WEBVTT 00:00:00.000 --> 00:00:06.487 align:middle line:90% [SQUEAKER] [SQUEAK] [SQUEAKIN] 00:00:06.487 --> 00:00:16.966 align:middle line:90% 00:00:16.966 --> 00:00:18.980 align:middle line:90% ESTHER DUFLO: Good morning. 00:00:18.980 --> 00:00:23.430 align:middle line:84% So in terms of the video and also in terms of you guys, 00:00:23.430 --> 00:00:27.830 align:middle line:84% I'm going to start by finishing the supply of health here, 00:00:27.830 --> 00:00:29.630 align:middle line:84% but then when whatever time is left, 00:00:29.630 --> 00:00:34.070 align:middle line:84% I'll start with the gender and family. 00:00:34.070 --> 00:00:39.320 align:middle line:90% 00:00:39.320 --> 00:00:42.317 align:middle line:84% and that will be basically, probably what 00:00:42.317 --> 00:00:43.900 align:middle line:84% I have time to cover today, so really, 00:00:43.900 --> 00:00:48.280 align:middle line:90% the end of Ben's lecture. 00:00:48.280 --> 00:00:53.110 align:middle line:84% So that's a bit of a hodge podge, mixed mash lecture today 00:00:53.110 --> 00:00:54.760 align:middle line:90% that this is. 00:00:54.760 --> 00:00:58.510 align:middle line:84% So where we left it last time is the idea 00:00:58.510 --> 00:01:02.290 align:middle line:84% that it's difficult for people to learn about health. 00:01:02.290 --> 00:01:08.230 align:middle line:84% Last time, we discussed the fact that self-limiting diseases are 00:01:08.230 --> 00:01:12.370 align:middle line:84% a problem and therefore, people also 00:01:12.370 --> 00:01:15.550 align:middle line:84% don't understand why a particular drug or compound 00:01:15.550 --> 00:01:17.890 align:middle line:84% treats them or doesn't treat them. 00:01:17.890 --> 00:01:21.250 align:middle line:84% In the lecture before that, we talked about externalities 00:01:21.250 --> 00:01:23.260 align:middle line:84% that makes it even harder to understand 00:01:23.260 --> 00:01:25.780 align:middle line:84% the impact of any drug that has extant, which 00:01:25.780 --> 00:01:28.390 align:middle line:84% is most drugs that have something 00:01:28.390 --> 00:01:32.320 align:middle line:90% to do with infectious diseases. 00:01:32.320 --> 00:01:35.110 align:middle line:84% And the consequences of all that is that, indeed, 00:01:35.110 --> 00:01:39.670 align:middle line:84% the private health care, as Arrow intuited, 00:01:39.670 --> 00:01:44.320 align:middle line:84% could be of a very, very, very bad quality. 00:01:44.320 --> 00:01:46.540 align:middle line:84% Just to give you a sense of how bad it is, 00:01:46.540 --> 00:01:51.520 align:middle line:84% so in India, Jishnu Das and a group of people 00:01:51.520 --> 00:01:54.430 align:middle line:84% did a survey in several states, which we 00:01:54.430 --> 00:01:56.350 align:middle line:90% are going to use a lot today. 00:01:56.350 --> 00:01:59.200 align:middle line:84% And they found that 77% of private providers 00:01:59.200 --> 00:02:03.820 align:middle line:84% in rural areas have no medical degree whatsoever. 00:02:03.820 --> 00:02:07.300 align:middle line:84% 18% of them have some other degree than a medical degree. 00:02:07.300 --> 00:02:09.820 align:middle line:90% 00:02:09.820 --> 00:02:12.850 align:middle line:84% So 18% have a medical degree, but that's not 00:02:12.850 --> 00:02:16.120 align:middle line:84% like your regular med school, which 00:02:16.120 --> 00:02:18.670 align:middle line:84% is MBBS in India, the equivalent of an MD. 00:02:18.670 --> 00:02:21.730 align:middle line:84% The rest have these BAMS, BIMS, BUMS, BHMS, 00:02:21.730 --> 00:02:24.430 align:middle line:84% which are some kind of alternative medicine 00:02:24.430 --> 00:02:28.010 align:middle line:90% or things like that. 00:02:28.010 --> 00:02:32.690 align:middle line:84% In the average village, there are 3.3 providers 00:02:32.690 --> 00:02:38.960 align:middle line:84% with no degree, less than 1 with some degree, and only 20% 00:02:38.960 --> 00:02:42.510 align:middle line:84% of the village have even one MBBS doctors. 00:02:42.510 --> 00:02:44.880 align:middle line:84% They are public providers who are more qualified 00:02:44.880 --> 00:02:47.670 align:middle line:84% and offer free services, but they have only 20% 00:02:47.670 --> 00:02:52.430 align:middle line:84% of the market share, up to 25% in villages 00:02:52.430 --> 00:02:57.270 align:middle line:84% where the primary health center is exactly located. 00:02:57.270 --> 00:03:00.240 align:middle line:84% Another thing is these private doctors 00:03:00.240 --> 00:03:04.260 align:middle line:84% that have 80% of the market share, know very little, 00:03:04.260 --> 00:03:07.470 align:middle line:84% and to establish that, they use vignettes, which I'll describe 00:03:07.470 --> 00:03:10.455 align:middle line:84% in a moment, and they do even less, 00:03:10.455 --> 00:03:12.580 align:middle line:84% so there is what's called in the medical literature 00:03:12.580 --> 00:03:15.700 align:middle line:84% and know-do gap, which is the gap between what you know 00:03:15.700 --> 00:03:17.360 align:middle line:90% and what you do. 00:03:17.360 --> 00:03:19.840 align:middle line:84% So how do we know what doctors know? 00:03:19.840 --> 00:03:23.980 align:middle line:84% So the way to do this that was pioneered 00:03:23.980 --> 00:03:28.030 align:middle line:84% by Jishnu Das and his group of colleagues 00:03:28.030 --> 00:03:30.700 align:middle line:84% is to use vignettes, very similar to what gets 00:03:30.700 --> 00:03:32.170 align:middle line:90% administered in board exams. 00:03:32.170 --> 00:03:35.530 align:middle line:84% If you have friends who took board exams for medicine, 00:03:35.530 --> 00:03:39.515 align:middle line:84% you are presented with some information about a patient. 00:03:39.515 --> 00:03:41.140 align:middle line:84% The doctor is told that the patient who 00:03:41.140 --> 00:03:44.500 align:middle line:84% will follow instruction, they are given the standard tests, 00:03:44.500 --> 00:03:48.820 align:middle line:84% and then you pose as a doctor to ask questions. 00:03:48.820 --> 00:03:50.560 align:middle line:84% And then you're graded on whether you 00:03:50.560 --> 00:03:51.730 align:middle line:90% asked the right questions. 00:03:51.730 --> 00:03:54.250 align:middle line:84% And then you give an answer and you ask further questions. 00:03:54.250 --> 00:03:57.950 align:middle line:84% You come up with a diagnosis at the end. 00:03:57.950 --> 00:03:59.660 align:middle line:84% So you're also graded, of course, 00:03:59.660 --> 00:04:01.380 align:middle line:84% whether you got to the right diagnosis. 00:04:01.380 --> 00:04:03.620 align:middle line:84% So here a woman, for example, a woman 00:04:03.620 --> 00:04:05.930 align:middle line:84% comes to the clinic with a child who has had 00:04:05.930 --> 00:04:08.250 align:middle line:90% diarrhea for the last two days. 00:04:08.250 --> 00:04:11.900 align:middle line:84% So the doctor must first ask about their stool, not stock, 00:04:11.900 --> 00:04:14.990 align:middle line:84% to figure out if she has dysentery, rather than a virus 00:04:14.990 --> 00:04:17.329 align:middle line:84% and check for signs of dehydration. 00:04:17.329 --> 00:04:21.140 align:middle line:84% In this particular vignette, it is not a bacteria, 00:04:21.140 --> 00:04:22.250 align:middle line:90% it is a virus. 00:04:22.250 --> 00:04:24.950 align:middle line:90% There is no need for antibiotic. 00:04:24.950 --> 00:04:28.460 align:middle line:84% The right treatment is oral rehydration solution. 00:04:28.460 --> 00:04:31.630 align:middle line:90% So they use that vignette. 00:04:31.630 --> 00:04:32.560 align:middle line:90% They've been used. 00:04:32.560 --> 00:04:37.270 align:middle line:84% So that's Hammer and Leohnardt have pioneered 00:04:37.270 --> 00:04:39.490 align:middle line:84% the use of the vignettes and used them 00:04:39.490 --> 00:04:40.750 align:middle line:90% in a number of settings. 00:04:40.750 --> 00:04:43.572 align:middle line:84% They have an overview of their work 00:04:43.572 --> 00:04:45.280 align:middle line:84% in their journal of economics perspective 00:04:45.280 --> 00:04:47.060 align:middle line:90% from a few years ago. 00:04:47.060 --> 00:04:50.680 align:middle line:84% So what they do then is, they use item response theory 00:04:50.680 --> 00:04:53.470 align:middle line:84% to code the answer to the question 00:04:53.470 --> 00:04:56.450 align:middle line:84% to create an index of competence, 00:04:56.450 --> 00:04:59.560 align:middle line:84% which solve for the optimal weight to each question. 00:04:59.560 --> 00:05:01.420 align:middle line:84% So basically, you weight less the question 00:05:01.420 --> 00:05:02.980 align:middle line:84% that everybody gets right anyway. 00:05:02.980 --> 00:05:05.530 align:middle line:84% You get more the questions that are more weighted. 00:05:05.530 --> 00:05:09.080 align:middle line:84% This is something, a technique that's coming from education. 00:05:09.080 --> 00:05:11.630 align:middle line:84% So for example, when you take the GRE, 00:05:11.630 --> 00:05:14.890 align:middle line:84% then graded according to item response theory, 00:05:14.890 --> 00:05:19.540 align:middle line:84% and then that's what is used to form your percentile ranks as 00:05:19.540 --> 00:05:22.698 align:middle line:90% opposed to just the whole score. 00:05:22.698 --> 00:05:24.490 align:middle line:84% So that's what they do here, so they create 00:05:24.490 --> 00:05:27.290 align:middle line:90% this index of competence. 00:05:27.290 --> 00:05:31.330 align:middle line:84% And so here people are graded by competence, 00:05:31.330 --> 00:05:36.160 align:middle line:84% and you can see that if we rank people by competence, then 00:05:36.160 --> 00:05:40.120 align:middle line:90% we can start asking questions. 00:05:40.120 --> 00:05:42.790 align:middle line:84% You can then ask descriptive question 00:05:42.790 --> 00:05:44.990 align:middle line:90% for people ranked by competence. 00:05:44.990 --> 00:05:48.340 align:middle line:84% So it's saying, for example, that among the most competent 00:05:48.340 --> 00:05:55.690 align:middle line:84% doctors, for the diarrhea question, only 50% of them 00:05:55.690 --> 00:05:57.610 align:middle line:84% asked about the stools, whether there 00:05:57.610 --> 00:05:59.482 align:middle line:84% was blood or mucus in the stool, which 00:05:59.482 --> 00:06:00.940 align:middle line:84% would have helped them to determine 00:06:00.940 --> 00:06:02.140 align:middle line:90% if there is disenteria. 00:06:02.140 --> 00:06:06.160 align:middle line:84% And among the least competent, basically, none of them, 00:06:06.160 --> 00:06:08.380 align:middle line:84% a few percentage points even asked the question. 00:06:08.380 --> 00:06:10.850 align:middle line:90% So that's an example. 00:06:10.850 --> 00:06:12.640 align:middle line:84% In some countries, doctors are better. 00:06:12.640 --> 00:06:15.370 align:middle line:84% So for example, in Indonesia, they ask-- 00:06:15.370 --> 00:06:18.350 align:middle line:90% so it depends on the thing. 00:06:18.350 --> 00:06:23.350 align:middle line:84% So check the fontenelle for dehydration, 00:06:23.350 --> 00:06:25.600 align:middle line:84% in India, no one asked that question. 00:06:25.600 --> 00:06:28.210 align:middle line:84% In Tanzania, they ask it more often. 00:06:28.210 --> 00:06:30.640 align:middle line:84% Whether there was fever, the question that asked enough, 00:06:30.640 --> 00:06:34.030 align:middle line:84% the frequency of the stool is a question that asks more. 00:06:34.030 --> 00:06:36.590 align:middle line:90% 00:06:36.590 --> 00:06:41.030 align:middle line:84% For tuberculosis, do they ask the right questions? 00:06:41.030 --> 00:06:44.030 align:middle line:84% Fraction of the doctors that asked the right question 00:06:44.030 --> 00:06:47.570 align:middle line:84% in vignette, for the good doctors, it's very high. 00:06:47.570 --> 00:06:50.090 align:middle line:90% For anybody else, it's very low. 00:06:50.090 --> 00:06:52.100 align:middle line:90% That's for Indonesia. 00:06:52.100 --> 00:06:56.120 align:middle line:84% And I forgot to put what the question was, 00:06:56.120 --> 00:07:01.280 align:middle line:84% question A, question B was whether it was asked or not so. 00:07:01.280 --> 00:07:03.680 align:middle line:84% There is a lot of variation, but the overall quality 00:07:03.680 --> 00:07:04.470 align:middle line:90% is very low. 00:07:04.470 --> 00:07:07.160 align:middle line:84% So what that says is that the doctor has 00:07:07.160 --> 00:07:08.870 align:middle line:84% to be more than median quality not 00:07:08.870 --> 00:07:11.010 align:middle line:84% to harm the patient with what they are doing. 00:07:11.010 --> 00:07:16.400 align:middle line:84% So if you remember the oath that the doctor 00:07:16.400 --> 00:07:20.390 align:middle line:84% take at the beginning of their career, do no harm, 00:07:20.390 --> 00:07:23.340 align:middle line:90% it's really not the case. 00:07:23.340 --> 00:07:28.530 align:middle line:84% In India, when you compare where people 00:07:28.530 --> 00:07:31.890 align:middle line:84% rank in terms of performance, from the vignette point 00:07:31.890 --> 00:07:35.580 align:middle line:84% of view, the top qualified doctors 00:07:35.580 --> 00:07:37.200 align:middle line:84% are better than the public doctors, 00:07:37.200 --> 00:07:38.910 align:middle line:84% but the public doctors are much better 00:07:38.910 --> 00:07:42.700 align:middle line:84% than the quacks, the ones who don't have a diploma. 00:07:42.700 --> 00:07:44.460 align:middle line:84% So you would expect people should 00:07:44.460 --> 00:07:47.430 align:middle line:84% come to the public doctor, when there is one available. 00:07:47.430 --> 00:07:51.980 align:middle line:84% It is both free, and they are more competent. 00:07:51.980 --> 00:07:53.280 align:middle line:90% But yet they don't. 00:07:53.280 --> 00:07:54.930 align:middle line:90% And so why is that the case? 00:07:54.930 --> 00:08:00.210 align:middle line:84% And that's related, probably to what they do once you show up. 00:08:00.210 --> 00:08:01.480 align:middle line:90% So how to find out. 00:08:01.480 --> 00:08:03.720 align:middle line:84% So the first method that they did 00:08:03.720 --> 00:08:07.410 align:middle line:84% is sent-- just leave someone in the clinic for some time 00:08:07.410 --> 00:08:12.090 align:middle line:84% and do direct observation of patients as they come. 00:08:12.090 --> 00:08:16.260 align:middle line:84% So it has the issue that people, the doctors might adjust 00:08:16.260 --> 00:08:18.780 align:middle line:84% their behavior to the fact that they 00:08:18.780 --> 00:08:21.240 align:middle line:90% have someone observing them. 00:08:21.240 --> 00:08:23.280 align:middle line:84% That doesn't seem to be much of a problem 00:08:23.280 --> 00:08:26.610 align:middle line:84% from descriptive data I'm going to show you in a minute. 00:08:26.610 --> 00:08:28.590 align:middle line:84% Even with someone present to observe them, 00:08:28.590 --> 00:08:30.790 align:middle line:84% they spend about 2 minutes with the patient. 00:08:30.790 --> 00:08:33.990 align:middle line:84% So if they do even less when they don't have someone, 00:08:33.990 --> 00:08:35.679 align:middle line:90% then maybe it's even less. 00:08:35.679 --> 00:08:41.640 align:middle line:84% But even with that, even with someone observing them 00:08:41.640 --> 00:08:42.990 align:middle line:90% that don't seem to be. 00:08:42.990 --> 00:08:45.510 align:middle line:84% Either their best behavior is truly terrible, indeed, 00:08:45.510 --> 00:08:49.630 align:middle line:84% or they are not really on their best behavior. 00:08:49.630 --> 00:08:52.000 align:middle line:84% But there is another problem with this, 00:08:52.000 --> 00:08:56.430 align:middle line:84% which is that if you have someone sitting there observing 00:08:56.430 --> 00:09:00.760 align:middle line:84% the patient, that person doesn't know what the patient has. 00:09:00.760 --> 00:09:03.480 align:middle line:84% So they don't know what the right course of action 00:09:03.480 --> 00:09:04.710 align:middle line:90% would be. 00:09:04.710 --> 00:09:10.430 align:middle line:84% And in addition, good doctors might attract 00:09:10.430 --> 00:09:12.660 align:middle line:90% a different mix of patients. 00:09:12.660 --> 00:09:15.110 align:middle line:84% So when you are sitting there observing what is going on, 00:09:15.110 --> 00:09:19.660 align:middle line:84% you're also sitting there, observing 00:09:19.660 --> 00:09:22.930 align:middle line:84% an endogenous response to an equilibrium where people have 00:09:22.930 --> 00:09:29.430 align:middle line:84% gone to that particular patient in ways that are, essentially, 00:09:29.430 --> 00:09:32.140 align:middle line:84% very, very difficult to control for. 00:09:32.140 --> 00:09:35.310 align:middle line:84% So to address these issues, all of these issues, potential 00:09:35.310 --> 00:09:37.727 align:middle line:84% of social desirability, not knowing 00:09:37.727 --> 00:09:39.810 align:middle line:84% what the right course of action because you're not 00:09:39.810 --> 00:09:41.352 align:middle line:84% a doctor, you don't know the patient, 00:09:41.352 --> 00:09:44.760 align:middle line:84% and the potential endogenous sorting of patients that 00:09:44.760 --> 00:09:47.400 align:middle line:84% is hard to deal with, what that did 00:09:47.400 --> 00:09:51.610 align:middle line:84% is that they started sending a standardized patient. 00:09:51.610 --> 00:09:54.390 align:middle line:84% So that's the equivalent of an audit study. 00:09:54.390 --> 00:09:57.180 align:middle line:84% You might have seen this audit study in other fields 00:09:57.180 --> 00:10:01.590 align:middle line:84% like, for example, to know whether car salesmen might 00:10:01.590 --> 00:10:05.040 align:middle line:84% discriminate against Black people or against women, 00:10:05.040 --> 00:10:11.460 align:middle line:84% you can send a woman or a Black person to the used car lot 00:10:11.460 --> 00:10:17.640 align:middle line:84% and have them do negotiation and see whether they are given 00:10:17.640 --> 00:10:20.860 align:middle line:84% less time, they are given higher prices, or so on and so forth. 00:10:20.860 --> 00:10:23.820 align:middle line:84% So that's the idea of audit studies. 00:10:23.820 --> 00:10:26.860 align:middle line:84% Audit studies have been used in various settings, 00:10:26.860 --> 00:10:28.360 align:middle line:90% including the labor market. 00:10:28.360 --> 00:10:35.010 align:middle line:84% And so here, the idea is to send a standardized patient who 00:10:35.010 --> 00:10:38.970 align:middle line:84% is afflicted by a particular ailment, 00:10:38.970 --> 00:10:42.228 align:middle line:90% that the advantage to know. 00:10:42.228 --> 00:10:44.520 align:middle line:84% Of course, the disadvantage is that they don't actually 00:10:44.520 --> 00:10:46.050 align:middle line:90% have that ailment. 00:10:46.050 --> 00:10:51.000 align:middle line:84% But goes and described his problems, so doctor, 00:10:51.000 --> 00:10:53.070 align:middle line:84% I woke up this morning with crushing chest pain 00:10:53.070 --> 00:10:56.130 align:middle line:90% and was feeling very anxious. 00:10:56.130 --> 00:10:58.050 align:middle line:84% Then he answers the question, completes 00:10:58.050 --> 00:11:03.510 align:middle line:84% the basic exams that provided the recommend a treatment. 00:11:03.510 --> 00:11:04.760 align:middle line:90% So there are issues with that. 00:11:04.760 --> 00:11:08.160 align:middle line:84% Number one, is maybe it's not ethical to send a fake patient. 00:11:08.160 --> 00:11:10.370 align:middle line:84% So it's generally done in a context, where 00:11:10.370 --> 00:11:12.890 align:middle line:84% you're telling them that they might 00:11:12.890 --> 00:11:17.118 align:middle line:90% be unrolling your doctors. 00:11:17.118 --> 00:11:19.660 align:middle line:84% Tell them that they might be a standard patient at some point 00:11:19.660 --> 00:11:23.650 align:middle line:84% or another, which increases or you reveal yourself at the end, 00:11:23.650 --> 00:11:27.210 align:middle line:84% and you're asking, can I use the data from that. 00:11:27.210 --> 00:11:29.970 align:middle line:84% The detection rates are actually low, despite that, 00:11:29.970 --> 00:11:37.740 align:middle line:84% and provider behavior seems to be consistent with believing 00:11:37.740 --> 00:11:39.030 align:middle line:90% the standardized patient. 00:11:39.030 --> 00:11:41.168 align:middle line:84% And they don't seem to come to the conclusion 00:11:41.168 --> 00:11:42.960 align:middle line:84% that the standardized patient is faking it. 00:11:42.960 --> 00:11:46.590 align:middle line:84% I don't know why all of my inverted commas 00:11:46.590 --> 00:11:47.880 align:middle line:90% became question marks. 00:11:47.880 --> 00:11:50.970 align:middle line:84% But in fact, the more they do with the patient, 00:11:50.970 --> 00:11:53.550 align:middle line:84% the more they are convinced that the standardized patient 00:11:53.550 --> 00:11:55.380 align:middle line:90% is real. 00:11:55.380 --> 00:11:57.370 align:middle line:84% And they do even more compelling, 00:11:57.370 --> 00:11:59.020 align:middle line:84% there is one case of general patient, 00:11:59.020 --> 00:12:00.990 align:middle line:90% which is child diarrhea. 00:12:00.990 --> 00:12:03.270 align:middle line:84% Obviously, you cannot bring a child. 00:12:03.270 --> 00:12:06.420 align:middle line:84% So someone comes to the doctor and said 00:12:06.420 --> 00:12:09.090 align:middle line:84% there is a child in a village that has diarrhea, 00:12:09.090 --> 00:12:12.960 align:middle line:84% and then answers to all the questions 00:12:12.960 --> 00:12:14.080 align:middle line:90% that are given to them. 00:12:14.080 --> 00:12:16.410 align:middle line:90% So did you check the blood? 00:12:16.410 --> 00:12:19.170 align:middle line:84% Did you check the stool to see if there was blood or mucus. 00:12:19.170 --> 00:12:22.540 align:middle line:84% They can give an answer to that, and so on and so forth. 00:12:22.540 --> 00:12:24.577 align:middle line:84% So the nice thing about this method, 00:12:24.577 --> 00:12:26.910 align:middle line:84% I mean, one of the nice things about this method, so you 00:12:26.910 --> 00:12:30.150 align:middle line:84% can combine standardized patient vignettes. 00:12:30.150 --> 00:12:33.878 align:middle line:84% So you get direct information about the know-do gaps. 00:12:33.878 --> 00:12:35.670 align:middle line:84% So you first send the standardized patient, 00:12:35.670 --> 00:12:38.790 align:middle line:84% then you go back to the doctor and say, hey, by the way, 00:12:38.790 --> 00:12:41.700 align:middle line:84% we had sent some standardized patient. 00:12:41.700 --> 00:12:44.400 align:middle line:90% Are OK for us to use this data? 00:12:44.400 --> 00:12:49.770 align:middle line:84% And if they are OK, they are given a vignette to answer. 00:12:49.770 --> 00:12:53.630 align:middle line:90% 00:12:53.630 --> 00:13:00.080 align:middle line:84% So here are some data from the direct observation. 00:13:00.080 --> 00:13:03.200 align:middle line:84% Again, the doctors are classified 00:13:03.200 --> 00:13:05.480 align:middle line:84% by a third of how much effort to exert 00:13:05.480 --> 00:13:09.960 align:middle line:84% between low effort, medium effort, high effort. 00:13:09.960 --> 00:13:14.780 align:middle line:84% So in India, in Delhi, they spend about 3.8 minutes 00:13:14.780 --> 00:13:17.150 align:middle line:90% with the patient. 00:13:17.150 --> 00:13:19.370 align:middle line:90% They ask three questions. 00:13:19.370 --> 00:13:23.730 align:middle line:84% They give one exam, and they give about 2.6, 00:13:23.730 --> 00:13:25.610 align:middle line:84% which we can round up to 3 minutes. 00:13:25.610 --> 00:13:27.260 align:middle line:90% So it's a three, three, three-- 00:13:27.260 --> 00:13:30.760 align:middle line:84% 3 minutes, three questions, three medicines. 00:13:30.760 --> 00:13:33.010 align:middle line:90% It's easy to remember. 00:13:33.010 --> 00:13:37.870 align:middle line:84% In Paraguay, they spend a little bit more time, 8 minutes. 00:13:37.870 --> 00:13:42.640 align:middle line:84% They ask more questions, and they give fewer medicine. 00:13:42.640 --> 00:13:45.070 align:middle line:84% And you can see, again, that the effort 00:13:45.070 --> 00:13:48.820 align:middle line:84% is related to the time that is being spent. 00:13:48.820 --> 00:13:52.480 align:middle line:84% In Germany, people spend closer to Paraguay. 00:13:52.480 --> 00:13:56.740 align:middle line:84% In Belgium, they spend 15 minutes with a patient. 00:13:56.740 --> 00:13:59.830 align:middle line:84% But the number for India are particularly extreme in terms 00:13:59.830 --> 00:14:03.430 align:middle line:90% of how low the effort is. 00:14:03.430 --> 00:14:09.460 align:middle line:84% And then from the standardized patient, 00:14:09.460 --> 00:14:13.520 align:middle line:84% we can look at the type of treatment that is being given. 00:14:13.520 --> 00:14:15.130 align:middle line:84% So this is a standard patient, so you 00:14:15.130 --> 00:14:21.160 align:middle line:84% know what the correct treatment would be of a treatment. 00:14:21.160 --> 00:14:24.070 align:middle line:90% And you can see that in most-- 00:14:24.070 --> 00:14:27.510 align:middle line:84% so this is various states in India and in China and Kenya. 00:14:27.510 --> 00:14:33.030 align:middle line:90% 00:14:33.030 --> 00:14:35.970 align:middle line:84% It's less than half the time, sometimes much less than half 00:14:35.970 --> 00:14:38.910 align:middle line:84% the time that you get the correct treatment 00:14:38.910 --> 00:14:40.980 align:middle line:84% that the standardized patient get 00:14:40.980 --> 00:14:44.700 align:middle line:84% the correct treatment for what they're presented with. 00:14:44.700 --> 00:14:47.520 align:middle line:84% Even in cases like, for example, one of the cases is a woman 00:14:47.520 --> 00:14:49.080 align:middle line:90% with pre-eclampsia-- 00:14:49.080 --> 00:14:51.930 align:middle line:84% that's a pregnant woman who has, essentially, 00:14:51.930 --> 00:14:53.580 align:middle line:84% high blood pressure during pregnancy. 00:14:53.580 --> 00:14:55.170 align:middle line:84% It's a life-threatening condition. 00:14:55.170 --> 00:14:59.620 align:middle line:84% Even there, they don't get the correct treatment. 00:14:59.620 --> 00:15:02.940 align:middle line:84% So any correct treatment is the correct plus some more. 00:15:02.940 --> 00:15:05.560 align:middle line:90% Just to correct is even lower. 00:15:05.560 --> 00:15:08.400 align:middle line:84% A lot of people get overtreated, so they get medicine 00:15:08.400 --> 00:15:11.820 align:middle line:84% that they don't need for this particular case. 00:15:11.820 --> 00:15:13.590 align:middle line:90% Incorrect treatment is current. 00:15:13.590 --> 00:15:17.340 align:middle line:84% One example of a treatment is antibiotic, 00:15:17.340 --> 00:15:19.740 align:middle line:84% so they think all the cases, none of these cases 00:15:19.740 --> 00:15:23.100 align:middle line:84% except possibly, the standard patient acquired antibiotic. 00:15:23.100 --> 00:15:31.470 align:middle line:84% And a lot of the cases, got it anyway, 68% in Patna, 00:15:31.470 --> 00:15:35.370 align:middle line:84% 55% in Kenya, 35% in Madhya Pradesh 00:15:35.370 --> 00:15:38.580 align:middle line:84% got an antibiotic that they didn't need. 00:15:38.580 --> 00:15:42.570 align:middle line:84% The use of steroids is less popular, particularly, 00:15:42.570 --> 00:15:43.380 align:middle line:90% outside of India. 00:15:43.380 --> 00:15:45.810 align:middle line:84% But you can see that in Bombay, they are very popular. 00:15:45.810 --> 00:15:48.030 align:middle line:84% 20% of people who don't need steroid, 00:15:48.030 --> 00:15:51.120 align:middle line:84% the only indication for steroid here was the asthma case. 00:15:51.120 --> 00:15:53.920 align:middle line:90% Go back with steroid. 00:15:53.920 --> 00:15:57.940 align:middle line:84% Referral is rare from this first line doctor, 00:15:57.940 --> 00:16:01.740 align:middle line:84% you're hoping that they would refer, but [INAUDIBLE].. 00:16:01.740 --> 00:16:05.400 align:middle line:84% Another thing that you know from combining this vignette 00:16:05.400 --> 00:16:07.830 align:middle line:84% to the standardized patient is this evidence 00:16:07.830 --> 00:16:10.680 align:middle line:90% of this know-do gap. 00:16:10.680 --> 00:16:14.340 align:middle line:84% So if you use the same vignette, you 00:16:14.340 --> 00:16:17.370 align:middle line:84% can compare what they to do because they are doing it 00:16:17.370 --> 00:16:19.500 align:middle line:84% on the vignette and what they actually 00:16:19.500 --> 00:16:21.720 align:middle line:84% do when the standardized patient is showing up 00:16:21.720 --> 00:16:23.310 align:middle line:90% at their doorstep. 00:16:23.310 --> 00:16:28.170 align:middle line:84% And you can see that there is always a know-do gap, 00:16:28.170 --> 00:16:29.880 align:middle line:84% that these are all the step, that they 00:16:29.880 --> 00:16:34.470 align:middle line:84% do more of the right thing, and they 00:16:34.470 --> 00:16:37.050 align:middle line:84% are more likely to get any correct treatment 00:16:37.050 --> 00:16:41.560 align:middle line:84% on the vignette than on the standardized patient. 00:16:41.560 --> 00:16:44.170 align:middle line:84% And while the knowledge, as we saw before, 00:16:44.170 --> 00:16:46.910 align:middle line:84% increase with qualification, for example, 00:16:46.910 --> 00:16:51.340 align:middle line:84% the MD public, MD private untrained, public guy 00:16:51.340 --> 00:16:53.050 align:middle line:90% know more than the private guy. 00:16:53.050 --> 00:16:55.660 align:middle line:90% 00:16:55.660 --> 00:16:59.320 align:middle line:84% So what they know increases a lot with qualification. 00:16:59.320 --> 00:17:06.730 align:middle line:84% You can see it from a private trained know the most. 00:17:06.730 --> 00:17:09.790 align:middle line:90% Public trained knows a bit less. 00:17:09.790 --> 00:17:13.770 align:middle line:84% Private untrained knows the least. 00:17:13.770 --> 00:17:18.000 align:middle line:84% But what they do, although there is the same U-shape pattern 00:17:18.000 --> 00:17:23.680 align:middle line:84% in this, that U-shape pattern is much flatter. 00:17:23.680 --> 00:17:27.450 align:middle line:84% So the know-do gap increases with qualification. 00:17:27.450 --> 00:17:29.500 align:middle line:84% The know gap increases with qualification, 00:17:29.500 --> 00:17:32.290 align:middle line:84% but the know-do gap also increases. 00:17:32.290 --> 00:17:36.670 align:middle line:84% And in fact, in a lot of cases, at the end of the day, 00:17:36.670 --> 00:17:39.600 align:middle line:84% the public providers don't do much better 00:17:39.600 --> 00:17:42.210 align:middle line:84% than the private provider because although they 00:17:42.210 --> 00:17:46.350 align:middle line:84% know much more, they're even farther from their front here. 00:17:46.350 --> 00:17:49.770 align:middle line:84% So the know-do gap is a thing to keep in mind. 00:17:49.770 --> 00:17:51.730 align:middle line:84% It's not just about training and knowledge. 00:17:51.730 --> 00:17:55.090 align:middle line:84% It's about incentive to actually do what is appropriate. 00:17:55.090 --> 00:17:55.590 align:middle line:90% Yep. 00:17:55.590 --> 00:17:57.517 align:middle line:84% AUDIENCE: When are the vignettes administered? 00:17:57.517 --> 00:17:59.975 align:middle line:84% So are they administered during the course that the doctors 00:17:59.975 --> 00:18:00.945 align:middle line:90% know [INAUDIBLE] day? 00:18:00.945 --> 00:18:03.930 align:middle line:84% Because one potential explanation you could think of 00:18:03.930 --> 00:18:05.970 align:middle line:84% is, I know what the right answer is 00:18:05.970 --> 00:18:09.990 align:middle line:84% in a fully, theoretical setup, but I'm 00:18:09.990 --> 00:18:11.230 align:middle line:90% dealing with many patients. 00:18:11.230 --> 00:18:12.647 align:middle line:84% I have a lot of demand on my time, 00:18:12.647 --> 00:18:14.003 align:middle line:90% so I just might not be able to-- 00:18:14.003 --> 00:18:15.420 align:middle line:84% ESTHER DUFLO: Attend to the thing. 00:18:15.420 --> 00:18:16.280 align:middle line:90% Yeah, it could be. 00:18:16.280 --> 00:18:19.740 align:middle line:84% If you spend 3 minutes, it's as much time as you have. 00:18:19.740 --> 00:18:23.120 align:middle line:84% So that's certainly one of the parameters 00:18:23.120 --> 00:18:27.020 align:middle line:84% is how much effort you're giving to each patient 00:18:27.020 --> 00:18:33.410 align:middle line:84% and to solving the problem that is in front of you. 00:18:33.410 --> 00:18:35.000 align:middle line:84% And the other, which we are going 00:18:35.000 --> 00:18:38.340 align:middle line:84% to spend a bit more time on is also, what are your incentive 00:18:38.340 --> 00:18:40.800 align:middle line:84% and some of how you're going to get paid. 00:18:40.800 --> 00:18:42.710 align:middle line:84% So in particular, in a lot of settings, 00:18:42.710 --> 00:18:46.220 align:middle line:84% you're going to see an over application of various kind 00:18:46.220 --> 00:18:50.060 align:middle line:84% of medicine, and a lot of the payment model 00:18:50.060 --> 00:18:54.410 align:middle line:84% is that people are paid both for a fee for the visit, 00:18:54.410 --> 00:18:57.240 align:middle line:84% and then they are selling the medicine at the same time. 00:18:57.240 --> 00:18:59.420 align:middle line:84% So there is also an incentive that way, 00:18:59.420 --> 00:19:01.817 align:middle line:84% and that's true of the private doctor and also, 00:19:01.817 --> 00:19:03.650 align:middle line:84% of the public doctor, who actually-- and pay 00:19:03.650 --> 00:19:08.000 align:middle line:84% the fee for the visit, but can sell medicine on the side. 00:19:08.000 --> 00:19:12.602 align:middle line:84% AUDIENCE: In terms of the patient, [INAUDIBLE] 00:19:12.602 --> 00:19:15.410 align:middle line:84% how they're going to differentiate between all 00:19:15.410 --> 00:19:17.945 align:middle line:90% the patients or providers. 00:19:17.945 --> 00:19:21.770 align:middle line:84% ESTHER DUFLO: Yes, so the patient 00:19:21.770 --> 00:19:27.350 align:middle line:84% tend to know that the private provider are less qualified 00:19:27.350 --> 00:19:29.130 align:middle line:90% than the public provider. 00:19:29.130 --> 00:19:31.610 align:middle line:84% So that's the mystery of why they go anyways. 00:19:31.610 --> 00:19:34.440 align:middle line:84% And one of the resolution of that mystery, 00:19:34.440 --> 00:19:36.800 align:middle line:84% it might be there, which is although they are more 00:19:36.800 --> 00:19:39.140 align:middle line:84% qualified and they know more, they actually 00:19:39.140 --> 00:19:43.418 align:middle line:84% don't perform better in front of real patients. 00:19:43.418 --> 00:19:45.820 align:middle line:84% AUDIENCE: This training they have, like one 00:19:45.820 --> 00:19:48.520 align:middle line:84% of these several degrees or I guess 00:19:48.520 --> 00:19:50.020 align:middle line:90% doing it, what is the best? 00:19:50.020 --> 00:19:54.250 align:middle line:84% Do we know what MBBS doctors do versus what is the max? 00:19:54.250 --> 00:19:55.713 align:middle line:90% ESTHER DUFLO: That's MBBS. 00:19:55.713 --> 00:19:56.380 align:middle line:90% That's the best. 00:19:56.380 --> 00:19:58.090 align:middle line:90% AUDIENCE: Or how do they do? 00:19:58.090 --> 00:20:00.330 align:middle line:90% What is the max do correctness? 00:20:00.330 --> 00:20:02.918 align:middle line:90% 00:20:02.918 --> 00:20:03.710 align:middle line:90% ESTHER DUFLO: Yeah. 00:20:03.710 --> 00:20:08.080 align:middle line:84% So in a sense, you have it here ranked by-- so you would want 00:20:08.080 --> 00:20:10.720 align:middle line:90% to see-- 00:20:10.720 --> 00:20:13.730 align:middle line:84% we kind of have it here in the vignette. 00:20:13.730 --> 00:20:16.070 align:middle line:84% This is people ranked by performance, 00:20:16.070 --> 00:20:19.470 align:middle line:84% not ranked by MBBS, but ranked by performance. 00:20:19.470 --> 00:20:22.190 align:middle line:84% And you can see that, for example, 00:20:22.190 --> 00:20:24.770 align:middle line:84% it depends on the question, so they are 00:20:24.770 --> 00:20:26.640 align:middle line:90% ranked by how good they are. 00:20:26.640 --> 00:20:30.320 align:middle line:84% And so the maximum for asking questions in the vignette 00:20:30.320 --> 00:20:36.600 align:middle line:84% is 45% for blood, and fontanelle is very low. 00:20:36.600 --> 00:20:43.670 align:middle line:84% So what you want is a summary of the correct treatment 00:20:43.670 --> 00:20:45.470 align:middle line:84% by ranking of standardized patient 00:20:45.470 --> 00:20:50.630 align:middle line:84% or by how good do they get with the standardized patient. 00:20:50.630 --> 00:20:52.340 align:middle line:84% They have an answer to this question. 00:20:52.340 --> 00:20:53.190 align:middle line:90% That's in the paper. 00:20:53.190 --> 00:20:54.630 align:middle line:84% I don't have it on the top of my mind, 00:20:54.630 --> 00:20:55.755 align:middle line:90% but that's a good question. 00:20:55.755 --> 00:20:59.600 align:middle line:84% That's exactly what you do with this item response theory is 00:20:59.600 --> 00:21:01.340 align:middle line:90% trying to make them better as-- 00:21:01.340 --> 00:21:03.320 align:middle line:84% ranked them by how good they are and saying, 00:21:03.320 --> 00:21:04.505 align:middle line:90% how good is the best doctor? 00:21:04.505 --> 00:21:07.560 align:middle line:90% 00:21:07.560 --> 00:21:09.840 align:middle line:84% And that's where they have a summary of this. 00:21:09.840 --> 00:21:12.195 align:middle line:84% You need to be at least median to not hurt the patient. 00:21:12.195 --> 00:21:16.100 align:middle line:90% 00:21:16.100 --> 00:21:19.150 align:middle line:84% So most people choose the private sector, 00:21:19.150 --> 00:21:23.080 align:middle line:84% despite that incorrect treatment cost the patient 00:21:23.080 --> 00:21:24.880 align:middle line:90% a lot of money. 00:21:24.880 --> 00:21:30.580 align:middle line:84% 70% of the cost of a visit in this study 00:21:30.580 --> 00:21:32.260 align:middle line:84% is the unnecessary treatment, the fact 00:21:32.260 --> 00:21:34.677 align:middle line:84% that they have to buy all these medicines they don't need. 00:21:34.677 --> 00:21:37.100 align:middle line:90% 00:21:37.100 --> 00:21:40.610 align:middle line:84% In Udaipur, where people are super, super poor, where 00:21:40.610 --> 00:21:43.430 align:middle line:84% we had done our health survey, 20% of the visits 00:21:43.430 --> 00:21:45.410 align:middle line:84% are to the public sector, which is free; 00:21:45.410 --> 00:21:49.850 align:middle line:84% 28% to the traditional healer, and the rest to the quacks. 00:21:49.850 --> 00:21:55.240 align:middle line:84% In Madhya Pradesh, 89% of visits are to a private doctor; 83%, 00:21:55.240 --> 00:21:59.230 align:middle line:84% even if there is a trained MBBS, public doctor 00:21:59.230 --> 00:22:01.040 align:middle line:90% right in the village. 00:22:01.040 --> 00:22:05.360 align:middle line:84% So even with the MBBS public doctor, 00:22:05.360 --> 00:22:07.160 align:middle line:84% which is free in the village, people 00:22:07.160 --> 00:22:10.250 align:middle line:84% still go to a private doctor, including 00:22:10.250 --> 00:22:14.610 align:middle line:84% 77% to the unqualified private provider. 00:22:14.610 --> 00:22:21.540 align:middle line:84% And so the idea of that the health sector 00:22:21.540 --> 00:22:23.190 align:middle line:84% should be mostly public because it's 00:22:23.190 --> 00:22:27.150 align:middle line:84% impossible to give quality, the first half is true. 00:22:27.150 --> 00:22:30.330 align:middle line:84% It's impossible to give quality, but the second is not true 00:22:30.330 --> 00:22:37.240 align:middle line:84% because most of the health care is delivered by private people. 00:22:37.240 --> 00:22:38.070 align:middle line:90% So why is that? 00:22:38.070 --> 00:22:41.130 align:middle line:90% 00:22:41.130 --> 00:22:46.770 align:middle line:84% So in Madhya Pradesh, they did this effort 00:22:46.770 --> 00:22:52.230 align:middle line:84% of standardized patient, trying to represent the symptoms 00:22:52.230 --> 00:22:53.352 align:middle line:90% for three diseases. 00:22:53.352 --> 00:22:54.810 align:middle line:84% Those are the angina thing, they're 00:22:54.810 --> 00:22:56.820 align:middle line:84% like the heart attack, the asthma, and dysentery 00:22:56.820 --> 00:23:00.190 align:middle line:90% in a child, who is not there. 00:23:00.190 --> 00:23:04.230 align:middle line:84% Then they have a sample of about 1,000 visits, 00:23:04.230 --> 00:23:08.220 align:middle line:84% and they have public doctors who are moonlighting 00:23:08.220 --> 00:23:09.420 align:middle line:90% in the private sector. 00:23:09.420 --> 00:23:12.750 align:middle line:84% So they can put a fixed effect for a sample that 00:23:12.750 --> 00:23:16.410 align:middle line:84% is both public and private and see whether they 00:23:16.410 --> 00:23:17.980 align:middle line:90% perform differently. 00:23:17.980 --> 00:23:23.940 align:middle line:84% And that gives some idea of how good, why it happened. 00:23:23.940 --> 00:23:26.880 align:middle line:84% So it's true in Madhya Pradesh, like in the data 00:23:26.880 --> 00:23:30.000 align:middle line:84% I showed you, that the public sector 00:23:30.000 --> 00:23:32.580 align:middle line:90% doctors are more likely to be-- 00:23:32.580 --> 00:23:37.660 align:middle line:84% doctor providers are more likely to have an MBBS degree. 00:23:37.660 --> 00:23:41.620 align:middle line:84% This is a sample of public MBBS provider, that's 00:23:41.620 --> 00:23:43.810 align:middle line:84% their fixed point sample, so of course, all of them 00:23:43.810 --> 00:23:44.680 align:middle line:90% have an MBBS. 00:23:44.680 --> 00:23:47.590 align:middle line:84% But in the sample at large, a quarter 00:23:47.590 --> 00:23:51.490 align:middle line:84% of the public and less than 10% of the private providers 00:23:51.490 --> 00:23:53.880 align:middle line:90% have a degree. 00:23:53.880 --> 00:23:55.030 align:middle line:90% They know a bit more. 00:23:55.030 --> 00:23:57.720 align:middle line:84% This is from the vignettes, they know a bit more, 00:23:57.720 --> 00:23:59.310 align:middle line:90% but they do even less. 00:23:59.310 --> 00:24:03.300 align:middle line:84% So this is the number of time they spent with the doctor. 00:24:03.300 --> 00:24:07.260 align:middle line:84% In the representative sample, they spent-- 00:24:07.260 --> 00:24:09.930 align:middle line:84% the private [INAUDIBLE] regression or private provider 00:24:09.930 --> 00:24:13.180 align:middle line:90% compared to public provider. 00:24:13.180 --> 00:24:16.450 align:middle line:84% They spend about 1.2 minutes, including case market district 00:24:16.450 --> 00:24:20.050 align:middle line:84% fixed effect 1.5 minutes more with the patient 00:24:20.050 --> 00:24:28.420 align:middle line:84% on a basis of a mean of 3.6% in the public. 00:24:28.420 --> 00:24:31.930 align:middle line:84% They complete more, the private doctor complete more, 00:24:31.930 --> 00:24:33.770 align:middle line:90% the checklist items. 00:24:33.770 --> 00:24:38.990 align:middle line:84% So actually, even though they know less, they do more. 00:24:38.990 --> 00:24:41.560 align:middle line:84% And so their IFT score, means how good 00:24:41.560 --> 00:24:44.530 align:middle line:84% is their total management of this case in terms 00:24:44.530 --> 00:24:46.690 align:middle line:84% of asking the question is higher, 00:24:46.690 --> 00:24:49.930 align:middle line:84% even though, again, they are less likely to be qualified 00:24:49.930 --> 00:24:51.280 align:middle line:90% and they know more. 00:24:51.280 --> 00:24:53.830 align:middle line:84% That's true in the sample at large, 00:24:53.830 --> 00:24:57.610 align:middle line:84% and that's true with very similar numbers, in fact, when 00:24:57.610 --> 00:25:01.720 align:middle line:84% you compare among MBBS doctors who are all 00:25:01.720 --> 00:25:06.640 align:middle line:84% in the public sector, their behavior in private and public, 00:25:06.640 --> 00:25:10.850 align:middle line:84% which reflects this mix of incentives, time spent, 00:25:10.850 --> 00:25:13.890 align:middle line:90% and so on. 00:25:13.890 --> 00:25:16.940 align:middle line:84% So that's a big underestimated of the difference 00:25:16.940 --> 00:25:19.610 align:middle line:84% in the service that someone can explain because this 00:25:19.610 --> 00:25:21.950 align:middle line:90% is conditional of being found. 00:25:21.950 --> 00:25:25.970 align:middle line:84% And in a lot of cases, the doctor is not even there, 00:25:25.970 --> 00:25:28.890 align:middle line:84% or the public facility is closed. 00:25:28.890 --> 00:25:32.940 align:middle line:84% As a result, they don't treat any better. 00:25:32.940 --> 00:25:38.040 align:middle line:84% If we compare whether the public people give 00:25:38.040 --> 00:25:41.310 align:middle line:84% the correct diagnosis or not, well, the private doctors 00:25:41.310 --> 00:25:46.740 align:middle line:84% are more likely to give a diagnosis. 00:25:46.740 --> 00:25:51.690 align:middle line:84% And then condition and diagnosis, they are no worse. 00:25:51.690 --> 00:25:53.950 align:middle line:90% The private are no worse. 00:25:53.950 --> 00:25:56.230 align:middle line:84% And then there is no difference at all 00:25:56.230 --> 00:26:02.400 align:middle line:84% when we compare the behavior in the fixed effect sample. 00:26:02.400 --> 00:26:05.960 align:middle line:84% So even though, again, so that would explain 00:26:05.960 --> 00:26:08.610 align:middle line:90% why the people go to them. 00:26:08.610 --> 00:26:12.060 align:middle line:84% There is one difference between them, 00:26:12.060 --> 00:26:18.710 align:middle line:84% which is, it's that the private sector tend to avoid Medicaid. 00:26:18.710 --> 00:26:23.960 align:middle line:84% The private doctors tend to give more medicine, about one 00:26:23.960 --> 00:26:26.870 align:middle line:84% more, on average, on a mean of 3. 00:26:26.870 --> 00:26:31.580 align:middle line:84% Although that's not true in the dual practice sample. 00:26:31.580 --> 00:26:34.643 align:middle line:90% So they don't do worse. 00:26:34.643 --> 00:26:35.810 align:middle line:90% They don't do better either. 00:26:35.810 --> 00:26:37.393 align:middle line:84% The private guys, they don't do worse, 00:26:37.393 --> 00:26:41.140 align:middle line:84% then they give a bit more medicine. 00:26:41.140 --> 00:26:43.390 align:middle line:84% So that would explain why people might continue 00:26:43.390 --> 00:26:46.390 align:middle line:84% going to the private sector, despite the fact 00:26:46.390 --> 00:26:48.220 align:middle line:84% that the private sector is not very good. 00:26:48.220 --> 00:26:49.720 align:middle line:90% They are treated better. 00:26:49.720 --> 00:26:52.600 align:middle line:84% The doctor will spend a few more minutes with him, 00:26:52.600 --> 00:26:54.250 align:middle line:90% ask them a bit more question. 00:26:54.250 --> 00:26:57.340 align:middle line:84% At the end of the day, they don't treat them 00:26:57.340 --> 00:26:58.510 align:middle line:90% in a worse way. 00:26:58.510 --> 00:27:00.610 align:middle line:84% I mean, the treatment is bad, but it's bad also, 00:27:00.610 --> 00:27:01.850 align:middle line:90% in the public sector. 00:27:01.850 --> 00:27:04.463 align:middle line:84% And they give them some more medicine to buy, 00:27:04.463 --> 00:27:06.130 align:middle line:84% which is a waste from our point of view, 00:27:06.130 --> 00:27:09.790 align:middle line:84% but it's something that there is a demand for, for the reason 00:27:09.790 --> 00:27:11.050 align:middle line:90% we explained. 00:27:11.050 --> 00:27:15.160 align:middle line:84% People feel good if they can do something 00:27:15.160 --> 00:27:19.090 align:middle line:84% against their ailment, even if something in question 00:27:19.090 --> 00:27:23.170 align:middle line:90% is not really necessary. 00:27:23.170 --> 00:27:24.240 align:middle line:90% So that's where we are. 00:27:24.240 --> 00:27:26.850 align:middle line:84% So compared to the Arrow situation, 00:27:26.850 --> 00:27:29.902 align:middle line:84% where you shouldn't have a private sector, 00:27:29.902 --> 00:27:32.360 align:middle line:84% you should have public sectors, because that's the only way 00:27:32.360 --> 00:27:37.460 align:middle line:84% to get good, quality medicine, or we have an awful quality, 00:27:37.460 --> 00:27:39.440 align:middle line:84% mostly delivered by the private sector. 00:27:39.440 --> 00:27:43.460 align:middle line:84% But whatever the public sector delivers is not much better. 00:27:43.460 --> 00:27:48.000 align:middle line:84% So what to do in front of this situation? 00:27:48.000 --> 00:27:50.600 align:middle line:84% So this is a place where, compared 00:27:50.600 --> 00:27:53.420 align:middle line:84% to education where when I got to this point 00:27:53.420 --> 00:27:56.960 align:middle line:84% on the supply of education, I was 00:27:56.960 --> 00:27:59.180 align:middle line:84% able to share with you at least an idea of what's 00:27:59.180 --> 00:28:00.810 align:middle line:84% the main problem, what to do about it. 00:28:00.810 --> 00:28:04.280 align:middle line:84% I think this is a situation where in health, we 00:28:04.280 --> 00:28:09.410 align:middle line:84% haven't reached this point yet, where many things have been 00:28:09.410 --> 00:28:12.050 align:middle line:84% tried, which I'm going to briefly show you 00:28:12.050 --> 00:28:14.240 align:middle line:84% in each category, but there is, I 00:28:14.240 --> 00:28:17.540 align:middle line:84% think, much less of a consensus that anything, in particular, 00:28:17.540 --> 00:28:19.430 align:middle line:84% is a good thing or a bad thing or has been 00:28:19.430 --> 00:28:22.850 align:middle line:90% successful or not successful. 00:28:22.850 --> 00:28:26.970 align:middle line:84% So that makes it exciting for a research topic, 00:28:26.970 --> 00:28:31.010 align:middle line:84% especially post COVID, where the health situation is even more 00:28:31.010 --> 00:28:33.980 align:middle line:84% important and urgent, interesting, 00:28:33.980 --> 00:28:37.170 align:middle line:84% and therefore, it's worth spending some time on it. 00:28:37.170 --> 00:28:38.360 align:middle line:90% So what can we do? 00:28:38.360 --> 00:28:43.280 align:middle line:84% So option 1, which she was pushed by a colleague, Larry 00:28:43.280 --> 00:28:44.870 align:middle line:84% Summers, for some time, although I 00:28:44.870 --> 00:28:49.040 align:middle line:84% think he has somewhat moved on from that, is more of the same. 00:28:49.040 --> 00:28:51.080 align:middle line:90% Universal health care. 00:28:51.080 --> 00:28:53.090 align:middle line:90% Worldwide universal health care. 00:28:53.090 --> 00:28:55.940 align:middle line:84% So basically, spend more budget, make 00:28:55.940 --> 00:28:57.950 align:middle line:84% sure that each country have more budget 00:28:57.950 --> 00:28:59.900 align:middle line:90% to spend on the public sector. 00:28:59.900 --> 00:29:02.510 align:middle line:90% That seems like a terrible idea. 00:29:02.510 --> 00:29:07.430 align:middle line:84% In the current situation, where absenteeism in public health 00:29:07.430 --> 00:29:11.300 align:middle line:84% facilities is 50%, even when you can get hold of someone, 00:29:11.300 --> 00:29:13.310 align:middle line:90% the treatment is there. 00:29:13.310 --> 00:29:16.610 align:middle line:84% It seems that pouring more money into the same system 00:29:16.610 --> 00:29:19.852 align:middle line:84% is unlikely to be super successful. 00:29:19.852 --> 00:29:21.560 align:middle line:84% I'm not going to spend much time on that. 00:29:21.560 --> 00:29:27.350 align:middle line:84% Although, that's in a sense, the status quo proposal 00:29:27.350 --> 00:29:32.730 align:middle line:84% in the international world is we just need to spend more money. 00:29:32.730 --> 00:29:33.990 align:middle line:90% In fact, India did that. 00:29:33.990 --> 00:29:37.430 align:middle line:84% They went from spending about 2% of GDP 00:29:37.430 --> 00:29:40.520 align:middle line:84% on public health care to spending 10% of GDP 00:29:40.520 --> 00:29:44.420 align:middle line:84% with very little to show for it in terms of health care 00:29:44.420 --> 00:29:45.740 align:middle line:90% quality. 00:29:45.740 --> 00:29:48.535 align:middle line:90% So we'll move that aside. 00:29:48.535 --> 00:29:50.660 align:middle line:84% Then you could try to make the public sector better 00:29:50.660 --> 00:29:53.590 align:middle line:90% by improving their incentives. 00:29:53.590 --> 00:29:56.380 align:middle line:84% I'll show you two papers showing that it 00:29:56.380 --> 00:30:00.580 align:middle line:84% seems difficult. We can try it, but it's difficult. 00:30:00.580 --> 00:30:01.780 align:middle line:90% Since this comes-- 00:30:01.780 --> 00:30:04.330 align:middle line:84% so that's one class of question, maybe you 00:30:04.330 --> 00:30:08.400 align:middle line:84% can do better than the papers I'm going to show you, 00:30:08.400 --> 00:30:10.320 align:middle line:84% another class of solution is that you 00:30:10.320 --> 00:30:13.240 align:middle line:84% could say, well, since it all come from the consumers-- 00:30:13.240 --> 00:30:16.590 align:middle line:84% in fact, I had a discussion with one of you 00:30:16.590 --> 00:30:21.030 align:middle line:84% last week after presenting the Bjorkman and co 00:30:21.030 --> 00:30:24.360 align:middle line:84% paper on the medicine in Uganda, it's solution 00:30:24.360 --> 00:30:27.420 align:middle line:84% seems to be to provide more information to teach 00:30:27.420 --> 00:30:30.270 align:middle line:84% the people what is good drug and bad drug 00:30:30.270 --> 00:30:31.920 align:middle line:90% and what a good drug should do. 00:30:31.920 --> 00:30:34.410 align:middle line:84% So basically, the idea is that since a lot 00:30:34.410 --> 00:30:36.390 align:middle line:84% of the root of the problem comes from the fact 00:30:36.390 --> 00:30:41.470 align:middle line:84% that people don't understand fully what makes you sick 00:30:41.470 --> 00:30:44.700 align:middle line:84% and what makes you heal and what to do about it, basically, 00:30:44.700 --> 00:30:47.310 align:middle line:84% try to make them more sophisticated. 00:30:47.310 --> 00:30:49.750 align:middle line:84% I'll show you one example of that. 00:30:49.750 --> 00:30:52.630 align:middle line:84% And then the third is try to work with the private sector. 00:30:52.630 --> 00:30:55.200 align:middle line:90% However bad it is, it is there. 00:30:55.200 --> 00:30:56.310 align:middle line:90% It is here. 00:30:56.310 --> 00:30:57.720 align:middle line:90% It is available. 00:30:57.720 --> 00:31:00.250 align:middle line:84% Try to make it better in some sense. 00:31:00.250 --> 00:31:03.930 align:middle line:84% So I'll give you an example in each of the three categories. 00:31:03.930 --> 00:31:07.550 align:middle line:84% There are studies in each of the three categories. 00:31:07.550 --> 00:31:09.440 align:middle line:84% So incentivize the public sector. 00:31:09.440 --> 00:31:16.450 align:middle line:84% So Abhijit, Rachel Glaser, and me did the first attempt there. 00:31:16.450 --> 00:31:18.310 align:middle line:84% On the heels of a successful project 00:31:18.310 --> 00:31:22.180 align:middle line:84% I had done with Rema Hannah of giving incentives to teacher 00:31:22.180 --> 00:31:25.518 align:middle line:84% to show up more regularly to their classes, which 00:31:25.518 --> 00:31:26.560 align:middle line:90% had been quite effective. 00:31:26.560 --> 00:31:31.180 align:middle line:84% We gave incentive, payment for their presence, and it worked. 00:31:31.180 --> 00:31:33.020 align:middle line:90% They came up. 00:31:33.020 --> 00:31:34.660 align:middle line:84% So we tried to do something similar 00:31:34.660 --> 00:31:38.590 align:middle line:84% in the same setting in Udaipur for nurses, which were absent. 00:31:38.590 --> 00:31:41.740 align:middle line:84% And absenteeism rate was crazy high at the beginning. 00:31:41.740 --> 00:31:45.430 align:middle line:84% We had surveyed the health center for a year, once a week, 00:31:45.430 --> 00:31:48.170 align:middle line:84% and they were absent 70% of the time. 00:31:48.170 --> 00:31:49.660 align:middle line:84% And it was completely unpredictable 00:31:49.660 --> 00:31:50.970 align:middle line:90% when they were open. 00:31:50.970 --> 00:31:52.720 align:middle line:84% It's not that they were open every Monday, 00:31:52.720 --> 00:31:54.340 align:middle line:90% so you at least could come. 00:31:54.340 --> 00:31:56.950 align:middle line:84% It was just some from time to time, 00:31:56.950 --> 00:31:59.110 align:middle line:84% a nurse showed up and opened the center, 00:31:59.110 --> 00:32:01.940 align:middle line:90% so we tried and changed that. 00:32:01.940 --> 00:32:06.450 align:middle line:84% So we did something, which you would characterize unambitious, 00:32:06.450 --> 00:32:09.530 align:middle line:84% which is we agreed with the government, something 00:32:09.530 --> 00:32:11.810 align:middle line:84% we did in collaboration with the government. 00:32:11.810 --> 00:32:16.340 align:middle line:84% The government told everyone in the place, treatment 00:32:16.340 --> 00:32:20.270 align:middle line:84% and control center, you have to be open, for sure, 00:32:20.270 --> 00:32:22.940 align:middle line:90% one day a week on Monday. 00:32:22.940 --> 00:32:25.740 align:middle line:84% Now, this is a new rule you cannot go to the field on that 00:32:25.740 --> 00:32:26.240 align:middle line:90% day. 00:32:26.240 --> 00:32:27.650 align:middle line:90% You have to be in your center. 00:32:27.650 --> 00:32:28.970 align:middle line:90% You cannot go to functions. 00:32:28.970 --> 00:32:30.680 align:middle line:90% You cannot go to training. 00:32:30.680 --> 00:32:32.530 align:middle line:90% You have to be there. 00:32:32.530 --> 00:32:36.660 align:middle line:84% And so that was common for the treatment and control, 00:32:36.660 --> 00:32:43.770 align:middle line:84% and then what we did is that we put in place a system 00:32:43.770 --> 00:32:47.040 align:middle line:90% to monitor presence. 00:32:47.040 --> 00:32:51.720 align:middle line:84% This was before the time of easy fingerprint recognition, 00:32:51.720 --> 00:32:53.640 align:middle line:84% so we had to make up some system where 00:32:53.640 --> 00:32:56.350 align:middle line:84% they had to sign, stamp, and sign again 00:32:56.350 --> 00:33:01.500 align:middle line:84% something on the register to show that they were, in fact, 00:33:01.500 --> 00:33:03.410 align:middle line:90% present on that day. 00:33:03.410 --> 00:33:09.260 align:middle line:84% And then we sent a person to do some random checks. 00:33:09.260 --> 00:33:12.710 align:middle line:84% We control the absent both on Monday and on other days, 00:33:12.710 --> 00:33:17.550 align:middle line:84% although the incentive was only based on Mondays. 00:33:17.550 --> 00:33:21.590 align:middle line:84% So what we find is that in treatment, 00:33:21.590 --> 00:33:24.560 align:middle line:84% initially, the absence rate really 00:33:24.560 --> 00:33:31.430 align:middle line:84% increased on Mondays and not as much on other days. 00:33:31.430 --> 00:33:35.030 align:middle line:84% So initially, by the way, there was a scope 00:33:35.030 --> 00:33:39.510 align:middle line:84% in this experiment, which is we didn't collect 00:33:39.510 --> 00:33:41.760 align:middle line:90% control data for some time. 00:33:41.760 --> 00:33:46.000 align:middle line:84% We started controlling control data after a few weeks. 00:33:46.000 --> 00:33:48.630 align:middle line:84% So this is the control group, and when we have both, 00:33:48.630 --> 00:33:51.240 align:middle line:84% the control group is, indeed, lower than the treatment group, 00:33:51.240 --> 00:33:52.245 align:middle line:90% particularly, on Monday. 00:33:52.245 --> 00:33:55.080 align:middle line:90% 00:33:55.080 --> 00:33:59.220 align:middle line:84% Then over time, the treatment group 00:33:59.220 --> 00:34:04.100 align:middle line:84% is less and less and less and less and less there. 00:34:04.100 --> 00:34:06.650 align:middle line:84% In fact, the control group is more and more and more and more 00:34:06.650 --> 00:34:10.909 align:middle line:84% there so that at the end of the experiment, 00:34:10.909 --> 00:34:13.320 align:middle line:84% about a year into the experiment, 00:34:13.320 --> 00:34:14.900 align:middle line:84% the control group is doing better 00:34:14.900 --> 00:34:19.510 align:middle line:84% on Monday and on other days than the treatment group. 00:34:19.510 --> 00:34:24.150 align:middle line:84% So this is not what I would call a successful experiment, 00:34:24.150 --> 00:34:29.460 align:middle line:84% where we actually managed to worsen absenteeism 00:34:29.460 --> 00:34:35.751 align:middle line:84% in our centers, compared to the status quo situation. 00:34:35.751 --> 00:34:36.459 align:middle line:90% So what happened? 00:34:36.459 --> 00:34:38.199 align:middle line:84% So I'll tell you what happened in the control group. 00:34:38.199 --> 00:34:40.491 align:middle line:84% I think what happened in the control group is that they 00:34:40.491 --> 00:34:43.600 align:middle line:84% progressively learned, this idea that you have 00:34:43.600 --> 00:34:47.260 align:middle line:90% to be open on Monday was new. 00:34:47.260 --> 00:34:52.199 align:middle line:84% And the message was progressively spreading. 00:34:52.199 --> 00:34:56.880 align:middle line:84% And so over time, the regular administration 00:34:56.880 --> 00:35:00.000 align:middle line:84% was putting some pressure on everyone to be there on Monday. 00:35:00.000 --> 00:35:02.353 align:middle line:84% They removed the training programs on Monday. 00:35:02.353 --> 00:35:04.770 align:middle line:84% They told them that emission days have to be on other days 00:35:04.770 --> 00:35:08.040 align:middle line:84% than Monday, so all of that made its way, 00:35:08.040 --> 00:35:09.990 align:middle line:90% and the control group improved. 00:35:09.990 --> 00:35:11.880 align:middle line:90% Fine. 00:35:11.880 --> 00:35:13.650 align:middle line:84% So that makes sense for why this should 00:35:13.650 --> 00:35:17.340 align:middle line:84% have improved, but why did this go down, meanwhile, 00:35:17.340 --> 00:35:18.330 align:middle line:90% in the treatment group? 00:35:18.330 --> 00:35:21.830 align:middle line:84% It started quite high, and then it went down, even on Monday. 00:35:21.830 --> 00:35:23.580 align:middle line:84% In your opinion, what could have happened? 00:35:23.580 --> 00:35:27.450 align:middle line:90% 00:35:27.450 --> 00:35:28.293 align:middle line:90% Yeah. 00:35:28.293 --> 00:35:30.751 align:middle line:84% AUDIENCE: Because they realized there were no consequences? 00:35:30.751 --> 00:35:31.780 align:middle line:90% ESTHER DUFLO: Exactly. 00:35:31.780 --> 00:35:35.000 align:middle line:84% So they realized there wasn't any consequences. 00:35:35.000 --> 00:35:37.820 align:middle line:90% And how did that happen? 00:35:37.820 --> 00:35:40.480 align:middle line:84% Well, they started being present more, 00:35:40.480 --> 00:35:43.130 align:middle line:84% but they were still absent quite a bit. 00:35:43.130 --> 00:35:46.390 align:middle line:84% And then there were three categories-- 00:35:46.390 --> 00:35:49.750 align:middle line:84% you were is there a present day, an absent day, or an excuse 00:35:49.750 --> 00:35:50.920 align:middle line:90% day? 00:35:50.920 --> 00:35:53.920 align:middle line:84% And over time, what happens is that the bosses 00:35:53.920 --> 00:35:56.710 align:middle line:84% of these nurses, which are the doctors and the PhDs, 00:35:56.710 --> 00:36:01.300 align:middle line:84% converted all of the absent days into excuse days. 00:36:01.300 --> 00:36:03.513 align:middle line:84% And they made all sorts of excuses for them, 00:36:03.513 --> 00:36:04.930 align:middle line:84% including that they were training, 00:36:04.930 --> 00:36:08.410 align:middle line:84% even though we knew there were no training on those days. 00:36:08.410 --> 00:36:11.065 align:middle line:84% But still, on the register, you had list of excuses. 00:36:11.065 --> 00:36:13.540 align:middle line:84% You had to do this, you had to do that, et cetera. 00:36:13.540 --> 00:36:18.730 align:middle line:84% So that, in fact, when you compare the official register, 00:36:18.730 --> 00:36:24.540 align:middle line:84% at the beginning, there is absence and very few exam days. 00:36:24.540 --> 00:36:26.910 align:middle line:84% And then overtime, there are no absences, 00:36:26.910 --> 00:36:28.720 align:middle line:90% and there are only exam days. 00:36:28.720 --> 00:36:30.990 align:middle line:84% So the exam does increase enormously, 00:36:30.990 --> 00:36:33.540 align:middle line:84% the absence goes down, and that's 00:36:33.540 --> 00:36:37.050 align:middle line:84% why, finally, the treatment is worse than the control. 00:36:37.050 --> 00:36:39.240 align:middle line:84% So basically, they learned progressively 00:36:39.240 --> 00:36:42.510 align:middle line:84% that the rule is not enforced anyway, so there is no reason 00:36:42.510 --> 00:36:44.880 align:middle line:90% to bother. 00:36:44.880 --> 00:36:47.350 align:middle line:84% And the fact that the rules was very, very clear, 00:36:47.350 --> 00:36:51.150 align:middle line:84% which is your absence, the absence, if it's not excused, 00:36:51.150 --> 00:36:56.880 align:middle line:84% is going to take away from your salary, makes it even obvious 00:36:56.880 --> 00:36:59.320 align:middle line:84% that actually, the rule is not enforced. 00:36:59.320 --> 00:37:02.100 align:middle line:84% So there is a rule, but it's not enforced, so we are good to go. 00:37:02.100 --> 00:37:04.380 align:middle line:84% Whereas in the control group, it's 00:37:04.380 --> 00:37:06.420 align:middle line:84% the normal thing in the bureaucracy, which 00:37:06.420 --> 00:37:10.230 align:middle line:84% is there is always this vague threat of punishment that 00:37:10.230 --> 00:37:14.590 align:middle line:84% might hold upon you, and you have no data to confirm, 00:37:14.590 --> 00:37:16.590 align:middle line:84% whether that vague threat or punishment is going 00:37:16.590 --> 00:37:18.465 align:middle line:90% to be carried out or not Yeah. 00:37:18.465 --> 00:37:21.620 align:middle line:84% AUDIENCE: Do you know what they do when they don't show up 00:37:21.620 --> 00:37:22.640 align:middle line:90% to the office? 00:37:22.640 --> 00:37:24.140 align:middle line:84% ESTHER DUFLO: What they do when they 00:37:24.140 --> 00:37:25.307 align:middle line:90% don't show up to the office? 00:37:25.307 --> 00:37:26.630 align:middle line:90% Yeah, they're home. 00:37:26.630 --> 00:37:31.640 align:middle line:84% These nurses, they are well-educated, 00:37:31.640 --> 00:37:33.900 align:middle line:84% they don't want to live in the middle of the village, 00:37:33.900 --> 00:37:35.630 align:middle line:84% they have kids that they send to school. 00:37:35.630 --> 00:37:38.108 align:middle line:84% So they live pretty far from their center, and to show up, 00:37:38.108 --> 00:37:39.650 align:middle line:84% they need to take a bus and all that. 00:37:39.650 --> 00:37:41.370 align:middle line:90% It's a hassle. 00:37:41.370 --> 00:37:44.640 align:middle line:84% So if you cannot show up, then why not? 00:37:44.640 --> 00:37:46.370 align:middle line:84% And so that's basically what they are. 00:37:46.370 --> 00:37:50.030 align:middle line:84% They are home, they take care of their kids, et cetera. 00:37:50.030 --> 00:37:53.480 align:middle line:84% And once we had sent a clear message that going to office 00:37:53.480 --> 00:37:56.675 align:middle line:84% was not requirement of the job, then that 00:37:56.675 --> 00:37:57.800 align:middle line:90% makes it even more obvious. 00:37:57.800 --> 00:37:58.300 align:middle line:90% Yeah. 00:37:58.300 --> 00:38:00.770 align:middle line:84% AUDIENCE: You said the control group doesn't 00:38:00.770 --> 00:38:03.860 align:middle line:84% have any data to know that they actually don't enforce it 00:38:03.860 --> 00:38:08.180 align:middle line:84% at the moment, but won't they just realize 00:38:08.180 --> 00:38:10.820 align:middle line:84% that they've been in some Mondays for this whole time 00:38:10.820 --> 00:38:14.550 align:middle line:90% and nothing has happened-- 00:38:14.550 --> 00:38:16.230 align:middle line:84% ESTHER DUFLO: But it's not as clear 00:38:16.230 --> 00:38:21.150 align:middle line:84% because the normal punishment in the absence of daily incentive 00:38:21.150 --> 00:38:24.360 align:middle line:84% is you might get sacked at some point. 00:38:24.360 --> 00:38:26.483 align:middle line:84% So there, they still have their supervisors 00:38:26.483 --> 00:38:28.400 align:middle line:84% that tell them if you don't show up on Monday, 00:38:28.400 --> 00:38:30.810 align:middle line:84% you might be sacked and all that, 00:38:30.810 --> 00:38:33.410 align:middle line:84% and so that's probably somewhere in the back of their mind. 00:38:33.410 --> 00:38:35.240 align:middle line:84% Whereas here, there is a pretty clear rule, which 00:38:35.240 --> 00:38:36.657 align:middle line:84% is if you don't show up on Monday, 00:38:36.657 --> 00:38:38.720 align:middle line:84% you're going to get pay deducted. 00:38:38.720 --> 00:38:41.000 align:middle line:84% And then you can see that your pay is never deducted 00:38:41.000 --> 00:38:42.830 align:middle line:84% because the absence is very regular. 00:38:42.830 --> 00:38:44.900 align:middle line:84% Here, in the control group, they don't know. 00:38:44.900 --> 00:38:46.580 align:middle line:84% Nobody is measuring their absence, 00:38:46.580 --> 00:38:52.490 align:middle line:84% but there is irregular visits by supervisor. 00:38:52.490 --> 00:38:55.260 align:middle line:84% They don't know when the supervisor might show up. 00:38:55.260 --> 00:38:57.350 align:middle line:84% So that's the kind of disadvantage 00:38:57.350 --> 00:39:01.400 align:middle line:84% of having a clear rule that is not enforced, making it easier 00:39:01.400 --> 00:39:05.337 align:middle line:84% to understand what the incentives are. 00:39:05.337 --> 00:39:06.920 align:middle line:84% So I think this is what happened here. 00:39:06.920 --> 00:39:10.590 align:middle line:90% 00:39:10.590 --> 00:39:13.800 align:middle line:84% But then the next question is, how could the government 00:39:13.800 --> 00:39:17.890 align:middle line:84% get away with not implementing their own rules? 00:39:17.890 --> 00:39:20.430 align:middle line:84% And one of the possible reasons is no one cares, 00:39:20.430 --> 00:39:22.545 align:middle line:84% because everybody has anywhere abandoned, 00:39:22.545 --> 00:39:25.620 align:middle line:84% and they are not giving up on the centers. 00:39:25.620 --> 00:39:28.830 align:middle line:90% 00:39:28.830 --> 00:39:30.630 align:middle line:84% Improving the quality of the centers not 00:39:30.630 --> 00:39:34.230 align:middle line:84% something that's very high on everybody's priority. 00:39:34.230 --> 00:39:37.170 align:middle line:84% And this is shown by the fact that it didn't bring back 00:39:37.170 --> 00:39:37.680 align:middle line:90% clients. 00:39:37.680 --> 00:39:39.450 align:middle line:84% So at the beginning of the experiment, 00:39:39.450 --> 00:39:41.970 align:middle line:84% when the facilities were open more often, 00:39:41.970 --> 00:39:44.700 align:middle line:84% you could see whether more clients came. 00:39:44.700 --> 00:39:47.430 align:middle line:84% In a given day, when we did these visits, 00:39:47.430 --> 00:39:52.470 align:middle line:84% you see someone, one person, not even all the time. 00:39:52.470 --> 00:39:56.880 align:middle line:84% 0.74 clients in treatment facility when it's open, 00:39:56.880 --> 00:39:59.640 align:middle line:84% 0.81 clients in private facilities, 00:39:59.640 --> 00:40:03.520 align:middle line:84% so people don't care, but they have given up. 00:40:03.520 --> 00:40:06.690 align:middle line:84% And that's why we call the people putting a Band-Aid 00:40:06.690 --> 00:40:12.320 align:middle line:84% on the corpse because basically, the system, 00:40:12.320 --> 00:40:16.790 align:middle line:84% without any demand for that system in the first place, 00:40:16.790 --> 00:40:20.760 align:middle line:84% it doesn't really rise or bubble up to the level of priorities. 00:40:20.760 --> 00:40:21.260 align:middle line:90% Yeah. 00:40:21.260 --> 00:40:23.635 align:middle line:84% AUDIENCE: I assume that we know that the relatives nearby 00:40:23.635 --> 00:40:27.140 align:middle line:84% the facilities knew of this new rule, and it be OK. 00:40:27.140 --> 00:40:31.630 align:middle line:84% This would be [INAUDIBLE],, whether we know that. 00:40:31.630 --> 00:40:35.330 align:middle line:84% ESTHER DUFLO: So the rule was put on the centers, 00:40:35.330 --> 00:40:39.140 align:middle line:84% but you have to have some minimal level of credibility 00:40:39.140 --> 00:40:41.070 align:middle line:84% that, actually, it's going to be enforced, 00:40:41.070 --> 00:40:43.370 align:middle line:90% which nobody had any trust. 00:40:43.370 --> 00:40:47.960 align:middle line:84% If you watch on YouTube, there is Abhijit's movie 00:40:47.960 --> 00:40:53.070 align:middle line:84% on the health care, and they call the name of the disease. 00:40:53.070 --> 00:40:55.130 align:middle line:84% And there is a very funny sequence 00:40:55.130 --> 00:40:57.710 align:middle line:84% where we interview someone about a girl, 00:40:57.710 --> 00:41:00.440 align:middle line:84% about the interview someone, the girl about the center. 00:41:00.440 --> 00:41:03.065 align:middle line:84% And she just gives all sorts of excuses to them. 00:41:03.065 --> 00:41:04.440 align:middle line:84% And then finally, she said, well, 00:41:04.440 --> 00:41:06.170 align:middle line:90% she's never out there anyways. 00:41:06.170 --> 00:41:08.915 align:middle line:84% So there is a pretty good understanding 00:41:08.915 --> 00:41:10.040 align:middle line:90% that that's the status quo. 00:41:10.040 --> 00:41:13.982 align:middle line:84% And you're right, there is a chicken and egg problem. 00:41:13.982 --> 00:41:16.190 align:middle line:84% The nurses are saying there's no point for me coming, 00:41:16.190 --> 00:41:17.270 align:middle line:90% nobody comes anyway. 00:41:17.270 --> 00:41:19.610 align:middle line:84% It's totally depressing to sit in my drab center 00:41:19.610 --> 00:41:22.550 align:middle line:84% with no medicine and no patients. 00:41:22.550 --> 00:41:25.190 align:middle line:84% And the patient said, why would I even go there 00:41:25.190 --> 00:41:28.730 align:middle line:84% when I can go to any of 10 private guy, who 00:41:28.730 --> 00:41:31.280 align:middle line:84% is unqualified, but not so much worse and at least 00:41:31.280 --> 00:41:32.950 align:middle line:90% is available. 00:41:32.950 --> 00:41:35.890 align:middle line:84% So reforming from the top without building 00:41:35.890 --> 00:41:40.710 align:middle line:84% a demand seems to be difficult. So that's that. 00:41:40.710 --> 00:41:43.530 align:middle line:84% So hence, there is the other idea of building a demand, 00:41:43.530 --> 00:41:45.000 align:middle line:84% maybe starting from a system, which 00:41:45.000 --> 00:41:46.920 align:middle line:84% is less defunct in the first place, which it 00:41:46.920 --> 00:41:49.610 align:middle line:90% is in most countries in Africa. 00:41:49.610 --> 00:41:50.990 align:middle line:90% The centers are open. 00:41:50.990 --> 00:41:53.730 align:middle line:90% There is more life there. 00:41:53.730 --> 00:41:58.310 align:middle line:84% So there is a very nice paper by the same Bjorkman and Svensson 00:41:58.310 --> 00:42:01.550 align:middle line:84% of medicine, that's actually in earlier paper. 00:42:01.550 --> 00:42:04.820 align:middle line:84% Similar problems, so the absence rate in the health center 00:42:04.820 --> 00:42:06.740 align:middle line:90% is 47%. 00:42:06.740 --> 00:42:08.480 align:middle line:84% I think it's not as bad in the sense 00:42:08.480 --> 00:42:11.940 align:middle line:84% that they are not different, and the centers are a bit larger. 00:42:11.940 --> 00:42:15.410 align:middle line:84% So there is always somewhere, but a lot of them are absent. 00:42:15.410 --> 00:42:17.840 align:middle line:90% The care is not great. 00:42:17.840 --> 00:42:21.270 align:middle line:84% So instead of using the top-down approach that we tried, 00:42:21.270 --> 00:42:24.320 align:middle line:84% they involve the community in monitoring the provider. 00:42:24.320 --> 00:42:27.070 align:middle line:84% So this was, at the time, this has slightly 00:42:27.070 --> 00:42:28.450 align:middle line:90% gone out of fashion. 00:42:28.450 --> 00:42:32.020 align:middle line:84% But about 10 years ago, 15 years ago, in the World Bank, 00:42:32.020 --> 00:42:33.160 align:middle line:90% this was everything. 00:42:33.160 --> 00:42:37.810 align:middle line:84% There was a public service, the World Bank Annual Report 00:42:37.810 --> 00:42:39.850 align:middle line:84% on Public Service Delivery, which 00:42:39.850 --> 00:42:42.640 align:middle line:84% was all about power to the people 00:42:42.640 --> 00:42:45.700 align:middle line:84% empowering the people, et cetera, community empowerment. 00:42:45.700 --> 00:42:47.620 align:middle line:84% And this is one of the few attempts 00:42:47.620 --> 00:42:50.440 align:middle line:84% to really take this idea seriously to the field 00:42:50.440 --> 00:42:51.920 align:middle line:90% and test it. 00:42:51.920 --> 00:42:53.440 align:middle line:84% So what they did is that they first 00:42:53.440 --> 00:42:55.060 align:middle line:84% went and collected their household 00:42:55.060 --> 00:42:58.540 align:middle line:84% survey with experience with people health facilities. 00:42:58.540 --> 00:43:01.180 align:middle line:84% Then they hired community-based organizations 00:43:01.180 --> 00:43:04.360 align:middle line:84% to facilitate three meetings-- a community meeting, 00:43:04.360 --> 00:43:07.390 align:middle line:84% a meeting at the health center, and an interface meeting 00:43:07.390 --> 00:43:10.360 align:middle line:84% involving both the nurses and the doctors and the people. 00:43:10.360 --> 00:43:11.980 align:middle line:84% They had to come up with an action 00:43:11.980 --> 00:43:14.620 align:middle line:84% plan about how to improve the situation 00:43:14.620 --> 00:43:18.340 align:middle line:84% and how the community member would come and do monitoring 00:43:18.340 --> 00:43:22.660 align:middle line:84% on the facilities, which is a bit worrisome, the Texas 00:43:22.660 --> 00:43:25.960 align:middle line:84% law of vigilantes coming to the facilities to monitor them, 00:43:25.960 --> 00:43:29.050 align:middle line:84% but it's probably not that sinister. 00:43:29.050 --> 00:43:33.100 align:middle line:84% So people volunteer to come and monitor 00:43:33.100 --> 00:43:36.880 align:middle line:84% that the nurses were there, and so on, and so forth. 00:43:36.880 --> 00:43:39.100 align:middle line:90% The results were pretty great. 00:43:39.100 --> 00:43:43.930 align:middle line:90% This is impressive results. 00:43:43.930 --> 00:43:49.330 align:middle line:84% Absenteeism went down from 47% to 34%. 00:43:49.330 --> 00:43:53.350 align:middle line:84% People did come more, either as a result of absenteeism 00:43:53.350 --> 00:43:57.010 align:middle line:84% or perhaps declining or perhaps because, unlike in our case, 00:43:57.010 --> 00:43:59.110 align:middle line:84% there was this public meeting, so everybody knew 00:43:59.110 --> 00:44:01.570 align:middle line:90% it was supposed to be better. 00:44:01.570 --> 00:44:04.390 align:middle line:84% They still continued to use the traditional healer but, 00:44:04.390 --> 00:44:10.810 align:middle line:84% on balance, which is the key litmus 00:44:10.810 --> 00:44:13.750 align:middle line:84% test of this intervention being effective, 00:44:13.750 --> 00:44:18.925 align:middle line:84% under five mortality rate went from 144 per 1,000 to 9 00:44:18.925 --> 00:44:20.680 align:middle line:90% per 1,000. 00:44:20.680 --> 00:44:25.540 align:middle line:84% Even more remarkably, the same team did a follow up, 00:44:25.540 --> 00:44:29.170 align:middle line:84% maybe 10 years later, in the same facilities, 00:44:29.170 --> 00:44:32.210 align:middle line:90% and they still see impact. 00:44:32.210 --> 00:44:34.330 align:middle line:84% So I really believe that this intervention 00:44:34.330 --> 00:44:37.500 align:middle line:90% worked in this set of villages. 00:44:37.500 --> 00:44:43.540 align:middle line:84% That said, it is not really a replicable intervention. 00:44:43.540 --> 00:44:47.860 align:middle line:84% It is extraordinarily expensive to conduct 00:44:47.860 --> 00:44:49.100 align:middle line:90% all of these meetings. 00:44:49.100 --> 00:44:50.890 align:middle line:90% It has to be done just so. 00:44:50.890 --> 00:44:53.935 align:middle line:84% Any attempt to replicate it in other contexts has not worked. 00:44:53.935 --> 00:44:59.512 align:middle line:90% 00:44:59.512 --> 00:45:00.970 align:middle line:84% I don't think this one was a fluke, 00:45:00.970 --> 00:45:03.280 align:middle line:84% but just because this one was particularly well done, 00:45:03.280 --> 00:45:05.240 align:middle line:84% nobody has managed to do it as well. 00:45:05.240 --> 00:45:07.240 align:middle line:84% So there is, for example a replication study 00:45:07.240 --> 00:45:09.490 align:middle line:90% by a political scientist. 00:45:09.490 --> 00:45:13.960 align:middle line:84% I think, also, in Uganda, trying to do just the same 00:45:13.960 --> 00:45:15.880 align:middle line:84% and not finding any effect, so they 00:45:15.880 --> 00:45:17.255 align:middle line:84% have a bit of a dispute with them 00:45:17.255 --> 00:45:19.338 align:middle line:84% where they're saying they didn't do just the same. 00:45:19.338 --> 00:45:20.390 align:middle line:90% It wasn't quite as good. 00:45:20.390 --> 00:45:22.690 align:middle line:84% But the point is that that's just so 00:45:22.690 --> 00:45:25.840 align:middle line:84% such an intense intervention that it's hard to replicate. 00:45:25.840 --> 00:45:28.450 align:middle line:84% So what Bjorkman and Svensson did 00:45:28.450 --> 00:45:32.770 align:middle line:84% is that they tried a lighter version of this intervention, 00:45:32.770 --> 00:45:36.550 align:middle line:84% where they give the information to people on the facilities, 00:45:36.550 --> 00:45:39.640 align:middle line:84% but they don't have all of the community meeting. 00:45:39.640 --> 00:45:43.710 align:middle line:84% Actually, now I have a doubt which 00:45:43.710 --> 00:45:45.180 align:middle line:84% element they got rid of, whether it 00:45:45.180 --> 00:45:47.358 align:middle line:84% was the full survey or the community meeting. 00:45:47.358 --> 00:45:49.650 align:middle line:84% But anyway, some lighter version that they were hopeful 00:45:49.650 --> 00:45:52.260 align:middle line:84% would work out, and then it didn't work. 00:45:52.260 --> 00:45:55.180 align:middle line:84% They have no impact on it at all. 00:45:55.180 --> 00:45:58.120 align:middle line:84% So their own effort, which has to be commended 00:45:58.120 --> 00:46:02.255 align:middle line:84% of realizing that that's a super nice proof of concept, 00:46:02.255 --> 00:46:03.630 align:middle line:84% but if you want to scale it up it 00:46:03.630 --> 00:46:06.130 align:middle line:84% has to be a little bit more manageable, 00:46:06.130 --> 00:46:09.450 align:middle line:84% the effort to make this more manageable have not worked out. 00:46:09.450 --> 00:46:14.970 align:middle line:84% So it's a nice proof of context, but it's not ready 00:46:14.970 --> 00:46:16.500 align:middle line:84% for primetime in terms of this is 00:46:16.500 --> 00:46:20.554 align:middle line:84% what we're going to scale up to improve the situation. 00:46:20.554 --> 00:46:23.630 align:middle line:84% So that's for reforming the public sector 00:46:23.630 --> 00:46:28.010 align:middle line:84% by providing incentive, either top down or bottom up. 00:46:28.010 --> 00:46:29.900 align:middle line:84% Second class of things, could you 00:46:29.900 --> 00:46:32.870 align:middle line:84% just make the consumer more sophisticated, 00:46:32.870 --> 00:46:37.220 align:middle line:84% teach them exactly how things work. 00:46:37.220 --> 00:46:40.520 align:middle line:84% And so one example of that is a paper 00:46:40.520 --> 00:46:44.840 align:middle line:84% by Jessica Cohen, Pascaline Dupas, and Simone Schaner, 00:46:44.840 --> 00:46:48.050 align:middle line:84% where they basically create sophistication 00:46:48.050 --> 00:46:53.660 align:middle line:84% by introducing a rapid diagnostic tests for malaria 00:46:53.660 --> 00:46:55.020 align:middle line:90% that wasn't there before. 00:46:55.020 --> 00:46:57.570 align:middle line:84% So people can understand one of the people. 00:46:57.570 --> 00:47:01.010 align:middle line:84% This is in Kenya, where there is routine of treatment 00:47:01.010 --> 00:47:02.360 align:middle line:90% for malaria. 00:47:02.360 --> 00:47:05.680 align:middle line:84% In particular, people take this ACT, the new drug, 00:47:05.680 --> 00:47:07.430 align:middle line:84% even when they don't have malaria, so they 00:47:07.430 --> 00:47:09.380 align:middle line:90% are trying to test that. 00:47:09.380 --> 00:47:14.050 align:middle line:84% But the first thing they do, as just background, 00:47:14.050 --> 00:47:16.870 align:middle line:84% so the ACT is this new item, is in combination therapy. 00:47:16.870 --> 00:47:18.400 align:middle line:90% It's effective today. 00:47:18.400 --> 00:47:23.000 align:middle line:84% All the anti-malaria medicine is not effective in Kenya anymore, 00:47:23.000 --> 00:47:24.830 align:middle line:84% and if this one is used too much, 00:47:24.830 --> 00:47:27.260 align:middle line:84% it's also going to become ineffective. 00:47:27.260 --> 00:47:30.370 align:middle line:84% This might all be solved if everybody get vaccinated 00:47:30.370 --> 00:47:31.970 align:middle line:84% against malaria, but at the time, 00:47:31.970 --> 00:47:35.120 align:middle line:84% there was no vaccines for malaria. 00:47:35.120 --> 00:47:38.590 align:middle line:84% So what they did is first, replicating this study 00:47:38.590 --> 00:47:44.440 align:middle line:84% on subsidizing treatment, so I've 00:47:44.440 --> 00:47:46.990 align:middle line:84% shown you a graph where we subsidize preventive care, 00:47:46.990 --> 00:47:49.180 align:middle line:90% showing high elasticity. 00:47:49.180 --> 00:47:54.930 align:middle line:84% And here, you're finding that when you move from no subsidy 00:47:54.930 --> 00:48:04.860 align:middle line:84% to complete subsidy, people are more likely to take ACT 00:48:04.860 --> 00:48:06.090 align:middle line:90% when they have some subsidy. 00:48:06.090 --> 00:48:07.465 align:middle line:84% They are more likely to take any, 00:48:07.465 --> 00:48:09.480 align:middle line:84% and they are more likely to use their voucher, 00:48:09.480 --> 00:48:12.750 align:middle line:90% but the impact is much less big. 00:48:12.750 --> 00:48:17.700 align:middle line:84% You go from a full subsidy to 80% subsidy. 00:48:17.700 --> 00:48:21.480 align:middle line:84% You get a 15 percentage point increase in take up. 00:48:21.480 --> 00:48:23.880 align:middle line:90% So there is an elasticity. 00:48:23.880 --> 00:48:27.750 align:middle line:84% It's not as brutal as for preventive care, 00:48:27.750 --> 00:48:31.380 align:middle line:84% and the problem is that it leads to the people 00:48:31.380 --> 00:48:35.010 align:middle line:84% that it convinced to take, we have the same issue 00:48:35.010 --> 00:48:38.000 align:middle line:84% as what was discussed for preventive care. 00:48:38.000 --> 00:48:41.580 align:middle line:84% So people who get convinced to take the ACT anyways 00:48:41.580 --> 00:48:44.430 align:middle line:84% are the people who don't have malaria. 00:48:44.430 --> 00:48:48.590 align:middle line:84% So basically, people are able to predict relatively well 00:48:48.590 --> 00:48:51.560 align:middle line:84% whether or not they actually have malaria, 00:48:51.560 --> 00:48:53.780 align:middle line:84% and the people who think they have malaria 00:48:53.780 --> 00:48:55.340 align:middle line:84% could see quite strongly that malaria 00:48:55.340 --> 00:48:59.030 align:middle line:84% takes the pill regardless of the subsidy, but the people who 00:48:59.030 --> 00:49:00.620 align:middle line:90% don't wait to see. 00:49:00.620 --> 00:49:03.170 align:middle line:90% So we can look at-- 00:49:03.170 --> 00:49:05.570 align:middle line:84% this predicted positivity is based 00:49:05.570 --> 00:49:08.750 align:middle line:84% on what the patient, could know about them-- 00:49:08.750 --> 00:49:12.890 align:middle line:84% their fever, shivering, by those kind of symptoms. 00:49:12.890 --> 00:49:15.080 align:middle line:84% They predicted to be positive, and this 00:49:15.080 --> 00:49:17.450 align:middle line:90% is compared to the 92% subsidy. 00:49:17.450 --> 00:49:22.620 align:middle line:84% So compared to the 92% subsidy when the subsidy is lower, 00:49:22.620 --> 00:49:25.860 align:middle line:84% people are more likely to be predicted positive for malaria 00:49:25.860 --> 00:49:27.350 align:middle line:90% when they take the medicine. 00:49:27.350 --> 00:49:30.810 align:middle line:84% So in other words, making it near free and there 00:49:30.810 --> 00:49:33.000 align:middle line:84% are people who think they don't have malaria, 00:49:33.000 --> 00:49:36.480 align:middle line:84% to do it anyways, to take the drug anyways. 00:49:36.480 --> 00:49:38.910 align:middle line:84% And in fact, when you test them, so what they do 00:49:38.910 --> 00:49:41.560 align:middle line:84% is that they go to the facility, people who come and say, 00:49:41.560 --> 00:49:43.650 align:middle line:90% can I give you a test now? 00:49:43.650 --> 00:49:53.130 align:middle line:84% And they also find that compared to the 92% subsidy, compared 00:49:53.130 --> 00:50:00.360 align:middle line:84% to nearly full subsidy, people who had to pay more 00:50:00.360 --> 00:50:05.350 align:middle line:84% are more likely to, in fact, be positive for malaria. 00:50:05.350 --> 00:50:09.690 align:middle line:84% So here you find in the case of this curative care, 00:50:09.690 --> 00:50:12.870 align:middle line:84% you have this effect of the extra subsidy drawing 00:50:12.870 --> 00:50:15.690 align:middle line:84% in a lot of people who don't really need it, 00:50:15.690 --> 00:50:18.820 align:middle line:84% which we didn't find in the case of preventive care. 00:50:18.820 --> 00:50:19.827 align:middle line:90% So that's 0.1. 00:50:19.827 --> 00:50:21.660 align:middle line:84% This is a little because what I want to talk 00:50:21.660 --> 00:50:23.723 align:middle line:90% to you is about testing. 00:50:23.723 --> 00:50:26.140 align:middle line:84% So the question is that, while we are in this world, where 00:50:26.140 --> 00:50:31.960 align:middle line:84% people basically overmedicate for malaria, on average, 00:50:31.960 --> 00:50:37.990 align:middle line:84% only 56% of people who test for malaria and have taken 00:50:37.990 --> 00:50:40.150 align:middle line:90% an ACT have malaria. 00:50:40.150 --> 00:50:43.750 align:middle line:84% That means 44% of people take an ACT, when they shouldn't, 00:50:43.750 --> 00:50:46.430 align:middle line:90% when it's not required. 00:50:46.430 --> 00:50:48.970 align:middle line:84% So they say, well, if we tested people for malaria, 00:50:48.970 --> 00:50:53.440 align:middle line:84% then maybe they wouldn't need to take 00:50:53.440 --> 00:50:57.400 align:middle line:84% the medicine if they didn't have it and we would have fewer. 00:50:57.400 --> 00:50:58.970 align:middle line:90% So what they offer is. 00:50:58.970 --> 00:51:00.820 align:middle line:84% And the problem is that-- and that's 00:51:00.820 --> 00:51:03.400 align:middle line:84% the source of the lack of sophistication here is not 00:51:03.400 --> 00:51:09.340 align:middle line:84% the patient's fault, is the fact that there exists 00:51:09.340 --> 00:51:11.650 align:middle line:84% no good tests, and the test that is proposed 00:51:11.650 --> 00:51:15.220 align:middle line:84% in the public health center has very high error 00:51:15.220 --> 00:51:18.280 align:middle line:84% rate of both Type 1 and Type 2, so they are not 00:51:18.280 --> 00:51:21.930 align:middle line:84% very discriminant, and people know that. 00:51:21.930 --> 00:51:23.330 align:middle line:84% So here what they do is that they 00:51:23.330 --> 00:51:26.900 align:middle line:84% buy a bunch of rapid diagnostic tests, which are much better 00:51:26.900 --> 00:51:29.730 align:middle line:84% and can be administered at the pharmacy. 00:51:29.730 --> 00:51:31.700 align:middle line:84% So it's basically thinking of getting 00:51:31.700 --> 00:51:33.830 align:middle line:84% your COVID test, an antigenic COVID test, 00:51:33.830 --> 00:51:35.750 align:middle line:84% you get the response in 15 minutes. 00:51:35.750 --> 00:51:39.140 align:middle line:84% Then only you take the medicine once the Merck medicine 00:51:39.140 --> 00:51:40.380 align:middle line:90% will be approved. 00:51:40.380 --> 00:51:43.280 align:middle line:84% So here, same thing, but for malaria. 00:51:43.280 --> 00:51:45.110 align:middle line:84% And what they find is that a lot of people 00:51:45.110 --> 00:51:52.160 align:middle line:84% are perfectly willing to try with the test. 00:51:52.160 --> 00:51:55.220 align:middle line:84% 20% of people, if they have to pay for it, 00:51:55.220 --> 00:52:02.960 align:middle line:84% and it goes to, on average, about 50% if it's subsidized. 00:52:02.960 --> 00:52:06.380 align:middle line:84% And if they have a higher predicted malaria, 00:52:06.380 --> 00:52:08.160 align:middle line:84% then they're more likely to take the test. 00:52:08.160 --> 00:52:11.210 align:middle line:84% So people seem to be happy to try. 00:52:11.210 --> 00:52:15.000 align:middle line:84% The problem is that they don't listen to the results. 00:52:15.000 --> 00:52:19.140 align:middle line:84% So if the test is positive, everyone takes the drug. 00:52:19.140 --> 00:52:20.930 align:middle line:84% So that's good in that direction. 00:52:20.930 --> 00:52:23.640 align:middle line:84% If the test is positive, everyone takes the drug. 00:52:23.640 --> 00:52:29.250 align:middle line:84% But if the test is negative, the compliance is not taking ACT. 00:52:29.250 --> 00:52:32.180 align:middle line:84% And you can see that even with the lowest subsidy 00:52:32.180 --> 00:52:36.920 align:middle line:84% level, 80%, half the people do not take the medicine, 00:52:36.920 --> 00:52:38.750 align:middle line:84% half the people do, despite the fact 00:52:38.750 --> 00:52:41.270 align:middle line:84% that they just learned the second that they 00:52:41.270 --> 00:52:43.450 align:middle line:90% don't have malaria. 00:52:43.450 --> 00:52:47.670 align:middle line:84% And if you subsidized them, then only 25% of them 00:52:47.670 --> 00:52:51.480 align:middle line:84% abstain to take the malaria medicine, so 75% of them 00:52:51.480 --> 00:52:54.480 align:middle line:84% take the medicine even though the test just told them 00:52:54.480 --> 00:52:56.460 align:middle line:90% that they don't have malaria. 00:52:56.460 --> 00:52:59.220 align:middle line:84% And it's not that they take the medicine and hide it, 00:52:59.220 --> 00:53:01.420 align:middle line:90% they take it and take it. 00:53:01.420 --> 00:53:02.620 align:middle line:90% So what's going on here? 00:53:02.620 --> 00:53:04.150 align:middle line:90% So that's the first problem. 00:53:04.150 --> 00:53:06.060 align:middle line:84% The second problem is that it doesn't really 00:53:06.060 --> 00:53:10.110 align:middle line:90% help with targeting. 00:53:10.110 --> 00:53:14.370 align:middle line:84% Even after the ADT, the extra people 00:53:14.370 --> 00:53:16.770 align:middle line:84% who are drawn in to take the medicine, 00:53:16.770 --> 00:53:19.320 align:middle line:84% despite the fact that they just tested negative, 00:53:19.320 --> 00:53:23.040 align:middle line:84% are mostly the people who don't have malaria. 00:53:23.040 --> 00:53:25.520 align:middle line:84% So it doesn't really help with your missed and missing. 00:53:25.520 --> 00:53:27.260 align:middle line:90% So what happened? 00:53:27.260 --> 00:53:30.110 align:middle line:84% What happened is that even this new shiny test is not 00:53:30.110 --> 00:53:34.790 align:middle line:84% trusted because and probably because the standard test 00:53:34.790 --> 00:53:36.420 align:middle line:90% has a lot of errors. 00:53:36.420 --> 00:53:39.620 align:middle line:84% And so people don't know if this one is really much better 00:53:39.620 --> 00:53:42.660 align:middle line:84% or it's new, and so they were happy to do it, 00:53:42.660 --> 00:53:46.460 align:middle line:84% but they prefer to rely on their intuition anyways. 00:53:46.460 --> 00:53:50.000 align:middle line:84% And this shows how difficult it is, how to mistrust, 00:53:50.000 --> 00:53:55.070 align:middle line:84% make it difficult to introduce a new device that 00:53:55.070 --> 00:53:59.640 align:middle line:90% will help people to do better. 00:53:59.640 --> 00:54:02.930 align:middle line:84% And in fact, I was talking to a graduate student in Michigan 00:54:02.930 --> 00:54:07.790 align:middle line:84% a few weeks ago, who is doing a project on Nigeria, where they 00:54:07.790 --> 00:54:10.910 align:middle line:84% have made a huge effort to improve the supply 00:54:10.910 --> 00:54:15.165 align:middle line:90% chain for malaria medications. 00:54:15.165 --> 00:54:17.540 align:middle line:84% And now, they are pretty sure that the malaria medication 00:54:17.540 --> 00:54:20.030 align:middle line:84% they have everywhere is of good quality, 00:54:20.030 --> 00:54:23.060 align:middle line:90% but people do not believe it. 00:54:23.060 --> 00:54:24.440 align:middle line:84% And they are trying to figure out 00:54:24.440 --> 00:54:30.680 align:middle line:84% how to increase the trust in the quality of the malaria medicine 00:54:30.680 --> 00:54:34.230 align:middle line:84% in a setting, where nobody trusts it in the first place. 00:54:34.230 --> 00:54:38.390 align:middle line:84% So that's an example of, I think, 00:54:38.390 --> 00:54:40.430 align:middle line:84% partially successful attempt, but not 00:54:40.430 --> 00:54:42.590 align:middle line:84% fully successful attempt of making consumer more 00:54:42.590 --> 00:54:43.670 align:middle line:90% sophisticated. 00:54:43.670 --> 00:54:47.030 align:middle line:84% I'm not saying this shows this is not a path, 00:54:47.030 --> 00:54:50.000 align:middle line:84% but it shows that there is more work to be done, 00:54:50.000 --> 00:54:56.390 align:middle line:84% maybe in conjunction with the supply chain effort, maybe 00:54:56.390 --> 00:54:58.638 align:middle line:84% you need to work at larger scale. 00:54:58.638 --> 00:54:59.180 align:middle line:90% I don't know. 00:54:59.180 --> 00:55:02.400 align:middle line:90% This remains quite open. 00:55:02.400 --> 00:55:07.710 align:middle line:84% Last one is working with a private health provider. 00:55:07.710 --> 00:55:16.250 align:middle line:84% So this is a project, which was done by Abhijit, Jishnu 00:55:16.250 --> 00:55:20.960 align:middle line:84% Das, Reshma Hussam, who was a student here at the time, 00:55:20.960 --> 00:55:29.580 align:middle line:84% and Abhijit Chowdhury, who is a doctor in West Bengal. 00:55:29.580 --> 00:55:32.910 align:middle line:84% And what it did is that, so it was mainly 00:55:32.910 --> 00:55:36.270 align:middle line:84% the Abhijit Chowdhury is a very well-known liver specialist, 00:55:36.270 --> 00:55:39.210 align:middle line:84% but on the side of that, he has an NGO 00:55:39.210 --> 00:55:41.130 align:middle line:84% that provides health care to people 00:55:41.130 --> 00:55:44.270 align:middle line:84% in [INAUDIBLE],, in a hospital there. 00:55:44.270 --> 00:55:47.900 align:middle line:84% And what these people did, called The Liver Foundation, 00:55:47.900 --> 00:55:50.900 align:middle line:84% is that they set up together a training program for the health 00:55:50.900 --> 00:55:53.220 align:middle line:84% care for the private doctor, saying, look, 00:55:53.220 --> 00:55:56.690 align:middle line:84% they are going to be there anyways. 00:55:56.690 --> 00:56:00.055 align:middle line:84% People trust them, people go to them, 00:56:00.055 --> 00:56:02.960 align:middle line:84% and they have a smaller know-do gap than the public doctors. 00:56:02.960 --> 00:56:07.520 align:middle line:84% Can we just try to train them, both to recognize what they 00:56:07.520 --> 00:56:13.530 align:middle line:84% cannot deal with in triage and deal with a number 00:56:13.530 --> 00:56:14.250 align:middle line:90% of the medicines? 00:56:14.250 --> 00:56:19.720 align:middle line:84% This is something that the medical profession absolutely 00:56:19.720 --> 00:56:21.010 align:middle line:90% hate. 00:56:21.010 --> 00:56:22.990 align:middle line:84% The idea that you would, in some sense, 00:56:22.990 --> 00:56:28.630 align:middle line:84% train and therefore, legitimate the quacks, so to speak, 00:56:28.630 --> 00:56:30.540 align:middle line:84% is something that is very, very, very 00:56:30.540 --> 00:56:33.220 align:middle line:84% unpopular with the regular medical profession, 00:56:33.220 --> 00:56:35.620 align:middle line:90% but they try it anyways. 00:56:35.620 --> 00:56:37.930 align:middle line:84% The main outcomes where quality of care as 00:56:37.930 --> 00:56:40.900 align:middle line:84% measured by the same standardized patient, as I 00:56:40.900 --> 00:56:43.000 align:middle line:90% showed you from Madhya Pradesh. 00:56:43.000 --> 00:56:44.740 align:middle line:90% And what do they find? 00:56:44.740 --> 00:56:46.750 align:middle line:84% The result as presented in science, 00:56:46.750 --> 00:56:51.220 align:middle line:84% they published in science, so we have nice, compact presentation 00:56:51.220 --> 00:56:53.840 align:middle line:90% of results. 00:56:53.840 --> 00:56:57.970 align:middle line:84% So first of all, the training was free, 00:56:57.970 --> 00:57:00.100 align:middle line:90% but it was very demanding. 00:57:00.100 --> 00:57:02.110 align:middle line:84% It was over several weeks, et cetera. 00:57:02.110 --> 00:57:05.800 align:middle line:84% They went to a lot of cases, and it was a bit long, 00:57:05.800 --> 00:57:08.320 align:middle line:84% so people had to take half a day of work away. 00:57:08.320 --> 00:57:12.010 align:middle line:84% So the attendance was only 56% on average. 00:57:12.010 --> 00:57:13.810 align:middle line:84% At the end of the day, they got nothing. 00:57:13.810 --> 00:57:15.550 align:middle line:84% They didn't even get a certificate 00:57:15.550 --> 00:57:18.280 align:middle line:84% because the medical profession hate the idea of certificates. 00:57:18.280 --> 00:57:22.680 align:middle line:84% So they could tell the patient have been trained, 00:57:22.680 --> 00:57:26.670 align:middle line:84% but they didn't have something to prove it. 00:57:26.670 --> 00:57:32.450 align:middle line:84% And so 56% of people showed up, and on average, they 00:57:32.450 --> 00:57:34.700 align:middle line:84% trained better in the treatment group 00:57:34.700 --> 00:57:37.340 align:middle line:84% than in the control group, bridging about half the gap 00:57:37.340 --> 00:57:40.490 align:middle line:90% with the public sector. 00:57:40.490 --> 00:57:43.320 align:middle line:90% 00:57:43.320 --> 00:57:46.020 align:middle line:84% And those who managed to go all the way till the end, 00:57:46.020 --> 00:57:46.800 align:middle line:90% improved more. 00:57:46.800 --> 00:57:50.960 align:middle line:84% This is not experimental anymore. 00:57:50.960 --> 00:57:53.180 align:middle line:84% Unfortunately, the training had no impact 00:57:53.180 --> 00:57:56.060 align:middle line:84% on the avoidance of unnecessary antibiotics. 00:57:56.060 --> 00:58:00.860 align:middle line:84% So what happened is that they performed, 00:58:00.860 --> 00:58:03.290 align:middle line:84% they identified better what the client has. 00:58:03.290 --> 00:58:05.600 align:middle line:84% They were more likely to provide correct treatment, 00:58:05.600 --> 00:58:08.000 align:middle line:84% but on top of that, they throw in some antibiotic 00:58:08.000 --> 00:58:12.270 align:middle line:90% as well to round up the visit. 00:58:12.270 --> 00:58:16.170 align:middle line:84% So too, interesting to say what happened, what is interesting 00:58:16.170 --> 00:58:21.263 align:middle line:84% is that when you look at what they know, just 00:58:21.263 --> 00:58:22.680 align:middle line:84% from the vignette, they don't seem 00:58:22.680 --> 00:58:26.400 align:middle line:84% to know much more in treatment and in control, 00:58:26.400 --> 00:58:28.650 align:middle line:84% despite the fact that they've just been trained. 00:58:28.650 --> 00:58:32.940 align:middle line:84% But what happened is that they do more. 00:58:32.940 --> 00:58:36.150 align:middle line:84% So basically, the treatment helps bridge the know-do gap. 00:58:36.150 --> 00:58:43.110 align:middle line:84% They do more of what they know and so in the treatment, guys, 00:58:43.110 --> 00:58:48.580 align:middle line:84% complete my time in the checklist. 00:58:48.580 --> 00:58:52.420 align:middle line:84% And what you see here is that, this is your correct treatment 00:58:52.420 --> 00:58:55.800 align:middle line:84% as a function of the number of checklist items you complete. 00:58:55.800 --> 00:58:58.300 align:middle line:84% And basically, the more items you complete in the checklist, 00:58:58.300 --> 00:59:01.990 align:middle line:84% the more you are likely to end up on the correct treatment. 00:59:01.990 --> 00:59:04.390 align:middle line:84% That's for a standardized patient. 00:59:04.390 --> 00:59:06.640 align:middle line:84% And that function didn't really change much 00:59:06.640 --> 00:59:08.750 align:middle line:84% in the treatment versus the control group. 00:59:08.750 --> 00:59:12.160 align:middle line:84% So it's not that they changed their production function, 00:59:12.160 --> 00:59:14.890 align:middle line:84% it's just that the treatment moved this density 00:59:14.890 --> 00:59:21.040 align:middle line:84% to the right, so they just put a bit more either effort or just 00:59:21.040 --> 00:59:24.130 align:middle line:84% knowing this is what I have to do for this particular patient, 00:59:24.130 --> 00:59:26.750 align:middle line:84% not knowing, but actually doing it. 00:59:26.750 --> 00:59:29.500 align:middle line:84% So basically, the treatment reinforced for them 00:59:29.500 --> 00:59:32.380 align:middle line:84% the importance of actually doing what they were supposed to do. 00:59:32.380 --> 00:59:34.300 align:middle line:84% It didn't change their propensity 00:59:34.300 --> 00:59:39.670 align:middle line:84% to treat with medication, and that 00:59:39.670 --> 00:59:41.470 align:middle line:84% is almost surely because they are 00:59:41.470 --> 00:59:43.480 align:middle line:84% paid by a function of the number of medication 00:59:43.480 --> 00:59:45.130 align:middle line:84% they give, and therefore, this would 00:59:45.130 --> 00:59:50.080 align:middle line:84% have been a too expensive proposition 00:59:50.080 --> 00:59:54.350 align:middle line:84% to not throw in a few expensive proposition. 00:59:54.350 --> 00:59:56.650 align:middle line:84% So to sum up, the private sector does 00:59:56.650 --> 01:00:00.590 align:middle line:84% provide low standard of care, the public sector as well. 01:00:00.590 --> 01:00:02.535 align:middle line:84% And on top of that, the public sector 01:00:02.535 --> 01:00:04.660 align:middle line:84% provides poor service, which is what people prefer, 01:00:04.660 --> 01:00:06.880 align:middle line:90% the private sector. 01:00:06.880 --> 01:00:09.580 align:middle line:84% Shutting down completely the unqualified private sector 01:00:09.580 --> 01:00:11.350 align:middle line:84% is the option that in a lot of cases, 01:00:11.350 --> 01:00:14.650 align:middle line:84% that's what the medical profession pushes, 01:00:14.650 --> 01:00:19.180 align:middle line:84% which is not realistic, not consistent with the number 01:00:19.180 --> 01:00:22.030 align:middle line:84% of doctors you can train in a particular year, 01:00:22.030 --> 01:00:26.290 align:middle line:84% and not consistent with the fact that doctors do not 01:00:26.290 --> 01:00:28.590 align:middle line:90% want to go in rural places. 01:00:28.590 --> 01:00:31.860 align:middle line:84% And therefore, unless you force them to go there, usually 01:00:31.860 --> 01:00:34.060 align:middle line:84% after medical college, most developing countries, 01:00:34.060 --> 01:00:36.030 align:middle line:84% you need to go for a year, but you 01:00:36.030 --> 01:00:38.700 align:middle line:84% would need to force people to go for 10 years, 01:00:38.700 --> 01:00:44.180 align:middle line:84% and nobody is very interested in that regulation. 01:00:44.180 --> 01:00:45.820 align:middle line:90% So what to do then? 01:00:45.820 --> 01:00:49.120 align:middle line:84% So mostly, there is lack of training in the private sector. 01:00:49.120 --> 01:00:50.950 align:middle line:84% We've seen that, but mostly, people 01:00:50.950 --> 01:00:53.860 align:middle line:90% don't even do what they know. 01:00:53.860 --> 01:00:55.870 align:middle line:84% And part of the problem is people 01:00:55.870 --> 01:00:57.910 align:middle line:90% don't necessarily trust them. 01:00:57.910 --> 01:00:59.410 align:middle line:84% They are skeptical of their advice. 01:00:59.410 --> 01:01:01.118 align:middle line:84% They know that they are not well trained. 01:01:01.118 --> 01:01:03.400 align:middle line:84% They suspect them to be corrupt, so they 01:01:03.400 --> 01:01:07.370 align:middle line:84% stay within their capacity as much as possible. 01:01:07.370 --> 01:01:11.480 align:middle line:84% So some certification and other help 01:01:11.480 --> 01:01:15.550 align:middle line:84% in building a reputation for being a good guy might help. 01:01:15.550 --> 01:01:16.900 align:middle line:90% Some checklists. 01:01:16.900 --> 01:01:21.940 align:middle line:84% So I don't if you've seen, there is a book called Checklist 01:01:21.940 --> 01:01:26.800 align:middle line:84% or something like that, which is actually from a former doctor. 01:01:26.800 --> 01:01:28.510 align:middle line:84% The book is not just about doctor, 01:01:28.510 --> 01:01:31.040 align:middle line:84% but it started with applying checking in hospital. 01:01:31.040 --> 01:01:33.580 align:middle line:84% So even in the US, introducing checklists 01:01:33.580 --> 01:01:36.250 align:middle line:84% increases the quality of care a lot. 01:01:36.250 --> 01:01:39.040 align:middle line:84% You would think that for a lot of these cases, 01:01:39.040 --> 01:01:42.460 align:middle line:84% like a tool to build a checklist would help. 01:01:42.460 --> 01:01:44.950 align:middle line:84% I don't know if people have tried that. 01:01:44.950 --> 01:01:48.910 align:middle line:84% And then not tying the revenue model with selling antibiotic 01:01:48.910 --> 01:01:53.090 align:middle line:84% would be good, and forcing some regulation might work well. 01:01:53.090 --> 01:01:56.050 align:middle line:90% 01:01:56.050 --> 01:01:58.460 align:middle line:84% Not the regulation saying cannot exercise, 01:01:58.460 --> 01:02:02.815 align:middle line:84% which is not possible, but you have to get a test on this 01:02:02.815 --> 01:02:04.690 align:middle line:84% vignette, and then you can get a certificate, 01:02:04.690 --> 01:02:06.840 align:middle line:84% which you can show on your window. 01:02:06.840 --> 01:02:10.690 align:middle line:84% So that would increase their training and their regulation. 01:02:10.690 --> 01:02:13.360 align:middle line:84% Requiring them to attend training on basic patient 01:02:13.360 --> 01:02:16.690 align:middle line:84% safety to limit the damage they can make, for example, 01:02:16.690 --> 01:02:19.120 align:middle line:84% not sharing needles, sterilization, CPR, 01:02:19.120 --> 01:02:20.020 align:middle line:90% and the like. 01:02:20.020 --> 01:02:23.380 align:middle line:84% And involve them in public health campaign, 01:02:23.380 --> 01:02:25.130 align:middle line:84% because they are connected to the patient. 01:02:25.130 --> 01:02:29.890 align:middle line:84% So in fact, during the COVID crisis, 01:02:29.890 --> 01:02:32.860 align:middle line:84% there was an effort in West Bengal 01:02:32.860 --> 01:02:34.397 align:middle line:84% and maybe in other places as well. 01:02:34.397 --> 01:02:35.980 align:middle line:84% But in West Bengal there was an effort 01:02:35.980 --> 01:02:40.990 align:middle line:84% to include the unqualified local provider 01:02:40.990 --> 01:02:43.690 align:middle line:84% as first screening and first response 01:02:43.690 --> 01:02:50.560 align:middle line:84% to help people deal with COVID during the various several 01:02:50.560 --> 01:02:53.710 align:middle line:90% waves that India had. 01:02:53.710 --> 01:02:54.970 align:middle line:90% What it did, I don't know. 01:02:54.970 --> 01:02:58.630 align:middle line:84% I haven't seen the study yet, but there was actually 01:02:58.630 --> 01:03:01.920 align:middle line:90% one effort along these lines. 01:03:01.920 --> 01:03:04.350 align:middle line:84% So that's what I had to say on the supply of health care. 01:03:04.350 --> 01:03:06.530 align:middle line:84% I'm happy to take questions, and then, otherwise, I 01:03:06.530 --> 01:03:12.260 align:middle line:84% can move on and complete what Ben 01:03:12.260 --> 01:03:14.180 align:middle line:90% didn't get to in his lecture. 01:03:14.180 --> 01:03:15.590 align:middle line:90% So yeah. 01:03:15.590 --> 01:03:18.140 align:middle line:84% AUDIENCE: Can you say a little more about the medical college 01:03:18.140 --> 01:03:18.640 align:middle line:90% thing? 01:03:18.640 --> 01:03:22.850 align:middle line:84% Like in the US, it's like the restriction of doctor's writing 01:03:22.850 --> 01:03:24.170 align:middle line:90% or whatever it is? 01:03:24.170 --> 01:03:27.308 align:middle line:84% What is the situation in terms of access to medical education? 01:03:27.308 --> 01:03:28.100 align:middle line:90% ESTHER DUFLO: Yeah. 01:03:28.100 --> 01:03:30.740 align:middle line:84% So like in the US, it's very restrictive. 01:03:30.740 --> 01:03:33.980 align:middle line:84% In the countries I know, it's very restricted and very 01:03:33.980 --> 01:03:36.530 align:middle line:84% competitive, so this is true in East Africa. 01:03:36.530 --> 01:03:37.538 align:middle line:90% It's true in Ghana. 01:03:37.538 --> 01:03:38.330 align:middle line:90% It's true in India. 01:03:38.330 --> 01:03:42.870 align:middle line:90% 01:03:42.870 --> 01:03:45.150 align:middle line:84% Partly for the same reason, which is the doctors 01:03:45.150 --> 01:03:49.200 align:middle line:84% have this way of enforcing their own reality, 01:03:49.200 --> 01:03:54.010 align:middle line:84% partly, also because it's hard to train doctors. 01:03:54.010 --> 01:03:59.710 align:middle line:84% And like training primary school teachers, that is relatively-- 01:03:59.710 --> 01:04:01.960 align:middle line:84% it's actually, very hard to train doctors. 01:04:01.960 --> 01:04:06.120 align:middle line:84% And to train doctors, you also need 01:04:06.120 --> 01:04:08.220 align:middle line:84% doctors who are willing to train other doctors 01:04:08.220 --> 01:04:13.957 align:middle line:84% and taking some time off of being a doctor, and presumably 01:04:13.957 --> 01:04:16.290 align:middle line:84% good doctors, where they are both needed where they are, 01:04:16.290 --> 01:04:17.748 align:middle line:84% and they could make a lot of money, 01:04:17.748 --> 01:04:20.070 align:middle line:90% so that's basically a bind. 01:04:20.070 --> 01:04:22.980 align:middle line:84% So there is this kind of inherent limit, 01:04:22.980 --> 01:04:27.990 align:middle line:84% not just because of putting a cap, numerus clausus, that you 01:04:27.990 --> 01:04:30.960 align:middle line:84% find in rich countries, France has the same thing, but also, 01:04:30.960 --> 01:04:33.300 align:middle line:84% because of the technology of producing doctors 01:04:33.300 --> 01:04:36.900 align:middle line:84% who are highly qualified people who, therefore, have 01:04:36.900 --> 01:04:39.700 align:middle line:90% a lot of other possibilities. 01:04:39.700 --> 01:04:42.220 align:middle line:84% Very much like we discussed for the case of secondary school 01:04:42.220 --> 01:04:43.255 align:middle line:90% teacher, but even more. 01:04:43.255 --> 01:04:45.940 align:middle line:90% 01:04:45.940 --> 01:04:48.370 align:middle line:84% Incidentally, this is something that it 01:04:48.370 --> 01:04:52.240 align:middle line:84% has a parallel in the US with the training of nurses. 01:04:52.240 --> 01:04:56.430 align:middle line:84% Where despite the fact that there is a shortage of nurses, 01:04:56.430 --> 01:04:59.370 align:middle line:84% there is a huge excess supply in the number 01:04:59.370 --> 01:05:02.460 align:middle line:84% of people who would like to be nurse, 01:05:02.460 --> 01:05:05.250 align:middle line:84% to apply for nursing degrees, who 01:05:05.250 --> 01:05:07.080 align:middle line:90% can't get to nursing schools. 01:05:07.080 --> 01:05:10.248 align:middle line:84% And one of the reasons why they can't get into nursing school 01:05:10.248 --> 01:05:11.790 align:middle line:84% is that there is nobody to teach them 01:05:11.790 --> 01:05:14.700 align:middle line:84% to be a nurse, because the nurses who are in exercise 01:05:14.700 --> 01:05:16.590 align:middle line:84% are required to actually be nurses. 01:05:16.590 --> 01:05:19.557 align:middle line:84% And they can make much, much more money trying to be nurses 01:05:19.557 --> 01:05:20.640 align:middle line:90% or training to be teacher. 01:05:20.640 --> 01:05:26.190 align:middle line:84% So they need to be a reform of the teaching sectors for nurses 01:05:26.190 --> 01:05:30.090 align:middle line:84% in the US, and I think it's a little bit similar. 01:05:30.090 --> 01:05:32.100 align:middle line:84% There's the extra problem in India 01:05:32.100 --> 01:05:33.930 align:middle line:84% that all the nurses that are trained 01:05:33.930 --> 01:05:39.267 align:middle line:84% are then absorbed in the US, because it's one sector where 01:05:39.267 --> 01:05:41.850 align:middle line:84% at least pre-COVID, I don't know if they're being liberalized, 01:05:41.850 --> 01:05:44.850 align:middle line:84% but being an immigrant, immigrating to be a nurse was 01:05:44.850 --> 01:05:47.880 align:middle line:90% always very easy in the US. 01:05:47.880 --> 01:05:50.310 align:middle line:84% That puts one more set of pressure 01:05:50.310 --> 01:05:56.730 align:middle line:84% on the supply of qualified nurses in developing countries. 01:05:56.730 --> 01:05:58.530 align:middle line:84% And also, on the outside option, you're 01:05:58.530 --> 01:06:00.960 align:middle line:90% asking what does a nurse do? 01:06:00.960 --> 01:06:06.480 align:middle line:84% Well, the people who are managing them perfectly 01:06:06.480 --> 01:06:09.390 align:middle line:84% know that the outside option is leaving and going 01:06:09.390 --> 01:06:10.150 align:middle line:90% somewhere else. 01:06:10.150 --> 01:06:13.350 align:middle line:84% And there is also, in this whether, that's 01:06:13.350 --> 01:06:16.890 align:middle line:84% the reason why nurses are treated 01:06:16.890 --> 01:06:20.835 align:middle line:84% well in India or in Uganda or elsewhere. 01:06:20.835 --> 01:06:22.950 align:middle line:84% Are we putting too much pressure on them, 01:06:22.950 --> 01:06:25.440 align:middle line:84% would actually they would just pack up and go to the US 01:06:25.440 --> 01:06:27.270 align:middle line:90% and make money. 01:06:27.270 --> 01:06:33.890 align:middle line:90% 01:06:33.890 --> 01:06:34.510 align:middle line:90% Great. 01:06:34.510 --> 01:06:36.430 align:middle line:84% So that's the end of the health lecture. 01:06:36.430 --> 01:06:37.720 align:middle line:90% I'm saying that to the camera. 01:06:37.720 --> 01:06:42.054 align:middle line:84% I'm going to now move to talk about gender and the feminine. 01:06:42.054 --> 01:06:45.020 align:middle line:90% 01:06:45.020 --> 01:06:49.940 align:middle line:84% And I just looked at what Ben had time to cover, 01:06:49.940 --> 01:06:54.350 align:middle line:84% and I figured what would make sense is to-- 01:06:54.350 --> 01:07:00.380 align:middle line:90% 01:07:00.380 --> 01:07:04.310 align:middle line:84% I made my own slides, but cover some of the papers 01:07:04.310 --> 01:07:06.770 align:middle line:84% that were at the end of his slides 01:07:06.770 --> 01:07:10.480 align:middle line:84% and that he didn't have time to get to. 01:07:10.480 --> 01:07:16.250 align:middle line:84% So a very natural segue into talking 01:07:16.250 --> 01:07:19.130 align:middle line:84% about gender issues, which I'm going to do, 01:07:19.130 --> 01:07:24.000 align:middle line:90% then finish on Monday. 01:07:24.000 --> 01:07:28.230 align:middle line:84% So you finished, last time, the lecture 01:07:28.230 --> 01:07:30.810 align:middle line:84% on talking about Chris Udry's paper, which 01:07:30.810 --> 01:07:33.540 align:middle line:84% is among the classic, showing that there 01:07:33.540 --> 01:07:36.490 align:middle line:90% is no efficiency in production. 01:07:36.490 --> 01:07:38.620 align:middle line:84% One more thing we could ask is whether there 01:07:38.620 --> 01:07:42.790 align:middle line:90% is efficiency in consumption. 01:07:42.790 --> 01:07:45.910 align:middle line:84% If households were able to insure 01:07:45.910 --> 01:07:50.500 align:middle line:84% each other with full commitment, that is they are able to-- 01:07:50.500 --> 01:07:53.200 align:middle line:84% if we had an efficient household model, 01:07:53.200 --> 01:07:58.720 align:middle line:84% efficiency requires that conditional on total household 01:07:58.720 --> 01:08:02.890 align:middle line:84% demand for eating, who gets what and what 01:08:02.890 --> 01:08:04.810 align:middle line:84% goods are being consumed and by whom, 01:08:04.810 --> 01:08:07.570 align:middle line:84% should be invariant of the actual variation 01:08:07.570 --> 01:08:10.610 align:middle line:84% because you know you're here for a long time. 01:08:10.610 --> 01:08:21.450 align:middle line:84% So even if I like potatoes and you like beef, 01:08:21.450 --> 01:08:23.130 align:middle line:84% the consumption of potato and beef 01:08:23.130 --> 01:08:24.569 align:middle line:84% should remain relatively constant. 01:08:24.569 --> 01:08:28.430 align:middle line:84% Because it's not because I earn a bit more money this period, 01:08:28.430 --> 01:08:30.140 align:middle line:84% next time it's going to be, and we 01:08:30.140 --> 01:08:34.109 align:middle line:84% should able to insure each other for this change. 01:08:34.109 --> 01:08:37.850 align:middle line:84% So the consumption of a particular commodity 01:08:37.850 --> 01:08:40.729 align:middle line:84% shouldn't depend on a shock that affects 01:08:40.729 --> 01:08:44.609 align:middle line:90% one person versus the other. 01:08:44.609 --> 01:08:46.950 align:middle line:84% In other words, the shocks should affect the consumption 01:08:46.950 --> 01:08:49.140 align:middle line:84% of each individual goods only to the extent 01:08:49.140 --> 01:08:50.800 align:middle line:84% that they affect total consumption. 01:08:50.800 --> 01:08:53.160 align:middle line:84% So the household might not be able to ensure 01:08:53.160 --> 01:08:55.229 align:middle line:90% it's entire set of consumption. 01:08:55.229 --> 01:09:00.797 align:middle line:84% That, in turn, is going to have implication what they consume. 01:09:00.797 --> 01:09:02.880 align:middle line:84% For example, if they are very poor, at some point, 01:09:02.880 --> 01:09:05.380 align:middle line:84% they might eat mostly potatoes because they don't have money 01:09:05.380 --> 01:09:07.950 align:middle line:84% for beef, but that should only transit from, 01:09:07.950 --> 01:09:10.859 align:middle line:84% that should not be over and above the impact 01:09:10.859 --> 01:09:12.450 align:middle line:84% of total consumption on the demand 01:09:12.450 --> 01:09:15.910 align:middle line:90% for the various kind of foods. 01:09:15.910 --> 01:09:18.399 align:middle line:84% So the consumption of the wife should not 01:09:18.399 --> 01:09:20.838 align:middle line:84% drop because of a particular blood draw 01:09:20.838 --> 01:09:22.630 align:middle line:84% and similarly, for the husband because they 01:09:22.630 --> 01:09:23.840 align:middle line:90% can enjoy each other later. 01:09:23.840 --> 01:09:26.080 align:middle line:84% So of course, you'll do much more of insurance 01:09:26.080 --> 01:09:27.529 align:middle line:90% in the spring semester. 01:09:27.529 --> 01:09:29.830 align:middle line:84% So it will become clear that it is correct, 01:09:29.830 --> 01:09:34.710 align:middle line:84% but the intrusion doesn't require that much. 01:09:34.710 --> 01:09:37.830 align:middle line:84% So household members, able to insure 01:09:37.830 --> 01:09:40.859 align:middle line:84% each other against transitory shock, and the answer to that 01:09:40.859 --> 01:09:41.800 align:middle line:90% is no. 01:09:41.800 --> 01:09:46.170 align:middle line:84% So it's first, I want to go to an experimental paper, 01:09:46.170 --> 01:09:48.630 align:middle line:90% the paper by John Robinson. 01:09:48.630 --> 01:09:50.399 align:middle line:84% In fact, it was his job market paper 01:09:50.399 --> 01:09:55.340 align:middle line:84% when he was in Princeton, graduate student. 01:09:55.340 --> 01:09:59.180 align:middle line:84% And it was a graduate student paper, so it was not very big, 01:09:59.180 --> 01:10:02.120 align:middle line:84% but he had 142 married couples and followed them 01:10:02.120 --> 01:10:05.390 align:middle line:90% for eight weeks in Kenya. 01:10:05.390 --> 01:10:09.080 align:middle line:84% And each week he met them and each individual in the couple 01:10:09.080 --> 01:10:15.800 align:middle line:84% had a 50% chance of receiving a 150 Shilling, 01:10:15.800 --> 01:10:18.200 align:middle line:84% equivalent to roughly 1 and 1/2 days 01:10:18.200 --> 01:10:23.300 align:middle line:84% income for men and one-week income for women, so something 01:10:23.300 --> 01:10:27.410 align:middle line:90% not irrelevant. 01:10:27.410 --> 01:10:29.890 align:middle line:84% So people received their payment, 01:10:29.890 --> 01:10:34.090 align:middle line:84% then both spouses knew what they had. 01:10:34.090 --> 01:10:37.810 align:middle line:84% And then what he did is that he got their data, 01:10:37.810 --> 01:10:43.210 align:middle line:84% very similar they did for their savings paper was basically-- 01:10:43.210 --> 01:10:50.710 align:middle line:84% in fact, same type of people doing cycle ratio of women 01:10:50.710 --> 01:10:55.960 align:middle line:84% selling on markets and stuff like that, this type of very 01:10:55.960 --> 01:11:00.910 align:middle line:84% regular work, did these diaries, And then he collected data 01:11:00.910 --> 01:11:03.535 align:middle line:84% on income and income shocks and labor supply from each member. 01:11:03.535 --> 01:11:06.750 align:middle line:90% 01:11:06.750 --> 01:11:09.210 align:middle line:84% So the shocks are like textbook shock. 01:11:09.210 --> 01:11:11.730 align:middle line:84% They are random, they are transitory, 01:11:11.730 --> 01:11:13.710 align:middle line:84% because every day, there is a new lottery, 01:11:13.710 --> 01:11:16.800 align:middle line:90% and they are idiosyncratic. 01:11:16.800 --> 01:11:19.470 align:middle line:84% They should not affect the bargaining power 01:11:19.470 --> 01:11:21.610 align:middle line:90% of each person. 01:11:21.610 --> 01:11:26.410 align:middle line:84% They are also public, so there is no incentive to lie. 01:11:26.410 --> 01:11:28.030 align:middle line:90% It is not possible anyways. 01:11:28.030 --> 01:11:29.380 align:middle line:90% There is no moral hazard. 01:11:29.380 --> 01:11:31.510 align:middle line:90% There is no labor supply. 01:11:31.510 --> 01:11:33.660 align:middle line:84% So you would expect complete sharing 01:11:33.660 --> 01:11:38.320 align:middle line:84% if the household was consumption efficient. 01:11:38.320 --> 01:11:42.320 align:middle line:84% And surprisingly, you don't, which is, 01:11:42.320 --> 01:11:46.890 align:middle line:84% people who get more money, they spend more. 01:11:46.890 --> 01:11:52.000 align:middle line:84% So when men get more money, also when a respondent gets 01:11:52.000 --> 01:11:53.500 align:middle line:90% more money-- it is now the men. 01:11:53.500 --> 01:11:55.960 align:middle line:84% If it's the respondent who get more money, 01:11:55.960 --> 01:11:58.370 align:middle line:84% they consume more on private good. 01:11:58.370 --> 01:11:59.690 align:middle line:90% And nothing else, basically. 01:11:59.690 --> 01:12:02.270 align:middle line:84% So they consume nothing else really change. 01:12:02.270 --> 01:12:07.270 align:middle line:84% So when they have money, when they get 1 Shilling more, 01:12:07.270 --> 01:12:12.700 align:middle line:84% they spend about 0.19 Shilling in total 01:12:12.700 --> 01:12:16.710 align:middle line:90% and all on private consumption. 01:12:16.710 --> 01:12:21.240 align:middle line:84% And for women, they don't do so much, so private consumption, 01:12:21.240 --> 01:12:26.190 align:middle line:84% like newspaper, snacks, bottled sodas, that kind of things. 01:12:26.190 --> 01:12:27.690 align:middle line:84% So women don't do that, but they buy 01:12:27.690 --> 01:12:30.135 align:middle line:84% themselves a lot of medical expenses, medical stuff. 01:12:30.135 --> 01:12:32.976 align:middle line:90% 01:12:32.976 --> 01:12:37.160 align:middle line:84% I know you might wonder, but they just got 1 Shilling. 01:12:37.160 --> 01:12:39.380 align:middle line:90% What happened to the rest? 01:12:39.380 --> 01:12:44.030 align:middle line:84% And the rest is saved, which makes a lot of sense, 01:12:44.030 --> 01:12:53.780 align:middle line:84% but it's not saved in the public way for the household. 01:12:53.780 --> 01:12:55.790 align:middle line:84% So you can see that it's saved total, 01:12:55.790 --> 01:12:58.865 align:middle line:84% but it's saved mostly in private-- 01:12:58.865 --> 01:13:00.980 align:middle line:84% I'm looking for the private note. 01:13:00.980 --> 01:13:04.960 align:middle line:84% It's saved mostly private vehicles. 01:13:04.960 --> 01:13:11.740 align:middle line:84% So they don't save in the form of their bank 01:13:11.740 --> 01:13:12.820 align:middle line:90% account, et cetera. 01:13:12.820 --> 01:13:16.900 align:middle line:84% They save basically, each under their own mattress. 01:13:16.900 --> 01:13:22.780 align:middle line:84% So when people get a shock, men save about 0.8 of it, 01:13:22.780 --> 01:13:25.540 align:middle line:84% and the rest is spent on themselves. 01:13:25.540 --> 01:13:27.290 align:middle line:90% Women save less of it. 01:13:27.290 --> 01:13:32.210 align:middle line:84% The rest is spent a little bit more or spread across things. 01:13:32.210 --> 01:13:34.720 align:middle line:84% In fact, they do buy some more food. 01:13:34.720 --> 01:13:40.060 align:middle line:84% They buy medical expenses, a little bit on the kids, 01:13:40.060 --> 01:13:43.400 align:middle line:90% so that's not consistent with-- 01:13:43.400 --> 01:13:51.130 align:middle line:90% And the men spend nothing-- 01:13:51.130 --> 01:13:54.970 align:middle line:84% when the spouse get more money, they get nothing. 01:13:54.970 --> 01:14:01.280 align:middle line:84% They don't transfer the money to the spouse, at least 01:14:01.280 --> 01:14:04.460 align:middle line:84% women transfer a bit of their money to the spouse. 01:14:04.460 --> 01:14:05.165 align:middle line:90% Men don't. 01:14:05.165 --> 01:14:08.560 align:middle line:90% 01:14:08.560 --> 01:14:11.400 align:middle line:84% But on the other hand, if their spouse give them money, 01:14:11.400 --> 01:14:14.830 align:middle line:84% they receive money, they give them lower transfers. 01:14:14.830 --> 01:14:17.830 align:middle line:84% So women and men behave a little bit differently, 01:14:17.830 --> 01:14:22.570 align:middle line:84% with women doing more of the shared things, but both of them 01:14:22.570 --> 01:14:25.360 align:middle line:84% basically save the majority of the money 01:14:25.360 --> 01:14:28.990 align:middle line:84% under their own private mattress, which 01:14:28.990 --> 01:14:45.180 align:middle line:84% is not what would be expected, given the efficient household 01:14:45.180 --> 01:14:45.680 align:middle line:90% model. 01:14:45.680 --> 01:14:49.510 align:middle line:84% So we already saw that they were not efficient in production, 01:14:49.510 --> 01:14:52.460 align:middle line:84% they are not efficient in consumption. 01:14:52.460 --> 01:14:55.340 align:middle line:84% So this is an experiment that is more experiment. 01:14:55.340 --> 01:14:58.750 align:middle line:84% Chris Udry and I have a paper famously unpublished. 01:14:58.750 --> 01:15:01.990 align:middle line:84% I can tell you one day what happened to that paper 01:15:01.990 --> 01:15:04.600 align:middle line:84% and why it's still a working paper. 01:15:04.600 --> 01:15:06.970 align:middle line:90% Maybe one day we'll publish it. 01:15:06.970 --> 01:15:09.430 align:middle line:90% It's in Cote d'Ivoire. 01:15:09.430 --> 01:15:13.360 align:middle line:84% It's the first paper I wrote after graduating, started 01:15:13.360 --> 01:15:16.180 align:middle line:84% a completely brand new paper I wrote after graduating, 01:15:16.180 --> 01:15:17.950 align:middle line:90% so to tell you it's old. 01:15:17.950 --> 01:15:21.100 align:middle line:84% And it uses the fact that women and men 01:15:21.100 --> 01:15:23.590 align:middle line:84% tend to grow different crops on their different farms. 01:15:23.590 --> 01:15:25.750 align:middle line:84% So you've already seen that in the Udry paper 01:15:25.750 --> 01:15:28.000 align:middle line:84% that they grow on different farms, but on top of that, 01:15:28.000 --> 01:15:29.620 align:middle line:90% they grow different crop. 01:15:29.620 --> 01:15:34.090 align:middle line:84% So they are women crop, men crop, and then yams. 01:15:34.090 --> 01:15:36.610 align:middle line:84% So yams are male crops, but they have 01:15:36.610 --> 01:15:41.470 align:middle line:84% to be used for feeding people, either by being directly eaten 01:15:41.470 --> 01:15:44.060 align:middle line:90% or by being sold to buy food. 01:15:44.060 --> 01:15:50.690 align:middle line:84% And there is a very, very strong social norm in this direction. 01:15:50.690 --> 01:15:55.630 align:middle line:84% So in Cote d'Ivoire, we know what people produce 01:15:55.630 --> 01:15:58.400 align:middle line:84% and so you can compute proxies for mail. 01:15:58.400 --> 01:16:00.280 align:middle line:84% You don't know what every individual produce, 01:16:00.280 --> 01:16:02.500 align:middle line:84% but you know what the households produces 01:16:02.500 --> 01:16:05.540 align:middle line:84% and how much money they make from their different crops. 01:16:05.540 --> 01:16:08.440 align:middle line:84% So you can compute proxies for male income, female income, 01:16:08.440 --> 01:16:11.620 align:middle line:84% and yam income by aggregating crop income 01:16:11.620 --> 01:16:13.120 align:middle line:84% across the different crops, knowing 01:16:13.120 --> 01:16:15.190 align:middle line:84% that probably the way that it's done in practice 01:16:15.190 --> 01:16:18.990 align:middle line:90% is different from the-- 01:16:18.990 --> 01:16:23.190 align:middle line:84% So this Cote d'Ivoire survey is a panel. 01:16:23.190 --> 01:16:25.195 align:middle line:84% In fact, it's an overlapping panel 01:16:25.195 --> 01:16:26.820 align:middle line:84% there are three years of it, but people 01:16:26.820 --> 01:16:30.710 align:middle line:84% are appearing there in year 1 and 2 are in year 2 and 3. 01:16:30.710 --> 01:16:34.310 align:middle line:84% So for each household, we stack the two, so for each household, 01:16:34.310 --> 01:16:36.500 align:middle line:90% we see them twice. 01:16:36.500 --> 01:16:39.410 align:middle line:84% We look at the change in their income 01:16:39.410 --> 01:16:45.140 align:middle line:84% from one year to the next, and then we predict. 01:16:45.140 --> 01:16:46.720 align:middle line:90% So this done in an OLS. 01:16:46.720 --> 01:16:48.440 align:middle line:84% It's done in a two-state least square, 01:16:48.440 --> 01:16:52.730 align:middle line:84% but imagine for the sake of pedagogy that we use 01:16:52.730 --> 01:16:54.740 align:middle line:90% predicted income. 01:16:54.740 --> 01:16:57.500 align:middle line:84% So we predict for each person how much 01:16:57.500 --> 01:16:59.090 align:middle line:84% better they are doing year 2 compared 01:16:59.090 --> 01:17:02.660 align:middle line:84% to year 1, the difference in year 2 income to year 1 income, 01:17:02.660 --> 01:17:05.480 align:middle line:84% for female income, male income, and yam income, 01:17:05.480 --> 01:17:11.480 align:middle line:84% and we run a regression of the change in consumption 01:17:11.480 --> 01:17:15.340 align:middle line:84% on the predicted female income, male income, and yam income. 01:17:15.340 --> 01:17:18.840 align:middle line:84% So the question is, in a power efficient model, 01:17:18.840 --> 01:17:24.670 align:middle line:84% would we expect any difference between beta, gamma, and delta, 01:17:24.670 --> 01:17:25.540 align:middle line:90% potentially? 01:17:25.540 --> 01:17:30.910 align:middle line:90% 01:17:30.910 --> 01:17:33.970 align:middle line:84% So we look at consumption of various goods, 01:17:33.970 --> 01:17:36.490 align:middle line:84% so we're going to look at food, private expenditure, 01:17:36.490 --> 01:17:37.750 align:middle line:90% public expenditure. 01:17:37.750 --> 01:17:43.460 align:middle line:90% 01:17:43.460 --> 01:17:46.070 align:middle line:84% Well, I would normally wait, but it's 11:55, 01:17:46.070 --> 01:17:47.690 align:middle line:90% so you're saved by the bell. 01:17:47.690 --> 01:17:54.420 align:middle line:84% You would because there might have different time pass. 01:17:54.420 --> 01:17:57.380 align:middle line:84% So for example, if a lot of the male income 01:17:57.380 --> 01:18:00.278 align:middle line:84% is coffee and the fact that you get good income today is 01:18:00.278 --> 01:18:02.195 align:middle line:84% probably a good prediction that you'll to get, 01:18:02.195 --> 01:18:06.800 align:middle line:84% again, good income tomorrow because that means your trees 01:18:06.800 --> 01:18:07.880 align:middle line:90% are doing well. 01:18:07.880 --> 01:18:11.720 align:middle line:84% And if the women are producing vegetable, this is perennial. 01:18:11.720 --> 01:18:14.240 align:middle line:90% This may be less information in. 01:18:14.240 --> 01:18:17.450 align:middle line:84% So we know that from the savings group 01:18:17.450 --> 01:18:20.420 align:middle line:84% that people should save transitory income 01:18:20.420 --> 01:18:22.440 align:middle line:90% and consume permanent income. 01:18:22.440 --> 01:18:26.060 align:middle line:84% So to the extent that there is different income 01:18:26.060 --> 01:18:27.530 align:middle line:84% process for these three groups, we 01:18:27.530 --> 01:18:32.030 align:middle line:84% might expect a different effect on Total consumption, 01:18:32.030 --> 01:18:35.270 align:middle line:84% and therefore, we might expect different impact 01:18:35.270 --> 01:18:39.780 align:middle line:90% on individual goods consumption. 01:18:39.780 --> 01:18:42.140 align:middle line:84% So what this suggests is that the consumption 01:18:42.140 --> 01:18:45.140 align:middle line:84% of any particular goods should change, but only to the extent 01:18:45.140 --> 01:18:49.270 align:middle line:90% that total expenditure change. 01:18:49.270 --> 01:18:51.030 align:middle line:84% So you want to run the same regression 01:18:51.030 --> 01:18:54.180 align:middle line:84% with total expenditure as the dependent variable 01:18:54.180 --> 01:18:57.330 align:middle line:84% and then calculate this ratio, and basically, they 01:18:57.330 --> 01:18:58.035 align:middle line:90% should be equal. 01:18:58.035 --> 01:19:00.565 align:middle line:90% 01:19:00.565 --> 01:19:03.880 align:middle line:84% And I'm sure there is a match, should we now 01:19:03.880 --> 01:19:05.140 align:middle line:90% decide to publish this paper. 01:19:05.140 --> 01:19:07.960 align:middle line:84% Instead, we would do GMM or some fancy thing 01:19:07.960 --> 01:19:09.710 align:middle line:84% like that, which would come to that. 01:19:09.710 --> 01:19:13.720 align:middle line:84% Which is basically saying, well, if female income, 01:19:13.720 --> 01:19:15.610 align:middle line:84% as a coefficient on total consumption, 01:19:15.610 --> 01:19:17.350 align:middle line:84% that is larger than male income, then 01:19:17.350 --> 01:19:20.590 align:middle line:84% it should also have a larger effect on everything else, 01:19:20.590 --> 01:19:21.700 align:middle line:90% but that's it. 01:19:21.700 --> 01:19:25.230 align:middle line:84% So all these ratios should be the same, 01:19:25.230 --> 01:19:26.960 align:middle line:84% and they're not is the conclusion. 01:19:26.960 --> 01:19:28.080 align:middle line:90% They are not at all. 01:19:28.080 --> 01:19:30.320 align:middle line:84% So the male income, yam income, female income 01:19:30.320 --> 01:19:32.360 align:middle line:84% is dependent on totally different ways 01:19:32.360 --> 01:19:39.140 align:middle line:84% and in ways that is consistent with the income are being sent 01:19:39.140 --> 01:19:42.620 align:middle line:84% to the various streams of income are being kept separate 01:19:42.620 --> 01:19:46.800 align:middle line:84% and being sent under various purpose. 01:19:46.800 --> 01:19:49.650 align:middle line:90% I'll just show you one slide. 01:19:49.650 --> 01:19:53.160 align:middle line:84% So these are all of the coefficients themselves. 01:19:53.160 --> 01:19:55.350 align:middle line:90% Total expenditure. 01:19:55.350 --> 01:20:00.060 align:middle line:84% Female income is slightly more responsibility to female income 01:20:00.060 --> 01:20:02.970 align:middle line:84% than to non-yam and to yam income. 01:20:02.970 --> 01:20:05.070 align:middle line:90% That's just what it is. 01:20:05.070 --> 01:20:07.230 align:middle line:84% But then if you're looking at the various goods, 01:20:07.230 --> 01:20:10.530 align:middle line:84% for example, the goods that are prestige goods, so these are 01:20:10.530 --> 01:20:12.390 align:middle line:84% stuff that people buy for themselves, 01:20:12.390 --> 01:20:15.660 align:middle line:84% like ceremonial clothes and that kind of things, 01:20:15.660 --> 01:20:22.410 align:middle line:84% are being spent, not that yam income has no impact on that. 01:20:22.410 --> 01:20:24.930 align:middle line:84% It's each individual income, female and males, 01:20:24.930 --> 01:20:26.650 align:middle line:90% both spend a lot of-- 01:20:26.650 --> 01:20:29.530 align:middle line:84% have a large propensity to spend out of female income. 01:20:29.530 --> 01:20:32.700 align:middle line:84% Now, if you look at the yam income, 01:20:32.700 --> 01:20:36.090 align:middle line:84% it's being spent on staple more than anything else. 01:20:36.090 --> 01:20:39.750 align:middle line:84% You might think it makes sense because they just 01:20:39.750 --> 01:20:41.710 align:middle line:90% eat the yam they have there. 01:20:41.710 --> 01:20:44.580 align:middle line:84% But it is also spent on stuff that's not a staple, 01:20:44.580 --> 01:20:49.450 align:middle line:84% but that's common goods, and in particular, education. 01:20:49.450 --> 01:20:53.240 align:middle line:84% So people spend their yam income on education and on staple. 01:20:53.240 --> 01:20:57.670 align:middle line:84% They spend their private income on prestige good. 01:20:57.670 --> 01:21:00.820 align:middle line:84% Women spend a lot of their income on food. 01:21:00.820 --> 01:21:03.430 align:middle line:84% Of course, they grow vegetables, so maybe that's 01:21:03.430 --> 01:21:06.220 align:middle line:84% correlated to the farming gets to market difference, 01:21:06.220 --> 01:21:09.520 align:middle line:84% but they also buy meat, they buy processed food, 01:21:09.520 --> 01:21:11.860 align:middle line:84% and in general, the purchased food only 01:21:11.860 --> 01:21:14.840 align:middle line:84% responds to the female income and not to the male income. 01:21:14.840 --> 01:21:17.710 align:middle line:84% So that's consistent with it is a job of the woman 01:21:17.710 --> 01:21:19.610 align:middle line:90% to put food on the table. 01:21:19.610 --> 01:21:22.300 align:middle line:84% So when their income goes up, food goes up, 01:21:22.300 --> 01:21:29.240 align:middle line:84% and the main income is not being channeled to this. 01:21:29.240 --> 01:21:34.570 align:middle line:84% So that's another example in real life, 01:21:34.570 --> 01:21:37.570 align:middle line:84% non-experimental real life example of the failure 01:21:37.570 --> 01:21:42.370 align:middle line:84% to do household optimization, and therefore, it 01:21:42.370 --> 01:21:47.180 align:middle line:84% doesn't seem to be that it's an efficient household model. 01:21:47.180 --> 01:21:49.980 align:middle line:90% So where does that leave us? 01:21:49.980 --> 01:21:51.920 align:middle line:84% And that's why I wanted to finish that today. 01:21:51.920 --> 01:21:55.610 align:middle line:84% That leaves us to, well clearly, the efficient household model 01:21:55.610 --> 01:21:58.350 align:middle line:84% is not a very good model of the family. 01:21:58.350 --> 01:22:02.330 align:middle line:84% So a lot of the modern literature on the family 01:22:02.330 --> 01:22:05.570 align:middle line:84% and on gender economics is about trying 01:22:05.570 --> 01:22:08.720 align:middle line:84% to understand what is going on, whether families are in any way 01:22:08.720 --> 01:22:12.020 align:middle line:84% different than a random collection of individuals 01:22:12.020 --> 01:22:16.550 align:middle line:84% that just met a minute ago, and if so, in what way. 01:22:16.550 --> 01:22:20.180 align:middle line:84% And what are the barriers that stand between the household 01:22:20.180 --> 01:22:21.650 align:middle line:90% and efficiency? 01:22:21.650 --> 01:22:25.020 align:middle line:84% Information, trust, social norms, and so on and so forth. 01:22:25.020 --> 01:22:28.160 align:middle line:84% So that's what I'll do on Monday and with an emphasis 01:22:28.160 --> 01:22:35.960 align:middle line:84% on women labor supply, which is a big policy area 01:22:35.960 --> 01:22:39.280 align:middle line:90% in its own right Thank you. 01:22:39.280 --> 01:22:59.000 align:middle line:90%