WEBVTT 00:00:00.000 --> 00:00:01.960 align:middle line:90% [SQUEAKING] 00:00:01.960 --> 00:00:03.430 align:middle line:90% [RUSTLING] 00:00:03.430 --> 00:00:06.370 align:middle line:90% [CLICKING] 00:00:06.370 --> 00:00:18.160 align:middle line:90% 00:00:18.160 --> 00:00:20.830 align:middle line:84% ESTHER DUFLO: So where we left it last time was that 00:00:20.830 --> 00:00:26.050 align:middle line:84% basically, the tests of the basic household 00:00:26.050 --> 00:00:29.050 align:middle line:84% models in particular, on the efficient household models, 00:00:29.050 --> 00:00:31.600 align:middle line:84% basically failed both in the production space 00:00:31.600 --> 00:00:33.430 align:middle line:90% and the consumption space. 00:00:33.430 --> 00:00:36.340 align:middle line:84% So we'd better think about something else in order 00:00:36.340 --> 00:00:38.350 align:middle line:90% to try to understand the family. 00:00:38.350 --> 00:00:41.440 align:middle line:84% And this is what a lot of the recent literature 00:00:41.440 --> 00:00:45.130 align:middle line:84% has tried to do without yet getting 00:00:45.130 --> 00:00:48.970 align:middle line:84% to another framework that would be a satisfactory alternative 00:00:48.970 --> 00:00:49.990 align:middle line:90% framework. 00:00:49.990 --> 00:00:54.250 align:middle line:90% But at least we have some clues. 00:00:54.250 --> 00:00:56.740 align:middle line:84% So I'll talk about some of this paper first. 00:00:56.740 --> 00:01:00.430 align:middle line:84% And then I'll do a unit on female labor 00:01:00.430 --> 00:01:04.314 align:middle line:84% for participation in gender on the labor market to complete. 00:01:04.314 --> 00:01:08.190 align:middle line:90% 00:01:08.190 --> 00:01:13.850 align:middle line:84% So one first set of issues that people have looked at 00:01:13.850 --> 00:01:16.490 align:middle line:90% is the information flows. 00:01:16.490 --> 00:01:18.230 align:middle line:84% Basically, one of the assumptions 00:01:18.230 --> 00:01:20.750 align:middle line:84% of the efficient household model is 00:01:20.750 --> 00:01:23.240 align:middle line:84% that people who are part of a family 00:01:23.240 --> 00:01:26.030 align:middle line:84% should do better than any random people playing 00:01:26.030 --> 00:01:29.060 align:middle line:84% some games simply because they know each other. 00:01:29.060 --> 00:01:32.090 align:middle line:84% And they have the ability to communicate easily. 00:01:32.090 --> 00:01:35.780 align:middle line:84% And this is something that has been put to the test most 00:01:35.780 --> 00:01:39.770 align:middle line:84% recently by a nice paper by Gautam Rao, 00:01:39.770 --> 00:01:43.340 align:middle line:84% Matthew Ridley, who is a student in the program who 00:01:43.340 --> 00:01:46.310 align:middle line:84% is on the job market this year, and Frank Schilbach, whom 00:01:46.310 --> 00:01:50.960 align:middle line:84% you know well, which looks at whether people 00:01:50.960 --> 00:01:54.290 align:middle line:84% share information that they both transparently know 00:01:54.290 --> 00:01:56.710 align:middle line:90% that they have. 00:01:56.710 --> 00:01:59.260 align:middle line:84% And what I'll show you in a minute 00:01:59.260 --> 00:02:01.570 align:middle line:84% is that people do not share information. 00:02:01.570 --> 00:02:03.490 align:middle line:84% In particular, husbands pay no attention, 00:02:03.490 --> 00:02:05.650 align:middle line:84% or pay little attention to the information 00:02:05.650 --> 00:02:10.310 align:middle line:84% that they know their wife has for a simple problem. 00:02:10.310 --> 00:02:12.650 align:middle line:84% So the experiment is the following. 00:02:12.650 --> 00:02:14.140 align:middle line:84% It's a social learning experiment 00:02:14.140 --> 00:02:18.670 align:middle line:84% with 400 couples and 500 strangers, 00:02:18.670 --> 00:02:21.730 align:middle line:84% which is trying to ask whether people respond similarly 00:02:21.730 --> 00:02:26.230 align:middle line:84% to info that was given to them or was given to their spouse, 00:02:26.230 --> 00:02:29.810 align:middle line:84% whether that's different from men and women, 00:02:29.810 --> 00:02:32.750 align:middle line:84% whether this is due to a lack of communication, 00:02:32.750 --> 00:02:35.570 align:middle line:84% or whether the information is not weighed appropriately, 00:02:35.570 --> 00:02:39.290 align:middle line:84% even if you know that it exists, and whether couples are 00:02:39.290 --> 00:02:42.710 align:middle line:84% behaving in any way differently from any random strangers 00:02:42.710 --> 00:02:45.750 align:middle line:90% on these dimensions. 00:02:45.750 --> 00:02:49.160 align:middle line:84% So what they did is that they gave people 00:02:49.160 --> 00:02:50.930 align:middle line:84% one of these games where you are trying 00:02:50.930 --> 00:02:53.600 align:middle line:84% to guess the share of red balls in an urn. 00:02:53.600 --> 00:02:57.930 align:middle line:84% And then you are making more money if you get closer. 00:02:57.930 --> 00:03:03.830 align:middle line:84% People then are given signals on what the share might be, 00:03:03.830 --> 00:03:09.710 align:middle line:84% either for themselves or their spouse in the spouse treatment. 00:03:09.710 --> 00:03:12.500 align:middle line:90% Their spouse can get a signal. 00:03:12.500 --> 00:03:14.660 align:middle line:84% And then they can discuss to find out whatever 00:03:14.660 --> 00:03:17.000 align:middle line:90% signal their spouse got. 00:03:17.000 --> 00:03:22.250 align:middle line:84% And then there is a third treatment 00:03:22.250 --> 00:03:24.350 align:middle line:84% where the spouse gets a signal they don't even 00:03:24.350 --> 00:03:25.610 align:middle line:90% need to discuss. 00:03:25.610 --> 00:03:28.260 align:middle line:84% Frank or Frank's [INAUDIBLE] is telling them, 00:03:28.260 --> 00:03:32.000 align:middle line:84% hey, it would be, for example, telling me, 00:03:32.000 --> 00:03:35.000 align:middle line:84% Abhijit has gotten the signal that there 00:03:35.000 --> 00:03:39.560 align:middle line:84% are 2/3 of blue on in the bowl or that in his own drawer, 00:03:39.560 --> 00:03:43.250 align:middle line:90% he drew mostly blue balls. 00:03:43.250 --> 00:03:48.710 align:middle line:84% So the key outcome of interest is in the Bayesian updating 00:03:48.710 --> 00:03:50.570 align:middle line:90% as a function of your signal-- 00:03:50.570 --> 00:03:52.850 align:middle line:84% how much weight you put on your own signals 00:03:52.850 --> 00:03:55.280 align:middle line:90% depending on who uncovered them. 00:03:55.280 --> 00:03:57.650 align:middle line:84% So clearly, if there is information 00:03:57.650 --> 00:04:01.220 align:middle line:84% pulling in a household, it's nothing very tricky here. 00:04:01.220 --> 00:04:03.650 align:middle line:84% You should weigh your information, your own 00:04:03.650 --> 00:04:06.840 align:middle line:84% and your spouse's information equally. 00:04:06.840 --> 00:04:09.600 align:middle line:84% So in the third treatment, you should clearly 00:04:09.600 --> 00:04:11.880 align:middle line:84% weigh the information of the spouse equally. 00:04:11.880 --> 00:04:13.830 align:middle line:84% And then maybe in the second treatment, 00:04:13.830 --> 00:04:15.960 align:middle line:84% if people find it difficult to communicate, 00:04:15.960 --> 00:04:19.709 align:middle line:84% you will lose something a little bit, 00:04:19.709 --> 00:04:21.930 align:middle line:84% not that here there is nothing like-- 00:04:21.930 --> 00:04:23.310 align:middle line:90% there is nothing strategic. 00:04:23.310 --> 00:04:27.000 align:middle line:84% The earnings, the proceeds of the experiment 00:04:27.000 --> 00:04:29.560 align:middle line:84% are shared between the husbands and the wife. 00:04:29.560 --> 00:04:33.180 align:middle line:84% So the dominant strategy is clearly 00:04:33.180 --> 00:04:34.680 align:middle line:90% to do as well as possible. 00:04:34.680 --> 00:04:38.880 align:middle line:84% There is nothing that you can gain by not paying attention 00:04:38.880 --> 00:04:39.970 align:middle line:90% to the full signal. 00:04:39.970 --> 00:04:41.970 align:middle line:84% So it's really focusing on this-- 00:04:41.970 --> 00:04:45.330 align:middle line:84% are you even able to communicate or to pay attention 00:04:45.330 --> 00:04:49.100 align:middle line:90% to what the other has to say? 00:04:49.100 --> 00:04:54.230 align:middle line:84% So the bottom line-- and I guess that's the key result-- 00:04:54.230 --> 00:04:59.180 align:middle line:84% is that husbands heavily discount the weight 00:04:59.180 --> 00:05:02.840 align:middle line:84% information, even when it's perfectly communicated. 00:05:02.840 --> 00:05:07.310 align:middle line:84% So you can see, as a function of what people eventually-- yes, 00:05:07.310 --> 00:05:10.070 align:middle line:90% you can, backward in-depth. 00:05:10.070 --> 00:05:12.950 align:middle line:84% What is the weight that they give to their own information, 00:05:12.950 --> 00:05:15.470 align:middle line:84% to the information given to the spouse, 00:05:15.470 --> 00:05:20.000 align:middle line:84% either consummated directly or by discussion? 00:05:20.000 --> 00:05:23.690 align:middle line:84% And basically, what they find is people put a lot of weight 00:05:23.690 --> 00:05:26.630 align:middle line:84% on any new information they get by signal, 00:05:26.630 --> 00:05:30.380 align:middle line:84% which suggests that their own prior is 00:05:30.380 --> 00:05:31.850 align:middle line:90% quite flat, as it should be. 00:05:31.850 --> 00:05:33.300 align:middle line:90% Because how would you know? 00:05:33.300 --> 00:05:35.240 align:middle line:84% So people put a lot of information 00:05:35.240 --> 00:05:38.870 align:middle line:84% on their own signal, but they put less information-- 00:05:38.870 --> 00:05:41.300 align:middle line:84% much less information-- on a signal 00:05:41.300 --> 00:05:44.480 align:middle line:84% that's given to their wife, and that they could discuss about. 00:05:44.480 --> 00:05:48.920 align:middle line:84% And strangely enough, when they are told, instead of here's 00:05:48.920 --> 00:05:51.560 align:middle line:84% your signal, here is your spouse's signal, 00:05:51.560 --> 00:05:53.580 align:middle line:84% where the really should be exactly the same-- 00:05:53.580 --> 00:05:55.100 align:middle line:84% there is no discussion issue-- they 00:05:55.100 --> 00:05:58.960 align:middle line:90% pay zero attention to this. 00:05:58.960 --> 00:06:03.550 align:middle line:84% So that's pretty striking in the sense of it's 00:06:03.550 --> 00:06:05.500 align:middle line:84% not even about lack of communication. 00:06:05.500 --> 00:06:09.940 align:middle line:84% It's just they're maybe so used to not paying attention 00:06:09.940 --> 00:06:13.900 align:middle line:84% to anything coming from the wife that even in this case, where 00:06:13.900 --> 00:06:19.200 align:middle line:84% it's signals about how many balls there is in the urns, 00:06:19.200 --> 00:06:21.220 align:middle line:90% they don't pay attention. 00:06:21.220 --> 00:06:24.370 align:middle line:84% For the wife, that's not the same thing. 00:06:24.370 --> 00:06:25.420 align:middle line:90% I used to pay-- 00:06:25.420 --> 00:06:29.170 align:middle line:84% the wives actually pay just as much attention 00:06:29.170 --> 00:06:31.870 align:middle line:84% to the information given to the husband 00:06:31.870 --> 00:06:34.780 align:middle line:84% and to the information that they get themselves, which 00:06:34.780 --> 00:06:37.400 align:middle line:90% is the rational thing to do. 00:06:37.400 --> 00:06:40.720 align:middle line:84% So you might wonder whether it's about gender in general, 00:06:40.720 --> 00:06:45.700 align:middle line:84% or just simply women being more smarter, or more open-minded, 00:06:45.700 --> 00:06:51.370 align:middle line:84% or more paying more attention to the information of others 00:06:51.370 --> 00:06:52.630 align:middle line:90% in general. 00:06:52.630 --> 00:06:56.800 align:middle line:84% But in fact, when you do the information with strangers-- 00:06:56.800 --> 00:06:59.290 align:middle line:84% so you can do the same thing-- instead of pairing you with 00:06:59.290 --> 00:07:01.690 align:middle line:84% your spouse, I'd pair you with a perfect stranger-- 00:07:01.690 --> 00:07:03.970 align:middle line:84% similarly, with a perfect stranger, 00:07:03.970 --> 00:07:08.680 align:middle line:84% you should be able to discuss if they have the signal. 00:07:08.680 --> 00:07:10.210 align:middle line:84% And you have to elicit it from them. 00:07:10.210 --> 00:07:13.360 align:middle line:84% Or if I give directly, this is the signal that Ricky got, 00:07:13.360 --> 00:07:17.020 align:middle line:84% you should just treat this exactly as your own. 00:07:17.020 --> 00:07:19.882 align:middle line:84% And then I didn't put back the graphs, I think. 00:07:19.882 --> 00:07:21.340 align:middle line:84% Yeah, I didn't put back the graphs. 00:07:21.340 --> 00:07:31.550 align:middle line:84% But the finding is that people pay little attention 00:07:31.550 --> 00:07:34.980 align:middle line:84% to the information given to strangers. 00:07:34.980 --> 00:07:39.430 align:middle line:84% And there, women do the same thing as men. 00:07:39.430 --> 00:07:42.790 align:middle line:84% So the fact that we see women paying attention 00:07:42.790 --> 00:07:46.930 align:middle line:84% to their husband is not due to the fact that women in general 00:07:46.930 --> 00:07:50.140 align:middle line:84% pay more attention to information that 00:07:50.140 --> 00:07:52.840 align:middle line:90% is transmitted by other people. 00:07:52.840 --> 00:07:55.810 align:middle line:84% But it's something about the household. 00:07:55.810 --> 00:07:58.000 align:middle line:84% Women are used to listening to their husband. 00:07:58.000 --> 00:08:01.000 align:middle line:84% And husbands are used to not to listen to their wife. 00:08:01.000 --> 00:08:03.570 align:middle line:90% And nobody listens to strangers. 00:08:03.570 --> 00:08:07.103 align:middle line:84% So irrespective of the household context, 00:08:07.103 --> 00:08:08.520 align:middle line:84% it's pretty interesting for social 00:08:08.520 --> 00:08:11.145 align:middle line:84% learning because that's going to limit a lot of social learning 00:08:11.145 --> 00:08:13.080 align:middle line:84% if nobody pays any attention to anything 00:08:13.080 --> 00:08:15.460 align:middle line:90% that someone else tells them. 00:08:15.460 --> 00:08:18.390 align:middle line:84% But for the point of view of the household, 00:08:18.390 --> 00:08:22.020 align:middle line:84% this really puts a clear wrench in the theory 00:08:22.020 --> 00:08:23.603 align:middle line:84% that the information communication 00:08:23.603 --> 00:08:25.020 align:middle line:84% is perfect in the household, which 00:08:25.020 --> 00:08:28.620 align:middle line:90% would go towards efficiency. 00:08:28.620 --> 00:08:32.740 align:middle line:84% So that might be one reason why efficiency is not 00:08:32.740 --> 00:08:34.360 align:middle line:90% achieved within the household. 00:08:34.360 --> 00:08:35.990 align:middle line:90% Any questions on this study? 00:08:35.990 --> 00:08:36.490 align:middle line:90% Yes. 00:08:36.490 --> 00:08:39.010 align:middle line:84% STUDENT: Did they offer interpretation for why 00:08:39.010 --> 00:08:42.820 align:middle line:90% they're perfectly communicated? 00:08:42.820 --> 00:08:45.220 align:middle line:84% You would think that the perfectly communicated 00:08:45.220 --> 00:08:46.592 align:middle line:90% in the discussion would-- 00:08:46.592 --> 00:08:47.342 align:middle line:90% ESTHER DUFLO: Yes. 00:08:47.342 --> 00:08:51.490 align:middle line:84% Yeah, this is quite surprising in that. 00:08:51.490 --> 00:08:54.220 align:middle line:84% Because it's not about the inability to communicate. 00:08:54.220 --> 00:08:56.470 align:middle line:84% It's just basically, you don't trust the information 00:08:56.470 --> 00:08:59.920 align:middle line:84% that comes from someone else even though it's information-- 00:08:59.920 --> 00:09:01.930 align:middle line:84% even though you're perfectly-- there's 00:09:01.930 --> 00:09:05.080 align:middle line:90% no agency of that other person. 00:09:05.080 --> 00:09:08.550 align:middle line:84% STUDENT: Why does the discussion have a dead end treatment then? 00:09:08.550 --> 00:09:12.360 align:middle line:90% ESTHER DUFLO: Then the direct-- 00:09:12.360 --> 00:09:13.155 align:middle line:90% I don't know. 00:09:13.155 --> 00:09:16.000 align:middle line:84% That's surprising and interesting. 00:09:16.000 --> 00:09:17.770 align:middle line:84% I don't think that we have a good answer. 00:09:17.770 --> 00:09:19.840 align:middle line:84% I don't think that they have, necessarily, a good answer. 00:09:19.840 --> 00:09:20.010 align:middle line:90% Yeah. 00:09:20.010 --> 00:09:20.510 align:middle line:90% Erica? 00:09:20.510 --> 00:09:24.252 align:middle line:84% STUDENT: Does the graph for when people have stranger-- 00:09:24.252 --> 00:09:26.460 align:middle line:84% when people are speaking with strangers look the same 00:09:26.460 --> 00:09:27.210 align:middle line:90% in terms of-- 00:09:27.210 --> 00:09:29.460 align:middle line:84% ESTHER DUFLO: Yeah, it looks like the husband's graph. 00:09:29.460 --> 00:09:30.970 align:middle line:90% STUDENT: OK. 00:09:30.970 --> 00:09:32.220 align:middle line:90% ESTHER DUFLO: I'm pretty sure. 00:09:32.220 --> 00:09:33.928 align:middle line:84% I certainly have a doubt of whether there 00:09:33.928 --> 00:09:35.950 align:middle line:90% is-- the order is the same. 00:09:35.950 --> 00:09:38.610 align:middle line:84% I don't know whether the jump is that big between directly 00:09:38.610 --> 00:09:41.316 align:middle line:84% communicated versus communication. 00:09:41.316 --> 00:09:45.142 align:middle line:90% 00:09:45.142 --> 00:09:47.350 align:middle line:84% You can probably check it by the end of this lecture. 00:09:47.350 --> 00:09:48.183 align:middle line:90% I have it somewhere. 00:09:48.183 --> 00:09:49.930 align:middle line:84% STUDENT: I guess it would be interesting 00:09:49.930 --> 00:09:52.240 align:middle line:84% if the spouse is signal by discussion 00:09:52.240 --> 00:09:54.631 align:middle line:84% is fatter than a perfect stranger, 00:09:54.631 --> 00:09:56.920 align:middle line:84% or if it's exactly the same as as a perfect stranger. 00:09:56.920 --> 00:09:58.712 align:middle line:84% ESTHER DUFLO: I'll check that for my phone. 00:09:58.712 --> 00:10:02.620 align:middle line:84% I'm very scared of moving my computer. 00:10:02.620 --> 00:10:04.994 align:middle line:84% Again, I have the slide somewhere. 00:10:04.994 --> 00:10:19.050 align:middle line:90% 00:10:19.050 --> 00:10:23.070 align:middle line:84% Yeah, that's the same except the stranger discussion 00:10:23.070 --> 00:10:28.170 align:middle line:84% is a little bit closer to the perfectly communicated signal, 00:10:28.170 --> 00:10:30.280 align:middle line:90% but the other is the same. 00:10:30.280 --> 00:10:35.230 align:middle line:84% And for women, again, the order is the same. 00:10:35.230 --> 00:10:38.350 align:middle line:84% And both bars are a little bit higher. 00:10:38.350 --> 00:10:41.410 align:middle line:84% So women don't apply zero information 00:10:41.410 --> 00:10:44.610 align:middle line:84% to the signals of others, but almost zero. 00:10:44.610 --> 00:10:50.340 align:middle line:90% 00:10:50.340 --> 00:10:53.120 align:middle line:90% So that's a cool paper. 00:10:53.120 --> 00:10:55.295 align:middle line:90% Very interesting, very creative. 00:10:55.295 --> 00:10:58.340 align:middle line:90% 00:10:58.340 --> 00:11:01.070 align:middle line:84% So in this case, we found woman paying attention 00:11:01.070 --> 00:11:05.010 align:middle line:90% to what their husband tell them. 00:11:05.010 --> 00:11:06.710 align:middle line:84% We find husband not paying attention 00:11:06.710 --> 00:11:08.420 align:middle line:90% to what their wives tell them. 00:11:08.420 --> 00:11:11.250 align:middle line:84% And then there is one more possible step, 00:11:11.250 --> 00:11:14.180 align:middle line:84% which is are you going to try and manipulate information 00:11:14.180 --> 00:11:16.290 align:middle line:90% when it's inconvenient for you? 00:11:16.290 --> 00:11:19.520 align:middle line:84% So in a paper by Nava Ashraf, Erica Field, 00:11:19.520 --> 00:11:23.900 align:middle line:84% and I think she's Jennifer Lee, women 00:11:23.900 --> 00:11:28.210 align:middle line:84% are proposed to jump the queue to get 00:11:28.210 --> 00:11:32.140 align:middle line:84% an injectable contraceptive that's valid for three months. 00:11:32.140 --> 00:11:41.500 align:middle line:84% Now this is in Zambia, where, like in many African countries, 00:11:41.500 --> 00:11:44.010 align:middle line:84% women and men have different views on how many children 00:11:44.010 --> 00:11:46.830 align:middle line:84% they would like to have, with men wanting 00:11:46.830 --> 00:11:49.193 align:middle line:84% to have more children than women, 00:11:49.193 --> 00:11:51.360 align:middle line:84% partly because, probably, they are not internalizing 00:11:51.360 --> 00:11:54.390 align:middle line:84% the fact that the chance of dying in childbirth 00:11:54.390 --> 00:11:56.360 align:middle line:90% is not at all zero. 00:11:56.360 --> 00:11:58.698 align:middle line:84% So they have these different preferences. 00:11:58.698 --> 00:12:00.365 align:middle line:84% That's well established in this context. 00:12:00.365 --> 00:12:02.950 align:middle line:90% 00:12:02.950 --> 00:12:04.890 align:middle line:84% So what they do in this experiment is 00:12:04.890 --> 00:12:07.320 align:middle line:84% that they are giving people-- so there is injectable 00:12:07.320 --> 00:12:10.540 align:middle line:84% contraceptive exists in Zambia, but there is a long queue 00:12:10.540 --> 00:12:11.550 align:middle line:90% to get it. 00:12:11.550 --> 00:12:14.290 align:middle line:84% They give people a voucher to jump the queue. 00:12:14.290 --> 00:12:15.660 align:middle line:90% And there are two treatments. 00:12:15.660 --> 00:12:18.502 align:middle line:84% Either it's the individual who gets the-- 00:12:18.502 --> 00:12:22.560 align:middle line:84% the individual woman who gets the voucher, or it's 00:12:22.560 --> 00:12:24.940 align:middle line:90% the couple who gets it together. 00:12:24.940 --> 00:12:27.220 align:middle line:90% You don't go to discuss it. 00:12:27.220 --> 00:12:32.760 align:middle line:84% And in fact, in many of the discussions in population 00:12:32.760 --> 00:12:35.760 align:middle line:84% council, et cetera, people who are discussing contraception, 00:12:35.760 --> 00:12:39.150 align:middle line:84% there is that view that involving their husbands, 00:12:39.150 --> 00:12:41.550 align:middle line:84% having a discussion with the husband on contraception 00:12:41.550 --> 00:12:42.990 align:middle line:90% is a good thing. 00:12:42.990 --> 00:12:44.910 align:middle line:90% Because you need their go-ahead. 00:12:44.910 --> 00:12:49.380 align:middle line:84% So in the policy world, the [? prior ?] on this 00:12:49.380 --> 00:12:52.590 align:middle line:84% would be that the effect on in terms of taking an appointment 00:12:52.590 --> 00:12:54.960 align:middle line:84% and ultimately getting the shot would 00:12:54.960 --> 00:12:58.680 align:middle line:84% be larger if there is a couple discussion than if given 00:12:58.680 --> 00:13:00.360 align:middle line:90% straight to the individual. 00:13:00.360 --> 00:13:03.360 align:middle line:84% But of course, it omits the fact that the woman could just 00:13:03.360 --> 00:13:04.710 align:middle line:90% decide to go get it done. 00:13:04.710 --> 00:13:07.260 align:middle line:90% And nobody would be any wiser. 00:13:07.260 --> 00:13:10.440 align:middle line:84% So if she anticipated a difference in opinion, 00:13:10.440 --> 00:13:12.300 align:middle line:84% she might just suppress the information 00:13:12.300 --> 00:13:15.300 align:middle line:84% and not tell anybody, and get it done. 00:13:15.300 --> 00:13:18.060 align:middle line:84% So that's basically a nice setup where 00:13:18.060 --> 00:13:19.380 align:middle line:90% you're running an experiment. 00:13:19.380 --> 00:13:26.670 align:middle line:84% And you don't know what the result might be. 00:13:26.670 --> 00:13:30.060 align:middle line:84% And what they find here very clearly is that-- 00:13:30.060 --> 00:13:36.120 align:middle line:84% so you are now comparing the individual versus couple. 00:13:36.120 --> 00:13:39.540 align:middle line:84% And people who were assigned to the couple treatment 00:13:39.540 --> 00:13:44.010 align:middle line:84% are less likely to redeem the voucher, 00:13:44.010 --> 00:13:49.560 align:middle line:84% and ultimately less likely to receive the injectable. 00:13:49.560 --> 00:13:52.950 align:middle line:84% So that's not quite significant when we take all of the women. 00:13:52.950 --> 00:13:58.680 align:middle line:84% But some women-- it involves a lot of 0 because many women 00:13:58.680 --> 00:13:59.700 align:middle line:90% never applied. 00:13:59.700 --> 00:14:03.390 align:middle line:84% If you look at the people who responded to the survey, 00:14:03.390 --> 00:14:09.900 align:middle line:84% there is a 26 percentage point decline in actually redeeming 00:14:09.900 --> 00:14:12.670 align:middle line:90% the voucher on a 65% basis. 00:14:12.670 --> 00:14:15.300 align:middle line:84% So it's something that people are interested in. 00:14:15.300 --> 00:14:19.330 align:middle line:90% 65% of the women do it. 00:14:19.330 --> 00:14:24.460 align:middle line:84% But this gets done by 26% percentage point 00:14:24.460 --> 00:14:27.610 align:middle line:84% if the men are involved for redeeming 00:14:27.610 --> 00:14:31.990 align:middle line:84% the voucher in terms of actually receiving the injectable. 00:14:31.990 --> 00:14:34.305 align:middle line:84% 30% of women actually receive the injectable 00:14:34.305 --> 00:14:35.740 align:middle line:90% in the individual treatment. 00:14:35.740 --> 00:14:39.810 align:middle line:84% But it goes down to 7% in the couple treatment. 00:14:39.810 --> 00:14:43.760 align:middle line:84% So here is clearly a case where the woman just 00:14:43.760 --> 00:14:45.850 align:middle line:84% didn't tell her husband and got it done. 00:14:45.850 --> 00:14:46.350 align:middle line:90% Yeah. 00:14:46.350 --> 00:14:51.230 align:middle line:84% STUDENT: So why, if the voucher is for skipping the-- 00:14:51.230 --> 00:14:52.910 align:middle line:84% moving forward in the queue, why do we 00:14:52.910 --> 00:14:55.310 align:middle line:84% expect there to be different outcomes 00:14:55.310 --> 00:14:59.213 align:middle line:84% before redeeming the voucher and receiving the injectable? 00:14:59.213 --> 00:15:01.630 align:middle line:84% ESTHER DUFLO: Yeah it's not till-- you redeem the voucher. 00:15:01.630 --> 00:15:02.890 align:middle line:90% You have your consultation. 00:15:02.890 --> 00:15:04.900 align:middle line:84% And then once you have your consultation, you discuss it. 00:15:04.900 --> 00:15:06.610 align:middle line:84% And then they might decide that they don't want to do it. 00:15:06.610 --> 00:15:08.320 align:middle line:90% Because it's three months. 00:15:08.320 --> 00:15:09.710 align:middle line:90% And there are side effects. 00:15:09.710 --> 00:15:13.100 align:middle line:84% And so the voucher is for the consultation. 00:15:13.100 --> 00:15:15.040 align:middle line:84% So of people who get the consultation, 00:15:15.040 --> 00:15:21.150 align:middle line:84% some people might get cold feet in the process. 00:15:21.150 --> 00:15:23.898 align:middle line:90% So that's the difference. 00:15:23.898 --> 00:15:26.530 align:middle line:90% 00:15:26.530 --> 00:15:31.020 align:middle line:84% So this is a setting where the most likely explanation 00:15:31.020 --> 00:15:33.630 align:middle line:84% is that the woman who got the [INAUDIBLE] never discussed 00:15:33.630 --> 00:15:35.340 align:middle line:90% it and then just got it done. 00:15:35.340 --> 00:15:38.640 align:middle line:90% 00:15:38.640 --> 00:15:41.730 align:middle line:84% So it's not necessarily the fact that households-- so this is 00:15:41.730 --> 00:15:44.640 align:middle line:90% woman suppressing information. 00:15:44.640 --> 00:15:46.933 align:middle line:84% Matt [? Lo ?] and Maddie McKelway, 00:15:46.933 --> 00:15:49.350 align:middle line:84% who were both students here when they were students here-- 00:15:49.350 --> 00:15:51.270 align:middle line:84% they're on the opposite experiment. 00:15:51.270 --> 00:15:53.995 align:middle line:84% And in fact, I wrote not a generalizable result. 00:15:53.995 --> 00:15:55.620 align:middle line:84% But now that I think about it, I should 00:15:55.620 --> 00:16:00.510 align:middle line:84% have written, that was information given to women 00:16:00.510 --> 00:16:02.040 align:middle line:90% and then hidden by women. 00:16:02.040 --> 00:16:03.570 align:middle line:84% And now this is information given 00:16:03.570 --> 00:16:06.025 align:middle line:90% to men on a different topic. 00:16:06.025 --> 00:16:07.900 align:middle line:84% So it's a different context, different topic. 00:16:07.900 --> 00:16:09.840 align:middle line:84% But one of the things that changes is the information 00:16:09.840 --> 00:16:10.740 align:middle line:90% is given to men. 00:16:10.740 --> 00:16:13.650 align:middle line:90% 00:16:13.650 --> 00:16:18.660 align:middle line:84% This is previewing women in the labor market. 00:16:18.660 --> 00:16:26.270 align:middle line:84% They gave either man or woman a voucher for access 00:16:26.270 --> 00:16:28.770 align:middle line:90% to a training program for women. 00:16:28.770 --> 00:16:30.920 align:middle line:84% So if they gave the voucher to the man, 00:16:30.920 --> 00:16:35.790 align:middle line:84% it was for his wife to access this training program. 00:16:35.790 --> 00:16:41.680 align:middle line:84% So they either had it given to the man privately-- 00:16:41.680 --> 00:16:44.860 align:middle line:84% so if he was not interested, he could just 00:16:44.860 --> 00:16:47.420 align:middle line:84% eat up the information and never discuss it. 00:16:47.420 --> 00:16:49.450 align:middle line:84% Or they gave it to the woman privately, 00:16:49.450 --> 00:16:53.200 align:middle line:84% or they gave it to them, forcing a discussion. 00:16:53.200 --> 00:16:55.540 align:middle line:84% So there again, the idea here in this case, 00:16:55.540 --> 00:16:57.380 align:middle line:84% the difference in opinion, as we will see, 00:16:57.380 --> 00:16:59.130 align:middle line:84% is that a lot of women would like to work. 00:16:59.130 --> 00:17:02.770 align:middle line:84% And most men prefer if their wives don't work. 00:17:02.770 --> 00:17:05.010 align:middle line:84% So the easiest thing for the husband 00:17:05.010 --> 00:17:08.230 align:middle line:84% would have been to just not even share it. 00:17:08.230 --> 00:17:11.970 align:middle line:84% In that case, the final take-up of the experiment 00:17:11.970 --> 00:17:14.069 align:middle line:84% might have been lower in the case 00:17:14.069 --> 00:17:19.319 align:middle line:84% where the voucher is given to the man only. 00:17:19.319 --> 00:17:23.560 align:middle line:84% In fact, what they find in terms of final enrollment 00:17:23.560 --> 00:17:29.740 align:middle line:84% is that the final enrollment of the woman if the husband gets 00:17:29.740 --> 00:17:34.930 align:middle line:84% the ticket is as large when the woman isn't 00:17:34.930 --> 00:17:38.330 align:middle line:84% even informed that the ticket exists as when she's informed. 00:17:38.330 --> 00:17:42.460 align:middle line:84% In fact, if anything, slightly larger if the woman 00:17:42.460 --> 00:17:44.860 align:middle line:90% is not informed. 00:17:44.860 --> 00:17:47.840 align:middle line:84% But let's say these two are the same. 00:17:47.840 --> 00:17:52.990 align:middle line:84% So it means that there is no additional-- 00:17:52.990 --> 00:17:55.990 align:middle line:84% there is no loss in allowing the husband 00:17:55.990 --> 00:18:00.640 align:middle line:84% to never discuss it with their wife 00:18:00.640 --> 00:18:04.260 align:middle line:84% because our wife is not even aware of the information. 00:18:04.260 --> 00:18:08.250 align:middle line:84% So even if a wife is aware of the information, 00:18:08.250 --> 00:18:10.130 align:middle line:84% finally, the outcome is the same, 00:18:10.130 --> 00:18:12.350 align:middle line:84% maybe because there is zero bargaining power 00:18:12.350 --> 00:18:14.060 align:middle line:90% that she has anyways. 00:18:14.060 --> 00:18:18.680 align:middle line:84% We're going back to the Frank and Matthew and Gautam thing. 00:18:18.680 --> 00:18:21.170 align:middle line:84% If you pay no attention to what she's going to say, 00:18:21.170 --> 00:18:22.073 align:middle line:90% you can let her know. 00:18:22.073 --> 00:18:23.240 align:middle line:90% Or you can not let her know. 00:18:23.240 --> 00:18:23.970 align:middle line:90% Doesn't matter. 00:18:23.970 --> 00:18:26.750 align:middle line:84% You are making the same decision eventually. 00:18:26.750 --> 00:18:30.240 align:middle line:84% But she doesn't seem to be hiding information. 00:18:30.240 --> 00:18:34.080 align:middle line:84% And then interestingly, in the discussion group, 00:18:34.080 --> 00:18:36.610 align:middle line:84% the final take-up actually goes down. 00:18:36.610 --> 00:18:39.690 align:middle line:84% So forcing a discussion on the issue-- 00:18:39.690 --> 00:18:42.300 align:middle line:84% which their intuition was it's going 00:18:42.300 --> 00:18:45.060 align:middle line:84% to increase the take-up because it's 00:18:45.060 --> 00:18:46.860 align:middle line:84% going to give a chance to the wife 00:18:46.860 --> 00:18:50.940 align:middle line:84% to say-- so it's a discussion with someone like a facilitator 00:18:50.940 --> 00:18:53.460 align:middle line:84% basically being present in the discussion. 00:18:53.460 --> 00:18:56.700 align:middle line:84% And according to the transcript of the facilitator, 00:18:56.700 --> 00:18:59.180 align:middle line:84% they are very polite and discussed 00:18:59.180 --> 00:19:01.150 align:middle line:90% the opportunity, and so on. 00:19:01.150 --> 00:19:05.070 align:middle line:84% And so the [? prior ?] here that it would make the thing go up 00:19:05.070 --> 00:19:07.620 align:middle line:84% because their husband could learn, 00:19:07.620 --> 00:19:10.620 align:middle line:84% the wife would have a chance to express her preferences, also 00:19:10.620 --> 00:19:12.930 align:middle line:84% to address, maybe, some of the concerns of the husband, 00:19:12.930 --> 00:19:14.110 align:middle line:90% and so on and so forth-- 00:19:14.110 --> 00:19:16.530 align:middle line:84% but in fact, the opposite happened. 00:19:16.530 --> 00:19:20.600 align:middle line:84% And when they discussed it, it crashes. 00:19:20.600 --> 00:19:24.290 align:middle line:84% So we don't have a good model for that. 00:19:24.290 --> 00:19:26.660 align:middle line:84% They proposed something where just 00:19:26.660 --> 00:19:31.400 align:middle line:84% discussing, having arguments is too costly. 00:19:31.400 --> 00:19:34.560 align:middle line:84% And people hate fighting about things. 00:19:34.560 --> 00:19:36.620 align:middle line:84% So if you force them to discuss, they actually 00:19:36.620 --> 00:19:37.580 align:middle line:90% shut down their mind. 00:19:37.580 --> 00:19:40.430 align:middle line:84% And they're like, don't even want to hear about it. 00:19:40.430 --> 00:19:45.890 align:middle line:84% When the wife get the ticket, the husband has to approve, 00:19:45.890 --> 00:19:47.400 align:middle line:90% at the end of the day. 00:19:47.400 --> 00:19:49.460 align:middle line:90% So the wife doesn't get-- 00:19:49.460 --> 00:19:51.050 align:middle line:90% it is asymmetric here. 00:19:51.050 --> 00:19:54.680 align:middle line:84% The wife doesn't get to enroll in the program 00:19:54.680 --> 00:19:56.840 align:middle line:90% without her husband knowing. 00:19:56.840 --> 00:20:00.200 align:middle line:84% So here we would expect that with or without information 00:20:00.200 --> 00:20:02.570 align:middle line:90% would be a similar enrollment. 00:20:02.570 --> 00:20:05.070 align:middle line:84% And that's more or less what we find. 00:20:05.070 --> 00:20:08.110 align:middle line:84% The discussion effect is lower in this case. 00:20:08.110 --> 00:20:10.650 align:middle line:84% It's clearly in the case of the husband, the fact 00:20:10.650 --> 00:20:14.210 align:middle line:84% that the discussion effect is lower. 00:20:14.210 --> 00:20:18.530 align:middle line:84% And interestingly, the overall take-up from the 00:20:18.530 --> 00:20:21.140 align:middle line:84% was getting the ticket from the husband getting the ticket is 00:20:21.140 --> 00:20:23.500 align:middle line:90% more or less similar. 00:20:23.500 --> 00:20:27.750 align:middle line:84% So maybe it's a little different than what 00:20:27.750 --> 00:20:30.210 align:middle line:84% we found in [INAUDIBLE],, where if the information gets 00:20:30.210 --> 00:20:33.210 align:middle line:84% from the wife, it's entirely discounted. 00:20:33.210 --> 00:20:38.880 align:middle line:84% Here it seems that the source of the information and the ticket 00:20:38.880 --> 00:20:42.490 align:middle line:84% doesn't seem to matter much for the final outcome. 00:20:42.490 --> 00:20:45.120 align:middle line:84% So just to say that these things are still-- 00:20:45.120 --> 00:20:48.090 align:middle line:84% I can't say that we have yet stylized patterns, 00:20:48.090 --> 00:20:53.290 align:middle line:84% but very interesting, very interesting set of facts. 00:20:53.290 --> 00:20:56.230 align:middle line:84% So let me stop here for general-- 00:20:56.230 --> 00:20:56.910 align:middle line:90% yeah. 00:20:56.910 --> 00:20:59.310 align:middle line:84% STUDENT: To what extent has this literature 00:20:59.310 --> 00:21:01.752 align:middle line:84% on gender norms and information sharing 00:21:01.752 --> 00:21:03.835 align:middle line:84% been replicated in other countries, other cultures 00:21:03.835 --> 00:21:04.750 align:middle line:90% and stuff like that? 00:21:04.750 --> 00:21:06.832 align:middle line:84% I think all these papers we looked at were India? 00:21:06.832 --> 00:21:07.540 align:middle line:90% ESTHER DUFLO: No. 00:21:07.540 --> 00:21:10.174 align:middle line:90% The Ashraf and Co. is in Zambia. 00:21:10.174 --> 00:21:10.674 align:middle line:90% STUDENT: OK. 00:21:10.674 --> 00:21:14.668 align:middle line:90% 00:21:14.668 --> 00:21:16.710 align:middle line:84% ESTHER DUFLO: There is some work in the US, which 00:21:16.710 --> 00:21:18.360 align:middle line:84% I'm going to discuss in a minute, 00:21:18.360 --> 00:21:20.010 align:middle line:90% on women in the labor supply. 00:21:20.010 --> 00:21:23.760 align:middle line:90% 00:21:23.760 --> 00:21:25.970 align:middle line:84% So I can't say the entire globe has been covered, 00:21:25.970 --> 00:21:28.110 align:middle line:90% but it's not just India. 00:21:28.110 --> 00:21:32.570 align:middle line:84% India has a particularly anti- female labor force 00:21:32.570 --> 00:21:33.810 align:middle line:90% participation. 00:21:33.810 --> 00:21:34.820 align:middle line:90% It does. 00:21:34.820 --> 00:21:38.990 align:middle line:84% The labor force participation is lowest in the world in India, 00:21:38.990 --> 00:21:41.480 align:middle line:84% lower in Saudi Arabia or other places 00:21:41.480 --> 00:21:44.220 align:middle line:84% that you would consider to be pretty antiwomen's work. 00:21:44.220 --> 00:21:45.800 align:middle line:84% So I think that's a good question 00:21:45.800 --> 00:21:48.080 align:middle line:84% that all of the work that is in India 00:21:48.080 --> 00:21:51.200 align:middle line:84% might be tainted by this particularly strong social norm 00:21:51.200 --> 00:21:52.610 align:middle line:90% against women working. 00:21:52.610 --> 00:21:56.228 align:middle line:84% But for example, a lot of the work 00:21:56.228 --> 00:21:58.770 align:middle line:84% that we've seen on the household actually was not from India. 00:21:58.770 --> 00:22:02.310 align:middle line:84% For example, you'd read my paper with Chris Udry 00:22:02.310 --> 00:22:04.160 align:middle line:90% is from Cote d'Ivoire. 00:22:04.160 --> 00:22:07.940 align:middle line:84% The [INAUDIBLE] paper is from Burkina Faso. 00:22:07.940 --> 00:22:11.750 align:middle line:84% So I actually do think there is a fair amount of things 00:22:11.750 --> 00:22:15.920 align:middle line:84% on the households that's not on India. 00:22:15.920 --> 00:22:18.140 align:middle line:84% And I think you're absolutely right that the culture 00:22:18.140 --> 00:22:20.540 align:middle line:90% setup matters a lot. 00:22:20.540 --> 00:22:23.540 align:middle line:84% In terms of women labor force participation, 00:22:23.540 --> 00:22:25.640 align:middle line:84% quite a bit of the work is on India, perhaps 00:22:25.640 --> 00:22:29.340 align:middle line:84% because this is such a surprisingly anti-women work 00:22:29.340 --> 00:22:29.840 align:middle line:90% setup. 00:22:29.840 --> 00:22:32.480 align:middle line:90% 00:22:32.480 --> 00:22:33.020 align:middle line:90% OK. 00:22:33.020 --> 00:22:35.540 align:middle line:84% Let me just continue talking about-- very relevant. 00:22:35.540 --> 00:22:37.460 align:middle line:84% I just put them in a different set of slides 00:22:37.460 --> 00:22:43.020 align:middle line:90% for having logical consistency. 00:22:43.020 --> 00:22:46.380 align:middle line:84% But it's very much the continuation 00:22:46.380 --> 00:22:49.950 align:middle line:84% of talking about the family with the angle of talking 00:22:49.950 --> 00:22:51.400 align:middle line:90% about women in the labor market. 00:22:51.400 --> 00:23:01.320 align:middle line:90% 00:23:01.320 --> 00:23:04.780 align:middle line:84% So this is a very active area of work right now. 00:23:04.780 --> 00:23:07.080 align:middle line:84% A lot of exciting stuff are being done. 00:23:07.080 --> 00:23:08.750 align:middle line:84% There's still more stuff to be done. 00:23:08.750 --> 00:23:13.610 align:middle line:84% But this is one of the places where there 00:23:13.610 --> 00:23:14.713 align:middle line:90% is a lot of recent papers. 00:23:14.713 --> 00:23:16.130 align:middle line:84% So I thought it made sense to have 00:23:16.130 --> 00:23:19.350 align:middle line:90% a whole lecture about that. 00:23:19.350 --> 00:23:25.580 align:middle line:84% So interestingly, if you read papers on bargaining power 00:23:25.580 --> 00:23:29.720 align:middle line:84% in a household in the US, the outcome 00:23:29.720 --> 00:23:32.450 align:middle line:84% that they show to show that women have more bargaining 00:23:32.450 --> 00:23:34.790 align:middle line:90% power is leisure. 00:23:34.790 --> 00:23:39.170 align:middle line:84% That is, female labor supply, which is seen as a sign of low. 00:23:39.170 --> 00:23:42.860 align:middle line:84% If women have to work, is seen as a sign of low bargaining 00:23:42.860 --> 00:23:44.000 align:middle line:90% power in the US. 00:23:44.000 --> 00:23:48.080 align:middle line:84% So for example, if you read Josh Angrist paper 00:23:48.080 --> 00:23:50.810 align:middle line:84% looking at the marriage market and showing 00:23:50.810 --> 00:23:55.100 align:middle line:84% that women who are from ethnic communities 00:23:55.100 --> 00:23:57.890 align:middle line:84% where there are more men relative to women 00:23:57.890 --> 00:24:01.290 align:middle line:84% have more bargaining power, the idea being 00:24:01.290 --> 00:24:03.000 align:middle line:84% that if the union breaks, it would 00:24:03.000 --> 00:24:05.670 align:middle line:90% be easier for them to remarry. 00:24:05.670 --> 00:24:08.220 align:middle line:84% Well, in that paper, the outcome is female labor force 00:24:08.220 --> 00:24:09.000 align:middle line:90% participation. 00:24:09.000 --> 00:24:11.042 align:middle line:84% What it shows is female labor force participation 00:24:11.042 --> 00:24:15.360 align:middle line:84% is lower in these groups where women have higher bargaining 00:24:15.360 --> 00:24:17.190 align:middle line:90% power. 00:24:17.190 --> 00:24:20.700 align:middle line:84% And in the [INAUDIBLE] papers from the US 00:24:20.700 --> 00:24:23.250 align:middle line:84% as well, this is pretty much always woman labor supply. 00:24:23.250 --> 00:24:25.710 align:middle line:84% They like that outcome because leisure 00:24:25.710 --> 00:24:27.930 align:middle line:84% is a good that is privately consumed 00:24:27.930 --> 00:24:29.400 align:middle line:90% and is a bit harder to share. 00:24:29.400 --> 00:24:30.750 align:middle line:90% So that makes sense. 00:24:30.750 --> 00:24:32.610 align:middle line:84% But of course, in most developing countries, 00:24:32.610 --> 00:24:37.230 align:middle line:84% that would be a silly outcome because female labor supply is 00:24:37.230 --> 00:24:41.610 align:middle line:84% actually considered to be, on the contrary, something 00:24:41.610 --> 00:24:44.830 align:middle line:84% that women want to do and their husband don't want them to do. 00:24:44.830 --> 00:24:53.010 align:middle line:84% So there is a nice review paper by Rohini Pande and co-authors, 00:24:53.010 --> 00:24:55.830 align:middle line:84% which is basically descriptive evidence of what women want 00:24:55.830 --> 00:24:57.510 align:middle line:90% to do and what men want to do. 00:24:57.510 --> 00:25:00.180 align:middle line:84% And what it shows is that most women in most 00:25:00.180 --> 00:25:02.370 align:middle line:84% developing countries, particularly in South Asia, 00:25:02.370 --> 00:25:04.410 align:middle line:84% but not only, women would like to work. 00:25:04.410 --> 00:25:07.320 align:middle line:84% And their husbands don't want them to work. 00:25:07.320 --> 00:25:09.590 align:middle line:84% And in fact, they are less likely to work. 00:25:09.590 --> 00:25:11.510 align:middle line:84% Just in terms of background, this 00:25:11.510 --> 00:25:15.740 align:middle line:84% is a bit anticipating what we're going to do in 772. 00:25:15.740 --> 00:25:19.190 align:middle line:84% But there is a recent paper by Hsieh 00:25:19.190 --> 00:25:21.650 align:middle line:90% published in Econometrica-- 00:25:21.650 --> 00:25:23.690 align:middle line:84% Hsieh and others, Chang-Tai Hsieh 00:25:23.690 --> 00:25:28.040 align:middle line:84% and a bunch of other people, which extend their framework, 00:25:28.040 --> 00:25:32.090 align:middle line:84% which you're going to become painfully familiar with in 772, 00:25:32.090 --> 00:25:36.770 align:middle line:84% on misallocation of resources to show that in the US, 00:25:36.770 --> 00:25:38.960 align:middle line:84% the entry of Black people and women 00:25:38.960 --> 00:25:42.170 align:middle line:84% in better paid occupations have led 00:25:42.170 --> 00:25:45.440 align:middle line:84% to a large increase in labor force productivity 00:25:45.440 --> 00:25:49.530 align:middle line:84% simply by allocating talent to where it should be going. 00:25:49.530 --> 00:25:52.110 align:middle line:84% And in particular for students in the second year who 00:25:52.110 --> 00:25:54.090 align:middle line:84% are thinking about their paper somehow, 00:25:54.090 --> 00:25:57.990 align:middle line:84% I absolutely love the first sentence of their abstract. 00:25:57.990 --> 00:26:01.320 align:middle line:84% It says, "In 1960, 94% of doctors and lawyers 00:26:01.320 --> 00:26:02.280 align:middle line:90% were white men. 00:26:02.280 --> 00:26:05.920 align:middle line:84% By 2010, the fraction was just 62%." 00:26:05.920 --> 00:26:09.570 align:middle line:84% So why do I love this way of writing the abstract? 00:26:09.570 --> 00:26:12.120 align:middle line:84% It gives you a very concrete fact 00:26:12.120 --> 00:26:15.480 align:middle line:84% that you might have known qualitatively, 00:26:15.480 --> 00:26:17.820 align:middle line:90% but not, maybe, quantitatively. 00:26:17.820 --> 00:26:20.940 align:middle line:84% And it sets the stage for their next question, which 00:26:20.940 --> 00:26:27.090 align:middle line:84% is presumably, it's not because they have become more talented 00:26:27.090 --> 00:26:28.560 align:middle line:90% at becoming lawyers. 00:26:28.560 --> 00:26:32.130 align:middle line:84% So this is related in part to barriers-- institutional, 00:26:32.130 --> 00:26:34.110 align:middle line:84% cultural, and social norms, et cetera-- 00:26:34.110 --> 00:26:37.320 align:middle line:84% that made it harder for them to access 00:26:37.320 --> 00:26:38.910 align:middle line:90% these jobs in the first place. 00:26:38.910 --> 00:26:41.850 align:middle line:84% And therefore, it should lead to a better allocation of talent 00:26:41.850 --> 00:26:44.670 align:middle line:90% and more productivity. 00:26:44.670 --> 00:26:46.530 align:middle line:84% So that's what the paper went on doing. 00:26:46.530 --> 00:26:49.650 align:middle line:84% And so they calibrate that 20% to 40% 00:26:49.650 --> 00:26:52.530 align:middle line:84% of output growth in the US can be explained by this. 00:26:52.530 --> 00:26:54.990 align:middle line:84% So given that it's a period where output growth was 00:26:54.990 --> 00:26:58.680 align:middle line:84% in general slower, this is saying 00:26:58.680 --> 00:27:01.830 align:middle line:84% it would have been even slower if women and Black people 00:27:01.830 --> 00:27:05.460 align:middle line:84% had not been able to access, in larger 00:27:05.460 --> 00:27:09.530 align:middle line:90% numbers, those occupations. 00:27:09.530 --> 00:27:12.760 align:middle line:84% So this first sentence of the abstract. 00:27:12.760 --> 00:27:18.970 align:middle line:90% It's wrap in one sentence. 00:27:18.970 --> 00:27:21.760 align:middle line:84% It just sets up the stage very nicely. 00:27:21.760 --> 00:27:23.950 align:middle line:84% It said, this is an important question. 00:27:23.950 --> 00:27:25.970 align:middle line:84% And of course, from there, he could have gone, 00:27:25.970 --> 00:27:27.130 align:middle line:90% I'm going to explain why. 00:27:27.130 --> 00:27:29.297 align:middle line:84% And in fact, that's not what they are interested in. 00:27:29.297 --> 00:27:30.820 align:middle line:84% It's explaining the consequences. 00:27:30.820 --> 00:27:33.530 align:middle line:84% But it sets up immediately why we should care. 00:27:33.530 --> 00:27:34.030 align:middle line:90% Yeah? 00:27:34.030 --> 00:27:35.530 align:middle line:84% STUDENT: So just to understand, it's 00:27:35.530 --> 00:27:40.050 align:middle line:90% just the striking quantitative. 00:27:40.050 --> 00:27:43.780 align:middle line:84% I feel a lot of abstracts start with a very concrete sentence, 00:27:43.780 --> 00:27:46.390 align:middle line:84% some facts, but maybe not as concrete as this. 00:27:46.390 --> 00:27:48.790 align:middle line:84% ESTHER DUFLO: I think it's a very concrete fact, somewhat 00:27:48.790 --> 00:27:51.220 align:middle line:84% unexpected in terms of the quantitative aspect. 00:27:51.220 --> 00:27:52.720 align:middle line:90% And that immediately shows-- 00:27:52.720 --> 00:27:55.030 align:middle line:84% I'm going to show you the impact of-- 00:27:55.030 --> 00:27:56.590 align:middle line:84% the next sentence goes, I'm going 00:27:56.590 --> 00:28:00.330 align:middle line:84% to show you the impact of this transition. 00:28:00.330 --> 00:28:03.650 align:middle line:84% And this is likely going to be large because that is a lot. 00:28:03.650 --> 00:28:08.700 align:middle line:90% 00:28:08.700 --> 00:28:16.870 align:middle line:84% Other paper from the US, which relates to this idea 00:28:16.870 --> 00:28:21.760 align:middle line:84% of woman in labor supply as actually something 00:28:21.760 --> 00:28:24.880 align:middle line:84% that they would like and that is constrained by their husbands 00:28:24.880 --> 00:28:27.460 align:middle line:90% in the US as well. 00:28:27.460 --> 00:28:31.840 align:middle line:84% Paper by Leo Bursztyn, Thomas Fujiwara, and Amanda Pallais 00:28:31.840 --> 00:28:34.210 align:middle line:84% in MBA at the top business school 00:28:34.210 --> 00:28:36.940 align:middle line:90% that's not very far from here. 00:28:36.940 --> 00:28:45.970 align:middle line:84% They had them work in their orientation session a very 00:28:45.970 --> 00:28:50.350 align:middle line:84% mixed group, mixed gender group, or in a single sex gender 00:28:50.350 --> 00:28:52.070 align:middle line:90% group. 00:28:52.070 --> 00:28:56.570 align:middle line:84% And then they had them say what type of job 00:28:56.570 --> 00:28:59.250 align:middle line:90% they want to apply to. 00:28:59.250 --> 00:29:03.330 align:middle line:84% And is there any way that they were supposed to say it. 00:29:03.330 --> 00:29:04.740 align:middle line:84% Either it was said, this is going 00:29:04.740 --> 00:29:07.170 align:middle line:84% to be discussed in your entire group, 00:29:07.170 --> 00:29:11.380 align:middle line:84% or this choice, this survey is going to remain private. 00:29:11.380 --> 00:29:17.880 align:middle line:84% So this is a typical Leo Bursztyn trick 00:29:17.880 --> 00:29:20.730 align:middle line:84% that you also have in their education 00:29:20.730 --> 00:29:26.880 align:middle line:84% paper on basically threatening to make 00:29:26.880 --> 00:29:31.080 align:middle line:84% an affirmation of value, public or not, which therefore 00:29:31.080 --> 00:29:33.660 align:middle line:84% makes it social, and potentially has 00:29:33.660 --> 00:29:37.480 align:middle line:90% it hit the social norms or not. 00:29:37.480 --> 00:29:42.210 align:middle line:84% But this is perhaps my favorite application of this trick. 00:29:42.210 --> 00:29:47.940 align:middle line:84% Because it seems very natural that you 00:29:47.940 --> 00:29:52.140 align:middle line:84% would either fill this survey privately or discuss it 00:29:52.140 --> 00:29:52.830 align:middle line:90% with the group. 00:29:52.830 --> 00:29:56.300 align:middle line:90% 00:29:56.300 --> 00:29:59.480 align:middle line:84% And this seems to be a large enough manipulation that people 00:29:59.480 --> 00:30:01.450 align:middle line:90% would be sensitive to it. 00:30:01.450 --> 00:30:04.060 align:middle line:84% Anyway, what they find is that when 00:30:04.060 --> 00:30:12.400 align:middle line:84% women are doing this choice of, are listing their choice 00:30:12.400 --> 00:30:13.630 align:middle line:90% of preferred careers-- 00:30:13.630 --> 00:30:18.130 align:middle line:84% this is very high stakes because it's information about careers. 00:30:18.130 --> 00:30:19.390 align:middle line:90% This is in the career field. 00:30:19.390 --> 00:30:24.730 align:middle line:84% So what they report is what they would have believed at that 00:30:24.730 --> 00:30:27.130 align:middle line:84% time-- subsequently, I assume, they had another chance 00:30:27.130 --> 00:30:28.450 align:middle line:90% to report-- 00:30:28.450 --> 00:30:31.210 align:middle line:84% is going to be what the counsellors are 00:30:31.210 --> 00:30:32.960 align:middle line:84% going to use to find them internships, 00:30:32.960 --> 00:30:35.590 align:middle line:90% which turns into job. 00:30:35.590 --> 00:30:38.380 align:middle line:84% So this is a very high stake decision. 00:30:38.380 --> 00:30:46.660 align:middle line:84% And in gender unified group, single gender group, 00:30:46.660 --> 00:30:52.100 align:middle line:84% it makes no difference to do it publicly or privately. 00:30:52.100 --> 00:30:55.550 align:middle line:84% But in groups where there are both men and women, 00:30:55.550 --> 00:31:00.830 align:middle line:84% women are less ambitious if they think that their choice is 00:31:00.830 --> 00:31:02.600 align:middle line:90% going to be publicly revealed. 00:31:02.600 --> 00:31:05.720 align:middle line:84% They display less ambitions in the treatment 00:31:05.720 --> 00:31:08.240 align:middle line:84% where they are made to believe that their choice is 00:31:08.240 --> 00:31:10.440 align:middle line:90% going to be publicly revealed. 00:31:10.440 --> 00:31:13.650 align:middle line:84% So that's why they call it acting wife because they 00:31:13.650 --> 00:31:18.110 align:middle line:90% think that the future wife-- 00:31:18.110 --> 00:31:20.250 align:middle line:90% sorry, let me finish-- 00:31:20.250 --> 00:31:22.110 align:middle line:84% that is true only for a single woman, 00:31:22.110 --> 00:31:23.370 align:middle line:90% but not for married women. 00:31:23.370 --> 00:31:26.100 align:middle line:90% 00:31:26.100 --> 00:31:33.400 align:middle line:84% So for a single woman in mixed group, 00:31:33.400 --> 00:31:35.830 align:middle line:84% they are much less-- they are revealing 00:31:35.830 --> 00:31:38.370 align:middle line:84% less ambitious choices if those choices 00:31:38.370 --> 00:31:40.870 align:middle line:84% are going to be public than if they are going to be private. 00:31:40.870 --> 00:31:44.570 align:middle line:90% 00:31:44.570 --> 00:31:47.840 align:middle line:84% So they relate that to the thing that if you 00:31:47.840 --> 00:31:51.450 align:middle line:84% don't want to appear to be an overambitious woman if you're 00:31:51.450 --> 00:31:51.950 align:middle line:90% single. 00:31:51.950 --> 00:31:54.650 align:middle line:84% That makes you less marriageable, 00:31:54.650 --> 00:31:56.910 align:middle line:84% less attractive on the marriage market. 00:31:56.910 --> 00:32:02.090 align:middle line:84% And presumably, the men around you in your MBA program 00:32:02.090 --> 00:32:05.220 align:middle line:90% are potentially mates. 00:32:05.220 --> 00:32:10.760 align:middle line:84% So that's to show that this phenomenon of maybe 00:32:10.760 --> 00:32:17.720 align:middle line:84% women anticipating the desire of their husbands 00:32:17.720 --> 00:32:20.180 align:middle line:84% or potential husbands that they shouldn't work 00:32:20.180 --> 00:32:24.500 align:middle line:84% or they should not be ambitious in their work is pervasive. 00:32:24.500 --> 00:32:26.040 align:middle line:90% So this last fact-- 00:32:26.040 --> 00:32:29.270 align:middle line:84% it's consistent with an efficient household 00:32:29.270 --> 00:32:31.280 align:middle line:90% model with limited commitment. 00:32:31.280 --> 00:32:35.660 align:middle line:84% That is, one where the person who has-- 00:32:35.660 --> 00:32:39.080 align:middle line:84% people are bargaining efficiency based on how much money they 00:32:39.080 --> 00:32:42.170 align:middle line:90% bring to the household. 00:32:42.170 --> 00:32:44.240 align:middle line:84% But then if, for some reason, you 00:32:44.240 --> 00:32:47.780 align:middle line:84% have an exogenous shock, and suddenly your bargaining power 00:32:47.780 --> 00:32:51.850 align:middle line:84% increases, you channel your share to-- 00:32:51.850 --> 00:32:55.220 align:middle line:84% your outside option increases, then the bargaining power 00:32:55.220 --> 00:32:56.360 align:middle line:90% is renegotiated. 00:32:56.360 --> 00:32:58.540 align:middle line:84% So in a fully efficient household model, 00:32:58.540 --> 00:33:00.290 align:middle line:84% there would be no renegotiation every time 00:33:00.290 --> 00:33:01.880 align:middle line:90% someone gets more money. 00:33:01.880 --> 00:33:08.040 align:middle line:84% But in a model with limited commitment-- which by the way, 00:33:08.040 --> 00:33:10.227 align:middle line:84% has become the workhorse model of insurance, 00:33:10.227 --> 00:33:12.560 align:middle line:84% both in developing countries and in developed countries, 00:33:12.560 --> 00:33:15.560 align:middle line:84% where bargaining power is affected by people's outside 00:33:15.560 --> 00:33:16.520 align:middle line:90% option-- 00:33:16.520 --> 00:33:20.450 align:middle line:84% means that a woman who works more, who earns more, 00:33:20.450 --> 00:33:23.630 align:middle line:84% has more bargaining power in the household. 00:33:23.630 --> 00:33:29.310 align:middle line:84% So she cannot-- this young woman finishing their MBA cannot say, 00:33:29.310 --> 00:33:32.630 align:middle line:84% OK, I'm committing to not bargain hard. 00:33:32.630 --> 00:33:34.100 align:middle line:84% I'm committing to a low bargaining 00:33:34.100 --> 00:33:35.000 align:middle line:90% power in a household. 00:33:35.000 --> 00:33:37.023 align:middle line:90% But let me earn more money. 00:33:37.023 --> 00:33:37.940 align:middle line:90% It's going to be good. 00:33:37.940 --> 00:33:40.970 align:middle line:84% It's going to be more money for you to spend, after all, right? 00:33:40.970 --> 00:33:43.280 align:middle line:84% They can't do that because once they have the money, 00:33:43.280 --> 00:33:45.170 align:middle line:84% they'll be in a position to say, fuck you. 00:33:45.170 --> 00:33:48.350 align:middle line:84% If we are not bargaining the way that I would want to bargain, 00:33:48.350 --> 00:33:50.460 align:middle line:90% I'm just leaving this household. 00:33:50.460 --> 00:33:52.630 align:middle line:90% So that's consistent with that. 00:33:52.630 --> 00:33:55.110 align:middle line:84% And then maybe that's why they don't 00:33:55.110 --> 00:33:59.370 align:middle line:84% want to display that they have this kind of ambition. 00:33:59.370 --> 00:34:03.660 align:middle line:84% It is also consistent with more general social norms 00:34:03.660 --> 00:34:08.280 align:middle line:84% that women shouldn't be working or that, as we'll see more 00:34:08.280 --> 00:34:13.630 align:middle line:84% in a minute, they should not earn more than their husbands. 00:34:13.630 --> 00:34:17.520 align:middle line:84% So let's spend some time talking about these social norms 00:34:17.520 --> 00:34:19.650 align:middle line:90% about women working. 00:34:19.650 --> 00:34:23.880 align:middle line:84% Even before the social norms, there is a legal framework. 00:34:23.880 --> 00:34:27.510 align:middle line:84% And often, we tend to forget as economists, we are [INAUDIBLE].. 00:34:27.510 --> 00:34:30.000 align:middle line:84% We consider that the legal framework is irrelevant. 00:34:30.000 --> 00:34:32.159 align:middle line:84% Any negotiation can happen anyways. 00:34:32.159 --> 00:34:34.235 align:middle line:84% But in reality, the legal framework-- 00:34:34.235 --> 00:34:36.360 align:middle line:84% what is legal, what is not, what are their rights-- 00:34:36.360 --> 00:34:40.710 align:middle line:84% do determine the context in which these negotiations are 00:34:40.710 --> 00:34:41.969 align:middle line:90% happening. 00:34:41.969 --> 00:34:45.150 align:middle line:84% And Penny Goldberg, when she was chief economist at the bank, 00:34:45.150 --> 00:34:47.730 align:middle line:84% led a project where they put together 00:34:47.730 --> 00:34:50.055 align:middle line:84% a new exciting database on gender rights. 00:34:50.055 --> 00:34:52.650 align:middle line:90% 00:34:52.650 --> 00:34:56.610 align:middle line:84% They have rights in labor market, 00:34:56.610 --> 00:35:00.550 align:middle line:84% in property, anti-discrimination. 00:35:00.550 --> 00:35:03.620 align:middle line:90% And that database is public. 00:35:03.620 --> 00:35:05.680 align:middle line:84% You can use it to do whatever you want. 00:35:05.680 --> 00:35:07.060 align:middle line:90% And it's updated every year. 00:35:07.060 --> 00:35:09.220 align:middle line:84% And they have a short paper in AER Insights 00:35:09.220 --> 00:35:13.480 align:middle line:84% where they are just putting together some descriptive data 00:35:13.480 --> 00:35:15.520 align:middle line:90% facts about this database. 00:35:15.520 --> 00:35:17.950 align:middle line:84% And just a few things-- first of all, the average 00:35:17.950 --> 00:35:19.360 align:middle line:90% is 75 out of 100. 00:35:19.360 --> 00:35:23.950 align:middle line:84% That is, for every four rights that the husband had, that men 00:35:23.950 --> 00:35:25.750 align:middle line:90% had, a woman has three. 00:35:25.750 --> 00:35:29.680 align:middle line:84% It's particularly low in Middle Eastern, North Africa, 00:35:29.680 --> 00:35:30.400 align:middle line:90% and South Asia. 00:35:30.400 --> 00:35:32.440 align:middle line:90% It's not going to surprise you. 00:35:32.440 --> 00:35:37.180 align:middle line:84% 40% in MENA-- so women have 40% of the right of men. 00:35:37.180 --> 00:35:39.140 align:middle line:90% And 60% in South Asia. 00:35:39.140 --> 00:35:42.070 align:middle line:84% So it's not just about social norms in India, in Bangladesh. 00:35:42.070 --> 00:35:43.670 align:middle line:90% It's also the law. 00:35:43.670 --> 00:35:45.800 align:middle line:84% And they also show a strong correlation 00:35:45.800 --> 00:35:49.160 align:middle line:84% between legal rights and female labor force participation 00:35:49.160 --> 00:35:50.970 align:middle line:90% and the gender gap. 00:35:50.970 --> 00:35:53.380 align:middle line:84% So it is not true that the laws on the [INAUDIBLE] 00:35:53.380 --> 00:35:54.350 align:middle line:90% are irrelevant. 00:35:54.350 --> 00:35:56.270 align:middle line:84% Except perhaps you could say, well, this 00:35:56.270 --> 00:35:57.740 align:middle line:90% is just all correlation. 00:35:57.740 --> 00:36:01.190 align:middle line:84% So in places where they are not interested in giving women 00:36:01.190 --> 00:36:03.690 align:middle line:84% access to the labor force anyway, 00:36:03.690 --> 00:36:07.050 align:middle line:84% it's also the places where the laws are less generous. 00:36:07.050 --> 00:36:09.010 align:middle line:84% So at this point, it's just a correlation. 00:36:09.010 --> 00:36:09.760 align:middle line:90% STUDENT: Question? 00:36:09.760 --> 00:36:10.552 align:middle line:90% ESTHER DUFLO: Yeah. 00:36:10.552 --> 00:36:17.156 align:middle line:84% STUDENT: How should we think about how legal [INAUDIBLE] 00:36:17.156 --> 00:36:21.600 align:middle line:84% say 40% [INAUDIBLE] relate with underlying economic conditions? 00:36:21.600 --> 00:36:24.900 align:middle line:84% So in Saudi Arabia, that number probably 00:36:24.900 --> 00:36:26.910 align:middle line:84% has gone up a lot recently because they 00:36:26.910 --> 00:36:28.218 align:middle line:90% have labor mandates. 00:36:28.218 --> 00:36:29.010 align:middle line:90% ESTHER DUFLO: Yeah. 00:36:29.010 --> 00:36:30.710 align:middle line:84% STUDENT: So I wonder how we should think about-- 00:36:30.710 --> 00:36:31.752 align:middle line:90% ESTHER DUFLO: Yeah, yeah. 00:36:31.752 --> 00:36:33.320 align:middle line:84% So all of this is super [INAUDIBLE].. 00:36:33.320 --> 00:36:33.820 align:middle line:90% Yeah. 00:36:33.820 --> 00:36:34.550 align:middle line:90% Yeah, absolutely. 00:36:34.550 --> 00:36:35.790 align:middle line:84% All of this should be on the [INAUDIBLE].. 00:36:35.790 --> 00:36:37.880 align:middle line:84% In fact, one of the things that they also showed, 00:36:37.880 --> 00:36:39.260 align:middle line:90% which I have not put it-- 00:36:39.260 --> 00:36:40.830 align:middle line:90% it's gotten better. 00:36:40.830 --> 00:36:44.060 align:middle line:84% So not just in Saudi Arabia, not just in MENA, 00:36:44.060 --> 00:36:47.030 align:middle line:84% but all over the world, they are basically 00:36:47.030 --> 00:36:49.070 align:middle line:84% documenting the reforms that have happened 00:36:49.070 --> 00:36:50.820 align:middle line:90% over the last several decades. 00:36:50.820 --> 00:36:52.352 align:middle line:84% And it has gotten better everywhere. 00:36:52.352 --> 00:36:53.060 align:middle line:90% And you're right. 00:36:53.060 --> 00:36:55.280 align:middle line:84% The fact that it gets better might 00:36:55.280 --> 00:36:58.550 align:middle line:84% be a product, in some extents, of labor market tightness, 00:36:58.550 --> 00:37:00.380 align:middle line:90% like in Saudi Arabia. 00:37:00.380 --> 00:37:02.787 align:middle line:84% That's not something that, to my knowledge, 00:37:02.787 --> 00:37:03.870 align:middle line:90% they discuss in the paper. 00:37:03.870 --> 00:37:05.975 align:middle line:84% But it's something someone could do. 00:37:05.975 --> 00:37:09.600 align:middle line:84% STUDENT: Do we expect labor market tightness 00:37:09.600 --> 00:37:11.940 align:middle line:84% to necessarily reduce gender wage gaps 00:37:11.940 --> 00:37:13.962 align:middle line:84% or we might expect maybe it helps 00:37:13.962 --> 00:37:16.217 align:middle line:90% the men more, but [INAUDIBLE]? 00:37:16.217 --> 00:37:18.050 align:middle line:84% ESTHER DUFLO: I don't think there would be-- 00:37:18.050 --> 00:37:20.300 align:middle line:90% I don't think there is-- 00:37:20.300 --> 00:37:23.150 align:middle line:84% I think it's going to depend a bit on the model. 00:37:23.150 --> 00:37:24.110 align:middle line:90% STUDENT: OK. 00:37:24.110 --> 00:37:26.540 align:middle line:90% ESTHER DUFLO: In general, yes. 00:37:26.540 --> 00:37:28.760 align:middle line:84% You would expect-- I assume you would expect that. 00:37:28.760 --> 00:37:31.100 align:middle line:84% But it depends on whether there are also 00:37:31.100 --> 00:37:34.730 align:middle line:84% restrictions on which market women can-- 00:37:34.730 --> 00:37:36.380 align:middle line:84% it's related to this misallocation 00:37:36.380 --> 00:37:39.230 align:middle line:84% point that is made by the Hsieh and others paper. 00:37:39.230 --> 00:37:42.920 align:middle line:84% Suppose that women can only work in textiles. 00:37:42.920 --> 00:37:47.480 align:middle line:84% Then labor market tightness might not 00:37:47.480 --> 00:37:54.790 align:middle line:84% do much to the gender gap because what you're 00:37:54.790 --> 00:37:58.630 align:middle line:84% going to see is to have more women employed in textile, 00:37:58.630 --> 00:38:03.400 align:middle line:84% and then all the men that are freed from textile moving 00:38:03.400 --> 00:38:09.660 align:middle line:84% to the other sectors and getting increased wages there. 00:38:09.660 --> 00:38:12.630 align:middle line:84% It's also related to something that I'm not 00:38:12.630 --> 00:38:16.140 align:middle line:84% going to talk about today, but that there 00:38:16.140 --> 00:38:19.800 align:middle line:84% is starting to be some work on, and I 00:38:19.800 --> 00:38:21.540 align:middle line:84% think you might do more with Ben, 00:38:21.540 --> 00:38:24.350 align:middle line:84% is monopsonies in the labor market. 00:38:24.350 --> 00:38:30.430 align:middle line:84% So both women and men might be subject 00:38:30.430 --> 00:38:34.840 align:middle line:84% to noncompetitive or imperfect competition in the labor 00:38:34.840 --> 00:38:36.560 align:middle line:90% market. 00:38:36.560 --> 00:38:40.100 align:middle line:84% Because it's hard to find another job once you have one. 00:38:40.100 --> 00:38:45.500 align:middle line:84% And this imperfect competition might be stronger for women 00:38:45.500 --> 00:38:48.050 align:middle line:84% because they are more restricted in the type of job 00:38:48.050 --> 00:38:50.510 align:middle line:90% that they can access. 00:38:50.510 --> 00:38:52.760 align:middle line:84% For example, they cannot move because they cannot move 00:38:52.760 --> 00:38:53.630 align:middle line:90% their family. 00:38:53.630 --> 00:38:58.250 align:middle line:84% They cannot travel far because they cannot go on buses. 00:38:58.250 --> 00:39:01.410 align:middle line:84% They have limitation on the sectors in which they work. 00:39:01.410 --> 00:39:04.190 align:middle line:84% So in this kind of a model version, 00:39:04.190 --> 00:39:11.030 align:middle line:84% you might expect that an increase in the labor market 00:39:11.030 --> 00:39:12.590 align:middle line:84% tightness in general is not going 00:39:12.590 --> 00:39:14.690 align:middle line:84% to lead to much of an improvement for the women's 00:39:14.690 --> 00:39:15.210 align:middle line:90% situation. 00:39:15.210 --> 00:39:17.300 align:middle line:84% So I think it's going to depend on the model 00:39:17.300 --> 00:39:20.680 align:middle line:90% and on the setting, obviously. 00:39:20.680 --> 00:39:23.860 align:middle line:84% So let's take a sidestep to think 00:39:23.860 --> 00:39:27.050 align:middle line:84% about culture and social norms, and in particular, culture 00:39:27.050 --> 00:39:29.780 align:middle line:84% and social norms in the labor market. 00:39:29.780 --> 00:39:34.420 align:middle line:84% So Nathan Nunn at Harvard is someone 00:39:34.420 --> 00:39:39.340 align:middle line:84% who has done a lot to bring back the study of culture 00:39:39.340 --> 00:39:42.160 align:middle line:84% and social norms as a legitimate object of study 00:39:42.160 --> 00:39:46.150 align:middle line:84% in economics in saying that it still has a lot of impact 00:39:46.150 --> 00:39:48.670 align:middle line:84% on the way that people live their lives, 00:39:48.670 --> 00:39:51.990 align:middle line:84% and academic outcomes, et cetera. 00:39:51.990 --> 00:39:59.370 align:middle line:84% One study is-- the paper is from a few years ago, 00:39:59.370 --> 00:40:04.860 align:middle line:84% but it's one of the great examples of his work. 00:40:04.860 --> 00:40:07.650 align:middle line:84% This is joined, by the way, with Alberto Alesina and Paula 00:40:07.650 --> 00:40:08.880 align:middle line:90% Giuliano. 00:40:08.880 --> 00:40:13.620 align:middle line:84% What they do is to test an old hypothesis of a Esther 00:40:13.620 --> 00:40:14.400 align:middle line:90% Boserup-- 00:40:14.400 --> 00:40:18.570 align:middle line:84% or Boserup, I guess-- which is in regions where the plow was 00:40:18.570 --> 00:40:21.180 align:middle line:84% dominant, males were more involved 00:40:21.180 --> 00:40:22.410 align:middle line:90% in working at the fields. 00:40:22.410 --> 00:40:30.460 align:middle line:90% And women were less valued. 00:40:30.460 --> 00:40:34.270 align:middle line:84% So for example, if you look at India today 00:40:34.270 --> 00:40:37.780 align:middle line:84% and you look at North India versus South India, 00:40:37.780 --> 00:40:40.300 align:middle line:84% a cliche which is correct is that women are better 00:40:40.300 --> 00:40:42.430 align:middle line:90% treated in the South of India. 00:40:42.430 --> 00:40:44.230 align:middle line:84% And the South of India is a culture 00:40:44.230 --> 00:40:48.550 align:middle line:84% of rice, where traditionally, it's done by hand. 00:40:48.550 --> 00:40:50.460 align:middle line:84% In fact, today, it's still done by hand. 00:40:50.460 --> 00:40:51.710 align:middle line:90% It's always been done by hand. 00:40:51.710 --> 00:40:55.430 align:middle line:84% It's labor intensive, but not in a particularly-- 00:40:55.430 --> 00:40:57.190 align:middle line:84% you don't need to be super strong. 00:40:57.190 --> 00:40:59.980 align:middle line:84% You need to be patient and just get it done. 00:40:59.980 --> 00:41:02.510 align:middle line:84% Whereas the North of India traditionally was not a rice 00:41:02.510 --> 00:41:03.010 align:middle line:90% place. 00:41:03.010 --> 00:41:04.870 align:middle line:84% Also now they aren't doing rice there. 00:41:04.870 --> 00:41:07.780 align:middle line:84% But it's not a place which is suitable to rice originally. 00:41:07.780 --> 00:41:12.490 align:middle line:84% So it was more of a plow culture and a wheat culture, 00:41:12.490 --> 00:41:15.460 align:middle line:90% where the labor supply-- 00:41:15.460 --> 00:41:18.940 align:middle line:84% where mostly men were involved, traditionally, in agriculture. 00:41:18.940 --> 00:41:22.540 align:middle line:84% And in fact, you still see this correlation today, 00:41:22.540 --> 00:41:26.393 align:middle line:84% where any dimension of women empowerment 00:41:26.393 --> 00:41:28.060 align:middle line:84% is worse in the North than in the South. 00:41:28.060 --> 00:41:32.170 align:middle line:84% So Esther Boserup's hypothesis is that that's a general fact. 00:41:32.170 --> 00:41:35.410 align:middle line:84% So what they do here to test the idea 00:41:35.410 --> 00:41:39.280 align:middle line:84% is that they use an FAO database on crop suitability 00:41:39.280 --> 00:41:42.580 align:middle line:84% to build an index of where the plow is more likely to be 00:41:42.580 --> 00:41:43.960 align:middle line:90% used in ancient times. 00:41:43.960 --> 00:41:46.420 align:middle line:84% And then they correlate it with today's social norm. 00:41:46.420 --> 00:41:48.880 align:middle line:84% And then they put a fixed effect and some such 00:41:48.880 --> 00:41:51.170 align:middle line:90% to try and test it. 00:41:51.170 --> 00:41:54.490 align:middle line:84% So I'll just show you one graph, which makes it clear. 00:41:54.490 --> 00:42:00.100 align:middle line:84% Traditional plow use and female labor force participation. 00:42:00.100 --> 00:42:01.230 align:middle line:90% So this is not real use. 00:42:01.230 --> 00:42:05.400 align:middle line:84% This is a expected use based on the crop 00:42:05.400 --> 00:42:07.170 align:middle line:90% suitability for the crop mixed. 00:42:07.170 --> 00:42:10.350 align:middle line:84% And you can see that the labor force participation 00:42:10.350 --> 00:42:13.560 align:middle line:84% is higher in places that were less likely to use the plow. 00:42:13.560 --> 00:42:18.370 align:middle line:84% So that's a cross-country interpretation. 00:42:18.370 --> 00:42:23.950 align:middle line:84% I find the paper quite convincing in terms of the very 00:42:23.950 --> 00:42:25.450 align:middle line:90% many tests they do, et cetera. 00:42:25.450 --> 00:42:28.150 align:middle line:84% It's cross-country so it's never going to be completely, 00:42:28.150 --> 00:42:29.170 align:middle line:90% completely watertight. 00:42:29.170 --> 00:42:30.878 align:middle line:84% On the other hand, it's the entire world. 00:42:30.878 --> 00:42:34.280 align:middle line:84% And so there is some benefits to that as well. 00:42:34.280 --> 00:42:40.120 align:middle line:84% So I strongly suggest you have a look at this paper. 00:42:40.120 --> 00:42:44.900 align:middle line:84% Now one more problem with the social norms is that-- 00:42:44.900 --> 00:42:49.330 align:middle line:84% and one reason why they might be very persistent 00:42:49.330 --> 00:42:52.330 align:middle line:84% is that even when they change, you might not 00:42:52.330 --> 00:42:53.710 align:middle line:90% know that they change. 00:42:53.710 --> 00:42:57.400 align:middle line:84% In particular, if, as we've seen in our first paper 00:42:57.400 --> 00:43:01.210 align:middle line:84% we discussed today, the discussions are-- 00:43:01.210 --> 00:43:06.370 align:middle line:84% you're not very good at discussing things with people. 00:43:06.370 --> 00:43:09.340 align:middle line:84% So if you find it difficult to learn from others 00:43:09.340 --> 00:43:13.060 align:middle line:84% about how many blue urns there is in a jar, 00:43:13.060 --> 00:43:14.710 align:middle line:84% you might find it even more difficult 00:43:14.710 --> 00:43:17.980 align:middle line:84% to discuss whether it's OK for your wife to work, 00:43:17.980 --> 00:43:20.200 align:middle line:84% or it's not OK for your wife to work, 00:43:20.200 --> 00:43:24.750 align:middle line:84% which might lead to a misperception of social norm. 00:43:24.750 --> 00:43:29.500 align:middle line:84% And hence, the second or Bursztyn trick, 00:43:29.500 --> 00:43:33.330 align:middle line:84% which is basically-- which he has performed in several 00:43:33.330 --> 00:43:33.880 align:middle line:90% papers-- 00:43:33.880 --> 00:43:38.130 align:middle line:84% which is trying to see if there is 00:43:38.130 --> 00:43:40.110 align:middle line:84% a bias in the perception of social norms, 00:43:40.110 --> 00:43:43.110 align:middle line:84% and if there is one, to debias people. 00:43:43.110 --> 00:43:46.440 align:middle line:84% So this particular experiment, which he presented in this very 00:43:46.440 --> 00:43:49.650 align:middle line:84% room, I remember-- so he was there presenting it from this 00:43:49.650 --> 00:43:51.240 align:middle line:90% space from where I talk-- 00:43:51.240 --> 00:43:54.780 align:middle line:84% was with 500 young men in Saudi Arabia. 00:43:54.780 --> 00:43:59.490 align:middle line:84% And they put them in a room, or by groups of about 30. 00:43:59.490 --> 00:44:02.460 align:middle line:84% And then they asked them first their opinion 00:44:02.460 --> 00:44:04.110 align:middle line:84% on whether women should be allowed 00:44:04.110 --> 00:44:06.130 align:middle line:90% to work outside the home. 00:44:06.130 --> 00:44:08.470 align:middle line:84% And surprisingly in this younger group, 00:44:08.470 --> 00:44:10.360 align:middle line:84% there is actually a fair number of people 00:44:10.360 --> 00:44:11.530 align:middle line:90% who think they would be OK. 00:44:11.530 --> 00:44:16.160 align:middle line:84% 87% of men agree with the statement, in my opinion, 00:44:16.160 --> 00:44:18.940 align:middle line:84% women should be allowed to work outside the home. 00:44:18.940 --> 00:44:24.250 align:middle line:84% But then they asked, consider these other men 00:44:24.250 --> 00:44:27.110 align:middle line:90% in this home who are with you. 00:44:27.110 --> 00:44:29.680 align:middle line:84% How many do you think will answer yes 00:44:29.680 --> 00:44:31.580 align:middle line:90% to that same question? 00:44:31.580 --> 00:44:37.600 align:middle line:84% And there 3/4 underestimate the numbers. 00:44:37.600 --> 00:44:41.530 align:middle line:84% So most of them say a number that 00:44:41.530 --> 00:44:44.920 align:middle line:84% is lower than 80%, that's less than 80%. 00:44:44.920 --> 00:44:47.150 align:middle line:84% Most of them say a lower number-- sometimes, 00:44:47.150 --> 00:44:49.200 align:middle line:90% much lower number. 00:44:49.200 --> 00:44:51.230 align:middle line:84% So what they do in the experiment is that 00:44:51.230 --> 00:44:53.570 align:middle line:84% then for half of these men after eliciting 00:44:53.570 --> 00:44:56.390 align:middle line:84% their own norm, their own view, as well 00:44:56.390 --> 00:44:58.880 align:middle line:84% as their view of other people's view, 00:44:58.880 --> 00:45:02.840 align:middle line:84% they give them the right number about other people's view. 00:45:02.840 --> 00:45:04.790 align:middle line:84% Obviously, the answer is not yes. 00:45:04.790 --> 00:45:07.580 align:middle line:90% It is OK, or it's not OK. 00:45:07.580 --> 00:45:09.170 align:middle line:90% But the answer is-- 00:45:09.170 --> 00:45:13.070 align:middle line:84% by the way, 87% of people we interviewed 00:45:13.070 --> 00:45:16.570 align:middle line:90% think that it's fine. 00:45:16.570 --> 00:45:21.360 align:middle line:84% And then they are given the choice between a $5 Amazon 00:45:21.360 --> 00:45:24.390 align:middle line:84% certificate and the opportunity to sign up 00:45:24.390 --> 00:45:26.940 align:middle line:84% their wife for a platform on the job, 00:45:26.940 --> 00:45:31.110 align:middle line:84% and then they give them follow-up calls 00:45:31.110 --> 00:45:34.570 align:middle line:84% to try to look at longer-term outcomes. 00:45:34.570 --> 00:45:41.700 align:middle line:84% And then after that, actually, they do even better than that. 00:45:41.700 --> 00:45:45.030 align:middle line:84% They do another experiment where they are doing the same thing, 00:45:45.030 --> 00:45:47.660 align:middle line:84% but in the context where people are-- 00:45:47.660 --> 00:45:51.780 align:middle line:84% women are given a job, a real job. 00:45:51.780 --> 00:45:54.600 align:middle line:84% Many women are-- they have a large sample of people 00:45:54.600 --> 00:45:57.070 align:middle line:90% who are given a real job. 00:45:57.070 --> 00:45:59.315 align:middle line:84% And this is marketing, phone marketing. 00:45:59.315 --> 00:46:00.940 align:middle line:84% And then they are given the opportunity 00:46:00.940 --> 00:46:06.640 align:middle line:84% to do it either from home or from office. 00:46:06.640 --> 00:46:09.930 align:middle line:90% And the office job pays more. 00:46:09.930 --> 00:46:13.728 align:middle line:84% And the question is, how many people take the office job? 00:46:13.728 --> 00:46:15.270 align:middle line:84% So that's the second experiment which 00:46:15.270 --> 00:46:18.690 align:middle line:84% I'm not going to show you today, but has exactly the same type 00:46:18.690 --> 00:46:25.290 align:middle line:84% of results and is a stronger, more heuristic thing. 00:46:25.290 --> 00:46:26.910 align:middle line:90% Here this is a little bit-- 00:46:26.910 --> 00:46:28.350 align:middle line:84% this is just signing up for a job. 00:46:28.350 --> 00:46:30.180 align:middle line:90% That's not so much. 00:46:30.180 --> 00:46:33.000 align:middle line:84% But for what they find here, on average, 00:46:33.000 --> 00:46:39.060 align:middle line:84% the people who are debiased are more likely to sign their wife 00:46:39.060 --> 00:46:40.920 align:middle line:90% up for the job platform-- 00:46:40.920 --> 00:46:43.230 align:middle line:90% 32% versus 23%. 00:46:43.230 --> 00:46:47.560 align:middle line:84% And of course, where it becomes more interesting-- 00:46:47.560 --> 00:46:49.180 align:middle line:90% so then we can look-- 00:46:49.180 --> 00:46:49.680 align:middle line:90% sorry. 00:46:49.680 --> 00:46:54.100 align:middle line:84% We can look at whether they, in fact, applied for the job. 00:46:54.100 --> 00:46:55.750 align:middle line:90% And you find a difference. 00:46:55.750 --> 00:46:58.210 align:middle line:84% You can look at whether they are interviewed for the job. 00:46:58.210 --> 00:46:59.840 align:middle line:90% And you still find a difference. 00:46:59.840 --> 00:47:01.570 align:middle line:84% And then you can look at whether they are 00:47:01.570 --> 00:47:04.408 align:middle line:90% employed or driving lessons. 00:47:04.408 --> 00:47:05.950 align:middle line:84% And there you don't find a difference 00:47:05.950 --> 00:47:09.160 align:middle line:90% in these lower outcomes. 00:47:09.160 --> 00:47:11.380 align:middle line:90% So this is why-- 00:47:11.380 --> 00:47:13.330 align:middle line:84% actually, I was editing the paper. 00:47:13.330 --> 00:47:15.480 align:middle line:90% And I got this-- 00:47:15.480 --> 00:47:17.160 align:middle line:84% I was editing the paper at the AER, 00:47:17.160 --> 00:47:19.680 align:middle line:90% and this is what they sent. 00:47:19.680 --> 00:47:21.015 align:middle line:90% And I was like, I don't know. 00:47:21.015 --> 00:47:24.870 align:middle line:84% At the end of the day, I don't know if this is-- 00:47:24.870 --> 00:47:27.890 align:middle line:84% the proof is the pudding is whether they are employed. 00:47:27.890 --> 00:47:30.170 align:middle line:84% And maybe there is an effect, but maybe not. 00:47:30.170 --> 00:47:32.990 align:middle line:90% The p value of 0.24. 00:47:32.990 --> 00:47:34.340 align:middle line:90% I don't know what to think. 00:47:34.340 --> 00:47:36.010 align:middle line:84% At the end of the day, does it matter? 00:47:36.010 --> 00:47:37.010 align:middle line:90% Or does it not matter? 00:47:37.010 --> 00:47:40.370 align:middle line:84% I need a real outcome, which is why they went and did 00:47:40.370 --> 00:47:47.177 align:middle line:84% this second experiment, where basically, they were thinking 00:47:47.177 --> 00:47:49.260 align:middle line:84% the problem with this one is that eventually, it's 00:47:49.260 --> 00:47:50.100 align:middle line:90% not very powered. 00:47:50.100 --> 00:47:53.185 align:middle line:84% Because there are so many steps between signing up 00:47:53.185 --> 00:47:55.560 align:middle line:84% for a job platform and actually getting the job that they 00:47:55.560 --> 00:47:57.570 align:middle line:90% were never going to be powered. 00:47:57.570 --> 00:48:00.550 align:middle line:84% So my objection here as an editor-- 00:48:00.550 --> 00:48:04.050 align:middle line:84% this is a point about how you want 00:48:04.050 --> 00:48:07.020 align:middle line:84% to think about designing your research project as well-- 00:48:07.020 --> 00:48:08.940 align:middle line:84% my objection here as an editor was 00:48:08.940 --> 00:48:11.310 align:middle line:84% that this particular project was not powered 00:48:11.310 --> 00:48:13.410 align:middle line:90% to find employment effect. 00:48:13.410 --> 00:48:16.260 align:middle line:84% Because given the number of things that have been given 00:48:16.260 --> 00:48:19.380 align:middle line:84% the few people who are employed-- about 7%-- 00:48:19.380 --> 00:48:22.080 align:middle line:84% it would have taken a much larger sample 00:48:22.080 --> 00:48:28.100 align:middle line:84% to be able to detect any effect that is not ginormous. 00:48:28.100 --> 00:48:31.090 align:middle line:84% So I was like, it's fine to find 0. 00:48:31.090 --> 00:48:32.710 align:middle line:90% I have no issue with 0. 00:48:32.710 --> 00:48:38.740 align:middle line:84% But your paper is not powered to find an impact on the outcome 00:48:38.740 --> 00:48:40.390 align:middle line:84% that you want to find an impact on. 00:48:40.390 --> 00:48:42.730 align:middle line:84% It's only powered to find an impact 00:48:42.730 --> 00:48:45.610 align:middle line:84% on these intermediate outcomes where we have it. 00:48:45.610 --> 00:48:46.600 align:middle line:90% So it's promising. 00:48:46.600 --> 00:48:49.697 align:middle line:90% But it's not enough for the AER. 00:48:49.697 --> 00:48:51.530 align:middle line:84% So at this point, they could have said fine. 00:48:51.530 --> 00:48:53.572 align:middle line:84% We're just going to go one level down and publish 00:48:53.572 --> 00:48:55.600 align:middle line:90% without intermediate outcome. 00:48:55.600 --> 00:48:57.160 align:middle line:84% Or they could have done what they 00:48:57.160 --> 00:48:59.290 align:middle line:84% did, which was the right move, in my opinion, which 00:48:59.290 --> 00:49:01.480 align:middle line:84% is to design a better power outcome. 00:49:01.480 --> 00:49:03.490 align:middle line:84% And to find a better power outcome, 00:49:03.490 --> 00:49:06.190 align:middle line:84% basically, that's why they did the second experiment 00:49:06.190 --> 00:49:09.670 align:middle line:84% by setting themselves in a context where already, people 00:49:09.670 --> 00:49:11.760 align:middle line:90% were working. 00:49:11.760 --> 00:49:15.570 align:middle line:84% And the decision was from home versus outside the home 00:49:15.570 --> 00:49:16.690 align:middle line:90% for more money. 00:49:16.690 --> 00:49:18.360 align:middle line:90% So that's super powerful. 00:49:18.360 --> 00:49:20.170 align:middle line:90% And if they had found-- 00:49:20.170 --> 00:49:24.930 align:middle line:84% in fact, in my return letter, I said, compose me your design 00:49:24.930 --> 00:49:25.740 align:middle line:90% before you run it. 00:49:25.740 --> 00:49:27.990 align:middle line:84% I'm going to tell you whether I'm 00:49:27.990 --> 00:49:32.310 align:middle line:84% going to sign up whether that's satisfactory, 00:49:32.310 --> 00:49:33.840 align:middle line:90% whatever you find. 00:49:33.840 --> 00:49:35.648 align:middle line:84% And so when he proposed the second design, 00:49:35.648 --> 00:49:36.690 align:middle line:90% I was like, that's great. 00:49:36.690 --> 00:49:39.630 align:middle line:84% Because if you find nothing, this is powerful. 00:49:39.630 --> 00:49:41.070 align:middle line:90% This is going to be a real 0. 00:49:41.070 --> 00:49:43.750 align:middle line:84% If you find an effect, then there's an effect. 00:49:43.750 --> 00:49:45.330 align:middle line:90% So go ahead and run it. 00:49:45.330 --> 00:49:47.700 align:middle line:84% And basically, I told him that I would publish no matter 00:49:47.700 --> 00:49:49.530 align:middle line:84% what, in a sense, no matter what he found 00:49:49.530 --> 00:49:53.400 align:middle line:84% in this second experiment because that was well powered 00:49:53.400 --> 00:49:56.070 align:middle line:90% for the outcome we cared about. 00:49:56.070 --> 00:49:59.630 align:middle line:84% But we can do a little better than that. 00:49:59.630 --> 00:50:02.690 align:middle line:84% Already more interesting than that is you have one more 00:50:02.690 --> 00:50:07.640 align:middle line:84% prediction on the sign-up, which is the sign-up should come from 00:50:07.640 --> 00:50:08.490 align:middle line:90% the fact-- 00:50:08.490 --> 00:50:11.690 align:middle line:84% who were originally biased about, 00:50:11.690 --> 00:50:14.630 align:middle line:84% underestimating the fraction of people. 00:50:14.630 --> 00:50:16.730 align:middle line:84% So some people were underestimating 00:50:16.730 --> 00:50:18.890 align:middle line:90% the fraction of people-- 00:50:18.890 --> 00:50:19.850 align:middle line:90% a lot of them. 00:50:19.850 --> 00:50:21.962 align:middle line:84% And some people were over-- but there are still 00:50:21.962 --> 00:50:24.170 align:middle line:84% some people who are estimating the fraction of people 00:50:24.170 --> 00:50:27.440 align:middle line:84% who found working for a job acceptable. 00:50:27.440 --> 00:50:30.920 align:middle line:84% And so the finding is that these guys, if anything, 00:50:30.920 --> 00:50:32.790 align:middle line:90% the control is slightly larger. 00:50:32.790 --> 00:50:38.540 align:middle line:84% So the opinion didn't really change. 00:50:38.540 --> 00:50:42.060 align:middle line:84% Or if anything, it changed in the direction you expect. 00:50:42.060 --> 00:50:44.360 align:middle line:84% And so the effect on sign-up comes 00:50:44.360 --> 00:50:50.240 align:middle line:84% from the people who were initially underestimating 00:50:50.240 --> 00:50:51.380 align:middle line:90% the social norm. 00:50:51.380 --> 00:50:54.950 align:middle line:84% So any experiments-- if you want to run this type of experiment 00:50:54.950 --> 00:50:56.840 align:middle line:84% where you're debiasing, it's really important 00:50:56.840 --> 00:51:00.050 align:middle line:84% to have the baseline value so that you can-- 00:51:00.050 --> 00:51:01.940 align:middle line:84% because the prediction is really not 00:51:01.940 --> 00:51:03.870 align:middle line:84% just about the effect of the treatment 00:51:03.870 --> 00:51:06.920 align:middle line:84% but how the effect of the treatment 00:51:06.920 --> 00:51:09.860 align:middle line:84% varies as a function of whether you're moving people up 00:51:09.860 --> 00:51:12.680 align:middle line:90% or down with your information. 00:51:12.680 --> 00:51:18.490 align:middle line:84% So that's important to keep in mind because this type of idea 00:51:18.490 --> 00:51:20.780 align:middle line:84% of I'm going to debias on something or the other-- 00:51:20.780 --> 00:51:23.530 align:middle line:84% it could be the social norm, the weather in California, 00:51:23.530 --> 00:51:28.030 align:middle line:84% or people's own ability in math-- 00:51:28.030 --> 00:51:28.720 align:middle line:90% I don't know. 00:51:28.720 --> 00:51:32.678 align:middle line:84% All of these things, you expect different results 00:51:32.678 --> 00:51:33.970 align:middle line:90% depending on what you're given. 00:51:33.970 --> 00:51:38.650 align:middle line:84% So you need to estimate the baseline. 00:51:38.650 --> 00:51:39.772 align:middle line:90% So that's the idea. 00:51:39.772 --> 00:51:41.230 align:middle line:84% So that's a social norm about women 00:51:41.230 --> 00:51:43.420 align:middle line:90% working outside the house. 00:51:43.420 --> 00:51:45.160 align:middle line:84% There are other social norms standing 00:51:45.160 --> 00:51:47.680 align:middle line:90% in the way of women success. 00:51:47.680 --> 00:51:52.900 align:middle line:84% So Marianne Bertrand, Jennifer Pan and Emir Kamenica 00:51:52.900 --> 00:51:56.110 align:middle line:84% have this paper looking at the norm 00:51:56.110 --> 00:51:59.080 align:middle line:84% that women must earn less than their husband. 00:51:59.080 --> 00:52:04.690 align:middle line:84% And what they show is that there is a discontinuity 00:52:04.690 --> 00:52:10.030 align:middle line:84% in the fraction of people that you're finding precisely 00:52:10.030 --> 00:52:12.550 align:middle line:84% at the place where women and men would be earning 00:52:12.550 --> 00:52:14.090 align:middle line:90% the same in the household. 00:52:14.090 --> 00:52:15.620 align:middle line:84% So basically, you have a big hole 00:52:15.620 --> 00:52:19.960 align:middle line:84% of women who earn just above their husband. 00:52:19.960 --> 00:52:24.550 align:middle line:84% And [INAUDIBLE] is trying to do a second 00:52:24.550 --> 00:52:27.970 align:middle line:84% your paper on this topic, trying to see 00:52:27.970 --> 00:52:31.830 align:middle line:84% whether this can be replicated in experimental work. 00:52:31.830 --> 00:52:34.820 align:middle line:84% So that's one, the social norm of women cannot earn more than 00:52:34.820 --> 00:52:36.650 align:middle line:84% their husband, which you find in the US. 00:52:36.650 --> 00:52:38.150 align:middle line:84% There is good reason to think you're 00:52:38.150 --> 00:52:42.070 align:middle line:84% finding it in other places that are even more conservative. 00:52:42.070 --> 00:52:44.980 align:middle line:90% But we'll find out maybe, soon. 00:52:44.980 --> 00:52:50.140 align:middle line:84% And then there is Claudia Goldin, 00:52:50.140 --> 00:52:52.480 align:middle line:84% who has, of course, made a career looking at women 00:52:52.480 --> 00:52:55.480 align:middle line:84% in the labor market and is now saying that, 00:52:55.480 --> 00:53:02.530 align:middle line:84% for example, in a recent work, that the norm 00:53:02.530 --> 00:53:08.230 align:middle line:84% of the rigid full-time, full week, many hours 00:53:08.230 --> 00:53:10.870 align:middle line:84% may be the last thing that's holding women now 00:53:10.870 --> 00:53:13.250 align:middle line:90% in the US labor market. 00:53:13.250 --> 00:53:15.760 align:middle line:84% So for example, when you look at lawyers-- 00:53:15.760 --> 00:53:17.350 align:middle line:90% work very, very long hours. 00:53:17.350 --> 00:53:19.810 align:middle line:84% And women eventually don't go into this profession, 00:53:19.810 --> 00:53:24.620 align:middle line:84% or they quit them early, or they choose mommy tracks situation. 00:53:24.620 --> 00:53:29.140 align:middle line:84% So here there are two social norms that in some sense 00:53:29.140 --> 00:53:35.020 align:middle line:84% could fight, could conspire to keep women, 00:53:35.020 --> 00:53:36.820 align:middle line:84% to hold women back in this situation. 00:53:36.820 --> 00:53:38.470 align:middle line:84% The one is that you're a good employee 00:53:38.470 --> 00:53:46.180 align:middle line:84% if you work all the time, if you work super, super long hours. 00:53:46.180 --> 00:53:49.150 align:middle line:84% And two is that if there is one person in the household that 00:53:49.150 --> 00:53:50.920 align:middle line:84% has to comply with that, it has to be 00:53:50.920 --> 00:53:54.940 align:middle line:84% the husband because the wife has to take care of the children. 00:53:54.940 --> 00:53:58.955 align:middle line:84% So that idea that women are multitasking 00:53:58.955 --> 00:54:00.330 align:middle line:84% is one I'm going to come back to. 00:54:00.330 --> 00:54:03.090 align:middle line:84% Because I think for developing countries, that's 00:54:03.090 --> 00:54:06.990 align:middle line:84% a pretty essential one, that sphere of women being 00:54:06.990 --> 00:54:12.000 align:middle line:84% in charge of children pretty much exclusively, 00:54:12.000 --> 00:54:15.780 align:middle line:84% as the residual claimant, is actually the one thing 00:54:15.780 --> 00:54:18.390 align:middle line:84% that you're refinding interacting 00:54:18.390 --> 00:54:24.940 align:middle line:84% with other parts of the labor market to hold women back. 00:54:24.940 --> 00:54:27.820 align:middle line:84% So we'll go back to that towards the end of this lecture. 00:54:27.820 --> 00:54:29.910 align:middle line:84% So one question then is that, OK, 00:54:29.910 --> 00:54:31.270 align:middle line:90% you have these social norms. 00:54:31.270 --> 00:54:34.140 align:middle line:90% They are very persistent. 00:54:34.140 --> 00:54:36.870 align:middle line:84% People might be, to some extent, misguided about them. 00:54:36.870 --> 00:54:40.530 align:middle line:84% But an interesting question is whether these norms 00:54:40.530 --> 00:54:43.500 align:middle line:90% can be changed by teaching. 00:54:43.500 --> 00:54:46.800 align:middle line:84% That's not a very economics-y way of thinking. 00:54:46.800 --> 00:54:50.220 align:middle line:84% We tend to follow the Baker [? Stigler ?] or [INAUDIBLE] 00:54:50.220 --> 00:54:55.030 align:middle line:84% known as [INAUDIBLE] idea is like, whatever preferences are, 00:54:55.030 --> 00:54:56.640 align:middle line:90% what they have, what they are. 00:54:56.640 --> 00:55:01.590 align:middle line:84% And indeed, you can tell people what the preferences of others 00:55:01.590 --> 00:55:03.610 align:middle line:90% are and inform them about that. 00:55:03.610 --> 00:55:09.240 align:middle line:84% But you can't really affect people's own view so easily. 00:55:09.240 --> 00:55:14.310 align:middle line:84% But so that's the traditional economics way of thinking. 00:55:14.310 --> 00:55:18.450 align:middle line:84% I already reminded you of the [INAUDIBLE] theorem 00:55:18.450 --> 00:55:20.010 align:middle line:84% that would think that laws doesn't 00:55:20.010 --> 00:55:22.530 align:middle line:84% matter because people bargain their way out of them anyways. 00:55:22.530 --> 00:55:25.950 align:middle line:84% And then that's the second one, second trope for the day is 00:55:25.950 --> 00:55:28.410 align:middle line:90% preferences cannot be changed. 00:55:28.410 --> 00:55:30.740 align:middle line:84% But we are learning that that's not the case. 00:55:30.740 --> 00:55:33.980 align:middle line:84% Because on many things, people's own sense 00:55:33.980 --> 00:55:37.970 align:middle line:84% of what their preferences are is super vague anyways. 00:55:37.970 --> 00:55:39.800 align:middle line:84% So they have these interesting experiments 00:55:39.800 --> 00:55:44.090 align:middle line:84% about trying to make people value things like mugs, 00:55:44.090 --> 00:55:47.870 align:middle line:84% or bottles of wines, and their judgment being influenced 00:55:47.870 --> 00:55:50.960 align:middle line:84% by the last digit of their Social Security number, 00:55:50.960 --> 00:55:52.670 align:middle line:84% if they just had to write it down. 00:55:52.670 --> 00:55:55.880 align:middle line:84% So people are first asked to write down their Social-- 00:55:55.880 --> 00:55:58.010 align:middle line:84% MBA students are first asked to write down 00:55:58.010 --> 00:56:01.700 align:middle line:84% the last two digits of their Social Security number. 00:56:01.700 --> 00:56:03.560 align:middle line:84% And then they are given a bottle of wine. 00:56:03.560 --> 00:56:05.602 align:middle line:84% And they are asked to bid for the bottle of wine. 00:56:05.602 --> 00:56:08.600 align:middle line:84% And people who have a higher Social Security number pay more 00:56:08.600 --> 00:56:11.230 align:middle line:84% for the bottle than people who don't. 00:56:11.230 --> 00:56:13.860 align:middle line:84% So that's an example of saying people have-- 00:56:13.860 --> 00:56:17.070 align:middle line:84% on many things, we just don't know. 00:56:17.070 --> 00:56:20.250 align:middle line:84% But we have to express an opinion, we express an opinion. 00:56:20.250 --> 00:56:23.490 align:middle line:84% And that might be true for things 00:56:23.490 --> 00:56:27.360 align:middle line:84% deeper than bottles of wine, but for things like values, 00:56:27.360 --> 00:56:28.655 align:middle line:90% and things like that. 00:56:28.655 --> 00:56:30.030 align:middle line:84% And if that's the case, then that 00:56:30.030 --> 00:56:33.810 align:middle line:84% means that perhaps norms are not a fatality. 00:56:33.810 --> 00:56:38.670 align:middle line:84% Perhaps it's possible to change social norms. 00:56:38.670 --> 00:56:40.740 align:middle line:84% And in fact, there is a literature 00:56:40.740 --> 00:56:50.250 align:middle line:84% on TV showing that TV entertainment can change 00:56:50.250 --> 00:56:53.595 align:middle line:84% people's social norms, so a great paper by Ileana La 00:56:53.595 --> 00:56:57.630 align:middle line:84% Ferrara on soap opera in Brazil to show that-- 00:56:57.630 --> 00:57:00.710 align:middle line:90% so in Brazil during the-- 00:57:00.710 --> 00:57:03.696 align:middle line:90% 00:57:03.696 --> 00:57:09.660 align:middle line:84% before the democracy came back, a lot of the film, 00:57:09.660 --> 00:57:12.720 align:middle line:90% cinema people moved to TV. 00:57:12.720 --> 00:57:16.230 align:middle line:84% Because this was less dangerous politically than doing films. 00:57:16.230 --> 00:57:18.810 align:middle line:84% And they were doing this soap opera. 00:57:18.810 --> 00:57:21.000 align:middle line:84% And the model of the soap opera was usually 00:57:21.000 --> 00:57:23.700 align:middle line:84% a woman with no children, or very few children 00:57:23.700 --> 00:57:26.730 align:middle line:84% living this modern life, and attractive, 00:57:26.730 --> 00:57:29.380 align:middle line:90% in an attractive setup. 00:57:29.380 --> 00:57:31.030 align:middle line:84% And what she does is that she looks 00:57:31.030 --> 00:57:34.270 align:middle line:84% at how the soap opera diffused via cable TV 00:57:34.270 --> 00:57:40.350 align:middle line:84% all over Brazil with the introduction of cable 00:57:40.350 --> 00:57:46.440 align:middle line:84% and does a definitive to show that as places get access 00:57:46.440 --> 00:57:53.260 align:middle line:84% to the soap opera, the number of kids declines in those places. 00:57:53.260 --> 00:57:56.260 align:middle line:84% And one interesting supporting piece 00:57:56.260 --> 00:58:00.940 align:middle line:84% of evidence, that might, in fact, be related to that 00:58:00.940 --> 00:58:03.685 align:middle line:84% is that the name given to the kids 00:58:03.685 --> 00:58:06.530 align:middle line:84% are increasingly the names of the people in the soap opera. 00:58:06.530 --> 00:58:07.030 align:middle line:90% [LAUGHTER] 00:58:07.030 --> 00:58:12.730 align:middle line:84% So that suggests that indeed, people are paying attention. 00:58:12.730 --> 00:58:17.640 align:middle line:84% And something as important as fertility is affected by TV. 00:58:17.640 --> 00:58:20.980 align:middle line:84% Abhijit and Ileana La Ferrara, in another recent paper, 00:58:20.980 --> 00:58:23.830 align:middle line:84% are doing a randomized experiment of an MTV program 00:58:23.830 --> 00:58:29.800 align:middle line:84% in Nigeria looking at basically risky sexual behavior, 00:58:29.800 --> 00:58:34.000 align:middle line:84% respectful attitude of boys vis-a-vis a girlfriend, 00:58:34.000 --> 00:58:38.770 align:middle line:84% and these type of things, which also shows an impact of TV. 00:58:38.770 --> 00:58:40.510 align:middle line:90% So TV is one. 00:58:40.510 --> 00:58:46.450 align:middle line:84% Another thing that seems to be changed by exposure 00:58:46.450 --> 00:58:51.160 align:middle line:84% is, I have a paper with Lori Beaman, Rohini Pande 00:58:51.160 --> 00:58:52.900 align:middle line:84% and Raghabendra Chattopadhyay by looking 00:58:52.900 --> 00:58:57.490 align:middle line:84% at women leaders in India through a series of papers. 00:58:57.490 --> 00:59:00.580 align:middle line:84% One of the papers is about what people think of-- 00:59:00.580 --> 00:59:03.800 align:middle line:84% whether it's appropriate for a woman to be leader. 00:59:03.800 --> 00:59:07.970 align:middle line:84% And what we find is that exposure to women-- 00:59:07.970 --> 00:59:12.110 align:middle line:90% it is imposed on you by quotas-- 00:59:12.110 --> 00:59:14.390 align:middle line:84% makes people more willing to accept 00:59:14.390 --> 00:59:16.370 align:middle line:84% the idea that a woman can be a good leader, 00:59:16.370 --> 00:59:19.760 align:middle line:84% and in fact, to vote for women, which again shows the fact 00:59:19.760 --> 00:59:23.950 align:middle line:90% that these ideas can change. 00:59:23.950 --> 00:59:27.370 align:middle line:84% And you might imagine that they might change-- 00:59:27.370 --> 00:59:29.440 align:middle line:84% it might be easier to change the social norm 00:59:29.440 --> 00:59:31.490 align:middle line:90% when children are younger. 00:59:31.490 --> 00:59:35.780 align:middle line:84% And therefore, education would be a good place. 00:59:35.780 --> 00:59:37.510 align:middle line:84% Schools would be a good time to do it. 00:59:37.510 --> 00:59:43.300 align:middle line:84% So Sima Jayachandran and Tarun Jain 00:59:43.300 --> 00:59:45.880 align:middle line:84% have done this paper in collaboration with an NGO 00:59:45.880 --> 00:59:48.400 align:middle line:84% in North India called Breakthrough 00:59:48.400 --> 00:59:52.210 align:middle line:84% to try to affect adolescent view of women and girls. 00:59:52.210 --> 00:59:54.040 align:middle line:90% It's a pretty intense program. 00:59:54.040 --> 00:59:57.820 align:middle line:84% It's a 45-minute classroom discussion every three weeks 00:59:57.820 --> 00:59:59.140 align:middle line:90% for two years. 00:59:59.140 --> 01:00:02.590 align:middle line:84% And they did RCT in 314 school in Haryana, 01:00:02.590 --> 01:00:04.180 align:middle line:84% which might be the place in the world 01:00:04.180 --> 01:00:08.350 align:middle line:84% where women are the least considered, maybe, 01:00:08.350 --> 01:00:11.050 align:middle line:90% with competition with UP-- 01:00:11.050 --> 01:00:20.020 align:middle line:84% extremely backward gender norms, huge discrepancy 01:00:20.020 --> 01:00:22.780 align:middle line:84% in the number of boys who are born versus girls. 01:00:22.780 --> 01:00:26.050 align:middle line:84% In any way you like, Haryana is a pretty horrible place 01:00:26.050 --> 01:00:27.460 align:middle line:90% for women. 01:00:27.460 --> 01:00:30.300 align:middle line:84% So if you find an effect there, that's encouraging. 01:00:30.300 --> 01:00:36.030 align:middle line:84% And what they find overall is some effect-- a 0.25 standard 01:00:36.030 --> 01:00:38.680 align:middle line:84% deviation improvement on self-reported gender norms-- 01:00:38.680 --> 01:00:40.720 align:middle line:84% so whether it's OK for women to work, 01:00:40.720 --> 01:00:44.550 align:middle line:84% whether boys need to leave enough space for women, 01:00:44.550 --> 01:00:47.940 align:middle line:84% whether if there is not enough job, 01:00:47.940 --> 01:00:50.400 align:middle line:84% it's men who should have the first job, and so on and so 01:00:50.400 --> 01:00:50.980 align:middle line:90% forth. 01:00:50.980 --> 01:00:56.220 align:middle line:84% And some effect on behavior, even among, in particular, 01:00:56.220 --> 01:00:58.750 align:middle line:90% among boys-- 01:00:58.750 --> 01:01:01.500 align:middle line:84% so this is the summary looking at their index. 01:01:01.500 --> 01:01:03.660 align:middle line:84% You should look at their paper to look at what 01:01:03.660 --> 01:01:06.120 align:middle line:90% goes into each of the index. 01:01:06.120 --> 01:01:08.700 align:middle line:84% So the general attitude I summarize-- the aspiration, 01:01:08.700 --> 01:01:11.610 align:middle line:84% or what you want to do with your career. 01:01:11.610 --> 01:01:13.470 align:middle line:84% The behaviors-- what to do in terms 01:01:13.470 --> 01:01:17.550 align:middle line:84% of chores, and stuff like that, both for girls and for boys. 01:01:17.550 --> 01:01:19.170 align:middle line:90% And so they find this effect. 01:01:19.170 --> 01:01:23.400 align:middle line:84% They find that it persists over several years 01:01:23.400 --> 01:01:24.850 align:middle line:90% after the intervention. 01:01:24.850 --> 01:01:32.410 align:middle line:84% So they are now trying to scale it up in the rest of the state. 01:01:32.410 --> 01:01:34.230 align:middle line:84% So that's very encouraging in the sense 01:01:34.230 --> 01:01:38.560 align:middle line:84% that it suggests that it might be possible to teach directly 01:01:38.560 --> 01:01:39.920 align:middle line:90% teach social norms. 01:01:39.920 --> 01:01:40.420 align:middle line:90% Yeah? 01:01:40.420 --> 01:01:42.870 align:middle line:84% STUDENT: Through a study like this, 01:01:42.870 --> 01:01:46.730 align:middle line:84% how important is it to have behavior measures? 01:01:46.730 --> 01:01:48.510 align:middle line:84% Could someone, say, let's demand effects? 01:01:48.510 --> 01:01:49.170 align:middle line:90% Or what do you think? 01:01:49.170 --> 01:01:49.830 align:middle line:90% ESTHER DUFLO: Yeah, yeah. 01:01:49.830 --> 01:01:50.788 align:middle line:90% Someone could say that. 01:01:50.788 --> 01:01:51.990 align:middle line:90% And someone has said that. 01:01:51.990 --> 01:01:54.805 align:middle line:90% 01:01:54.805 --> 01:01:57.480 align:middle line:84% This is also a paper I edited at the AER 01:01:57.480 --> 01:02:00.150 align:middle line:84% and I've decided to publish because I think 01:02:00.150 --> 01:02:01.620 align:middle line:90% the result is so important. 01:02:01.620 --> 01:02:05.235 align:middle line:84% But it's clearly come up in all of the reports saying, well, 01:02:05.235 --> 01:02:07.500 align:middle line:84% you just told the kids for two years 01:02:07.500 --> 01:02:13.290 align:middle line:84% that they have to be nice to their women. 01:02:13.290 --> 01:02:15.750 align:middle line:90% So they are parroting that back. 01:02:15.750 --> 01:02:24.150 align:middle line:84% So there is an interesting twist in the paper, where 01:02:24.150 --> 01:02:29.190 align:middle line:84% they develop a test, or they use a test for testing 01:02:29.190 --> 01:02:34.590 align:middle line:84% for social desirability, trying to catch social desirability 01:02:34.590 --> 01:02:37.710 align:middle line:90% bias that they implement. 01:02:37.710 --> 01:02:42.210 align:middle line:84% It also helps that they are looking for long-term outcomes. 01:02:42.210 --> 01:02:44.520 align:middle line:84% So a long time after the intervention 01:02:44.520 --> 01:02:46.980 align:middle line:90% itself was stopped-- 01:02:46.980 --> 01:02:49.970 align:middle line:84% so if people still remember one year after the fact 01:02:49.970 --> 01:02:52.790 align:middle line:84% that you want them to say that one has to be nice to women, 01:02:52.790 --> 01:02:58.170 align:middle line:84% maybe there is no difference anymore between people 01:02:58.170 --> 01:03:01.590 align:middle line:84% wanting to say that and people actually believing it. 01:03:01.590 --> 01:03:04.020 align:middle line:84% And then of course, there is a self-reported behavior. 01:03:04.020 --> 01:03:05.260 align:middle line:90% That's, again, self-reported. 01:03:05.260 --> 01:03:06.690 align:middle line:90% So it could be demand effect. 01:03:06.690 --> 01:03:09.810 align:middle line:90% 01:03:09.810 --> 01:03:16.800 align:middle line:84% It's whether you help at home and that type of things 01:03:16.800 --> 01:03:21.900 align:middle line:84% would be lying, which they might also do just to be nice. 01:03:21.900 --> 01:03:31.640 align:middle line:84% But that remains-- that finding a hard measure of impact 01:03:31.640 --> 01:03:36.835 align:middle line:84% on behavior would be the cool thing to do there. 01:03:36.835 --> 01:03:38.210 align:middle line:84% Of course, these are adolescents. 01:03:38.210 --> 01:03:40.640 align:middle line:84% So you could try and trace them until they 01:03:40.640 --> 01:03:45.330 align:middle line:84% get married, like in our paper in Ghana, 01:03:45.330 --> 01:03:47.000 align:middle line:90% where we traced people forever. 01:03:47.000 --> 01:03:51.170 align:middle line:84% But it's a pretty big commitment. 01:03:51.170 --> 01:03:58.050 align:middle line:90% 01:03:58.050 --> 01:04:06.660 align:middle line:84% So now given the disagreement, thinking now the social norm as 01:04:06.660 --> 01:04:11.190 align:middle line:84% given for women, to get women to participate in the labor force, 01:04:11.190 --> 01:04:14.070 align:middle line:84% you could either change husbands' opinion 01:04:14.070 --> 01:04:17.890 align:middle line:84% or change the wives' ability to advocate for themselves 01:04:17.890 --> 01:04:19.180 align:middle line:90% if there is this-- 01:04:19.180 --> 01:04:23.850 align:middle line:84% as long as people still have these different shared opinion. 01:04:23.850 --> 01:04:26.750 align:middle line:84% So a paper by Maddie McKelway looks at that. 01:04:26.750 --> 01:04:29.320 align:middle line:90% 01:04:29.320 --> 01:04:32.350 align:middle line:84% That actually was a job market paper-- so in the experiments 01:04:32.350 --> 01:04:37.300 align:middle line:84% she conducted on her own on a shoestring budget in UP. 01:04:37.300 --> 01:04:40.390 align:middle line:90% 01:04:40.390 --> 01:04:47.300 align:middle line:84% She hails from Virginia, so UP was quite a different setup. 01:04:47.300 --> 01:04:48.490 align:middle line:90% So it's well worth reading. 01:04:48.490 --> 01:04:50.032 align:middle line:84% First of all, it's a very good paper, 01:04:50.032 --> 01:04:52.240 align:middle line:84% second of all just to have a sense of what 01:04:52.240 --> 01:04:54.610 align:middle line:84% you can do as a graduate student if that's 01:04:54.610 --> 01:04:56.910 align:middle line:90% what you decide to do. 01:04:56.910 --> 01:04:59.460 align:middle line:84% She teamed up with a large carpet manufacturer 01:04:59.460 --> 01:05:04.120 align:middle line:84% in India, the largest carpet manufacturer in India. 01:05:04.120 --> 01:05:08.800 align:middle line:84% And they wanted to recruit women in part, 01:05:08.800 --> 01:05:12.670 align:middle line:84% and probably mostly because of tightness in the labor market. 01:05:12.670 --> 01:05:16.150 align:middle line:84% They just can't hire enough weavers anymore. 01:05:16.150 --> 01:05:18.640 align:middle line:84% And weaving is traditionally actually 01:05:18.640 --> 01:05:19.975 align:middle line:90% a male occupation in UP. 01:05:19.975 --> 01:05:23.410 align:middle line:90% 01:05:23.410 --> 01:05:25.690 align:middle line:84% It's funny because in the 192 class, 01:05:25.690 --> 01:05:27.640 align:middle line:84% I think it's Arthur who mentioned this 01:05:27.640 --> 01:05:29.410 align:middle line:84% as an archetypal female occupation. 01:05:29.410 --> 01:05:33.020 align:middle line:84% Well, it turns out, in UP, it's a male occupation. 01:05:33.020 --> 01:05:36.460 align:middle line:84% And they are trying to get women in it 01:05:36.460 --> 01:05:40.760 align:middle line:84% because there are no reasons why women couldn't do weaving. 01:05:40.760 --> 01:05:43.640 align:middle line:84% And so in fact, it's with this same partner 01:05:43.640 --> 01:05:47.150 align:middle line:84% that Matt and Maddie did their first experiment 01:05:47.150 --> 01:05:48.920 align:middle line:84% that I mentioned before, where they gave 01:05:48.920 --> 01:05:50.250 align:middle line:90% the voucher for the training. 01:05:50.250 --> 01:05:54.440 align:middle line:84% And so it's a second experiment with the same partner. 01:05:54.440 --> 01:05:57.230 align:middle line:90% And there they did two things. 01:05:57.230 --> 01:05:58.070 align:middle line:90% She did two things. 01:05:58.070 --> 01:06:01.820 align:middle line:84% She did a psychosocial intervention 01:06:01.820 --> 01:06:07.620 align:middle line:84% called generalized self-efficacy training over several weeks. 01:06:07.620 --> 01:06:08.820 align:middle line:90% That's one intervention. 01:06:08.820 --> 01:06:10.980 align:middle line:84% And the other is a promotion of the job 01:06:10.980 --> 01:06:13.800 align:middle line:84% to the husband and the in-laws, which 01:06:13.800 --> 01:06:16.020 align:middle line:84% is a six-minute video that was meant 01:06:16.020 --> 01:06:20.650 align:middle line:84% to address some of the concern that husbands might have had. 01:06:20.650 --> 01:06:22.680 align:middle line:84% So they did some focus group to figure out 01:06:22.680 --> 01:06:24.030 align:middle line:90% what people were worried about. 01:06:24.030 --> 01:06:25.320 align:middle line:90% And then they did this video. 01:06:25.320 --> 01:06:30.510 align:middle line:84% And then they showed the video on the iPad 01:06:30.510 --> 01:06:32.430 align:middle line:90% when going from house to house. 01:06:32.430 --> 01:06:37.440 align:middle line:84% So the video had explained that it was women-only workplaces 01:06:37.440 --> 01:06:42.940 align:middle line:84% and what people would be doing there and the money they 01:06:42.940 --> 01:06:45.760 align:middle line:90% could make, and all of that. 01:06:45.760 --> 01:06:48.070 align:middle line:84% And then of course, so you have two treatments 01:06:48.070 --> 01:06:51.070 align:middle line:84% and the interaction because you might 01:06:51.070 --> 01:06:53.380 align:middle line:84% expect that the self-efficacy intervention would 01:06:53.380 --> 01:07:02.320 align:middle line:84% have different impact when the husband might be predisposed 01:07:02.320 --> 01:07:05.860 align:middle line:84% to be interested in the job versus when they are not. 01:07:05.860 --> 01:07:09.130 align:middle line:90% 01:07:09.130 --> 01:07:12.000 align:middle line:84% So the general self-efficacy training-- so this 01:07:12.000 --> 01:07:15.360 align:middle line:84% is an old concept by Bandura from '77, 01:07:15.360 --> 01:07:18.180 align:middle line:84% which is actually very, very classic in psychology. 01:07:18.180 --> 01:07:21.480 align:middle line:84% But it's one of the concepts that has not yet 01:07:21.480 --> 01:07:24.180 align:middle line:84% made its way into economics and behavioral economics 01:07:24.180 --> 01:07:26.310 align:middle line:90% in a large way. 01:07:26.310 --> 01:07:28.800 align:middle line:84% Maybe after Maddie's paper, it will. 01:07:28.800 --> 01:07:35.790 align:middle line:84% And it's the idea that you can affect the outcomes, 01:07:35.790 --> 01:07:39.630 align:middle line:84% you can affect your outcomes by your agency. 01:07:39.630 --> 01:07:42.990 align:middle line:84% You have agency on the outcomes that are around you. 01:07:42.990 --> 01:07:46.890 align:middle line:84% So they're basically going from case studies 01:07:46.890 --> 01:07:53.830 align:middle line:84% and discussing two situations or problems that women might have, 01:07:53.830 --> 01:07:57.300 align:middle line:90% and then how she can address it. 01:07:57.300 --> 01:08:04.500 align:middle line:84% And there is a standard scale for general self-efficacy. 01:08:04.500 --> 01:08:07.710 align:middle line:84% So generalizes as opposed to in any particular area. 01:08:07.710 --> 01:08:09.900 align:middle line:84% You could have self-efficacy for child rearing, 01:08:09.900 --> 01:08:14.310 align:middle line:84% and no self-efficacy for labor market. 01:08:14.310 --> 01:08:19.020 align:middle line:84% General self-efficacy is across all areas. 01:08:19.020 --> 01:08:20.370 align:middle line:90% So there is a scale for it. 01:08:20.370 --> 01:08:23.750 align:middle line:84% So she measured at the end and then at the beginning. 01:08:23.750 --> 01:08:28.069 align:middle line:84% And then the training took place. 01:08:28.069 --> 01:08:31.609 align:middle line:84% Importantly, in the training, she was sure to pick-- 01:08:31.609 --> 01:08:34.040 align:middle line:84% so she worked with an organization 01:08:34.040 --> 01:08:37.790 align:middle line:84% that was with psychologists who were specialists in that. 01:08:37.790 --> 01:08:40.007 align:middle line:84% And she changed a little bit the curriculum 01:08:40.007 --> 01:08:41.840 align:middle line:84% to make sure that there was no examples that 01:08:41.840 --> 01:08:44.580 align:middle line:84% had anything to do with labor markets 01:08:44.580 --> 01:08:48.090 align:middle line:84% because she didn't want to directly lead 01:08:48.090 --> 01:08:49.109 align:middle line:90% in this direction. 01:08:49.109 --> 01:08:51.740 align:middle line:90% 01:08:51.740 --> 01:08:54.950 align:middle line:90% So in terms of the-- 01:08:54.950 --> 01:08:57.680 align:middle line:84% so the GSE questions are of the form, 01:08:57.680 --> 01:08:59.060 align:middle line:90% can you affect this outcome? 01:08:59.060 --> 01:09:00.380 align:middle line:90% Can you affect that outcome? 01:09:00.380 --> 01:09:03.710 align:middle line:84% Generally, you would say that if a child is sick, 01:09:03.710 --> 01:09:11.760 align:middle line:84% you are able to do what it takes to help them, et cetera. 01:09:11.760 --> 01:09:15.390 align:middle line:84% So she looks at the impact of the treatment 01:09:15.390 --> 01:09:16.140 align:middle line:90% and the promotion. 01:09:16.140 --> 01:09:18.557 align:middle line:84% You can look at the impact of the treatment, the promotion 01:09:18.557 --> 01:09:19.500 align:middle line:90% or the two interacted. 01:09:19.500 --> 01:09:21.330 align:middle line:84% Given what we discussed about interaction, 01:09:21.330 --> 01:09:24.760 align:middle line:84% let's focus on the interaction specification immediately. 01:09:24.760 --> 01:09:25.930 align:middle line:90% And you can see that-- 01:09:25.930 --> 01:09:29.943 align:middle line:90% 01:09:29.943 --> 01:09:34.950 align:middle line:84% you can look at the impact of the intervention 01:09:34.950 --> 01:09:37.649 align:middle line:90% after several weeks. 01:09:37.649 --> 01:09:43.540 align:middle line:84% And starting towards at the end of the program, 01:09:43.540 --> 01:09:47.520 align:middle line:90% the GSE leads to-- 01:09:47.520 --> 01:09:54.100 align:middle line:84% the two GSE specification leads to more GSE. 01:09:54.100 --> 01:10:02.250 align:middle line:84% And the promotion itself doesn't lead to more GSE. 01:10:02.250 --> 01:10:04.680 align:middle line:84% It's not, perhaps, powered very much. 01:10:04.680 --> 01:10:08.400 align:middle line:84% So it's difficult to show that these things are significantly 01:10:08.400 --> 01:10:12.270 align:middle line:84% different, which is something she has been struggling with. 01:10:12.270 --> 01:10:16.610 align:middle line:90% The sample size is not enormous. 01:10:16.610 --> 01:10:20.330 align:middle line:84% But you do find that over time, it 01:10:20.330 --> 01:10:25.460 align:middle line:84% seems relatively stable, that the GSE treatment seems 01:10:25.460 --> 01:10:27.350 align:middle line:90% to increase GSE. 01:10:27.350 --> 01:10:28.850 align:middle line:84% But you can see why she preferred 01:10:28.850 --> 01:10:33.200 align:middle line:84% to look at an interactive specification 01:10:33.200 --> 01:10:36.170 align:middle line:84% because then it looks like it's pulling these two, 01:10:36.170 --> 01:10:38.580 align:middle line:90% so you get some extra precision. 01:10:38.580 --> 01:10:40.880 align:middle line:84% And you're finding, then, across the board 01:10:40.880 --> 01:10:43.700 align:middle line:84% that being treated with GSE increases your GSE. 01:10:43.700 --> 01:10:47.910 align:middle line:90% 01:10:47.910 --> 01:10:52.500 align:middle line:84% Then when we look at participation 01:10:52.500 --> 01:10:55.340 align:middle line:90% in the farm program-- 01:10:55.340 --> 01:10:57.730 align:middle line:84% so the main outcome was, is there 01:10:57.730 --> 01:10:59.620 align:middle line:84% participation in the farm program 01:10:59.620 --> 01:11:03.970 align:middle line:90% of working outside of the farm? 01:11:03.970 --> 01:11:08.470 align:middle line:84% You find that the promotion and the robust increase in signing 01:11:08.470 --> 01:11:11.840 align:middle line:84% up for the farm program and for doing any work, 01:11:11.840 --> 01:11:15.370 align:middle line:84% including this one, at least in the first five months-- 01:11:15.370 --> 01:11:18.320 align:middle line:84% though it was not durable over time-- 01:11:18.320 --> 01:11:23.090 align:middle line:84% the GSE had an impact that's not significant for signing up 01:11:23.090 --> 01:11:24.860 align:middle line:90% on the farm's program. 01:11:24.860 --> 01:11:29.210 align:middle line:84% It's significant for working on farms in general. 01:11:29.210 --> 01:11:32.330 align:middle line:84% Remember that the GSE was not about-- 01:11:32.330 --> 01:11:34.710 align:middle line:84% the GSE intervention had nothing to do with the program. 01:11:34.710 --> 01:11:37.340 align:middle line:84% So it was not particularly promoting the program. 01:11:37.340 --> 01:11:39.860 align:middle line:84% It was just giving people, women agency 01:11:39.860 --> 01:11:41.850 align:middle line:84% to do whatever they want in their life. 01:11:41.850 --> 01:11:43.670 align:middle line:90% It was not even about work. 01:11:43.670 --> 01:11:45.840 align:middle line:84% So that's probably a better outcome. 01:11:45.840 --> 01:11:49.520 align:middle line:84% So GSE alone increased the female labor force 01:11:49.520 --> 01:11:50.460 align:middle line:90% participation. 01:11:50.460 --> 01:11:53.010 align:middle line:84% So this is something that women tend to want to do. 01:11:53.010 --> 01:11:55.640 align:middle line:84% And in fact, you see that they want to do about as much 01:11:55.640 --> 01:11:59.660 align:middle line:84% as the promotion treatment directly to the husband. 01:11:59.660 --> 01:12:03.170 align:middle line:84% But when you combine the two, instead of having a larger 01:12:03.170 --> 01:12:06.140 align:middle line:84% effect, which is what I would have expected, 01:12:06.140 --> 01:12:11.280 align:middle line:84% you find no significant effect or a smaller effect. 01:12:11.280 --> 01:12:13.280 align:middle line:84% And in fact, at six weeks, the effect 01:12:13.280 --> 01:12:19.340 align:middle line:84% is even significantly different from having just a promo. 01:12:19.340 --> 01:12:21.650 align:middle line:84% If you add the GSE to the promotion, 01:12:21.650 --> 01:12:24.400 align:middle line:90% you have the effect crashing. 01:12:24.400 --> 01:12:30.910 align:middle line:84% So that's a bit surprising and maybe a bit reminiscent of this 01:12:30.910 --> 01:12:34.510 align:middle line:84% discussion treatment that led to, actually, 01:12:34.510 --> 01:12:38.260 align:middle line:84% less take-up than when either the wife alone or the husband 01:12:38.260 --> 01:12:41.980 align:middle line:84% alone was given the ticket, and again suggests that there is 01:12:41.980 --> 01:12:47.400 align:middle line:84% something about basically women-- 01:12:47.400 --> 01:12:49.710 align:middle line:84% one possible interpretation-- I'm not saying it's 01:12:49.710 --> 01:12:52.560 align:middle line:84% the right one, but one possible interpretation of this result 01:12:52.560 --> 01:12:56.610 align:middle line:90% is that women want to work. 01:12:56.610 --> 01:13:01.380 align:middle line:84% With the GSE treatment, they are given the power 01:13:01.380 --> 01:13:03.630 align:middle line:90% to advocate for what they want. 01:13:03.630 --> 01:13:07.560 align:middle line:90% And then they do it. 01:13:07.560 --> 01:13:09.990 align:middle line:84% Or a husband that's in the GSE treatment alone, 01:13:09.990 --> 01:13:13.560 align:middle line:84% or husbands find out that actually, she can make money. 01:13:13.560 --> 01:13:14.710 align:middle line:90% And it's not a bad thing. 01:13:14.710 --> 01:13:19.660 align:middle line:84% Let's give this a try, and push them to work. 01:13:19.660 --> 01:13:23.440 align:middle line:84% But when she has self-efficacy to argue for herself, 01:13:23.440 --> 01:13:26.610 align:middle line:84% she's like, I'm not going to take this new job that you just 01:13:26.610 --> 01:13:29.730 align:middle line:84% decided is interesting or something like that. 01:13:29.730 --> 01:13:34.050 align:middle line:84% She's sufficiently-- she's less-- 01:13:34.050 --> 01:13:40.440 align:middle line:84% she's resistant to this new idea that has been imposed to her. 01:13:40.440 --> 01:13:45.330 align:middle line:84% That's just an interpretation, which is hard to prove, 01:13:45.330 --> 01:13:49.830 align:middle line:84% but again shows how the things are complicated because again, 01:13:49.830 --> 01:13:53.430 align:middle line:84% my [? prior ?] would have been that combining the two 01:13:53.430 --> 01:13:56.595 align:middle line:84% would have a larger effect than doing either individually. 01:13:56.595 --> 01:14:00.060 align:middle line:90% 01:14:00.060 --> 01:14:04.130 align:middle line:84% So one of the reason why men might 01:14:04.130 --> 01:14:06.350 align:middle line:84% be reluctant to see wives working 01:14:06.350 --> 01:14:08.300 align:middle line:84% is because it might increase bargaining 01:14:08.300 --> 01:14:10.250 align:middle line:90% power, their bargaining power. 01:14:10.250 --> 01:14:15.887 align:middle line:84% And in a limited commitment model, 01:14:15.887 --> 01:14:16.970 align:middle line:90% you don't want to do that. 01:14:16.970 --> 01:14:21.620 align:middle line:84% You want to keep your-- so men might want to keep women home 01:14:21.620 --> 01:14:23.510 align:middle line:84% in order to make sure that they remain-- 01:14:23.510 --> 01:14:28.370 align:middle line:84% they don't become snooty, and the man 01:14:28.370 --> 01:14:31.500 align:middle line:90% remains the boss of the house. 01:14:31.500 --> 01:14:36.470 align:middle line:84% So there is an experiment, or a project, ambitious project 01:14:36.470 --> 01:14:41.300 align:middle line:84% by Rohini Pande, Natalia Rigol, Simon Schaner, Charity T. Moore 01:14:41.300 --> 01:14:48.480 align:middle line:84% and Erica Field, which is in Madhya Pradesh, where they 01:14:48.480 --> 01:14:52.390 align:middle line:90% didn't increase access to work. 01:14:52.390 --> 01:14:56.770 align:middle line:84% But they increased the control of the woman 01:14:56.770 --> 01:15:00.250 align:middle line:84% to the money coming from work that they were doing anyways. 01:15:00.250 --> 01:15:03.580 align:middle line:84% So this is in the context of the NREGA program. 01:15:03.580 --> 01:15:06.760 align:middle line:84% The NREGA is a workfare program in India 01:15:06.760 --> 01:15:11.650 align:middle line:84% whereby women and men, everyone, has access 01:15:11.650 --> 01:15:17.210 align:middle line:84% to a certain number of labor every year. 01:15:17.210 --> 01:15:19.360 align:middle line:84% So this is quite well paid because it 01:15:19.360 --> 01:15:21.550 align:middle line:90% sets the official minimum wage. 01:15:21.550 --> 01:15:26.740 align:middle line:84% And in some states, it's actually a big part 01:15:26.740 --> 01:15:30.760 align:middle line:90% of the household sustenance. 01:15:30.760 --> 01:15:36.010 align:middle line:84% So what they look at here is a program where people, women-- 01:15:36.010 --> 01:15:38.650 align:middle line:84% the government gives access to bank accounts 01:15:38.650 --> 01:15:41.210 align:middle line:90% in randomly selected GPs. 01:15:41.210 --> 01:15:44.870 align:middle line:84% And in one treatment, the NREGA payment of the woman 01:15:44.870 --> 01:15:49.550 align:middle line:84% was directly sent to the bank account of the woman. 01:15:49.550 --> 01:15:53.240 align:middle line:84% So it doesn't change the access to job, which in a sense, 01:15:53.240 --> 01:15:54.260 align:middle line:90% they already have. 01:15:54.260 --> 01:15:57.790 align:middle line:84% But it makes sure that they control their own job. 01:15:57.790 --> 01:16:00.130 align:middle line:84% And what they find in terms of impacts 01:16:00.130 --> 01:16:05.180 align:middle line:84% is that it increases a labor supply. 01:16:05.180 --> 01:16:07.870 align:middle line:84% So women who have now access to-- 01:16:07.870 --> 01:16:10.330 align:middle line:84% who can now retain their earnings 01:16:10.330 --> 01:16:13.630 align:middle line:84% are more likely to work in the short run, 01:16:13.630 --> 01:16:16.730 align:middle line:90% although not in the long run. 01:16:16.730 --> 01:16:20.200 align:middle line:84% And so this is just when this is happening because 01:16:20.200 --> 01:16:21.650 align:middle line:90% at some point, it stopped. 01:16:21.650 --> 01:16:23.380 align:middle line:90% So this is while it's happening. 01:16:23.380 --> 01:16:26.335 align:middle line:84% They have more access to the program. 01:16:26.335 --> 01:16:30.520 align:middle line:90% 01:16:30.520 --> 01:16:34.180 align:middle line:84% This is true both in the NREGA program 01:16:34.180 --> 01:16:36.430 align:middle line:84% and in other forms of work, which 01:16:36.430 --> 01:16:40.020 align:middle line:84% is not directly deposited to this account. 01:16:40.020 --> 01:16:46.670 align:middle line:84% So their interpretation is that women found out about this-- 01:16:46.670 --> 01:16:49.760 align:middle line:84% that now they would be able to retain their earnings, 01:16:49.760 --> 01:16:51.080 align:middle line:90% they started working. 01:16:51.080 --> 01:16:52.850 align:middle line:84% That increases their bargaining power 01:16:52.850 --> 01:16:55.560 align:middle line:84% and gave them the possibility to work even more. 01:16:55.560 --> 01:16:58.250 align:middle line:84% And so that's why you have an increase 01:16:58.250 --> 01:17:03.830 align:middle line:84% in labor supply in NREGA and in a non-NREGA program as well. 01:17:03.830 --> 01:17:08.040 align:middle line:84% And they show some evidence for that, 01:17:08.040 --> 01:17:13.880 align:middle line:84% which is that if you look at other dimensions of women 01:17:13.880 --> 01:17:14.960 align:middle line:90% empowerment-- 01:17:14.960 --> 01:17:17.720 align:middle line:84% for example, mobility, et cetera-- 01:17:17.720 --> 01:17:20.480 align:middle line:84% among the women who are the most constrained, 01:17:20.480 --> 01:17:22.580 align:middle line:84% among the women who are the least empowered 01:17:22.580 --> 01:17:29.330 align:middle line:84% at the beginning, you find some evidence that other measures 01:17:29.330 --> 01:17:32.280 align:middle line:84% of empowerment, like how much they can purchase, 01:17:32.280 --> 01:17:35.520 align:middle line:84% they have a say on decision-making, 01:17:35.520 --> 01:17:40.950 align:middle line:84% they can move to visit their parents or stuff like that-- 01:17:40.950 --> 01:17:42.980 align:middle line:84% and this is an aggregate of all of the measures 01:17:42.980 --> 01:17:46.410 align:middle line:84% that they have-- is affected by the program. 01:17:46.410 --> 01:17:49.760 align:middle line:84% So this is the idea that indeed, the bargaining power 01:17:49.760 --> 01:17:53.760 align:middle line:84% might be an impact of how much you're working. 01:17:53.760 --> 01:17:58.910 align:middle line:84% So here you keep more of your earnings. 01:17:58.910 --> 01:18:00.410 align:middle line:90% And therefore, you work more. 01:18:00.410 --> 01:18:02.490 align:middle line:84% The two combined gives you more bargaining power. 01:18:02.490 --> 01:18:03.440 align:middle line:90% You work even more. 01:18:03.440 --> 01:18:07.460 align:middle line:84% And that also increases your decision power 01:18:07.460 --> 01:18:09.350 align:middle line:90% in other spheres of the family. 01:18:09.350 --> 01:18:12.580 align:middle line:90% So that's the optimistic take. 01:18:12.580 --> 01:18:22.420 align:middle line:90% This is the how the norms-- 01:18:22.420 --> 01:18:24.640 align:middle line:84% male and female reports of various things 01:18:24.640 --> 01:18:28.960 align:middle line:84% that are affected, and again, showing some impact 01:18:28.960 --> 01:18:30.150 align:middle line:90% on those social norms. 01:18:30.150 --> 01:18:33.330 align:middle line:90% So that's the optimistic take. 01:18:33.330 --> 01:18:37.940 align:middle line:84% The less optimistic take is from a third paper 01:18:37.940 --> 01:18:42.350 align:middle line:84% by Maddie, which looks at the long-run effect of the access 01:18:42.350 --> 01:18:47.030 align:middle line:84% to this employment opportunity due to the exogenously-- 01:18:47.030 --> 01:18:49.410 align:middle line:90% due to the promotion sample. 01:18:49.410 --> 01:18:54.030 align:middle line:84% So now forgetting the general self-efficacy, 01:18:54.030 --> 01:18:57.710 align:middle line:84% which itself would have changed bargaining power-- 01:18:57.710 --> 01:19:00.110 align:middle line:84% and in fact, she does find some impact 01:19:00.110 --> 01:19:03.750 align:middle line:84% on savings and some impact on own decisions. 01:19:03.750 --> 01:19:05.760 align:middle line:84% The promotion gave access to women 01:19:05.760 --> 01:19:08.220 align:middle line:84% to this outside employment opportunity. 01:19:08.220 --> 01:19:10.530 align:middle line:84% So a natural question is whether there 01:19:10.530 --> 01:19:13.740 align:middle line:84% is an impact of that over time of this new work 01:19:13.740 --> 01:19:19.430 align:middle line:84% that they have on their on their decision-making. 01:19:19.430 --> 01:19:23.180 align:middle line:84% So she follows decision-making in households where husbands 01:19:23.180 --> 01:19:26.650 align:middle line:84% were given the promotion compared to the control group, 01:19:26.650 --> 01:19:28.720 align:middle line:90% forgetting the GSE. 01:19:28.720 --> 01:19:30.700 align:middle line:84% And what she finds is that in the short run, 01:19:30.700 --> 01:19:34.690 align:middle line:84% women spend more time working, but just as much time 01:19:34.690 --> 01:19:36.340 align:middle line:90% on doing chores. 01:19:36.340 --> 01:19:39.950 align:middle line:84% So that they, in fact, have less leisure. 01:19:39.950 --> 01:19:41.930 align:middle line:84% At four months, they think that they 01:19:41.930 --> 01:19:45.980 align:middle line:84% have more decision-making power, but their family does not. 01:19:45.980 --> 01:19:47.740 align:middle line:84% And then quickly, they just drop out 01:19:47.740 --> 01:19:53.950 align:middle line:84% of the job because it's incompatible with 01:19:53.950 --> 01:19:57.360 align:middle line:90% their other responsibilities. 01:19:57.360 --> 01:20:02.370 align:middle line:84% So that shows that this expected effect, 01:20:02.370 --> 01:20:05.940 align:middle line:84% this hoped-for effect of working on increasing your bargaining 01:20:05.940 --> 01:20:06.630 align:middle line:90% power-- 01:20:06.630 --> 01:20:08.260 align:middle line:90% it doesn't always happen. 01:20:08.260 --> 01:20:10.790 align:middle line:90% It didn't happen in our people. 01:20:10.790 --> 01:20:12.740 align:middle line:84% So interestingly, they dropped out 01:20:12.740 --> 01:20:16.050 align:middle line:84% because that's incompatible with their other responsibilities. 01:20:16.050 --> 01:20:18.680 align:middle line:84% That linked to what we were seeing at the very beginning 01:20:18.680 --> 01:20:23.570 align:middle line:84% of the lecture, which is women often report that they would 01:20:23.570 --> 01:20:24.860 align:middle line:90% like to work, but they cannot. 01:20:24.860 --> 01:20:26.360 align:middle line:84% Because their husbands don't want, 01:20:26.360 --> 01:20:29.750 align:middle line:84% or they cannot because it's incompatible with 01:20:29.750 --> 01:20:31.580 align:middle line:84% their childcare responsibilities. 01:20:31.580 --> 01:20:34.440 align:middle line:84% And they would like a part-time job. 01:20:34.440 --> 01:20:38.070 align:middle line:84% So that's related to this Claudia Goldin idea 01:20:38.070 --> 01:20:42.360 align:middle line:84% that the problem is mainly that the jobs don't have 01:20:42.360 --> 01:20:44.376 align:middle line:90% to all be structured full-time. 01:20:44.376 --> 01:20:47.700 align:middle line:90% 01:20:47.700 --> 01:20:51.660 align:middle line:84% So perhaps, the strongest norm is that women must take care-- 01:20:51.660 --> 01:20:55.350 align:middle line:84% the social norms that hold women back-- 01:20:55.350 --> 01:21:00.180 align:middle line:84% you find everywhere, and helps tying to a lot of the things 01:21:00.180 --> 01:21:03.260 align:middle line:84% that it is women who must take care of children. 01:21:03.260 --> 01:21:06.120 align:middle line:84% We saw it in Cote d'Ivoire, where it's women who-- 01:21:06.120 --> 01:21:08.110 align:middle line:84% when women income goes up, there is more food 01:21:08.110 --> 01:21:11.500 align:middle line:84% on the table, which seems a bit absurd from the efficiency 01:21:11.500 --> 01:21:14.350 align:middle line:84% of the household point of view, but is related to this idea 01:21:14.350 --> 01:21:17.850 align:middle line:84% that it's women who have to feed kids. 01:21:17.850 --> 01:21:21.240 align:middle line:84% And this might explain another stylised fact, 01:21:21.240 --> 01:21:23.940 align:middle line:84% which was first established by the De Mel, McKenzie, 01:21:23.940 --> 01:21:26.310 align:middle line:90% and Woodruff. 01:21:26.310 --> 01:21:31.740 align:middle line:84% They show that when you give the same-sized grant randomly 01:21:31.740 --> 01:21:34.810 align:middle line:84% assigned to female business, there's 01:21:34.810 --> 01:21:38.910 align:middle line:84% no impact on the profitability of female business. 01:21:38.910 --> 01:21:40.950 align:middle line:84% The same grant to male business-- and there 01:21:40.950 --> 01:21:42.690 align:middle line:90% are large effects. 01:21:42.690 --> 01:21:45.350 align:middle line:84% So basically, it seems that the profitability, 01:21:45.350 --> 01:21:48.470 align:middle line:84% the marginal impact on money on female businesses 01:21:48.470 --> 01:21:51.380 align:middle line:84% is zero when it's very high on male businesses. 01:21:51.380 --> 01:21:54.980 align:middle line:84% So it's this fact of maybe female businesses 01:21:54.980 --> 01:21:59.080 align:middle line:84% are less profitable than male businesses. 01:21:59.080 --> 01:21:59.940 align:middle line:90% So why is that? 01:21:59.940 --> 01:22:01.020 align:middle line:90% And I'll finish on that. 01:22:01.020 --> 01:22:03.940 align:middle line:90% I am mindful of the time. 01:22:03.940 --> 01:22:06.180 align:middle line:84% Another paper by Rohini Pande, Natalia Rigol, 01:22:06.180 --> 01:22:07.435 align:middle line:90% and Erica Field-- 01:22:07.435 --> 01:22:10.260 align:middle line:84% it's this trio plus another co-author-- 01:22:10.260 --> 01:22:13.440 align:middle line:84% argued by revisiting the same data plus other data 01:22:13.440 --> 01:22:15.900 align:middle line:84% that they bring to the table that it's not 01:22:15.900 --> 01:22:18.720 align:middle line:84% the female business that isn't productive. 01:22:18.720 --> 01:22:20.880 align:middle line:84% It's that the money is never invested 01:22:20.880 --> 01:22:23.710 align:middle line:84% in the female business, that in fact, 01:22:23.710 --> 01:22:28.080 align:middle line:84% if you looked at single women and you give them 01:22:28.080 --> 01:22:31.172 align:middle line:84% fellowship of money, grant money, 01:22:31.172 --> 01:22:32.380 align:middle line:90% the effect would be the same. 01:22:32.380 --> 01:22:34.920 align:middle line:84% In fact, they don't really have very many single women 01:22:34.920 --> 01:22:36.030 align:middle line:90% to look at. 01:22:36.030 --> 01:22:40.680 align:middle line:84% But what happens, they show, is that the family income 01:22:40.680 --> 01:22:42.960 align:middle line:84% increases just as much if you give 01:22:42.960 --> 01:22:46.230 align:middle line:84% the money to the female business versus money 01:22:46.230 --> 01:22:47.400 align:middle line:90% to the male business. 01:22:47.400 --> 01:22:51.120 align:middle line:84% It's just that when you give money to the female business, 01:22:51.120 --> 01:22:53.650 align:middle line:84% it's invested in the male business. 01:22:53.650 --> 01:22:57.165 align:middle line:84% So it looks like the female business is not productive. 01:22:57.165 --> 01:22:59.290 align:middle line:84% But it's because the money was never invested there 01:22:59.290 --> 01:23:00.082 align:middle line:90% in the first place. 01:23:00.082 --> 01:23:02.600 align:middle line:90% 01:23:02.600 --> 01:23:07.960 align:middle line:84% Solène Delecourt have a paper where they run an experiment 01:23:07.960 --> 01:23:12.130 align:middle line:84% with vegetable vendors, also in India. 01:23:12.130 --> 01:23:15.260 align:middle line:84% And they show that if you give them the same input-- 01:23:15.260 --> 01:23:20.080 align:middle line:84% so they replicate that fact that women-owned vegetable stalls 01:23:20.080 --> 01:23:23.800 align:middle line:84% appear to be less productive than male-owned vegetable 01:23:23.800 --> 01:23:25.110 align:middle line:90% stalls. 01:23:25.110 --> 01:23:27.360 align:middle line:84% But then what they do is this fun experiment 01:23:27.360 --> 01:23:30.210 align:middle line:90% where they give traders-- 01:23:30.210 --> 01:23:33.450 align:middle line:84% they give them vegetables to sell. 01:23:33.450 --> 01:23:35.300 align:middle line:90% And they observe. 01:23:35.300 --> 01:23:39.270 align:middle line:84% They sell for one day or for several days. 01:23:39.270 --> 01:23:42.120 align:middle line:84% And then they calculate how much money they are making. 01:23:42.120 --> 01:23:44.910 align:middle line:84% And they show that starting with the same endowment, 01:23:44.910 --> 01:23:48.560 align:middle line:84% women make the same amount of money as men. 01:23:48.560 --> 01:23:51.190 align:middle line:84% So it's not that women are less good traders, 01:23:51.190 --> 01:23:54.500 align:middle line:84% or it's not that customers discriminate against women. 01:23:54.500 --> 01:23:56.570 align:middle line:84% It's just that they probably start 01:23:56.570 --> 01:23:58.550 align:middle line:90% with a much lower endowment. 01:23:58.550 --> 01:24:00.830 align:middle line:84% And that's very consistent with this idea 01:24:00.830 --> 01:24:03.890 align:middle line:84% of, in the Rigol Pande Field, paper 01:24:03.890 --> 01:24:08.030 align:middle line:84% that oh, the money was not productive because it was never 01:24:08.030 --> 01:24:10.520 align:middle line:90% invested in the first place. 01:24:10.520 --> 01:24:14.740 align:middle line:84% So then the question is, why is it that-- 01:24:14.740 --> 01:24:16.900 align:middle line:84% the lower investment in the female business 01:24:16.900 --> 01:24:20.140 align:middle line:84% is not, in turn, endogenously relating to the fact 01:24:20.140 --> 01:24:21.970 align:middle line:84% that the female business isn't productive. 01:24:21.970 --> 01:24:22.750 align:middle line:84% Because you could say, of course, 01:24:22.750 --> 01:24:23.800 align:middle line:90% you don't invest in women. 01:24:23.800 --> 01:24:25.258 align:middle line:84% Female businesses are unproductive. 01:24:25.258 --> 01:24:26.720 align:middle line:90% That's why you don't invest. 01:24:26.720 --> 01:24:29.080 align:middle line:84% But the Delecourt paper suggests that's not 01:24:29.080 --> 01:24:33.190 align:middle line:84% that because if we invest the same, we get the same. 01:24:33.190 --> 01:24:35.670 align:middle line:84% So maybe another reason why people are not 01:24:35.670 --> 01:24:37.800 align:middle line:84% investing in the female businesses 01:24:37.800 --> 01:24:39.300 align:middle line:84% is that the female business is never 01:24:39.300 --> 01:24:41.610 align:middle line:84% made to make much money in the first place. 01:24:41.610 --> 01:24:45.720 align:middle line:84% It's meant to be something that women can do while minding 01:24:45.720 --> 01:24:48.490 align:middle line:90% the kids near them. 01:24:48.490 --> 01:24:51.405 align:middle line:84% And so therefore, you can't really 01:24:51.405 --> 01:24:52.780 align:middle line:84% grow it because then you won't be 01:24:52.780 --> 01:24:54.160 align:middle line:90% able to take care of the kids. 01:24:54.160 --> 01:24:57.220 align:middle line:84% And the same constraint is precisely what 01:24:57.220 --> 01:24:59.470 align:middle line:84% makes it difficult for them to get or keep 01:24:59.470 --> 01:25:04.240 align:middle line:84% jobs, and also the reason why promotions are constrained. 01:25:04.240 --> 01:25:06.430 align:middle line:84% So there is another paper by Woodruff 01:25:06.430 --> 01:25:08.380 align:middle line:84% and others in Bangladesh showing that even 01:25:08.380 --> 01:25:13.480 align:middle line:84% in factories, the perfectly good tailors are not 01:25:13.480 --> 01:25:15.070 align:middle line:84% promoted at the same rate as men, 01:25:15.070 --> 01:25:16.600 align:middle line:84% even though if they are promoted, 01:25:16.600 --> 01:25:18.340 align:middle line:90% they are just as productive. 01:25:18.340 --> 01:25:20.028 align:middle line:84% And there are many reasons for that. 01:25:20.028 --> 01:25:22.570 align:middle line:84% But one of the reasons could be that there is an anticipation 01:25:22.570 --> 01:25:23.650 align:middle line:90% that they will not stay. 01:25:23.650 --> 01:25:25.540 align:middle line:84% There is, in fact, much larger turnover 01:25:25.540 --> 01:25:26.930 align:middle line:90% for women than for men. 01:25:26.930 --> 01:25:28.660 align:middle line:84% And [INAUDIBLE] has a paper on that 01:25:28.660 --> 01:25:30.490 align:middle line:90% also for women enterprises. 01:25:30.490 --> 01:25:35.910 align:middle line:84% And that may all be related to this child care norm. 01:25:35.910 --> 01:25:38.940 align:middle line:84% So let me stop there, which I think is a good place to stop. 01:25:38.940 --> 01:25:42.240 align:middle line:84% There is a very active agenda on gender, labor market. 01:25:42.240 --> 01:25:43.380 align:middle line:90% Just scratched the surface. 01:25:43.380 --> 01:25:48.180 align:middle line:84% But it's a good place to do it because you have everything 01:25:48.180 --> 01:25:53.610 align:middle line:84% a little bit coming there-- the norms, the family situation, 01:25:53.610 --> 01:25:55.400 align:middle line:90% et cetera. 01:25:55.400 --> 01:26:08.000 align:middle line:90%