WEBVTT 00:00:00.000 --> 00:00:01.960 align:middle line:90% [SQUEAKING] 00:00:01.960 --> 00:00:03.430 align:middle line:90% [RUSTLING] 00:00:03.430 --> 00:00:05.880 align:middle line:90% [CLICKING] 00:00:05.880 --> 00:00:18.647 align:middle line:90% 00:00:18.647 --> 00:00:19.480 align:middle line:90% ESTHER DUFLO: Hello. 00:00:19.480 --> 00:00:21.470 align:middle line:90% AUDIENCE: Hi. 00:00:21.470 --> 00:00:23.710 align:middle line:84% ESTHER DUFLO: So today and Wednesday, 00:00:23.710 --> 00:00:26.410 align:middle line:84% we are going to be talking about health. 00:00:26.410 --> 00:00:29.380 align:middle line:84% Of course, this is something that 00:00:29.380 --> 00:00:33.190 align:middle line:84% has received heightened attention in the last couple 00:00:33.190 --> 00:00:36.790 align:middle line:90% of years for some reason. 00:00:36.790 --> 00:00:41.500 align:middle line:90% And I've be talking about-- 00:00:41.500 --> 00:00:44.570 align:middle line:84% roughly speaking, I have one lecture on demand 00:00:44.570 --> 00:00:47.440 align:middle line:90% and one lecture on supply. 00:00:47.440 --> 00:00:49.240 align:middle line:84% Although it's possible that today's lecture 00:00:49.240 --> 00:00:53.620 align:middle line:84% runs a little longer, and we'll finish it over Wednesday. 00:00:53.620 --> 00:00:56.080 align:middle line:84% In the demand for health, one could 00:00:56.080 --> 00:00:58.300 align:middle line:90% talk about many, many things. 00:00:58.300 --> 00:01:00.880 align:middle line:84% But the one thing I want to focus 00:01:00.880 --> 00:01:06.280 align:middle line:84% is why people seem to want treatments, like antibiotics 00:01:06.280 --> 00:01:10.800 align:middle line:84% and steroid, but not vaccination or the preventive care. 00:01:10.800 --> 00:01:13.980 align:middle line:84% Something we are seeing in the context of the COVID vaccine, 00:01:13.980 --> 00:01:19.720 align:middle line:84% where it's the same people who want to consume ivermectin 00:01:19.720 --> 00:01:23.200 align:middle line:84% and who refuse to get vaccinated. 00:01:23.200 --> 00:01:29.010 align:middle line:84% So somehow there seems to be a disconnect between. 00:01:29.010 --> 00:01:31.620 align:middle line:84% So it's not, for example, some people who 00:01:31.620 --> 00:01:34.845 align:middle line:84% don't want to get vaccinated against various things 00:01:34.845 --> 00:01:37.470 align:middle line:84% are against all sorts of treatment, 00:01:37.470 --> 00:01:41.770 align:middle line:84% like Jehovah's Witnesses or Christian Science followers. 00:01:41.770 --> 00:01:47.310 align:middle line:84% But the vast majority of people don't want to 00:01:47.310 --> 00:01:53.970 align:middle line:84% or are not so excited about getting vaccinated or taking 00:01:53.970 --> 00:01:59.790 align:middle line:84% iron pills or doing all sorts of other preventive things 00:01:59.790 --> 00:02:04.180 align:middle line:90% but love curative medicine. 00:02:04.180 --> 00:02:09.780 align:middle line:84% That's something that we are, perhaps, are discovering anew 00:02:09.780 --> 00:02:14.280 align:middle line:84% in the US, today, with respect to the COVID vaccination 00:02:14.280 --> 00:02:15.780 align:middle line:90% or even the flu shot. 00:02:15.780 --> 00:02:18.930 align:middle line:84% But it's something that, if you work on development, 00:02:18.930 --> 00:02:22.230 align:middle line:90% you've encountered many times. 00:02:22.230 --> 00:02:27.150 align:middle line:84% For example, we have a survey in Udaipur. 00:02:27.150 --> 00:02:34.830 align:middle line:84% Very ancient now, but the main factor is still true. 00:02:34.830 --> 00:02:39.960 align:middle line:84% It still holds, which shows that the super-poor households 00:02:39.960 --> 00:02:47.370 align:middle line:84% in Udaipur spend 8% of their budget on health. 00:02:47.370 --> 00:02:52.080 align:middle line:90% And the people spend more. 00:02:52.080 --> 00:02:57.900 align:middle line:84% The people are willing to spend like 5,000 rupees on health, 00:02:57.900 --> 00:03:02.400 align:middle line:84% on some extreme event, which is much more than what 00:03:02.400 --> 00:03:04.500 align:middle line:90% they consume usually. 00:03:04.500 --> 00:03:06.660 align:middle line:84% So they are willing to go into debt in order 00:03:06.660 --> 00:03:09.690 align:middle line:90% to get antibiotics. 00:03:09.690 --> 00:03:12.480 align:middle line:84% We find that, in that same survey, 00:03:12.480 --> 00:03:15.360 align:middle line:84% that these super-poor households have consulted 00:03:15.360 --> 00:03:19.680 align:middle line:84% someone for health care episode several times 00:03:19.680 --> 00:03:21.300 align:middle line:90% in the last month. 00:03:21.300 --> 00:03:23.340 align:middle line:84% They are going to all sorts of people. 00:03:23.340 --> 00:03:26.490 align:middle line:84% They are going to traditional healer. 00:03:26.490 --> 00:03:29.280 align:middle line:84% They are going to regular allopathic doctor. 00:03:29.280 --> 00:03:32.280 align:middle line:84% They are going to quacks, like pharmacists 00:03:32.280 --> 00:03:34.290 align:middle line:90% who call themselves doctors. 00:03:34.290 --> 00:03:36.420 align:middle line:84% They are quite sophisticated about managing 00:03:36.420 --> 00:03:37.887 align:middle line:90% these different sources. 00:03:37.887 --> 00:03:39.720 align:middle line:84% Although the models might be whatever it is. 00:03:39.720 --> 00:03:43.320 align:middle line:84% But then, in that same sample, there 00:03:43.320 --> 00:03:48.750 align:middle line:84% is a surprisingly low take up for preventive medicine. 00:03:48.750 --> 00:03:52.810 align:middle line:84% For example, that same sample of Udaipur, a few years ago 00:03:52.810 --> 00:03:54.990 align:middle line:84% less than 5% of children were fully 00:03:54.990 --> 00:03:59.170 align:middle line:84% immunized against the basic childhood diseases-- 00:03:59.170 --> 00:04:03.180 align:middle line:84% so they had received the five extended medicine. 00:04:03.180 --> 00:04:05.520 align:middle line:84% Another fact which is related to it 00:04:05.520 --> 00:04:11.910 align:middle line:84% is super-high price elasticity for technologies 00:04:11.910 --> 00:04:15.990 align:middle line:84% that are known to be widely effective-- bed nets, 00:04:15.990 --> 00:04:18.029 align:middle line:84% immunization, deworming, et cetera. 00:04:18.029 --> 00:04:23.790 align:middle line:84% So looking first at a graph put together by Pascaline Dupas, 00:04:23.790 --> 00:04:28.080 align:middle line:84% for a paper in Science, that looks at elasticity 00:04:28.080 --> 00:04:30.750 align:middle line:84% for positive prices, for all sorts 00:04:30.750 --> 00:04:36.600 align:middle line:84% of different goods and different places, from various studies. 00:04:36.600 --> 00:04:38.970 align:middle line:84% And what you see is that, in particular, 00:04:38.970 --> 00:04:44.700 align:middle line:84% at very, very low prices, a steep fall in demand. 00:04:44.700 --> 00:04:47.430 align:middle line:84% Not necessarily exactly at 0, so it's not necessarily 00:04:47.430 --> 00:04:49.743 align:middle line:90% from 0 to another price. 00:04:49.743 --> 00:04:51.660 align:middle line:84% At the beginning, this is what people believe, 00:04:51.660 --> 00:04:55.950 align:middle line:84% like you have this huge, huge drop at 0. 00:04:55.950 --> 00:04:59.130 align:middle line:84% From 0, 60% of people take clorin when it's 0. 00:04:59.130 --> 00:05:02.640 align:middle line:84% And it turns to 10% when it's $0.20. 00:05:02.640 --> 00:05:03.930 align:middle line:90% But it's not just that. 00:05:03.930 --> 00:05:06.630 align:middle line:84% For example, here, you have for bed nets, 00:05:06.630 --> 00:05:09.160 align:middle line:84% you have the steep drop a little higher. 00:05:09.160 --> 00:05:13.840 align:middle line:84% What is interesting is that a lot of this falls-- 00:05:13.840 --> 00:05:18.370 align:middle line:84% if you exclude these two, a lot of these falls, 00:05:18.370 --> 00:05:20.710 align:middle line:84% the levels are relatively comparable 00:05:20.710 --> 00:05:23.110 align:middle line:84% even though the things are completely different 00:05:23.110 --> 00:05:25.210 align:middle line:84% and they have very different costs. 00:05:25.210 --> 00:05:30.010 align:middle line:84% So people are equally elastic for an iron pill that literally 00:05:30.010 --> 00:05:32.800 align:middle line:84% costs a few cents as for a bed net that 00:05:32.800 --> 00:05:36.940 align:middle line:90% costs many dollars or $7 or $8. 00:05:36.940 --> 00:05:39.700 align:middle line:84% And well, they probably would be able to tell you 00:05:39.700 --> 00:05:42.580 align:middle line:84% that the cost to deliver the households to them 00:05:42.580 --> 00:05:43.300 align:middle line:90% is different. 00:05:43.300 --> 00:05:46.090 align:middle line:84% And the value is different, because, from the bed nets, 00:05:46.090 --> 00:05:50.440 align:middle line:84% you would get a stream of services over a long time. 00:05:50.440 --> 00:05:55.480 align:middle line:84% It's a durable versus a pill that you take just once. 00:05:55.480 --> 00:06:01.610 align:middle line:84% So this doesn't seem true that it could come out 00:06:01.610 --> 00:06:08.150 align:middle line:84% of any reasonable model, all of this super-sharp drop of take 00:06:08.150 --> 00:06:10.790 align:middle line:84% up at very high-- at very low prices. 00:06:10.790 --> 00:06:13.340 align:middle line:90% 00:06:13.340 --> 00:06:16.760 align:middle line:84% In addition to the elasticity, the price 00:06:16.760 --> 00:06:21.800 align:middle line:84% elasticity being higher for any positive prices, 00:06:21.800 --> 00:06:23.960 align:middle line:84% you also have high price elasticity 00:06:23.960 --> 00:06:26.142 align:middle line:90% for any negative prices. 00:06:26.142 --> 00:06:40.990 align:middle line:84% So looking at immunization, so, yeah-- 00:06:40.990 --> 00:06:44.097 align:middle line:90% So let me get there in a minute. 00:06:44.097 --> 00:06:45.430 align:middle line:90% So I'll go back to immunization. 00:06:45.430 --> 00:06:50.560 align:middle line:84% But with the COVID 19 vaccine, you get these tiny incentives. 00:06:50.560 --> 00:06:52.990 align:middle line:84% People are given lottery tickets, 00:06:52.990 --> 00:06:55.190 align:middle line:90% so that might be big if you win. 00:06:55.190 --> 00:06:56.830 align:middle line:84% But your point of winning being so low 00:06:56.830 --> 00:06:58.840 align:middle line:90% it's actually a tiny incentive. 00:06:58.840 --> 00:07:00.790 align:middle line:84% Or a beer or something like that. 00:07:00.790 --> 00:07:02.930 align:middle line:90% Now, I don't know that-- 00:07:02.930 --> 00:07:06.400 align:middle line:84% There is, actually, at least one study, from the Netherlands, 00:07:06.400 --> 00:07:09.790 align:middle line:84% on the impact of incentive, showing that these incentives 00:07:09.790 --> 00:07:11.620 align:middle line:90% seem to have a very large-- 00:07:11.620 --> 00:07:12.880 align:middle line:90% reasonably large effect. 00:07:12.880 --> 00:07:17.470 align:middle line:84% I'll show evidence of impact of incentives on immunization 00:07:17.470 --> 00:07:18.340 align:middle line:90% in a moment. 00:07:18.340 --> 00:07:20.860 align:middle line:84% But, certainly, there is a widespread belief 00:07:20.860 --> 00:07:23.600 align:middle line:84% that these small incentives move behavior. 00:07:23.600 --> 00:07:26.290 align:middle line:84% So you can think of them as a price-- 00:07:26.290 --> 00:07:28.510 align:middle line:84% as a negative price that you give to people. 00:07:28.510 --> 00:07:30.217 align:middle line:90% So it's symmetric around 0. 00:07:30.217 --> 00:07:32.050 align:middle line:84% You pay people a little bit to do something, 00:07:32.050 --> 00:07:36.240 align:middle line:84% and then they do it in a large part. 00:07:36.240 --> 00:07:38.460 align:middle line:84% So there are two counter-arguments 00:07:38.460 --> 00:07:43.920 align:middle line:84% to this idea that people are very price elastic. 00:07:43.920 --> 00:07:48.300 align:middle line:84% One is that, well, maybe people are not so price elastic. 00:07:48.300 --> 00:07:52.200 align:middle line:84% But if you offer things for 0, a bunch 00:07:52.200 --> 00:07:55.770 align:middle line:84% of people who have no interest in the thing 00:07:55.770 --> 00:07:57.030 align:middle line:90% are going to get it anyway. 00:07:57.030 --> 00:08:00.990 align:middle line:84% Because at 0, your valuation is always 00:08:00.990 --> 00:08:03.270 align:middle line:84% going to be a little bit higher than 0. 00:08:03.270 --> 00:08:06.210 align:middle line:84% Even if, for example, even potentially 00:08:06.210 --> 00:08:10.350 align:middle line:90% to do something else with it. 00:08:10.350 --> 00:08:12.600 align:middle line:84% So for example, with bed nets, there 00:08:12.600 --> 00:08:17.340 align:middle line:84% was this idea that people would take the bed nets, not 00:08:17.340 --> 00:08:19.378 align:middle line:84% because they are interested in the bed nets, 00:08:19.378 --> 00:08:21.420 align:middle line:84% but because they could use it for something else. 00:08:21.420 --> 00:08:23.560 align:middle line:84% They could use it as a piece of cloth. 00:08:23.560 --> 00:08:25.350 align:middle line:84% So you could make a fishing net out of it 00:08:25.350 --> 00:08:28.420 align:middle line:90% or something like that. 00:08:28.420 --> 00:08:32.159 align:middle line:84% So in that case, when you give things for free 00:08:32.159 --> 00:08:34.409 align:middle line:84% as opposed to putting a small cost, 00:08:34.409 --> 00:08:37.200 align:middle line:84% then you draw in a lot of people. 00:08:37.200 --> 00:08:40.350 align:middle line:84% But you don't draw in people who are interested in health. 00:08:40.350 --> 00:08:44.400 align:middle line:84% You are drawing in people who are-- 00:08:44.400 --> 00:08:48.180 align:middle line:84% because, even if you have no interest in the health, 00:08:48.180 --> 00:08:51.240 align:middle line:84% you might have another interest in that thing. 00:08:51.240 --> 00:08:55.420 align:middle line:84% That is just big enough that for 0, you would take it. 00:08:55.420 --> 00:08:58.140 align:middle line:84% So in that case, there would be no mystery. 00:08:58.140 --> 00:09:00.270 align:middle line:84% It's just that there is just a bunch of people 00:09:00.270 --> 00:09:05.250 align:middle line:84% who are not interested but, at 0, would come in just in case. 00:09:05.250 --> 00:09:09.360 align:middle line:84% So there is actually several people that look at that. 00:09:09.360 --> 00:09:14.760 align:middle line:84% And the test, there, is whether, once people, for example, 00:09:14.760 --> 00:09:19.100 align:middle line:84% acquire a bed net, do they use it? 00:09:19.100 --> 00:09:24.060 align:middle line:84% Or in Ashraf-Berry-Shapiro, another device-- this thing, 00:09:24.060 --> 00:09:27.280 align:middle line:90% it's water treatment. 00:09:27.280 --> 00:09:28.740 align:middle line:84% Once you get the water treatment, 00:09:28.740 --> 00:09:30.730 align:middle line:84% are you actually putting it in the water? 00:09:30.730 --> 00:09:33.450 align:middle line:84% So if your valuation of the bed nets 00:09:33.450 --> 00:09:41.040 align:middle line:84% for the goal of-- for the purpose of health is 0, 00:09:41.040 --> 00:09:43.920 align:middle line:84% then you're not going to use it as a bed net 00:09:43.920 --> 00:09:46.410 align:middle line:90% once you get it for free. 00:09:46.410 --> 00:09:48.030 align:middle line:84% In addition, there is another argument 00:09:48.030 --> 00:09:51.810 align:middle line:84% people make, which is the very fact that you paid 0 for a good 00:09:51.810 --> 00:09:54.210 align:middle line:84% makes you think it's worthless, so there 00:09:54.210 --> 00:09:56.490 align:middle line:84% might be some behavioral thing on top of it. 00:09:56.490 --> 00:09:58.740 align:middle line:84% That would also mean that if you don't pay for a good, 00:09:58.740 --> 00:10:04.260 align:middle line:84% it also reduces your valuation of that good as a health good. 00:10:04.260 --> 00:10:06.660 align:middle line:84% The two combined have the prediction 00:10:06.660 --> 00:10:13.660 align:middle line:84% that if, in fact, the real elasticity is lower 00:10:13.660 --> 00:10:16.680 align:middle line:84% but there are just a lot of people who are not interested, 00:10:16.680 --> 00:10:18.870 align:middle line:84% we should find that people who get a good for free 00:10:18.870 --> 00:10:21.940 align:middle line:90% are less likely to use it. 00:10:21.940 --> 00:10:27.900 align:middle line:84% So the first paper to test that was Cohen and Pascaline Dupas 00:10:27.900 --> 00:10:33.510 align:middle line:84% and Jessica Cohen, who did a fast study distributing 00:10:33.510 --> 00:10:37.890 align:middle line:84% free bed nets or bed nets at some prices 00:10:37.890 --> 00:10:42.660 align:middle line:84% along with the first maternity checkups. 00:10:42.660 --> 00:10:48.210 align:middle line:84% And what they did, later, is that they did random checks 00:10:48.210 --> 00:10:51.420 align:middle line:84% to see whether the people were used at follow-up. 00:10:51.420 --> 00:10:56.760 align:middle line:84% And what they find is that, conditional on having a net, 00:10:56.760 --> 00:10:59.940 align:middle line:84% so including both the treatment effect, potentially, as well as 00:10:59.940 --> 00:11:02.070 align:middle line:84% the selection effect, which is what 00:11:02.070 --> 00:11:05.100 align:middle line:84% we are interested in today, the use of the bed nets 00:11:05.100 --> 00:11:09.840 align:middle line:84% actually is the same regardless of the prices you get, you pay. 00:11:09.840 --> 00:11:13.650 align:middle line:84% Now, of course it's conditional on getting a net. 00:11:13.650 --> 00:11:16.500 align:middle line:84% But since you are much more likely to have a net if it 00:11:16.500 --> 00:11:19.740 align:middle line:84% was for free, the effective coverage by a net 00:11:19.740 --> 00:11:21.648 align:middle line:84% is, in fact, highly price elastic, 00:11:21.648 --> 00:11:23.440 align:middle line:84% the same way that I was showing you before. 00:11:23.440 --> 00:11:27.090 align:middle line:84% So if I showed you the effective coverage, which 00:11:27.090 --> 00:11:30.930 align:middle line:84% is the probability to have a net multiplied by the probability 00:11:30.930 --> 00:11:37.260 align:middle line:84% to use it, this is, in fact, highly dependent on how much 00:11:37.260 --> 00:11:38.550 align:middle line:90% people need to pay. 00:11:38.550 --> 00:11:42.750 align:middle line:84% And even at 50 Kenyan shillings per net, 00:11:42.750 --> 00:11:49.140 align:middle line:84% we are now at 1/3 of the or 1/6 of the full price. 00:11:49.140 --> 00:11:51.750 align:middle line:84% So it's still much, much lower than the full price. 00:11:51.750 --> 00:11:54.630 align:middle line:84% Nobody is willing to pay any more. 00:11:54.630 --> 00:12:02.930 align:middle line:84% So that price elasticity seems real for the use of a bed net. 00:12:02.930 --> 00:12:03.980 align:middle line:90% They also make-- 00:12:03.980 --> 00:12:09.670 align:middle line:90% 00:12:09.670 --> 00:12:12.940 align:middle line:84% So likewise, as I was saying, a large reaction 00:12:12.940 --> 00:12:16.360 align:middle line:84% to small positive incentives for immunization, 00:12:16.360 --> 00:12:20.770 align:middle line:84% such as the ones that are proposed for the COVID 19 00:12:20.770 --> 00:12:21.590 align:middle line:90% vaccine. 00:12:21.590 --> 00:12:27.460 align:middle line:84% The first study I know about is a paper 00:12:27.460 --> 00:12:31.420 align:middle line:84% that Abhijit, Rachel Glennerster, Dhruva Kothari, 00:12:31.420 --> 00:12:35.530 align:middle line:84% and myself worked on in Udaipur, the same area 00:12:35.530 --> 00:12:37.770 align:middle line:90% where we did the survey. 00:12:37.770 --> 00:12:41.410 align:middle line:84% We were really struck by this 5% full immunization rate 00:12:41.410 --> 00:12:43.640 align:middle line:84% and wanted to do something about it. 00:12:43.640 --> 00:12:46.240 align:middle line:84% So we did what Dean Karlan has called 00:12:46.240 --> 00:12:49.960 align:middle line:84% "conditional lentil transfer," instead of "conditional cash 00:12:49.960 --> 00:12:51.100 align:middle line:90% transfer." 00:12:51.100 --> 00:12:54.550 align:middle line:84% So we did a small experiment in 130 villages. 00:12:54.550 --> 00:12:57.460 align:middle line:84% 60 got randomly assigned to receive regular immunization 00:12:57.460 --> 00:12:58.750 align:middle line:90% camp. 00:12:58.750 --> 00:13:00.940 align:middle line:84% And so that's already a big improvement 00:13:00.940 --> 00:13:03.148 align:middle line:90% over what they had normally. 00:13:03.148 --> 00:13:04.690 align:middle line:84% Because, normally, they were supposed 00:13:04.690 --> 00:13:07.780 align:middle line:84% to go, this time, to the health clinic, which 00:13:07.780 --> 00:13:09.710 align:middle line:90% was closed a lot of the time. 00:13:09.710 --> 00:13:13.210 align:middle line:84% And so maybe it's just hard for them to get immunized. 00:13:13.210 --> 00:13:14.830 align:middle line:84% Because you show up with your kid. 00:13:14.830 --> 00:13:16.073 align:middle line:90% Walk two kilometer. 00:13:16.073 --> 00:13:16.990 align:middle line:90% Show up with your kid. 00:13:16.990 --> 00:13:18.220 align:middle line:90% The nurse is not there. 00:13:18.220 --> 00:13:19.870 align:middle line:84% Maybe you're not going to do it, again. 00:13:19.870 --> 00:13:23.810 align:middle line:84% Actually maybe it was, in fact, extraordinarily expensive. 00:13:23.810 --> 00:13:25.600 align:middle line:84% So the first thing we wanted to do 00:13:25.600 --> 00:13:29.230 align:middle line:84% is to sort of simulate what would happen if the camps were 00:13:29.230 --> 00:13:31.000 align:middle line:90% really ultra regular. 00:13:31.000 --> 00:13:34.810 align:middle line:84% Would it be sufficient to bring everyone to immunization camp? 00:13:34.810 --> 00:13:36.850 align:middle line:84% So again, in the COVID 19 vaccine, 00:13:36.850 --> 00:13:39.280 align:middle line:84% questions have been raised, in particular, 00:13:39.280 --> 00:13:42.190 align:middle line:84% for people from minorities, around access and making 00:13:42.190 --> 00:13:43.610 align:middle line:90% it as simple as possible. 00:13:43.610 --> 00:13:46.640 align:middle line:84% So here, suppose we make it as simple and regular as possible. 00:13:46.640 --> 00:13:48.970 align:middle line:84% So the NGO we work with, Seva Mandir, 00:13:48.970 --> 00:13:52.690 align:middle line:84% made an agreement with the government of Rajasthan 00:13:52.690 --> 00:13:57.400 align:middle line:84% to take over immunization services in 60 villages. 00:13:57.400 --> 00:13:58.750 align:middle line:90% They hired a nurse. 00:13:58.750 --> 00:14:01.180 align:middle line:84% They were there every month, rain or shine. 00:14:01.180 --> 00:14:02.440 align:middle line:90% All of the camps were held. 00:14:02.440 --> 00:14:04.990 align:middle line:84% They were in a central location in the village. 00:14:04.990 --> 00:14:06.910 align:middle line:90% So make it quite easy. 00:14:06.910 --> 00:14:11.110 align:middle line:84% Then in half of those, we offered small incentives 00:14:11.110 --> 00:14:12.160 align:middle line:90% to get immunized. 00:14:12.160 --> 00:14:16.330 align:middle line:84% And the small incentive was a kilo of lentil per shot 00:14:16.330 --> 00:14:19.250 align:middle line:84% plus one set of plates for finishing the shot. 00:14:19.250 --> 00:14:26.590 align:middle line:84% So it's actually-- there's kind of a slope to it. 00:14:26.590 --> 00:14:28.480 align:middle line:84% It's convex, which is going to be 00:14:28.480 --> 00:14:32.500 align:middle line:84% relevant for the rest of what we're going to do today. 00:14:32.500 --> 00:14:36.440 align:middle line:84% And we found pretty large impact. 00:14:36.440 --> 00:14:36.940 align:middle line:90% Oops. 00:14:36.940 --> 00:14:40.280 align:middle line:90% 00:14:40.280 --> 00:14:43.430 align:middle line:84% This is immunization in the control hamlets. 00:14:43.430 --> 00:14:47.030 align:middle line:84% So at the end line, they were at 6%. 00:14:47.030 --> 00:14:51.860 align:middle line:84% In intervention A hamlet, that's the ones that 00:14:51.860 --> 00:14:54.590 align:middle line:84% just got the camps, so that's already 00:14:54.590 --> 00:14:57.650 align:middle line:90% about doubling to about 12%. 00:14:57.650 --> 00:15:01.760 align:middle line:84% And then in dimension B hamlet, when we added the incentive, 00:15:01.760 --> 00:15:05.400 align:middle line:90% it comes to about 37%. 00:15:05.400 --> 00:15:11.150 align:middle line:84% So in fact, that draws in so many kids that it is cheaper-- 00:15:11.150 --> 00:15:14.990 align:middle line:84% To get the kid fully immunized, it 00:15:14.990 --> 00:15:17.240 align:middle line:84% turns out to be cheaper to offer the incentive 00:15:17.240 --> 00:15:19.820 align:middle line:90% than not offer the incentive. 00:15:19.820 --> 00:15:23.340 align:middle line:84% Because the nurse was more occupied. 00:15:23.340 --> 00:15:25.400 align:middle line:90% So the cost per shot goes down. 00:15:25.400 --> 00:15:28.100 align:middle line:84% The full cost increases, but you have more kids. 00:15:28.100 --> 00:15:31.970 align:middle line:84% But it means even counting the transfer of the costs, 00:15:31.970 --> 00:15:34.160 align:middle line:84% it's actually cheaper to provide incentive 00:15:34.160 --> 00:15:37.200 align:middle line:84% than not to provide them in this case. 00:15:37.200 --> 00:15:42.380 align:middle line:84% It has to be a function of not very many people 00:15:42.380 --> 00:15:45.020 align:middle line:84% getting immunized without the incentive 00:15:45.020 --> 00:15:48.830 align:middle line:90% and the large treatment effect. 00:15:48.830 --> 00:15:52.670 align:middle line:84% Furthermore, we find very little spillover 00:15:52.670 --> 00:15:56.000 align:middle line:84% to the neighboring villages of just having the camp. 00:15:56.000 --> 00:15:59.630 align:middle line:84% So people do not seem to be walking nearby 00:15:59.630 --> 00:16:01.040 align:middle line:90% in order to get immunized. 00:16:01.040 --> 00:16:02.930 align:middle line:84% On the other hand, we find spillovers 00:16:02.930 --> 00:16:06.570 align:middle line:90% when the incentive is there. 00:16:06.570 --> 00:16:09.780 align:middle line:84% So this bar is bigger than this one, for example. 00:16:09.780 --> 00:16:14.510 align:middle line:84% So showing that people will not move to get 00:16:14.510 --> 00:16:18.070 align:middle line:84% immunized, but they'll come for the kilo of lentils. 00:16:18.070 --> 00:16:20.780 align:middle line:90% 00:16:20.780 --> 00:16:25.220 align:middle line:84% So that suggests that maybe the COVID vaccine 00:16:25.220 --> 00:16:28.490 align:middle line:90% incentives might work. 00:16:28.490 --> 00:16:31.670 align:middle line:84% But there is, however, one big difference 00:16:31.670 --> 00:16:35.030 align:middle line:84% in the context, which is where we ask people 00:16:35.030 --> 00:16:39.650 align:middle line:84% in Udaipur whether they think immunization is a good thing. 00:16:39.650 --> 00:16:41.360 align:middle line:90% They all think it's great. 00:16:41.360 --> 00:16:43.730 align:middle line:84% And they all would like to do it. 00:16:43.730 --> 00:16:47.180 align:middle line:84% So 98% of people-- and we did the same survey in Haryana. 00:16:47.180 --> 00:16:51.660 align:middle line:84% You read the [INAUDIBLE] paper, where we did the immunization 00:16:51.660 --> 00:16:52.160 align:middle line:90% context. 00:16:52.160 --> 00:16:53.900 align:middle line:84% In this context, everybody tells you 00:16:53.900 --> 00:16:57.550 align:middle line:84% that they would really like to have their kids immunized. 00:16:57.550 --> 00:17:00.850 align:middle line:84% So when people have asked me about the incentives 00:17:00.850 --> 00:17:02.930 align:middle line:84% for the COVID vaccine, I was like, well, 00:17:02.930 --> 00:17:04.990 align:middle line:84% there are some people who might want to do it 00:17:04.990 --> 00:17:06.250 align:middle line:90% and don't get around to do it. 00:17:06.250 --> 00:17:08.380 align:middle line:84% And then the incentive might work. 00:17:08.380 --> 00:17:10.000 align:middle line:84% But it is less clear that someone 00:17:10.000 --> 00:17:11.920 align:middle line:84% might be convinced by a small incentive 00:17:11.920 --> 00:17:15.109 align:middle line:84% if they don't want it in the first place. 00:17:15.109 --> 00:17:19.240 align:middle line:84% So that creates a very different context. 00:17:19.240 --> 00:17:23.260 align:middle line:90% 00:17:23.260 --> 00:17:25.869 align:middle line:84% And that's why we have to think about what 00:17:25.869 --> 00:17:30.640 align:middle line:84% causes these very large responses to the incentives. 00:17:30.640 --> 00:17:31.720 align:middle line:90% So why are this? 00:17:31.720 --> 00:17:33.400 align:middle line:90% Why are those? 00:17:33.400 --> 00:17:35.197 align:middle line:90% These are just a few examples. 00:17:35.197 --> 00:17:36.280 align:middle line:90% But there are many others. 00:17:36.280 --> 00:17:38.890 align:middle line:90% 00:17:38.890 --> 00:17:43.270 align:middle line:84% Rebecca Thornton has a paper on giving people incentives, 00:17:43.270 --> 00:17:47.410 align:middle line:84% tiny incentives, to get their HIV test results after the test 00:17:47.410 --> 00:17:48.880 align:middle line:90% had already been taken anyway. 00:17:48.880 --> 00:17:51.220 align:middle line:84% So it doesn't even involve the shots. 00:17:51.220 --> 00:17:52.870 align:middle line:90% It just getting the results-- 00:17:52.870 --> 00:17:58.810 align:middle line:84% and find very large impact of getting a small incentive 00:17:58.810 --> 00:18:03.490 align:middle line:84% to collect your HIV/AIDS test results. 00:18:03.490 --> 00:18:06.760 align:middle line:84% She also finds that, if the tent to collect the result 00:18:06.760 --> 00:18:09.200 align:middle line:84% is a little further away, people don't go. 00:18:09.200 --> 00:18:11.020 align:middle line:84% So in a sense, in the same paper, 00:18:11.020 --> 00:18:14.950 align:middle line:84% she has a response to the negative-- the cost 00:18:14.950 --> 00:18:18.970 align:middle line:84% of having to travel, the walking cost versus 00:18:18.970 --> 00:18:21.340 align:middle line:84% the cost of the small incentive, and find 00:18:21.340 --> 00:18:23.600 align:middle line:90% the same thing in both paper. 00:18:23.600 --> 00:18:25.540 align:middle line:90% So these are just two example. 00:18:25.540 --> 00:18:27.890 align:middle line:90% But there are many, many others. 00:18:27.890 --> 00:18:30.100 align:middle line:84% And now, in particular, there are replication 00:18:30.100 --> 00:18:33.820 align:middle line:84% of this small incentive for immunization in many contexts, 00:18:33.820 --> 00:18:38.500 align:middle line:84% most recently in Pakistan, in a paper in which I was involved, 00:18:38.500 --> 00:18:42.680 align:middle line:84% but also in Ethiopia and in other places. 00:18:42.680 --> 00:18:44.950 align:middle line:84% And it's the basis for the conditional cash transfer 00:18:44.950 --> 00:18:49.270 align:middle line:84% approach, which has become very popular in many countries. 00:18:49.270 --> 00:18:52.790 align:middle line:84% But of course, those cash transfers are much larger. 00:18:52.790 --> 00:18:55.870 align:middle line:84% So although it's very popular, it's 00:18:55.870 --> 00:19:01.030 align:middle line:84% pretty puzzling in light of a demand for health model 00:19:01.030 --> 00:19:03.250 align:middle line:84% where people are responsive to benefits 00:19:03.250 --> 00:19:04.600 align:middle line:90% and responsive to costs. 00:19:04.600 --> 00:19:07.300 align:middle line:84% Because the benefits to preventive health 00:19:07.300 --> 00:19:09.850 align:middle line:84% are fairly large-- are very large. 00:19:09.850 --> 00:19:12.280 align:middle line:84% And that's why I was pointing out that, 00:19:12.280 --> 00:19:14.290 align:middle line:84% OK, are people aware of these benefits? 00:19:14.290 --> 00:19:15.670 align:middle line:90% Maybe they are not. 00:19:15.670 --> 00:19:17.920 align:middle line:84% In the case of the immunization, in the US, 00:19:17.920 --> 00:19:20.345 align:middle line:84% maybe the remaining people, who are not vaccinated, 00:19:20.345 --> 00:19:21.595 align:middle line:90% are not aware of the benefits. 00:19:21.595 --> 00:19:26.810 align:middle line:84% Or for them, these benefits are balanced by very large costs. 00:19:26.810 --> 00:19:30.850 align:middle line:84% But in the case of Rajasthan or Haryana, 00:19:30.850 --> 00:19:33.623 align:middle line:84% it seems that people are aware of the benefits to some extent. 00:19:33.623 --> 00:19:35.290 align:middle line:84% At least they are aware enough that they 00:19:35.290 --> 00:19:37.498 align:middle line:84% are willing to tell you that there are such benefits. 00:19:37.498 --> 00:19:39.378 align:middle line:90% 00:19:39.378 --> 00:19:40.920 align:middle line:84% By the way, I don't know what their-- 00:19:40.920 --> 00:19:43.180 align:middle line:84% The answer about COVID vaccines are different. 00:19:43.180 --> 00:19:47.970 align:middle line:84% There is also a great amount of hesitancy for COVID vaccine, 00:19:47.970 --> 00:19:53.130 align:middle line:84% in India, that you don't see for childhood vaccination. 00:19:53.130 --> 00:19:54.630 align:middle line:84% And that I don't know whether it's 00:19:54.630 --> 00:19:57.210 align:middle line:84% because it's the first adult vaccine that they have, 00:19:57.210 --> 00:19:59.700 align:middle line:84% since there are no regular flu shot vaccine. 00:19:59.700 --> 00:20:01.770 align:middle line:90% So it's something unusual. 00:20:01.770 --> 00:20:04.740 align:middle line:84% Or whether it is because the same kind of disinformation 00:20:04.740 --> 00:20:08.520 align:middle line:84% that floats in the US somehow makes its way in all 00:20:08.520 --> 00:20:09.790 align:middle line:90% of these countries, as well. 00:20:09.790 --> 00:20:12.360 align:middle line:90% I don't think we know that yet. 00:20:12.360 --> 00:20:15.030 align:middle line:84% But in the case of childhood vaccination, or maybe they 00:20:15.030 --> 00:20:16.300 align:middle line:90% just-- 00:20:16.300 --> 00:20:18.450 align:middle line:84% we'll talk a bit about social norms in a minute. 00:20:18.450 --> 00:20:20.880 align:middle line:84% Maybe if people feel that you want 00:20:20.880 --> 00:20:23.820 align:middle line:84% them to hear that the childhood vaccination is 00:20:23.820 --> 00:20:25.570 align:middle line:84% safe and effective, and they should do it. 00:20:25.570 --> 00:20:27.570 align:middle line:90% So that's what they tell you. 00:20:27.570 --> 00:20:31.440 align:middle line:84% But they think it's OK to say that the vaccines for COVID 00:20:31.440 --> 00:20:33.750 align:middle line:84% are not safe, because that's what everybody around them 00:20:33.750 --> 00:20:34.400 align:middle line:90% is saying. 00:20:34.400 --> 00:20:36.150 align:middle line:84% It's possible that it's the case, as well. 00:20:36.150 --> 00:20:39.250 align:middle line:90% 00:20:39.250 --> 00:20:41.145 align:middle line:84% So all of these goods have large benefits. 00:20:41.145 --> 00:20:42.770 align:middle line:84% People seem to care about their health, 00:20:42.770 --> 00:20:44.895 align:middle line:84% since they are spending so much money on treatment. 00:20:44.895 --> 00:20:47.128 align:middle line:84% It's not that they care, in particular. 00:20:47.128 --> 00:20:48.920 align:middle line:84% They won't immunize their kids for measles. 00:20:48.920 --> 00:20:50.600 align:middle line:84% But if the kid has measles, they will 00:20:50.600 --> 00:20:55.370 align:middle line:84% spend tons of money trying to get them better. 00:20:55.370 --> 00:20:57.620 align:middle line:84% And the prices or the opportunity costs 00:20:57.620 --> 00:20:59.150 align:middle line:90% are not that high to begin with. 00:20:59.150 --> 00:21:01.940 align:middle line:84% A lot of these goods are heavily subsidized or they are free, 00:21:01.940 --> 00:21:05.630 align:middle line:84% and all you have to do is to spend some time-- 00:21:05.630 --> 00:21:09.010 align:middle line:84% take some time off to get the kid vaccinated. 00:21:09.010 --> 00:21:11.730 align:middle line:84% So what are the possible explanations? 00:21:11.730 --> 00:21:14.400 align:middle line:84% So one of the most popular explanation in economics 00:21:14.400 --> 00:21:15.350 align:middle line:90% is present bias. 00:21:15.350 --> 00:21:22.200 align:middle line:84% So if you remember the model we just saw for savings. 00:21:22.200 --> 00:21:25.470 align:middle line:84% People might have very different preference in the present 00:21:25.470 --> 00:21:26.740 align:middle line:90% and in the future. 00:21:26.740 --> 00:21:31.740 align:middle line:84% So here, in the case of immunization, 00:21:31.740 --> 00:21:38.340 align:middle line:84% think of being immunized, that's something 00:21:38.340 --> 00:21:48.060 align:middle line:84% you value in the future, and the time that you spent in front 00:21:48.060 --> 00:21:50.190 align:middle line:84% of the television, instead of getting immunized, 00:21:50.190 --> 00:21:51.840 align:middle line:90% as a temptation good. 00:21:51.840 --> 00:21:55.770 align:middle line:84% So it's in the space of time or inconvenience 00:21:55.770 --> 00:22:00.720 align:middle line:84% as opposed to being in the space of money. 00:22:00.720 --> 00:22:05.440 align:middle line:84% So you think that you would like-- 00:22:05.440 --> 00:22:09.330 align:middle line:84% So today, you think about the benefits of immunization. 00:22:09.330 --> 00:22:11.250 align:middle line:90% In the future, they look good. 00:22:11.250 --> 00:22:13.890 align:middle line:84% But you don't experience any today, because today your kid 00:22:13.890 --> 00:22:16.140 align:middle line:90% is not exposed to the measles. 00:22:16.140 --> 00:22:19.890 align:middle line:84% Or today, your kid is not-- there is not any benefit today. 00:22:19.890 --> 00:22:21.930 align:middle line:84% There'll be benefits in the future. 00:22:21.930 --> 00:22:25.260 align:middle line:84% But the temptation of staying home is today. 00:22:25.260 --> 00:22:30.810 align:middle line:84% The temptation of using some of your time is today. 00:22:30.810 --> 00:22:35.380 align:middle line:84% And you don't value that time for tomorrow. 00:22:35.380 --> 00:22:41.300 align:middle line:84% So you will always want to wait and do it some other time. 00:22:41.300 --> 00:22:45.790 align:middle line:84% So in that case, a small cost today discouraged actions. 00:22:45.790 --> 00:22:47.710 align:middle line:84% But it could be undone by a small benefit 00:22:47.710 --> 00:22:52.630 align:middle line:84% today, for example, the bag of lentils or the plates. 00:22:52.630 --> 00:22:57.190 align:middle line:84% And that would give you an argument for preventive care 00:22:57.190 --> 00:23:01.960 align:middle line:84% subsidies, not just positive prices 00:23:01.960 --> 00:23:05.770 align:middle line:84% but also subsidies, not just because 00:23:05.770 --> 00:23:08.860 align:middle line:84% of externalities reason, which, in and of itself, 00:23:08.860 --> 00:23:11.360 align:middle line:84% would justify that we pay household to get immunized, 00:23:11.360 --> 00:23:13.870 align:middle line:84% so they internalize the benefits to other people 00:23:13.870 --> 00:23:15.140 align:middle line:90% in the community. 00:23:15.140 --> 00:23:18.220 align:middle line:84% But also for internalities, because the small benefit 00:23:18.220 --> 00:23:23.181 align:middle line:84% received today is going to outweigh 00:23:23.181 --> 00:23:26.290 align:middle line:90% the small cost of today. 00:23:26.290 --> 00:23:28.270 align:middle line:84% And at some level, you should give ice cream 00:23:28.270 --> 00:23:32.140 align:middle line:84% in the camps as opposed to money in this context. 00:23:32.140 --> 00:23:35.000 align:middle line:84% Because you'd have a temptation against your temptation, 00:23:35.000 --> 00:23:37.390 align:middle line:90% and that would be great. 00:23:37.390 --> 00:23:41.710 align:middle line:84% So in that sense, the beer might be 00:23:41.710 --> 00:23:43.540 align:middle line:90% the ideal form of an incentive. 00:23:43.540 --> 00:23:49.910 align:middle line:90% 00:23:49.910 --> 00:23:55.340 align:middle line:84% If indeed present bias was a big reason 00:23:55.340 --> 00:23:58.760 align:middle line:84% why people don't get immunized, then incentives 00:23:58.760 --> 00:24:03.470 align:middle line:84% would not be the only way that you could go about it. 00:24:03.470 --> 00:24:06.080 align:middle line:84% For example, one thing that would be powerful 00:24:06.080 --> 00:24:11.450 align:middle line:84% is default option, which is, basically, you have to opt out 00:24:11.450 --> 00:24:13.700 align:middle line:90% to not get your kid immunized. 00:24:13.700 --> 00:24:16.910 align:middle line:84% So the form of mandates that are present for childhood 00:24:16.910 --> 00:24:20.120 align:middle line:84% immunization in many countries including the US. 00:24:20.120 --> 00:24:25.520 align:middle line:84% It's not that you cannot refuse to get your child immunized, 00:24:25.520 --> 00:24:28.190 align:middle line:84% but, if you do, it's going to be inconvenient. 00:24:28.190 --> 00:24:32.240 align:middle line:84% You have to find a doctor that is willing to sign you a form. 00:24:32.240 --> 00:24:36.350 align:middle line:84% Or you have to find a religious leader who is going to give you 00:24:36.350 --> 00:24:39.120 align:middle line:90% the religious exemption. 00:24:39.120 --> 00:24:42.510 align:middle line:90% And so that's a bit of a pain. 00:24:42.510 --> 00:24:46.160 align:middle line:84% So that can be interpreted in this framework 00:24:46.160 --> 00:24:48.470 align:middle line:84% by saying, well, by default, your kid 00:24:48.470 --> 00:24:52.320 align:middle line:84% is going to be immunized for all of these diseases. 00:24:52.320 --> 00:24:54.290 align:middle line:84% And if you want to get out of the default, 00:24:54.290 --> 00:24:57.860 align:middle line:84% you have to do something slightly painful. 00:24:57.860 --> 00:25:01.480 align:middle line:84% So that's this idea of nudge or choice infrastructure 00:25:01.480 --> 00:25:07.540 align:middle line:84% is that the default is going to push you in that direction. 00:25:07.540 --> 00:25:11.720 align:middle line:84% Not just because it shows you what to do, 00:25:11.720 --> 00:25:14.600 align:middle line:84% which might also be part of the impact of default, by the way, 00:25:14.600 --> 00:25:20.175 align:middle line:84% but also because it sets up a barrier. 00:25:20.175 --> 00:25:22.300 align:middle line:84% So that's why in most-- for childhood immunization, 00:25:22.300 --> 00:25:24.498 align:middle line:84% I think, in most of the developed countries, 00:25:24.498 --> 00:25:25.540 align:middle line:90% we don't have incentives. 00:25:25.540 --> 00:25:29.680 align:middle line:84% We have mandates that are not very hard mandates. 00:25:29.680 --> 00:25:32.600 align:middle line:84% We can see that, in the case of the COVID 19 vaccine, 00:25:32.600 --> 00:25:35.290 align:middle line:84% the mandates are actually much harder. 00:25:35.290 --> 00:25:37.000 align:middle line:84% And they are now only for adults. 00:25:37.000 --> 00:25:40.150 align:middle line:84% But in places like, for example, if you are a health care 00:25:40.150 --> 00:25:42.650 align:middle line:90% worker, you'd lose your job. 00:25:42.650 --> 00:25:44.860 align:middle line:90% So that's not a nudge. 00:25:44.860 --> 00:25:48.250 align:middle line:84% So it's a different kind of mandate. 00:25:48.250 --> 00:25:49.540 align:middle line:90% And they don't give-- 00:25:49.540 --> 00:25:52.630 align:middle line:84% at least in, for example, my pediatrician's office, 00:25:52.630 --> 00:25:55.540 align:middle line:84% there is a big placard saying, we 00:25:55.540 --> 00:25:59.540 align:middle line:84% are not giving medical exemption for the COVID 19 vaccine. 00:25:59.540 --> 00:26:02.290 align:middle line:90% So I think they are making-- 00:26:02.290 --> 00:26:05.050 align:middle line:84% partly, because the reason why people are refusing COVID 00:26:05.050 --> 00:26:06.430 align:middle line:90% 19 vaccines are different. 00:26:06.430 --> 00:26:13.390 align:middle line:84% They are a refusal not an inertia, not this present bias 00:26:13.390 --> 00:26:13.970 align:middle line:90% framework. 00:26:13.970 --> 00:26:15.980 align:middle line:84% Then if you want your mandate to be effective, 00:26:15.980 --> 00:26:17.530 align:middle line:90% it's actually a proper mandate. 00:26:17.530 --> 00:26:21.810 align:middle line:90% 00:26:21.810 --> 00:26:24.630 align:middle line:84% But for childhood immunization, if the issue is really 00:26:24.630 --> 00:26:27.100 align:middle line:84% hyperbolic discounting issues, that you want to do it, 00:26:27.100 --> 00:26:30.640 align:middle line:84% but it's kind of never quite the right time, 00:26:30.640 --> 00:26:34.090 align:middle line:84% then a default option would help. 00:26:34.090 --> 00:26:37.750 align:middle line:84% Another thing that might help is to help people to commit. 00:26:37.750 --> 00:26:41.550 align:middle line:84% So in the same way that, for the savings, 00:26:41.550 --> 00:26:48.160 align:middle line:84% we saw Ashraf, Karlan, and Gine looking at giving people 00:26:48.160 --> 00:26:50.970 align:middle line:90% a commitment savings option. 00:26:50.970 --> 00:26:56.250 align:middle line:84% Dean Karlan and Xavier Gine have a paper in the Philippine, 00:26:56.250 --> 00:26:58.530 align:middle line:84% where they offer a program called CARES-- 00:26:58.530 --> 00:27:03.810 align:middle line:84% not to be confused with the unemployment relief in the US-- 00:27:03.810 --> 00:27:07.950 align:middle line:84% where people are offered this incredibly attractive savings 00:27:07.950 --> 00:27:13.170 align:middle line:84% account, which is an account that pays 0 interest, 00:27:13.170 --> 00:27:20.430 align:middle line:84% where you agree, if you put the money, to submit yourself 00:27:20.430 --> 00:27:22.210 align:middle line:90% to a smoking test. 00:27:22.210 --> 00:27:26.130 align:middle line:84% And if you fail the smoking test, you lose the money. 00:27:26.130 --> 00:27:28.500 align:middle line:84% So that incredibly attractive product 00:27:28.500 --> 00:27:30.240 align:middle line:84% should really be taken by no one, 00:27:30.240 --> 00:27:34.680 align:middle line:84% unless they are very keen to stop smoking, 00:27:34.680 --> 00:27:37.500 align:middle line:90% because you might fail the test. 00:27:37.500 --> 00:27:40.303 align:middle line:84% In fact, 10% of people take up the program 00:27:40.303 --> 00:27:41.220 align:middle line:90% when they are offered. 00:27:41.220 --> 00:27:43.320 align:middle line:90% So it's not everyone. 00:27:43.320 --> 00:27:45.210 align:middle line:84% But it's comparable to the take up 00:27:45.210 --> 00:27:50.040 align:middle line:84% of smoking cessation offers, when they are made in the US, 00:27:50.040 --> 00:27:52.210 align:middle line:90% for example. 00:27:52.210 --> 00:27:56.490 align:middle line:84% And of the people who were offered the program, 00:27:56.490 --> 00:27:59.880 align:middle line:84% they were 3% more likely to stop smoking 00:27:59.880 --> 00:28:03.120 align:middle line:84% than people who were not offered the program. 00:28:03.120 --> 00:28:07.320 align:middle line:84% So assuming that being offered the program 00:28:07.320 --> 00:28:10.300 align:middle line:84% doesn't have a direct effect, and, therefore, 00:28:10.300 --> 00:28:14.130 align:middle line:84% we can use the offer of the program as an instrument 00:28:14.130 --> 00:28:16.080 align:middle line:90% for stopping smoking. 00:28:16.080 --> 00:28:18.630 align:middle line:84% And now, I'll let you, as homework, think about 00:28:18.630 --> 00:28:20.530 align:middle line:84% whether that's a good assumption or not. 00:28:20.530 --> 00:28:22.620 align:middle line:84% But if you are willing to assume it, 00:28:22.620 --> 00:28:24.330 align:middle line:84% that gives you a treatment effect 00:28:24.330 --> 00:28:29.580 align:middle line:84% of being actually signed up of 33%, then divided by 3, 00:28:29.580 --> 00:28:32.580 align:middle line:90% which is a pretty large account. 00:28:32.580 --> 00:28:36.420 align:middle line:84% So the fact that anybody is willing to do that, 00:28:36.420 --> 00:28:39.930 align:middle line:84% and, of those people, there was actually some success involved 00:28:39.930 --> 00:28:43.950 align:middle line:84% is also consistent with these commitment problems. 00:28:43.950 --> 00:28:46.890 align:middle line:84% And of course, the paper on alcohol, 00:28:46.890 --> 00:28:54.340 align:middle line:84% that you saw with Frank, again goes in this direction. 00:28:54.340 --> 00:29:00.040 align:middle line:84% I do have one misgiving with this type of commitment-based, 00:29:00.040 --> 00:29:04.120 align:middle line:84% commitment problem, time inconsistency problem based 00:29:04.120 --> 00:29:06.490 align:middle line:84% explanation for preventive health, 00:29:06.490 --> 00:29:09.880 align:middle line:84% in particular vaccination, which is that it really 00:29:09.880 --> 00:29:13.990 align:middle line:84% requires people to keep being fooled by themselves. 00:29:13.990 --> 00:29:17.890 align:middle line:84% You have to think, I'm not going to do it today. 00:29:17.890 --> 00:29:20.410 align:middle line:84% But I value it, and I'm going to do it tomorrow. 00:29:20.410 --> 00:29:24.160 align:middle line:84% But then when tomorrow rolls along, you again not do it. 00:29:24.160 --> 00:29:26.920 align:middle line:84% And take for example, the example of our immunization 00:29:26.920 --> 00:29:27.670 align:middle line:90% camp. 00:29:27.670 --> 00:29:31.540 align:middle line:84% Every single month there is an immunization camp. 00:29:31.540 --> 00:29:34.840 align:middle line:84% Every single month, they fail to go. 00:29:34.840 --> 00:29:38.860 align:middle line:84% I mean, after a while, they should have gotten the memo 00:29:38.860 --> 00:29:42.640 align:middle line:90% and figured something out. 00:29:42.640 --> 00:29:44.830 align:middle line:84% And at the same time, we're saying 00:29:44.830 --> 00:29:47.200 align:middle line:84% people are sophisticated enough to sign up 00:29:47.200 --> 00:29:49.090 align:middle line:90% for commitment devices. 00:29:49.090 --> 00:29:50.860 align:middle line:84% Well, then I have trouble thinking 00:29:50.860 --> 00:29:52.060 align:middle line:90% they are not sophisticated-- 00:29:52.060 --> 00:29:54.330 align:middle line:84% If they can sign up for a CARES commitment device 00:29:54.330 --> 00:29:56.830 align:middle line:84% to stop smoking-- of course, these are different households. 00:29:56.830 --> 00:30:01.630 align:middle line:84% But in the context of Udaipur, we could find, 00:30:01.630 --> 00:30:04.150 align:middle line:84% I'm sure, a lot of evidence showing a certain degree 00:30:04.150 --> 00:30:06.500 align:middle line:90% of sophistication. 00:30:06.500 --> 00:30:11.630 align:middle line:84% Then how could they continue not going, month after month 00:30:11.630 --> 00:30:15.020 align:middle line:84% after month, till the kid finally gets measles? 00:30:15.020 --> 00:30:17.240 align:middle line:84% That doesn't-- it's kind of hard. 00:30:17.240 --> 00:30:21.620 align:middle line:84% If they really understand the benefit, the huge benefit 00:30:21.620 --> 00:30:27.740 align:middle line:84% that we think they are, then it seems a little bit unlikely. 00:30:27.740 --> 00:30:29.900 align:middle line:90% Do you see my point? 00:30:29.900 --> 00:30:33.200 align:middle line:90% So even though I-- 00:30:33.200 --> 00:30:36.050 align:middle line:84% So this type of explanation, that intuitively 00:30:36.050 --> 00:30:40.580 align:middle line:84% is very sensible and a lot of pieces of evidence kind of gel, 00:30:40.580 --> 00:30:43.370 align:middle line:84% but, quantitatively, it makes no sense at some level. 00:30:43.370 --> 00:30:45.740 align:middle line:84% And I haven't given you a quantitative calibration. 00:30:45.740 --> 00:30:48.210 align:middle line:84% But I've made one in my head for you. 00:30:48.210 --> 00:30:51.950 align:middle line:84% Like somehow it goes in the right direction, 00:30:51.950 --> 00:30:55.940 align:middle line:90% but it doesn't pass. 00:30:55.940 --> 00:30:57.080 align:middle line:90% So that's my issue-- 00:30:57.080 --> 00:31:00.600 align:middle line:84% as long as the benefits are in fact as large as possible. 00:31:00.600 --> 00:31:09.000 align:middle line:84% So one other possibility is that people are just not informed. 00:31:09.000 --> 00:31:11.400 align:middle line:84% So one of the thing is that these huge benefits, 00:31:11.400 --> 00:31:13.590 align:middle line:90% they exist. 00:31:13.590 --> 00:31:15.600 align:middle line:84% They are not balanced by huge cost. 00:31:15.600 --> 00:31:18.300 align:middle line:84% By the way, another explanation for low take-up 00:31:18.300 --> 00:31:20.730 align:middle line:84% would be people are scared or there are 00:31:20.730 --> 00:31:22.630 align:middle line:90% huge costs associated with it. 00:31:22.630 --> 00:31:25.430 align:middle line:84% But in that case, they wouldn't be swayed by small incentives. 00:31:25.430 --> 00:31:29.130 align:middle line:84% Right, this is the discussion we had on the COVID vaccine. 00:31:29.130 --> 00:31:31.558 align:middle line:84% You have to have a real mandate to move people 00:31:31.558 --> 00:31:32.850 align:middle line:90% who really don't want to do it. 00:31:32.850 --> 00:31:34.920 align:middle line:84% Because they think there is a big cost. 00:31:34.920 --> 00:31:35.930 align:middle line:90% Yeah? 00:31:35.930 --> 00:31:39.820 align:middle line:84% AUDIENCE: On the last point you made, 00:31:39.820 --> 00:31:42.580 align:middle line:84% in most of these types of experiments and things, 00:31:42.580 --> 00:31:44.338 align:middle line:84% are these camps things that are going 00:31:44.338 --> 00:31:45.880 align:middle line:84% to happen in one month in the village 00:31:45.880 --> 00:31:47.380 align:middle line:84% and not come back for several years? 00:31:47.380 --> 00:31:49.990 align:middle line:84% Or is it a situation where vaccination camp 00:31:49.990 --> 00:31:51.730 align:middle line:84% will happen month after month or year 00:31:51.730 --> 00:31:53.230 align:middle line:84% after year and just this one happens 00:31:53.230 --> 00:31:55.880 align:middle line:84% to be run by Seva Mandir and has an incentive? 00:31:55.880 --> 00:31:59.040 align:middle line:84% I'm just what is that outside option if I don't take up 00:31:59.040 --> 00:32:00.072 align:middle line:90% one of these? 00:32:00.072 --> 00:32:00.780 align:middle line:90% ESTHER DUFLO: Oh. 00:32:00.780 --> 00:32:02.670 align:middle line:84% So in this case, I think they could 00:32:02.670 --> 00:32:04.500 align:middle line:84% have expected Seva Mandir to continue 00:32:04.500 --> 00:32:05.715 align:middle line:90% to do-- to run the camps. 00:32:05.715 --> 00:32:06.480 align:middle line:90% AUDIENCE: Yeah. 00:32:06.480 --> 00:32:10.260 align:middle line:84% ESTHER DUFLO: Because it was not done as an experiment. 00:32:10.260 --> 00:32:11.670 align:middle line:90% So from their point of view. 00:32:11.670 --> 00:32:13.260 align:middle line:90% It's a program by Seva Mandir. 00:32:13.260 --> 00:32:17.310 align:middle line:84% In fact, as far as I know, they are still running those camps. 00:32:17.310 --> 00:32:18.690 align:middle line:90% 20 years later, they are still-- 00:32:18.690 --> 00:32:20.430 align:middle line:84% AUDIENCE: The reason I ask is, is there 00:32:20.430 --> 00:32:23.210 align:middle line:84% anything in the [INAUDIBLE] theory that you're saying 00:32:23.210 --> 00:32:28.158 align:middle line:84% is [INAUDIBLE],, that has to do with that outside option? 00:32:28.158 --> 00:32:28.950 align:middle line:90% ESTHER DUFLO: Yeah. 00:32:28.950 --> 00:32:32.130 align:middle line:84% Yeah, so that's a good point, which is something 00:32:32.130 --> 00:32:35.310 align:middle line:84% that I think-- and [INAUDIBLE] suggested 00:32:35.310 --> 00:32:38.970 align:middle line:84% what we do is to have a deadline and to say, 00:32:38.970 --> 00:32:41.220 align:middle line:84% if you want to get your kids immunized at the camp, 00:32:41.220 --> 00:32:43.030 align:middle line:84% in this village, that's this month. 00:32:43.030 --> 00:32:45.010 align:middle line:84% And then after that, we are not coming back. 00:32:45.010 --> 00:32:48.130 align:middle line:90% So this is without deadline. 00:32:48.130 --> 00:32:50.520 align:middle line:90% So this is just-- 00:32:50.520 --> 00:32:54.598 align:middle line:90% this would-- Yes? 00:32:54.598 --> 00:32:56.140 align:middle line:84% AUDIENCE: And couldn't it be the case 00:32:56.140 --> 00:33:01.030 align:middle line:84% that people are not getting this preventive health because they 00:33:01.030 --> 00:33:02.913 align:middle line:90% don't have time for it. 00:33:02.913 --> 00:33:05.330 align:middle line:84% In Chile, one of the biggest issues with the COVID vaccine 00:33:05.330 --> 00:33:09.930 align:middle line:84% is that vaccination camps were open when people were working. 00:33:09.930 --> 00:33:13.400 align:middle line:84% So they had no permission from their bosses 00:33:13.400 --> 00:33:15.860 align:middle line:84% to leave work and get vaccinated. 00:33:15.860 --> 00:33:17.940 align:middle line:84% Because, otherwise, they would get fired. 00:33:17.940 --> 00:33:20.630 align:middle line:84% So they didn't get vaccinations, because there 00:33:20.630 --> 00:33:25.280 align:middle line:84% wasn't the right amount of time during the day to do that. 00:33:25.280 --> 00:33:26.030 align:middle line:90% ESTHER DUFLO: Yes. 00:33:26.030 --> 00:33:28.630 align:middle line:84% So here, it's going to take them some time. 00:33:28.630 --> 00:33:31.337 align:middle line:84% A lot of these people, particularly the moms are. 00:33:31.337 --> 00:33:33.170 align:middle line:84% And it's not that they're working somewhere, 00:33:33.170 --> 00:33:34.563 align:middle line:90% but they are very busy. 00:33:34.563 --> 00:33:35.480 align:middle line:90% They have stuff to do. 00:33:35.480 --> 00:33:37.380 align:middle line:84% So you need to take some time off. 00:33:37.380 --> 00:33:40.185 align:middle line:84% So think of that as the opportunity cost. 00:33:40.185 --> 00:33:41.810 align:middle line:84% The description you are giving in Chile 00:33:41.810 --> 00:33:43.910 align:middle line:90% is super-high opportunity cost. 00:33:43.910 --> 00:33:47.960 align:middle line:84% But I think these people have quite low opportunity cost. 00:33:47.960 --> 00:33:51.590 align:middle line:84% And that's low opportunity cost, including-- 00:33:51.590 --> 00:33:53.983 align:middle line:84% plus the fact that the kids might-- they get a bit sick 00:33:53.983 --> 00:33:54.650 align:middle line:90% and all of that. 00:33:54.650 --> 00:33:57.380 align:middle line:84% Those are the costs you're going to experience today 00:33:57.380 --> 00:33:59.060 align:middle line:90% in exchange of a benefit-- 00:33:59.060 --> 00:34:01.800 align:middle line:84% a flow of benefits in the future. 00:34:01.800 --> 00:34:04.730 align:middle line:84% And then these are not very big costs, 00:34:04.730 --> 00:34:10.370 align:middle line:84% so we can out compensate them, balance them out 00:34:10.370 --> 00:34:12.719 align:middle line:90% with the small benefit. 00:34:12.719 --> 00:34:15.889 align:middle line:84% But in your example, for Chile, if the cost happened 00:34:15.889 --> 00:34:17.360 align:middle line:90% to be large then-- 00:34:17.360 --> 00:34:21.320 align:middle line:84% AUDIENCE: They even had to write a law that said that if you get 00:34:21.320 --> 00:34:25.010 align:middle line:84% fired by the boss, because you want to get a COVID vaccine, 00:34:25.010 --> 00:34:28.760 align:middle line:84% then they're going to sue that firm. 00:34:28.760 --> 00:34:30.650 align:middle line:84% ESTHER DUFLO: And so should they. 00:34:30.650 --> 00:34:32.270 align:middle line:90% This is a-- yes. 00:34:32.270 --> 00:34:33.906 align:middle line:84% So, yeah, there might be situations-- 00:34:33.906 --> 00:34:35.989 align:middle line:84% That's also what they were saying at the beginning 00:34:35.989 --> 00:34:37.250 align:middle line:90% of the vaccine here. 00:34:37.250 --> 00:34:39.710 align:middle line:84% Some people had real issues with access. 00:34:39.710 --> 00:34:42.480 align:middle line:90% And in India-- 00:34:42.480 --> 00:34:45.440 align:middle line:84% So I don't want to belittle the access issue. 00:34:45.440 --> 00:34:47.510 align:middle line:84% In India, for childhood vaccination, 00:34:47.510 --> 00:34:49.280 align:middle line:84% for years, this was the only thing 00:34:49.280 --> 00:34:51.080 align:middle line:84% the government was trying to do is 00:34:51.080 --> 00:34:52.610 align:middle line:90% to really improve the access. 00:34:52.610 --> 00:34:53.989 align:middle line:90% And it has improved. 00:34:53.989 --> 00:34:56.179 align:middle line:90% It's much better than it was. 00:34:56.179 --> 00:34:59.270 align:middle line:84% And the vaccination take-up is still low. 00:34:59.270 --> 00:35:02.720 align:middle line:90% So it's not the only part of it. 00:35:02.720 --> 00:35:05.300 align:middle line:84% So one possibility is that people-- 00:35:05.300 --> 00:35:08.240 align:middle line:84% it's not that there are huge benefits and then huge costs, 00:35:08.240 --> 00:35:11.450 align:middle line:84% but that the benefits-- the cost are low and also the benefits 00:35:11.450 --> 00:35:12.140 align:middle line:90% are low-- 00:35:12.140 --> 00:35:16.460 align:middle line:84% not the real one but the perceived one. 00:35:16.460 --> 00:35:20.660 align:middle line:84% And in fact, there is plenty of evidence that people do not-- 00:35:20.660 --> 00:35:23.270 align:middle line:84% that there is a lot of relevant pieces of information people 00:35:23.270 --> 00:35:25.170 align:middle line:90% don't have. 00:35:25.170 --> 00:35:27.020 align:middle line:84% And so they might not realize what 00:35:27.020 --> 00:35:30.960 align:middle line:84% are the benefits associated with a particular behavior. 00:35:30.960 --> 00:35:37.310 align:middle line:84% So for example, I mentioned HIV education 00:35:37.310 --> 00:35:44.340 align:middle line:84% in Kenya, which is insisting on this ABCD program, 00:35:44.340 --> 00:35:47.640 align:middle line:84% Abstain, Be faithful, use a Condom or you Die. 00:35:47.640 --> 00:35:51.180 align:middle line:84% And it's not-- that program doesn't 00:35:51.180 --> 00:35:54.100 align:middle line:84% have one key piece of information, 00:35:54.100 --> 00:35:58.200 align:middle line:84% which is that the rate of getting HIV is much higher. 00:35:58.200 --> 00:36:00.690 align:middle line:84% The rate of being contaminated by HIV 00:36:00.690 --> 00:36:04.260 align:middle line:84% is much higher for older men than for younger boys. 00:36:04.260 --> 00:36:06.300 align:middle line:84% As it would be, since it doesn't go away. 00:36:06.300 --> 00:36:08.310 align:middle line:84% Once you're contaminated, you're contaminated. 00:36:08.310 --> 00:36:11.560 align:middle line:84% And also because older men, in fact, 00:36:11.560 --> 00:36:16.110 align:middle line:84% have risky sexual behavior, often 00:36:16.110 --> 00:36:21.520 align:middle line:84% have many girls that they pay for sex and so on and so forth. 00:36:21.520 --> 00:36:24.030 align:middle line:84% So they are called sugar daddies. 00:36:24.030 --> 00:36:29.100 align:middle line:84% So Pascaline Dupas, for her job market paper, 00:36:29.100 --> 00:36:37.110 align:middle line:84% did an experiment where she just gave girls and boys 00:36:37.110 --> 00:36:44.280 align:middle line:84% a piece of information that the rate of HIV is much higher. 00:36:44.280 --> 00:36:45.690 align:middle line:84% Just it was basically giving them 00:36:45.690 --> 00:36:48.720 align:middle line:84% information on rate of HIV positivity in Kenya 00:36:48.720 --> 00:36:52.830 align:middle line:84% among older men, younger boys, and then, also, women. 00:36:52.830 --> 00:36:54.990 align:middle line:84% And by the way, one other thing you find 00:36:54.990 --> 00:36:57.270 align:middle line:84% is girls are more likely to be contaminated, 00:36:57.270 --> 00:36:59.230 align:middle line:84% which is something they didn't know. 00:36:59.230 --> 00:37:03.670 align:middle line:84% So that is something that kids did not know. 00:37:03.670 --> 00:37:05.890 align:middle line:84% And then the way she looked at the impact 00:37:05.890 --> 00:37:08.770 align:middle line:90% is by looking at pregnancy. 00:37:08.770 --> 00:37:12.370 align:middle line:84% Because HIV is rare and hard to test. 00:37:12.370 --> 00:37:15.610 align:middle line:84% But pregnancy is a pretty good indication 00:37:15.610 --> 00:37:18.740 align:middle line:90% of risky sexual behavior. 00:37:18.740 --> 00:37:28.600 align:middle line:84% And she finds a decline in pregnancy of 31%. 00:37:28.600 --> 00:37:30.790 align:middle line:84% It's not that they had 100 pregnancy 00:37:30.790 --> 00:37:33.280 align:middle line:84% but it's normalized to 100 for some reason. 00:37:33.280 --> 00:37:38.650 align:middle line:84% So it's normal at a 100, and then 31 of this 100 00:37:38.650 --> 00:37:40.060 align:middle line:90% were averted. 00:37:40.060 --> 00:37:43.780 align:middle line:84% So it's a pretty large impact of one piece of information 00:37:43.780 --> 00:37:45.820 align:middle line:90% that people didn't have. 00:37:45.820 --> 00:37:49.450 align:middle line:84% So one possibility is that somehow the information 00:37:49.450 --> 00:37:53.610 align:middle line:84% doesn't go to-- people don't know the impact. 00:37:53.610 --> 00:37:57.180 align:middle line:84% And another paper by Pascaline also 00:37:57.180 --> 00:38:01.590 align:middle line:84% pushes this information thing, information exploration 00:38:01.590 --> 00:38:04.290 align:middle line:84% by looking at what happens if people get a first bed 00:38:04.290 --> 00:38:08.160 align:middle line:84% net for free or the first offer of a bed net for free, 00:38:08.160 --> 00:38:09.960 align:middle line:84% they are more likely to actually get 00:38:09.960 --> 00:38:13.770 align:middle line:84% a bed net than people who got an offer 00:38:13.770 --> 00:38:15.270 align:middle line:90% and were asked to pay something. 00:38:15.270 --> 00:38:17.640 align:middle line:84% So this is not from the first experiment I showed you. 00:38:17.640 --> 00:38:19.350 align:middle line:84% This is from a [? second ?] paper 00:38:19.350 --> 00:38:21.480 align:middle line:84% where she randomized, at the individual level, 00:38:21.480 --> 00:38:27.000 align:middle line:84% by giving people vouchers to buy bed nets. 00:38:27.000 --> 00:38:30.060 align:middle line:84% So in the first study, it was done in the health clinic 00:38:30.060 --> 00:38:31.980 align:middle line:84% and randomized at the health clinic level. 00:38:31.980 --> 00:38:35.790 align:middle line:84% This one, it was done in just distributing people voucher 00:38:35.790 --> 00:38:38.910 align:middle line:84% and said, you can buy yourself a long lasting net, 00:38:38.910 --> 00:38:42.990 align:middle line:84% either for free or by paying something. 00:38:42.990 --> 00:38:47.220 align:middle line:84% And the full price is about 500 or 600 shillings, 00:38:47.220 --> 00:38:53.710 align:middle line:84% so it goes-- the subsidies were about from 50% to 100%. 00:38:53.710 --> 00:38:56.040 align:middle line:84% And you find the same results that we already 00:38:56.040 --> 00:38:57.540 align:middle line:84% saw several times today, which is 00:38:57.540 --> 00:39:04.750 align:middle line:84% people are less likely to get a net if they have to pay for it. 00:39:04.750 --> 00:39:06.570 align:middle line:84% But then what is interesting is that she 00:39:06.570 --> 00:39:11.340 align:middle line:84% looked at-- she went back to this paper six months later. 00:39:11.340 --> 00:39:13.560 align:middle line:84% And she asked them, if they wanted 00:39:13.560 --> 00:39:16.260 align:middle line:84% to buy a net at a price that were-- 00:39:16.260 --> 00:39:19.260 align:middle line:84% I think it was somewhere in the middle-- 00:39:19.260 --> 00:39:24.070 align:middle line:84% at the price of 150 shillings, so kind of the mid-range price. 00:39:24.070 --> 00:39:26.100 align:middle line:84% So in the second period, everyone 00:39:26.100 --> 00:39:29.010 align:middle line:90% is offered the same discount. 00:39:29.010 --> 00:39:32.670 align:middle line:84% And then what we show in this-- what she shows in this graph is 00:39:32.670 --> 00:39:37.980 align:middle line:84% depending on the offer they got in the first period, 00:39:37.980 --> 00:39:41.610 align:middle line:84% are they more likely to get the second net? 00:39:41.610 --> 00:39:43.500 align:middle line:84% And what you find is people who got it 00:39:43.500 --> 00:39:46.950 align:middle line:84% for free or for very little money 00:39:46.950 --> 00:39:49.500 align:middle line:84% are more likely to purchase a net than people 00:39:49.500 --> 00:39:54.480 align:middle line:84% who were offered the net at some price, which, in a sense, 00:39:54.480 --> 00:39:55.530 align:middle line:90% is pretty striking. 00:39:55.530 --> 00:39:58.565 align:middle line:90% Because they already have one. 00:39:58.565 --> 00:39:59.940 align:middle line:84% So for them, that's a second net. 00:39:59.940 --> 00:40:03.120 align:middle line:84% So the marginal benefit of a second net 00:40:03.120 --> 00:40:05.100 align:middle line:84% should be lower than the marginal benefits 00:40:05.100 --> 00:40:06.600 align:middle line:90% of a first net. 00:40:06.600 --> 00:40:10.080 align:middle line:84% And despite that, they are more likely to get one. 00:40:10.080 --> 00:40:14.325 align:middle line:84% This, by the way, again, goes against another argument 00:40:14.325 --> 00:40:16.200 align:middle line:84% of the people who think that we should always 00:40:16.200 --> 00:40:18.300 align:middle line:90% make people pay for things. 00:40:18.300 --> 00:40:21.600 align:middle line:84% Which is that people get used to handouts. 00:40:21.600 --> 00:40:23.430 align:middle line:84% And once you make people pay for something, 00:40:23.430 --> 00:40:26.620 align:middle line:84% they refuse to pay for anything else. 00:40:26.620 --> 00:40:28.260 align:middle line:84% And what this graph shows is that, 00:40:28.260 --> 00:40:30.420 align:middle line:84% instead, people get used to nets rather 00:40:30.420 --> 00:40:32.850 align:middle line:90% than getting used to hand outs. 00:40:32.850 --> 00:40:35.460 align:middle line:90% So it could be learning. 00:40:35.460 --> 00:40:38.380 align:middle line:84% It could be habit formation, as well. 00:40:38.380 --> 00:40:42.690 align:middle line:84% So the sleeping under a net is a little painful. 00:40:42.690 --> 00:40:43.920 align:middle line:90% You have to set it up. 00:40:43.920 --> 00:40:47.340 align:middle line:90% It's a little bit hot inside. 00:40:47.340 --> 00:40:52.630 align:middle line:84% And so it has to become part of your routine. 00:40:52.630 --> 00:40:56.070 align:middle line:84% So one possibility also is, just by having your first net, 00:40:56.070 --> 00:40:59.310 align:middle line:84% you got into this habit of using the net. 00:40:59.310 --> 00:41:01.510 align:middle line:90% And therefore, you continue. 00:41:01.510 --> 00:41:05.610 align:middle line:84% So if you wanted to separate the fact that you provide 00:41:05.610 --> 00:41:08.640 align:middle line:84% an incentive for behavior, for a while, and then it goes away 00:41:08.640 --> 00:41:12.270 align:middle line:84% and people continue, it can be interpreted either 00:41:12.270 --> 00:41:14.900 align:middle line:84% as learning about the benefits or the costs 00:41:14.900 --> 00:41:16.650 align:middle line:84% of this technology, saying, actually, it's 00:41:16.650 --> 00:41:18.390 align:middle line:84% not that hard to sleep under a net 00:41:18.390 --> 00:41:21.300 align:middle line:84% or it's really super-cool not to have mosquitos 00:41:21.300 --> 00:41:24.250 align:middle line:84% or my kid didn't get malaria for the whole six months. 00:41:24.250 --> 00:41:25.360 align:middle line:90% So this is really great. 00:41:25.360 --> 00:41:28.140 align:middle line:84% That's the scale in favor of the explanation. 00:41:28.140 --> 00:41:29.820 align:middle line:84% Or it could be a habit formation, 00:41:29.820 --> 00:41:32.750 align:middle line:84% which is it becomes less costly to you 00:41:32.750 --> 00:41:35.530 align:middle line:90% as you are getting used to it. 00:41:35.530 --> 00:41:40.770 align:middle line:84% So there is a paper that I'm not going to go through in detail, 00:41:40.770 --> 00:41:45.550 align:middle line:84% now, by Reshmaan Hussam, who looks at trying to build up 00:41:45.550 --> 00:41:49.980 align:middle line:84% a habit for hand washing, where she showed that-- where 00:41:49.980 --> 00:41:53.880 align:middle line:84% she tries to set up tests that can allow you to distinguish 00:41:53.880 --> 00:41:56.970 align:middle line:84% between these two explanations in the case of hand washing. 00:41:56.970 --> 00:41:58.110 align:middle line:90% Yep? 00:41:58.110 --> 00:42:01.252 align:middle line:84% AUDIENCE: What does it actually take to set up a bed net? 00:42:01.252 --> 00:42:04.410 align:middle line:84% What are the actual steps to setting it up? 00:42:04.410 --> 00:42:06.755 align:middle line:84% ESTHER DUFLO: How long does it take to set up a bed net? 00:42:06.755 --> 00:42:07.380 align:middle line:90% AUDIENCE: Yeah. 00:42:07.380 --> 00:42:09.600 align:middle line:84% ESTHER DUFLO: Well, depending how you said set it up 00:42:09.600 --> 00:42:10.270 align:middle line:90% in your house. 00:42:10.270 --> 00:42:17.010 align:middle line:84% So typically, people attach it to a hook on their roof. 00:42:17.010 --> 00:42:20.520 align:middle line:84% Once it's attached to the hook on their roof, 00:42:20.520 --> 00:42:21.960 align:middle line:90% it's really not much. 00:42:21.960 --> 00:42:25.210 align:middle line:84% You have to unfold it every evening. 00:42:25.210 --> 00:42:27.030 align:middle line:90% So it's a few minutes. 00:42:27.030 --> 00:42:31.500 align:middle line:84% And then so if you do that, if you hook it properly and put 00:42:31.500 --> 00:42:33.370 align:middle line:90% it, it's really not that bad. 00:42:33.370 --> 00:42:35.940 align:middle line:84% AUDIENCE: So if you have two nets, once you are operating 00:42:35.940 --> 00:42:37.770 align:middle line:84% and it crosses [INAUDIBLE] one net, 00:42:37.770 --> 00:42:40.596 align:middle line:84% the second one would take a shorter amount of time 00:42:40.596 --> 00:42:42.923 align:middle line:90% relative to the first one. 00:42:42.923 --> 00:42:44.840 align:middle line:84% ESTHER DUFLO: I would say it would be similar. 00:42:44.840 --> 00:42:47.480 align:middle line:84% Because you would put it elsewhere in your house. 00:42:47.480 --> 00:42:51.110 align:middle line:90% And you put a second hook. 00:42:51.110 --> 00:42:53.990 align:middle line:84% But yeah, you also would have known how to use it 00:42:53.990 --> 00:42:56.120 align:middle line:84% or would have learned how to do it, 00:42:56.120 --> 00:42:58.440 align:middle line:84% efficiently, in a way that works for you. 00:42:58.440 --> 00:43:00.432 align:middle line:90% Yeah? 00:43:00.432 --> 00:43:02.390 align:middle line:84% AUDIENCE: This can be shown having been offered 00:43:02.390 --> 00:43:04.028 align:middle line:90% a net for free not having used? 00:43:04.028 --> 00:43:05.570 align:middle line:84% ESTHER DUFLO: Its conditional of them 00:43:05.570 --> 00:43:09.300 align:middle line:84% being offered the net for free and not even having it. 00:43:09.300 --> 00:43:11.897 align:middle line:84% So these guys are more likely to have-- 00:43:11.897 --> 00:43:13.730 align:middle line:84% these guys are much more likely to have one. 00:43:13.730 --> 00:43:16.430 align:middle line:84% And of the people who have one, most people 00:43:16.430 --> 00:43:20.990 align:middle line:84% use it, in fact, regardless of how much they've paid for it. 00:43:20.990 --> 00:43:22.410 align:middle line:84% And it's not conditional on that. 00:43:22.410 --> 00:43:25.370 align:middle line:84% So those guys are actually more likely to have one and be 00:43:25.370 --> 00:43:27.236 align:middle line:90% using it. 00:43:27.236 --> 00:43:30.420 align:middle line:84% AUDIENCE: Could there be like a deadline story, something 00:43:30.420 --> 00:43:32.970 align:middle line:84% similar to a deadline story, where 00:43:32.970 --> 00:43:37.470 align:middle line:84% by having gotten an offer at $0 and then seeing 00:43:37.470 --> 00:43:39.900 align:middle line:84% the difference in the price increase, 00:43:39.900 --> 00:43:44.520 align:middle line:84% you might forecast the next option I have to buy a net is 00:43:44.520 --> 00:43:47.270 align:middle line:90% going to be 100% whereas-- 00:43:47.270 --> 00:43:49.770 align:middle line:84% ESTHER DUFLO: Oh, that's right, you might forecast inflation 00:43:49.770 --> 00:43:52.710 align:middle line:84% in bed net prices and saying, this is my last option. 00:43:52.710 --> 00:43:54.090 align:middle line:90% That's totally possible. 00:43:54.090 --> 00:43:55.950 align:middle line:90% That's a good point. 00:43:55.950 --> 00:43:57.690 align:middle line:84% Saying if you forecast inflation, 00:43:57.690 --> 00:43:59.400 align:middle line:90% you're just going to buy now. 00:43:59.400 --> 00:44:00.840 align:middle line:90% They are true. 00:44:00.840 --> 00:44:04.500 align:middle line:84% But this good deal is definitely going away. 00:44:04.500 --> 00:44:05.430 align:middle line:90% That's good. 00:44:05.430 --> 00:44:06.640 align:middle line:90% I had not thought of that. 00:44:06.640 --> 00:44:07.920 align:middle line:90% That's a good idea. 00:44:07.920 --> 00:44:11.870 align:middle line:90% 00:44:11.870 --> 00:44:14.210 align:middle line:84% So these two-- just again, two pieces 00:44:14.210 --> 00:44:17.300 align:middle line:84% of evidence that suggest that there is a lot of things people 00:44:17.300 --> 00:44:22.130 align:middle line:84% don't know and when they learn, they actually 00:44:22.130 --> 00:44:24.080 align:middle line:90% adjust their behavior. 00:44:24.080 --> 00:44:27.230 align:middle line:84% And the other problem we have is that learning about health 00:44:27.230 --> 00:44:30.200 align:middle line:84% is really difficult. In particular, 00:44:30.200 --> 00:44:32.360 align:middle line:84% learning about preventive health and the difference 00:44:32.360 --> 00:44:35.490 align:middle line:84% between curative and preventive health is very difficult. 00:44:35.490 --> 00:44:40.880 align:middle line:84% So there is a very nice paper by Jishnu Das and Monica Sanchez, 00:44:40.880 --> 00:44:44.315 align:middle line:84% where what they are looking at is curative care. 00:44:44.315 --> 00:44:47.300 align:middle line:90% 00:44:47.300 --> 00:44:50.990 align:middle line:84% And again, this could be a model of information 00:44:50.990 --> 00:44:59.330 align:middle line:84% and why ivermectin became a drug of choice in the case of COVID, 00:44:59.330 --> 00:45:02.270 align:middle line:84% even if, in fact, it doesn't work. 00:45:02.270 --> 00:45:03.770 align:middle line:84% Assuming that, in fact, it doesn't 00:45:03.770 --> 00:45:07.430 align:middle line:84% work, why could people think that it works even when 00:45:07.430 --> 00:45:08.330 align:middle line:90% it doesn't work? 00:45:08.330 --> 00:45:12.200 align:middle line:84% Well, it's because COVID is like the flu. 00:45:12.200 --> 00:45:15.200 align:middle line:84% For a lot of people, it's self-limiting. 00:45:15.200 --> 00:45:18.780 align:middle line:84% That means that a lot of people will get better, regardless. 00:45:18.780 --> 00:45:23.840 align:middle line:84% And so they take ivermectin or pink Gatorade, 00:45:23.840 --> 00:45:26.660 align:middle line:90% and then they get better. 00:45:26.660 --> 00:45:30.710 align:middle line:84% And then they can very easily attribute 00:45:30.710 --> 00:45:35.250 align:middle line:84% the fact of getting better to ivermectin or to pink Gatorade. 00:45:35.250 --> 00:45:40.082 align:middle line:84% So in the case of India, and it's same thing 00:45:40.082 --> 00:45:41.540 align:middle line:84% in many countries in Africa, people 00:45:41.540 --> 00:45:46.980 align:middle line:84% overuse antibiotic and in very creative way. 00:45:46.980 --> 00:45:49.560 align:middle line:84% So for example, people will prescribe you 00:45:49.560 --> 00:45:52.200 align:middle line:84% one short course, what they call a short course of antibiotic, 00:45:52.200 --> 00:45:53.870 align:middle line:90% which is like one pill. 00:45:53.870 --> 00:45:56.360 align:middle line:84% You take antibiotic once, that doesn't help you. 00:45:56.360 --> 00:45:58.040 align:middle line:90% In fact, that doesn't help you. 00:45:58.040 --> 00:45:59.270 align:middle line:90% It creates resistance. 00:45:59.270 --> 00:46:00.590 align:middle line:90% It hurts everyone else. 00:46:00.590 --> 00:46:03.922 align:middle line:84% But you'll feel better if what you had was self-limiting. 00:46:03.922 --> 00:46:05.630 align:middle line:84% And then you think, well, that was great. 00:46:05.630 --> 00:46:10.280 align:middle line:84% This one shot-- preferably a shot, an actual shot. 00:46:10.280 --> 00:46:14.370 align:middle line:84% This one shot of antibiotic had a great effect on me. 00:46:14.370 --> 00:46:17.070 align:middle line:84% So next time I'm sick, I'm going to do this again. 00:46:17.070 --> 00:46:19.190 align:middle line:84% And if you do that again, a few times, every time 00:46:19.190 --> 00:46:22.370 align:middle line:84% you confirm your opinion that it really works out very nicely. 00:46:22.370 --> 00:46:24.900 align:middle line:90% 00:46:24.900 --> 00:46:28.400 align:middle line:84% So learning about doctor quality is really hard and especially 00:46:28.400 --> 00:46:34.910 align:middle line:84% if it is hard for people to think about a model 00:46:34.910 --> 00:46:38.030 align:middle line:84% where you do nothing and you got better. 00:46:38.030 --> 00:46:40.880 align:middle line:84% You don't attribute it to doing nothing. 00:46:40.880 --> 00:46:43.640 align:middle line:84% Whereas, if you do something and you get better, and especially 00:46:43.640 --> 00:46:45.410 align:middle line:84% if someone has told you, has given you 00:46:45.410 --> 00:46:48.560 align:middle line:84% the causal model saying, I'm doing this 00:46:48.560 --> 00:46:49.890 align:middle line:90% and you will get better. 00:46:49.890 --> 00:46:52.430 align:middle line:84% And in fact, you get better, then it's hard to-- 00:46:52.430 --> 00:46:57.500 align:middle line:84% So the kind of fake snake oil-type medicine 00:46:57.500 --> 00:47:01.100 align:middle line:84% are more likely to take hold with medicines 00:47:01.100 --> 00:47:04.760 align:middle line:84% that have either this fluctuating tendency 00:47:04.760 --> 00:47:08.180 align:middle line:90% or are self-limiting. 00:47:08.180 --> 00:47:11.150 align:middle line:84% Now, with preventive care, trying 00:47:11.150 --> 00:47:13.305 align:middle line:84% to understand the benefits of preventive care 00:47:13.305 --> 00:47:15.680 align:middle line:84% is really difficult. Because you do something and nothing 00:47:15.680 --> 00:47:17.570 align:middle line:90% happens. 00:47:17.570 --> 00:47:21.950 align:middle line:84% And then if the vaccine works and protects you, 00:47:21.950 --> 00:47:28.220 align:middle line:84% you don't get COVID or you don't get measles. 00:47:28.220 --> 00:47:30.150 align:middle line:84% At least there are a lot of COVID. 00:47:30.150 --> 00:47:33.200 align:middle line:84% So if you're the one vaccinated person 00:47:33.200 --> 00:47:35.190 align:middle line:84% in a group of unvaccinated people, 00:47:35.190 --> 00:47:37.190 align:middle line:84% and you don't get it when everyone else gets it, 00:47:37.190 --> 00:47:38.500 align:middle line:90% maybe you learn something. 00:47:38.500 --> 00:47:41.000 align:middle line:90% But even there, it's harder. 00:47:41.000 --> 00:47:44.460 align:middle line:84% But measles is too frequent and kids die from it, 00:47:44.460 --> 00:47:45.860 align:middle line:90% but it's not that frequent. 00:47:45.860 --> 00:47:49.040 align:middle line:84% And so you might never really get an opportunity 00:47:49.040 --> 00:47:50.600 align:middle line:90% to experience it. 00:47:50.600 --> 00:47:53.660 align:middle line:84% And it's particularly difficult to link 00:47:53.660 --> 00:47:57.290 align:middle line:84% cause and effect in the cases like immunization, 00:47:57.290 --> 00:47:59.910 align:middle line:90% which is contagious. 00:47:59.910 --> 00:48:03.860 align:middle line:84% So for example, in the case of deworming, 00:48:03.860 --> 00:48:06.170 align:middle line:84% if everybody around you is dewormed, 00:48:06.170 --> 00:48:11.700 align:middle line:84% you won't get worms regardless of whether you're dewormed 00:48:11.700 --> 00:48:12.520 align:middle line:90% or not. 00:48:12.520 --> 00:48:14.082 align:middle line:84% So trying to understand the mother 00:48:14.082 --> 00:48:16.290 align:middle line:84% and whether this deworming medicine works, and you're 00:48:16.290 --> 00:48:18.570 align:middle line:84% comparing yourself, who hasn't taken it, 00:48:18.570 --> 00:48:20.100 align:middle line:84% to everyone else who has taken it, 00:48:20.100 --> 00:48:21.683 align:middle line:84% you're saying this thing is really not 00:48:21.683 --> 00:48:23.980 align:middle line:90% what it's made out to be. 00:48:23.980 --> 00:48:28.380 align:middle line:84% And the way that the benefits were sold to me were oversold. 00:48:28.380 --> 00:48:34.110 align:middle line:84% So in some sense, every time we try 00:48:34.110 --> 00:48:37.590 align:middle line:84% to explain the benefit of a vaccine to people, 00:48:37.590 --> 00:48:40.170 align:middle line:90% we are overselling it to them. 00:48:40.170 --> 00:48:44.700 align:middle line:84% Because we are trying to say this is the benefits this. 00:48:44.700 --> 00:48:47.310 align:middle line:84% Your private benefits are going to be huge. 00:48:47.310 --> 00:48:49.350 align:middle line:84% But it could be that the private benefits 00:48:49.350 --> 00:48:56.760 align:middle line:84% are huge in a way that is impossible to assess 00:48:56.760 --> 00:48:58.200 align:middle line:90% for people. 00:48:58.200 --> 00:49:05.080 align:middle line:84% So that would explain why, without anything else, 00:49:05.080 --> 00:49:07.350 align:middle line:84% the standard information story would 00:49:07.350 --> 00:49:14.100 align:middle line:84% make people quite skeptical of the benefits 00:49:14.100 --> 00:49:17.430 align:middle line:84% and, therefore, more indifferent than we think 00:49:17.430 --> 00:49:19.320 align:middle line:84% they ought to be and, therefore, removing 00:49:19.320 --> 00:49:20.907 align:middle line:90% the puzzle, in a sense. 00:49:20.907 --> 00:49:22.740 align:middle line:84% And so, now, if they are mostly indifferent, 00:49:22.740 --> 00:49:23.790 align:middle line:90% then we can reinterpret. 00:49:23.790 --> 00:49:25.020 align:middle line:90% We can go back-- 00:49:25.020 --> 00:49:29.100 align:middle line:84% going back to a little bit of time inconsistency 00:49:29.100 --> 00:49:32.460 align:middle line:84% on top of a huge zone of indifference, 00:49:32.460 --> 00:49:35.400 align:middle line:84% anyways, will create a large number 00:49:35.400 --> 00:49:38.790 align:middle line:84% of people whose behavior will look very inert. 00:49:38.790 --> 00:49:41.900 align:middle line:84% And then the puzzle somehow just disappears. 00:49:41.900 --> 00:49:46.580 align:middle line:84% On top of it, in this context, where 00:49:46.580 --> 00:49:48.830 align:middle line:84% forming your own models of the world 00:49:48.830 --> 00:49:51.260 align:middle line:84% is really hard, because you don't have the right data, 00:49:51.260 --> 00:49:55.340 align:middle line:84% and it's complicated and all that, you would imagine 00:49:55.340 --> 00:49:58.940 align:middle line:84% this is the place where trust would be the most important 00:49:58.940 --> 00:50:00.260 align:middle line:90% word you have to have. 00:50:00.260 --> 00:50:02.810 align:middle line:84% If you cannot make up the story yourself, 00:50:02.810 --> 00:50:05.960 align:middle line:84% you have to have trusted the person who is telling you to do 00:50:05.960 --> 00:50:07.010 align:middle line:90% it. 00:50:07.010 --> 00:50:09.320 align:middle line:84% Frankly, I have no real understanding 00:50:09.320 --> 00:50:12.800 align:middle line:90% of how mRNA vaccines work. 00:50:12.800 --> 00:50:16.760 align:middle line:84% And I have no reason to do it, from my own experience, 00:50:16.760 --> 00:50:19.880 align:middle line:84% whether this will implant a chip in my head 00:50:19.880 --> 00:50:23.120 align:middle line:84% that is going to make me an agent of the CIA. 00:50:23.120 --> 00:50:27.980 align:middle line:84% But I trust the people, the system, 00:50:27.980 --> 00:50:32.060 align:middle line:84% et cetera, to both reassure me that they 00:50:32.060 --> 00:50:34.670 align:middle line:84% in fact work and the randomized controlled trial 00:50:34.670 --> 00:50:37.130 align:middle line:84% have been done properly and so on and so forth, 00:50:37.130 --> 00:50:41.160 align:middle line:84% and that they don't have those side effects. 00:50:41.160 --> 00:50:42.530 align:middle line:90% But if you don't-- 00:50:42.530 --> 00:50:44.780 align:middle line:84% And so this is where the trust in the government 00:50:44.780 --> 00:50:47.400 align:middle line:84% plays a key role, and our trust in official sources, 00:50:47.400 --> 00:50:49.770 align:middle line:84% the government, the nurses, the doctors, et cetera. 00:50:49.770 --> 00:50:54.510 align:middle line:84% But both in OECD countries and in developing countries, 00:50:54.510 --> 00:50:57.200 align:middle line:84% there has been a huge decline in the trust of government 00:50:57.200 --> 00:51:00.170 align:middle line:90% over the last 10, 20 years. 00:51:00.170 --> 00:51:02.540 align:middle line:84% In the US, it's at an extremely low level. 00:51:02.540 --> 00:51:07.067 align:middle line:84% So it is not entirely surprising that people cannot get 00:51:07.067 --> 00:51:09.650 align:middle line:84% the whole story of the benefits, because they don't understand 00:51:09.650 --> 00:51:12.140 align:middle line:84% the mechanism, as they shouldn't be expected to. 00:51:12.140 --> 00:51:14.480 align:middle line:84% And they don't trust the messengers. 00:51:14.480 --> 00:51:19.910 align:middle line:84% In developing countries, this mistrust in health services 00:51:19.910 --> 00:51:24.920 align:middle line:84% has, in many cases, some roots, because some public health 00:51:24.920 --> 00:51:27.680 align:middle line:84% campaigns have been botched or some people have 00:51:27.680 --> 00:51:30.920 align:middle line:84% been using public health campaigns to do other things 00:51:30.920 --> 00:51:32.760 align:middle line:90% than what they should be doing. 00:51:32.760 --> 00:51:37.820 align:middle line:84% So going back to our friend, ABCD, or the ABCD campaign 00:51:37.820 --> 00:51:44.770 align:middle line:84% against HIV, quite explicitly, the government people 00:51:44.770 --> 00:51:52.360 align:middle line:84% will tell you that it's not just about HIV. 00:51:52.360 --> 00:51:54.850 align:middle line:84% But in addition, if kids were less promiscuous, 00:51:54.850 --> 00:51:57.040 align:middle line:84% it would have a lot of good side effects. 00:51:57.040 --> 00:52:00.350 align:middle line:84% We wouldn't have early pregnancies and all that. 00:52:00.350 --> 00:52:02.820 align:middle line:84% So we might as well scare them to death on HIV. 00:52:02.820 --> 00:52:04.820 align:middle line:84% And, therefore, we'll get these things, as well. 00:52:04.820 --> 00:52:07.270 align:middle line:84% So we know, of course, that using a condom 00:52:07.270 --> 00:52:10.147 align:middle line:90% would be good enough. 00:52:10.147 --> 00:52:11.230 align:middle line:90% You don't need to abstain. 00:52:11.230 --> 00:52:16.030 align:middle line:84% We know that if they had a sexual relationship only 00:52:16.030 --> 00:52:20.200 align:middle line:84% among themselves, it would be much safer for HIV. 00:52:20.200 --> 00:52:22.780 align:middle line:84% But we are not going to tell them, because, well, we might 00:52:22.780 --> 00:52:24.980 align:middle line:90% as well get them to behave. 00:52:24.980 --> 00:52:27.520 align:middle line:84% So this is just as an example of this. 00:52:27.520 --> 00:52:29.350 align:middle line:84% It's not necessarily sinister motives, 00:52:29.350 --> 00:52:34.210 align:middle line:84% but there is always a tendency of slightly-- or misusing 00:52:34.210 --> 00:52:38.290 align:middle line:84% your power, information power over other people 00:52:38.290 --> 00:52:39.760 align:middle line:84% that you are supposed to instruct. 00:52:39.760 --> 00:52:43.600 align:middle line:84% Because, well then, you might as well get some benefits. 00:52:43.600 --> 00:52:48.340 align:middle line:84% This mistrust in the health system 00:52:48.340 --> 00:52:51.550 align:middle line:84% has a huge impact in people believing 00:52:51.550 --> 00:52:56.200 align:middle line:84% things like COVID vaccines but also childhood vaccinations. 00:52:56.200 --> 00:52:59.380 align:middle line:84% For example, in India, there has been-- 00:52:59.380 --> 00:53:02.740 align:middle line:90% 00:53:02.740 --> 00:53:06.760 align:middle line:84% during the emergency period in the late '70s, 00:53:06.760 --> 00:53:08.410 align:middle line:84% when Mrs. Gandhi was in power, there 00:53:08.410 --> 00:53:11.530 align:middle line:84% was a huge campaign of forced sterilization. 00:53:11.530 --> 00:53:14.440 align:middle line:84% And it was disguised in many ways. 00:53:14.440 --> 00:53:15.820 align:middle line:90% And people were sent to camps. 00:53:15.820 --> 00:53:18.770 align:middle line:90% And people remember that. 00:53:18.770 --> 00:53:21.490 align:middle line:84% And so I've long believed-- although I still 00:53:21.490 --> 00:53:23.530 align:middle line:84% haven't seen the study yet, I have long 00:53:23.530 --> 00:53:27.460 align:middle line:84% believed that the forced sterilization, that 00:53:27.460 --> 00:53:30.310 align:middle line:84% in areas where you have more forced sterilization, 00:53:30.310 --> 00:53:34.750 align:middle line:84% there is more mistrust in the health system today 00:53:34.750 --> 00:53:36.730 align:middle line:84% and, therefore, less take-up of services. 00:53:36.730 --> 00:53:40.190 align:middle line:84% So although this is a study I've always wanted to do, 00:53:40.190 --> 00:53:43.250 align:middle line:84% and I've never seen a good version of it, 00:53:43.250 --> 00:53:47.020 align:middle line:84% although I've seen some not so good version of it. 00:53:47.020 --> 00:53:49.960 align:middle line:84% For India, there is one paper that comes very, very close 00:53:49.960 --> 00:53:57.490 align:middle line:84% to it, which is a paper by Sara Lowes and Montero, 00:53:57.490 --> 00:54:02.710 align:middle line:84% where they look at Cameroon, which used to be separated 00:54:02.710 --> 00:54:07.960 align:middle line:84% between France and England during the colonial period 00:54:07.960 --> 00:54:11.060 align:middle line:90% and is now one country. 00:54:11.060 --> 00:54:13.870 align:middle line:84% And the French and the British had different ideas 00:54:13.870 --> 00:54:15.160 align:middle line:90% on different things. 00:54:15.160 --> 00:54:16.990 align:middle line:84% And in particular, at some point, 00:54:16.990 --> 00:54:22.330 align:middle line:84% the French got super worried about sleeping sickness. 00:54:22.330 --> 00:54:25.450 align:middle line:84% And they undertook a massive campaign 00:54:25.450 --> 00:54:29.890 align:middle line:90% against sleeping sickness. 00:54:29.890 --> 00:54:33.190 align:middle line:90% So they vaccinated everyone. 00:54:33.190 --> 00:54:36.640 align:middle line:84% And the vaccine was not that good, and a lot of kids died. 00:54:36.640 --> 00:54:40.130 align:middle line:90% And it was just a bad idea. 00:54:40.130 --> 00:54:42.130 align:middle line:84% And those places still have low trust 00:54:42.130 --> 00:54:45.130 align:middle line:84% in medicine today compared to comparable places 00:54:45.130 --> 00:54:48.040 align:middle line:84% in the same countries that were under British rule 00:54:48.040 --> 00:54:52.980 align:middle line:84% and that did not receive the same vaccination. 00:54:52.980 --> 00:54:56.850 align:middle line:84% So that's an example of long shadow of mistrust. 00:54:56.850 --> 00:55:01.560 align:middle line:84% Another example from the US is Marcella Alsan and Marianne 00:55:01.560 --> 00:55:06.390 align:middle line:84% Wanamaker study on the Tuskegee trials, where 00:55:06.390 --> 00:55:11.130 align:middle line:84% places that are closer to Tuskegee County, that 00:55:11.130 --> 00:55:14.460 align:middle line:84% are closer to Tuskegee have less trust in government today. 00:55:14.460 --> 00:55:18.090 align:middle line:90% But this is exactly the idea. 00:55:18.090 --> 00:55:20.760 align:middle line:90% Tuskegee was awful. 00:55:20.760 --> 00:55:24.330 align:middle line:90% And this was just misguided. 00:55:24.330 --> 00:55:26.160 align:middle line:90% If you see the-- 00:55:26.160 --> 00:55:29.830 align:middle line:84% This Tuskegee was actual bad behavior-- people lying, 00:55:29.830 --> 00:55:30.330 align:middle line:90% et cetera. 00:55:30.330 --> 00:55:33.390 align:middle line:84% This is just they were trying to do a public health campaign, 00:55:33.390 --> 00:55:35.910 align:middle line:84% and it turns out their product was not great. 00:55:35.910 --> 00:55:37.530 align:middle line:90% And this is still-- 00:55:37.530 --> 00:55:41.100 align:middle line:84% kids today are less vaccinated in those places compared 00:55:41.100 --> 00:55:48.300 align:middle line:84% to very, very close by places, which were under British rule. 00:55:48.300 --> 00:55:51.360 align:middle line:84% Another paper along the same line, still on vaccination, 00:55:51.360 --> 00:55:55.020 align:middle line:84% is a paper, by Monica Martinez-Bravo and Andreas 00:55:55.020 --> 00:55:59.410 align:middle line:84% Stegmann, looking at the CIA vaccine ruse. 00:55:59.410 --> 00:56:03.480 align:middle line:84% So when the Americans wanted to get Bin Laden, 00:56:03.480 --> 00:56:06.720 align:middle line:84% they needed to collect information about, I think, 00:56:06.720 --> 00:56:08.598 align:middle line:90% the DNA of his kids. 00:56:08.598 --> 00:56:12.300 align:middle line:84% And they send a guy to do some vaccination campaign 00:56:12.300 --> 00:56:15.990 align:middle line:84% in Pakistan, where he was hiding, in order 00:56:15.990 --> 00:56:20.040 align:middle line:84% to collect swabs, to have swabs of the kids, 00:56:20.040 --> 00:56:24.510 align:middle line:84% so they could find out that it was in fact him. 00:56:24.510 --> 00:56:25.150 align:middle line:90% They found him. 00:56:25.150 --> 00:56:25.650 align:middle line:90% Whatever. 00:56:25.650 --> 00:56:26.317 align:middle line:90% They killed him. 00:56:26.317 --> 00:56:30.630 align:middle line:84% And after that, if you look at the places where that happened, 00:56:30.630 --> 00:56:34.050 align:middle line:84% there is less vaccination today than in other places. 00:56:34.050 --> 00:56:37.440 align:middle line:90% So that's another example. 00:56:37.440 --> 00:56:41.490 align:middle line:84% So this mistrust in the government-- 00:56:41.490 --> 00:56:45.210 align:middle line:84% And therefore, contrary to the impact 00:56:45.210 --> 00:56:48.580 align:middle line:84% of very well-targeted pieces of information, 00:56:48.580 --> 00:56:50.730 align:middle line:84% for example, how costly it is to use a net 00:56:50.730 --> 00:56:55.620 align:middle line:84% or older men are more likely to be HIV positive than younger 00:56:55.620 --> 00:57:03.030 align:middle line:84% men, you have a low impact of general kind of "trust me, 00:57:03.030 --> 00:57:07.770 align:middle line:84% you should do this," without any concrete pieces of information 00:57:07.770 --> 00:57:09.600 align:middle line:90% attached to them. 00:57:09.600 --> 00:57:13.740 align:middle line:84% So in our study of the ABCD campaign, 00:57:13.740 --> 00:57:18.900 align:middle line:84% we find no impact of the ABCD campaign on behavior. 00:57:18.900 --> 00:57:20.850 align:middle line:84% In the same setting, where Pascaline 00:57:20.850 --> 00:57:24.360 align:middle line:84% find large effect of that more concrete piece of information. 00:57:24.360 --> 00:57:27.480 align:middle line:84% And also in Kenya, Kremer and Miguel, 00:57:27.480 --> 00:57:30.240 align:middle line:84% along with a deworming program, have a campaign 00:57:30.240 --> 00:57:32.100 align:middle line:84% where they're trying to tell people, 00:57:32.100 --> 00:57:38.430 align:middle line:84% kids, don't walk barefoot, don't go fishing in the lake. 00:57:38.430 --> 00:57:42.930 align:middle line:90% And then that has zero impact. 00:57:42.930 --> 00:57:45.900 align:middle line:84% That's because it doesn't have a new information. 00:57:45.900 --> 00:57:48.480 align:middle line:84% It's just some general preaching message. 00:57:48.480 --> 00:57:52.050 align:middle line:90% And that is not effective. 00:57:52.050 --> 00:57:56.210 align:middle line:90% 00:57:56.210 --> 00:57:59.190 align:middle line:90% So that's the trust part. 00:57:59.190 --> 00:58:04.100 align:middle line:84% So now if we are in this kind of setting, 00:58:04.100 --> 00:58:07.880 align:middle line:84% then you would imagine that the social norm, 00:58:07.880 --> 00:58:11.570 align:middle line:84% the social equilibrium would play a pretty important role, 00:58:11.570 --> 00:58:16.760 align:middle line:84% first of all, because it can substitute or, on the contrary, 00:58:16.760 --> 00:58:19.790 align:middle line:84% annihilate whatever trust exists in the system 00:58:19.790 --> 00:58:27.470 align:middle line:84% if a community agrees to a common behavior. 00:58:27.470 --> 00:58:30.410 align:middle line:84% Second of all, because-- since people understand-- 00:58:30.410 --> 00:58:35.150 align:middle line:84% if people understand that the benefits to immunization 00:58:35.150 --> 00:58:41.900 align:middle line:84% is mainly social, then whether or not 00:58:41.900 --> 00:58:47.180 align:middle line:84% there is a social sanction or social reward associated 00:58:47.180 --> 00:58:53.210 align:middle line:84% with taking a social action can create the benefits. 00:58:53.210 --> 00:58:56.060 align:middle line:84% Now people see that, OK, I'm mainly 00:58:56.060 --> 00:58:58.970 align:middle line:84% doing this for the community, as far as I can understand. 00:58:58.970 --> 00:59:02.240 align:middle line:84% Is the community going to be grateful to me 00:59:02.240 --> 00:59:05.075 align:middle line:84% and going to manifest this gratefulness or not? 00:59:05.075 --> 00:59:08.390 align:middle line:90% 00:59:08.390 --> 00:59:14.980 align:middle line:84% So one way in which you could help 00:59:14.980 --> 00:59:18.850 align:middle line:84% the community reward people, who are doing 00:59:18.850 --> 00:59:22.570 align:middle line:84% the right thing for the community, is by helping people 00:59:22.570 --> 00:59:25.820 align:middle line:84% signal that they are the "right kind" of people. 00:59:25.820 --> 00:59:29.320 align:middle line:84% So when you go to vote, you get a "I voted" sticker. 00:59:29.320 --> 00:59:33.430 align:middle line:84% When people get vaccinated, sometimes, also, they 00:59:33.430 --> 00:59:34.180 align:middle line:90% get sticker. 00:59:34.180 --> 00:59:37.120 align:middle line:84% I don't if you got "I got vaccinated" stickers. 00:59:37.120 --> 00:59:39.250 align:middle line:90% I didn't get one. 00:59:39.250 --> 00:59:41.440 align:middle line:84% In France, they don't give you one. 00:59:41.440 --> 00:59:46.605 align:middle line:84% But for example, when doctors get-- 00:59:46.605 --> 00:59:48.730 align:middle line:84% also in France, when a health worker get vaccinated 00:59:48.730 --> 00:59:51.070 align:middle line:84% for the flu, they get a badge saying "I'm vaccinated. 00:59:51.070 --> 00:59:54.860 align:middle line:84% That's to protect you--" that kind of things. 00:59:54.860 --> 00:59:58.270 align:middle line:84% So Anne Karing did a very nice experiment. 00:59:58.270 --> 01:00:00.160 align:middle line:84% It was a job market paper, by the way, 01:00:00.160 --> 01:00:02.050 align:middle line:84% so it's worth reading it in order 01:00:02.050 --> 01:00:07.750 align:middle line:84% to get a sense of what you could do as a student. 01:00:07.750 --> 01:00:11.770 align:middle line:84% She said let's try and help people 01:00:11.770 --> 01:00:17.272 align:middle line:84% signal that they are the type of people who immunize their kids, 01:00:17.272 --> 01:00:18.730 align:middle line:84% either because they're good parents 01:00:18.730 --> 01:00:20.522 align:middle line:84% or because they are good for the community. 01:00:20.522 --> 01:00:25.240 align:middle line:84% So they give silicon bracelet to show that you have completed 01:00:25.240 --> 01:00:27.282 align:middle line:90% the sequence of immunization. 01:00:27.282 --> 01:00:28.990 align:middle line:84% She randomized that at the village level. 01:00:28.990 --> 01:00:31.090 align:middle line:84% And she found-- or the health center level. 01:00:31.090 --> 01:00:35.110 align:middle line:84% And she finds that, in places where the bracelets were given 01:00:35.110 --> 01:00:38.290 align:middle line:84% at the end of the sequence, bracelets were very popular, 01:00:38.290 --> 01:00:41.250 align:middle line:84% and they do increase immunization. 01:00:41.250 --> 01:00:43.030 align:middle line:90% Now, one question for you. 01:00:43.030 --> 01:00:45.660 align:middle line:90% How do you-- how does she know? 01:00:45.660 --> 01:00:50.100 align:middle line:84% How do you know that it's signaling, per se, that you're 01:00:50.100 --> 01:00:52.320 align:middle line:90% a good person? 01:00:52.320 --> 01:00:55.290 align:middle line:90% And what else could it be? 01:00:55.290 --> 01:00:55.950 align:middle line:90% Yeah? 01:00:55.950 --> 01:00:58.110 align:middle line:84% AUDIENCE: You want to make sure that-- 01:00:58.110 --> 01:01:01.140 align:middle line:84% if these are a form of verifying that he's actually 01:01:01.140 --> 01:01:03.600 align:middle line:84% been immunized, whereas, relative to before, there's 01:01:03.600 --> 01:01:06.732 align:middle line:84% no way of verifying it, so that businesses couldn't act on it. 01:01:06.732 --> 01:01:16.018 align:middle line:84% If you want to make sure that business aren't [INAUDIBLE] 01:01:16.018 --> 01:01:16.810 align:middle line:90% ESTHER DUFLO: Yeah. 01:01:16.810 --> 01:01:18.980 align:middle line:84% So you would see-- so here, by the way, 01:01:18.980 --> 01:01:20.060 align:middle line:90% it's not just businesses. 01:01:20.060 --> 01:01:21.568 align:middle line:84% It's like basically people around. 01:01:21.568 --> 01:01:23.860 align:middle line:84% It's not that people use that, like we use it in France 01:01:23.860 --> 01:01:24.880 align:middle line:90% to go anywhere. 01:01:24.880 --> 01:01:26.515 align:middle line:90% But it's a way of saying-- 01:01:26.515 --> 01:01:29.140 align:middle line:84% well, in her mind, it's a way of showing you've been immunized, 01:01:29.140 --> 01:01:30.390 align:middle line:90% to your friend and everything. 01:01:30.390 --> 01:01:32.680 align:middle line:84% And as you say, well, the point is, 01:01:32.680 --> 01:01:34.570 align:middle line:84% how do you know that people understand 01:01:34.570 --> 01:01:37.030 align:middle line:84% that it's, in fact, signaling that it's, 01:01:37.030 --> 01:01:39.290 align:middle line:84% in fact, a verification for that. 01:01:39.290 --> 01:01:42.880 align:middle line:84% And if it's not, why would people-- 01:01:42.880 --> 01:01:45.760 align:middle line:84% why would you, anyway, see potentially an impact 01:01:45.760 --> 01:01:47.710 align:middle line:90% of the bracelets? 01:01:47.710 --> 01:01:48.470 align:middle line:90% Yeah. 01:01:48.470 --> 01:01:50.418 align:middle line:84% AUDIENCE: People just like getting bracelets. 01:01:50.418 --> 01:01:51.210 align:middle line:90% ESTHER DUFLO: Yeah. 01:01:51.210 --> 01:01:53.358 align:middle line:84% Maybe people just like the bracelets. 01:01:53.358 --> 01:01:54.900 align:middle line:84% So you think, well, it's really tiny. 01:01:54.900 --> 01:01:57.630 align:middle line:84% But remember, people went and got vaccinated 01:01:57.630 --> 01:01:59.460 align:middle line:90% for a kilo of lentils. 01:01:59.460 --> 01:02:02.790 align:middle line:84% And a silicone bracelet is more fun than a-- 01:02:02.790 --> 01:02:05.400 align:middle line:84% for your kid is more fun than a kilo of lentils. 01:02:05.400 --> 01:02:10.260 align:middle line:84% So people might just like it, right. 01:02:10.260 --> 01:02:12.720 align:middle line:90% So what she does to test that? 01:02:12.720 --> 01:02:14.910 align:middle line:84% And that's a nice feature of the experiment 01:02:14.910 --> 01:02:16.650 align:middle line:84% is that she introduced a treatment, which 01:02:16.650 --> 01:02:19.410 align:middle line:84% is uninformative bracelets, where everybody 01:02:19.410 --> 01:02:23.170 align:middle line:90% gets a bracelet, anyways. 01:02:23.170 --> 01:02:27.330 align:middle line:84% And then you can compare the uninformative bracelets 01:02:27.330 --> 01:02:29.430 align:middle line:84% to the informative bracelets that you 01:02:29.430 --> 01:02:32.200 align:middle line:84% get after having completing the sequence. 01:02:32.200 --> 01:02:36.720 align:middle line:84% And in fact, although it's not quite how the paper is written, 01:02:36.720 --> 01:02:40.020 align:middle line:84% but if you look at the table, in fact, 01:02:40.020 --> 01:02:44.640 align:middle line:84% the uninformative bracelet-- so that's the part where, I think, 01:02:44.640 --> 01:02:49.780 align:middle line:84% she got maybe very happy with her results, 01:02:49.780 --> 01:02:51.400 align:middle line:90% and she wanted to-- 01:02:51.400 --> 01:02:54.960 align:middle line:84% But if you look at the table, the uninformative results 01:02:54.960 --> 01:02:58.440 align:middle line:84% are about as effective as the informative bracelets. 01:02:58.440 --> 01:03:03.510 align:middle line:84% So one way to read this is that you can't really 01:03:03.510 --> 01:03:05.910 align:middle line:84% reject that people just like bracelets, 01:03:05.910 --> 01:03:08.470 align:middle line:84% and that it's not the signal that works out. 01:03:08.470 --> 01:03:11.920 align:middle line:84% But it was super-well designed to figure it out. 01:03:11.920 --> 01:03:15.330 align:middle line:84% And now she's doing another one on deworming, 01:03:15.330 --> 01:03:19.800 align:middle line:84% where maybe we'll find more of a difference between signaling. 01:03:19.800 --> 01:03:26.810 align:middle line:90% 01:03:26.810 --> 01:03:32.090 align:middle line:84% So with many of these explanations, the information, 01:03:32.090 --> 01:03:35.760 align:middle line:84% trust, conveying the social norms, 01:03:35.760 --> 01:03:40.890 align:middle line:84% the social networks are likely to play a key role. 01:03:40.890 --> 01:03:46.430 align:middle line:84% So the idea then is how do you activate the social network, 01:03:46.430 --> 01:03:49.280 align:middle line:84% in the case of immunization or preventive good, 01:03:49.280 --> 01:03:52.265 align:middle line:90% to your benefits? 01:03:52.265 --> 01:03:54.740 align:middle line:84% And of course, the common idea is 01:03:54.740 --> 01:04:01.760 align:middle line:84% to activate it via people who are influential in the network. 01:04:01.760 --> 01:04:05.540 align:middle line:84% And in the COVID 19 people, it was 01:04:05.540 --> 01:04:08.270 align:middle line:84% done both by trying to activate stars 01:04:08.270 --> 01:04:13.130 align:middle line:84% and by trying to activate more locally influential people, 01:04:13.130 --> 01:04:16.550 align:middle line:84% like church leader, community leader, et cetera. 01:04:16.550 --> 01:04:25.130 align:middle line:84% So starting with stars, Vivi Alatas, [INAUDIBLE],, 01:04:25.130 --> 01:04:26.900 align:middle line:84% and a bunch of other people on this paper, 01:04:26.900 --> 01:04:32.390 align:middle line:84% other paper on the role of stars in promoting vaccination 01:04:32.390 --> 01:04:33.680 align:middle line:90% messages-- 01:04:33.680 --> 01:04:36.260 align:middle line:84% again, childhood vaccination messages. 01:04:36.260 --> 01:04:39.020 align:middle line:84% And what they do is that they use local stars, 01:04:39.020 --> 01:04:44.420 align:middle line:84% so that they have geographical differences and randomize 01:04:44.420 --> 01:04:47.390 align:middle line:90% at that level. 01:04:47.390 --> 01:04:49.130 align:middle line:84% There are many stars, which are on there, 01:04:49.130 --> 01:04:53.990 align:middle line:84% different kind of groups of people on Twitter, 01:04:53.990 --> 01:04:58.220 align:middle line:84% either write in their own voice or just 01:04:58.220 --> 01:05:01.070 align:middle line:90% retweet the messages by WHO. 01:05:01.070 --> 01:05:03.260 align:middle line:84% And what they find is that the most effective 01:05:03.260 --> 01:05:07.100 align:middle line:84% is when a star writes in their own voice. 01:05:07.100 --> 01:05:10.550 align:middle line:84% "I am so glad that my kid got immunized." 01:05:10.550 --> 01:05:12.410 align:middle line:84% Or "I think immunization is important, 01:05:12.410 --> 01:05:13.520 align:middle line:90% or something like that. 01:05:13.520 --> 01:05:16.550 align:middle line:84% And those messages are much more likely to be liked 01:05:16.550 --> 01:05:18.950 align:middle line:84% and to be retweeted and so on and so forth. 01:05:18.950 --> 01:05:20.600 align:middle line:84% We don't know-- they don't have-- 01:05:20.600 --> 01:05:23.510 align:middle line:84% They can't test immunization status, unfortunately. 01:05:23.510 --> 01:05:26.720 align:middle line:84% So it's all about that communication by star 01:05:26.720 --> 01:05:30.500 align:middle line:90% on vaccines get powerful. 01:05:30.500 --> 01:05:38.400 align:middle line:84% Abhijit, and the same, Ben, Rema, Arun, 01:05:38.400 --> 01:05:43.380 align:middle line:84% have a paper that Larry Katz called the Banerjee effect, 01:05:43.380 --> 01:05:50.060 align:middle line:84% which we did in May of 2020 at the very beginning of the COVID 01:05:50.060 --> 01:05:53.510 align:middle line:90% 19 pandemic. 01:05:53.510 --> 01:05:59.030 align:middle line:84% We had Abhijit, who at the time was 01:05:59.030 --> 01:06:01.490 align:middle line:84% a star for tourism in West Bengal, first, 01:06:01.490 --> 01:06:02.900 align:middle line:90% because of the Nobel Prize. 01:06:02.900 --> 01:06:05.090 align:middle line:84% In India, they take their Nobel prizes seriously. 01:06:05.090 --> 01:06:10.820 align:middle line:84% So he's like literally on billboards and magazines 01:06:10.820 --> 01:06:11.480 align:middle line:90% and all that. 01:06:11.480 --> 01:06:15.950 align:middle line:90% It's something crazy, actually. 01:06:15.950 --> 01:06:18.080 align:middle line:84% When he goes to India, when he arrives 01:06:18.080 --> 01:06:21.560 align:middle line:84% at the airport in Kolkata, they are like people thronging. 01:06:21.560 --> 01:06:26.850 align:middle line:84% And he has to be put in a police car and all that. 01:06:26.850 --> 01:06:29.480 align:middle line:84% If I go to India with him, I say, let him go, 01:06:29.480 --> 01:06:31.940 align:middle line:84% and, with the kids, we go some other way. 01:06:31.940 --> 01:06:35.060 align:middle line:90% Say, we don't know that guy. 01:06:35.060 --> 01:06:38.180 align:middle line:84% Anyway, so we try and put it to good use. 01:06:38.180 --> 01:06:43.730 align:middle line:84% And another reason why he was potentially famous, relevantly 01:06:43.730 --> 01:06:45.080 align:middle line:90% famous in this context. 01:06:45.080 --> 01:06:46.670 align:middle line:90% Is because he was-- 01:06:46.670 --> 01:06:50.240 align:middle line:84% still is, actually-- the FAO chief for West Bengal, 01:06:50.240 --> 01:06:51.530 align:middle line:90% without the medical expertise. 01:06:51.530 --> 01:06:58.280 align:middle line:84% He is the chairman of the COVID 19 prevention commission. 01:06:58.280 --> 01:07:00.440 align:middle line:90% So he recorded short videos. 01:07:00.440 --> 01:07:08.750 align:middle line:84% And they were sent to people via SMS. 01:07:08.750 --> 01:07:11.120 align:middle line:84% It was randomized at the geographical level 01:07:11.120 --> 01:07:16.130 align:middle line:84% to about 20 million people in West Bengal-- 01:07:16.130 --> 01:07:18.380 align:middle line:84% randomized at the zip code, equivalent of the zip code 01:07:18.380 --> 01:07:19.850 align:middle line:90% level. 01:07:19.850 --> 01:07:22.490 align:middle line:90% And the videos? 01:07:22.490 --> 01:07:26.180 align:middle line:84% So we tried different variation in the message. 01:07:26.180 --> 01:07:29.390 align:middle line:90% And we also-- 01:07:29.390 --> 01:07:31.220 align:middle line:90% And the main outcome was-- so 01:07:31.220 --> 01:07:33.260 align:middle line:84% But all of the message encouraged people 01:07:33.260 --> 01:07:37.250 align:middle line:84% to practice social distancing and to report symptoms 01:07:37.250 --> 01:07:40.700 align:middle line:84% to the ASHA, to the local health worker. 01:07:40.700 --> 01:07:45.650 align:middle line:84% And then we looked at whether the ASHA reported that anyone 01:07:45.650 --> 01:07:48.290 align:middle line:84% reported symptoms to her, in the first few days 01:07:48.290 --> 01:07:55.280 align:middle line:84% after the messages were sent, and, therefore, 01:07:55.280 --> 01:07:57.650 align:middle line:84% we could be sure it's not that we increased COVID 19 01:07:57.650 --> 01:08:00.180 align:middle line:84% prevalence with this message and, therefore, more symptoms 01:08:00.180 --> 01:08:02.000 align:middle line:90% were there to be reported. 01:08:02.000 --> 01:08:04.070 align:middle line:84% Anyway, what we found is a doubling of rate 01:08:04.070 --> 01:08:06.620 align:middle line:84% at which symptoms were reported to the ASHA, 01:08:06.620 --> 01:08:12.470 align:middle line:84% as well as more self-reported social distancing 01:08:12.470 --> 01:08:14.900 align:middle line:90% of various kinds. 01:08:14.900 --> 01:08:17.510 align:middle line:84% Which doesn't seem to come from information, because people 01:08:17.510 --> 01:08:19.910 align:middle line:84% receive about 20 messages a day, anyways, 01:08:19.910 --> 01:08:22.069 align:middle line:84% and seem to be pretty well informed, but more 01:08:22.069 --> 01:08:26.779 align:middle line:84% about an extra nudge of doing the work. 01:08:26.779 --> 01:08:31.430 align:middle line:84% So it does seem that these stars, public figures have, 01:08:31.430 --> 01:08:34.460 align:middle line:90% potentially, an impact. 01:08:34.460 --> 01:08:39.109 align:middle line:84% The idea of the Biden immunization corps 01:08:39.109 --> 01:08:43.819 align:middle line:84% was to use locally influential people. 01:08:43.819 --> 01:08:45.319 align:middle line:84% And this is something which you find 01:08:45.319 --> 01:08:48.560 align:middle line:84% a lot of in developing countries, where you try to-- 01:08:48.560 --> 01:08:51.270 align:middle line:84% Whatever, when you have something to sell-- 01:08:51.270 --> 01:08:53.600 align:middle line:84% a public health campaign or a new product-- 01:08:53.600 --> 01:08:55.850 align:middle line:84% you try and identify, in the community, 01:08:55.850 --> 01:09:00.180 align:middle line:90% people that are influential. 01:09:00.180 --> 01:09:03.229 align:middle line:84% So the problem is, who is influential? 01:09:03.229 --> 01:09:05.180 align:middle line:84% And the way that-- the traditional way to get 01:09:05.180 --> 01:09:08.970 align:middle line:90% at that is you know in advance. 01:09:08.970 --> 01:09:12.895 align:middle line:84% So for example, in the case of the micro-credit organization 01:09:12.895 --> 01:09:15.020 align:middle line:84% that you write for the paper, you write on gossips. 01:09:15.020 --> 01:09:19.550 align:middle line:84% You could think it is the people who are the shopkeeper, 01:09:19.550 --> 01:09:22.340 align:middle line:84% the teachers, the self-help group members, et cetera, 01:09:22.340 --> 01:09:23.510 align:middle line:90% the religious leaders-- 01:09:23.510 --> 01:09:26.120 align:middle line:84% are the ones that were used by the Biden campaign. 01:09:26.120 --> 01:09:35.000 align:middle line:84% But what we found in that paper, for example, 01:09:35.000 --> 01:09:37.970 align:middle line:84% on diffusion of micro-finance, what we found 01:09:37.970 --> 01:09:40.010 align:middle line:84% is that it doesn't necessarily work. 01:09:40.010 --> 01:09:42.770 align:middle line:84% A lot of these people, who are supposed to be influential, 01:09:42.770 --> 01:09:44.340 align:middle line:90% are really not that influential. 01:09:44.340 --> 01:09:46.520 align:middle line:84% And if you hit one of these person 01:09:46.520 --> 01:09:49.010 align:middle line:84% that you think is influential and you spread information 01:09:49.010 --> 01:09:52.200 align:middle line:84% to them, then information goes nowhere. 01:09:52.200 --> 01:09:55.370 align:middle line:84% So you would like to find people who are influential. 01:09:55.370 --> 01:09:57.990 align:middle line:90% And it's hard to do. 01:09:57.990 --> 01:10:03.050 align:middle line:84% And so that's where the gossips idea came up just 01:10:03.050 --> 01:10:05.360 align:middle line:84% to find a way to find these influential people. 01:10:05.360 --> 01:10:10.543 align:middle line:90% So what's the idea of gossips? 01:10:10.543 --> 01:10:12.460 align:middle line:84% That's the paper, or the first paper you read. 01:10:12.460 --> 01:10:17.420 align:middle line:90% 01:10:17.420 --> 01:10:17.972 align:middle line:90% Yeah? 01:10:17.972 --> 01:10:19.430 align:middle line:84% AUDIENCE: Identify people who would 01:10:19.430 --> 01:10:21.170 align:middle line:84% be good at spreading information by just 01:10:21.170 --> 01:10:23.553 align:middle line:84% asking people who would be good at spreading information? 01:10:23.553 --> 01:10:24.470 align:middle line:90% ESTHER DUFLO: Exactly. 01:10:24.470 --> 01:10:26.600 align:middle line:90% So it's asking in the community. 01:10:26.600 --> 01:10:29.270 align:middle line:84% Finding a few people and asking them, well, who do you-- 01:10:29.270 --> 01:10:32.300 align:middle line:84% if I have a piece of information, 01:10:32.300 --> 01:10:35.780 align:middle line:84% who do you think I should give it to? 01:10:35.780 --> 01:10:37.640 align:middle line:90% And the reason why? 01:10:37.640 --> 01:10:39.980 align:middle line:84% So that seems pretty straightforward-- 01:10:39.980 --> 01:10:43.130 align:middle line:90% just ask people. 01:10:43.130 --> 01:10:45.740 align:middle line:84% We know that community members have information. 01:10:45.740 --> 01:10:50.270 align:middle line:84% For example, in a paper by Ben and Vivi and Rema and Abhijit, 01:10:50.270 --> 01:10:51.312 align:middle line:90% we know that they are. 01:10:51.312 --> 01:10:52.770 align:middle line:84% They know who is poor, for example. 01:10:52.770 --> 01:10:55.940 align:middle line:84% So they know some stuff about their village. 01:10:55.940 --> 01:10:59.690 align:middle line:84% We know that they are good at identifying 01:10:59.690 --> 01:11:01.290 align:middle line:90% people who are productive. 01:11:01.290 --> 01:11:04.760 align:middle line:84% So that's a paper by Ben Roth, Natalia Rigol, and Reshmaan 01:11:04.760 --> 01:11:08.195 align:middle line:84% Hussam, who is basically suggesting that people knew, 01:11:08.195 --> 01:11:11.270 align:middle line:84% in their community, who would make good use of 1,000 rupees 01:11:11.270 --> 01:11:14.930 align:middle line:84% or 10,000 rupees of a loan-- of a grant. 01:11:14.930 --> 01:11:17.000 align:middle line:84% And they know better than we could 01:11:17.000 --> 01:11:20.330 align:middle line:84% by trying to run a regression, do machine learning, 01:11:20.330 --> 01:11:22.040 align:middle line:90% or whatever. 01:11:22.040 --> 01:11:24.480 align:middle line:84% So that would be that first order. 01:11:24.480 --> 01:11:27.380 align:middle line:90% That's not a crazy thing to say. 01:11:27.380 --> 01:11:32.240 align:middle line:84% But the thing with being influential in a network, that 01:11:32.240 --> 01:11:33.080 align:middle line:90% might be harder. 01:11:33.080 --> 01:11:37.250 align:middle line:84% Because getting a full view of your network, 01:11:37.250 --> 01:11:38.240 align:middle line:90% people don't have. 01:11:38.240 --> 01:11:40.070 align:middle line:84% For example, I know who my friends are. 01:11:40.070 --> 01:11:42.380 align:middle line:84% I might know who the friends of my friends are. 01:11:42.380 --> 01:11:45.890 align:middle line:84% But I don't know who the friend of my friend of my friends are. 01:11:45.890 --> 01:11:47.870 align:middle line:84% And you would think, in order to know 01:11:47.870 --> 01:11:50.720 align:middle line:84% who is influential in a network, if that's 01:11:50.720 --> 01:11:54.410 align:middle line:84% related to your central location in a network, 01:11:54.410 --> 01:12:00.000 align:middle line:90% you would be quite lost. 01:12:00.000 --> 01:12:02.450 align:middle line:84% Especially as many of you have mentioned, 01:12:02.450 --> 01:12:04.550 align:middle line:90% networks get divided. 01:12:04.550 --> 01:12:06.990 align:middle line:84% There is some segregation in the network. 01:12:06.990 --> 01:12:10.610 align:middle line:84% So you might only have the idea of about half the village 01:12:10.610 --> 01:12:11.210 align:middle line:90% at best. 01:12:11.210 --> 01:12:14.280 align:middle line:90% 01:12:14.280 --> 01:12:17.630 align:middle line:90% So why would they know? 01:12:17.630 --> 01:12:21.360 align:middle line:84% So one way to think about this is to say, well, 01:12:21.360 --> 01:12:24.570 align:middle line:84% what does it mean to be influential in the network? 01:12:24.570 --> 01:12:28.470 align:middle line:84% So in a diffusion sense, someone wanted a graph-- 01:12:28.470 --> 01:12:30.440 align:middle line:84% someone, one of you, who commented, 01:12:30.440 --> 01:12:34.050 align:middle line:84% wanted to see a graph of centrality. 01:12:34.050 --> 01:12:35.180 align:middle line:90% So here is one. 01:12:35.180 --> 01:12:39.440 align:middle line:84% So here is what it means to be influential in a diffusion 01:12:39.440 --> 01:12:40.070 align:middle line:90% sense. 01:12:40.070 --> 01:12:43.760 align:middle line:84% If you have a model where people simply pass information, 01:12:43.760 --> 01:12:46.430 align:middle line:84% with some probability, to the people in their edge, 01:12:46.430 --> 01:12:48.940 align:middle line:84% and the process is repeated for a number of period. 01:12:48.940 --> 01:12:51.080 align:middle line:84% So this is, if you have a piece of information, 01:12:51.080 --> 01:12:54.740 align:middle line:84% you'll pass it half the time, let's say, 01:12:54.740 --> 01:12:56.460 align:middle line:90% and for four periods. 01:12:56.460 --> 01:12:59.090 align:middle line:84% So first I give information to them. 01:12:59.090 --> 01:13:01.460 align:middle line:84% Then they give it to one of their two friends. 01:13:01.460 --> 01:13:03.140 align:middle line:90% Then the process is done again. 01:13:03.140 --> 01:13:05.030 align:middle line:90% He gives it to another friend. 01:13:05.030 --> 01:13:07.700 align:middle line:84% And this one gives it to two of their friends. 01:13:07.700 --> 01:13:10.010 align:middle line:84% And then they, again-- that continues again. 01:13:10.010 --> 01:13:11.630 align:middle line:90% And then this is over. 01:13:11.630 --> 01:13:17.150 align:middle line:84% And I count, in expectation, how many people, in that, 01:13:17.150 --> 01:13:19.170 align:middle line:90% have learned it from them. 01:13:19.170 --> 01:13:21.540 align:middle line:90% So that's diffusion centrality. 01:13:21.540 --> 01:13:23.210 align:middle line:84% Here is another example of someone 01:13:23.210 --> 01:13:25.700 align:middle line:84% who is a little bit more remote here. 01:13:25.700 --> 01:13:29.390 align:middle line:84% So after four periods, in expectation, only 6% 01:13:29.390 --> 01:13:32.580 align:middle line:90% have heard the news. 01:13:32.580 --> 01:13:36.090 align:middle line:84% So diffusion centrality is the total expected number 01:13:36.090 --> 01:13:38.700 align:middle line:84% of times information, starting at point I, 01:13:38.700 --> 01:13:40.270 align:middle line:90% has reached everybody else. 01:13:40.270 --> 01:13:41.730 align:middle line:84% So it's the probability of passing 01:13:41.730 --> 01:13:46.380 align:middle line:84% the information times each of the links 01:13:46.380 --> 01:13:51.200 align:middle line:84% to the power of the number of periods and then summed up. 01:13:51.200 --> 01:13:53.905 align:middle line:84% It's related to other measure of centrality 01:13:53.905 --> 01:13:54.780 align:middle line:90% you might have heard. 01:13:54.780 --> 01:13:56.450 align:middle line:90% For example, people who have-- 01:13:56.450 --> 01:13:58.280 align:middle line:84% if there was only one period, it would 01:13:58.280 --> 01:14:01.220 align:middle line:84% be proportional to degree-- your number of friends. 01:14:01.220 --> 01:14:03.320 align:middle line:84% If there was an infinite number of period, 01:14:03.320 --> 01:14:06.110 align:middle line:84% it would become eigenvector centrality. 01:14:06.110 --> 01:14:11.210 align:middle line:84% And for any number of periods, if the process die out 01:14:11.210 --> 01:14:13.250 align:middle line:84% after a few period of transmission, 01:14:13.250 --> 01:14:17.870 align:middle line:84% people get tired of conveying this particular piece of news, 01:14:17.870 --> 01:14:23.170 align:middle line:84% then it's going to be somewhere in between the two. 01:14:23.170 --> 01:14:25.740 align:middle line:84% So the question now becomes more precise. 01:14:25.740 --> 01:14:28.440 align:middle line:84% It's whether we think that, despite having 01:14:28.440 --> 01:14:34.230 align:middle line:84% a fairly poor representation of the network architecture, 01:14:34.230 --> 01:14:37.935 align:middle line:84% people might be able to identify people who have high DC. 01:14:37.935 --> 01:14:41.760 align:middle line:90% 01:14:41.760 --> 01:14:45.420 align:middle line:84% First of all, in theory, we could 01:14:45.420 --> 01:14:47.460 align:middle line:84% think that hitting people with high diffusion 01:14:47.460 --> 01:14:51.460 align:middle line:84% centrality would be effective to diffuse information. 01:14:51.460 --> 01:14:54.590 align:middle line:84% As it turns out, in practice that's true as well. 01:14:54.590 --> 01:15:00.070 align:middle line:84% So for example, Mushfiq Mobarak, Lori Beaman, and Ariell Reshef 01:15:00.070 --> 01:15:01.630 align:middle line:84% have a paper, where they are looking 01:15:01.630 --> 01:15:05.410 align:middle line:84% at introducing a new agricultural technology 01:15:05.410 --> 01:15:06.400 align:middle line:90% in a village. 01:15:06.400 --> 01:15:10.540 align:middle line:84% And they show that if they hit people 01:15:10.540 --> 01:15:16.375 align:middle line:84% who are eigenvector central, the information circulates faster. 01:15:16.375 --> 01:15:19.600 align:middle line:90% 01:15:19.600 --> 01:15:22.540 align:middle line:84% In our paper, our first diffusion paper, 01:15:22.540 --> 01:15:24.580 align:middle line:84% where we looked at these kind of seeds 01:15:24.580 --> 01:15:29.320 align:middle line:84% that were supposedly influential people, if they in fact 01:15:29.320 --> 01:15:32.650 align:middle line:84% are diffusion central, the information circulates 01:15:32.650 --> 01:15:34.540 align:middle line:84% and people take up micro-finance. 01:15:34.540 --> 01:15:38.120 align:middle line:84% But if they are not, then the information doesn't. 01:15:38.120 --> 01:15:41.330 align:middle line:84% So there seems to be a correlation there, as well. 01:15:41.330 --> 01:15:43.052 align:middle line:84% So that seems that if you could hit 01:15:43.052 --> 01:15:45.010 align:middle line:84% the diffusion central people, it would be good. 01:15:45.010 --> 01:15:52.930 align:middle line:84% But do people have a reason to know that? 01:15:52.930 --> 01:15:59.260 align:middle line:84% And the way that they could find out 01:15:59.260 --> 01:16:06.850 align:middle line:84% is because, although they don't know the centrality of someone, 01:16:06.850 --> 01:16:08.800 align:middle line:84% they already know how many times or they 01:16:08.800 --> 01:16:10.300 align:middle line:84% might have an idea of how many times 01:16:10.300 --> 01:16:13.300 align:middle line:84% they have heard about that person. 01:16:13.300 --> 01:16:16.210 align:middle line:84% So now it's not just hearing a piece of news 01:16:16.210 --> 01:16:20.260 align:middle line:84% coming from that person but hearing about that person. 01:16:20.260 --> 01:16:24.520 align:middle line:84% So suppose that there are some news, like I got a new goat 01:16:24.520 --> 01:16:33.460 align:middle line:84% or might change jobs, and the probability that this gossip 01:16:33.460 --> 01:16:38.750 align:middle line:84% circulates is passed from one node to another. 01:16:38.750 --> 01:16:40.360 align:middle line:84% And the only thing people can keep 01:16:40.360 --> 01:16:44.500 align:middle line:84% track of is how many times they've heard that news. 01:16:44.500 --> 01:16:46.900 align:middle line:84% And if the news is of interest, then it 01:16:46.900 --> 01:16:49.240 align:middle line:84% will be circulating by many edge. 01:16:49.240 --> 01:16:52.630 align:middle line:84% And people are more likely to hear it 01:16:52.630 --> 01:16:56.890 align:middle line:84% if the original person, about whom the news is, 01:16:56.890 --> 01:16:59.080 align:middle line:84% has more friends, who have found out about the news, 01:16:59.080 --> 01:17:01.030 align:middle line:84% and themselves have more friends to whom 01:17:01.030 --> 01:17:03.650 align:middle line:90% they can spread this news. 01:17:03.650 --> 01:17:09.220 align:middle line:84% So the formula is different than the formula 01:17:09.220 --> 01:17:10.570 align:middle line:90% for diffusion centrality. 01:17:10.570 --> 01:17:12.700 align:middle line:90% But it's related. 01:17:12.700 --> 01:17:15.850 align:middle line:84% It's now the number of times I've 01:17:15.850 --> 01:17:18.880 align:middle line:90% heard about a piece of news. 01:17:18.880 --> 01:17:27.910 align:middle line:84% And so the diffusion centrality tracks how the info 01:17:27.910 --> 01:17:29.650 align:middle line:90% spread from a given node. 01:17:29.650 --> 01:17:32.440 align:middle line:84% And the network gossip tracks how often 01:17:32.440 --> 01:17:37.850 align:middle line:84% I hear about an info that originated from other nodes. 01:17:37.850 --> 01:17:43.510 align:middle line:84% So if you compare this formula to the other ones, 01:17:43.510 --> 01:17:46.420 align:middle line:84% it's not the same but it's related. 01:17:46.420 --> 01:17:48.580 align:middle line:84% And the intuition for why it's related 01:17:48.580 --> 01:17:52.390 align:middle line:84% is because, if people have a lot of friends, who themselves have 01:17:52.390 --> 01:17:57.160 align:middle line:84% a lot of friends, on the one hand, anything they say 01:17:57.160 --> 01:18:01.090 align:middle line:84% will circulate a lot, and, also, any news about them 01:18:01.090 --> 01:18:02.440 align:middle line:90% will circulate a lot. 01:18:02.440 --> 01:18:07.130 align:middle line:90% 01:18:07.130 --> 01:18:09.710 align:middle line:84% So if you have enough communication period, 01:18:09.710 --> 01:18:15.410 align:middle line:84% the two will actually converge to each other 01:18:15.410 --> 01:18:19.550 align:middle line:90% and remain very correlated. 01:18:19.550 --> 01:18:21.470 align:middle line:84% And in fact, with infinite number of periods, 01:18:21.470 --> 01:18:23.240 align:middle line:90% they'll just be the same. 01:18:23.240 --> 01:18:27.540 align:middle line:84% They'll eventually converge to be exactly the same. 01:18:27.540 --> 01:18:33.410 align:middle line:84% And the intuition is, one, that people who are popular 01:18:33.410 --> 01:18:36.935 align:middle line:84% can convey a piece of news and also pieces of news about them 01:18:36.935 --> 01:18:37.435 align:middle line:90% convey. 01:18:37.435 --> 01:18:40.340 align:middle line:90% 01:18:40.340 --> 01:18:44.030 align:middle line:84% I have heard many times about Arun 01:18:44.030 --> 01:18:47.630 align:middle line:84% also means that, when Arun has a piece of news, 01:18:47.630 --> 01:18:50.795 align:middle line:90% many people have heard it. 01:18:50.795 --> 01:18:52.545 align:middle line:84% You can see that it's not the same object. 01:18:52.545 --> 01:18:54.960 align:middle line:84% It's I have heard about him versus 01:18:54.960 --> 01:18:57.330 align:middle line:90% he has reached a lot of people. 01:18:57.330 --> 01:18:59.010 align:middle line:84% But the two are correlated, because they 01:18:59.010 --> 01:19:01.470 align:middle line:84% are related to the information passing from one edge 01:19:01.470 --> 01:19:03.690 align:middle line:90% to the other. 01:19:03.690 --> 01:19:05.030 align:middle line:90% So that's the idea of gossips. 01:19:05.030 --> 01:19:08.030 align:middle line:84% And so what this suggests, if that theory is correct, 01:19:08.030 --> 01:19:09.530 align:middle line:84% is that, first of all, people should 01:19:09.530 --> 01:19:13.220 align:middle line:84% have relative agreement about who the central people are. 01:19:13.220 --> 01:19:16.372 align:middle line:90% 01:19:16.372 --> 01:19:18.080 align:middle line:84% There should be quite a bit of consensus, 01:19:18.080 --> 01:19:20.520 align:middle line:84% because they are the same within the network. 01:19:20.520 --> 01:19:23.660 align:middle line:84% So you can introspect whether, in your class, 01:19:23.660 --> 01:19:26.078 align:middle line:90% it's the same person. 01:19:26.078 --> 01:19:27.620 align:middle line:84% If you have a piece of news you don't 01:19:27.620 --> 01:19:29.360 align:middle line:84% want to spread, who you should go see 01:19:29.360 --> 01:19:31.610 align:middle line:84% and who you should not tell it and vice versa. 01:19:31.610 --> 01:19:41.760 align:middle line:90% 01:19:41.760 --> 01:19:44.610 align:middle line:84% And that if an information gets transmitted 01:19:44.610 --> 01:19:50.310 align:middle line:84% to one of these people, it should, in fact, spread. 01:19:50.310 --> 01:19:52.170 align:middle line:84% Then you could even have an advantage 01:19:52.170 --> 01:19:58.110 align:middle line:84% over the properly calculated value of eigenvector 01:19:58.110 --> 01:19:59.610 align:middle line:84% centrality, of diffusion centrality, 01:19:59.610 --> 01:20:06.840 align:middle line:84% because, in addition, you ask the question, who 01:20:06.840 --> 01:20:10.260 align:middle line:84% is good at transmitting information. 01:20:10.260 --> 01:20:12.522 align:middle line:84% If you can calculate how many times you've 01:20:12.522 --> 01:20:13.980 align:middle line:84% heard about them-- in addition, you 01:20:13.980 --> 01:20:15.813 align:middle line:84% might know that these are gregarious people. 01:20:15.813 --> 01:20:18.130 align:middle line:90% Who like to talk a lot. 01:20:18.130 --> 01:20:19.650 align:middle line:84% And so that is something that's not 01:20:19.650 --> 01:20:21.990 align:middle line:84% captured by this very simple model, where everyone 01:20:21.990 --> 01:20:24.660 align:middle line:84% has the same Q, but could be incorporated 01:20:24.660 --> 01:20:26.220 align:middle line:90% by someone about them. 01:20:26.220 --> 01:20:28.920 align:middle line:84% So you could do even better if you can do this simply. 01:20:28.920 --> 01:20:31.350 align:middle line:84% If the simple exercise of counting the piece of news 01:20:31.350 --> 01:20:34.800 align:middle line:84% is, in fact, going to give you a good sense of people, places, 01:20:34.800 --> 01:20:36.390 align:middle line:84% and network, you can do even better 01:20:36.390 --> 01:20:46.150 align:middle line:84% by adding other things you end up knowing about them. 01:20:46.150 --> 01:20:47.140 align:middle line:90% So that's the idea. 01:20:47.140 --> 01:20:50.730 align:middle line:84% So on Wednesday, we are going to see how that fares. 01:20:50.730 --> 01:20:54.300 align:middle line:84% And then how that compares with other things 01:20:54.300 --> 01:20:56.550 align:middle line:84% that one could also do in one context. 01:20:56.550 --> 01:21:02.700 align:middle line:84% And think also a bit more methodologically about one even 01:21:02.700 --> 01:21:03.930 align:middle line:90% answer these questions. 01:21:03.930 --> 01:21:07.830 align:middle line:84% And then we'll shift to, given this model of demands, 01:21:07.830 --> 01:21:14.190 align:middle line:84% thinking about the mess that health care supply becomes. 01:21:14.190 --> 01:21:16.130 align:middle line:90% Thank you so much. 01:21:16.130 --> 01:21:31.000 align:middle line:90%