. 00:51:17.980 --> 00:51:20.610 And this is what the utility function looks like. 00:51:20.610 --> 00:51:22.658 But if you think about it, in many cases, 00:51:22.658 --> 00:51:24.450 actually, we don't know what is the utility 00:51:24.450 --> 00:51:24.930 function [INAUDIBLE]. 00:51:24.930 --> 00:51:26.597 If somebody asked [INAUDIBLE] particular 00:51:26.597 --> 00:51:30.240 about new goods, who knows what our preferences actually are. 00:51:30.240 --> 00:51:33.700 So for example, if you say, you're trying to buy a monitor, 00:51:33.700 --> 00:51:36.120 and there's like a 30 inch monitor versus a 24 inch 00:51:36.120 --> 00:51:40.140 monitor, who knows how much you're willing to pay for that. 00:51:40.140 --> 00:51:43.770 Maybe $50, maybe $100, maybe $20, maybe nothing. 00:51:43.770 --> 00:51:47.460 It's very hard to sort of figure this out. 00:51:47.460 --> 00:51:50.190 And sort of when asked to make these kinds of decisions, 00:51:50.190 --> 00:51:52.877 people tend to sort of construct their preferences 00:51:52.877 --> 00:51:53.460 on this stuff. 00:51:53.460 --> 00:51:56.010 They essentially make stuff up. 00:51:56.010 --> 00:51:59.167 And so because of that, because people 00:51:59.167 --> 00:52:00.750 are sort of fundamentally in some ways 00:52:00.750 --> 00:52:04.980 unsure about their preferences, this construction 00:52:04.980 --> 00:52:08.070 of their valuation is very much easily manipulated 00:52:08.070 --> 00:52:12.600 by, often, cues that should be really irrelevant. 00:52:12.600 --> 00:52:16.470 And so one example that I showed you a little bit-- 00:52:16.470 --> 00:52:19.560 I think in the second class, or from the survey, 00:52:19.560 --> 00:52:20.910 is what's called anchoring. 00:52:20.910 --> 00:52:23.795 This is also in the Ariely paper, which essentially 00:52:23.795 --> 00:52:25.170 is, what they were doing is, they 00:52:25.170 --> 00:52:30.420 were asking people about their willingness 00:52:30.420 --> 00:52:32.460 to [INAUDIBLE] their Social Security number. 00:52:32.460 --> 00:52:34.043 And then afterwards, their willingness 00:52:34.043 --> 00:52:38.310 to pay for different things, about like, wine, design books, 00:52:38.310 --> 00:52:39.735 chocolates, and so on. 00:52:39.735 --> 00:52:41.610 Notice that these are somewhat unusual things 00:52:41.610 --> 00:52:44.070 that people wouldn't necessarily buy every day. 00:52:44.070 --> 00:52:45.930 And essentially what you see is people 00:52:45.930 --> 00:52:48.690 who are in the highest quintile of their Social Security 00:52:48.690 --> 00:52:49.350 number-- 00:52:49.350 --> 00:52:51.960 this is like year number five-- 00:52:51.960 --> 00:52:53.760 have much higher willingness to pay. 00:52:53.760 --> 00:52:55.590 And this is also all incentive-compatible 00:52:55.590 --> 00:52:57.840 [INAUDIBLE] methods essentially of using 00:52:57.840 --> 00:53:00.300 some form of BDM methods. 00:53:00.300 --> 00:53:03.750 People's willingness to pay is way higher when their Social 00:53:03.750 --> 00:53:04.890 Security number-- 00:53:04.890 --> 00:53:08.460 the last digits of their Social Security number are higher. 00:53:08.460 --> 00:53:10.290 Now of course, that shouldn't be the case. 00:53:10.290 --> 00:53:13.410 Your Social Security number has nothing to do with your tastes 00:53:13.410 --> 00:53:14.700 for wine. 00:53:14.700 --> 00:53:17.280 Because that's [INAUDIBLE] explicitly random. 00:53:17.280 --> 00:53:20.070 Yet, people are easily manipulable. 00:53:20.070 --> 00:53:23.610 And the difference in sort of valuations is like, [? huge. ?] 00:53:23.610 --> 00:53:26.340 If you happen to have a low Social Security number, 00:53:26.340 --> 00:53:33.050 or the last digit are low, then you are willing to pay $11.73. 00:53:33.050 --> 00:53:35.900 If it's high, you're willing to pay like three times or even 00:53:35.900 --> 00:53:37.250 more as much. 00:53:37.250 --> 00:53:41.630 So the huge differences show that people are pretty easily 00:53:41.630 --> 00:53:42.620 manipulable. 00:53:42.620 --> 00:53:45.580 Now notice that these questions that I asked here 00:53:45.580 --> 00:53:49.040 are sort of somewhat unusual sort of items. 00:53:49.040 --> 00:53:50.540 This is not asking like how much are 00:53:50.540 --> 00:53:52.490 you willing to pay for food at the food truck, 00:53:52.490 --> 00:53:57.660 or whatever stuff that you do every day, or for a pizza. 00:53:57.660 --> 00:53:59.780 Because there, people kind of know already 00:53:59.780 --> 00:54:01.200 how much they're paying anyway. 00:54:01.200 --> 00:54:03.090 And if I just ask you something else that's 00:54:03.090 --> 00:54:05.090 different from the market price, people probably 00:54:05.090 --> 00:54:09.650 would not want to be paying like $30 for it, 00:54:09.650 --> 00:54:14.010 even if their Social Security number tends to be quite high. 00:54:14.010 --> 00:54:17.120 Now, Ariely et al go further with this. 00:54:17.120 --> 00:54:21.413 In some sense they have sort of more radical deviations. 00:54:21.413 --> 00:54:23.330 So they elicit people's willingness to accept. 00:54:23.330 --> 00:54:26.450 And this is how much they have to be 00:54:26.450 --> 00:54:29.540 paid to endure an unpleasant sound 00:54:29.540 --> 00:54:32.940 for a different length of time. 00:54:32.940 --> 00:54:36.800 Now, why did they pick like an unpleasant sound? 00:54:36.800 --> 00:54:38.630 In part, because they could provide people 00:54:38.630 --> 00:54:39.890 like a sample of it. 00:54:39.890 --> 00:54:43.040 In part because people don't have any experience with that. 00:54:43.040 --> 00:54:45.330 It's completely unclear how much you are 00:54:45.330 --> 00:54:47.090 supposed to be willing to pay-- 00:54:47.090 --> 00:54:50.060 are willing to accept for listening to a sound. 00:54:50.060 --> 00:54:52.103 So there's no price or market price for it. 00:54:52.103 --> 00:54:54.020 And people sort of had to sort of in some ways 00:54:54.020 --> 00:54:56.250 rely on their own preferences for that. 00:54:56.250 --> 00:54:58.800 And then they know it's very easy to change the quantity. 00:54:58.800 --> 00:55:01.320 You can do like 10 seconds, 30 seconds, 50 seconds. 00:55:01.320 --> 00:55:05.320 It's very easy to manipulate that. 00:55:05.320 --> 00:55:07.030 Now what is the procedure? 00:55:07.030 --> 00:55:12.320 Subjects we're listening to a 30 second sample of the noise. 00:55:12.320 --> 00:55:16.390 Then there were answering whether they hypothetically 00:55:16.390 --> 00:55:19.210 would be willing to listen to the noise for another 30 00:55:19.210 --> 00:55:20.970 seconds for x cents. 00:55:20.970 --> 00:55:22.720 And then they were asked their willingness 00:55:22.720 --> 00:55:27.700 to accept for 10, 30 and 60 seconds afterwards. 00:55:27.700 --> 00:55:32.170 Notice that number two is only a hypothetical question, 00:55:32.170 --> 00:55:35.870 and is essentially entirely irrelevant for number three. 00:55:35.870 --> 00:55:36.370 Sorry. 00:55:36.370 --> 00:55:39.040 This is supposed to say number three, not number one. 00:55:39.040 --> 00:55:42.200 So number two here is like completely irrelevant. 00:55:42.200 --> 00:55:45.100 This should really have no effect whatsoever 00:55:45.100 --> 00:55:48.910 in affecting your willingness to accept for the 10, 30 and 60 00:55:48.910 --> 00:55:49.670 seconds. 00:55:49.670 --> 00:55:51.580 Again, this is just a hypothetical question 00:55:51.580 --> 00:55:53.890 that really does not matter, and is not 00:55:53.890 --> 00:55:57.440 going to be implemented at all. 00:55:57.440 --> 00:55:59.050 And then the experimental variation 00:55:59.050 --> 00:56:04.390 here is that x was varied across subjects. 00:56:04.390 --> 00:56:05.650 Now, why would x matter? 00:56:05.650 --> 00:56:07.310 What's going on here? 00:56:07.310 --> 00:56:10.380 What's the experiment trying to do? 00:56:10.380 --> 00:56:13.592 People are very unsure about what their valuation is. 00:56:13.592 --> 00:56:15.300 So now what they do is, essentially, they 00:56:15.300 --> 00:56:20.580 use the x-- the amount that's offered as like an anchor. 00:56:20.580 --> 00:56:24.570 The same way as in, usually in markets, when you go to a store 00:56:24.570 --> 00:56:27.000 and look at how much do certain things 00:56:27.000 --> 00:56:29.700 cost, and you don't know the quality of the underlying 00:56:29.700 --> 00:56:33.060 items, often people sort of try to infer quality or lack 00:56:33.060 --> 00:56:35.190 of quality from the price. 00:56:35.190 --> 00:56:37.380 And if we sort of say, if I'm telling you 00:56:37.380 --> 00:56:39.210 I'm going to pay you like $100 for this, 00:56:39.210 --> 00:56:40.630 it must be really, really painful. 00:56:40.630 --> 00:56:43.440 So people try to sort of infer in some ways something 00:56:43.440 --> 00:56:44.340 from that. 00:56:44.340 --> 00:56:47.920 Now notice that this is explicitly hypothetical. 00:56:47.920 --> 00:56:48.420 [SOUND] 00:56:48.420 --> 00:56:48.990 And so on. 00:56:48.990 --> 00:56:50.700 And so really it shouldn't matter. 00:56:50.700 --> 00:56:52.200 If people were sure about like-- 00:56:52.200 --> 00:56:53.580 I just played you the sound. 00:56:53.580 --> 00:56:55.580 So you shouldn't be able to tell me how much you 00:56:55.580 --> 00:56:56.620 like it or dislike it. 00:56:56.620 --> 00:56:58.240 So this should really be irrelevant. 00:56:58.240 --> 00:57:00.490 But people seem to sort of essentially just not really 00:57:00.490 --> 00:57:02.640 know what's appropriate, or how much they're 00:57:02.640 --> 00:57:05.820 willing to accept or not to do that. 00:57:05.820 --> 00:57:08.600 And x essentially is sort of then anchoring [INAUDIBLE].. 00:57:08.600 --> 00:57:09.100 OK. 00:57:09.100 --> 00:57:11.560 So now we get these somewhat messy graphs. 00:57:11.560 --> 00:57:15.410 Can somebody explain to me what this graph shows? 00:57:15.410 --> 00:57:17.920 What do we find? 00:57:17.920 --> 00:57:20.950 These are [INAUDIBLE] certain subjects. 00:57:20.950 --> 00:57:24.080 And then there's like 10 seconds, 30 seconds, 00:57:24.080 --> 00:57:25.420 and 60 seconds here. 00:57:25.420 --> 00:57:26.950 So people are always-- 00:57:26.950 --> 00:57:29.470 for each of these lines, on average, at least, 00:57:29.470 --> 00:57:31.600 willingness to accept goes up, in the way 00:57:31.600 --> 00:57:34.250 you expect, in the sense of, for more time, people 00:57:34.250 --> 00:57:35.890 are asking for more. 00:57:35.890 --> 00:57:37.760 That's very reasonable. 00:57:37.760 --> 00:57:40.360 So 30 seconds, presumably, are worse than 10 seconds. 00:57:40.360 --> 00:57:43.240 And 60 seconds are worse than 10 seconds. 00:57:43.240 --> 00:57:46.120 But the levels seem to be completely arbitrary. 00:57:46.120 --> 00:57:48.430 Essentially, giving people a high anchor 00:57:48.430 --> 00:57:52.570 increases the levels by a lot for each duration. 00:57:52.570 --> 00:57:55.130 And giving people a low anchor decreases their willingness 00:57:55.130 --> 00:57:57.255 to pay at least a little bit compared to no anchor, 00:57:57.255 --> 00:57:59.890 and surely compared to the [? high anchor. ?] 00:57:59.890 --> 00:58:00.702 That is to say-- 00:58:00.702 --> 00:58:02.660 and this is sort of where this term comes from, 00:58:02.660 --> 00:58:06.130 coherent arbitrariness, which is, essentially, people 00:58:06.130 --> 00:58:08.230 seem to be coherent in the sense of like, 00:58:08.230 --> 00:58:11.890 once you fix a certain level, based on that level, 00:58:11.890 --> 00:58:14.570 if you ask them, OK, if you tell me about 10 seconds 00:58:14.570 --> 00:58:17.020 and I ask you about 30 seconds, about 60 seconds, 00:58:17.020 --> 00:58:19.900 this demand curve or supply curve, if you want, 00:58:19.900 --> 00:58:21.400 looks pretty reasonable. 00:58:21.400 --> 00:58:23.620 But the actual level to start with 00:58:23.620 --> 00:58:27.520 is completely arbitrary, because I can essentially 00:58:27.520 --> 00:58:29.380 manipulate you by quite a bit. 00:58:29.380 --> 00:58:32.090 Look at the differences in magnitudes. 00:58:32.090 --> 00:58:33.460 This is like 50 versus 30. 00:58:33.460 --> 00:58:34.840 That's almost like twice as much, 00:58:34.840 --> 00:58:37.937 depending on just this [INAUDIBLE] anchor. 00:58:37.937 --> 00:58:39.770 And there's sort of different rounds in that 00:58:39.770 --> 00:58:41.853 that you can sort of do this one way or the other. 00:58:41.853 --> 00:58:45.010 And this is like essentially increasing or decreasing, 00:58:45.010 --> 00:58:48.360 starting in 10 seconds versus-- and then going up. 00:58:48.360 --> 00:58:52.170 People's willingness to accept goes up the longer it is. 00:58:52.170 --> 00:58:54.420 Or if you start with 60 seconds and go down, 00:58:54.420 --> 00:58:56.220 essentially, people's willingness to accept 00:58:56.220 --> 00:58:57.880 goes down the shorter it is. 00:58:57.880 --> 00:58:59.800 So the direction is very much coherent. 00:58:59.800 --> 00:59:01.770 The direction of essentially this sort 00:59:01.770 --> 00:59:05.340 of change in terms of relative to duration 00:59:05.340 --> 00:59:06.720 is very much coherent. 00:59:06.720 --> 00:59:09.930 The level seems very much arbitrary. 00:59:09.930 --> 00:59:14.370 And so this is where the title of the paper comes from. 00:59:14.370 --> 00:59:17.880 Arbitrariness, essentially, people's willingness to accept 00:59:17.880 --> 00:59:20.390 depended strongly on x. 00:59:20.390 --> 00:59:22.140 For x equals 50, the willingness to accept 00:59:22.140 --> 00:59:25.350 is like about $0.59 on average. 00:59:25.350 --> 00:59:29.040 For x equals 10, it's only $0.40. 00:59:29.040 --> 00:59:32.760 But it's also coherent in very sensible ways. 00:59:32.760 --> 00:59:34.770 Their willingness to accept are highly 00:59:34.770 --> 00:59:36.390 sensitive to the duration in the-- 00:59:36.390 --> 00:59:38.670 very much in the expected direction. 00:59:38.670 --> 00:59:43.640 Longer is always perceived to be more painful. 00:59:43.640 --> 00:59:44.180 OK. 00:59:44.180 --> 00:59:45.430 So how do we think about this? 00:59:45.430 --> 00:59:50.612 Well, so one, preferences can be influenced by relevant cues. 00:59:50.612 --> 00:59:52.820 For instance, [INAUDIBLE] arbitrary initial question, 00:59:52.820 --> 00:59:54.862 or the Social Security number, or whatever that's 00:59:54.862 --> 00:59:56.490 being elicited to start with. 00:59:56.490 --> 01:00:00.440 But once people have stated a preference, 01:00:00.440 --> 01:00:03.080 related preferences, they're like, essentially, 01:00:03.080 --> 01:00:07.010 surrounding preferences are consistent in a sense of like, 01:00:07.010 --> 01:00:10.380 if you think listening to the sound is painful, 01:00:10.380 --> 01:00:14.910 asking you to listen to it twice as long will be more painful. 01:00:14.910 --> 01:00:16.460 Therefore, I have to pay you more 01:00:16.460 --> 01:00:21.860 to do that compared to what you said in the first place. 01:00:21.860 --> 01:00:23.350 Any questions or is that clear? 01:00:25.980 --> 01:00:27.490 OK. 01:00:27.490 --> 01:00:29.530 So now there are some concerns about x 01:00:29.530 --> 01:00:32.800 might be viewed as a hint from the experimenters, 01:00:32.800 --> 01:00:34.390 how bad the sound is. 01:00:34.390 --> 01:00:36.730 But you know, people just listened to the sound 01:00:36.730 --> 01:00:38.200 for 30 seconds to start with. 01:00:38.200 --> 01:00:40.100 Like, I gave you the sound for 30 seconds. 01:00:40.100 --> 01:00:41.930 I told you exactly what it is. 01:00:41.930 --> 01:00:47.120 So in some sense, that seems like not really a concern 01:00:47.120 --> 01:00:47.630 anyway. 01:00:47.630 --> 01:00:49.250 But there's also-- the very end of this experiment, 01:00:49.250 --> 01:00:51.440 where x was generated by the last two digits 01:00:51.440 --> 01:00:55.130 of the subject's Social Security number, 01:00:55.130 --> 01:00:58.190 and explicitly so, and still there was a correlation between 01:00:58.190 --> 01:01:00.680 essentially x and subjects' willingness to accept, 01:01:00.680 --> 01:01:03.740 which really shouldn't be in any other-- 01:01:03.740 --> 01:01:06.070 in the absence of such anchoring effects. 01:01:06.070 --> 01:01:07.280 Are stakes too low? 01:01:07.280 --> 01:01:09.740 I told you about like, $0.30, and $0.50. 01:01:09.740 --> 01:01:11.592 Maybe people just don't care, and so on. 01:01:11.592 --> 01:01:13.550 But there's also another experiment [INAUDIBLE] 01:01:13.550 --> 01:01:14.390 10-fold stakes. 01:01:14.390 --> 01:01:16.350 And they got essentially the same results. 01:01:20.110 --> 01:01:22.950 So this evidence now is very much consistent with the idea 01:01:22.950 --> 01:01:25.380 that subjects are searching for their preferences. 01:01:25.380 --> 01:01:27.240 They don't quite know their true willingness 01:01:27.240 --> 01:01:30.940 to accept for the sound, and this is essentially arbitrary. 01:01:30.940 --> 01:01:33.000 But they know their willingness to accept 01:01:33.000 --> 01:01:35.370 should relate to each other in a coherent way. 01:01:35.370 --> 01:01:38.310 And in fact, once you fix the level in some ways, 01:01:38.310 --> 01:01:42.150 they are in fact coherent, or essentially they're 01:01:42.150 --> 01:01:43.320 sort of making sense. 01:01:43.320 --> 01:01:45.240 They're [INAUDIBLE] consistent with that, 01:01:45.240 --> 01:01:48.060 once you fix a certain level. 01:01:48.060 --> 01:01:52.930 Now this is now getting back to Tom Sawyer. 01:01:52.930 --> 01:01:56.300 Now, the paper goes-- so this essentially is sort of saying, 01:01:56.300 --> 01:01:59.630 your level of willingness to accept for a certain good 01:01:59.630 --> 01:02:02.510 that's unpleasant is malleable. 01:02:02.510 --> 01:02:04.130 Essentially, I can manipulate you 01:02:04.130 --> 01:02:08.960 to ask for a lower or a higher price, 01:02:08.960 --> 01:02:11.450 depending on some irrelevant question that I ask you, 01:02:11.450 --> 01:02:14.990 for a given thing that is perceived to be unpleasant. 01:02:14.990 --> 01:02:19.010 Similarly, I can manipulate you to pay more or less 01:02:19.010 --> 01:02:21.530 for a certain good that's perceived 01:02:21.530 --> 01:02:23.750 to be a good, something that you want. 01:02:23.750 --> 01:02:25.880 So the stuff on the Social Security number 01:02:25.880 --> 01:02:29.130 that I showed you previously, going back, 01:02:29.130 --> 01:02:30.390 these are all kind of things-- 01:02:30.390 --> 01:02:32.760 presumably, there's some use for these things. 01:02:32.760 --> 01:02:35.550 Belgian chocolates are supposed to be delicious. 01:02:35.550 --> 01:02:37.230 Red wine, even if you don't like wine, 01:02:37.230 --> 01:02:39.190 you can sell it to somebody else or something. 01:02:39.190 --> 01:02:42.600 These are all things that seem to be worth willing 01:02:42.600 --> 01:02:45.210 to pay for something. 01:02:45.210 --> 01:02:47.460 And now essentially, given that you 01:02:47.460 --> 01:02:50.430 are willing to pay something, or some positive amount for it, 01:02:50.430 --> 01:02:53.580 I can manipulate now-- or the experimenters, in this case, 01:02:53.580 --> 01:02:58.230 can manipulate people into higher or lower willingness 01:02:58.230 --> 01:02:59.690 to pay. 01:02:59.690 --> 01:03:02.650 And so that's true for positive or negative things. 01:03:02.650 --> 01:03:05.680 Now, what's amazing about the Tom Sawyer story, 01:03:05.680 --> 01:03:08.540 however, isn't the-- 01:03:08.540 --> 01:03:09.820 it's ain't that work? 01:03:09.820 --> 01:03:12.340 It's sort of like, that's the expression he sort of says 01:03:12.340 --> 01:03:13.870 towards the end. 01:03:13.870 --> 01:03:16.690 The amazing part here is that not only 01:03:16.690 --> 01:03:20.770 is Tom Sawyer able to manipulate his friend's willingness 01:03:20.770 --> 01:03:25.090 to accept, or willingness to pay in one direction or-- 01:03:25.090 --> 01:03:27.670 increase or decrease it, but he's even able 01:03:27.670 --> 01:03:30.155 to flip it from something that he 01:03:30.155 --> 01:03:32.530 hates doing, that's essentially something that you really 01:03:32.530 --> 01:03:34.510 do not want to do, where essentially one has 01:03:34.510 --> 01:03:37.840 to pay somebody to do it, but instead, his friend 01:03:37.840 --> 01:03:41.135 is willing to pay Tom Sawyer for being able to do it, right? 01:03:41.135 --> 01:03:42.760 The friend, now, instead of like having 01:03:42.760 --> 01:03:45.260 the friend to pay for it, you can say, I have to pay you $10 01:03:45.260 --> 01:03:48.190 to do it, the friend is very happy to say, Tom, 01:03:48.190 --> 01:03:49.838 I'm going to pay you some amount. 01:03:49.838 --> 01:03:51.880 So I think he gives him some apples or some candy 01:03:51.880 --> 01:03:55.570 or whatever, so that he can actually do it himself. 01:03:55.570 --> 01:03:57.670 So essentially, what the manipulation here does 01:03:57.670 --> 01:04:00.280 is not only changing the level for something 01:04:00.280 --> 01:04:02.890 that's good or bad, but it's flipping the sign 01:04:02.890 --> 01:04:06.130 from willingness to accept to willingness to pay, which 01:04:06.130 --> 01:04:08.387 is a more radical deviation. 01:04:08.387 --> 01:04:10.470 You might sort of think we know what's good for us 01:04:10.470 --> 01:04:11.630 and what's bad for us. 01:04:11.630 --> 01:04:13.630 But it seems to be that what Tom Sawyer is doing 01:04:13.630 --> 01:04:17.764 is manipulating sort of the social perception of the item 01:04:17.764 --> 01:04:20.380 or the activity, and that sort of flips essentially 01:04:20.380 --> 01:04:22.930 the sign of the item. 01:04:22.930 --> 01:04:27.880 So then Ariely now does that in a very beautiful experiment 01:04:27.880 --> 01:04:28.420 in class. 01:04:28.420 --> 01:04:31.180 And I always have wanted to sort of replicate this, 01:04:31.180 --> 01:04:34.000 but I'm not sure I should. 01:04:34.000 --> 01:04:37.510 So what he does is, or at the time, this is with MBAs-- 01:04:37.510 --> 01:04:42.340 where he did a poetry reading from Walt Whitman's 01:04:42.340 --> 01:04:44.020 Leaves of Grass. 01:04:44.020 --> 01:04:47.725 And there, he has an experiment, where half of the class 01:04:47.725 --> 01:04:49.588 is asked hypothetically-- 01:04:49.588 --> 01:04:51.130 again, hypothetically, whether they'd 01:04:51.130 --> 01:04:53.590 be willing to pay $10 to listen to Ariely 01:04:53.590 --> 01:04:55.450 recite poetry for 10 minutes. 01:04:55.450 --> 01:04:59.440 So he's like, OK, in class, some of you 01:04:59.440 --> 01:05:01.120 are able to listen to this. 01:05:01.120 --> 01:05:02.320 But it could be like-- 01:05:02.320 --> 01:05:05.230 I think it's outside of class, because otherwise [INAUDIBLE].. 01:05:05.230 --> 01:05:09.460 And are you willing to pay $10 for attending this poetry 01:05:09.460 --> 01:05:10.220 reading? 01:05:10.220 --> 01:05:12.220 And I think actually, Ariely is a very good sort 01:05:12.220 --> 01:05:13.120 of story reader. 01:05:13.120 --> 01:05:15.040 So it might actually be fun to listen to it. 01:05:15.040 --> 01:05:17.050 But some of you know to ask hypothetically, 01:05:17.050 --> 01:05:19.690 again, are you willing to pay $10 for it? 01:05:19.690 --> 01:05:21.700 The other half then are asked hypothetically 01:05:21.700 --> 01:05:25.660 are they're willing to accept $10 to listen to Ariely recite 01:05:25.660 --> 01:05:26.920 the poetry for 10 minutes. 01:05:26.920 --> 01:05:30.190 That's to say, it sounds like it's pretty painful 01:05:30.190 --> 01:05:31.570 to listen to Ariely. 01:05:31.570 --> 01:05:33.430 He's going to pay you $10 to do it. 01:05:33.430 --> 01:05:36.920 And are you willing to do that? 01:05:36.920 --> 01:05:39.110 Notice that here is only hypothetical. 01:05:39.110 --> 01:05:40.700 These are all hypothetical choices 01:05:40.700 --> 01:05:42.455 that will not be implemented. 01:05:42.455 --> 01:05:44.330 So these hypothetical questions should really 01:05:44.330 --> 01:05:47.810 have no effect whatsoever on what's asked afterwards. 01:05:47.810 --> 01:05:50.510 Because afterwards, people are in fact indicating 01:05:50.510 --> 01:05:53.210 their monetary valuations for one, three or six 01:05:53.210 --> 01:05:56.230 minutes of poetry reading. 01:05:56.230 --> 01:05:59.340 Now you can already guess what's happening here is here, 01:05:59.340 --> 01:06:00.780 the first condition is essentially 01:06:00.780 --> 01:06:02.460 asking for people's willingness to pay. 01:06:02.460 --> 01:06:03.930 That's essentially selling the item 01:06:03.930 --> 01:06:06.180 as like a good, that's something that you really want. 01:06:06.180 --> 01:06:09.060 And you're sort of essentially encouraging people 01:06:09.060 --> 01:06:11.880 to say, well, you should be willing to pay 01:06:11.880 --> 01:06:16.440 something for it, even as much as like $10 for 10 minutes. 01:06:16.440 --> 01:06:18.270 And that'll [INAUDIBLE] encourage people, 01:06:18.270 --> 01:06:22.140 since they don't quite know, listening to your professor 01:06:22.140 --> 01:06:24.480 is like doing some weird poetry could be really great, 01:06:24.480 --> 01:06:27.060 or could be really bad, they don't quite know. 01:06:27.060 --> 01:06:30.270 But here, essentially, they're manipulated into paying for it, 01:06:30.270 --> 01:06:34.440 while here they're manipulated into being paid for it, 01:06:34.440 --> 01:06:37.890 or asking for money to be paid for it. 01:06:37.890 --> 01:06:43.765 So now, what they then get is their willingness to pay, 01:06:43.765 --> 01:06:45.390 and their willingness to pay conditions 01:06:45.390 --> 01:06:47.160 when the first condition here, where 01:06:47.160 --> 01:06:51.870 they are given the hypothetical $10 for 10 minutes. 01:06:51.870 --> 01:06:56.260 Here, the willingness to pay goes up and is positive. 01:06:56.260 --> 01:06:59.110 And here, it's negative and goes down. 01:06:59.110 --> 01:07:03.100 So what we see essentially is that not only is it 01:07:03.100 --> 01:07:06.760 the case that when you make it essential, when you ask people 01:07:06.760 --> 01:07:08.200 how much are they willing to pay, 01:07:08.200 --> 01:07:10.890 people are in fact afterwards willing to pay for it. 01:07:10.890 --> 01:07:13.120 But notice that everybody could have just said zero 01:07:13.120 --> 01:07:14.650 if they didn't want to pay. 01:07:14.650 --> 01:07:18.190 But in addition, once you sort of 01:07:18.190 --> 01:07:21.490 fix their willingness to pay, the demand curve 01:07:21.490 --> 01:07:24.790 is very much sensible in a sense. 01:07:24.790 --> 01:07:27.195 Once you're willing to pay some amount for one minute, 01:07:27.195 --> 01:07:28.570 you're going to be willing to pay 01:07:28.570 --> 01:07:31.240 more money for more minutes. 01:07:31.240 --> 01:07:33.280 Essentially, once we define this good to be 01:07:33.280 --> 01:07:35.830 a good for you, or like this item or this activity 01:07:35.830 --> 01:07:37.570 to be like a good thing for you, people 01:07:37.570 --> 01:07:39.190 say more must be better, and sort of 01:07:39.190 --> 01:07:42.020 are willing to pay more for more minutes. 01:07:42.020 --> 01:07:47.320 On the other hand, once the item is determined as like a bad, 01:07:47.320 --> 01:07:49.450 in the sense of this is really a bad activity. 01:07:49.450 --> 01:07:52.000 Listening to your professor reciting 01:07:52.000 --> 01:07:55.993 poetry is really awful, and you have to be paid for that, 01:07:55.993 --> 01:07:57.910 then they sort of say, well one minute is bad, 01:07:57.910 --> 01:07:59.680 but like six minutes surely is really bad. 01:07:59.680 --> 01:08:02.410 And you have to pay me a lot more for six minutes than 01:08:02.410 --> 01:08:04.270 for one minute. 01:08:04.270 --> 01:08:05.920 And that's essentially saying subjects 01:08:05.920 --> 01:08:09.430 don't know whether this reading is good or bad. 01:08:09.430 --> 01:08:11.180 But they do know, essentially, either way, 01:08:11.180 --> 01:08:13.180 more requires more money. 01:08:13.180 --> 01:08:16.000 And that's essentially exactly the coherent arbitrariness, 01:08:16.000 --> 01:08:19.180 where essentially, the level or even like 01:08:19.180 --> 01:08:20.680 the sign is really unclear. 01:08:20.680 --> 01:08:22.180 But once the sign is sort of fixed, 01:08:22.180 --> 01:08:26.340 people behave in pretty reasonable ways. 01:08:26.340 --> 01:08:27.670 I actually don't remember. 01:08:27.670 --> 01:08:29.406 I should look it up-- 01:08:29.406 --> 01:08:33.569 surely in the paper, whether it's possible in this condition 01:08:33.569 --> 01:08:34.439 to actually have-- 01:08:34.439 --> 01:08:36.147 [INAUDIBLE] so the [INAUDIBLE] experiment 01:08:36.147 --> 01:08:39.060 would surely always ask you a whole range where 01:08:39.060 --> 01:08:43.540 you can always say, between say, minus six or minus 10, 01:08:43.540 --> 01:08:44.790 and plus 10. 01:08:44.790 --> 01:08:47.569 Now tell me your willingness to accept versus willingness 01:08:47.569 --> 01:08:48.069 to pay. 01:08:48.069 --> 01:08:49.944 So if it's positive, it's willingness to pay. 01:08:49.944 --> 01:08:52.020 If it's negative, it's willingness to accept. 01:08:52.020 --> 01:08:53.189 I think that's what's-- 01:08:53.189 --> 01:08:54.689 so surely, that's how the experiment 01:08:54.689 --> 01:08:56.010 should have been done. 01:08:56.010 --> 01:08:58.838 I would like to think and hope that's also 01:08:58.838 --> 01:08:59.880 how it was actually done. 01:08:59.880 --> 01:09:01.080 I don't remember. 01:09:01.080 --> 01:09:02.543 This is quite a while ago. 01:09:02.543 --> 01:09:04.210 But it's really in the paper, so I don't 01:09:04.210 --> 01:09:05.609 know if anybody remembers. 01:09:05.609 --> 01:09:07.312 But I think it's exactly right. 01:09:07.312 --> 01:09:09.270 In some sense, the experiment is very much sort 01:09:09.270 --> 01:09:11.340 of manipulating people in the way of saying like, 01:09:11.340 --> 01:09:16.319 well, so the prompt here really gets you into like, OK, you're 01:09:16.319 --> 01:09:17.463 supposed to pay for this. 01:09:17.463 --> 01:09:19.380 Now it's kind of a little bit of a weird thing 01:09:19.380 --> 01:09:21.630 to say I'm willing to accept. 01:09:21.630 --> 01:09:26.189 Professor says, are you willing to pay $10 for it? 01:09:26.189 --> 01:09:27.689 And then you say, well, how much are 01:09:27.689 --> 01:09:30.390 you willing to pay for like six minutes? 01:09:30.390 --> 01:09:32.580 And then you say, well you have to pay me like $5, 01:09:32.580 --> 01:09:35.340 is a bit of a weird thing to say and the other way around. 01:09:35.340 --> 01:09:38.390 So it's very much sort of like I think a set up 01:09:38.390 --> 01:09:40.950 to sort of generate the effects that they're looking for. 01:09:40.950 --> 01:09:43.719 I think the underlying-- so I think the underlying essence-- 01:09:43.719 --> 01:09:46.399 and there are some other experiments that show somewhat 01:09:46.399 --> 01:09:47.580 similar results-- 01:09:47.580 --> 01:09:50.600 the underlying essence is right that particularly 01:09:50.600 --> 01:09:53.240 for unknown goods, people's preferences 01:09:53.240 --> 01:09:54.770 are very much malleable. 01:09:54.770 --> 01:09:57.182 And people just don't know what they are. 01:09:57.182 --> 01:09:58.640 And there's also some other things. 01:09:58.640 --> 01:10:01.970 For example, if you think about, for example, social activities, 01:10:01.970 --> 01:10:05.090 where maybe your friends like them, or they don't like them, 01:10:05.090 --> 01:10:07.880 and people are very much like-- and this is exactly 01:10:07.880 --> 01:10:10.490 the Tom Sawyer example, where people are very much sort 01:10:10.490 --> 01:10:12.020 of malleable. 01:10:12.020 --> 01:10:14.660 And you can vastly shape what people 01:10:14.660 --> 01:10:18.080 want versus not in many cases. 01:10:18.080 --> 01:10:20.730 And for example, one example would be like, 01:10:20.730 --> 01:10:23.060 there's some very nice work by Loewenstein 01:10:23.060 --> 01:10:27.020 and coauthors that looks at education. 01:10:27.020 --> 01:10:30.730 And sort of, do students want to study versus not. 01:10:30.730 --> 01:10:34.130 And do you want to be a nerd versus not in class. 01:10:34.130 --> 01:10:37.550 And so depending very much on your environment as a kid, 01:10:37.550 --> 01:10:40.310 when you grow up, if all of your friends 01:10:40.310 --> 01:10:42.830 are sort of essentially not working hard, and sort of 01:10:42.830 --> 01:10:45.740 want to be cool, and sort of do not want to study, 01:10:45.740 --> 01:10:48.380 and it's not a cool thing in your environment to study, 01:10:48.380 --> 01:10:50.450 people might just not want that. 01:10:50.450 --> 01:10:53.742 But on the other hand, if you have like five nerd friends who 01:10:53.742 --> 01:10:55.700 are all working really hard, or like, if you're 01:10:55.700 --> 01:10:57.990 running around MIT, studying is a cool thing. 01:10:57.990 --> 01:11:00.050 And everybody sort of is working hard. 01:11:00.050 --> 01:11:04.890 And that becomes suddenly very much like a desirable activity. 01:11:04.890 --> 01:11:06.560 So I think more generally, while this 01:11:06.560 --> 01:11:08.660 is a bit of a contrived experiment, 01:11:08.660 --> 01:11:11.900 surely people's preferences are very much 01:11:11.900 --> 01:11:16.310 malleable through their social environments, 01:11:16.310 --> 01:11:19.790 and what they think others think, 01:11:19.790 --> 01:11:21.380 and how they perceive it. 01:11:21.380 --> 01:11:23.035 And that can be [INAUDIBLE] shaped by-- 01:11:26.470 --> 01:11:29.090 manipulated in certain ways, and potentially affected. 01:11:29.090 --> 01:11:31.520 That could be very much like a policy angle 01:11:31.520 --> 01:11:33.710 if you wanted to change people's behavior profoundly 01:11:33.710 --> 01:11:38.340 without spending much money, in fact. 01:11:38.340 --> 01:11:41.600 Let me-- I think that's mostly what I have to say. 01:11:41.600 --> 01:11:44.240 Let me summarize, and then tell you about next week 01:11:44.240 --> 01:11:46.170 for a second. 01:11:46.170 --> 01:11:48.110 So we asked the question, whether people 01:11:48.110 --> 01:11:49.340 have stable preferences. 01:11:49.340 --> 01:11:50.870 And it seems to me that people don't 01:11:50.870 --> 01:11:53.450 have clear preferences for goods and experiences, 01:11:53.450 --> 01:11:56.330 and sort of construct their preferences on the spot. 01:11:56.330 --> 01:12:00.020 Now, they're influenced by environmental cues 01:12:00.020 --> 01:12:03.050 that, in a way, that doesn't necessarily 01:12:03.050 --> 01:12:05.870 reflect the true utility from the good or experience. 01:12:05.870 --> 01:12:07.610 And that's very much also what companies 01:12:07.610 --> 01:12:08.527 are doing [INAUDIBLE]. 01:12:08.527 --> 01:12:12.230 And sometimes, some items are made really, really expensive. 01:12:12.230 --> 01:12:13.820 Like, why is this thing expensive? 01:12:13.820 --> 01:12:16.640 And somehow companies are able to create 01:12:16.640 --> 01:12:19.580 some fads or some way of making things desirable by just making 01:12:19.580 --> 01:12:23.095 them expensive and making people want them in that way. 01:12:23.095 --> 01:12:24.470 That very much relies on the fact 01:12:24.470 --> 01:12:26.643 that people are in fact sort of malleable. 01:12:26.643 --> 01:12:28.310 And you can sort of introspect and think 01:12:28.310 --> 01:12:31.400 about what things in the world make you generally happy. 01:12:31.400 --> 01:12:32.827 What things do you really like? 01:12:32.827 --> 01:12:34.910 And what things are sort of more things that like, 01:12:34.910 --> 01:12:36.860 well, other people like them, and you kind of 01:12:36.860 --> 01:12:39.193 get sort of manipulated, or you get sort of like tricked 01:12:39.193 --> 01:12:40.920 into doing that. 01:12:40.920 --> 01:12:42.410 And in some sense, and we'll talk 01:12:42.410 --> 01:12:44.420 about this next week a little bit about happiness-- 01:12:44.420 --> 01:12:46.712 it's important to sort of try and figure it out and try 01:12:46.712 --> 01:12:49.910 to sort of not perhaps be as much influenced by others, 01:12:49.910 --> 01:12:53.960 and rather figuring out what you generally and truly like. 01:12:53.960 --> 01:12:56.390 So there's a nice series of experiments 01:12:56.390 --> 01:12:59.510 that sort of demonstrate this coherent arbitrariness, 01:12:59.510 --> 01:13:02.480 in very sort of like clean variation, 01:13:02.480 --> 01:13:06.080 and somewhat contrived context. 01:13:06.080 --> 01:13:08.900 So you might wonder then, does it matter in the real world? 01:13:08.900 --> 01:13:11.772 Well it's perhaps less important in settings 01:13:11.772 --> 01:13:12.980 where people have experience. 01:13:12.980 --> 01:13:14.897 Again, like if I had done the same experiments 01:13:14.897 --> 01:13:17.488 with like pizza, you know how much you're 01:13:17.488 --> 01:13:18.530 willing to pay for pizza. 01:13:18.530 --> 01:13:19.613 You know the market price. 01:13:19.613 --> 01:13:21.260 And you know also, and so on. 01:13:21.260 --> 01:13:23.810 And so there, probably, your willingness to pay 01:13:23.810 --> 01:13:26.990 and your preferences are very much like set. 01:13:26.990 --> 01:13:29.630 But in some other cases, in particular like new 01:13:29.630 --> 01:13:32.310 environments, [INAUDIBLE] sort of new-- 01:13:32.310 --> 01:13:34.057 when preferences are shaped-- 01:13:34.057 --> 01:13:36.140 for example, think about maybe back when you first 01:13:36.140 --> 01:13:40.610 started at MIT, sort of seeing lots of other students, 01:13:40.610 --> 01:13:42.890 when norms are shaped, people's preferences 01:13:42.890 --> 01:13:46.820 are very much malleable and can be influenced profoundly. 01:13:46.820 --> 01:13:52.370 Now, there's not so much like actual field evidence 01:13:52.370 --> 01:13:55.663 in high stakes settings that are sort of this clean nature. 01:13:55.663 --> 01:13:57.080 Having said that, the examples of, 01:13:57.080 --> 01:13:58.970 for example, environmental-- sorry, 01:13:58.970 --> 01:14:01.982 educational choices is perhaps the most compelling one 01:14:01.982 --> 01:14:03.440 because that's really a high stakes 01:14:03.440 --> 01:14:05.160 setting that matters a lot. 01:14:05.160 --> 01:14:08.450 But people's preferences or their choices 01:14:08.450 --> 01:14:10.940 are very much malleable. 01:14:10.940 --> 01:14:14.920 That's all I have to say on coherent arbitrariness. 01:14:14.920 --> 01:14:19.370 So next time, Monday, we're going to talk about poverty. 01:14:19.370 --> 01:14:21.895 Please read the paper by Mani et al. 01:14:21.895 --> 01:14:23.270 And then on Wednesday we're going 01:14:23.270 --> 01:14:25.460 to talk about happiness and mental health. 01:14:25.460 --> 01:14:30.290 Now, I promised you a guest lecturer. 01:14:30.290 --> 01:14:32.990 And so you know, I was thinking about what's 01:14:32.990 --> 01:14:33.810 a good thing to do. 01:14:33.810 --> 01:14:37.080 And so you may have heard about this at UC Berkeley. 01:14:37.080 --> 01:14:40.560 They have llamas come to campus to make students happy, 01:14:40.560 --> 01:14:42.370 and destress them in some ways. 01:14:42.370 --> 01:14:45.623 Now, of course, I can't bring any real llamas. 01:14:45.623 --> 01:14:47.040 But what I found, and you may have 01:14:47.040 --> 01:14:48.460 seen this on The Daily Show. 01:14:48.460 --> 01:14:50.210 There's this thing called Goat To Meeting, 01:14:50.210 --> 01:14:53.280 where you can essentially get like a goat or a llama 01:14:53.280 --> 01:14:56.040 to come to your meeting for 10 minutes. 01:14:56.040 --> 01:15:00.600 And I did ask for like a llama to arrive. 01:15:00.600 --> 01:15:02.340 But there's apparently a chance that it 01:15:02.340 --> 01:15:03.990 might be a goat or a cow. 01:15:03.990 --> 01:15:08.490 So we'll have a visitor at the end of the lecture. 01:15:08.490 --> 01:15:14.280 I think I asked for 2:20, or between 2:20 and 2:30. 01:15:14.280 --> 01:15:17.400 And which we'll have a guest lecturer that will hopefully 01:15:17.400 --> 01:15:20.640 either tell us about happiness or maybe make some of you, 01:15:20.640 --> 01:15:23.370 or at least myself, happy. 01:15:23.370 --> 01:15:27.040 That's all I have for today. 01:15:27.040 --> 01:15:30.740 I'm happy to answer any questions that you might have.