WEBVTT 00:00:00.000 --> 00:00:06.930 align:middle line:84% [SQUEAKING] [RUSTLING] [CLICKING] 00:00:06.930 --> 00:00:18.820 align:middle line:90% 00:00:18.820 --> 00:00:21.140 align:middle line:84% PROFESSOR: OK, so nice to see you all again. 00:00:21.140 --> 00:00:23.420 align:middle line:84% So we're going to talk about the family. 00:00:23.420 --> 00:00:27.070 align:middle line:84% So I'm going to today talk about basic intra-household 00:00:27.070 --> 00:00:30.820 align:middle line:84% bargaining models, sort of building-- in some sense 00:00:30.820 --> 00:00:32.950 align:middle line:84% kind of building us up to the Oudry reading 00:00:32.950 --> 00:00:34.610 align:middle line:84% that you guys did for today on sort 00:00:34.610 --> 00:00:37.063 align:middle line:84% of testing for [INAUDIBLE] efficiency in the household. 00:00:37.063 --> 00:00:38.980 align:middle line:84% And then Esther is going to come back and talk 00:00:38.980 --> 00:00:43.120 align:middle line:84% about other gender and within the family issues next time. 00:00:43.120 --> 00:00:51.350 align:middle line:84% So just by way of background, right, so in some sense, 00:00:51.350 --> 00:00:55.660 align:middle line:84% so far, we've abstracted from households, right. 00:00:55.660 --> 00:01:00.100 align:middle line:84% So what I meant by that is, we assumed 00:01:00.100 --> 00:01:02.440 align:middle line:84% that a household had like a single utility function, 00:01:02.440 --> 00:01:04.785 align:middle line:84% u of x, where x is like everything 00:01:04.785 --> 00:01:05.660 align:middle line:90% that can be consumed. 00:01:05.660 --> 00:01:08.210 align:middle line:84% And we had some wealth or whatever. 00:01:08.210 --> 00:01:10.480 align:middle line:84% And we had the price of the pX is less than W, 00:01:10.480 --> 00:01:11.830 align:middle line:90% some budget constraint, right. 00:01:11.830 --> 00:01:13.750 align:middle line:84% So that's the unitary household model, OK. 00:01:13.750 --> 00:01:17.440 align:middle line:90% 00:01:17.440 --> 00:01:19.150 align:middle line:84% But of course, as we know, the real world 00:01:19.150 --> 00:01:22.420 align:middle line:84% is more complicated than that, right. 00:01:22.420 --> 00:01:27.130 align:middle line:84% Households consist typically-- many households, their spouses, 00:01:27.130 --> 00:01:27.640 align:middle line:90% right. 00:01:27.640 --> 00:01:29.170 align:middle line:90% Often, people are married. 00:01:29.170 --> 00:01:31.450 align:middle line:84% And those spouses need to make decisions together 00:01:31.450 --> 00:01:33.910 align:middle line:84% about their consumption choices, OK. 00:01:33.910 --> 00:01:36.070 align:middle line:84% And that can be both statically and dynamically. 00:01:36.070 --> 00:01:38.237 align:middle line:84% I'm going to mostly talk about static choices today, 00:01:38.237 --> 00:01:40.810 align:middle line:84% but I'll mention, at least hopefully at the very end, 00:01:40.810 --> 00:01:42.843 align:middle line:90% an example of dynamic choice. 00:01:42.843 --> 00:01:45.010 align:middle line:84% And also, you may imagine that different spouses may 00:01:45.010 --> 00:01:46.900 align:middle line:84% have different control over different types of assets 00:01:46.900 --> 00:01:48.520 align:middle line:84% or different consumption decisions. 00:01:48.520 --> 00:01:54.250 align:middle line:84% And this may be relevant for some development issues. 00:01:54.250 --> 00:01:55.630 align:middle line:90% So this is the benchmark. 00:01:55.630 --> 00:01:58.190 align:middle line:84% And the question is, is that kind of a useful benchmark, 00:01:58.190 --> 00:02:00.315 align:middle line:84% and when do we actually need to take into account 00:02:00.315 --> 00:02:02.440 align:middle line:84% the fact that actually there are different decision 00:02:02.440 --> 00:02:04.657 align:middle line:84% makers in the household, and they may be interacting 00:02:04.657 --> 00:02:05.740 align:middle line:90% in kind of different ways? 00:02:05.740 --> 00:02:08.590 align:middle line:84% Or when is a simplification kind of the good one? 00:02:08.590 --> 00:02:11.800 align:middle line:84% OK, so there's sort of a long literature that 00:02:11.800 --> 00:02:14.470 align:middle line:84% started with these papers by Chiappori, Browning, 00:02:14.470 --> 00:02:17.040 align:middle line:84% and Bourguignon, which were about testing 00:02:17.040 --> 00:02:18.790 align:middle line:84% this unitary model and trying to see if we 00:02:18.790 --> 00:02:20.180 align:middle line:90% can accept it or reject it. 00:02:20.180 --> 00:02:22.660 align:middle line:84% And the paper you guys read for today is one example of one 00:02:22.660 --> 00:02:25.720 align:middle line:84% that's sort of about an example of testing the unitary model, 00:02:25.720 --> 00:02:27.710 align:middle line:90% but there are others. 00:02:27.710 --> 00:02:30.970 align:middle line:84% So here's a model they're going to start with. 00:02:30.970 --> 00:02:33.490 align:middle line:90% They're going to start with a-- 00:02:33.490 --> 00:02:35.140 align:middle line:84% or here's how you build this up, right. 00:02:35.140 --> 00:02:36.432 align:middle line:90% So how would you build this up? 00:02:36.432 --> 00:02:40.638 align:middle line:84% You'd imagine, basically, that there's a-- 00:02:40.638 --> 00:02:41.930 align:middle line:90% I'll show you this in a second. 00:02:41.930 --> 00:02:44.320 align:middle line:84% We're going to have a joint utility function that's 00:02:44.320 --> 00:02:47.780 align:middle line:84% going to be some weighted sum of individual utility functions, 00:02:47.780 --> 00:02:48.280 align:middle line:90% right. 00:02:48.280 --> 00:02:49.330 align:middle line:90% And that can be very general. 00:02:49.330 --> 00:02:51.730 align:middle line:84% Like I can care about my spouse, or my spouse can care about me, 00:02:51.730 --> 00:02:52.230 align:middle line:90% or whatever. 00:02:52.230 --> 00:02:54.730 align:middle line:84% But it's individual utility functions. 00:02:54.730 --> 00:02:57.340 align:middle line:84% And there are going to be some kind of weights 00:02:57.340 --> 00:02:59.230 align:middle line:84% between those two utility functions, OK. 00:02:59.230 --> 00:03:01.313 align:middle line:84% And those are going to be determined by bargaining 00:03:01.313 --> 00:03:02.510 align:middle line:90% or something like that. 00:03:02.510 --> 00:03:07.610 align:middle line:84% And what we're going to see is that we're going to see that 00:03:07.610 --> 00:03:09.680 align:middle line:84% basically, households are going to be maximizing 00:03:09.680 --> 00:03:10.755 align:middle line:90% this sum of weigted-- 00:03:10.755 --> 00:03:13.130 align:middle line:84% we're [? actually ?] going to do sum of weighted utility. 00:03:13.130 --> 00:03:15.530 align:middle line:84% And that's going to imply a Pareto efficiency, right? 00:03:15.530 --> 00:03:17.822 align:middle line:84% The Pareto efficiency means that if you can think about 00:03:17.822 --> 00:03:20.345 align:middle line:90% that there shouldn't be-- 00:03:20.345 --> 00:03:25.210 align:middle line:90% 00:03:25.210 --> 00:03:27.710 align:middle line:84% for example, as a test you guys saw in the [? Udry ?] paper, 00:03:27.710 --> 00:03:29.270 align:middle line:84% the particular allocation of control 00:03:29.270 --> 00:03:31.603 align:middle line:84% over certain types of production in the household-- that 00:03:31.603 --> 00:03:33.320 align:middle line:84% shouldn't change the total amount 00:03:33.320 --> 00:03:35.255 align:middle line:84% produced in the household, for example. 00:03:35.255 --> 00:03:37.590 align:middle line:84% OK, but let's just show what this means, OK? 00:03:37.590 --> 00:03:39.260 align:middle line:84% So imagine that basically, there's 00:03:39.260 --> 00:03:41.030 align:middle line:84% two household members, a man and a woman, 00:03:41.030 --> 00:03:43.910 align:middle line:90% just as an example, right? 00:03:43.910 --> 00:03:49.850 align:middle line:84% So the utility of each is going to be um and uf, where 00:03:49.850 --> 00:03:52.040 align:middle line:84% qm and qf are the private consumption 00:03:52.040 --> 00:03:54.290 align:middle line:90% of the man and the woman. 00:03:54.290 --> 00:03:56.400 align:middle line:90% And Q is public consumption, OK? 00:03:56.400 --> 00:03:58.910 align:middle line:84% So note, for example, this is very general, right? 00:03:58.910 --> 00:04:03.260 align:middle line:84% This can allow me to care about wife's consumption and vice 00:04:03.260 --> 00:04:04.200 align:middle line:90% versa, OK? 00:04:04.200 --> 00:04:06.840 align:middle line:84% But these are individual utility functions. 00:04:06.840 --> 00:04:08.540 align:middle line:84% So we have some vector of consumption q 00:04:08.540 --> 00:04:09.890 align:middle line:90% and some vector of prices p. 00:04:09.890 --> 00:04:14.100 align:middle line:84% We also can allow the prices paid to be different, maybe. 00:04:14.100 --> 00:04:17.642 align:middle line:84% We set up some bargaining weights, lambda-- 00:04:17.642 --> 00:04:19.600 align:middle line:84% by the way, if I'm going too fast, let me know. 00:04:19.600 --> 00:04:21.779 align:middle line:84% And also, the slides should all be posted on Canvas. 00:04:21.779 --> 00:04:24.036 align:middle line:90% You should be able to have them. 00:04:24.036 --> 00:04:26.790 align:middle line:84% OK, so we saved some bargaining weights-- 00:04:26.790 --> 00:04:29.610 align:middle line:90% lambda m, lambda f. 00:04:29.610 --> 00:04:33.780 align:middle line:84% So the household utility is just given by the sum 00:04:33.780 --> 00:04:36.720 align:middle line:90% of the utilities, right? 00:04:36.720 --> 00:04:38.610 align:middle line:84% So the household together is going 00:04:38.610 --> 00:04:40.890 align:middle line:84% to maximize some of these utilities subject 00:04:40.890 --> 00:04:43.835 align:middle line:90% to this joint budget constraint. 00:04:43.835 --> 00:04:45.210 align:middle line:84% And the joint budget constraint-- 00:04:45.210 --> 00:04:47.880 align:middle line:84% we're going to have the total consumption 00:04:47.880 --> 00:04:49.680 align:middle line:90% is equal to the total incomes. 00:04:49.680 --> 00:04:53.250 align:middle line:84% So note a couple of things that are important about this very 00:04:53.250 --> 00:04:54.330 align:middle line:90% simple setup, right? 00:04:54.330 --> 00:04:58.890 align:middle line:84% One is the incomes just add together here 00:04:58.890 --> 00:05:00.630 align:middle line:90% on the right-hand side, right? 00:05:00.630 --> 00:05:02.820 align:middle line:84% There's nothing special about husband's income 00:05:02.820 --> 00:05:03.570 align:middle line:90% or wife's income. 00:05:03.570 --> 00:05:06.600 align:middle line:84% They're all just-- it goes into this collective income, right? 00:05:06.600 --> 00:05:08.490 align:middle line:84% And then they bargain over it, right? 00:05:08.490 --> 00:05:11.242 align:middle line:84% But so this is going to be one of the tests 00:05:11.242 --> 00:05:12.450 align:middle line:90% that we're going to look for. 00:05:12.450 --> 00:05:16.050 align:middle line:84% That's going to be a test of income pooling, OK? 00:05:16.050 --> 00:05:17.700 align:middle line:84% The other thing that you should note 00:05:17.700 --> 00:05:23.860 align:middle line:84% is, tsh tsh tsh tsh tsh tsh tsh tsh tsh tsh, this point, which 00:05:23.860 --> 00:05:29.300 align:middle line:84% is if we hold the bargaining weights fixed, 00:05:29.300 --> 00:05:32.480 align:middle line:84% then actually, even though this model 00:05:32.480 --> 00:05:34.010 align:middle line:84% had two different individuals who 00:05:34.010 --> 00:05:38.600 align:middle line:84% are making their decisions separately, 00:05:38.600 --> 00:05:41.300 align:middle line:84% this is actually going to look just like the unitary household 00:05:41.300 --> 00:05:42.820 align:middle line:90% again, right? 00:05:42.820 --> 00:05:43.600 align:middle line:90% Why? 00:05:43.600 --> 00:05:48.370 align:middle line:84% Because if I hold the bargaining weights fixed, lambda m 00:05:48.370 --> 00:05:53.050 align:middle line:84% and lambda f, I can just define a new utility function, which 00:05:53.050 --> 00:06:00.030 align:middle line:84% is equal to the weighted sum of these utilities, right? 00:06:00.030 --> 00:06:05.130 align:middle line:84% And so this actually can be a challenge 00:06:05.130 --> 00:06:07.590 align:middle line:84% for testing these things in the data 00:06:07.590 --> 00:06:11.580 align:middle line:84% because they actually end up looking very similar. 00:06:11.580 --> 00:06:14.860 align:middle line:84% They can look very similar in a lot of contexts, right? 00:06:14.860 --> 00:06:16.690 align:middle line:84% So even though here we might have 00:06:16.690 --> 00:06:19.300 align:middle line:84% different individual utilities and people 00:06:19.300 --> 00:06:22.390 align:middle line:84% care about different things, if the bargaining weights are 00:06:22.390 --> 00:06:27.650 align:middle line:84% fixed, then it ends up looking a lot like the unitary household. 00:06:27.650 --> 00:06:29.485 align:middle line:90% So what do we do with this? 00:06:29.485 --> 00:06:35.550 align:middle line:84% OK, so basically, this is going to lead to two types of tests 00:06:35.550 --> 00:06:38.520 align:middle line:84% for whether or not we should be thinking of a unitary household 00:06:38.520 --> 00:06:40.060 align:middle line:90% decision-making. 00:06:40.060 --> 00:06:43.230 align:middle line:84% The first kind of test is going to come 00:06:43.230 --> 00:06:47.410 align:middle line:84% from this income pooling idea, which is to say, 00:06:47.410 --> 00:06:51.060 align:middle line:90% does income from different-- 00:06:51.060 --> 00:06:55.180 align:middle line:84% if, say, the man or the spouse, or one spouse 00:06:55.180 --> 00:06:58.530 align:middle line:84% or the other spouse gets income, does that 00:06:58.530 --> 00:07:01.140 align:middle line:84% affect the patterns of consumption, right? 00:07:01.140 --> 00:07:04.390 align:middle line:84% And the income pooling test says it shouldn't. 00:07:04.390 --> 00:07:06.140 align:middle line:84% Second thing that we're going to come from 00:07:06.140 --> 00:07:09.275 align:middle line:84% is we're going to say, well, maybe 00:07:09.275 --> 00:07:10.650 align:middle line:84% we can think about something that 00:07:10.650 --> 00:07:12.360 align:middle line:90% affects the bargaining weights. 00:07:12.360 --> 00:07:16.260 align:middle line:84% And if I can think about different types 00:07:16.260 --> 00:07:19.470 align:middle line:84% of interventions, or policies, or things 00:07:19.470 --> 00:07:23.310 align:middle line:84% like that that are going to affect the lambda m's 00:07:23.310 --> 00:07:26.610 align:middle line:84% and lambda f's, and I say, well, if I change something that 00:07:26.610 --> 00:07:28.800 align:middle line:84% looks like it's going to increase the, 00:07:28.800 --> 00:07:31.890 align:middle line:84% say, the wife's bargaining power in a variety of ways, 00:07:31.890 --> 00:07:34.625 align:middle line:84% does that change the pattern of consumption? 00:07:34.625 --> 00:07:37.382 align:middle line:84% OK, and that's also going to be a test of this. 00:07:37.382 --> 00:07:41.540 align:middle line:84% OK, so those are going to be two classes of tests. 00:07:41.540 --> 00:07:44.327 align:middle line:84% The third class of test is going to be the one in the Udry 00:07:44.327 --> 00:07:46.910 align:middle line:84% example, which is going to be a Pareto efficiency, which we'll 00:07:46.910 --> 00:07:48.490 align:middle line:90% talk about later. 00:07:48.490 --> 00:07:52.660 align:middle line:84% OK, so that's just the basic setup. 00:07:52.660 --> 00:07:54.370 align:middle line:84% I went through that really fast, I think. 00:07:54.370 --> 00:07:55.765 align:middle line:90% But is that clear? 00:07:55.765 --> 00:07:56.890 align:middle line:90% Do you guys have questions? 00:07:56.890 --> 00:08:00.210 align:middle line:84% You should stop me if I'm going too fast. 00:08:00.210 --> 00:08:02.490 align:middle line:90% Good? 00:08:02.490 --> 00:08:03.720 align:middle line:90% OK. 00:08:03.720 --> 00:08:08.365 align:middle line:84% So I think this is basically what I just said. 00:08:08.365 --> 00:08:10.870 align:middle line:90% 00:08:10.870 --> 00:08:14.975 align:middle line:90% So here are some examples of-- 00:08:14.975 --> 00:08:17.940 align:middle line:84% so I said we're not going to observe the bargaining weights. 00:08:17.940 --> 00:08:23.050 align:middle line:84% We're going to make assumptions about them. 00:08:23.050 --> 00:08:25.592 align:middle line:84% And this is the income pooling test I already talked about. 00:08:25.592 --> 00:08:26.092 align:middle line:90% OK. 00:08:26.092 --> 00:08:31.570 align:middle line:90% 00:08:31.570 --> 00:08:35.190 align:middle line:84% Oh, the other thing which is going to be relevant for making 00:08:35.190 --> 00:08:37.530 align:middle line:84% these tests, for the empirical tests we're going to talk 00:08:37.530 --> 00:08:42.470 align:middle line:84% about, is we're going to have to go back to these q's. 00:08:42.470 --> 00:08:48.320 align:middle line:90% So we had qf and qm. 00:08:48.320 --> 00:08:50.540 align:middle line:84% Technically, if you just look at this model, 00:08:50.540 --> 00:08:55.850 align:middle line:84% the identity of what is a qf and what is a qm is irrelevant. 00:08:55.850 --> 00:09:00.170 align:middle line:84% Imagine I had a good exogenous shock in bargaining weights. 00:09:00.170 --> 00:09:03.950 align:middle line:84% If that changes the distribution of the qf's and the qm's, I 00:09:03.950 --> 00:09:07.160 align:middle line:84% can already reject a single unitary house model, right? 00:09:07.160 --> 00:09:11.330 align:middle line:84% It doesn't matter what the different q's are. 00:09:11.330 --> 00:09:13.752 align:middle line:84% But it might be nice to be a little more precise 00:09:13.752 --> 00:09:14.960 align:middle line:90% about our predictions, right? 00:09:14.960 --> 00:09:17.240 align:middle line:84% If we had certain ideas that there 00:09:17.240 --> 00:09:20.720 align:middle line:84% were, quote "women's goods" or quote "men's goods," 00:09:20.720 --> 00:09:24.140 align:middle line:84% then we might be able to generate a more specific test, 00:09:24.140 --> 00:09:26.540 align:middle line:84% a more directional test, which is to say, oh, 00:09:26.540 --> 00:09:28.832 align:middle line:84% if we have something that increases women in bargaining 00:09:28.832 --> 00:09:31.250 align:middle line:84% power and we know that women like such and such, 00:09:31.250 --> 00:09:35.660 align:middle line:84% we can look at these particular types of goods. 00:09:35.660 --> 00:09:41.560 align:middle line:84% And I should say, actually, identifying 00:09:41.560 --> 00:09:43.480 align:middle line:90% that is hard, right? 00:09:43.480 --> 00:09:47.373 align:middle line:90% 00:09:47.373 --> 00:09:48.790 align:middle line:84% People make assumptions, sometimes 00:09:48.790 --> 00:09:52.240 align:middle line:84% about what are some goods that look like men's goods. 00:09:52.240 --> 00:09:55.670 align:middle line:84% But actually, there are many types 00:09:55.670 --> 00:09:58.590 align:middle line:84% of goods that we may have different preferences on, 00:09:58.590 --> 00:10:00.590 align:middle line:84% that men or women have different preferences on. 00:10:00.590 --> 00:10:02.132 align:middle line:84% A lot of the tests will be like that. 00:10:02.132 --> 00:10:03.680 align:middle line:84% If you want to look at the number, 00:10:03.680 --> 00:10:06.830 align:middle line:84% the amount of goods that are strictly consumed by men 00:10:06.830 --> 00:10:08.690 align:middle line:84% or strictly consumed by women, for example, 00:10:08.690 --> 00:10:10.550 align:middle line:84% that can actually be pretty small, right? 00:10:10.550 --> 00:10:12.140 align:middle line:84% There are examples like that that people have used, 00:10:12.140 --> 00:10:14.330 align:middle line:84% but there's not a huge set that looks like that. 00:10:14.330 --> 00:10:17.060 align:middle line:84% Plus, it also can be tricky if you have 00:10:17.060 --> 00:10:21.590 align:middle line:90% the altruistic utility, right? 00:10:21.590 --> 00:10:23.270 align:middle line:84% You might say, for example, maybe women 00:10:23.270 --> 00:10:24.260 align:middle line:90% in the household wear jewelry. 00:10:24.260 --> 00:10:25.400 align:middle line:84% And men tend not to wear jewelry. 00:10:25.400 --> 00:10:27.983 align:middle line:84% But of course, we all know that men can sometimes give jewelry 00:10:27.983 --> 00:10:29.550 align:middle line:90% to their wives, or vice versa. 00:10:29.550 --> 00:10:32.880 align:middle line:90% So it can be complicated. 00:10:32.880 --> 00:10:36.852 align:middle line:84% But if we do have these more specific tests, sometimes, 00:10:36.852 --> 00:10:38.310 align:middle line:84% people will take the view of, oh, I 00:10:38.310 --> 00:10:39.960 align:middle line:84% can identify the following as goods 00:10:39.960 --> 00:10:41.400 align:middle line:84% that are associated with particular types of household 00:10:41.400 --> 00:10:41.900 align:middle line:90% members. 00:10:41.900 --> 00:10:43.620 align:middle line:84% And I can generate a directional test. 00:10:43.620 --> 00:10:45.870 align:middle line:84% If the income of that person increases, 00:10:45.870 --> 00:10:48.240 align:middle line:84% do we get more of their consumption? 00:10:48.240 --> 00:10:51.480 align:middle line:84% That's a joint test of both income pooling 00:10:51.480 --> 00:10:58.350 align:middle line:90% and this different consumption. 00:10:58.350 --> 00:11:00.510 align:middle line:90% OK. 00:11:00.510 --> 00:11:02.380 align:middle line:84% So here are some examples, by the way, 00:11:02.380 --> 00:11:09.670 align:middle line:84% that people have used in the literature to think about some 00:11:09.670 --> 00:11:11.530 align:middle line:90% of these issues, right? 00:11:11.530 --> 00:11:12.490 align:middle line:90% And I'm curious. 00:11:12.490 --> 00:11:13.460 align:middle line:84% Actually, maybe we can pause for a sec. 00:11:13.460 --> 00:11:15.127 align:middle line:84% So here are different things people have 00:11:15.127 --> 00:11:17.290 align:middle line:84% used as either sources of variation 00:11:17.290 --> 00:11:19.207 align:middle line:84% in the bargaining power of the household-- 00:11:19.207 --> 00:11:21.790 align:middle line:84% so different amounts of earned income from different household 00:11:21.790 --> 00:11:24.340 align:middle line:84% members, say maybe the age gap, resources 00:11:24.340 --> 00:11:27.317 align:middle line:84% brought to the marriage, unearned income, divorce laws. 00:11:27.317 --> 00:11:28.900 align:middle line:84% There's a whole literature-- actually, 00:11:28.900 --> 00:11:32.980 align:middle line:84% it started, I think, in the US on changing from making divorce 00:11:32.980 --> 00:11:36.820 align:middle line:84% easier as a change in bargaining power, 00:11:36.820 --> 00:11:39.380 align:middle line:84% where one is about the scarcity in the marriage market. 00:11:39.380 --> 00:11:43.387 align:middle line:84% So what do you guys think of these as examples of things 00:11:43.387 --> 00:11:45.220 align:middle line:84% that you'd use to test for bargaining power? 00:11:45.220 --> 00:11:46.870 align:middle line:84% Do you have any thoughts on this? 00:11:46.870 --> 00:11:48.280 align:middle line:90% Seem sensible, not sensible? 00:11:48.280 --> 00:11:50.412 align:middle line:84% Are there other ones you'd put on this list? 00:11:50.412 --> 00:11:52.870 align:middle line:84% First of all, do you understand why these things might be-- 00:11:52.870 --> 00:11:55.320 align:middle line:90% why you might think of-- 00:11:55.320 --> 00:11:56.820 align:middle line:84% so someone give me an example of how 00:11:56.820 --> 00:11:58.487 align:middle line:84% you might think of one of these as being 00:11:58.487 --> 00:12:02.630 align:middle line:84% related to bargaining power in the example I just set up. 00:12:02.630 --> 00:12:05.130 align:middle line:90% 00:12:05.130 --> 00:12:05.630 align:middle line:90% Somebody. 00:12:05.630 --> 00:12:06.208 align:middle line:90% Yeah? 00:12:06.208 --> 00:12:07.250 align:middle line:90% Remind me of name, sorry. 00:12:07.250 --> 00:12:08.000 align:middle line:90% AUDIENCE: Whitney. 00:12:08.000 --> 00:12:09.230 align:middle line:90% BEN OLKEN: Whitney. 00:12:09.230 --> 00:12:11.222 align:middle line:84% AUDIENCE: So with the divorce law example, 00:12:11.222 --> 00:12:12.680 align:middle line:84% maybe you can think that if there's 00:12:12.680 --> 00:12:15.560 align:middle line:84% a law that makes it easier for women to get a divorce, 00:12:15.560 --> 00:12:18.320 align:middle line:84% then they'll have increased bargaining 00:12:18.320 --> 00:12:20.250 align:middle line:84% power in the household because they can say, 00:12:20.250 --> 00:12:22.250 align:middle line:84% if they don't like the way the things are going, 00:12:22.250 --> 00:12:23.480 align:middle line:90% they'll just leave. 00:12:23.480 --> 00:12:23.750 align:middle line:90% BEN OLKEN: Right. 00:12:23.750 --> 00:12:25.880 align:middle line:84% AUDIENCE: So they'll change the lambda that we have. 00:12:25.880 --> 00:12:26.390 align:middle line:90% BEN OLKEN: Yeah, exactly. 00:12:26.390 --> 00:12:28.015 align:middle line:84% So if you think about-- how many of you 00:12:28.015 --> 00:12:34.460 align:middle line:84% guys have seen bargaining models in game theory, or in micro? 00:12:34.460 --> 00:12:35.510 align:middle line:90% Some of you. 00:12:35.510 --> 00:12:36.930 align:middle line:90% Only some of you. 00:12:36.930 --> 00:12:37.430 align:middle line:90% Right. 00:12:37.430 --> 00:12:39.560 align:middle line:84% The way a typical bargaining model might work 00:12:39.560 --> 00:12:43.015 align:middle line:84% is to say we each get our outside option and some share 00:12:43.015 --> 00:12:43.640 align:middle line:90% of the surplus. 00:12:43.640 --> 00:12:46.370 align:middle line:84% That's the standard Nash bargaining solution, right? 00:12:46.370 --> 00:12:52.130 align:middle line:84% So imagine that basically, if divorce basically says 00:12:52.130 --> 00:12:56.060 align:middle line:84% my outside option has changed, if it's giving me a better 00:12:56.060 --> 00:12:57.500 align:middle line:84% outside option, if we don't agree 00:12:57.500 --> 00:12:59.792 align:middle line:84% on how we're going to share resources in this marriage, 00:12:59.792 --> 00:13:00.620 align:middle line:90% I can leave, right? 00:13:00.620 --> 00:13:03.320 align:middle line:84% If that differentially increases the outside option 00:13:03.320 --> 00:13:05.302 align:middle line:84% of one side versus the other side, 00:13:05.302 --> 00:13:07.760 align:middle line:84% you could think of that as an increase in bargaining power, 00:13:07.760 --> 00:13:08.260 align:middle line:90% right? 00:13:08.260 --> 00:13:11.090 align:middle line:84% But it would have to be a divorce law that-- 00:13:11.090 --> 00:13:15.410 align:middle line:90% 00:13:15.410 --> 00:13:18.020 align:middle line:84% what's complicated is that you have think 00:13:18.020 --> 00:13:20.630 align:middle line:84% of it as-- from the model's perspective, if you're 00:13:20.630 --> 00:13:24.380 align:middle line:84% both stuck in this marriage, it's not totally obvious 00:13:24.380 --> 00:13:29.800 align:middle line:84% how moving to unilateral divorce necessarily 00:13:29.800 --> 00:13:31.800 align:middle line:84% increases-- it definitely means that both sides' 00:13:31.800 --> 00:13:33.342 align:middle line:84% outside options become more relevant. 00:13:33.342 --> 00:13:38.963 align:middle line:84% But it's not obviously signed, I think. 00:13:38.963 --> 00:13:40.380 align:middle line:84% As an empirical matter, I think it 00:13:40.380 --> 00:13:42.760 align:middle line:84% does turn out to increase the woman's bargaining power. 00:13:42.760 --> 00:13:45.600 align:middle line:84% But I think from the model perspective, 00:13:45.600 --> 00:13:47.650 align:middle line:90% it's not totally obvious to me. 00:13:47.650 --> 00:13:48.400 align:middle line:90% Yeah, Wesley? 00:13:48.400 --> 00:13:49.615 align:middle line:84% AUDIENCE: Can I ask a similar question about the age gap? 00:13:49.615 --> 00:13:49.960 align:middle line:90% BEN OLKEN: Yeah. 00:13:49.960 --> 00:13:51.585 align:middle line:84% AUDIENCE: It's not entirely clear to me 00:13:51.585 --> 00:13:54.280 align:middle line:84% which direction you would expect that to go in bargaining power. 00:13:54.280 --> 00:13:58.420 align:middle line:84% I could envision maybe you would expect that a larger age 00:13:58.420 --> 00:14:03.635 align:middle line:84% gap causes an increase in the bargaining power of a younger 00:14:03.635 --> 00:14:05.260 align:middle line:84% individual in some circumstances, maybe 00:14:05.260 --> 00:14:07.430 align:middle line:84% of the older individual in other circumstances. 00:14:07.430 --> 00:14:14.920 align:middle line:84% What's the take that I guess the age gap, or literature 00:14:14.920 --> 00:14:15.890 align:middle line:90% to try to [INAUDIBLE]? 00:14:15.890 --> 00:14:17.265 align:middle line:84% BEN OLKEN: That's a good example. 00:14:17.265 --> 00:14:18.560 align:middle line:90% Actually, it's a good example. 00:14:18.560 --> 00:14:20.643 align:middle line:84% I'm trying to think of one off the top of my head. 00:14:20.643 --> 00:14:25.855 align:middle line:84% You might imagine that, I think, in a bigger age 00:14:25.855 --> 00:14:27.730 align:middle line:84% gap between spouses-- particularly if spouses 00:14:27.730 --> 00:14:30.910 align:middle line:84% are getting married where one of them is very young, 00:14:30.910 --> 00:14:34.060 align:middle line:84% you might imagine that they're not as good bargainers, 00:14:34.060 --> 00:14:36.220 align:middle line:84% or as in a strong bargaining position. 00:14:36.220 --> 00:14:40.540 align:middle line:84% Their outside option may not be as good if they can't leave. 00:14:40.540 --> 00:14:42.460 align:middle line:84% Or you may just imagine that-- yeah, 00:14:42.460 --> 00:14:44.168 align:middle line:84% I think that's probably what I would say. 00:14:44.168 --> 00:14:46.520 align:middle line:90% 00:14:46.520 --> 00:14:47.750 align:middle line:90% Other comments? 00:14:47.750 --> 00:14:49.450 align:middle line:90% Other thoughts? 00:14:49.450 --> 00:14:50.050 align:middle line:90% Yeah, Ahmed. 00:14:50.050 --> 00:14:51.920 align:middle line:84% AUDIENCE: [INAUDIBLE] following up on that. 00:14:51.920 --> 00:14:53.580 align:middle line:90% We don't know the-- 00:14:53.580 --> 00:14:59.230 align:middle line:84% or canwe say-- can we [? spot ?] the age gap, for instance? 00:14:59.230 --> 00:15:01.870 align:middle line:84% Because I've heard stories about-- it can go either way. 00:15:01.870 --> 00:15:02.480 align:middle line:90% It can be-- 00:15:02.480 --> 00:15:05.230 align:middle line:84% BEN OLKEN: So give me an example of a story where it goes-- 00:15:05.230 --> 00:15:08.773 align:middle line:84% where a larger age gap-- so what do you mean? 00:15:08.773 --> 00:15:13.090 align:middle line:84% AUDIENCE: I would assume, say, if it's a 30-year-old worse 00:15:13.090 --> 00:15:15.310 align:middle line:84% than 18-year-old, I would assume that age gap would 00:15:15.310 --> 00:15:18.470 align:middle line:90% be in favor of the 30-year-old. 00:15:18.470 --> 00:15:21.130 align:middle line:84% But if it's, say, a 30-year-old versus a 75-year-old, 00:15:21.130 --> 00:15:24.123 align:middle line:84% I would assume, at some point, it should go the other way. 00:15:24.123 --> 00:15:24.790 align:middle line:90% BEN OLKEN: Yeah. 00:15:24.790 --> 00:15:28.805 align:middle line:84% So actually, so here's my basic takeaway from all this. 00:15:28.805 --> 00:15:30.430 align:middle line:84% The way I would want to think about all 00:15:30.430 --> 00:15:32.800 align:middle line:84% of this is to say, let's set up a bargaining model. 00:15:32.800 --> 00:15:35.950 align:middle line:84% And let's think about how these things affect your bargaining 00:15:35.950 --> 00:15:36.933 align:middle line:90% position. 00:15:36.933 --> 00:15:38.350 align:middle line:84% A simple bargaining model where we 00:15:38.350 --> 00:15:40.780 align:middle line:84% each get our outside option plus some share of the surplus, 00:15:40.780 --> 00:15:41.280 align:middle line:90% right? 00:15:41.280 --> 00:15:44.500 align:middle line:84% That's not necessarily formally the same. 00:15:44.500 --> 00:15:46.750 align:middle line:84% But that's how I would think about changing this. 00:15:46.750 --> 00:15:50.350 align:middle line:84% And I think the question is, how would you 00:15:50.350 --> 00:15:52.133 align:middle line:90% think about each of these? 00:15:52.133 --> 00:15:53.800 align:middle line:84% Which of these are really clear in that? 00:15:53.800 --> 00:15:55.270 align:middle line:84% And I agree with you, it's not totally obvious. 00:15:55.270 --> 00:15:56.812 align:middle line:84% Maybe the 75-year-old, you're saying, 00:15:56.812 --> 00:16:00.230 align:middle line:90% has no outside option, right? 00:16:00.230 --> 00:16:04.210 align:middle line:84% But which of these do you think actually really 00:16:04.210 --> 00:16:07.060 align:middle line:84% does map very clearly into that threshold? 00:16:07.060 --> 00:16:07.870 align:middle line:90% Yeah, Patrick? 00:16:07.870 --> 00:16:08.930 align:middle line:90% AUDIENCE: Earned income. 00:16:08.930 --> 00:16:09.460 align:middle line:90% BEN OLKEN: Which one? 00:16:09.460 --> 00:16:10.150 align:middle line:90% AUDIENCE: Earned income. 00:16:10.150 --> 00:16:11.192 align:middle line:90% BEN OLKEN: Earned income? 00:16:11.192 --> 00:16:12.160 align:middle line:90% So say why. 00:16:12.160 --> 00:16:13.570 align:middle line:84% AUDIENCE: You make the money, you call the shots. 00:16:13.570 --> 00:16:13.872 align:middle line:90% BEN OLKEN: What? 00:16:13.872 --> 00:16:14.480 align:middle line:90% Sorry? 00:16:14.480 --> 00:16:16.280 align:middle line:84% AUDIENCE: You make the money, you call the shots. 00:16:16.280 --> 00:16:17.980 align:middle line:84% BEN OLKEN: Yeah, but try to put that into the bargaining 00:16:17.980 --> 00:16:18.500 align:middle line:90% perspective. 00:16:18.500 --> 00:16:21.000 align:middle line:84% So imagine-- how do you think about that from the bargaining 00:16:21.000 --> 00:16:21.648 align:middle line:90% perspective? 00:16:21.648 --> 00:16:23.690 align:middle line:84% So why is it the case that if you earn the money, 00:16:23.690 --> 00:16:24.580 align:middle line:90% you earn the shots? 00:16:24.580 --> 00:16:25.122 align:middle line:90% AUDIENCE: OK. 00:16:25.122 --> 00:16:26.120 align:middle line:90% So if I try to-- 00:16:26.120 --> 00:16:26.930 align:middle line:90% BEN OLKEN: You call the shots. 00:16:26.930 --> 00:16:29.138 align:middle line:84% AUDIENCE: --hold my money, or if I started bargaining 00:16:29.138 --> 00:16:36.270 align:middle line:84% with my money, I expect the utility out of this bargain 00:16:36.270 --> 00:16:42.610 align:middle line:84% model, having given in some of that into the pooling. 00:16:42.610 --> 00:16:45.173 align:middle line:84% It comes with a higher opportunity cost. 00:16:45.173 --> 00:16:45.840 align:middle line:90% BEN OLKEN: Yeah. 00:16:45.840 --> 00:16:47.882 align:middle line:84% But let's try to be a little more precise, right? 00:16:47.882 --> 00:16:49.620 align:middle line:84% So imagine that basically, we have the-- 00:16:49.620 --> 00:16:52.710 align:middle line:84% let's call it the outside option of the man 00:16:52.710 --> 00:16:54.930 align:middle line:84% and the outside option of the woman. 00:16:54.930 --> 00:16:58.020 align:middle line:90% And we have, for example-- 00:16:58.020 --> 00:17:01.050 align:middle line:84% let's imagine we're going to have some, in this case, fixed 00:17:01.050 --> 00:17:05.430 align:middle line:90% bargaining weight alpha, right? 00:17:05.430 --> 00:17:07.480 align:middle line:84% So what's going to happen is they're going to-- 00:17:07.480 --> 00:17:10.710 align:middle line:84% and if they can agree on everything together, 00:17:10.710 --> 00:17:13.680 align:middle line:84% they get some joint surplus s, right? 00:17:13.680 --> 00:17:19.319 align:middle line:84% So the man's outcome is going to be s 00:17:19.319 --> 00:17:27.800 align:middle line:84% minus om minus of times alpha plus om. 00:17:27.800 --> 00:17:34.440 align:middle line:84% And the woman's outcome is going to be 1 minus theta s 00:17:34.440 --> 00:17:39.090 align:middle line:84% minus om minus of plus of, right? 00:17:39.090 --> 00:17:41.340 align:middle line:84% So this is the standard Nash bargaining solution 00:17:41.340 --> 00:17:47.910 align:middle line:84% with a constant, with simple case constant fixed bargaining 00:17:47.910 --> 00:17:51.870 align:middle line:84% weight alpha and a surplus from agreement that's 00:17:51.870 --> 00:17:55.590 align:middle line:90% just fixed or whatever, OK? 00:17:55.590 --> 00:17:57.540 align:middle line:84% So this is a simple setup I was describing. 00:17:57.540 --> 00:18:00.270 align:middle line:84% That basically says if we disagree 00:18:00.270 --> 00:18:03.720 align:middle line:84% and can't agree on anything, the man gets of, 00:18:03.720 --> 00:18:04.920 align:middle line:90% the woman gets of-- 00:18:04.920 --> 00:18:06.960 align:middle line:84% sorry, the man gets om, the woman gets of. 00:18:06.960 --> 00:18:11.310 align:middle line:84% If we agree, then they each get their outside option-- om 00:18:11.310 --> 00:18:12.460 align:middle line:90% and of, respectively. 00:18:12.460 --> 00:18:14.790 align:middle line:84% And then they just share the surplus from agreement. 00:18:14.790 --> 00:18:17.260 align:middle line:84% It's s minus each of their outside options. 00:18:17.260 --> 00:18:21.250 align:middle line:84% And they just share that alpha 1 minus alpha, OK? 00:18:21.250 --> 00:18:22.887 align:middle line:84% So that's the simple bargaining setup. 00:18:22.887 --> 00:18:25.470 align:middle line:84% So Patrick, going back to your example, how do you think about 00:18:25.470 --> 00:18:26.720 align:middle line:90% earned income in this example? 00:18:26.720 --> 00:18:30.980 align:middle line:90% 00:18:30.980 --> 00:18:34.180 align:middle line:84% AUDIENCE: I would think that, for example, the [INAUDIBLE] 00:18:34.180 --> 00:18:35.620 align:middle line:90% much higher income. 00:18:35.620 --> 00:18:36.940 align:middle line:84% BEN OLKEN: Sorry, you've got to speak up because of the mask. 00:18:36.940 --> 00:18:37.780 align:middle line:90% It's hard for me to hear you. 00:18:37.780 --> 00:18:40.240 align:middle line:84% AUDIENCE: The male has a higher income [INAUDIBLE].. 00:18:40.240 --> 00:18:42.160 align:middle line:84% BEN OLKEN: Or one spouse, we can call it. 00:18:42.160 --> 00:18:44.080 align:middle line:84% AUDIENCE: OK, one spouse has higher income. 00:18:44.080 --> 00:18:45.070 align:middle line:90% om is much larger. 00:18:45.070 --> 00:18:47.910 align:middle line:90% 00:18:47.910 --> 00:18:50.900 align:middle line:84% So if om is much larger, then the [INAUDIBLE] 00:18:50.900 --> 00:18:54.763 align:middle line:90% would be relatively smaller. 00:18:54.763 --> 00:18:55.430 align:middle line:90% BEN OLKEN: Yeah. 00:18:55.430 --> 00:18:58.190 align:middle line:84% So by the way, I'm not being-- let me just say, 00:18:58.190 --> 00:19:00.140 align:middle line:84% I'm not being totally precise here. 00:19:00.140 --> 00:19:01.790 align:middle line:84% Because in the model I had over there, 00:19:01.790 --> 00:19:03.530 align:middle line:84% I had fixed bargaining weights over-- 00:19:03.530 --> 00:19:06.032 align:middle line:84% I had fixed bargaining weights over the utilities. 00:19:06.032 --> 00:19:08.240 align:middle line:84% And here I'm actually having fixed bargaining weights 00:19:08.240 --> 00:19:09.782 align:middle line:84% and they're just sharing the surplus. 00:19:09.782 --> 00:19:12.360 align:middle line:84% So this is not a precise mapping 1 to 1. 00:19:12.360 --> 00:19:15.472 align:middle line:84% And if I had thought this through more clearly 00:19:15.472 --> 00:19:17.180 align:middle line:84% in advance, I probably would have come up 00:19:17.180 --> 00:19:19.580 align:middle line:84% with a micro [INAUDIBLE] of that model over there with this one. 00:19:19.580 --> 00:19:21.980 align:middle line:84% But I didn't think it through in advance clearly enough. 00:19:21.980 --> 00:19:24.087 align:middle line:90% So my apologies. 00:19:24.087 --> 00:19:25.670 align:middle line:84% But anyway, you can see, yes, exactly. 00:19:25.670 --> 00:19:28.490 align:middle line:84% So imagine you have more unearned income-- 00:19:28.490 --> 00:19:30.230 align:middle line:90% more earned income, om. 00:19:30.230 --> 00:19:31.590 align:middle line:90% What does that basically mean? 00:19:31.590 --> 00:19:33.715 align:middle line:84% It means that in the event of the world in which we 00:19:33.715 --> 00:19:37.400 align:middle line:84% can't agree, I can go take my earned income 00:19:37.400 --> 00:19:38.840 align:middle line:90% and live a rich life. 00:19:38.840 --> 00:19:42.230 align:middle line:84% And if my spouse has a lower earned income, 00:19:42.230 --> 00:19:44.300 align:middle line:84% my spouse is going to go off and be poor. 00:19:44.300 --> 00:19:45.920 align:middle line:84% That basically means that I'm going 00:19:45.920 --> 00:19:53.130 align:middle line:84% to be able to extract more from the joint, 00:19:53.130 --> 00:19:55.980 align:middle line:90% more from the relationship. 00:19:55.980 --> 00:20:00.110 align:middle line:84% And in this very simple model, this 00:20:00.110 --> 00:20:08.420 align:middle line:84% is not an income pooling model because the amount-- 00:20:08.420 --> 00:20:10.250 align:middle line:84% imagine that basically, this is the amount 00:20:10.250 --> 00:20:15.900 align:middle line:84% that the man, the utility that's going to be received by the man 00:20:15.900 --> 00:20:18.210 align:middle line:84% is going to depend on what fraction of the total income 00:20:18.210 --> 00:20:20.122 align:middle line:84% the man is bringing to the marriage and vice 00:20:20.122 --> 00:20:20.955 align:middle line:90% versa for the woman. 00:20:20.955 --> 00:20:24.680 align:middle line:90% 00:20:24.680 --> 00:20:27.400 align:middle line:90% Is that clear? 00:20:27.400 --> 00:20:27.900 align:middle line:90% OK. 00:20:27.900 --> 00:20:31.860 align:middle line:84% So I guess my point in having you think through this 00:20:31.860 --> 00:20:36.610 align:middle line:84% is when you see this stuff, I agree this one is, I think, 00:20:36.610 --> 00:20:37.830 align:middle line:90% pretty clear. 00:20:37.830 --> 00:20:39.672 align:middle line:84% Unearned income also-- well, actually, no. 00:20:39.672 --> 00:20:41.130 align:middle line:84% Earned income is very clear, right? 00:20:41.130 --> 00:20:43.680 align:middle line:84% Because if we disagree, I can go off in the labor market 00:20:43.680 --> 00:20:45.513 align:middle line:84% and continue earning that income separately. 00:20:45.513 --> 00:20:47.940 align:middle line:84% So it's very, very easily mapped to my outside option. 00:20:47.940 --> 00:20:49.660 align:middle line:84% How do you think about unearned income? 00:20:49.660 --> 00:20:51.827 align:middle line:84% So first of all, what do we mean by unearned income? 00:20:51.827 --> 00:20:54.320 align:middle line:90% 00:20:54.320 --> 00:20:55.410 align:middle line:90% AUDIENCE: Wealth? 00:20:55.410 --> 00:20:56.172 align:middle line:90% BEN OLKEN: Huh? 00:20:56.172 --> 00:20:57.410 align:middle line:90% AUDIENCE: Maybe wealth? 00:20:57.410 --> 00:20:58.705 align:middle line:90% BEN OLKEN: Wealth, you said? 00:20:58.705 --> 00:20:59.330 align:middle line:90% AUDIENCE: Yeah. 00:20:59.330 --> 00:21:01.323 align:middle line:90% BEN OLKEN: But what do you mean? 00:21:01.323 --> 00:21:02.240 align:middle line:90% AUDIENCE: Your estate. 00:21:02.240 --> 00:21:03.123 align:middle line:90% BEN OLKEN: Huh? 00:21:03.123 --> 00:21:04.040 align:middle line:90% AUDIENCE: Your estate. 00:21:04.040 --> 00:21:05.210 align:middle line:90% BEN OLKEN: Your estate? 00:21:05.210 --> 00:21:05.300 align:middle line:90% AUDIENCE: Yeah. 00:21:05.300 --> 00:21:07.220 align:middle line:84% BEN OLKEN: Yeah, meaning what you inherit, for example. 00:21:07.220 --> 00:21:07.720 align:middle line:90% Yeah. 00:21:07.720 --> 00:21:09.350 align:middle line:90% So that's right. 00:21:09.350 --> 00:21:11.700 align:middle line:90% So OK. 00:21:11.700 --> 00:21:15.190 align:middle line:84% Someday, my computer will come back. 00:21:15.190 --> 00:21:15.690 align:middle line:90% Uh-oh. 00:21:15.690 --> 00:21:18.430 align:middle line:90% 00:21:18.430 --> 00:21:20.230 align:middle line:90% So what do you mean by-- 00:21:20.230 --> 00:21:22.990 align:middle line:84% AUDIENCE: So I guess it would depend on 00:21:22.990 --> 00:21:30.187 align:middle line:84% whether you've got a prenup, whether your unearned income is 00:21:30.187 --> 00:21:32.604 align:middle line:84% not going to be split between the two of you when you have 00:21:32.604 --> 00:21:33.460 align:middle line:90% that-- there's divorce. 00:21:33.460 --> 00:21:34.543 align:middle line:90% BEN OLKEN: Exactly, right? 00:21:34.543 --> 00:21:36.190 align:middle line:84% So divorce law, for example, differs. 00:21:36.190 --> 00:21:37.953 align:middle line:84% In the US, for example, in some states-- 00:21:37.953 --> 00:21:40.120 align:middle line:84% and this certainly also differs by country-- we have 00:21:40.120 --> 00:21:43.690 align:middle line:84% what's called community property, which basically means 00:21:43.690 --> 00:21:46.570 align:middle line:84% everything is shared equally regardless 00:21:46.570 --> 00:21:48.100 align:middle line:90% of where it came from. 00:21:48.100 --> 00:21:50.590 align:middle line:84% But in other places, they keep track and say, 00:21:50.590 --> 00:21:52.957 align:middle line:84% if you were to divorce, you would get the-- 00:21:52.957 --> 00:21:55.540 align:middle line:84% but I came to the marriage with a lot more than my spouse did. 00:21:55.540 --> 00:21:57.280 align:middle line:90% Then I would get more-- 00:21:57.280 --> 00:21:58.550 align:middle line:90% I would get that back. 00:21:58.550 --> 00:22:01.690 align:middle line:84% So I think this unearned income-- 00:22:01.690 --> 00:22:02.947 align:middle line:90% you have to think of it-- 00:22:02.947 --> 00:22:04.780 align:middle line:84% if you want to use it from this perspective, 00:22:04.780 --> 00:22:06.530 align:middle line:84% you have to understand what would actually 00:22:06.530 --> 00:22:09.400 align:middle line:84% happen in the event of a divorce, how that would affect 00:22:09.400 --> 00:22:10.720 align:middle line:90% your bargaining power, right? 00:22:10.720 --> 00:22:12.040 align:middle line:90% And it's not obvious it would-- 00:22:12.040 --> 00:22:14.260 align:middle line:84% if you're in a world where the states splits stuff up equally, 00:22:14.260 --> 00:22:15.130 align:middle line:90% then it's necessarily clear. 00:22:15.130 --> 00:22:16.210 align:middle line:90% It wouldn't matter. 00:22:16.210 --> 00:22:18.790 align:middle line:84% So we're going to look at a paper in a second that 00:22:18.790 --> 00:22:21.040 align:middle line:84% looks at unearned income as opposed to earned income. 00:22:21.040 --> 00:22:24.580 align:middle line:84% Why might you prefer to do unearned income 00:22:24.580 --> 00:22:25.840 align:middle line:90% rather than earned income? 00:22:25.840 --> 00:22:28.062 align:middle line:84% If the empirical test is clearer for earned income, 00:22:28.062 --> 00:22:30.520 align:middle line:84% why do you think these papers might look at unearned income 00:22:30.520 --> 00:22:31.020 align:middle line:90% instead? 00:22:31.020 --> 00:22:33.500 align:middle line:90% 00:22:33.500 --> 00:22:34.000 align:middle line:90% Yeah. 00:22:34.000 --> 00:22:36.880 align:middle line:84% AUDIENCE: [INAUDIBLE] from the [INAUDIBLE] on somebody? 00:22:36.880 --> 00:22:38.130 align:middle line:90% BEN OLKEN: OK, exactly, right? 00:22:38.130 --> 00:22:39.922 align:middle line:84% So certainly, you can drop a cash transfer. 00:22:39.922 --> 00:22:41.712 align:middle line:84% And if I get to all the slides here, 00:22:41.712 --> 00:22:43.920 align:middle line:84% I have at least one paper on dropping cash transfers. 00:22:43.920 --> 00:22:44.880 align:middle line:90% So yes, exactly. 00:22:44.880 --> 00:22:48.302 align:middle line:84% But suppose it was wealth, right? 00:22:48.302 --> 00:22:50.135 align:middle line:84% It was the wealth you bring to the marriage. 00:22:50.135 --> 00:22:55.720 align:middle line:90% 00:22:55.720 --> 00:22:56.560 align:middle line:90% Yeah, Wes? 00:22:56.560 --> 00:22:57.820 align:middle line:84% AUDIENCE: Earned income assumes that everyone's 00:22:57.820 --> 00:22:58.770 align:middle line:90% in the labor market. 00:22:58.770 --> 00:22:59.260 align:middle line:90% BEN OLKEN: What, sorry? 00:22:59.260 --> 00:22:59.620 align:middle line:90% Earning 00:22:59.620 --> 00:23:00.400 align:middle line:84% AUDIENCE: Earned income would assume 00:23:00.400 --> 00:23:01.750 align:middle line:90% that both parties are working. 00:23:01.750 --> 00:23:04.137 align:middle line:90% In order to earn some income-- 00:23:04.137 --> 00:23:05.220 align:middle line:90% BEN OLKEN: Yeah, so right. 00:23:05.220 --> 00:23:06.303 align:middle line:90% So this is actually right. 00:23:06.303 --> 00:23:08.820 align:middle line:84% So as I think you were saying, this is totally endogenous. 00:23:08.820 --> 00:23:09.870 align:middle line:90% Earned income is endogenous. 00:23:09.870 --> 00:23:12.270 align:middle line:84% What you actually need is a measure of your earning power, 00:23:12.270 --> 00:23:12.770 align:middle line:90% right? 00:23:12.770 --> 00:23:16.920 align:middle line:84% So for example, I could have great earning power as an MIT 00:23:16.920 --> 00:23:19.865 align:middle line:84% professor but be choosing not to work, for example. 00:23:19.865 --> 00:23:21.240 align:middle line:84% What we actually want to know is, 00:23:21.240 --> 00:23:22.710 align:middle line:84% what's my outside option if I were to go back 00:23:22.710 --> 00:23:23.627 align:middle line:90% into the labor market? 00:23:23.627 --> 00:23:25.210 align:middle line:84% What would my earned income be, right? 00:23:25.210 --> 00:23:26.980 align:middle line:84% So you don't want actual earned income. 00:23:26.980 --> 00:23:31.410 align:middle line:84% But you want something proxy for earned income. 00:23:31.410 --> 00:23:33.900 align:middle line:84% For unearned income, what's the problem 00:23:33.900 --> 00:23:36.540 align:middle line:84% with this stuff, with resources brought to the marriage, 00:23:36.540 --> 00:23:38.320 align:middle line:90% for example? 00:23:38.320 --> 00:23:40.900 align:middle line:84% What's an empirical challenge, if you wanted to look at that? 00:23:40.900 --> 00:23:45.560 align:middle line:90% 00:23:45.560 --> 00:23:48.493 align:middle line:84% AUDIENCE: You might not have data at all. 00:23:48.493 --> 00:23:50.160 align:middle line:84% BEN OLKEN: Well, that's definitely true. 00:23:50.160 --> 00:23:51.580 align:middle line:90% [LAUGHTER] 00:23:51.580 --> 00:23:52.500 align:middle line:90% But suppose you did. 00:23:52.500 --> 00:23:57.500 align:middle line:90% 00:23:57.500 --> 00:23:58.730 align:middle line:90% So it seems exogenous, right? 00:23:58.730 --> 00:24:00.917 align:middle line:90% I don't control what I inherit. 00:24:00.917 --> 00:24:01.664 align:middle line:90% That's true. 00:24:01.664 --> 00:24:10.140 align:middle line:90% 00:24:10.140 --> 00:24:13.650 align:middle line:84% AUDIENCE: [INAUDIBLE] to come dictate who you [INAUDIBLE].. 00:24:13.650 --> 00:24:15.180 align:middle line:90% BEN OLKEN: Absolutely, right? 00:24:15.180 --> 00:24:15.990 align:middle line:90% So exactly. 00:24:15.990 --> 00:24:19.860 align:middle line:84% So we're not going to talk a lot about marriage markets 00:24:19.860 --> 00:24:20.370 align:middle line:90% with prices. 00:24:20.370 --> 00:24:21.930 align:middle line:84% But certainly, there are certainly 00:24:21.930 --> 00:24:23.370 align:middle line:84% lots of marriage markets in the world, 00:24:23.370 --> 00:24:24.240 align:middle line:84% particularly in the developing world. 00:24:24.240 --> 00:24:27.030 align:middle line:84% Explicitly, there are dowries and bride prices, right? 00:24:27.030 --> 00:24:33.085 align:middle line:84% And so there's an explicit marriage market 00:24:33.085 --> 00:24:33.960 align:middle line:90% that looks like that. 00:24:33.960 --> 00:24:35.460 align:middle line:84% Even without that, the marriage market, 00:24:35.460 --> 00:24:37.200 align:middle line:84% the matching market is pretty complicated. 00:24:37.200 --> 00:24:38.400 align:middle line:84% And people may have different preferences 00:24:38.400 --> 00:24:39.600 align:middle line:84% in terms of who they're going to match with. 00:24:39.600 --> 00:24:41.370 align:middle line:84% And those things may be related to things 00:24:41.370 --> 00:24:42.985 align:middle line:84% that look like your determinants, 00:24:42.985 --> 00:24:44.610 align:middle line:84% look like your consumption preferences. 00:24:44.610 --> 00:24:44.940 align:middle line:90% Yeah. 00:24:44.940 --> 00:24:47.148 align:middle line:84% AUDIENCE: So the concern here would be, for instance, 00:24:47.148 --> 00:24:51.450 align:middle line:84% that if you had, for instance, a disability of some sort that 00:24:51.450 --> 00:24:53.520 align:middle line:84% meant that you couldn't join the labor market, 00:24:53.520 --> 00:24:55.477 align:middle line:84% your parents might give you a higher dowry? 00:24:55.477 --> 00:24:57.310 align:middle line:84% BEN OLKEN: No, that's not what I'm thinking. 00:24:57.310 --> 00:24:58.980 align:middle line:90% I'm thinking about-- 00:24:58.980 --> 00:25:00.930 align:middle line:84% I'm thinking of a much simpler thing. 00:25:00.930 --> 00:25:02.430 align:middle line:84% Sorry, that could be certainly true. 00:25:02.430 --> 00:25:04.120 align:middle line:90% Yes, that certainly could be. 00:25:04.120 --> 00:25:07.200 align:middle line:90% I was thinking more-- 00:25:07.200 --> 00:25:11.280 align:middle line:84% who are the types of people who might, say, want to marry 00:25:11.280 --> 00:25:13.332 align:middle line:90% a really rich spouse, right? 00:25:13.332 --> 00:25:15.040 align:middle line:84% And who are people who don't care so much 00:25:15.040 --> 00:25:16.810 align:middle line:90% about marrying a rich spouse? 00:25:16.810 --> 00:25:19.240 align:middle line:84% They may have different consumption preferences. 00:25:19.240 --> 00:25:22.450 align:middle line:84% So if I want to estimate a model where I'm going 00:25:22.450 --> 00:25:23.980 align:middle line:90% to-- what's the test, right? 00:25:23.980 --> 00:25:26.980 align:middle line:84% The empirical test was going to be something like-- 00:25:26.980 --> 00:25:33.720 align:middle line:90% 00:25:33.720 --> 00:25:39.277 align:middle line:90% I'm going to regress, say-- 00:25:39.277 --> 00:25:40.610 align:middle line:90% this is not really on the board. 00:25:40.610 --> 00:25:42.050 align:middle line:84% But I'm going to regress something 00:25:42.050 --> 00:25:45.300 align:middle line:90% about the q's of the-- 00:25:45.300 --> 00:25:47.150 align:middle line:84% I'll go back here so we're on the board. 00:25:47.150 --> 00:25:50.195 align:middle line:84% But I'm going to regret, say, the man's consumption 00:25:50.195 --> 00:25:52.820 align:middle line:84% versus women's consumption as a function of how much bargaining 00:25:52.820 --> 00:25:54.567 align:middle line:84% power they have, or how much unearned-- 00:25:54.567 --> 00:25:56.150 align:middle line:84% what fraction-- for example, I'm going 00:25:56.150 --> 00:25:59.300 align:middle line:84% to run a regression of things that 00:25:59.300 --> 00:26:01.855 align:middle line:84% look like women's consumption on how much resources the woman 00:26:01.855 --> 00:26:04.230 align:middle line:84% brought to the marriage-- well, for the man, for example. 00:26:04.230 --> 00:26:07.460 align:middle line:84% But imagine that people's preferences over consumption 00:26:07.460 --> 00:26:09.770 align:middle line:84% are correlated with their preferences 00:26:09.770 --> 00:26:11.930 align:middle line:84% for who they're going to match with, right? 00:26:11.930 --> 00:26:13.730 align:middle line:90% I want a really fancy life. 00:26:13.730 --> 00:26:16.520 align:middle line:84% Maybe I marry a really rich partner, for example, right? 00:26:16.520 --> 00:26:19.370 align:middle line:84% That could be an example where the endogenous composition 00:26:19.370 --> 00:26:22.275 align:middle line:84% of the marriage could be related to the preferences. 00:26:22.275 --> 00:26:23.680 align:middle line:90% You see what I'm saying? 00:26:23.680 --> 00:26:26.060 align:middle line:84% AUDIENCE: Isn't that just for earned [INAUDIBLE]?? 00:26:26.060 --> 00:26:27.510 align:middle line:90% BEN OLKEN: Yes. 00:26:27.510 --> 00:26:28.010 align:middle line:90% Yeah. 00:26:28.010 --> 00:26:31.720 align:middle line:90% 00:26:31.720 --> 00:26:33.320 align:middle line:90% Yes, yes, it is. 00:26:33.320 --> 00:26:35.320 align:middle line:84% It's not necessarily better for unearned income. 00:26:35.320 --> 00:26:36.620 align:middle line:84% But yes, it's problem for earned income, too. 00:26:36.620 --> 00:26:37.120 align:middle line:90% Yeah. 00:26:37.120 --> 00:26:40.320 align:middle line:84% AUDIENCE: So then a totally separate channel, 00:26:40.320 --> 00:26:42.150 align:middle line:84% but also a potential concern would 00:26:42.150 --> 00:26:46.590 align:middle line:84% be if your unearned income is somehow scaling inversely 00:26:46.590 --> 00:26:50.417 align:middle line:90% with your earning potential? 00:26:50.417 --> 00:26:52.500 align:middle line:84% BEN OLKEN: Yes, that also could be a problem, too. 00:26:52.500 --> 00:26:53.610 align:middle line:84% AUDIENCE: It would be a bad problem. 00:26:53.610 --> 00:26:54.460 align:middle line:90% BEN OLKEN: Yes, yes. 00:26:54.460 --> 00:26:54.960 align:middle line:90% Correct. 00:26:54.960 --> 00:26:57.530 align:middle line:84% If those things were correlated too, that could be a problem. 00:26:57.530 --> 00:26:59.500 align:middle line:84% So I'm going to show you a couple of examples 00:26:59.500 --> 00:27:00.500 align:middle line:90% of some specific papers. 00:27:00.500 --> 00:27:02.210 align:middle line:84% We can talk about this in specific context. 00:27:02.210 --> 00:27:03.590 align:middle line:84% Actually, the main point I wanted 00:27:03.590 --> 00:27:04.820 align:middle line:90% you to take away from this-- 00:27:04.820 --> 00:27:06.112 align:middle line:90% I'm putting this on the board-- 00:27:06.112 --> 00:27:08.330 align:middle line:84% is to start thinking about it from this perspective. 00:27:08.330 --> 00:27:10.920 align:middle line:84% When we say these things impact bargaining power, 00:27:10.920 --> 00:27:12.162 align:middle line:90% what do we mean? 00:27:12.162 --> 00:27:14.120 align:middle line:84% The simplest version of that is what am I going 00:27:14.120 --> 00:27:15.720 align:middle line:90% to get in the outside option? 00:27:15.720 --> 00:27:17.150 align:middle line:84% And so whenever you see something 00:27:17.150 --> 00:27:18.290 align:middle line:84% like that, you should be thinking through, 00:27:18.290 --> 00:27:19.400 align:middle line:90% what is the outside option? 00:27:19.400 --> 00:27:21.320 align:middle line:84% How do these things map to the outside option? 00:27:21.320 --> 00:27:22.653 align:middle line:90% That was actually my main point. 00:27:22.653 --> 00:27:27.280 align:middle line:90% 00:27:27.280 --> 00:27:28.820 align:middle line:90% OK. 00:27:28.820 --> 00:27:29.330 align:middle line:90% Yeah. 00:27:29.330 --> 00:27:32.102 align:middle line:84% AUDIENCE: Sorry, how should we think of outside in this model? 00:27:32.102 --> 00:27:32.810 align:middle line:90% You said it was-- 00:27:32.810 --> 00:27:34.393 align:middle line:84% BEN OLKEN: So I said it was exogenous. 00:27:34.393 --> 00:27:35.390 align:middle line:90% AUDIENCE: Yeah, but-- 00:27:35.390 --> 00:27:36.080 align:middle line:84% BEN OLKEN: Where does it come from? 00:27:36.080 --> 00:27:36.740 align:middle line:90% AUDIENCE: Yeah. 00:27:36.740 --> 00:27:38.690 align:middle line:84% BEN OLKEN: So I don't have a good theory 00:27:38.690 --> 00:27:39.890 align:middle line:90% of alpha in this model. 00:27:39.890 --> 00:27:43.440 align:middle line:84% In some bargaining models, it comes from your discount rate. 00:27:43.440 --> 00:27:46.790 align:middle line:84% So this is the standard Nash bargaining model. 00:27:46.790 --> 00:27:51.920 align:middle line:84% Some models microfound this with a series of alternating offers. 00:27:51.920 --> 00:27:54.008 align:middle line:90% So I make you an offer. 00:27:54.008 --> 00:27:56.300 align:middle line:84% And you can choose to accept it or make a counteroffer. 00:27:56.300 --> 00:27:58.675 align:middle line:84% Or we go back and forth and back and forth till we agree. 00:27:58.675 --> 00:28:01.400 align:middle line:84% And in those models, I think that, if I remember correctly, 00:28:01.400 --> 00:28:05.240 align:middle line:84% the differences in the discount rates between the two parties 00:28:05.240 --> 00:28:08.820 align:middle line:84% are going to determine the alpha. 00:28:08.820 --> 00:28:11.335 align:middle line:84% That may not have been a clear-- that may not have been 00:28:11.335 --> 00:28:12.600 align:middle line:90% the answer you're looking for. 00:28:12.600 --> 00:28:14.710 align:middle line:84% But we treat it as-- from the Nash perspective, 00:28:14.710 --> 00:28:15.510 align:middle line:90% we treat it as exogenous. 00:28:15.510 --> 00:28:16.920 align:middle line:84% And I think there are some models that microfound it. 00:28:16.920 --> 00:28:18.295 align:middle line:84% And I know that's an example of-- 00:28:18.295 --> 00:28:20.940 align:middle line:84% I think that's the Rubenstein model, I believe. 00:28:20.940 --> 00:28:23.432 align:middle line:90% 00:28:23.432 --> 00:28:24.390 align:middle line:90% OK, someone is nodding. 00:28:24.390 --> 00:28:27.505 align:middle line:90% So OK. 00:28:27.505 --> 00:28:32.760 align:middle line:90% 00:28:32.760 --> 00:28:33.815 align:middle line:90% I already said that. 00:28:33.815 --> 00:28:36.420 align:middle line:84% OK, so here's the first paper I wanted to talk about in a bit 00:28:36.420 --> 00:28:36.920 align:middle line:90% of detail. 00:28:36.920 --> 00:28:39.480 align:middle line:90% This is by Duncan Thomas, 1990. 00:28:39.480 --> 00:28:44.670 align:middle line:84% So Thomas is going to be an example of looking 00:28:44.670 --> 00:28:51.350 align:middle line:84% at mothers' and fathers' income on child health. 00:28:51.350 --> 00:28:52.400 align:middle line:90% OK? 00:28:52.400 --> 00:28:56.820 align:middle line:84% So you can interpret this in two ways. 00:28:56.820 --> 00:29:00.080 align:middle line:84% Either moms care more about kids, right? 00:29:00.080 --> 00:29:01.428 align:middle line:90% That would be the ex ante way. 00:29:01.428 --> 00:29:02.720 align:middle line:90% Maybe moms care what more kids. 00:29:02.720 --> 00:29:05.150 align:middle line:84% We're going to say that child's health is mom's good. 00:29:05.150 --> 00:29:07.340 align:middle line:84% Or maybe we could just say, we don't 00:29:07.340 --> 00:29:09.840 align:middle line:84% actually a priority of what's a mom's good 00:29:09.840 --> 00:29:11.330 align:middle line:90% and what's a dad's good. 00:29:11.330 --> 00:29:15.740 align:middle line:84% But we can just look at differences in the q's 00:29:15.740 --> 00:29:18.850 align:middle line:90% as a test of income pooling. 00:29:18.850 --> 00:29:20.200 align:middle line:90% OK. 00:29:20.200 --> 00:29:22.575 align:middle line:84% And he's going to focus on nonlabor income for the reason 00:29:22.575 --> 00:29:25.075 align:middle line:84% that we talked about, which is that at least he doesn't have 00:29:25.075 --> 00:29:27.820 align:middle line:84% to deal with the fact that the labor income is endogenous, 00:29:27.820 --> 00:29:31.690 align:middle line:84% that the labor leisure decision is itself a choice. 00:29:31.690 --> 00:29:32.860 align:middle line:90% We don't want that in there. 00:29:32.860 --> 00:29:34.563 align:middle line:90% We want the unearned income. 00:29:34.563 --> 00:29:36.230 align:middle line:84% So he's going to run a regression, which 00:29:36.230 --> 00:29:42.650 align:middle line:84% is going to look like some outcome variable 00:29:42.650 --> 00:29:46.130 align:middle line:84% as a function of the woman's unearned income and the man's 00:29:46.130 --> 00:29:48.157 align:middle line:90% unearned income. 00:29:48.157 --> 00:29:50.240 align:middle line:84% And I think we talked about some of the challenges 00:29:50.240 --> 00:29:52.260 align:middle line:90% with this regression. 00:29:52.260 --> 00:29:56.060 align:middle line:84% So if I were to show you this, what do you think of this? 00:29:56.060 --> 00:30:00.095 align:middle line:84% So we talked with marriage market issues, matching issues. 00:30:00.095 --> 00:30:02.660 align:middle line:90% 00:30:02.660 --> 00:30:07.080 align:middle line:84% What else might you be worried about in this example? 00:30:07.080 --> 00:30:07.740 align:middle line:90% Yeah, Ahmed. 00:30:07.740 --> 00:30:09.240 align:middle line:84% AUDIENCE: So is this a way of trying 00:30:09.240 --> 00:30:13.380 align:middle line:84% to get at the logging parameter in bold? 00:30:13.380 --> 00:30:17.610 align:middle line:84% I'm assuming the child is a common good for the marriage? 00:30:17.610 --> 00:30:21.210 align:middle line:84% BEN OLKEN: This is a test of the pooled income hypothesis, OK? 00:30:21.210 --> 00:30:26.240 align:middle line:84% So this is a test which basically says-- 00:30:26.240 --> 00:30:29.705 align:middle line:90% so yes, in two respects. 00:30:29.705 --> 00:30:30.410 align:middle line:90% OK, sorry. 00:30:30.410 --> 00:30:32.570 align:middle line:90% Yes, it is a test of that. 00:30:32.570 --> 00:30:38.415 align:middle line:84% The strict test of this is it's a test of the prediction that-- 00:30:38.415 --> 00:30:44.527 align:middle line:90% 00:30:44.527 --> 00:30:46.110 align:middle line:84% or I'll just show it up on the slides. 00:30:46.110 --> 00:30:53.700 align:middle line:84% It's a test of this prediction that fixing the bargaining 00:30:53.700 --> 00:30:57.270 align:middle line:84% weights, the identity of the income doesn't matter, OK? 00:30:57.270 --> 00:30:58.740 align:middle line:90% So it's a test of that, right? 00:30:58.740 --> 00:31:01.140 align:middle line:84% It's testing, do we get different responses 00:31:01.140 --> 00:31:03.055 align:middle line:90% from Ym or Yf? 00:31:03.055 --> 00:31:05.510 align:middle line:90% Yf or Ym? 00:31:05.510 --> 00:31:12.410 align:middle line:84% And yes, number 2, in terms of the outside option, what would 00:31:12.410 --> 00:31:14.000 align:middle line:90% be the alternative hypothesis? 00:31:14.000 --> 00:31:16.280 align:middle line:84% Yes, it's this idea here over here, exactly, 00:31:16.280 --> 00:31:18.408 align:middle line:84% that my outside option is going to depend on 00:31:18.408 --> 00:31:19.700 align:middle line:90% how much of the income is mine. 00:31:19.700 --> 00:31:25.098 align:middle line:90% 00:31:25.098 --> 00:31:27.140 align:middle line:84% AUDIENCE: I'm not sure what you're trying to say. 00:31:27.140 --> 00:31:29.530 align:middle line:84% But the conceptual problem I'm having 00:31:29.530 --> 00:31:32.950 align:middle line:84% is that if it's true to model, still I'm 00:31:32.950 --> 00:31:36.130 align:middle line:90% thinking that bargaining-- 00:31:36.130 --> 00:31:38.980 align:middle line:84% given the differential bargaining power-- 00:31:38.980 --> 00:31:40.410 align:middle line:90% say it's fixed, it's exogenous-- 00:31:40.410 --> 00:31:43.120 align:middle line:90% I don't know-- 0.8, 0.2. 00:31:43.120 --> 00:31:45.520 align:middle line:84% It still feels like, for some reason, 00:31:45.520 --> 00:31:51.610 align:middle line:84% the income to the agent with the higher bargaining power 00:31:51.610 --> 00:31:55.690 align:middle line:84% could have a different effect on the common good. 00:31:55.690 --> 00:32:01.370 align:middle line:84% Are we assuming common good is equally valued by the agents? 00:32:01.370 --> 00:32:03.655 align:middle line:90% BEN OLKEN: Yeah-- no. 00:32:03.655 --> 00:32:04.780 align:middle line:90% AUDIENCE: They [? don't? ?] 00:32:04.780 --> 00:32:05.630 align:middle line:90% BEN OLKEN: But OK. 00:32:05.630 --> 00:32:07.000 align:middle line:90% But this is a great question. 00:32:07.000 --> 00:32:08.060 align:middle line:84% Actually, I was just thinking about this. 00:32:08.060 --> 00:32:09.102 align:middle line:90% This is a great question. 00:32:09.102 --> 00:32:11.630 align:middle line:84% OK, so let's go back to this very big setup, OK? 00:32:11.630 --> 00:32:17.868 align:middle line:84% So in this setup, what happens if I change the fraction of Y-- 00:32:17.868 --> 00:32:19.660 align:middle line:84% if I change the fraction of unearned income 00:32:19.660 --> 00:32:21.160 align:middle line:84% that goes from the man to the woman? 00:32:21.160 --> 00:32:23.461 align:middle line:90% 00:32:23.461 --> 00:32:24.544 align:middle line:90% AUDIENCE: For this model-- 00:32:24.544 --> 00:32:25.169 align:middle line:90% BEN OLKEN: Huh? 00:32:25.169 --> 00:32:28.187 align:middle line:90% 00:32:28.187 --> 00:32:30.020 align:middle line:84% AUDIENCE: But it [? still ?] doesn't change. 00:32:30.020 --> 00:32:30.500 align:middle line:90% Nothing-- 00:32:30.500 --> 00:32:31.167 align:middle line:90% BEN OLKEN: Yeah. 00:32:31.167 --> 00:32:33.610 align:middle line:84% So if the bargaining weights are fixed, 00:32:33.610 --> 00:32:36.110 align:middle line:84% so it's not in a world where they're determined endogenously 00:32:36.110 --> 00:32:38.980 align:middle line:84% like that-- if the bargaining weights are fixed, 00:32:38.980 --> 00:32:41.270 align:middle line:90% then nothing happens, right? 00:32:41.270 --> 00:32:43.970 align:middle line:84% That's the test they want to reject. 00:32:43.970 --> 00:32:44.470 align:middle line:90% OK? 00:32:44.470 --> 00:32:45.962 align:middle line:90% And then we can reject that. 00:32:45.962 --> 00:32:48.420 align:middle line:84% One way we can reject it is the bargaining weights changed. 00:32:48.420 --> 00:32:51.003 align:middle line:84% I should have actually stopped and paused for other questions. 00:32:51.003 --> 00:32:53.548 align:middle line:84% Are there other ways in which we could reject this 00:32:53.548 --> 00:32:54.840 align:middle line:90% besides the bargaining weights? 00:32:54.840 --> 00:32:57.465 align:middle line:84% I guess at some level, it's all the bargaining weight changing. 00:32:57.465 --> 00:32:59.490 align:middle line:84% But you seem to have something else in mind 00:32:59.490 --> 00:33:00.540 align:middle line:84% for some other way you could reject 00:33:00.540 --> 00:33:02.415 align:middle line:84% this besides the bargaining weights changing. 00:33:02.415 --> 00:33:06.205 align:middle line:84% AUDIENCE: It seems like budgets can also be bargained over. 00:33:06.205 --> 00:33:08.580 align:middle line:84% You're bargaining over your [INAUDIBLE] if you don't have 00:33:08.580 --> 00:33:10.330 align:middle line:90% the budget in this model-- 00:33:10.330 --> 00:33:12.538 align:middle line:84% BEN OLKEN: What do you mean, bargaining over budgets? 00:33:12.538 --> 00:33:15.750 align:middle line:90% 00:33:15.750 --> 00:33:19.380 align:middle line:84% AUDIENCE: The bargaining power can also determine-- 00:33:19.380 --> 00:33:21.840 align:middle line:84% say they have three different consumptions-- 00:33:21.840 --> 00:33:25.920 align:middle line:84% individual consumption, marriage consumption. 00:33:25.920 --> 00:33:28.200 align:middle line:84% My bargaining power in the marriage 00:33:28.200 --> 00:33:31.500 align:middle line:84% can determine, say, how much of the income of the other person 00:33:31.500 --> 00:33:35.700 align:middle line:84% I can get for the couple good and for myself as well. 00:33:35.700 --> 00:33:37.170 align:middle line:90% Say I don't care about my child. 00:33:37.170 --> 00:33:39.537 align:middle line:90% I have more bargaining power. 00:33:39.537 --> 00:33:41.370 align:middle line:84% BEN OLKEN: But those are all-- it's all just 00:33:41.370 --> 00:33:43.110 align:middle line:90% a vector of consumption goods. 00:33:43.110 --> 00:33:44.970 align:middle line:90% It's x1 to x-whatever. 00:33:44.970 --> 00:33:48.603 align:middle line:90% 00:33:48.603 --> 00:33:50.270 align:middle line:84% AUDIENCE: I have to think more about it. 00:33:50.270 --> 00:33:53.103 align:middle line:84% BEN OLKEN: Yeah, I don't think it matters. 00:33:53.103 --> 00:33:54.520 align:middle line:84% Another thing, by the way, which-- 00:33:54.520 --> 00:33:57.130 align:middle line:84% this is one microfoundation, by the way, for why it might fail. 00:33:57.130 --> 00:33:58.780 align:middle line:84% Another microfoundation for why it might fail 00:33:58.780 --> 00:34:00.340 align:middle line:84% might have to do with control rights. 00:34:00.340 --> 00:34:04.300 align:middle line:84% Maybe if we actually think about the mechanics 00:34:04.300 --> 00:34:08.278 align:middle line:84% of decision-making, it might be that actually, 00:34:08.278 --> 00:34:09.820 align:middle line:84% maybe we're not bargaining like this. 00:34:09.820 --> 00:34:13.330 align:middle line:84% But maybe whoever has the cash actually makes 00:34:13.330 --> 00:34:17.800 align:middle line:84% decisions and presents the spouse with a fait accompli, 00:34:17.800 --> 00:34:18.610 align:middle line:90% right? 00:34:18.610 --> 00:34:19.900 align:middle line:90% We've all heard-- 00:34:19.900 --> 00:34:23.440 align:middle line:84% I feel like literature is filled with examples of this thing 00:34:23.440 --> 00:34:27.418 align:middle line:90% happening, as is real life. 00:34:27.418 --> 00:34:29.710 align:middle line:84% And that would be a slightly different microfoundation. 00:34:29.710 --> 00:34:31.360 align:middle line:90% I don't think it would-- 00:34:31.360 --> 00:34:33.458 align:middle line:84% that also you could think about ways 00:34:33.458 --> 00:34:34.750 align:middle line:90% in which this model could fail. 00:34:34.750 --> 00:34:35.802 align:middle line:90% Yeah? 00:34:35.802 --> 00:34:37.510 align:middle line:84% AUDIENCE: Is the change in control rights 00:34:37.510 --> 00:34:40.210 align:middle line:90% model change in om or of? 00:34:40.210 --> 00:34:41.530 align:middle line:90% BEN OLKEN: What, sorry? 00:34:41.530 --> 00:34:43.385 align:middle line:84% AUDIENCE: Is the change in control rights 00:34:43.385 --> 00:34:45.760 align:middle line:84% different than our change in parameters in the bargaining 00:34:45.760 --> 00:34:47.440 align:middle line:90% model you covered [INAUDIBLE]? 00:34:47.440 --> 00:34:48.219 align:middle line:90% BEN OLKEN: Yes. 00:34:48.219 --> 00:34:53.340 align:middle line:84% So in this example, so I haven't written down 00:34:53.340 --> 00:34:55.919 align:middle line:84% the change in the control rights version of this. 00:34:55.919 --> 00:35:03.470 align:middle line:90% 00:35:03.470 --> 00:35:07.220 align:middle line:84% AUDIENCE: Wages-- the women's wages are deposited into a bank 00:35:07.220 --> 00:35:09.680 align:middle line:84% account that is managed by the man, 00:35:09.680 --> 00:35:11.720 align:middle line:84% and she knew that if her outside option-- 00:35:11.720 --> 00:35:13.970 align:middle line:84% that would similarly be the case in another marriage-- 00:35:13.970 --> 00:35:14.945 align:middle line:84% BEN OLKEN: Then it's going to be the same. 00:35:14.945 --> 00:35:16.260 align:middle line:90% AUDIENCE: Right, yeah. 00:35:16.260 --> 00:35:17.120 align:middle line:90% BEN OLKEN: Yeah. 00:35:17.120 --> 00:35:18.670 align:middle line:84% So I'm trying to think through exactly how you 00:35:18.670 --> 00:35:19.587 align:middle line:90% would write that down. 00:35:19.587 --> 00:35:21.830 align:middle line:84% It could end up looking similar, right? 00:35:21.830 --> 00:35:23.880 align:middle line:84% Because basically, in the outside option, 00:35:23.880 --> 00:35:25.130 align:middle line:90% I have control over the money. 00:35:25.130 --> 00:35:26.672 align:middle line:84% I just do whatever I'm going to want. 00:35:26.672 --> 00:35:29.540 align:middle line:84% So maybe it actually is going to look very similar, actually. 00:35:29.540 --> 00:35:31.790 align:middle line:84% AUDIENCE: Yeah, it's the definition of control, right? 00:35:31.790 --> 00:35:32.180 align:middle line:90% BEN OLKEN: Yeah, right? 00:35:32.180 --> 00:35:34.490 align:middle line:84% It's basically, if we don't agree, I just get to-- 00:35:34.490 --> 00:35:36.620 align:middle line:90% I think it's very similar. 00:35:36.620 --> 00:35:40.850 align:middle line:84% If we disagree, then I get to consume my stuff. 00:35:40.850 --> 00:35:43.240 align:middle line:84% And I get to decide how to spend my money. 00:35:43.240 --> 00:35:44.990 align:middle line:84% You get to decide how to spend your money. 00:35:44.990 --> 00:35:46.407 align:middle line:84% And those are our outside options. 00:35:46.407 --> 00:35:48.260 align:middle line:84% So actually, sometimes, I think actually, 00:35:48.260 --> 00:35:50.720 align:middle line:84% it's a slightly different microfoundation 00:35:50.720 --> 00:35:52.970 align:middle line:84% in the sense of, in this one, in the example 00:35:52.970 --> 00:35:54.830 align:middle line:84% I gave before, I was thinking of the marriage splits up. 00:35:54.830 --> 00:35:56.300 align:middle line:84% But the other one is, we don't agree 00:35:56.300 --> 00:35:56.990 align:middle line:90% on how to spend the money. 00:35:56.990 --> 00:35:57.950 align:middle line:90% And I just go spend my money. 00:35:57.950 --> 00:35:59.300 align:middle line:90% And you spend your money, right? 00:35:59.300 --> 00:36:00.150 align:middle line:90% That's another example. 00:36:00.150 --> 00:36:01.608 align:middle line:84% But it's going to look very similar 00:36:01.608 --> 00:36:04.363 align:middle line:90% in the formula in that sense. 00:36:04.363 --> 00:36:05.530 align:middle line:90% That's all I'm going to say. 00:36:05.530 --> 00:36:07.770 align:middle line:90% OK. 00:36:07.770 --> 00:36:09.187 align:middle line:90% OK. 00:36:09.187 --> 00:36:10.770 align:middle line:84% Other challenges with this regression? 00:36:10.770 --> 00:36:12.312 align:middle line:84% Other things you want to think about? 00:36:12.312 --> 00:36:14.800 align:middle line:90% 00:36:14.800 --> 00:36:15.300 align:middle line:90% OK, right. 00:36:15.300 --> 00:36:17.550 align:middle line:90% So we talked about-- 00:36:17.550 --> 00:36:20.100 align:middle line:84% there could be individual-level [? minumum ?] variables. 00:36:20.100 --> 00:36:24.660 align:middle line:84% So for example, another thing about this regression 00:36:24.660 --> 00:36:32.200 align:middle line:90% is the utilities may be related. 00:36:32.200 --> 00:36:34.000 align:middle line:84% And this is a cross-sectional regression. 00:36:34.000 --> 00:36:38.310 align:middle line:84% So the utilities may be related to these income shocks, right? 00:36:38.310 --> 00:36:39.810 align:middle line:84% So imagine that someone who grows up 00:36:39.810 --> 00:36:43.560 align:middle line:84% really rich and they're expecting 00:36:43.560 --> 00:36:45.300 align:middle line:84% to inherit a lot of money or whatever-- 00:36:45.300 --> 00:36:46.890 align:middle line:84% they may have different preferences than someone who 00:36:46.890 --> 00:36:47.880 align:middle line:90% doesn't grow up rich. 00:36:47.880 --> 00:36:52.530 align:middle line:84% And that might affect their consumption choices, right? 00:36:52.530 --> 00:36:59.160 align:middle line:84% So the ideal examples will have better variation than this, 00:36:59.160 --> 00:37:02.018 align:middle line:84% in just cross-sectional because you need to hold the utility-- 00:37:02.018 --> 00:37:04.560 align:middle line:84% the right test is we're holding the utility function constant 00:37:04.560 --> 00:37:06.750 align:middle line:90% and just shocking the Y's. 00:37:06.750 --> 00:37:07.333 align:middle line:90% Yeah, Patrick? 00:37:07.333 --> 00:37:08.708 align:middle line:84% AUDIENCE: So in this case, do you 00:37:08.708 --> 00:37:11.400 align:middle line:84% have a shock in the fraction of shared income from each party? 00:37:11.400 --> 00:37:13.830 align:middle line:84% Or do you have a shock in total income, too? 00:37:13.830 --> 00:37:16.740 align:middle line:84% BEN OLKEN: This one, they don't really have a shock. 00:37:16.740 --> 00:37:19.800 align:middle line:84% So the first version of this, they don't have a shock. 00:37:19.800 --> 00:37:20.970 align:middle line:90% It's just cross-sectional. 00:37:20.970 --> 00:37:23.217 align:middle line:84% I'll show you a shock in a second, OK? 00:37:23.217 --> 00:37:25.050 align:middle line:84% So this one, they don't have a shock, right? 00:37:25.050 --> 00:37:26.250 align:middle line:84% So this is just going to be cross-sectional. 00:37:26.250 --> 00:37:27.070 align:middle line:90% And that's what I was saying. 00:37:27.070 --> 00:37:29.220 align:middle line:84% The problem is of endogenous marriage formation. 00:37:29.220 --> 00:37:31.470 align:middle line:84% And the problem's that the utilities may be correlated 00:37:31.470 --> 00:37:32.655 align:middle line:90% with the income shocks. 00:37:32.655 --> 00:37:35.910 align:middle line:90% 00:37:35.910 --> 00:37:36.410 align:middle line:90% Right? 00:37:36.410 --> 00:37:39.500 align:middle line:84% So I was going to start by saying yes. 00:37:39.500 --> 00:37:42.710 align:middle line:84% This was the first paper that did this in this form-- 00:37:42.710 --> 00:37:46.950 align:middle line:84% at least in development that I know of. 00:37:46.950 --> 00:37:49.380 align:middle line:84% He does find that basically, there 00:37:49.380 --> 00:37:50.770 align:middle line:90% are different income effects. 00:37:50.770 --> 00:37:52.770 align:middle line:84% So he looks at the ratio of the income effects 00:37:52.770 --> 00:37:55.230 align:middle line:84% for man's income versus woman's income 00:37:55.230 --> 00:37:58.890 align:middle line:84% and finds very, very substantial differences in those incomes. 00:37:58.890 --> 00:38:02.220 align:middle line:84% But as you were saying, what you'd like to do 00:38:02.220 --> 00:38:05.280 align:middle line:84% is you'd like to have better identified shocks 00:38:05.280 --> 00:38:07.080 align:middle line:84% to the different income that are not 00:38:07.080 --> 00:38:11.110 align:middle line:84% going to be correlated with the utility shocks, 00:38:11.110 --> 00:38:14.278 align:middle line:84% with the preference parameters, OK? 00:38:14.278 --> 00:38:16.570 align:middle line:84% So let me not go through the results of this in detail. 00:38:16.570 --> 00:38:19.540 align:middle line:84% Let me go to the next paper, which was actually 00:38:19.540 --> 00:38:21.130 align:middle line:84% a paper by Esther, which she wrote 00:38:21.130 --> 00:38:24.670 align:middle line:84% in 2003, which looks at grandparents spending 00:38:24.670 --> 00:38:26.320 align:middle line:90% on their grandchildren. 00:38:26.320 --> 00:38:29.080 align:middle line:84% And the reason for this, I think, 00:38:29.080 --> 00:38:31.570 align:middle line:84% not that grandparents and grandchildren are particularly 00:38:31.570 --> 00:38:34.330 align:middle line:84% of interest-- although, they of course, are interesting-- 00:38:34.330 --> 00:38:36.038 align:middle line:84% you could have equally written this paper 00:38:36.038 --> 00:38:38.650 align:middle line:84% with moms and fathers and their kids. 00:38:38.650 --> 00:38:41.770 align:middle line:84% But she had a policy change she could look at, 00:38:41.770 --> 00:38:46.800 align:middle line:84% which was a large unexpected change 00:38:46.800 --> 00:38:48.810 align:middle line:90% in people's unearned incomes. 00:38:48.810 --> 00:38:49.310 align:middle line:90% OK? 00:38:49.310 --> 00:38:52.490 align:middle line:84% So in particular, at the end of apartheid, 00:38:52.490 --> 00:38:55.550 align:middle line:84% South Africa expands their pension programs 00:38:55.550 --> 00:38:57.110 align:middle line:90% to include Blacks. 00:38:57.110 --> 00:39:00.200 align:middle line:90% And there are a lot of kids-- 00:39:00.200 --> 00:39:03.230 align:middle line:84% 17% of kids who are living in multigenerational households 00:39:03.230 --> 00:39:06.560 align:middle line:90% that have a pension earner, OK? 00:39:06.560 --> 00:39:10.760 align:middle line:84% And she compares living with children who are living with 00:39:10.760 --> 00:39:14.930 align:middle line:84% an eligible grandparent-- so age 60 for females, 00:39:14.930 --> 00:39:16.220 align:middle line:90% or age 65 for males-- 00:39:16.220 --> 00:39:17.510 align:middle line:90% that's the eligibility rule-- 00:39:17.510 --> 00:39:20.700 align:middle line:84% to people who are living with a grandparent, 00:39:20.700 --> 00:39:24.740 align:middle line:84% but the grandparent is just younger, right? 00:39:24.740 --> 00:39:25.580 align:middle line:90% Yeah, Patrick? 00:39:25.580 --> 00:39:27.955 align:middle line:84% AUDIENCE: The comparison takes place before they actually 00:39:27.955 --> 00:39:29.700 align:middle line:90% get the money, right? 00:39:29.700 --> 00:39:32.323 align:middle line:84% BEN OLKEN: I thought it takes place after. 00:39:32.323 --> 00:39:34.698 align:middle line:84% AUDIENCE: Wouldn't you be concerned about income effects? 00:39:34.698 --> 00:39:37.510 align:middle line:90% 00:39:37.510 --> 00:39:38.890 align:middle line:90% BEN OLKEN: Right. 00:39:38.890 --> 00:39:39.640 align:middle line:90% Correct. 00:39:39.640 --> 00:39:43.000 align:middle line:84% So she's going to compare what happens if the grandmother gets 00:39:43.000 --> 00:39:45.280 align:middle line:84% the pension versus the grandfather gets the pension. 00:39:45.280 --> 00:39:46.250 align:middle line:90% AUDIENCE: OK. 00:39:46.250 --> 00:39:46.750 align:middle line:90% OK. 00:39:46.750 --> 00:39:48.130 align:middle line:84% BEN OLKEN: So it's going to be [INAUDIBLE],, sorry. 00:39:48.130 --> 00:39:49.630 align:middle line:84% AUDIENCE: If they find out they were 00:39:49.630 --> 00:39:51.610 align:middle line:84% going to be earning this income, could 00:39:51.610 --> 00:39:54.870 align:middle line:84% you look at before they got the actual-- 00:39:54.870 --> 00:39:57.370 align:middle line:84% BEN OLKEN: That would also be cool, as an anticipated thing. 00:39:57.370 --> 00:40:04.370 align:middle line:84% But I think the key thing is going to be comparing just-- 00:40:04.370 --> 00:40:08.000 align:middle line:84% so she wants to compare what is the treatment effect-- 00:40:08.000 --> 00:40:10.700 align:middle line:84% so there's two things going on in this regression, right? 00:40:10.700 --> 00:40:17.930 align:middle line:84% You have, say, 60-year-old female grandmothers 00:40:17.930 --> 00:40:19.910 align:middle line:90% and 59-year-old grandmothers. 00:40:19.910 --> 00:40:21.860 align:middle line:84% And you have 65-year-old grandfathers 00:40:21.860 --> 00:40:23.420 align:middle line:90% and 64-year-old grandfathers. 00:40:23.420 --> 00:40:26.070 align:middle line:84% So we want to restrict ourselves to kids who 00:40:26.070 --> 00:40:27.320 align:middle line:90% are living with a grandparent. 00:40:27.320 --> 00:40:28.340 align:middle line:84% Because obviously, kids who are not 00:40:28.340 --> 00:40:29.210 align:middle line:84% living with a grandparent-- those 00:40:29.210 --> 00:40:30.752 align:middle line:84% are in different kinds of households. 00:40:30.752 --> 00:40:34.402 align:middle line:84% The treatment effect of living with a grandmother is going 00:40:34.402 --> 00:40:36.110 align:middle line:84% to-- she's going to get that by comparing 00:40:36.110 --> 00:40:39.340 align:middle line:84% 65 or 64-year-old grandmothers, just eligible 00:40:39.340 --> 00:40:41.360 align:middle line:84% versus just not eligible-- kids living with. 00:40:41.360 --> 00:40:43.568 align:middle line:84% And the treatment effect of living with a grandfather 00:40:43.568 --> 00:40:46.220 align:middle line:84% is 65-year-old grandfathers versus 64-year-old 00:40:46.220 --> 00:40:47.780 align:middle line:90% grandfathers, right? 00:40:47.780 --> 00:40:51.330 align:middle line:84% So those are the two different treatment effects. 00:40:51.330 --> 00:40:54.037 align:middle line:84% Both of those capture the income effect. 00:40:54.037 --> 00:40:56.370 align:middle line:84% Then the thing she wants to look at from the perspective 00:40:56.370 --> 00:41:00.600 align:middle line:84% of this topic is does it matter if the money goes 00:41:00.600 --> 00:41:04.190 align:middle line:84% to the grandmother or the grandfather? 00:41:04.190 --> 00:41:06.110 align:middle line:84% That's the kind of unitary-- that's 00:41:06.110 --> 00:41:16.920 align:middle line:90% the decision points here. 00:41:16.920 --> 00:41:19.660 align:middle line:90% OK, question. 00:41:19.660 --> 00:41:21.490 align:middle line:84% First of all, does that seem sensible? 00:41:21.490 --> 00:41:25.150 align:middle line:84% And does that map into the framework I was just 00:41:25.150 --> 00:41:26.660 align:middle line:90% talking about? 00:41:26.660 --> 00:41:27.160 align:middle line:90% Yeah. 00:41:27.160 --> 00:41:28.300 align:middle line:90% Remind me your name, sorry. 00:41:28.300 --> 00:41:28.870 align:middle line:90% AUDIENCE: Erin. 00:41:28.870 --> 00:41:29.537 align:middle line:90% BEN OLKEN: Erin. 00:41:29.537 --> 00:41:34.540 align:middle line:84% AUDIENCE: Yeah, I guess the one threat could be that sometimes, 00:41:34.540 --> 00:41:38.102 align:middle line:84% norms around whether or not an elderly grandparent can come 00:41:38.102 --> 00:41:39.660 align:middle line:84% back and live with a family might 00:41:39.660 --> 00:41:40.952 align:middle line:90% depend on what their gender is. 00:41:40.952 --> 00:41:42.440 align:middle line:90% BEN OLKEN: The what, sorry? 00:41:42.440 --> 00:41:44.440 align:middle line:84% AUDIENCE: Norms around whether a grandparent can 00:41:44.440 --> 00:41:49.930 align:middle line:84% come to return and live with the family, for their children, 00:41:49.930 --> 00:41:53.970 align:middle line:84% depends on the gender of the grandparent. 00:41:53.970 --> 00:41:56.725 align:middle line:84% I know at least in India, we were seeing that-- 00:41:56.725 --> 00:42:00.240 align:middle line:84% that's something we did with elderly females, who 00:42:00.240 --> 00:42:03.600 align:middle line:90% were more likely to live alone. 00:42:03.600 --> 00:42:04.500 align:middle line:90% BEN OLKEN: Exactly. 00:42:04.500 --> 00:42:05.010 align:middle line:90% Right. 00:42:05.010 --> 00:42:06.670 align:middle line:90% So what we'd like to do-- 00:42:06.670 --> 00:42:08.190 align:middle line:90% so I think the cleanest test-- 00:42:08.190 --> 00:42:10.410 align:middle line:90% I was just thinking about this. 00:42:10.410 --> 00:42:12.510 align:middle line:84% I think the cleanest test of this 00:42:12.510 --> 00:42:16.215 align:middle line:84% would be kids who are living with both grandparents, right? 00:42:16.215 --> 00:42:18.090 align:middle line:84% So we're going to hold all of those household 00:42:18.090 --> 00:42:19.860 align:middle line:90% composition things the same. 00:42:19.860 --> 00:42:25.405 align:middle line:84% And we'd exogenously vary which grandparent is just eligible. 00:42:25.405 --> 00:42:27.030 align:middle line:84% And she includes a lot of demographics. 00:42:27.030 --> 00:42:29.280 align:middle line:84% And I think this is in there, but we can double-check. 00:42:29.280 --> 00:42:35.010 align:middle line:84% The cleanest test is households with grandparents 00:42:35.010 --> 00:42:39.257 align:middle line:84% of both genders comparing the case when the grandfather is 00:42:39.257 --> 00:42:41.340 align:middle line:84% just eligible to the case where the grandmother is 00:42:41.340 --> 00:42:41.880 align:middle line:90% just eligible. 00:42:41.880 --> 00:42:42.750 align:middle line:90% That would be the cleanest test. 00:42:42.750 --> 00:42:43.482 align:middle line:90% Yeah. 00:42:43.482 --> 00:42:44.940 align:middle line:84% AUDIENCE: And the concern with that 00:42:44.940 --> 00:42:48.000 align:middle line:84% would be that there might be some specific dynamic 00:42:48.000 --> 00:42:50.430 align:middle line:84% in their relationship if the grandmother is 00:42:50.430 --> 00:42:52.190 align:middle line:84% older than the grandfather, potentially. 00:42:52.190 --> 00:42:52.690 align:middle line:90% Well 00:42:52.690 --> 00:42:55.232 align:middle line:84% BEN OLKEN: They, don't have to be because of the differential 00:42:55.232 --> 00:42:56.840 align:middle line:90% age rule. 00:42:56.840 --> 00:42:58.100 align:middle line:90% But yes, it could be an issue. 00:42:58.100 --> 00:42:58.830 align:middle line:90% But I'm saying-- 00:42:58.830 --> 00:43:00.590 align:middle line:84% AUDIENCE: The fact that the grandmother became eligible 00:43:00.590 --> 00:43:02.540 align:middle line:84% before the grandfather implies that they're 00:43:02.540 --> 00:43:04.897 align:middle line:84% in the type of marriage where the woman would be older. 00:43:04.897 --> 00:43:05.480 align:middle line:90% BEN OLKEN: No. 00:43:05.480 --> 00:43:07.660 align:middle line:84% Because the age difference-- the eligibility. 00:43:07.660 --> 00:43:09.410 align:middle line:84% The women are eligible five years younger. 00:43:09.410 --> 00:43:10.990 align:middle line:90% AUDIENCE: I see. 00:43:10.990 --> 00:43:13.297 align:middle line:90% BEN OLKEN: Yeah. 00:43:13.297 --> 00:43:14.880 align:middle line:84% And by the way, you are right to think 00:43:14.880 --> 00:43:16.200 align:middle line:84% that there could be discontinuities 00:43:16.200 --> 00:43:16.900 align:middle line:90% and that sort of thing. 00:43:16.900 --> 00:43:18.525 align:middle line:84% There's a paper in the US, for example, 00:43:18.525 --> 00:43:21.515 align:middle line:84% about earnings of spouses that seems 00:43:21.515 --> 00:43:23.640 align:middle line:84% to suggest that there is an important discontinuity 00:43:23.640 --> 00:43:24.180 align:middle line:90% at 50/50. 00:43:24.180 --> 00:43:26.268 align:middle line:84% So there could be a discontinuity exactly older 00:43:26.268 --> 00:43:27.060 align:middle line:90% or exactly younger. 00:43:27.060 --> 00:43:27.880 align:middle line:84% So you're right to think about that. 00:43:27.880 --> 00:43:28.880 align:middle line:90% But it's not a context-- 00:43:28.880 --> 00:43:29.820 align:middle line:90% not a concern here. 00:43:29.820 --> 00:43:31.020 align:middle line:84% AUDIENCE: It would take five years-- 00:43:31.020 --> 00:43:31.687 align:middle line:90% BEN OLKEN: What? 00:43:31.687 --> 00:43:33.903 align:middle line:84% AUDIENCE: So [? fewer ?] than five years is enough 00:43:33.903 --> 00:43:35.280 align:middle line:90% of a buffer zone so that-- 00:43:35.280 --> 00:43:36.863 align:middle line:84% BEN OLKEN: Five years is probably not, 00:43:36.863 --> 00:43:38.460 align:middle line:84% I think, a particularly salient-- 00:43:38.460 --> 00:43:39.180 align:middle line:90% AUDIENCE: Discontinuity. 00:43:39.180 --> 00:43:39.960 align:middle line:90% BEN OLKEN: Discontinuity. 00:43:39.960 --> 00:43:40.350 align:middle line:90% Right. 00:43:40.350 --> 00:43:42.360 align:middle line:84% It technically is, but I think in practice, it probably 00:43:42.360 --> 00:43:43.235 align:middle line:90% is not a salient one. 00:43:43.235 --> 00:43:45.140 align:middle line:90% Yeah. 00:43:45.140 --> 00:43:47.130 align:middle line:90% OK. 00:43:47.130 --> 00:43:47.630 align:middle line:90% All right. 00:43:47.630 --> 00:43:49.103 align:middle line:90% So yeah. 00:43:49.103 --> 00:43:50.270 align:middle line:90% So this is what she's doing. 00:43:50.270 --> 00:43:52.340 align:middle line:84% She's going to look at, in particular, 00:43:52.340 --> 00:43:54.680 align:middle line:84% she's going to control for these household 00:43:54.680 --> 00:43:56.180 align:middle line:84% composition things for the reason we 00:43:56.180 --> 00:43:58.490 align:middle line:90% talked about [INAUDIBLE]. 00:43:58.490 --> 00:44:00.200 align:middle line:90% But yes, that's the idea. 00:44:00.200 --> 00:44:01.750 align:middle line:90% So what she finds-- 00:44:01.750 --> 00:44:04.077 align:middle line:84% so by the way, so why is this better than the-- 00:44:04.077 --> 00:44:06.160 align:middle line:84% why is this an advance on the previous literature? 00:44:06.160 --> 00:44:08.410 align:middle line:84% It's because now we have a plausibly exogenous 00:44:08.410 --> 00:44:10.120 align:middle line:90% shock to income, right? 00:44:10.120 --> 00:44:16.390 align:middle line:84% And so she's going to look at the weight for height 00:44:16.390 --> 00:44:19.390 align:middle line:84% for the kids as a measure of kids' nutrition. 00:44:19.390 --> 00:44:21.820 align:middle line:84% And she's going to look at it for girls and boys. 00:44:21.820 --> 00:44:26.080 align:middle line:84% And so what she's going to find is that basically-- 00:44:26.080 --> 00:44:27.790 align:middle line:84% so she's going to compare households 00:44:27.790 --> 00:44:32.110 align:middle line:84% where the woman is eligible to the [? one ?] case where 00:44:32.110 --> 00:44:34.500 align:middle line:90% the man's eligible. 00:44:34.500 --> 00:44:37.230 align:middle line:84% And what she finds it she finds impacts of a woman being 00:44:37.230 --> 00:44:42.360 align:middle line:90% eligible on girls' nutrition. 00:44:42.360 --> 00:44:44.670 align:middle line:84% She finds this impact of women being eligible 00:44:44.670 --> 00:44:47.730 align:middle line:84% on girls' nutrition and no impact of a man being 00:44:47.730 --> 00:44:49.140 align:middle line:90% eligible on girls' nutrition. 00:44:49.140 --> 00:44:52.680 align:middle line:84% She actually finds-- although this is not really the focus 00:44:52.680 --> 00:44:54.325 align:middle line:90% of the paper-- 00:44:54.325 --> 00:44:56.760 align:middle line:84% I don't know if she tests this-- in some specifications, 00:44:56.760 --> 00:44:59.610 align:middle line:84% it looks like the woman versus the man being eligible 00:44:59.610 --> 00:45:02.488 align:middle line:84% may also have near them as individually significant 00:45:02.488 --> 00:45:03.030 align:middle line:90% for the boys. 00:45:03.030 --> 00:45:04.500 align:middle line:84% Although I wonder if you test the difference 00:45:04.500 --> 00:45:06.810 align:middle line:84% of these coefficients, that the difference between them 00:45:06.810 --> 00:45:08.730 align:middle line:90% might be significant. 00:45:08.730 --> 00:45:09.763 align:middle line:90% I'm not actually sure. 00:45:09.763 --> 00:45:12.180 align:middle line:84% Maybe not, actually, given the size those standard errors. 00:45:12.180 --> 00:45:15.270 align:middle line:84% But certainly, she finds this differential effect 00:45:15.270 --> 00:45:22.486 align:middle line:84% of men versus women on girls' nutrition. 00:45:22.486 --> 00:45:24.610 align:middle line:90% OK, so that also suggests that-- 00:45:24.610 --> 00:45:26.200 align:middle line:84% so there's something going on where 00:45:26.200 --> 00:45:28.030 align:middle line:84% who you give the income to is affecting 00:45:28.030 --> 00:45:30.200 align:middle line:84% the patterns of consumption in the household. 00:45:30.200 --> 00:45:30.920 align:middle line:90% Yeah? 00:45:30.920 --> 00:45:33.503 align:middle line:84% AUDIENCE: The difference of the coefficients being significant 00:45:33.503 --> 00:45:35.200 align:middle line:90% would mean what [INAUDIBLE]? 00:45:35.200 --> 00:45:36.730 align:middle line:90% BEN OLKEN: That's a test. 00:45:36.730 --> 00:45:39.930 align:middle line:84% So the key test-- so going back to, I think, 00:45:39.930 --> 00:45:42.750 align:middle line:84% what somebody said that we don't want just income effects 00:45:42.750 --> 00:45:44.040 align:middle line:90% of someone being eligible. 00:45:44.040 --> 00:45:45.960 align:middle line:84% We want to test whether there's a difference 00:45:45.960 --> 00:45:48.605 align:middle line:84% of whether the man's eligible versus the woman's eligible. 00:45:48.605 --> 00:45:49.147 align:middle line:90% AUDIENCE: Oh. 00:45:49.147 --> 00:45:52.670 align:middle line:84% So it's odd that that wasn't included in the table. 00:45:52.670 --> 00:45:54.990 align:middle line:90% BEN OLKEN: Yes. 00:45:54.990 --> 00:45:57.780 align:middle line:84% Maybe it's in the paper and I missed it, but yes. 00:45:57.780 --> 00:46:01.020 align:middle line:84% If I was presenting this table, I would include P-value tests-- 00:46:01.020 --> 00:46:02.888 align:middle line:84% man coefficient equals woman coefficient. 00:46:02.888 --> 00:46:05.720 align:middle line:90% 00:46:05.720 --> 00:46:07.940 align:middle line:90% AUDIENCE: So each coefficient-- 00:46:07.940 --> 00:46:10.440 align:middle line:84% it's a discontinuity in the [INAUDIBLE],, correct? 00:46:10.440 --> 00:46:11.230 align:middle line:90% BEN OLKEN: Huh? 00:46:11.230 --> 00:46:14.260 align:middle line:84% AUDIENCE: So the coefficient that women [? eligibile ?] is 00:46:14.260 --> 00:46:16.990 align:middle line:84% like a discontinuity in age and [? something. ?] Is that 00:46:16.990 --> 00:46:17.517 align:middle line:90% correct? 00:46:17.517 --> 00:46:18.100 align:middle line:90% BEN OLKEN: OK. 00:46:18.100 --> 00:46:19.802 align:middle line:90% So-- 00:46:19.802 --> 00:46:21.260 align:middle line:84% AUDIENCE: [? Tied ?] together, this 00:46:21.260 --> 00:46:25.040 align:middle line:84% is like a straight line, RD or something like that? 00:46:25.040 --> 00:46:25.910 align:middle line:90% BEN OLKEN: OK. 00:46:25.910 --> 00:46:27.830 align:middle line:84% So I don't think this paper is written 00:46:27.830 --> 00:46:29.780 align:middle line:84% as an RD in the sense of-- we're going to talk 00:46:29.780 --> 00:46:31.190 align:middle line:90% about RDs a little bit later. 00:46:31.190 --> 00:46:33.482 align:middle line:84% Actually, I think the plan is for you to talk about RDs 00:46:33.482 --> 00:46:35.270 align:middle line:90% on Friday in recitation. 00:46:35.270 --> 00:46:37.940 align:middle line:84% As far as I remember-- and you can correct me if I'm wrong. 00:46:37.940 --> 00:46:39.230 align:middle line:84% You can go check and correct me if I'm wrong. 00:46:39.230 --> 00:46:41.688 align:middle line:84% Or if Esther was here, she could correct me if I was wrong. 00:46:41.688 --> 00:46:43.280 align:middle line:90% You can ask her on Wednesday. 00:46:43.280 --> 00:46:45.140 align:middle line:84% I don't think this paper is written as an RD 00:46:45.140 --> 00:46:47.060 align:middle line:84% with a continuous running variable in age. 00:46:47.060 --> 00:46:49.100 align:middle line:84% I think she doesn't have enough observations, for example, 00:46:49.100 --> 00:46:50.030 align:middle line:90% to do that, right? 00:46:50.030 --> 00:46:52.470 align:middle line:84% So there's only 1,500 observations here. 00:46:52.470 --> 00:46:54.530 align:middle line:84% So I think it's just dummies for being eligible. 00:46:54.530 --> 00:46:57.690 align:middle line:84% The thought experiment is 100% an RD. 00:46:57.690 --> 00:47:00.277 align:middle line:84% But I don't think this paper is set up as an RD. 00:47:00.277 --> 00:47:01.860 align:middle line:84% But it could have been with more data. 00:47:01.860 --> 00:47:05.030 align:middle line:90% 00:47:05.030 --> 00:47:08.710 align:middle line:90% Other questions? 00:47:08.710 --> 00:47:09.730 align:middle line:90% But yes. 00:47:09.730 --> 00:47:13.000 align:middle line:84% If you had infinite data and you were writing this paper today, 00:47:13.000 --> 00:47:18.860 align:middle line:84% I think you would set this up as a running variable. 00:47:18.860 --> 00:47:20.620 align:middle line:84% I think if I was writing the paper today, 00:47:20.620 --> 00:47:22.990 align:middle line:84% I would set this up-- and I had infinite data-- 00:47:22.990 --> 00:47:25.300 align:middle line:84% I would restrict myself to households 00:47:25.300 --> 00:47:28.480 align:middle line:84% with both a grandfather and a grandmother. 00:47:28.480 --> 00:47:30.730 align:middle line:84% And I would put in running variables 00:47:30.730 --> 00:47:33.850 align:middle line:90% for near the cutoff of age. 00:47:33.850 --> 00:47:35.710 align:middle line:84% I would put in linear running variables 00:47:35.710 --> 00:47:40.380 align:middle line:84% for being on both sides of the discontinuity-- 00:47:40.380 --> 00:47:42.830 align:middle line:90% for birthday, for example. 00:47:42.830 --> 00:47:45.088 align:middle line:84% And I would have a discontinuous break in eligibility 00:47:45.088 --> 00:47:45.880 align:middle line:90% at the age cutoffs. 00:47:45.880 --> 00:47:51.640 align:middle line:90% 00:47:51.640 --> 00:47:54.430 align:middle line:84% And maybe that would be actually a good example 00:47:54.430 --> 00:47:56.440 align:middle line:84% to talk about on Friday, for example. 00:47:56.440 --> 00:48:01.060 align:middle line:90% 00:48:01.060 --> 00:48:02.970 align:middle line:84% So is everyone following what I just said 00:48:02.970 --> 00:48:04.273 align:middle line:90% about regression discontinuity? 00:48:04.273 --> 00:48:06.690 align:middle line:84% Has anyone seen this before in at least one other context? 00:48:06.690 --> 00:48:07.440 align:middle line:90% Yes? 00:48:07.440 --> 00:48:08.020 align:middle line:90% Yes, OK. 00:48:08.020 --> 00:48:08.520 align:middle line:90% Great. 00:48:08.520 --> 00:48:09.000 align:middle line:90% We'll go through it. 00:48:09.000 --> 00:48:11.583 align:middle line:84% We'll still go through it more detail in recitation on Friday. 00:48:11.583 --> 00:48:15.500 align:middle line:90% But that's what I mean. 00:48:15.500 --> 00:48:18.250 align:middle line:90% OK. 00:48:18.250 --> 00:48:23.458 align:middle line:84% So given this finding, you might say, well, maybe policymakers 00:48:23.458 --> 00:48:25.000 align:middle line:84% may want to think about-- there might 00:48:25.000 --> 00:48:28.990 align:middle line:84% be policy implications with this in terms of giving resources 00:48:28.990 --> 00:48:30.640 align:middle line:84% differentially to men versus women. 00:48:30.640 --> 00:48:33.570 align:middle line:90% 00:48:33.570 --> 00:48:36.210 align:middle line:84% For example, there are lots of transfer programs 00:48:36.210 --> 00:48:39.220 align:middle line:84% in the world, which give resources to households to help 00:48:39.220 --> 00:48:40.470 align:middle line:90% them with a variety of things. 00:48:40.470 --> 00:48:43.510 align:middle line:84% I'm going to talk about a bunch of them in the PF section. 00:48:43.510 --> 00:48:47.960 align:middle line:84% If you believe A, if you reject the unitary model, and B, 00:48:47.960 --> 00:48:49.960 align:middle line:84% if you believe some of these findings that women 00:48:49.960 --> 00:48:54.610 align:middle line:84% maybe, on average, channel more of their resources 00:48:54.610 --> 00:48:57.790 align:middle line:84% to their kids, and C, if you're a paternalistic decision maker 00:48:57.790 --> 00:48:59.710 align:middle line:84% or policymaker whose goal is actually 00:48:59.710 --> 00:49:02.110 align:middle line:84% to help the kids in the household, which you could be 00:49:02.110 --> 00:49:04.630 align:middle line:84% for reasons that we'll talk about in a few lectures, 00:49:04.630 --> 00:49:06.640 align:middle line:84% then all that suggests you might care about 00:49:06.640 --> 00:49:09.302 align:middle line:84% whether you give the transfer to the man or the woman. 00:49:09.302 --> 00:49:11.090 align:middle line:90% OK? 00:49:11.090 --> 00:49:17.660 align:middle line:84% So here's an example of a recent randomized trial 00:49:17.660 --> 00:49:18.710 align:middle line:90% that looked at this. 00:49:18.710 --> 00:49:21.810 align:middle line:84% So this is in the context of the GiveDirectly trial, 00:49:21.810 --> 00:49:24.650 align:middle line:84% which was a trial that gave transfers to households. 00:49:24.650 --> 00:49:27.290 align:middle line:84% This is going to be a $400 one-time transfer 00:49:27.290 --> 00:49:29.190 align:middle line:90% to households in Kenya. 00:49:29.190 --> 00:49:31.313 align:middle line:84% And they did a bunch of things, some of which 00:49:31.313 --> 00:49:32.230 align:middle line:90% I'll talk about later. 00:49:32.230 --> 00:49:33.390 align:middle line:84% But one of the things they did is 00:49:33.390 --> 00:49:34.830 align:middle line:84% they randomized whether the transfer went 00:49:34.830 --> 00:49:36.030 align:middle line:84% to the men or the women in the household, 00:49:36.030 --> 00:49:37.920 align:middle line:84% in the case of a married household 00:49:37.920 --> 00:49:39.820 align:middle line:90% with a man and a woman, OK? 00:49:39.820 --> 00:49:43.405 align:middle line:84% So as I said, there's a bunch of stuff. 00:49:43.405 --> 00:49:46.580 align:middle line:90% 00:49:46.580 --> 00:49:51.870 align:middle line:84% These are different outcome variables in the column. 00:49:51.870 --> 00:49:54.870 align:middle line:84% And this is the additional effect 00:49:54.870 --> 00:49:59.280 align:middle line:84% of randomizing who the recipient is to the woman. 00:49:59.280 --> 00:50:05.430 align:middle line:84% And in general, they get a joint P-value of 0.11 00:50:05.430 --> 00:50:10.920 align:middle line:84% on a lot of things, like food security, health, education. 00:50:10.920 --> 00:50:15.360 align:middle line:84% They see nothing on a bunch of these. 00:50:15.360 --> 00:50:18.270 align:middle line:84% Now they see nothing on a bunch of these. 00:50:18.270 --> 00:50:21.150 align:middle line:84% They do see some-- they have a psychological well-being 00:50:21.150 --> 00:50:23.405 align:middle line:84% index, which I think is a measure in a survey. 00:50:23.405 --> 00:50:25.530 align:middle line:84% And they have some questions on female empowerment. 00:50:25.530 --> 00:50:26.730 align:middle line:84% Both of those seem to go up if we 00:50:26.730 --> 00:50:28.022 align:middle line:90% give the transfer to the women. 00:50:28.022 --> 00:50:31.950 align:middle line:84% But the other stuff, like health and education, 00:50:31.950 --> 00:50:33.890 align:middle line:90% don't seem to change. 00:50:33.890 --> 00:50:38.482 align:middle line:84% Now it is also worth noting that these 00:50:38.482 --> 00:50:40.440 align:middle line:84% don't seem to change from the treatment effect, 00:50:40.440 --> 00:50:42.600 align:middle line:90% either, of giving this transfer. 00:50:42.600 --> 00:50:44.973 align:middle line:84% So maybe this was a case where the transfer 00:50:44.973 --> 00:50:47.640 align:middle line:84% didn't have differential effects on these incomes to begin with. 00:50:47.640 --> 00:50:49.470 align:middle line:84% And maybe it's therefore-- maybe it's not affecting 00:50:49.470 --> 00:50:50.190 align:middle line:90% these incomes to begin with. 00:50:50.190 --> 00:50:51.815 align:middle line:84% So maybe it's not so surprising that it 00:50:51.815 --> 00:50:53.180 align:middle line:90% had differential effects-- 00:50:53.180 --> 00:50:54.860 align:middle line:90% maybe. 00:50:54.860 --> 00:50:59.150 align:middle line:84% But certainly, if you had just seen the previous papers-- 00:50:59.150 --> 00:51:02.540 align:middle line:84% the Thomas paper that says that the income going 00:51:02.540 --> 00:51:09.500 align:middle line:84% to the woman in the household yields substantially better 00:51:09.500 --> 00:51:11.090 align:middle line:84% nutritional outcomes for the kids, 00:51:11.090 --> 00:51:14.585 align:middle line:84% and likewise, the grandmothers and grandfathers paper-- 00:51:14.585 --> 00:51:16.760 align:middle line:84% a reasonable hypothesis might have 00:51:16.760 --> 00:51:19.322 align:middle line:84% been that even if you don't get much on average, 00:51:19.322 --> 00:51:20.780 align:middle line:84% or when you give it to the men, you 00:51:20.780 --> 00:51:22.655 align:middle line:84% get big effects if you give it to the woman-- 00:51:22.655 --> 00:51:24.310 align:middle line:84% they're not seeing that in this paper. 00:51:24.310 --> 00:51:27.910 align:middle line:84% So I think it's also just worth pointing out-- 00:51:27.910 --> 00:51:30.300 align:middle line:84% it could be that this program didn't 00:51:30.300 --> 00:51:31.800 align:middle line:84% have effects on health and education 00:51:31.800 --> 00:51:33.640 align:middle line:90% for a variety of reasons. 00:51:33.640 --> 00:51:36.120 align:middle line:84% It could just be a different example. 00:51:36.120 --> 00:51:40.410 align:middle line:84% I think more work could be done on this question. 00:51:40.410 --> 00:51:41.280 align:middle line:90% Yeah, Patrick. 00:51:41.280 --> 00:51:44.258 align:middle line:84% AUDIENCE: They talk about take-up rates between females? 00:51:44.258 --> 00:51:45.300 align:middle line:90% BEN OLKEN: What's, sorry? 00:51:45.300 --> 00:51:47.383 align:middle line:84% AUDIENCE: Take-up rates between males and females? 00:51:47.383 --> 00:51:48.240 align:middle line:90% BEN OLKEN: Yeah. 00:51:48.240 --> 00:51:49.050 align:middle line:90% Oh, do they? 00:51:49.050 --> 00:51:52.170 align:middle line:84% I think the take-up rate is 100% in this example. 00:51:52.170 --> 00:51:53.730 align:middle line:84% I think this is all conditional on-- 00:51:53.730 --> 00:51:56.790 align:middle line:84% I don't remember this, but I don't 00:51:56.790 --> 00:51:59.472 align:middle line:84% think the take-up rates are an issue in this context. 00:51:59.472 --> 00:52:01.930 align:middle line:84% I could be wrong, but I don't remember that being an issue. 00:52:01.930 --> 00:52:06.220 align:middle line:84% AUDIENCE: So everyone that got offered took it? 00:52:06.220 --> 00:52:08.480 align:middle line:84% BEN OLKEN: OK, I don't remember for a fact. 00:52:08.480 --> 00:52:11.530 align:middle line:84% I don't remember low take-up as an issue in this paper. 00:52:11.530 --> 00:52:12.610 align:middle line:90% But you're right, sure. 00:52:12.610 --> 00:52:14.380 align:middle line:84% If the take-up rate was different-- 00:52:14.380 --> 00:52:19.650 align:middle line:90% 00:52:19.650 --> 00:52:26.570 align:middle line:84% but you're not seeing a substantially lower-- 00:52:26.570 --> 00:52:31.270 align:middle line:84% on the did I spend the money variables, 00:52:31.270 --> 00:52:33.020 align:middle line:84% none of these look any different depending 00:52:33.020 --> 00:52:34.850 align:middle line:84% on whether they give it to the woman or the man. 00:52:34.850 --> 00:52:36.530 align:middle line:84% So if you thought the woman's take-it was a lot lower, 00:52:36.530 --> 00:52:38.450 align:middle line:84% you'd expect negative coefficients here. 00:52:38.450 --> 00:52:39.620 align:middle line:84% This is the additional effect of giving it 00:52:39.620 --> 00:52:40.700 align:middle line:90% to the women or the men. 00:52:40.700 --> 00:52:41.510 align:middle line:90% Yeah? 00:52:41.510 --> 00:52:46.490 align:middle line:84% AUDIENCE: Could there be some sort of control thing at play? 00:52:46.490 --> 00:52:49.340 align:middle line:90% If you were to-- 00:52:49.340 --> 00:52:50.990 align:middle line:84% so in the previous example, I feel 00:52:50.990 --> 00:52:52.940 align:middle line:84% like part of the difference is that you're 00:52:52.940 --> 00:52:54.050 align:middle line:90% announcing the policy. 00:52:54.050 --> 00:52:56.450 align:middle line:84% And even if it's the case that women are just 00:52:56.450 --> 00:52:58.760 align:middle line:84% becoming eligible before the men would become eligible, 00:52:58.760 --> 00:53:01.910 align:middle line:84% the men can expect that they'll receive the 70% bonus 00:53:01.910 --> 00:53:04.030 align:middle line:84% at the time of becoming eligible. 00:53:04.030 --> 00:53:04.880 align:middle line:90% So-- 00:53:04.880 --> 00:53:05.755 align:middle line:90% BEN OLKEN: I'm sorry. 00:53:05.755 --> 00:53:06.620 align:middle line:90% Wait, I'm sorry. 00:53:06.620 --> 00:53:07.760 align:middle line:90% Say it again, sorry? 00:53:07.760 --> 00:53:09.015 align:middle line:84% Oh, you went back to the previous examples. 00:53:09.015 --> 00:53:09.270 align:middle line:90% Sorry. 00:53:09.270 --> 00:53:10.010 align:middle line:84% AUDIENCE: In the previous example-- 00:53:10.010 --> 00:53:10.280 align:middle line:90% BEN OLKEN: Yes. 00:53:10.280 --> 00:53:12.020 align:middle line:84% AUDIENCE: --you had an effect where 00:53:12.020 --> 00:53:14.360 align:middle line:84% you were announcing that you were going to give 00:53:14.360 --> 00:53:16.650 align:middle line:90% a transfer to both parties. 00:53:16.650 --> 00:53:17.820 align:middle line:90% BEN OLKEN: Yes. 00:53:17.820 --> 00:53:20.220 align:middle line:84% AUDIENCE: The transfer arrived to the woman 00:53:20.220 --> 00:53:22.095 align:middle line:84% faster than it did to the man in the case 00:53:22.095 --> 00:53:23.220 align:middle line:90% where she became eligible-- 00:53:23.220 --> 00:53:24.750 align:middle line:90% BEN OLKEN: Correct. 00:53:24.750 --> 00:53:26.460 align:middle line:84% AUDIENCE: And so there's a notion 00:53:26.460 --> 00:53:29.220 align:middle line:84% that if the man waits sufficiently long, 00:53:29.220 --> 00:53:30.600 align:middle line:90% he'll get what's owed to him? 00:53:30.600 --> 00:53:31.380 align:middle line:90% BEN OLKEN: Yes. 00:53:31.380 --> 00:53:33.370 align:middle line:84% AUDIENCE: But in this particular example, 00:53:33.370 --> 00:53:36.840 align:middle line:84% you have this exogenous body that's just throwing money 00:53:36.840 --> 00:53:38.195 align:middle line:90% at a household. 00:53:38.195 --> 00:53:39.570 align:middle line:84% Could you make an argument that-- 00:53:39.570 --> 00:53:39.780 align:middle line:90% BEN OLKEN: Oh. 00:53:39.780 --> 00:53:41.655 align:middle line:84% AUDIENCE: --there's a different control thing 00:53:41.655 --> 00:53:46.560 align:middle line:84% at play of this money, if it's being given totally randomly 00:53:46.560 --> 00:53:49.408 align:middle line:84% to one party or the other, is going to be pooled 00:53:49.408 --> 00:53:51.450 align:middle line:84% or is going to be put in the hands of one person, 00:53:51.450 --> 00:53:53.670 align:middle line:84% whereas that would less be the case? 00:53:53.670 --> 00:53:54.370 align:middle line:90% BEN OLKEN: Yes. 00:53:54.370 --> 00:53:54.870 align:middle line:90% Yes. 00:53:54.870 --> 00:53:55.590 align:middle line:90% OK, sorry. 00:53:55.590 --> 00:53:56.730 align:middle line:90% Yes, I agree with you. 00:53:56.730 --> 00:53:58.680 align:middle line:90% So totally plausible. 00:53:58.680 --> 00:54:00.330 align:middle line:84% So just to be clear, the argument 00:54:00.330 --> 00:54:03.600 align:middle line:84% was in the pension example, we're all 00:54:03.600 --> 00:54:05.070 align:middle line:90% going to get a pension someday. 00:54:05.070 --> 00:54:07.470 align:middle line:84% So maybe a reasonable bargaining solution 00:54:07.470 --> 00:54:09.910 align:middle line:84% is you do what you want with your pension. 00:54:09.910 --> 00:54:11.410 align:middle line:84% I'll do what I want with my pension. 00:54:11.410 --> 00:54:12.780 align:middle line:84% It's all going to work out roughly 00:54:12.780 --> 00:54:14.430 align:middle line:84% because we're all going to get pensions eventually, 00:54:14.430 --> 00:54:16.380 align:middle line:84% whereas in this one, it's a one-time thing. 00:54:16.380 --> 00:54:18.215 align:middle line:90% And only one of us got it. 00:54:18.215 --> 00:54:20.340 align:middle line:84% And therefore, we really have to duke this one out. 00:54:20.340 --> 00:54:21.120 align:middle line:90% Is that basically your argument? 00:54:21.120 --> 00:54:21.840 align:middle line:90% AUDIENCE: Yes. 00:54:21.840 --> 00:54:22.800 align:middle line:90% BEN OLKEN: Yeah, it could be. 00:54:22.800 --> 00:54:23.592 align:middle line:90% That's interesting. 00:54:23.592 --> 00:54:25.280 align:middle line:90% Yeah. 00:54:25.280 --> 00:54:25.790 align:middle line:90% Yeah? 00:54:25.790 --> 00:54:28.790 align:middle line:84% AUDIENCE: One [INAUDIBLE] this clearly not in this 00:54:28.790 --> 00:54:32.120 align:middle line:84% [? section ?] that the transfer is for female or male rather 00:54:32.120 --> 00:54:36.473 align:middle line:84% than for the whole household when they do the experiment, 00:54:36.473 --> 00:54:37.640 align:middle line:90% when you give the transfer-- 00:54:37.640 --> 00:54:38.120 align:middle line:90% BEN OLKEN: Yeah? 00:54:38.120 --> 00:54:40.280 align:middle line:84% AUDIENCE: [? Is it ?] possible that the female or the male 00:54:40.280 --> 00:54:41.905 align:middle line:84% because they're already household, they 00:54:41.905 --> 00:54:43.400 align:middle line:90% think it is [INAUDIBLE]. 00:54:43.400 --> 00:54:45.150 align:middle line:84% So they are more likely to pool money 00:54:45.150 --> 00:54:47.430 align:middle line:84% together because it is not something really occurring, 00:54:47.430 --> 00:54:48.155 align:middle line:90% right? 00:54:48.155 --> 00:54:49.280 align:middle line:90% BEN OLKEN: I think it's a-- 00:54:49.280 --> 00:54:49.820 align:middle line:90% yeah. 00:54:49.820 --> 00:54:50.840 align:middle line:84% I think that's a related context. 00:54:50.840 --> 00:54:52.298 align:middle line:84% What is the framing, in some sense? 00:54:52.298 --> 00:54:54.440 align:middle line:84% And how are they setting this up, right? 00:54:54.440 --> 00:54:55.997 align:middle line:90% Is that a reasonable-- 00:54:55.997 --> 00:54:57.830 align:middle line:84% AUDIENCE: Yeah, I'm just curious why there-- 00:54:57.830 --> 00:54:58.970 align:middle line:90% although it's given to female-- 00:54:58.970 --> 00:54:59.720 align:middle line:84% BEN OLKEN: So this was actually-- 00:54:59.720 --> 00:55:01.095 align:middle line:84% I think the way it was done is it 00:55:01.095 --> 00:55:02.400 align:middle line:90% was done through mobile money. 00:55:02.400 --> 00:55:03.775 align:middle line:84% And so the question is literally, 00:55:03.775 --> 00:55:07.170 align:middle line:90% whose phone does it go to? 00:55:07.170 --> 00:55:10.250 align:middle line:84% Your account or your husband's account? 00:55:10.250 --> 00:55:11.790 align:middle line:84% AUDIENCE: I'm thinking like if I-- 00:55:11.790 --> 00:55:13.740 align:middle line:84% is it my earned income or is it my pension? 00:55:13.740 --> 00:55:16.080 align:middle line:90% It is clear it is for me. 00:55:16.080 --> 00:55:19.290 align:middle line:84% But for this case, if someone drops off money, and they met 00:55:19.290 --> 00:55:20.815 align:middle line:90% [INAUDIBLE] drop the money. 00:55:20.815 --> 00:55:21.690 align:middle line:90% BEN OLKEN: Oh, I see. 00:55:21.690 --> 00:55:22.240 align:middle line:90% Yeah, yeah. 00:55:22.240 --> 00:55:22.740 align:middle line:90% No, exactly. 00:55:22.740 --> 00:55:24.957 align:middle line:90% It could be framed differently. 00:55:24.957 --> 00:55:26.040 align:middle line:90% That's what you're saying. 00:55:26.040 --> 00:55:27.722 align:middle line:90% It could be framed differently. 00:55:27.722 --> 00:55:29.430 align:middle line:84% I think what you're basically saying is-- 00:55:29.430 --> 00:55:35.540 align:middle line:90% 00:55:35.540 --> 00:55:40.550 align:middle line:84% so in the simplest bargaining model I set up, 00:55:40.550 --> 00:55:42.530 align:middle line:90% it's not different, right? 00:55:42.530 --> 00:55:43.452 align:middle line:90% It's clear. 00:55:43.452 --> 00:55:45.410 align:middle line:84% Think about the version of the bargaining model 00:55:45.410 --> 00:55:46.910 align:middle line:84% I talked about where basically, it's 00:55:46.910 --> 00:55:50.870 align:middle line:84% like, I get to just make the consumption decision myself. 00:55:50.870 --> 00:55:53.240 align:middle line:84% If we don't agree, I guess we each get 00:55:53.240 --> 00:55:55.070 align:middle line:90% to spend our own money, right? 00:55:55.070 --> 00:55:57.867 align:middle line:84% In that case, it's different because it 00:55:57.867 --> 00:56:00.200 align:middle line:84% went on my mobile money account versus your mobile money 00:56:00.200 --> 00:56:01.920 align:middle line:90% account, right? 00:56:01.920 --> 00:56:04.730 align:middle line:84% So in that version, I think it's still very different. 00:56:04.730 --> 00:56:06.007 align:middle line:90% I do think you are right. 00:56:06.007 --> 00:56:08.090 align:middle line:84% And both of these comments I think are related to, 00:56:08.090 --> 00:56:10.320 align:middle line:84% what is the framing of this stuff 00:56:10.320 --> 00:56:12.240 align:middle line:84% and how this is presented to the households 00:56:12.240 --> 00:56:18.290 align:middle line:84% could matter in terms of what the implications are. 00:56:18.290 --> 00:56:21.683 align:middle line:84% That being said, I do think that we should update from this 00:56:21.683 --> 00:56:24.100 align:middle line:84% in the sense that you might have seen the previous results 00:56:24.100 --> 00:56:26.767 align:middle line:84% and drawn the totally reasonable policy inference that we should 00:56:26.767 --> 00:56:29.398 align:middle line:84% give the money to the moms if we care about the kids. 00:56:29.398 --> 00:56:30.940 align:middle line:84% And this says, actually, no, maybe it 00:56:30.940 --> 00:56:36.150 align:middle line:84% doesn't actually matter as much as you think in this example. 00:56:36.150 --> 00:56:37.900 align:middle line:90% No, you disagree? 00:56:37.900 --> 00:56:38.790 align:middle line:90% Oh. 00:56:38.790 --> 00:56:41.123 align:middle line:84% It's very hard to read people's faces with the masks on. 00:56:41.123 --> 00:56:43.093 align:middle line:90% I'm doing my best. 00:56:43.093 --> 00:56:45.010 align:middle line:84% AUDIENCE: Would an interesting comparison test 00:56:45.010 --> 00:56:48.730 align:middle line:84% be if you took the same income and then delivered it 00:56:48.730 --> 00:56:52.300 align:middle line:84% to both parents individually, and then 00:56:52.300 --> 00:56:54.930 align:middle line:84% slightly changed the proportion that you 00:56:54.930 --> 00:56:58.920 align:middle line:84% awarded to both parents, assuming 00:56:58.920 --> 00:57:00.145 align:middle line:90% you had many, many, many? 00:57:00.145 --> 00:57:02.520 align:middle line:84% BEN OLKEN: Yeah, that would be a good test of your theory 00:57:02.520 --> 00:57:05.100 align:middle line:84% that basically-- that might be a good test of it. 00:57:05.100 --> 00:57:07.060 align:middle line:84% But again, it depends on the framing. 00:57:07.060 --> 00:57:09.810 align:middle line:84% And there are a variety of other-- 00:57:09.810 --> 00:57:12.540 align:middle line:84% I think Esther will talk about more of these next time. 00:57:12.540 --> 00:57:14.850 align:middle line:84% There are a variety of other more subtle tests. 00:57:14.850 --> 00:57:16.607 align:middle line:84% We give you slightly more of something. 00:57:16.607 --> 00:57:18.690 align:middle line:84% We change the decision rights or we change the way 00:57:18.690 --> 00:57:21.240 align:middle line:84% we talk about it, or explicitly going into the details 00:57:21.240 --> 00:57:22.440 align:middle line:90% of this bargaining question. 00:57:22.440 --> 00:57:23.670 align:middle line:84% There is a bunch of additional work 00:57:23.670 --> 00:57:25.020 align:middle line:84% on that, which I'm not going to talk about today, 00:57:25.020 --> 00:57:26.760 align:middle line:84% but which I think Esther will talk about at least some 00:57:26.760 --> 00:57:27.750 align:middle line:90% of that next time. 00:57:27.750 --> 00:57:31.470 align:middle line:90% 00:57:31.470 --> 00:57:32.280 align:middle line:90% OK. 00:57:32.280 --> 00:57:34.423 align:middle line:84% So now I want to get to the paper 00:57:34.423 --> 00:57:36.840 align:middle line:84% you guys read for today, which is about Pareto efficiency. 00:57:36.840 --> 00:57:39.930 align:middle line:90% 00:57:39.930 --> 00:57:43.680 align:middle line:84% So in the Burkina Faso example that you 00:57:43.680 --> 00:57:45.810 align:middle line:84% got-- this is another test of these models, right? 00:57:45.810 --> 00:57:49.010 align:middle line:84% It's do we get to the efficient outcome? 00:57:49.010 --> 00:57:51.010 align:middle line:84% In this example over here, the efficient outcome 00:57:51.010 --> 00:57:53.392 align:middle line:84% is somehow we managed to earn S and then 00:57:53.392 --> 00:57:54.850 align:middle line:84% share it as opposed to each getting 00:57:54.850 --> 00:57:56.611 align:middle line:90% our own individual [INAUDIBLE]. 00:57:56.611 --> 00:57:59.440 align:middle line:90% 00:57:59.440 --> 00:58:02.500 align:middle line:84% In the Burkina Faso example that you guys read for today, 00:58:02.500 --> 00:58:04.510 align:middle line:84% individuals control different plots. 00:58:04.510 --> 00:58:07.180 align:middle line:84% And so the Pareto optimal behavior 00:58:07.180 --> 00:58:10.480 align:middle line:84% is going to be to maximize total income 00:58:10.480 --> 00:58:14.200 align:middle line:84% and then share it according to some weights. 00:58:14.200 --> 00:58:15.120 align:middle line:90% OK? 00:58:15.120 --> 00:58:17.310 align:middle line:84% At least, that's going to be the benchmark. 00:58:17.310 --> 00:58:18.210 align:middle line:84% And we should talk about whether or not 00:58:18.210 --> 00:58:19.320 align:middle line:84% we think that's a reasonable benchmark. 00:58:19.320 --> 00:58:21.920 align:middle line:84% Are there are other benchmarks you should think about as well? 00:58:21.920 --> 00:58:22.420 align:middle line:90% OK? 00:58:22.420 --> 00:58:27.560 align:middle line:84% And then the alternative is if I have control 00:58:27.560 --> 00:58:29.090 align:middle line:84% over the income from my plot and you 00:58:29.090 --> 00:58:30.923 align:middle line:84% have the control over income from your plot, 00:58:30.923 --> 00:58:34.935 align:middle line:84% we might not efficiently share resources across the two. 00:58:34.935 --> 00:58:36.560 align:middle line:84% And I have control over some resources. 00:58:36.560 --> 00:58:40.040 align:middle line:84% And I might say, well, I'm going to overinvest in my own plot 00:58:40.040 --> 00:58:42.290 align:middle line:84% because I want the resources from my plot 00:58:42.290 --> 00:58:43.970 align:middle line:84% even if it was more efficient to put 00:58:43.970 --> 00:58:45.900 align:middle line:84% some of the resources in someone else's plot. 00:58:45.900 --> 00:58:49.820 align:middle line:84% So that's the question is are we jointly maximizing 00:58:49.820 --> 00:58:50.870 align:middle line:90% the production? 00:58:50.870 --> 00:58:53.455 align:middle line:84% Or somehow, is the identity of who 00:58:53.455 --> 00:58:54.830 align:middle line:84% controls the income from the plot 00:58:54.830 --> 00:58:56.780 align:middle line:84% affect the production choices on the plot? 00:58:56.780 --> 00:58:59.840 align:middle line:84% That's the basic question, right? 00:58:59.840 --> 00:59:00.340 align:middle line:90% OK. 00:59:00.340 --> 00:59:07.800 align:middle line:90% So what's the setup? 00:59:07.800 --> 00:59:10.420 align:middle line:90% We have labor supply. 00:59:10.420 --> 00:59:11.670 align:middle line:90% We have consumption decisions. 00:59:11.670 --> 00:59:12.878 align:middle line:90% We have an area of each plot. 00:59:12.878 --> 00:59:15.180 align:middle line:90% 00:59:15.180 --> 00:59:18.042 align:middle line:84% We have a production function, right? 00:59:18.042 --> 00:59:20.500 align:middle line:84% And we're going to assume this production function is going 00:59:20.500 --> 00:59:22.125 align:middle line:84% to be the same for all the individuals, 00:59:22.125 --> 00:59:24.962 align:middle line:84% although he's going to test that later in the paper, right? 00:59:24.962 --> 00:59:26.920 align:middle line:84% This is, I think, a key assumption, by the way. 00:59:26.920 --> 00:59:28.398 align:middle line:90% And we can talk about that. 00:59:28.398 --> 00:59:30.190 align:middle line:84% I don't know what you guys thought of that. 00:59:30.190 --> 00:59:35.728 align:middle line:84% To me, when I was rereading the paper this morning, to me, 00:59:35.728 --> 00:59:37.270 align:middle line:84% a lot of my questions had to do with, 00:59:37.270 --> 00:59:42.480 align:middle line:84% is that assumption reasonable or not? 00:59:42.480 --> 00:59:52.770 align:middle line:84% We have total labor income, total labor supply. 00:59:52.770 --> 01:00:02.900 align:middle line:84% And we have some joint decision making over consumption. 01:00:02.900 --> 01:00:04.730 align:middle line:84% So the Pareto efficient allocation 01:00:04.730 --> 01:00:06.560 align:middle line:90% is like we had before, right? 01:00:06.560 --> 01:00:11.357 align:middle line:84% We're going to maximize some weighted sum of the utilities 01:00:11.357 --> 01:00:13.940 align:middle line:84% subject to the budget constraint and the production functions. 01:00:13.940 --> 01:00:18.320 align:middle line:84% And the key point-- and I'm going to come back to this 01:00:18.320 --> 01:00:19.060 align:middle line:90% idea-- 01:00:19.060 --> 01:00:20.810 align:middle line:84% the key point is that production decisions 01:00:20.810 --> 01:00:22.550 align:middle line:90% are independent of preferences. 01:00:22.550 --> 01:00:23.990 align:middle line:84% And so I'm going to actually talk 01:00:23.990 --> 01:00:30.402 align:middle line:84% about this in more detail again when I talk about the labor 01:00:30.402 --> 01:00:32.860 align:middle line:84% lectures, this idea that our production and our consumption 01:00:32.860 --> 01:00:34.540 align:middle line:90% decisions should be separable. 01:00:34.540 --> 01:00:39.190 align:middle line:84% It's a key test of a bunch of different models. 01:00:39.190 --> 01:00:44.920 align:middle line:84% But basically, the idea is we're going to solve the-- 01:00:44.920 --> 01:00:50.650 align:middle line:84% we're going to maximize the total output by optimally 01:00:50.650 --> 01:00:53.980 align:middle line:84% allocating our resources across these two different plots 01:00:53.980 --> 01:00:55.750 align:middle line:90% to maximize output. 01:00:55.750 --> 01:00:59.140 align:middle line:84% And really, the key idea is it shouldn't 01:00:59.140 --> 01:01:02.505 align:middle line:84% matter for the household whose plot is the man's plot 01:01:02.505 --> 01:01:03.880 align:middle line:84% and whose plot is a woman's plot. 01:01:03.880 --> 01:01:06.172 align:middle line:84% We should just be maximizing output over the two plots. 01:01:06.172 --> 01:01:09.355 align:middle line:84% And then we should get the income and split it up. 01:01:09.355 --> 01:01:10.660 align:middle line:90% OK. 01:01:10.660 --> 01:01:13.290 align:middle line:84% So that's basically the framework. 01:01:13.290 --> 01:01:17.250 align:middle line:84% He's going to define the yield on a particular plot 01:01:17.250 --> 01:01:24.180 align:middle line:84% as the amount of output per area planted. 01:01:24.180 --> 01:01:29.280 align:middle line:84% And he'll look at the deviation on some particular plot k-- 01:01:29.280 --> 01:01:32.955 align:middle line:90% or particular plot i, rather. 01:01:32.955 --> 01:01:36.780 align:middle line:90% 01:01:36.780 --> 01:01:39.000 align:middle line:90% Crop k is going to be the-- 01:01:39.000 --> 01:01:41.250 align:middle line:84% they're going to look at the deviation from the output 01:01:41.250 --> 01:01:43.230 align:middle line:84% per acre on that particular plot compared 01:01:43.230 --> 01:01:46.216 align:middle line:84% to the average across all their plots. 01:01:46.216 --> 01:01:48.970 align:middle line:90% OK? 01:01:48.970 --> 01:01:53.200 align:middle line:84% And they're going to do this within crop, right? 01:01:53.200 --> 01:01:55.450 align:middle line:84% So they're going to say, we plant six different crops. 01:01:55.450 --> 01:01:57.340 align:middle line:90% But actually, I plant peanuts. 01:01:57.340 --> 01:01:58.330 align:middle line:90% And you plant peanuts. 01:01:58.330 --> 01:01:59.290 align:middle line:90% And I plant rice. 01:01:59.290 --> 01:02:00.273 align:middle line:90% And you plant rice. 01:02:00.273 --> 01:02:01.690 align:middle line:84% And we're going to look at my rice 01:02:01.690 --> 01:02:06.070 align:middle line:84% versus your rice is going to be the key we want to look at. 01:02:06.070 --> 01:02:06.640 align:middle line:90% OK? 01:02:06.640 --> 01:02:09.070 align:middle line:84% And so then they're basically going 01:02:09.070 --> 01:02:10.630 align:middle line:84% to run this regression, where they're 01:02:10.630 --> 01:02:13.660 align:middle line:84% going to run the output on whether it's 01:02:13.660 --> 01:02:18.790 align:middle line:84% the women's plot putting in household times 01:02:18.790 --> 01:02:20.450 align:middle line:90% time times crop fix effects. 01:02:20.450 --> 01:02:23.020 align:middle line:84% So in this year, we produced this. 01:02:23.020 --> 01:02:25.150 align:middle line:90% We both produced rice. 01:02:25.150 --> 01:02:27.120 align:middle line:84% We both produced groundnuts, whatever. 01:02:27.120 --> 01:02:29.120 align:middle line:84% And we both-- it was a good year, or a bad year, 01:02:29.120 --> 01:02:30.550 align:middle line:90% or whatever. 01:02:30.550 --> 01:02:33.920 align:middle line:84% But we're going to look across our different plots. 01:02:33.920 --> 01:02:34.420 align:middle line:90% OK? 01:02:34.420 --> 01:02:36.910 align:middle line:84% And the key test is going to be whether the output of the man's 01:02:36.910 --> 01:02:37.790 align:middle line:84% plot is different than the woman's plot. 01:02:37.790 --> 01:02:38.290 align:middle line:90% Yeah. 01:02:38.290 --> 01:02:39.880 align:middle line:84% AUDIENCE: So on the previous slide, 01:02:39.880 --> 01:02:42.850 align:middle line:84% when you said, why not use between the [INAUDIBLE].. 01:02:42.850 --> 01:02:44.070 align:middle line:90% BEN OLKEN: Yeah. 01:02:44.070 --> 01:02:45.282 align:middle line:90% AUDIENCE: [INAUDIBLE] 01:02:45.282 --> 01:02:46.240 align:middle line:90% BEN OLKEN: Oh, why not? 01:02:46.240 --> 01:02:49.000 align:middle line:84% AUDIENCE: Well if you measure g in terms of-- 01:02:49.000 --> 01:02:55.222 align:middle line:84% instead of yield and things like dollar rewards of yield-- 01:02:55.222 --> 01:02:56.055 align:middle line:90% BEN OLKEN: Oh, yeah. 01:02:56.055 --> 01:02:59.250 align:middle line:90% 01:02:59.250 --> 01:03:01.547 align:middle line:90% So yeah. 01:03:01.547 --> 01:03:02.630 align:middle line:90% So that's a good question. 01:03:02.630 --> 01:03:05.810 align:middle line:90% So why not do it all per dollar? 01:03:05.810 --> 01:03:10.555 align:middle line:90% 01:03:10.555 --> 01:03:11.930 align:middle line:84% That's a good question, actually. 01:03:11.930 --> 01:03:13.513 align:middle line:90% Do you have an answer? 01:03:13.513 --> 01:03:15.180 align:middle line:84% AUDIENCE: Well, I was wondering if there 01:03:15.180 --> 01:03:18.479 align:middle line:84% was a-- you don't want to overstrip the soil and plant 01:03:18.479 --> 01:03:19.460 align:middle line:90% too much corn. 01:03:19.460 --> 01:03:20.793 align:middle line:90% So you'd vary it by planting-- 01:03:20.793 --> 01:03:22.085 align:middle line:90% BEN OLKEN: In different places? 01:03:22.085 --> 01:03:24.230 align:middle line:84% AUDIENCE: --sometimes, different-- yeah. 01:03:24.230 --> 01:03:25.520 align:middle line:84% BEN OLKEN: Yeah, actually, that's a good question. 01:03:25.520 --> 01:03:26.978 align:middle line:84% I was trying to think that through. 01:03:26.978 --> 01:03:33.350 align:middle line:90% 01:03:33.350 --> 01:03:34.250 align:middle line:90% Why not do it? 01:03:34.250 --> 01:03:34.940 align:middle line:90% Yeah, go ahead. 01:03:34.940 --> 01:03:36.510 align:middle line:84% AUDIENCE: I think it's between [INAUDIBLE].. 01:03:36.510 --> 01:03:37.370 align:middle line:90% BEN OLKEN: What, sorry? 01:03:37.370 --> 01:03:39.578 align:middle line:84% AUDIENCE: Maybe they consume a lot of different crops 01:03:39.578 --> 01:03:41.983 align:middle line:84% to feed their home family rather than [INAUDIBLE] 01:03:41.983 --> 01:03:43.400 align:middle line:84% at market so that they really want 01:03:43.400 --> 01:03:45.510 align:middle line:90% to diversify the crop type. 01:03:45.510 --> 01:03:48.110 align:middle line:84% So they really put their decision-making 01:03:48.110 --> 01:03:50.690 align:middle line:84% if they want to get his share of his crop, his share 01:03:50.690 --> 01:03:54.710 align:middle line:84% of that crop so that they try to see, well, you switch fast 01:03:54.710 --> 01:03:56.460 align:middle line:90% and make a certain type of crop. 01:03:56.460 --> 01:03:57.520 align:middle line:90% BEN OLKEN: Yeah. 01:03:57.520 --> 01:03:58.443 align:middle line:90% Yeah, no. 01:03:58.443 --> 01:04:00.110 align:middle line:84% The more I think about it, actually, I'm 01:04:00.110 --> 01:04:01.520 align:middle line:84% having a harder time answering it than I thought. 01:04:01.520 --> 01:04:02.540 align:middle line:84% So I was thinking, well, you want 01:04:02.540 --> 01:04:04.873 align:middle line:84% to be able to see about fixed effects, like for example, 01:04:04.873 --> 01:04:06.560 align:middle line:84% it's a bad year for this kind of crop, 01:04:06.560 --> 01:04:08.060 align:middle line:84% or a bad year for this kind of crop. 01:04:08.060 --> 01:04:10.340 align:middle line:84% But why would one of them be systematically correlated 01:04:10.340 --> 01:04:12.150 align:middle line:84% with the woman's crop, for example? 01:04:12.150 --> 01:04:16.260 align:middle line:84% So why would we have the woman differentially plant? 01:04:16.260 --> 01:04:20.830 align:middle line:84% Well, one thing could be, for example-- 01:04:20.830 --> 01:04:22.580 align:middle line:84% no, I don't have a good example, actually. 01:04:22.580 --> 01:04:26.100 align:middle line:84% Sorry, I don't have a good answer for you, actually. 01:04:26.100 --> 01:04:29.530 align:middle line:90% Someone else have a good answer? 01:04:29.530 --> 01:04:30.920 align:middle line:90% I think I see your point. 01:04:30.920 --> 01:04:32.710 align:middle line:84% You should be able to dollarize it. 01:04:32.710 --> 01:04:35.830 align:middle line:84% It does seem like you're comparing more 01:04:35.830 --> 01:04:39.180 align:middle line:90% like to like in this example. 01:04:39.180 --> 01:04:41.492 align:middle line:84% Under the model, it's a good question. 01:04:41.492 --> 01:04:43.200 align:middle line:84% Under the model, I can see what you mean. 01:04:43.200 --> 01:04:45.120 align:middle line:84% Under the model, we should have no differences 01:04:45.120 --> 01:04:48.460 align:middle line:84% in average-- everything should be equalized up to the margin. 01:04:48.460 --> 01:04:54.125 align:middle line:84% We should end-- well, if we had marginals, 01:04:54.125 --> 01:04:55.250 align:middle line:90% so we don't have marginals. 01:04:55.250 --> 01:04:56.040 align:middle line:90% We have averages. 01:04:56.040 --> 01:04:58.587 align:middle line:84% So one reason-- maybe that could be related to it. 01:04:58.587 --> 01:05:00.170 align:middle line:84% But if we had marginals, you're right. 01:05:00.170 --> 01:05:03.470 align:middle line:84% I think we should be equalizing our output to the point where 01:05:03.470 --> 01:05:06.380 align:middle line:84% the marginal return in dollar terms 01:05:06.380 --> 01:05:07.580 align:middle line:90% is the same on every plot. 01:05:07.580 --> 01:05:10.232 align:middle line:84% So I think if we had marginals, I think that would be right. 01:05:10.232 --> 01:05:11.690 align:middle line:84% It certainly can't hurt to do this. 01:05:11.690 --> 01:05:13.800 align:middle line:84% But I don't see a good reason not to do that. 01:05:13.800 --> 01:05:14.300 align:middle line:90% Yeah. 01:05:14.300 --> 01:05:16.130 align:middle line:84% AUDIENCE: Why could we [INAUDIBLE]?? 01:05:16.130 --> 01:05:18.470 align:middle line:84% I was going to ask a similar question. 01:05:18.470 --> 01:05:20.050 align:middle line:84% Wherever you do the misallocation, 01:05:20.050 --> 01:05:22.370 align:middle line:84% it seems like marginal or [INAUDIBLE] 01:05:22.370 --> 01:05:24.260 align:middle line:84% capital and the right people [INAUDIBLE].. 01:05:24.260 --> 01:05:26.330 align:middle line:90% BEN OLKEN: Well, sorry. 01:05:26.330 --> 01:05:27.020 align:middle line:90% That's the goal. 01:05:27.020 --> 01:05:28.070 align:middle line:90% AUDIENCE: Yeah, yeah. 01:05:28.070 --> 01:05:31.070 align:middle line:84% So sometimes I think, what cases generally can 01:05:31.070 --> 01:05:33.425 align:middle line:90% we use averages to [INAUDIBLE]? 01:05:33.425 --> 01:05:35.550 align:middle line:84% Because it seems like that's what's heppening here. 01:05:35.550 --> 01:05:36.390 align:middle line:90% BEN OLKEN: Yeah, OK. 01:05:36.390 --> 01:05:37.598 align:middle line:90% So that's another good point. 01:05:37.598 --> 01:05:44.750 align:middle line:90% So OK, a couple of points. 01:05:44.750 --> 01:05:47.890 align:middle line:84% Number 1-- so in the misallocation literature, 01:05:47.890 --> 01:05:49.570 align:middle line:84% they also end up looking at averages 01:05:49.570 --> 01:05:50.890 align:middle line:90% more often than you may think. 01:05:50.890 --> 01:05:54.520 align:middle line:84% They write down a theory, which is about marginals, right? 01:05:54.520 --> 01:05:56.950 align:middle line:84% But actually, they have to assume 01:05:56.950 --> 01:05:58.540 align:middle line:90% a functional form in order to-- 01:05:58.540 --> 01:06:00.018 align:middle line:84% what they observe is just output. 01:06:00.018 --> 01:06:01.810 align:middle line:84% So you have to write down a functional form 01:06:01.810 --> 01:06:03.643 align:middle line:84% to get back the marginals from the averages. 01:06:03.643 --> 01:06:07.517 align:middle line:84% So I think a general point that I think all of you 01:06:07.517 --> 01:06:09.850 align:middle line:84% are making, which is totally right, which you're making, 01:06:09.850 --> 01:06:12.120 align:middle line:84% which is totally right is both in that literature 01:06:12.120 --> 01:06:14.280 align:middle line:84% and in this literature, the thing we really 01:06:14.280 --> 01:06:16.350 align:middle line:84% want is to estimate the marginals 01:06:16.350 --> 01:06:18.390 align:middle line:84% and randomly shock different things to estimate 01:06:18.390 --> 01:06:19.780 align:middle line:90% the marginal returns. 01:06:19.780 --> 01:06:24.360 align:middle line:84% So the misallocation literature does that in functional form. 01:06:24.360 --> 01:06:26.040 align:middle line:84% This paper, I think, also essentially 01:06:26.040 --> 01:06:28.050 align:middle line:84% deals with it in functional form, right? 01:06:28.050 --> 01:06:29.860 align:middle line:90% So what are they going to do? 01:06:29.860 --> 01:06:32.082 align:middle line:84% They're going to do two things, I think. 01:06:32.082 --> 01:06:33.040 align:middle line:90% What is he going to do? 01:06:33.040 --> 01:06:35.020 align:middle line:84% He's going to do two things to answer that. 01:06:35.020 --> 01:06:37.187 align:middle line:84% The first is he's going to estimate these production 01:06:37.187 --> 01:06:38.220 align:middle line:90% functions, right? 01:06:38.220 --> 01:06:41.010 align:middle line:84% And so they look similar on the two different things. 01:06:41.010 --> 01:06:42.510 align:middle line:90% And that's, I think, going to-- 01:06:42.510 --> 01:06:46.050 align:middle line:84% and unless there's some big, unobserved term that 01:06:46.050 --> 01:06:51.332 align:middle line:84% would change the marginals in some way-- 01:06:51.332 --> 01:06:53.290 align:middle line:84% the fact that production functions look similar 01:06:53.290 --> 01:06:54.575 align:middle line:90% may help you. 01:06:54.575 --> 01:06:56.200 align:middle line:84% The second thing, though, is he's going 01:06:56.200 --> 01:06:57.880 align:middle line:90% to directly look at inputs. 01:06:57.880 --> 01:07:01.690 align:middle line:84% And he's going to make this argument that the inputs look 01:07:01.690 --> 01:07:04.010 align:middle line:90% way way, way less. 01:07:04.010 --> 01:07:06.950 align:middle line:84% And the returns to the inputs, at least 01:07:06.950 --> 01:07:12.590 align:middle line:84% estimated cross-sectionally, look still high and concave. 01:07:12.590 --> 01:07:13.910 align:middle line:90% They look concave. 01:07:13.910 --> 01:07:16.280 align:middle line:84% So he's going to basically say, if we think 01:07:16.280 --> 01:07:18.710 align:middle line:84% that the returns to these things are concave 01:07:18.710 --> 01:07:21.980 align:middle line:84% and we see much lower inputs on one versus the other, 01:07:21.980 --> 01:07:24.110 align:middle line:84% it is probably the case that the marginals are not 01:07:24.110 --> 01:07:26.030 align:middle line:90% being equated. 01:07:26.030 --> 01:07:29.557 align:middle line:84% That's basically the argument he's going to make. 01:07:29.557 --> 01:07:31.640 align:middle line:84% Is that a reasonable answer, or is that the answer 01:07:31.640 --> 01:07:32.570 align:middle line:90% to your question? 01:07:32.570 --> 01:07:32.900 align:middle line:90% AUDIENCE: Yeah. 01:07:32.900 --> 01:07:34.050 align:middle line:84% When I asked the question, I was-- 01:07:34.050 --> 01:07:34.690 align:middle line:90% BEN OLKEN: What, sorry? 01:07:34.690 --> 01:07:36.091 align:middle line:84% AUDIENCE: I just had [? a quick ?] 01:07:36.091 --> 01:07:37.049 align:middle line:90% [? question on that. ?] 01:07:37.049 --> 01:07:37.870 align:middle line:90% BEN OLKEN: Yeah. 01:07:37.870 --> 01:07:39.662 align:middle line:84% So I think that's what they're going to do. 01:07:39.662 --> 01:07:43.210 align:middle line:84% But I think you're absolutely right to ask that question 01:07:43.210 --> 01:07:46.900 align:middle line:84% because I think lots of these papers 01:07:46.900 --> 01:07:50.320 align:middle line:84% make arguments about equating things 01:07:50.320 --> 01:07:52.750 align:middle line:90% which are true in marginals. 01:07:52.750 --> 01:07:55.750 align:middle line:84% But unless you have random shocks of things, 01:07:55.750 --> 01:07:59.890 align:middle line:84% you don't necessarily actually estimate the marginals. 01:07:59.890 --> 01:08:01.180 align:middle line:90% Yeah. 01:08:01.180 --> 01:08:02.338 align:middle line:90% Other questions? 01:08:02.338 --> 01:08:04.970 align:middle line:90% 01:08:04.970 --> 01:08:05.590 align:middle line:90% OK. 01:08:05.590 --> 01:08:09.490 align:middle line:84% So the key feature in this context 01:08:09.490 --> 01:08:13.940 align:middle line:84% is that men and women control different plots, all right? 01:08:13.940 --> 01:08:15.650 align:middle line:90% By the way, and he has data. 01:08:15.650 --> 01:08:18.310 align:middle line:84% As at least one of you pointed out in the comments, 01:08:18.310 --> 01:08:21.290 align:middle line:84% the data here is incredibly rich. 01:08:21.290 --> 01:08:23.118 align:middle line:84% To both have multiple plots and to have 01:08:23.118 --> 01:08:25.160 align:middle line:84% data on all the different plots is kind of really 01:08:25.160 --> 01:08:26.585 align:middle line:90% pretty remarkable. 01:08:26.585 --> 01:08:27.085 align:middle line:90% OK? 01:08:27.085 --> 01:08:32.240 align:middle line:90% 01:08:32.240 --> 01:08:34.290 align:middle line:90% And he's going to focus on-- 01:08:34.290 --> 01:08:36.040 align:middle line:84% half the households, the man and the wives 01:08:36.040 --> 01:08:38.109 align:middle line:84% are harvesting the same crop on different plots 01:08:38.109 --> 01:08:39.109 align:middle line:90% of the same year. 01:08:39.109 --> 01:08:41.319 align:middle line:84% So that allows him to put in these fixed effects, 01:08:41.319 --> 01:08:45.609 align:middle line:84% although as per our previous point, if you could do it in-- 01:08:45.609 --> 01:08:47.330 align:middle line:84% oh, here's another reason, by the way, 01:08:47.330 --> 01:08:53.290 align:middle line:84% is that to do the thing in dollars, you need profits. 01:08:53.290 --> 01:08:57.368 align:middle line:84% So the marginal-- the thing you want to think about-- 01:08:57.368 --> 01:08:58.660 align:middle line:90% first of all, you need profits. 01:08:58.660 --> 01:08:59.840 align:middle line:90% And you need marginal profits. 01:08:59.840 --> 01:09:01.149 align:middle line:84% So the marginals, we already talked about. 01:09:01.149 --> 01:09:02.680 align:middle line:84% The other thing which is tricky is profits. 01:09:02.680 --> 01:09:04.120 align:middle line:84% You have to track off input cost. 01:09:04.120 --> 01:09:05.245 align:middle line:90% And that's actually tricky. 01:09:05.245 --> 01:09:09.880 align:middle line:84% So it's much easier to measure output than profits. 01:09:09.880 --> 01:09:10.993 align:middle line:90% So put that in mind, too. 01:09:10.993 --> 01:09:12.410 align:middle line:84% Because you also need the inputs-- 01:09:12.410 --> 01:09:13.862 align:middle line:90% the inputs and the input costs. 01:09:13.862 --> 01:09:15.779 align:middle line:84% So that's the other thing that makes it easier 01:09:15.779 --> 01:09:17.200 align:middle line:90% to do it in output. 01:09:17.200 --> 01:09:19.590 align:middle line:84% And so in some sense, what is the key result, right? 01:09:19.590 --> 01:09:24.520 align:middle line:84% The key result is that in general, 01:09:24.520 --> 01:09:28.330 align:middle line:84% that there's a negative coefficient on output 01:09:28.330 --> 01:09:37.529 align:middle line:84% for female plots controlling very flexibly for plot size 01:09:37.529 --> 01:09:41.550 align:middle line:84% and other plot characteristics, right? 01:09:41.550 --> 01:09:42.928 align:middle line:90% And I think one thing another-- 01:09:42.928 --> 01:09:44.470 align:middle line:84% something else you guys picked up on, 01:09:44.470 --> 01:09:46.095 align:middle line:84% which I think is nice about this paper, 01:09:46.095 --> 01:09:50.843 align:middle line:84% is they're controlling for these things very flexibly and not 01:09:50.843 --> 01:09:51.760 align:middle line:90% parametrically, right? 01:09:51.760 --> 01:09:54.730 align:middle line:84% We don't really have a strong view of how soil 01:09:54.730 --> 01:09:58.840 align:middle line:84% type 12 versus 37 effects output or whatever. 01:09:58.840 --> 01:10:02.380 align:middle line:84% So he's going to control for all that pretty flexibly. 01:10:02.380 --> 01:10:04.150 align:middle line:90% OK. 01:10:04.150 --> 01:10:07.352 align:middle line:90% So let me pause here. 01:10:07.352 --> 01:10:08.810 align:middle line:84% So what did you guys think of this? 01:10:08.810 --> 01:10:11.300 align:middle line:84% Did you find it convincing, not convincing? 01:10:11.300 --> 01:10:12.550 align:middle line:90% We talked about the marginals. 01:10:12.550 --> 01:10:17.005 align:middle line:84% We talked about a couple of other issues. 01:10:17.005 --> 01:10:18.130 align:middle line:90% What are some other issues? 01:10:18.130 --> 01:10:20.080 align:middle line:84% What are some other thoughts you guys had about this paper? 01:10:20.080 --> 01:10:22.455 align:middle line:84% And generally, it seemed like you guys more or less liked 01:10:22.455 --> 01:10:26.260 align:middle line:84% it, but did other people have other concerns from reading it? 01:10:26.260 --> 01:10:33.780 align:middle line:90% 01:10:33.780 --> 01:10:34.280 align:middle line:90% Come on. 01:10:34.280 --> 01:10:35.810 align:middle line:90% You all sent in comments. 01:10:35.810 --> 01:10:40.080 align:middle line:84% So does anyone have any other concerns about it? 01:10:40.080 --> 01:10:43.470 align:middle line:90% 01:10:43.470 --> 01:10:44.250 align:middle line:90% Yeah, Patrick. 01:10:44.250 --> 01:10:46.520 align:middle line:84% AUDIENCE: [INAUDIBLE] the control steps were-- 01:10:46.520 --> 01:10:49.063 align:middle line:84% fully absorbed the quality of the plot. 01:10:49.063 --> 01:10:49.730 align:middle line:90% BEN OLKEN: Yeah. 01:10:49.730 --> 01:10:50.480 align:middle line:90% No, right? 01:10:50.480 --> 01:10:54.710 align:middle line:84% So what you have to believe is that the controls are-- 01:10:54.710 --> 01:10:57.270 align:middle line:90% what he has is-- 01:10:57.270 --> 01:10:58.807 align:middle line:90% did I bring this? 01:10:58.807 --> 01:11:03.280 align:middle line:84% Tsh tsh tsh tsh tsh tsh tsh tsh tsh tsh tsh tsh. 01:11:03.280 --> 01:11:04.150 align:middle line:90% Did I bring this? 01:11:04.150 --> 01:11:06.160 align:middle line:90% No, I didn't bring this slide. 01:11:06.160 --> 01:11:10.805 align:middle line:84% But what he does is he shows you that basically, 01:11:10.805 --> 01:11:13.180 align:middle line:84% the results look pretty similar with and with-- actually, 01:11:13.180 --> 01:11:14.230 align:middle line:90% I think they look-- 01:11:14.230 --> 01:11:15.430 align:middle line:90% I don't [INAUDIBLE] look. 01:11:15.430 --> 01:11:17.710 align:middle line:84% But he shows you the results with him without the controls, 01:11:17.710 --> 01:11:18.210 align:middle line:90% right? 01:11:18.210 --> 01:11:28.070 align:middle line:84% And I think you'd have to believe that the results still 01:11:28.070 --> 01:11:30.080 align:middle line:84% go through with the really rich controls. 01:11:30.080 --> 01:11:30.920 align:middle line:84% And you'd have to believe that there's 01:11:30.920 --> 01:11:32.587 align:middle line:84% a lot of other unobservable stuff that's 01:11:32.587 --> 01:11:34.195 align:middle line:90% not captured by these controls. 01:11:34.195 --> 01:11:35.570 align:middle line:84% But yes, I think that's something 01:11:35.570 --> 01:11:36.740 align:middle line:90% you could think about. 01:11:36.740 --> 01:11:39.388 align:middle line:90% It would have to be that the-- 01:11:39.388 --> 01:11:41.930 align:middle line:84% but the other thing is he also shows you these direct results 01:11:41.930 --> 01:11:43.280 align:middle line:90% on input use. 01:11:43.280 --> 01:11:48.020 align:middle line:84% So you'd have to believe not just that the land was worse, 01:11:48.020 --> 01:11:50.330 align:middle line:84% but the marginal returns to the input use 01:11:50.330 --> 01:11:52.507 align:middle line:84% were a lot lower, which could be, right? 01:11:52.507 --> 01:11:54.590 align:middle line:84% But that would be the thing you'd have to believe. 01:11:54.590 --> 01:11:55.040 align:middle line:90% Oh, right. 01:11:55.040 --> 01:11:55.550 align:middle line:90% Sorry. 01:11:55.550 --> 01:11:56.540 align:middle line:90% That's the other point. 01:11:56.540 --> 01:11:58.010 align:middle line:84% Actually, based on the observables, 01:11:58.010 --> 01:12:00.030 align:middle line:84% the women have better quality plots. 01:12:00.030 --> 01:12:02.670 align:middle line:84% So you'd have to believe that the women's plots have 01:12:02.670 --> 01:12:04.980 align:middle line:84% some higher quality, but somehow the marginals are 01:12:04.980 --> 01:12:05.800 align:middle line:90% super duper low. 01:12:05.800 --> 01:12:08.467 align:middle line:84% And that's why they're not-- and the marginal returns and inputs 01:12:08.467 --> 01:12:09.390 align:middle line:90% are super duper low. 01:12:09.390 --> 01:12:19.480 align:middle line:90% 01:12:19.480 --> 01:12:20.130 align:middle line:90% Other thoughts? 01:12:20.130 --> 01:12:20.755 align:middle line:90% Other comments? 01:12:20.755 --> 01:12:26.840 align:middle line:90% 01:12:26.840 --> 01:12:27.340 align:middle line:90% Yeah? 01:12:27.340 --> 01:12:29.173 align:middle line:84% AUDIENCE: [INAUDIBLE] think about the travel 01:12:29.173 --> 01:12:30.250 align:middle line:90% cost [INAUDIBLE]. 01:12:30.250 --> 01:12:32.322 align:middle line:90% Managing one plot seems-- 01:12:32.322 --> 01:12:33.280 align:middle line:90% BEN OLKEN: What, sorry? 01:12:33.280 --> 01:12:34.280 align:middle line:90% AUDIENCE: --less costly. 01:12:34.280 --> 01:12:38.050 align:middle line:84% Managing one plot might be less costly than managing two plots, 01:12:38.050 --> 01:12:40.155 align:middle line:90% although the market [INAUDIBLE]. 01:12:40.155 --> 01:12:41.530 align:middle line:84% BEN OLKEN: Yeah, but the question 01:12:41.530 --> 01:12:46.300 align:middle line:84% is why does that explain this stuff, where you're basically 01:12:46.300 --> 01:12:51.114 align:middle line:84% using less of other inputs on the woman's plot? 01:12:51.114 --> 01:12:53.490 align:middle line:84% AUDIENCE: Yeah, the location of the plots 01:12:53.490 --> 01:12:56.130 align:middle line:84% might be farther away from the husband, for example. 01:12:56.130 --> 01:12:59.470 align:middle line:84% So it's more costly to travel there, 01:12:59.470 --> 01:13:02.170 align:middle line:84% to travel to the best plot rather than just 01:13:02.170 --> 01:13:04.450 align:middle line:84% towards the plots around the house 01:13:04.450 --> 01:13:06.425 align:middle line:84% to reach my [? disproportionality ?] 01:13:06.425 --> 01:13:09.043 align:middle line:90% [INAUDIBLE]. 01:13:09.043 --> 01:13:09.710 align:middle line:90% BEN OLKEN: Yeah. 01:13:09.710 --> 01:13:10.890 align:middle line:90% So that could be. 01:13:10.890 --> 01:13:14.190 align:middle line:84% Does anyone remember if he controls for distance 01:13:14.190 --> 01:13:14.788 align:middle line:90% to the plot? 01:13:14.788 --> 01:13:16.580 align:middle line:84% AUDIENCE: There's one version where he just 01:13:16.580 --> 01:13:18.402 align:middle line:90% said one's very close. 01:13:18.402 --> 01:13:19.490 align:middle line:90% BEN OLKEN: OK, right. 01:13:19.490 --> 01:13:21.900 align:middle line:90% [INAUDIBLE] 01:13:21.900 --> 01:13:22.400 align:middle line:90% Yeah? 01:13:22.400 --> 01:13:23.025 align:middle line:90% AUDIENCE: Wait. 01:13:23.025 --> 01:13:24.395 align:middle line:90% I forgot. 01:13:24.395 --> 01:13:28.630 align:middle line:84% Can plots be reallocate-- switched, reallocated? 01:13:28.630 --> 01:13:30.930 align:middle line:90% Or are they-- 01:13:30.930 --> 01:13:32.430 align:middle line:84% BEN OLKEN: So remember, this came up 01:13:32.430 --> 01:13:33.430 align:middle line:90% in some of the comments. 01:13:33.430 --> 01:13:40.683 align:middle line:90% 01:13:40.683 --> 01:13:42.350 align:middle line:84% AUDIENCE: You mean moving the household? 01:13:42.350 --> 01:13:42.640 align:middle line:90% Yeah 01:13:42.640 --> 01:13:43.390 align:middle line:90% BEN OLKEN: Yeah. 01:13:43.390 --> 01:13:45.990 align:middle line:84% AUDIENCE: I think he was saying that it was allocated 01:13:45.990 --> 01:13:47.508 align:middle line:90% by the type of marriage. 01:13:47.508 --> 01:13:48.300 align:middle line:90% BEN OLKEN: Exactly. 01:13:48.300 --> 01:13:49.600 align:middle line:84% AUDIENCE: So I think it was [INAUDIBLE].. 01:13:49.600 --> 01:13:50.267 align:middle line:90% BEN OLKEN: Yeah. 01:13:50.267 --> 01:13:52.730 align:middle line:90% 01:13:52.730 --> 01:13:58.760 align:middle line:84% AUDIENCE: And do we know if you use the plot more, 01:13:58.760 --> 01:14:01.990 align:middle line:90% the quality goes down? 01:14:01.990 --> 01:14:03.140 align:middle line:90% BEN OLKEN: That is true. 01:14:03.140 --> 01:14:07.090 align:middle line:84% So well, I think you need to at least lay them 01:14:07.090 --> 01:14:08.890 align:middle line:90% fallow for some period of time. 01:14:08.890 --> 01:14:12.110 align:middle line:84% But there can be this notion of overuse. 01:14:12.110 --> 01:14:14.480 align:middle line:84% But would that be a problem for this? 01:14:14.480 --> 01:14:18.140 align:middle line:84% AUDIENCE: I guess I'm wondering if the notion of-- 01:14:18.140 --> 01:14:22.370 align:middle line:84% if you're measuring that there's higher quality 01:14:22.370 --> 01:14:25.850 align:middle line:84% at a given time slice of the woman's plot 01:14:25.850 --> 01:14:28.115 align:middle line:90% that's just being used less-- 01:14:28.115 --> 01:14:31.432 align:middle line:90% 01:14:31.432 --> 01:14:32.642 align:middle line:90% that might-- 01:14:32.642 --> 01:14:33.600 align:middle line:90% BEN OLKEN: Yeah, right. 01:14:33.600 --> 01:14:36.300 align:middle line:90% So but right. 01:14:36.300 --> 01:14:38.850 align:middle line:84% But you'd have to tell some funky story, where 01:14:38.850 --> 01:14:41.100 align:middle line:84% basically the year they happened to do the survey was 01:14:41.100 --> 01:14:43.475 align:middle line:84% the year that all the women decided to let their plots be 01:14:43.475 --> 01:14:45.402 align:middle line:90% fallow or whatever, right? 01:14:45.402 --> 01:14:47.610 align:middle line:84% There's no reason to think that would be particularly 01:14:47.610 --> 01:14:48.390 align:middle line:90% true in the survey year. 01:14:48.390 --> 01:14:50.098 align:middle line:84% And so you'd think it should average out. 01:14:50.098 --> 01:14:53.670 align:middle line:90% 01:14:53.670 --> 01:14:56.010 align:middle line:90% Other thoughts? 01:14:56.010 --> 01:14:57.180 align:middle line:90% OK, I'm almost out of time. 01:14:57.180 --> 01:14:59.100 align:middle line:84% But I think you guys had a bunch of interesting comments 01:14:59.100 --> 01:14:59.600 align:middle line:90% on this. 01:14:59.600 --> 01:15:00.900 align:middle line:90% So I would encourage you to-- 01:15:00.900 --> 01:15:03.030 align:middle line:84% I think there's lots of interesting things going on 01:15:03.030 --> 01:15:03.530 align:middle line:90% here. 01:15:03.530 --> 01:15:07.990 align:middle line:90% 01:15:07.990 --> 01:15:17.170 align:middle line:84% The last thing I wanted to mention, doo-doo-doo-doo-doo, 01:15:17.170 --> 01:15:21.073 align:middle line:84% was I just wanted to mention two things before I conclude 01:15:21.073 --> 01:15:23.740 align:middle line:84% because I think there's a lot of stuff here that I won't be able 01:15:23.740 --> 01:15:26.065 align:middle line:84% to talk to because Esther has more gender stuff she's going 01:15:26.065 --> 01:15:27.482 align:middle line:84% to talk about in the last lecture. 01:15:27.482 --> 01:15:29.060 align:middle line:90% I wanted to mention two things. 01:15:29.060 --> 01:15:34.210 align:middle line:84% Number 1 is I wanted to mention-- 01:15:34.210 --> 01:15:36.760 align:middle line:84% some of you mentioned this idea of randomly dropping income 01:15:36.760 --> 01:15:37.270 align:middle line:90% on people. 01:15:37.270 --> 01:15:42.857 align:middle line:84% This paper by Robinson, I think, is a little bit closer 01:15:42.857 --> 01:15:44.440 align:middle line:84% to what some of you were talking about 01:15:44.440 --> 01:15:46.310 align:middle line:84% vis-a-vis the GiveDirectly example. 01:15:46.310 --> 01:15:50.200 align:middle line:90% So he's going to have random-- 01:15:50.200 --> 01:15:53.740 align:middle line:84% both spouses have random chance of receiving smaller income 01:15:53.740 --> 01:15:54.640 align:middle line:90% drops. 01:15:54.640 --> 01:15:59.800 align:middle line:84% And look at the web of these things are different. 01:15:59.800 --> 01:16:03.820 align:middle line:84% That's just another randomized example I wanted to mention. 01:16:03.820 --> 01:16:06.850 align:middle line:84% And he does-- so I just wanted to mention 01:16:06.850 --> 01:16:08.295 align:middle line:90% that one as one example. 01:16:08.295 --> 01:16:09.670 align:middle line:84% And the final thing I also wanted 01:16:09.670 --> 01:16:12.670 align:middle line:84% to mention that I'm not going to talk about is dynamics. 01:16:12.670 --> 01:16:16.170 align:middle line:84% So everything we've talked about has been static, right? 01:16:16.170 --> 01:16:17.670 align:middle line:84% And I just wanted to mention there's 01:16:17.670 --> 01:16:19.185 align:middle line:84% a whole additional literature on-- 01:16:19.185 --> 01:16:21.060 align:middle line:84% or there is a bit of an additional literature 01:16:21.060 --> 01:16:22.980 align:middle line:90% on dynamics. 01:16:22.980 --> 01:16:26.160 align:middle line:84% And I wanted to make sure to have an opportunity 01:16:26.160 --> 01:16:30.900 align:middle line:84% to display my favorite cartoon on this subject, which 01:16:30.900 --> 01:16:35.610 align:middle line:84% is about dynamic allocation between parents and kids. 01:16:35.610 --> 01:16:40.860 align:middle line:84% But more seriously, there is this other idea about-- 01:16:40.860 --> 01:16:44.580 align:middle line:84% this Jensen and Miller paper, which is about whether or not I 01:16:44.580 --> 01:16:47.340 align:middle line:90% would-- 01:16:47.340 --> 01:16:50.160 align:middle line:84% if parents and kids can't contract, whether or not 01:16:50.160 --> 01:16:53.910 align:middle line:84% I want to do things to my kids to make sure that later, 01:16:53.910 --> 01:16:56.940 align:middle line:84% they'll behave nicely to me when they're grown up. 01:16:56.940 --> 01:16:58.890 align:middle line:84% And you can think of that as about 01:16:58.890 --> 01:17:04.903 align:middle line:84% intergenerational friction, where I might-- in particular, 01:17:04.903 --> 01:17:06.570 align:middle line:84% the idea of this Jensen and Miller paper 01:17:06.570 --> 01:17:11.250 align:middle line:84% is that I may differentially underinvest in my kids 01:17:11.250 --> 01:17:15.000 align:middle line:84% because I want them to stay on the farm and stay close to me 01:17:15.000 --> 01:17:18.950 align:middle line:84% and provide old age support for me. 01:17:18.950 --> 01:17:21.777 align:middle line:84% I will not invest in them or give them the ability 01:17:21.777 --> 01:17:24.110 align:middle line:84% to go off and move off to the city, which might maximize 01:17:24.110 --> 01:17:27.380 align:middle line:84% our total income as a collective multigenerational household, 01:17:27.380 --> 01:17:29.400 align:middle line:84% but might mean that I might get less of it. 01:17:29.400 --> 01:17:32.390 align:middle line:84% And so I just wanted to flag that, that these issues are not 01:17:32.390 --> 01:17:33.230 align:middle line:90% just static. 01:17:33.230 --> 01:17:34.220 align:middle line:84% We've talked about them statically, 01:17:34.220 --> 01:17:36.303 align:middle line:84% about man versus woman's income, and how does that 01:17:36.303 --> 01:17:37.700 align:middle line:90% affect choices. 01:17:37.700 --> 01:17:39.440 align:middle line:84% But it could also be parents and kids. 01:17:39.440 --> 01:17:43.260 align:middle line:84% And understanding that this heterogeneity, 01:17:43.260 --> 01:17:46.460 align:middle line:84% this dynamic optimization frictions 01:17:46.460 --> 01:17:49.050 align:middle line:84% is another interesting thing we can talk about. 01:17:49.050 --> 01:17:49.550 align:middle line:90% OK. 01:17:49.550 --> 01:17:50.300 align:middle line:90% Sorry there was a lot there. 01:17:50.300 --> 01:17:51.290 align:middle line:90% I didn't get to finish. 01:17:51.290 --> 01:17:53.590 align:middle line:84% But Esther will take it over next time, 01:17:53.590 --> 01:17:55.692 align:middle line:90% then I'll come back after that. 01:17:55.692 --> 01:17:57.600 align:middle line:90% OK, thanks. 01:17:57.600 --> 01:18:09.000 align:middle line:90%