WEBVTT 00:00:00.000 --> 00:00:06.818 align:middle line:84% [SQUEAKING] [RUSTLING] [CLICKING] 00:00:06.818 --> 00:00:18.225 align:middle line:90% 00:00:18.225 --> 00:00:19.350 align:middle line:90% FRANK SCHILBACH: All right. 00:00:19.350 --> 00:00:22.560 align:middle line:84% Welcome, everyone, to my second lecture 00:00:22.560 --> 00:00:25.560 align:middle line:84% on behavioral development economics. 00:00:25.560 --> 00:00:28.470 align:middle line:84% Let me just recap briefly what we talked about last time 00:00:28.470 --> 00:00:31.240 align:middle line:84% and how this time will be different. 00:00:31.240 --> 00:00:34.155 align:middle line:84% So last time, we talked about high return investment 00:00:34.155 --> 00:00:37.080 align:middle line:84% opportunities and wrote down an oil equation 00:00:37.080 --> 00:00:41.370 align:middle line:84% to try to understand potential reasons why people might not 00:00:41.370 --> 00:00:43.090 align:middle line:90% take advantage of them. 00:00:43.090 --> 00:00:45.423 align:middle line:84% And so the idea is, essentially, there's 00:00:45.423 --> 00:00:47.340 align:middle line:84% high rates of returns of many different things 00:00:47.340 --> 00:00:48.780 align:middle line:90% that people could invest in. 00:00:48.780 --> 00:00:52.140 align:middle line:84% And what a part of a large share of behavioral development 00:00:52.140 --> 00:00:55.620 align:middle line:84% economics has tried to do in recent years is 00:00:55.620 --> 00:00:58.245 align:middle line:84% try to understand using universal psychological 00:00:58.245 --> 00:01:02.190 align:middle line:84% features, such as present bias, or loss aversion, 00:01:02.190 --> 00:01:05.040 align:middle line:84% or perhaps biased beliefs in some ways, 00:01:05.040 --> 00:01:07.710 align:middle line:84% to try and understand why people are not taking advantage 00:01:07.710 --> 00:01:09.990 align:middle line:90% of such opportunities, OK? 00:01:09.990 --> 00:01:12.870 align:middle line:84% So here's a fertilizer that you could invest in. 00:01:12.870 --> 00:01:15.030 align:middle line:90% Here's machines you could buy. 00:01:15.030 --> 00:01:18.300 align:middle line:84% Here's investments in health, et cetera, that people are not 00:01:18.300 --> 00:01:19.460 align:middle line:90% taking advantage of. 00:01:19.460 --> 00:01:22.080 align:middle line:84% And we try to understand, can behavioral factors, 00:01:22.080 --> 00:01:26.250 align:middle line:84% psychological factors, or other types of factors 00:01:26.250 --> 00:01:28.950 align:middle line:84% help us understand why people might not be doing that? 00:01:28.950 --> 00:01:33.930 align:middle line:84% Those are universal factors that the rich and the poor 00:01:33.930 --> 00:01:35.945 align:middle line:90% might be similarly affected. 00:01:35.945 --> 00:01:37.320 align:middle line:84% There are some questions on is it 00:01:37.320 --> 00:01:39.450 align:middle line:84% different for the rich and the poor in the sense 00:01:39.450 --> 00:01:41.880 align:middle line:84% that in rich settings, there might 00:01:41.880 --> 00:01:45.000 align:middle line:84% be a better infrastructure that nudges you into the right 00:01:45.000 --> 00:01:45.750 align:middle line:90% decisions. 00:01:45.750 --> 00:01:48.750 align:middle line:84% Anyway, but these are not factors that are specifically 00:01:48.750 --> 00:01:55.733 align:middle line:84% to a developing country context or low-income populations. 00:01:55.733 --> 00:01:57.150 align:middle line:84% Separately from this-- and this is 00:01:57.150 --> 00:02:00.390 align:middle line:84% what we're going to talk about today-- is the question of, 00:02:00.390 --> 00:02:02.790 align:middle line:84% is there something specific about the psychology 00:02:02.790 --> 00:02:03.390 align:middle line:90% of poverty? 00:02:03.390 --> 00:02:06.600 align:middle line:84% Or does poverty itself affect people's behavior 00:02:06.600 --> 00:02:07.740 align:middle line:90% in certain ways? 00:02:07.740 --> 00:02:11.520 align:middle line:84% And could it be that because people are poor, 00:02:11.520 --> 00:02:15.660 align:middle line:84% that has some treatment effect on people's choices 00:02:15.660 --> 00:02:16.900 align:middle line:90% and behaviors? 00:02:16.900 --> 00:02:18.900 align:middle line:84% And that's what we're going to talk about today. 00:02:18.900 --> 00:02:24.120 align:middle line:84% And we're going to talk about three sets of papers-- 00:02:24.120 --> 00:02:28.750 align:middle line:84% excuse me-- three sets of papers. 00:02:28.750 --> 00:02:31.240 align:middle line:84% First, we talk about scarcity and financial strain. 00:02:31.240 --> 00:02:32.652 align:middle line:90% Second, we talk about sleep. 00:02:32.652 --> 00:02:34.360 align:middle line:84% And this is the paper that you have read. 00:02:34.360 --> 00:02:37.420 align:middle line:84% And third, we're going to talk a little bit about mental health, 00:02:37.420 --> 00:02:41.110 align:middle line:84% more like a broad overview of that area. 00:02:41.110 --> 00:02:44.410 align:middle line:84% Any questions about last time or anything else before I start? 00:02:44.410 --> 00:02:48.402 align:middle line:90% 00:02:48.402 --> 00:02:50.400 align:middle line:90% OK. 00:02:50.400 --> 00:02:50.970 align:middle line:90% All right. 00:02:50.970 --> 00:02:53.370 align:middle line:84% So similarly to last time, I wanted 00:02:53.370 --> 00:02:57.400 align:middle line:84% to start with the motivation for this type of topic, 00:02:57.400 --> 00:03:01.560 align:middle line:84% which is the observation that there's quite a few-- 00:03:01.560 --> 00:03:03.435 align:middle line:84% what I'm going to call-- seemingly suboptimal 00:03:03.435 --> 00:03:06.487 align:middle line:84% behaviors among the poor, among low-income populations. 00:03:06.487 --> 00:03:08.070 align:middle line:84% We talked about some of these already, 00:03:08.070 --> 00:03:09.940 align:middle line:84% which are investment savings decision. 00:03:09.940 --> 00:03:13.590 align:middle line:84% These are the hybrids of return on the investment opportunities 00:03:13.590 --> 00:03:15.060 align:middle line:84% that people could take advantage of 00:03:15.060 --> 00:03:17.040 align:middle line:84% and are not necessarily doing that. 00:03:17.040 --> 00:03:20.280 align:middle line:84% Similarly, a credit market behavior as people 00:03:20.280 --> 00:03:23.670 align:middle line:84% take high interest rate loans and then have to pay them back 00:03:23.670 --> 00:03:27.030 align:middle line:84% and often cycle over and over from loan to loan. 00:03:27.030 --> 00:03:29.670 align:middle line:84% But there's also other domains where 00:03:29.670 --> 00:03:35.280 align:middle line:84% you see, for example, lower productivity, punctuality, 00:03:35.280 --> 00:03:40.680 align:middle line:84% medical adherence, food and drug consumption, healthy eating, 00:03:40.680 --> 00:03:44.160 align:middle line:84% parenting behavior, et cetera, where there's 00:03:44.160 --> 00:03:45.960 align:middle line:84% lots of correlational evidence that 00:03:45.960 --> 00:03:48.900 align:middle line:84% shows that low-income populations tend 00:03:48.900 --> 00:03:51.030 align:middle line:84% to do worse in a lot of these metrics that 00:03:51.030 --> 00:03:53.170 align:middle line:90% are being collected. 00:03:53.170 --> 00:03:59.530 align:middle line:84% Now, if you step away for a bit and forget what I told you 00:03:59.530 --> 00:04:01.780 align:middle line:84% for the last two minutes, what is 00:04:01.780 --> 00:04:04.570 align:middle line:84% a typical, neoclassical explanation 00:04:04.570 --> 00:04:06.070 align:middle line:90% for these types of behaviors? 00:04:06.070 --> 00:04:09.280 align:middle line:84% What might you say if you see these types of patterns? 00:04:09.280 --> 00:04:14.050 align:middle line:84% What might you say could explain these types of phenomena? 00:04:14.050 --> 00:04:18.049 align:middle line:90% 00:04:18.049 --> 00:04:18.549 align:middle line:90% Yeah. 00:04:18.549 --> 00:04:19.987 align:middle line:90% AUDIENCE: Misinformation. 00:04:19.987 --> 00:04:21.320 align:middle line:90% FRANK SCHILBACH: Misinformation. 00:04:21.320 --> 00:04:23.320 align:middle line:84% And so on what is the underlying, 00:04:23.320 --> 00:04:25.950 align:middle line:90% omitted variable, if you want. 00:04:25.950 --> 00:04:27.810 align:middle line:84% Why is it that low-income populations 00:04:27.810 --> 00:04:29.254 align:middle line:90% are less well-informed? 00:04:29.254 --> 00:04:32.158 align:middle line:90% 00:04:32.158 --> 00:04:35.708 align:middle line:84% AUDIENCE: Perhaps, there's [INAUDIBLE] education. 00:04:35.708 --> 00:04:36.750 align:middle line:90% FRANK SCHILBACH: Exactly. 00:04:36.750 --> 00:04:41.250 align:middle line:84% Education, schooling, maybe also just peer effects in some ways. 00:04:41.250 --> 00:04:43.200 align:middle line:84% But it could essentially be that people just 00:04:43.200 --> 00:04:44.685 align:middle line:90% have less information. 00:04:44.685 --> 00:04:47.382 align:middle line:84% You can explain at least some of these behaviors. 00:04:47.382 --> 00:04:49.590 align:middle line:84% For example, medical adherence-- you might say people 00:04:49.590 --> 00:04:52.860 align:middle line:84% understand less how useful certain types of medications 00:04:52.860 --> 00:04:55.530 align:middle line:84% are, food and drug consumption as well. 00:04:55.530 --> 00:04:58.500 align:middle line:84% Healthy eating, you have to understand calorie content, 00:04:58.500 --> 00:05:01.730 align:middle line:90% et cetera, and so on. 00:05:01.730 --> 00:05:02.375 align:middle line:90% Anything else? 00:05:02.375 --> 00:05:10.202 align:middle line:90% 00:05:10.202 --> 00:05:12.035 align:middle line:84% AUDIENCE: There are institutional structures 00:05:12.035 --> 00:05:17.975 align:middle line:84% that benefit from people [INAUDIBLE] a certain way, 00:05:17.975 --> 00:05:22.373 align:middle line:84% like payment loans through [INAUDIBLE] [? profits. ?] 00:05:22.373 --> 00:05:23.290 align:middle line:90% FRANK SCHILBACH: Yeah. 00:05:23.290 --> 00:05:25.810 align:middle line:84% So I'm going to call these institutional environmental 00:05:25.810 --> 00:05:30.340 align:middle line:84% factors, which are might say, for example, if you have 00:05:30.340 --> 00:05:33.580 align:middle line:84% universal health insurance, or if you have savings schemes, 00:05:33.580 --> 00:05:34.630 align:middle line:90% 401(k) savings. 00:05:34.630 --> 00:05:37.480 align:middle line:84% Your employer signs you up for savings schemes. 00:05:37.480 --> 00:05:40.150 align:middle line:84% Well, for me, it's much easier to save because I just 00:05:40.150 --> 00:05:41.680 align:middle line:90% don't have to do very much. 00:05:41.680 --> 00:05:43.347 align:middle line:84% I have some health insurance, and then-- 00:05:43.347 --> 00:05:48.745 align:middle line:84% so I kow MIT, I guess, contributes to my retirement 00:05:48.745 --> 00:05:49.510 align:middle line:90% savings. 00:05:49.510 --> 00:05:52.060 align:middle line:84% And then, it's relatively easy to set this up. 00:05:52.060 --> 00:05:54.790 align:middle line:84% If on the other hand, you are a rickshaw driver in Chennai, 00:05:54.790 --> 00:06:01.150 align:middle line:84% or other places, or just other daily worker, 00:06:01.150 --> 00:06:03.700 align:middle line:84% you're going to get cash every day you work for every trip 00:06:03.700 --> 00:06:04.287 align:middle line:90% that you take. 00:06:04.287 --> 00:06:06.370 align:middle line:84% And then, you have to figure out how to save that. 00:06:06.370 --> 00:06:08.980 align:middle line:84% You need a lot more discipline to save because there's 00:06:08.980 --> 00:06:12.520 align:middle line:84% lots of temptations on the way and maybe lots of other ways 00:06:12.520 --> 00:06:15.040 align:middle line:84% in which you might want to spend your money potentially. 00:06:15.040 --> 00:06:17.878 align:middle line:90% 00:06:17.878 --> 00:06:19.670 align:middle line:84% There might not even like a bank available. 00:06:19.670 --> 00:06:22.300 align:middle line:84% And again, Esther is going to talk with you about savings. 00:06:22.300 --> 00:06:24.550 align:middle line:84% So essentially, environmental features 00:06:24.550 --> 00:06:26.650 align:middle line:84% might make it just much more difficult for people 00:06:26.650 --> 00:06:28.428 align:middle line:90% to make the right choices. 00:06:28.428 --> 00:06:30.220 align:middle line:84% And there's other environmental conditions, 00:06:30.220 --> 00:06:31.690 align:middle line:90% for example, transportation. 00:06:31.690 --> 00:06:34.660 align:middle line:84% If you live two hours away from the city. 00:06:34.660 --> 00:06:35.560 align:middle line:90% You don't have a car. 00:06:35.560 --> 00:06:37.000 align:middle line:84% And you have public transportation 00:06:37.000 --> 00:06:38.050 align:middle line:90% that's unreliable. 00:06:38.050 --> 00:06:41.560 align:middle line:84% Well, of course, you're going to show up less reliably, 00:06:41.560 --> 00:06:44.210 align:middle line:84% or it's much harder to be punctual, for example. 00:06:44.210 --> 00:06:47.170 align:middle line:84% There might be predatory lending that tries to trick people 00:06:47.170 --> 00:06:50.260 align:middle line:90% into high interest rate loans. 00:06:50.260 --> 00:06:52.870 align:middle line:84% And that might be just specifically 00:06:52.870 --> 00:06:55.600 align:middle line:84% targeting certain types of populations, perhaps 00:06:55.600 --> 00:06:58.000 align:middle line:84% populations with low education or the like. 00:06:58.000 --> 00:07:00.310 align:middle line:84% And then, other types of institutional structure, 00:07:00.310 --> 00:07:04.600 align:middle line:84% defaults or the like, where essentially society sets up 00:07:04.600 --> 00:07:07.330 align:middle line:84% certain structures that make it easy to make right decisions 00:07:07.330 --> 00:07:08.200 align:middle line:90% for people. 00:07:08.200 --> 00:07:12.160 align:middle line:84% And those are often more developed or more pronounced 00:07:12.160 --> 00:07:14.810 align:middle line:90% in richer population. 00:07:14.810 --> 00:07:17.500 align:middle line:84% A second broad class of explanations 00:07:17.500 --> 00:07:19.900 align:middle line:84% is selection or omitted variables, 00:07:19.900 --> 00:07:25.590 align:middle line:84% which are things like education, intelligence-- 00:07:25.590 --> 00:07:28.540 align:middle line:84% by no means endorsing these views-- 00:07:28.540 --> 00:07:31.720 align:middle line:84% effort or mistakes and say people 00:07:31.720 --> 00:07:33.400 align:middle line:90% might make more mistakes. 00:07:33.400 --> 00:07:37.390 align:middle line:84% And those mistakes might lead to poverty. 00:07:37.390 --> 00:07:39.340 align:middle line:84% And therefore, you might see that correlation. 00:07:39.340 --> 00:07:40.840 align:middle line:84% To be very clear, this is not my view. 00:07:40.840 --> 00:07:42.340 align:middle line:84% But in some sense, in principle, you 00:07:42.340 --> 00:07:43.960 align:middle line:90% could have that explanation. 00:07:43.960 --> 00:07:47.013 align:middle line:84% Now, today, we're going to discuss 00:07:47.013 --> 00:07:49.180 align:middle line:84% a different explanation, as I just told you already. 00:07:49.180 --> 00:07:50.805 align:middle line:84% We're going to talk about the treatment 00:07:50.805 --> 00:07:52.340 align:middle line:90% effect of poverty itself. 00:07:52.340 --> 00:07:53.530 align:middle line:90% So it's not that people-- 00:07:53.530 --> 00:07:56.350 align:middle line:84% because of other factors, because they're 00:07:56.350 --> 00:07:58.840 align:middle line:90% different in certain ways-- 00:07:58.840 --> 00:08:01.930 align:middle line:84% end up being poor and that explains this correlation. 00:08:01.930 --> 00:08:03.430 align:middle line:84% The idea is that they happen to be 00:08:03.430 --> 00:08:08.260 align:middle line:84% poor because of bad luck, essentially, or the family 00:08:08.260 --> 00:08:10.810 align:middle line:90% structure, et cetera. 00:08:10.810 --> 00:08:13.630 align:middle line:84% And then features of their environment 00:08:13.630 --> 00:08:16.660 align:middle line:90% might affect people's choices-- 00:08:16.660 --> 00:08:19.370 align:middle line:84% their productivity, et cetera, OK? 00:08:19.370 --> 00:08:21.550 align:middle line:84% And that's the broad the core of this hypothesis 00:08:21.550 --> 00:08:26.350 align:middle line:84% that people have proposed and started to try and test. 00:08:26.350 --> 00:08:28.660 align:middle line:84% And this is a fundamentally different view 00:08:28.660 --> 00:08:30.190 align:middle line:84% than what we discussed on Monday. 00:08:30.190 --> 00:08:32.980 align:middle line:84% Because on Monday, we were saying, look, there's people-- 00:08:32.980 --> 00:08:37.179 align:middle line:84% the rich and the poor have certain behavioral biases, 00:08:37.179 --> 00:08:40.450 align:middle line:84% and misinformation, or wrong beliefs, et cetera. 00:08:40.450 --> 00:08:47.830 align:middle line:84% And we happen to be particularly interested in developing 00:08:47.830 --> 00:08:49.030 align:middle line:90% country population. 00:08:49.030 --> 00:08:51.160 align:middle line:84% And using universal psychology might 00:08:51.160 --> 00:08:53.763 align:middle line:84% be able to help us understand these patterns. 00:08:53.763 --> 00:08:54.680 align:middle line:90% This one is different. 00:08:54.680 --> 00:08:57.910 align:middle line:84% This is saying there's something specific about poverty that 00:08:57.910 --> 00:09:01.660 align:middle line:84% might affect or might cause certain intermediary outcome. 00:09:01.660 --> 00:09:06.250 align:middle line:84% And those, in turn, might affect people's choices. 00:09:06.250 --> 00:09:10.240 align:middle line:84% A slightly different motivation is coming from the work 00:09:10.240 --> 00:09:13.330 align:middle line:84% that you discussed with Esther last week about poverty traps. 00:09:13.330 --> 00:09:16.120 align:middle line:84% So recent work has shown some evidence of poverty traps, 00:09:16.120 --> 00:09:21.040 align:middle line:84% in particular, or evidence of long-run impacts on earnings 00:09:21.040 --> 00:09:24.220 align:middle line:84% consumption, et cetera of these ultra poor type 00:09:24.220 --> 00:09:26.305 align:middle line:84% intervention, these multifaceted, anti-poverty 00:09:26.305 --> 00:09:28.720 align:middle line:90% programs. 00:09:28.720 --> 00:09:30.760 align:middle line:84% And we have some evidence that seems 00:09:30.760 --> 00:09:34.150 align:middle line:84% to be something unlocked or something that is essentially 00:09:34.150 --> 00:09:38.500 align:middle line:84% happening in people's lives after they 00:09:38.500 --> 00:09:40.090 align:middle line:90% have received these programs. 00:09:40.090 --> 00:09:41.770 align:middle line:84% But we don't have a great understanding 00:09:41.770 --> 00:09:44.020 align:middle line:90% what the underlying channel is. 00:09:44.020 --> 00:09:47.890 align:middle line:84% Is it occupational choice, or is it something else? 00:09:47.890 --> 00:09:51.130 align:middle line:84% And one broad hypothesis might be that psychological factor 00:09:51.130 --> 00:09:53.077 align:middle line:90% could play at least some role. 00:09:53.077 --> 00:09:54.910 align:middle line:84% It could be that, for example, mental health 00:09:54.910 --> 00:09:57.580 align:middle line:84% has improved through these types of programs. 00:09:57.580 --> 00:10:02.740 align:middle line:84% And then, that leads to improved use of money, productivity, 00:10:02.740 --> 00:10:04.945 align:middle line:90% decision-making, et cetera. 00:10:04.945 --> 00:10:07.300 align:middle line:84% And so that the broad idea then, as well, there's 00:10:07.300 --> 00:10:09.190 align:middle line:84% a behavioral poverty trap if you want 00:10:09.190 --> 00:10:11.120 align:middle line:84% that poverty affects cognition, stress, 00:10:11.120 --> 00:10:12.490 align:middle line:90% mental health, et cetera. 00:10:12.490 --> 00:10:15.910 align:middle line:84% And these factors, in turn, could impede economic decision 00:10:15.910 --> 00:10:17.330 align:middle line:90% making and productivity. 00:10:17.330 --> 00:10:19.570 align:middle line:84% So for example, savings behavior, 00:10:19.570 --> 00:10:23.600 align:middle line:84% insurance choices, but also labor supply, productivity 00:10:23.600 --> 00:10:24.860 align:middle line:90% earnings, et cetera. 00:10:24.860 --> 00:10:28.040 align:middle line:84% And these then in turn might affect and exacerbate 00:10:28.040 --> 00:10:30.710 align:middle line:90% or sustain future poverty. 00:10:30.710 --> 00:10:31.210 align:middle line:90% OK. 00:10:31.210 --> 00:10:32.650 align:middle line:90% Any questions on this so far? 00:10:32.650 --> 00:10:36.990 align:middle line:90% 00:10:36.990 --> 00:10:39.080 align:middle line:84% So then there is different lines of research 00:10:39.080 --> 00:10:41.660 align:middle line:84% that in some sense parallel and the very similar overall 00:10:41.660 --> 00:10:45.030 align:middle line:84% and look at quite related hypotheses. 00:10:45.030 --> 00:10:48.260 align:middle line:84% So the first one, and that's like among the earliest 00:10:48.260 --> 00:10:50.510 align:middle line:84% work in this area, which we want to talk about 00:10:50.510 --> 00:10:52.760 align:middle line:90% first a little bit is scarcity. 00:10:52.760 --> 00:10:55.280 align:middle line:84% This is work by, in particular, Sendhill Mullainathan 00:10:55.280 --> 00:10:57.710 align:middle line:84% and Eldar Shafir are coauthors, which 00:10:57.710 --> 00:11:01.160 align:middle line:84% is the idea that monetary concerns capture 00:11:01.160 --> 00:11:02.250 align:middle line:90% cognitive function. 00:11:02.250 --> 00:11:04.790 align:middle line:84% So essentially, poverty captures people's minds. 00:11:04.790 --> 00:11:08.630 align:middle line:84% When people have worries about paying their bills, 00:11:08.630 --> 00:11:11.600 align:middle line:84% about feeding their children, about being 00:11:11.600 --> 00:11:14.630 align:middle line:84% able to pay school fees, about being able to pay rent, 00:11:14.630 --> 00:11:17.330 align:middle line:84% that is very taxing in terms of cognition. 00:11:17.330 --> 00:11:22.890 align:middle line:84% And that's, in turn, affecting people's choices, productivity, 00:11:22.890 --> 00:11:24.140 align:middle line:90% and so on. 00:11:24.140 --> 00:11:25.700 align:middle line:84% A related hypothesis is, and that's 00:11:25.700 --> 00:11:30.530 align:middle line:84% kind of in fact, quite difficult to differentiate or separate 00:11:30.530 --> 00:11:31.650 align:middle line:90% is stress. 00:11:31.650 --> 00:11:35.670 align:middle line:84% So poverty causes stress and stress causes future poverty. 00:11:35.670 --> 00:11:38.030 align:middle line:84% There's some evidence of the former, so like, 00:11:38.030 --> 00:11:41.450 align:middle line:84% there's research by Johannes Haushofer and coauthors 00:11:41.450 --> 00:11:44.780 align:middle line:84% that shows that when people have negative shocks, 00:11:44.780 --> 00:11:46.520 align:middle line:84% we see increased levels of stress 00:11:46.520 --> 00:11:49.550 align:middle line:84% as measured by cortisol and other types of levels. 00:11:49.550 --> 00:11:51.710 align:middle line:84% There's less evidence of the latter, at least 00:11:51.710 --> 00:11:52.470 align:middle line:90% in the short run. 00:11:52.470 --> 00:11:54.770 align:middle line:84% So when you stress people out in the lab, 00:11:54.770 --> 00:11:58.100 align:middle line:84% you don't actually find lots of evidence of people doing 00:11:58.100 --> 00:12:01.430 align:middle line:90% worse in certain behaviors. 00:12:01.430 --> 00:12:03.890 align:middle line:84% And that might be, in part, because stress actually 00:12:03.890 --> 00:12:05.760 align:middle line:84% mobilizes resources and attention. 00:12:05.760 --> 00:12:07.790 align:middle line:84% And so in some way in the short run, 00:12:07.790 --> 00:12:10.550 align:middle line:84% this hypothesis might not be quite right. 00:12:10.550 --> 00:12:12.860 align:middle line:84% However, it might well be that long term stress 00:12:12.860 --> 00:12:14.130 align:middle line:90% can impair mental health. 00:12:14.130 --> 00:12:15.672 align:middle line:84% So it might be that in the short run, 00:12:15.672 --> 00:12:18.590 align:middle line:84% stress is actually beneficial for having attention or being-- 00:12:18.590 --> 00:12:21.320 align:middle line:84% so this is kind of the idea, if a tiger runs after you, 00:12:21.320 --> 00:12:23.850 align:middle line:84% you are going to be really focused and run really fast. 00:12:23.850 --> 00:12:25.070 align:middle line:90% That's good. 00:12:25.070 --> 00:12:28.160 align:middle line:84% And you might make actually good decisions, et cetera. 00:12:28.160 --> 00:12:31.980 align:middle line:84% But if you're stressed out for a long time, 00:12:31.980 --> 00:12:33.680 align:middle line:84% for many months or years, that might 00:12:33.680 --> 00:12:36.320 align:middle line:84% be really bad for your physical or mental health. 00:12:36.320 --> 00:12:39.620 align:middle line:84% And that, in turn, might then lead to future poverty. 00:12:39.620 --> 00:12:41.450 align:middle line:84% That part of the hypothesis hasn't really 00:12:41.450 --> 00:12:44.630 align:middle line:84% been tested that well, in part, because it's 00:12:44.630 --> 00:12:46.760 align:middle line:84% really hard to test, because it's 00:12:46.760 --> 00:12:48.710 align:middle line:84% really hard to separate this from other types 00:12:48.710 --> 00:12:50.443 align:middle line:90% of interventions. 00:12:50.443 --> 00:12:52.110 align:middle line:84% Third, we are going to talk a little bit 00:12:52.110 --> 00:12:53.110 align:middle line:90% about sleep deprivation. 00:12:53.110 --> 00:12:56.310 align:middle line:84% This is the paper you read, so that the poor are 00:12:56.310 --> 00:12:57.810 align:middle line:84% sorry-- this is a typo, but the poor 00:12:57.810 --> 00:12:58.980 align:middle line:90% are more likely to be tired. 00:12:58.980 --> 00:13:01.830 align:middle line:84% They're also more likely to be in pain or drink excessively. 00:13:01.830 --> 00:13:05.820 align:middle line:84% But so broadly, that's to say, people's sleeping conditions, 00:13:05.820 --> 00:13:11.490 align:middle line:84% in particular among the urban poor, are very bad, 00:13:11.490 --> 00:13:12.990 align:middle line:84% and that in turn might affect sleep, 00:13:12.990 --> 00:13:15.510 align:middle line:84% and sleep might have certain effects. 00:13:15.510 --> 00:13:17.880 align:middle line:84% The paper that you read is not supporting that either, 00:13:17.880 --> 00:13:19.830 align:middle line:84% but there might be other directions, 00:13:19.830 --> 00:13:23.850 align:middle line:84% in particular, when it comes to sleep quality that might 00:13:23.850 --> 00:13:27.820 align:middle line:90% lead to more evidence on this. 00:13:27.820 --> 00:13:32.190 align:middle line:84% And then, the fourth hypothesis is broadly 00:13:32.190 --> 00:13:34.578 align:middle line:84% mental health in particular, depression and anxiety. 00:13:34.578 --> 00:13:36.870 align:middle line:84% And this is kind of in some ways where lots of research 00:13:36.870 --> 00:13:38.640 align:middle line:90% recently has been done. 00:13:38.640 --> 00:13:40.860 align:middle line:84% Poverty is associated with mental ill health. 00:13:40.860 --> 00:13:43.680 align:middle line:84% It's essentially, not having money is really 00:13:43.680 --> 00:13:45.300 align:middle line:90% bad for your mental health. 00:13:45.300 --> 00:13:47.970 align:middle line:84% There's lots of evidence of that from cash transfers 00:13:47.970 --> 00:13:49.410 align:middle line:84% and other types of interventions, 00:13:49.410 --> 00:13:51.477 align:middle line:84% or from like negative income shocks, 00:13:51.477 --> 00:13:54.060 align:middle line:84% in particular, when people lose their job, for example, that's 00:13:54.060 --> 00:13:55.410 align:middle line:90% bad for their mental health. 00:13:55.410 --> 00:13:58.680 align:middle line:84% And so then the question is, well does that lead then 00:13:58.680 --> 00:14:02.310 align:middle line:90% in turn to worse outcome? 00:14:02.310 --> 00:14:04.260 align:middle line:84% Then there's a bunch of other factors 00:14:04.260 --> 00:14:08.520 align:middle line:84% that have been studied more or less in different settings, 00:14:08.520 --> 00:14:10.440 align:middle line:84% things like sadness, lack of hope, 00:14:10.440 --> 00:14:14.070 align:middle line:84% self-efficacy, substance abuse, pain, et cetera. 00:14:14.070 --> 00:14:15.763 align:middle line:84% These have been studied to some degree, 00:14:15.763 --> 00:14:17.430 align:middle line:84% but these could also be quite important, 00:14:17.430 --> 00:14:19.410 align:middle line:84% but I'm not going to talk about them too much. 00:14:19.410 --> 00:14:21.660 align:middle line:84% It's important to understand that these things are not 00:14:21.660 --> 00:14:23.760 align:middle line:84% mutually exclusive, so they could be all going 00:14:23.760 --> 00:14:27.630 align:middle line:84% on for a range of people, different people 00:14:27.630 --> 00:14:32.950 align:middle line:84% could be affected by different types of factors. 00:14:32.950 --> 00:14:35.580 align:middle line:84% And we need to understand a little bit 00:14:35.580 --> 00:14:39.250 align:middle line:84% better how to aggregate these types of factors. 00:14:39.250 --> 00:14:42.732 align:middle line:84% So like the type of idea of how to test for poverty traps 00:14:42.732 --> 00:14:44.940 align:middle line:84% that you talked about last week with Esther might not 00:14:44.940 --> 00:14:48.480 align:middle line:84% necessarily apply here, because if you see only 5% or 10% 00:14:48.480 --> 00:14:51.060 align:middle line:84% of people are affected by each of these factors, 00:14:51.060 --> 00:14:55.380 align:middle line:84% if you measure these average as a elasticities of income, 00:14:55.380 --> 00:14:58.500 align:middle line:84% unlike say sleep, and then sleep on other factors, that 00:14:58.500 --> 00:15:01.150 align:middle line:84% might not be the right thing, the right metric to consider. 00:15:01.150 --> 00:15:03.420 align:middle line:84% So I'll return to this at the very end. 00:15:03.420 --> 00:15:04.230 align:middle line:90% Yeah? 00:15:04.230 --> 00:15:06.660 align:middle line:84% AUDIENCE: How exactly do you distinguish-- or is there 00:15:06.660 --> 00:15:09.150 align:middle line:84% a clear distinction between number two and number four 00:15:09.150 --> 00:15:11.810 align:middle line:90% here? 00:15:11.810 --> 00:15:15.587 align:middle line:84% Is stress kind of one subcomponent of mental health 00:15:15.587 --> 00:15:17.670 align:middle line:84% broadly, and it seems like there's similar factors 00:15:17.670 --> 00:15:19.033 align:middle line:90% at play that [INAUDIBLE]. 00:15:19.033 --> 00:15:21.450 align:middle line:84% You're both more stressed out and [INAUDIBLE] the question 00:15:21.450 --> 00:15:24.133 align:middle line:84% like, is this something that we can interpret [INAUDIBLE]?? 00:15:24.133 --> 00:15:26.050 align:middle line:84% FRANK SCHILBACH: Yeah, that's a good question. 00:15:26.050 --> 00:15:28.157 align:middle line:84% In fact, one and two are actually 00:15:28.157 --> 00:15:29.740 align:middle line:84% quite hard to empirically distinguish, 00:15:29.740 --> 00:15:31.210 align:middle line:84% because if you think about money, 00:15:31.210 --> 00:15:33.700 align:middle line:84% that's stressful and worrisome and worries. 00:15:33.700 --> 00:15:39.650 align:middle line:84% So I think here, stress one, you can have a strict definition, 00:15:39.650 --> 00:15:41.433 align:middle line:84% which is like a biological reaction 00:15:41.433 --> 00:15:43.600 align:middle line:84% that you can measure through cortisol or other types 00:15:43.600 --> 00:15:46.420 align:middle line:90% of biomarkers. 00:15:46.420 --> 00:15:49.930 align:middle line:84% While here I mean specifically depression and anxiety. 00:15:49.930 --> 00:15:52.360 align:middle line:84% Of course, there could be other mental health conditions. 00:15:52.360 --> 00:15:55.150 align:middle line:84% And of course, stress this is what I was talking here 00:15:55.150 --> 00:15:57.340 align:middle line:84% about stress, long term stress, in fact, 00:15:57.340 --> 00:16:00.890 align:middle line:90% leads to mental health problems. 00:16:00.890 --> 00:16:03.010 align:middle line:84% So I wouldn't necessarily see these as mutually 00:16:03.010 --> 00:16:03.990 align:middle line:90% exclusive at all. 00:16:03.990 --> 00:16:05.365 align:middle line:84% I'm more kind of in some sense, I 00:16:05.365 --> 00:16:06.740 align:middle line:84% was giving sort of like a history 00:16:06.740 --> 00:16:08.800 align:middle line:84% of like the research that's done in this area. 00:16:08.800 --> 00:16:13.420 align:middle line:84% In some sense, these two were kind of like first 00:16:13.420 --> 00:16:16.360 align:middle line:84% in the studies by Mullainathan and Shafir here 00:16:16.360 --> 00:16:19.360 align:middle line:84% and by Johannes Haushofer and colleagues here. 00:16:19.360 --> 00:16:24.430 align:middle line:84% And then mental health, which has come up most recently, 00:16:24.430 --> 00:16:26.740 align:middle line:84% mostly focusing on depression and anxiety, 00:16:26.740 --> 00:16:29.050 align:middle line:84% but even depression and anxiety hard to separate 00:16:29.050 --> 00:16:34.150 align:middle line:84% the very correlated and very happen in very, often, 00:16:34.150 --> 00:16:35.740 align:middle line:84% like effect of same types of people. 00:16:35.740 --> 00:16:37.157 align:middle line:84% Now, there are different concepts, 00:16:37.157 --> 00:16:39.490 align:middle line:84% like depression is very different from anxiety 00:16:39.490 --> 00:16:42.280 align:middle line:90% in what it does. 00:16:42.280 --> 00:16:44.200 align:middle line:84% But empirically, it's actually very hard 00:16:44.200 --> 00:16:47.140 align:middle line:84% to separate them because they're so correlated. 00:16:47.140 --> 00:16:49.060 align:middle line:84% Yeah, and I think I'm going to talk 00:16:49.060 --> 00:16:52.610 align:middle line:84% to you about the next paper, which is in some ways, 00:16:52.610 --> 00:16:55.810 align:middle line:84% you can think of like as evidence in favor 00:16:55.810 --> 00:16:57.820 align:middle line:90% if you want or scarcity. 00:16:57.820 --> 00:17:00.190 align:middle line:84% In fact, we are not able to distinguish 00:17:00.190 --> 00:17:02.962 align:middle line:84% other psychological channels, in part, because that's actually 00:17:02.962 --> 00:17:03.670 align:middle line:90% quite hard to do. 00:17:03.670 --> 00:17:05.960 align:middle line:84% I'm going to get back to that in a second. 00:17:05.960 --> 00:17:06.740 align:middle line:90% Yeah? 00:17:06.740 --> 00:17:07.440 align:middle line:84% AUDIENCE: Feel free to defer this 00:17:07.440 --> 00:17:09.140 align:middle line:84% if it's not directly relevant, but how 00:17:09.140 --> 00:17:15.020 align:middle line:84% do these types of studies speak to the happy poor hypothesis? 00:17:15.020 --> 00:17:17.178 align:middle line:84% It's often commented that people who 00:17:17.178 --> 00:17:20.970 align:middle line:84% are poor-- [? to a ?] particular shock are more resilient 00:17:20.970 --> 00:17:23.352 align:middle line:84% or don't react the same way because they're used to it 00:17:23.352 --> 00:17:24.310 align:middle line:90% or something like that? 00:17:24.310 --> 00:17:25.727 align:middle line:84% FRANK SCHILBACH: Yeah, so that's-- 00:17:25.727 --> 00:17:28.490 align:middle line:84% I think that's like in some sense, 00:17:28.490 --> 00:17:32.470 align:middle line:84% like a myth that has persisted for a long time that turns out 00:17:32.470 --> 00:17:33.310 align:middle line:90% to be just wrong. 00:17:33.310 --> 00:17:37.420 align:middle line:84% So there's lots of evidence, and there's in particular, 00:17:37.420 --> 00:17:40.340 align:middle line:84% there's a very nice paper by Johannes Haushofer and Ernst 00:17:40.340 --> 00:17:43.780 align:middle line:84% Fehr that looks at the psychology of poverty. 00:17:43.780 --> 00:17:50.410 align:middle line:84% And the work by Justin Wolfers and Betsey Stevenson 00:17:50.410 --> 00:17:56.360 align:middle line:84% shows essentially that both in the cross section when 00:17:56.360 --> 00:17:58.990 align:middle line:84% you look at across countries, if you look at within countries, 00:17:58.990 --> 00:18:02.140 align:middle line:84% if you look at changes over time, income 00:18:02.140 --> 00:18:06.550 align:middle line:84% is very strongly associated with psychological well-being. 00:18:06.550 --> 00:18:10.660 align:middle line:84% So essentially, being richer makes 00:18:10.660 --> 00:18:13.090 align:middle line:90% you happier in various ways. 00:18:13.090 --> 00:18:16.000 align:middle line:84% And there's lots and lots of evidence of that, 00:18:16.000 --> 00:18:20.680 align:middle line:84% including cash transfers make people happy. 00:18:20.680 --> 00:18:22.330 align:middle line:84% And it's a very robust body of evidence 00:18:22.330 --> 00:18:25.060 align:middle line:84% that shows that cash transfers improve 00:18:25.060 --> 00:18:26.590 align:middle line:84% psychological well-being as measured 00:18:26.590 --> 00:18:29.290 align:middle line:84% by happiness or other types of measures 00:18:29.290 --> 00:18:32.830 align:middle line:84% as measured by depression and anxiety and so on. 00:18:32.830 --> 00:18:37.510 align:middle line:84% And so I think part of the reason 00:18:37.510 --> 00:18:39.580 align:middle line:84% is, as you're alluding to, is like the poor are 00:18:39.580 --> 00:18:42.790 align:middle line:90% exposed to a lot more shocks. 00:18:42.790 --> 00:18:46.540 align:middle line:84% Now it might be that each single shock has 00:18:46.540 --> 00:18:48.910 align:middle line:84% less of an effect on people, but I 00:18:48.910 --> 00:18:51.430 align:middle line:84% think the effect of just being exposed to so many 00:18:51.430 --> 00:18:55.720 align:middle line:84% more adverse factors just dwarfs any effect of saying 00:18:55.720 --> 00:18:57.463 align:middle line:84% like, maybe people are more resilient. 00:18:57.463 --> 00:18:59.380 align:middle line:84% I don't actually know whether there's evidence 00:18:59.380 --> 00:19:02.800 align:middle line:90% for that resilience hypothesis. 00:19:02.800 --> 00:19:06.100 align:middle line:84% But there's clear evidence that as a whole, 00:19:06.100 --> 00:19:08.290 align:middle line:84% being rich or having more money is 00:19:08.290 --> 00:19:11.140 align:middle line:84% good for your happiness and psychological well-being 00:19:11.140 --> 00:19:12.100 align:middle line:90% on average. 00:19:12.100 --> 00:19:15.370 align:middle line:90% 00:19:15.370 --> 00:19:16.960 align:middle line:90% OK. 00:19:16.960 --> 00:19:19.570 align:middle line:84% So let me give you a very brief overview 00:19:19.570 --> 00:19:23.990 align:middle line:84% of the work on scarcity and the idea behind that. 00:19:23.990 --> 00:19:26.860 align:middle line:84% And so the idea is that essentially scarcity is not 00:19:26.860 --> 00:19:28.720 align:middle line:90% having enough of something. 00:19:28.720 --> 00:19:30.010 align:middle line:90% And you can talk about money. 00:19:30.010 --> 00:19:33.510 align:middle line:84% You can talk about can think about time, friends, hair, et 00:19:33.510 --> 00:19:34.960 align:middle line:84% cetera, you name it, just anything 00:19:34.960 --> 00:19:37.780 align:middle line:90% that you don't have enough of. 00:19:37.780 --> 00:19:40.240 align:middle line:84% If you read that book, the definition of scarcity 00:19:40.240 --> 00:19:44.080 align:middle line:84% is a little bit loose in the sense that it says essentially 00:19:44.080 --> 00:19:45.910 align:middle line:84% it's a feeling of not having enough 00:19:45.910 --> 00:19:49.582 align:middle line:84% of something, which if you think about it, 00:19:49.582 --> 00:19:51.790 align:middle line:84% that's empirically kind of a little bit hard to test, 00:19:51.790 --> 00:19:53.260 align:middle line:90% because how do I know? 00:19:53.260 --> 00:19:54.970 align:middle line:84% I can tell you how many friends you have, 00:19:54.970 --> 00:19:57.850 align:middle line:84% or I can tell you how much time you have 00:19:57.850 --> 00:19:59.150 align:middle line:90% or how much money you have. 00:19:59.150 --> 00:20:01.630 align:middle line:84% It's much harder to measure whether you feel that you 00:20:01.630 --> 00:20:03.530 align:middle line:90% have enough of something. 00:20:03.530 --> 00:20:06.910 align:middle line:84% And so the idea then is that scarcity captures people's 00:20:06.910 --> 00:20:13.270 align:middle line:84% mental or cognitive capacity, which they call bandwidth. 00:20:13.270 --> 00:20:15.650 align:middle line:84% So importantly, that happens automatically. 00:20:15.650 --> 00:20:17.590 align:middle line:84% So it's not like people decide to think 00:20:17.590 --> 00:20:21.310 align:middle line:84% about these types of things, but instead, scarcity captures 00:20:21.310 --> 00:20:22.330 align:middle line:90% people's minds. 00:20:22.330 --> 00:20:26.620 align:middle line:84% So that's to say, when people are unhappy, 00:20:26.620 --> 00:20:28.060 align:middle line:90% or suppose people, for example-- 00:20:28.060 --> 00:20:31.420 align:middle line:84% one example that they give in the book 00:20:31.420 --> 00:20:33.550 align:middle line:84% or in some of their papers is that 00:20:33.550 --> 00:20:39.160 align:middle line:84% suppose you get into a bad fight or break up with a loved one 00:20:39.160 --> 00:20:43.610 align:middle line:84% or boyfriend or girlfriend or the like, 00:20:43.610 --> 00:20:45.440 align:middle line:84% and then you have a job interview. 00:20:45.440 --> 00:20:48.040 align:middle line:84% It's really, really not productive doing that job 00:20:48.040 --> 00:20:51.850 align:middle line:84% interview to think about what went wrong during that breakup 00:20:51.850 --> 00:20:54.220 align:middle line:84% or what could you have done differently or the like, 00:20:54.220 --> 00:20:56.470 align:middle line:84% but these thoughts might just invade your mind. 00:20:56.470 --> 00:20:59.890 align:middle line:84% You might not be able to block them overall. 00:20:59.890 --> 00:21:01.653 align:middle line:84% That's not productive thinking in a sense, 00:21:01.653 --> 00:21:03.070 align:middle line:84% you should probably figure out how 00:21:03.070 --> 00:21:06.773 align:middle line:84% to make amends with your friends or partner or the like. 00:21:06.773 --> 00:21:09.190 align:middle line:84% But during that job interview, it's really not a good idea 00:21:09.190 --> 00:21:10.100 align:middle line:90% to do that. 00:21:10.100 --> 00:21:12.160 align:middle line:84% And so similarly the idea is here 00:21:12.160 --> 00:21:14.320 align:middle line:84% that while it might be very productive to think 00:21:14.320 --> 00:21:18.070 align:middle line:84% about your finances if you're poor, how to pay your bills 00:21:18.070 --> 00:21:20.710 align:middle line:84% or how to pay back your neighbor or how 00:21:20.710 --> 00:21:23.260 align:middle line:84% to pay school fees and so on, it's 00:21:23.260 --> 00:21:25.165 align:middle line:84% not useful to do that while you are at work. 00:21:25.165 --> 00:21:28.180 align:middle line:90% 00:21:28.180 --> 00:21:30.790 align:middle line:84% And so poverty, specifically then 00:21:30.790 --> 00:21:33.380 align:middle line:84% makes monetary concerns top of mind because of course, 00:21:33.380 --> 00:21:35.920 align:middle line:84% poverty by definition is not having enough money, 00:21:35.920 --> 00:21:38.950 align:middle line:84% so it implies, in some ways, that the poor act more 00:21:38.950 --> 00:21:40.630 align:middle line:84% rational in financial choices, have 00:21:40.630 --> 00:21:42.970 align:middle line:84% greater focus and improved decision making in some ways. 00:21:42.970 --> 00:21:44.830 align:middle line:84% And there's some evidence in support of that, 00:21:44.830 --> 00:21:47.080 align:middle line:84% that the poor, for example, are better at like knowing 00:21:47.080 --> 00:21:49.840 align:middle line:90% prices and the like. 00:21:49.840 --> 00:21:52.670 align:middle line:84% But of course, overall cognitive capacity is limited. 00:21:52.670 --> 00:21:55.800 align:middle line:84% There's only so much you can think about at one time. 00:21:55.800 --> 00:21:58.000 align:middle line:84% So now if you think more about monetary concerns 00:21:58.000 --> 00:21:59.375 align:middle line:84% that essentially implies that you 00:21:59.375 --> 00:22:01.750 align:middle line:84% have less cognitive resources available to think 00:22:01.750 --> 00:22:02.600 align:middle line:90% about other things. 00:22:02.600 --> 00:22:04.540 align:middle line:84% And so then scarcity deteriorates 00:22:04.540 --> 00:22:07.090 align:middle line:84% performance at other cognitive tasks, 00:22:07.090 --> 00:22:09.910 align:middle line:84% and therefore, might impede with your decision making, 00:22:09.910 --> 00:22:11.780 align:middle line:90% your productivity, and so on. 00:22:11.780 --> 00:22:13.570 align:middle line:84% And that's the core of the hypothesis 00:22:13.570 --> 00:22:14.950 align:middle line:90% of this type of work. 00:22:14.950 --> 00:22:17.270 align:middle line:84% Importantly, this is not rational inattention. 00:22:17.270 --> 00:22:20.920 align:middle line:84% So it's not about, you decide to think about monetary issues 00:22:20.920 --> 00:22:23.390 align:middle line:84% because you want to solve them, that's really important. 00:22:23.390 --> 00:22:25.240 align:middle line:84% But rather, it's like, essentially it 00:22:25.240 --> 00:22:30.140 align:middle line:84% captures your mind and distracts you in some inefficient way. 00:22:30.140 --> 00:22:33.525 align:middle line:84% Now, what evidence do we have in favor of this hypothesis? 00:22:33.525 --> 00:22:35.150 align:middle line:84% So there is this very influential paper 00:22:35.150 --> 00:22:37.970 align:middle line:84% by Mani et al. that looks at or presents 00:22:37.970 --> 00:22:40.607 align:middle line:84% two complementary studies, pieces of evidence. 00:22:40.607 --> 00:22:42.440 align:middle line:84% One is what's called the "Mall study," which 00:22:42.440 --> 00:22:46.430 align:middle line:84% is, in a mall in Trenton, New Jersey, 00:22:46.430 --> 00:22:50.750 align:middle line:84% make people think about monetary troubles and find that doing so 00:22:50.750 --> 00:22:54.770 align:middle line:84% lowers measured IQ-type cognitive function 00:22:54.770 --> 00:22:58.080 align:middle line:90% measures by about 10 IQ points. 00:22:58.080 --> 00:23:03.350 align:middle line:84% So essentially, when you do that with lower income participants, 00:23:03.350 --> 00:23:04.460 align:middle line:90% there's large effects. 00:23:04.460 --> 00:23:06.780 align:middle line:84% For richer participants, nothing happens. 00:23:06.780 --> 00:23:09.710 align:middle line:84% And these effects are large and pronounced. 00:23:09.710 --> 00:23:12.770 align:middle line:84% Similarly, in what they call the "Harvest Study," 00:23:12.770 --> 00:23:15.830 align:middle line:84% they look at farmers in rural Tamil Nadu in India 00:23:15.830 --> 00:23:17.295 align:middle line:84% before and after harvest, and then 00:23:17.295 --> 00:23:19.670 align:middle line:84% find that after harvest when people have more money, when 00:23:19.670 --> 00:23:22.880 align:middle line:84% they're flush with cash, they have higher cognitive function 00:23:22.880 --> 00:23:24.000 align:middle line:90% than before. 00:23:24.000 --> 00:23:27.170 align:middle line:84% So these studies are very much complementary 00:23:27.170 --> 00:23:29.090 align:middle line:84% because one is more lobby, the other one 00:23:29.090 --> 00:23:31.890 align:middle line:84% is more a field study in sorta view 00:23:31.890 --> 00:23:34.580 align:middle line:84% of what settings have been extremely 00:23:34.580 --> 00:23:38.150 align:middle line:84% influential overall in affecting people's thinking. 00:23:38.150 --> 00:23:41.690 align:middle line:84% At the same, time, we have seen very few successful replication 00:23:41.690 --> 00:23:45.260 align:middle line:84% or actually evidence that this is going on in other settings. 00:23:45.260 --> 00:23:47.330 align:middle line:84% For example, Cavalho et al find no differences 00:23:47.330 --> 00:23:51.461 align:middle line:84% around paydays among US households. 00:23:51.461 --> 00:23:54.320 align:middle line:84% Dietmar Fehr and coauthors find no differences in cognition 00:23:54.320 --> 00:23:56.480 align:middle line:84% before and after harvest in Zambia. 00:23:56.480 --> 00:24:00.590 align:middle line:84% And in fact, some evidence of when people are poorer, 00:24:00.590 --> 00:24:02.750 align:middle line:84% they make better decisions in some ways. 00:24:02.750 --> 00:24:05.630 align:middle line:84% There's some evidence recently by Heather Schofield 00:24:05.630 --> 00:24:09.380 align:middle line:84% and coauthors that's consistent with the Mall study, 00:24:09.380 --> 00:24:11.780 align:middle line:84% but by and large, it's fair to say 00:24:11.780 --> 00:24:13.400 align:middle line:84% there's not a lot of evidence that 00:24:13.400 --> 00:24:17.360 align:middle line:84% in addition to the original study supports this hypothesis. 00:24:17.360 --> 00:24:19.550 align:middle line:84% And in particular, one broad question 00:24:19.550 --> 00:24:23.630 align:middle line:84% is whether scarcity affects economic behavior. 00:24:23.630 --> 00:24:25.520 align:middle line:84% And the paper that I'm going to talk about 00:24:25.520 --> 00:24:28.850 align:middle line:84% is trying to provide some evidence 00:24:28.850 --> 00:24:36.120 align:middle line:84% of scarcity or other factors affecting economic behaviors. 00:24:36.120 --> 00:24:38.540 align:middle line:84% And so the starting point of this paper 00:24:38.540 --> 00:24:42.080 align:middle line:84% is what I already talked to you about before, which is to say, 00:24:42.080 --> 00:24:45.440 align:middle line:84% financial strain or periods of cash shortfalls, 00:24:45.440 --> 00:24:47.480 align:middle line:84% these might come from unexpected shocks, 00:24:47.480 --> 00:24:50.120 align:middle line:84% but also through predictable cycles. 00:24:50.120 --> 00:24:53.360 align:middle line:84% And when people can't meet their expenses, 00:24:53.360 --> 00:24:54.620 align:middle line:90% that is mentally costly. 00:24:54.620 --> 00:24:57.740 align:middle line:84% It might cause worry, stress, anxiety, sadness et cetera. 00:24:57.740 --> 00:25:00.270 align:middle line:84% And it affects a large fraction of people, 00:25:00.270 --> 00:25:01.970 align:middle line:84% including in India and [? Arisa ?] 00:25:01.970 --> 00:25:05.750 align:middle line:84% rural setting in the lean season, 86% of people 00:25:05.750 --> 00:25:09.530 align:middle line:84% feel being worried about their finances at that time. 00:25:09.530 --> 00:25:11.150 align:middle line:84% Now as I said, there's a literature 00:25:11.150 --> 00:25:13.760 align:middle line:84% that looks at the cognitive consequences of this 00:25:13.760 --> 00:25:16.820 align:middle line:84% in lab type of settings, mostly looking 00:25:16.820 --> 00:25:19.910 align:middle line:84% at things like Raven's matrices type of IQ test. 00:25:19.910 --> 00:25:22.520 align:middle line:84% And what we're trying to do in this paper 00:25:22.520 --> 00:25:24.500 align:middle line:84% is trying to test whether there is 00:25:24.500 --> 00:25:26.180 align:middle line:90% economically important effects. 00:25:26.180 --> 00:25:28.330 align:middle line:84% At the end of the day, you might say, well, 00:25:28.330 --> 00:25:30.943 align:middle line:84% the cognitive measures or the IQ tests 00:25:30.943 --> 00:25:33.110 align:middle line:84% are important of course for people's everyday lives. 00:25:33.110 --> 00:25:35.940 align:middle line:84% But as economists, we kind want to know, does it in fact, 00:25:35.940 --> 00:25:38.970 align:middle line:84% have real world economic consequences? 00:25:38.970 --> 00:25:41.780 align:middle line:84% And so one particularly important measure 00:25:41.780 --> 00:25:45.680 align:middle line:84% is real world worker productivity labor income 00:25:45.680 --> 00:25:48.650 align:middle line:84% overall, which of course then might affect future poverty 00:25:48.650 --> 00:25:52.258 align:middle line:84% through your ability to work and earn money for your family. 00:25:52.258 --> 00:25:54.800 align:middle line:84% And so the question that we're trying to answer in this paper 00:25:54.800 --> 00:25:56.550 align:middle line:84% is, can relieving financial stress 00:25:56.550 --> 00:25:58.730 align:middle line:84% make a person more productive at work? 00:25:58.730 --> 00:26:02.540 align:middle line:84% This is a field experiment with 408 manufacturing workers 00:26:02.540 --> 00:26:05.240 align:middle line:84% in India, where we alleviate financial constraints 00:26:05.240 --> 00:26:07.340 align:middle line:84% by varying the timing of income and look 00:26:07.340 --> 00:26:09.530 align:middle line:90% at effects on productivity. 00:26:09.530 --> 00:26:11.780 align:middle line:84% We're going to have some measure of attention, which 00:26:11.780 --> 00:26:15.080 align:middle line:84% is like in some sense consistent with the hypothesis that 00:26:15.080 --> 00:26:17.060 align:middle line:90% scarcity is at play. 00:26:17.060 --> 00:26:19.280 align:middle line:84% But actually we cannot, and this is getting back 00:26:19.280 --> 00:26:23.780 align:middle line:84% to your question, we cannot really disentangle different 00:26:23.780 --> 00:26:27.290 align:middle line:84% psychological mechanisms such as stress, worry effect, sadness, 00:26:27.290 --> 00:26:28.160 align:middle line:90% even sleep. 00:26:28.160 --> 00:26:29.630 align:middle line:84% Because at the end of the day, we 00:26:29.630 --> 00:26:31.880 align:middle line:84% can say people have lower attention at work, 00:26:31.880 --> 00:26:33.590 align:middle line:84% but we can't rule out that this is 00:26:33.590 --> 00:26:37.410 align:middle line:84% coming from like stress or sadness et cetera and so on. 00:26:37.410 --> 00:26:40.430 align:middle line:84% And if you think about it, it's not entirely clear 00:26:40.430 --> 00:26:43.040 align:middle line:84% why we necessarily care that much, in the sense 00:26:43.040 --> 00:26:45.800 align:middle line:84% these things are all correlated at play. 00:26:45.800 --> 00:26:47.812 align:middle line:84% And in some ways, at the end of the day, 00:26:47.812 --> 00:26:49.520 align:middle line:84% maybe we care more about the bottom line, 00:26:49.520 --> 00:26:51.770 align:middle line:84% does it actually matter for people's behavior 00:26:51.770 --> 00:26:53.700 align:middle line:90% in the real world? 00:26:53.700 --> 00:26:56.600 align:middle line:84% So let me show you some motivational evidence why 00:26:56.600 --> 00:26:58.830 align:middle line:84% this is a reasonable hypothesis to test. 00:26:58.830 --> 00:27:03.740 align:middle line:84% So this is, in again, rural India and Arisa. 00:27:03.740 --> 00:27:06.710 align:middle line:84% These are workers in the lean season. 00:27:06.710 --> 00:27:09.680 align:middle line:90% 71% have outstanding loans. 00:27:09.680 --> 00:27:13.040 align:middle line:84% People have very low levels of liquidity. 00:27:13.040 --> 00:27:15.050 align:middle line:84% When you ask them would you be able to come up 00:27:15.050 --> 00:27:18.380 align:middle line:84% with 1,000 rupees for an emergency, 61% of them 00:27:18.380 --> 00:27:20.900 align:middle line:84% say they're not-- this would be very difficult for them 00:27:20.900 --> 00:27:21.830 align:middle line:90% to do that. 00:27:21.830 --> 00:27:25.580 align:middle line:84% And daily expenses, loans, health issues et cetera 00:27:25.580 --> 00:27:29.550 align:middle line:84% loom really large in people's minds. 00:27:29.550 --> 00:27:31.130 align:middle line:84% When you ask them, what causes you 00:27:31.130 --> 00:27:33.410 align:middle line:84% financial strain, that's the kind of answers 00:27:33.410 --> 00:27:35.840 align:middle line:90% that you might get. 00:27:35.840 --> 00:27:37.490 align:middle line:84% Lots of people, as I told you already, 00:27:37.490 --> 00:27:41.390 align:middle line:84% are worried about money and about their finances. 00:27:41.390 --> 00:27:43.460 align:middle line:84% Importantly, also lots of people are 00:27:43.460 --> 00:27:47.220 align:middle line:84% in fact worried about those finances during work. 00:27:47.220 --> 00:27:49.100 align:middle line:84% So we asked them at the end on one day, 00:27:49.100 --> 00:27:51.230 align:middle line:84% what did you think about during work today? 00:27:51.230 --> 00:27:54.000 align:middle line:84% Many people say they think about the work task, which is great. 00:27:54.000 --> 00:27:56.000 align:middle line:84% And they should be thinking about the work task, 00:27:56.000 --> 00:27:57.020 align:middle line:90% because they were working. 00:27:57.020 --> 00:27:58.520 align:middle line:84% But about half of them, in fact, say 00:27:58.520 --> 00:28:02.960 align:middle line:84% they think about their finances during work. 00:28:02.960 --> 00:28:05.070 align:middle line:84% And this sort of then motivates our experiments, 00:28:05.070 --> 00:28:08.000 align:middle line:84% we try to see, OK, can we reduce people's financial strain 00:28:08.000 --> 00:28:12.000 align:middle line:84% and does that lead to improved productivity? 00:28:12.000 --> 00:28:14.510 align:middle line:84% So a little bit more on the setting, 00:28:14.510 --> 00:28:18.830 align:middle line:84% it's the agricultural lean season in rural Arisa India. 00:28:18.830 --> 00:28:21.560 align:middle line:84% People have intermittent, short-term contract jobs, 00:28:21.560 --> 00:28:23.900 align:middle line:84% between a day and about two weeks. 00:28:23.900 --> 00:28:26.390 align:middle line:84% The work task is a task where we ask 00:28:26.390 --> 00:28:29.130 align:middle line:84% people to make disposable plates for restaurants, 00:28:29.130 --> 00:28:30.620 align:middle line:90% which are like solitary leaves. 00:28:30.620 --> 00:28:32.700 align:middle line:90% I'll show you that in a second. 00:28:32.700 --> 00:28:34.498 align:middle line:84% This is a real world task in the sense 00:28:34.498 --> 00:28:36.290 align:middle line:84% that this is the thing that people actually 00:28:36.290 --> 00:28:37.370 align:middle line:90% do in that area. 00:28:37.370 --> 00:28:40.380 align:middle line:84% When you go to a restaurant in Arisa in rural areas, 00:28:40.380 --> 00:28:42.740 align:middle line:84% you might actually get such a leaf 00:28:42.740 --> 00:28:46.250 align:middle line:90% plates served to eat your food. 00:28:46.250 --> 00:28:48.620 align:middle line:84% People are incentivized, get piece rate for production. 00:28:48.620 --> 00:28:50.690 align:middle line:84% And this is the primary source of income 00:28:50.690 --> 00:28:52.200 align:middle line:90% during the experiment. 00:28:52.200 --> 00:28:53.570 align:middle line:90% So this is quite a bit of money. 00:28:53.570 --> 00:28:56.120 align:middle line:84% These stakes are pretty high for people. 00:28:56.120 --> 00:28:59.970 align:middle line:84% So people should be quite motivated to be productive. 00:28:59.970 --> 00:29:01.820 align:middle line:84% So now what we're trying to do is 00:29:01.820 --> 00:29:06.200 align:middle line:84% we vary people's timing of when workers receive their expected 00:29:06.200 --> 00:29:09.470 align:middle line:84% wage payments, so we'll have essentially variation. 00:29:09.470 --> 00:29:11.480 align:middle line:84% And some people have money earlier than others. 00:29:11.480 --> 00:29:13.480 align:middle line:84% And we can look at that and when you have money, 00:29:13.480 --> 00:29:15.290 align:middle line:84% when you have cash in hand, does it 00:29:15.290 --> 00:29:17.700 align:middle line:90% affect your work productivity? 00:29:17.700 --> 00:29:21.500 align:middle line:84% Now you might say, rightly, why should cash arrival 00:29:21.500 --> 00:29:23.540 align:middle line:84% ease mental load despite prior money, 00:29:23.540 --> 00:29:24.780 align:middle line:90% knowing that money is coming? 00:29:24.780 --> 00:29:26.197 align:middle line:84% So we set it up such that you know 00:29:26.197 --> 00:29:27.830 align:middle line:90% actually the money is coming. 00:29:27.830 --> 00:29:30.230 align:middle line:84% Now, essentially, what it allows people to do, 00:29:30.230 --> 00:29:32.570 align:middle line:84% is it allows people to pay for urgent expenses 00:29:32.570 --> 00:29:35.480 align:middle line:84% to reduce debt overhang that weighs on people's minds 00:29:35.480 --> 00:29:38.930 align:middle line:84% or just feel more vulnerable about potential shocks that 00:29:38.930 --> 00:29:42.320 align:middle line:84% might be coming and perhaps also less embarrassed 00:29:42.320 --> 00:29:45.770 align:middle line:84% towards your neighbors or people that you owe money to. 00:29:45.770 --> 00:29:50.450 align:middle line:84% Now, if you think about it your student life when 00:29:50.450 --> 00:29:52.640 align:middle line:84% at some point towards the end of the month 00:29:52.640 --> 00:29:54.320 align:middle line:84% or towards the end of the semester 00:29:54.320 --> 00:29:56.700 align:middle line:84% or the beginning of the semester, 00:29:56.700 --> 00:30:00.950 align:middle line:84% even if you know that you get a check at the end of the month 00:30:00.950 --> 00:30:02.565 align:middle line:84% or the beginning of the month, having 00:30:02.565 --> 00:30:04.190 align:middle line:84% actually the money in your bank account 00:30:04.190 --> 00:30:05.990 align:middle line:84% and being able to expenses, makes 00:30:05.990 --> 00:30:09.120 align:middle line:84% a surprisingly large difference, at least in my life. 00:30:09.120 --> 00:30:11.090 align:middle line:84% So anyway, so that's kind of like the idea 00:30:11.090 --> 00:30:13.460 align:middle line:90% that we're trying to test. 00:30:13.460 --> 00:30:16.020 align:middle line:84% The experiment in some ways is very simple. 00:30:16.020 --> 00:30:19.070 align:middle line:84% What we do is we have a control group and a treatment group. 00:30:19.070 --> 00:30:22.190 align:middle line:84% Everybody has the same structure of the experiment. 00:30:22.190 --> 00:30:24.380 align:middle line:84% Everybody works for 12 consecutive days. 00:30:24.380 --> 00:30:26.240 align:middle line:84% On day one, we get some first payment 00:30:26.240 --> 00:30:29.300 align:middle line:84% to insure trusts that help people understand 00:30:29.300 --> 00:30:32.720 align:middle line:84% that we actually following through with our payments. 00:30:32.720 --> 00:30:36.328 align:middle line:84% And on day 12, everybody is paid the remaining payments, 00:30:36.328 --> 00:30:38.120 align:middle line:84% whatever they haven't been paid until then. 00:30:38.120 --> 00:30:39.710 align:middle line:90% And everybody works every day. 00:30:39.710 --> 00:30:42.080 align:middle line:84% Everybody is told how much they have produced every day, 00:30:42.080 --> 00:30:45.470 align:middle line:84% so there's no uncertainty about how much they 00:30:45.470 --> 00:30:47.300 align:middle line:90% get for any single day. 00:30:47.300 --> 00:30:49.970 align:middle line:84% And then the main treatment is that on day eight, 00:30:49.970 --> 00:30:52.910 align:middle line:84% the treatment group gets a cash infusion. 00:30:52.910 --> 00:30:56.090 align:middle line:84% They get essentially money for whatever they have worked 00:30:56.090 --> 00:30:58.880 align:middle line:90% from day two to day seven. 00:30:58.880 --> 00:31:00.680 align:middle line:84% And that's quite a bit of money, in part, 00:31:00.680 --> 00:31:02.120 align:middle line:84% because we're in the lean season so people 00:31:02.120 --> 00:31:03.080 align:middle line:90% have very little cash. 00:31:03.080 --> 00:31:04.830 align:middle line:84% So they get quite a bit of money, 00:31:04.830 --> 00:31:07.335 align:middle line:84% which is about three weeks typical earnings 00:31:07.335 --> 00:31:08.960 align:middle line:84% in the lean season, just because people 00:31:08.960 --> 00:31:10.730 align:middle line:84% have so few jobs during that time, 00:31:10.730 --> 00:31:13.640 align:middle line:84% they work something like eight, nine days usually, and somewhat 00:31:13.640 --> 00:31:15.900 align:middle line:90% lower paid during that time. 00:31:15.900 --> 00:31:17.660 align:middle line:84% So what that allows us to do then 00:31:17.660 --> 00:31:20.240 align:middle line:84% we can look at the post pay period 00:31:20.240 --> 00:31:22.877 align:middle line:84% where we can compare the treatment group compared 00:31:22.877 --> 00:31:24.710 align:middle line:84% to the control group, and look at now people 00:31:24.710 --> 00:31:28.760 align:middle line:84% who are flush with cash, are they working more productively? 00:31:28.760 --> 00:31:32.930 align:middle line:84% Now, as I said, this is not unannounced, 00:31:32.930 --> 00:31:36.500 align:middle line:84% so we can actually look-- so we have an announcement day on day 00:31:36.500 --> 00:31:38.870 align:middle line:84% five where we can tell people essentially 00:31:38.870 --> 00:31:41.480 align:middle line:84% when they're being paid so that this does not 00:31:41.480 --> 00:31:42.840 align:middle line:90% come as a surprise. 00:31:42.840 --> 00:31:44.750 align:middle line:84% And then we can-- this allows us to test 00:31:44.750 --> 00:31:46.400 align:middle line:84% for some other mechanisms that I don't 00:31:46.400 --> 00:31:47.720 align:middle line:90% want to get too much into. 00:31:47.720 --> 00:31:50.720 align:middle line:84% We can test for fairness, gift exchange, et cetera, 00:31:50.720 --> 00:31:56.360 align:middle line:84% to try to understand whether the effects that we find 00:31:56.360 --> 00:32:01.310 align:middle line:84% are due to the effects of having more money, 00:32:01.310 --> 00:32:03.890 align:middle line:84% versus the effect of like disliking your employer more 00:32:03.890 --> 00:32:06.632 align:middle line:84% or the like because you're grateful, et cetera. 00:32:06.632 --> 00:32:08.090 align:middle line:84% We spend a lot of time in the paper 00:32:08.090 --> 00:32:12.330 align:middle line:84% trying to address these types of concerns. 00:32:12.330 --> 00:32:13.520 align:middle line:90% So what do we find. 00:32:13.520 --> 00:32:15.782 align:middle line:84% So first, as a sanity check, and these 00:32:15.782 --> 00:32:17.990 align:middle line:84% are all kind of difference and difference regressions 00:32:17.990 --> 00:32:19.940 align:middle line:90% where essentially this is-- 00:32:19.940 --> 00:32:22.730 align:middle line:84% so this is just in the cross section. 00:32:22.730 --> 00:32:26.090 align:middle line:84% When we measure people's cash expenditures, 00:32:26.090 --> 00:32:31.520 align:middle line:84% what do people spend their money on right following 00:32:31.520 --> 00:32:34.325 align:middle line:90% the early cash infusion? 00:32:34.325 --> 00:32:36.200 align:middle line:84% People essentially spend quite a bit of money 00:32:36.200 --> 00:32:38.760 align:middle line:90% to repay their loans. 00:32:38.760 --> 00:32:41.720 align:middle line:84% They also increase their household expenditures. 00:32:41.720 --> 00:32:44.510 align:middle line:84% And a lot of this happens essentially 00:32:44.510 --> 00:32:46.910 align:middle line:84% on the day of payment or the day after. 00:32:46.910 --> 00:32:49.430 align:middle line:84% So essentially, immediately after people get cash, 00:32:49.430 --> 00:32:54.860 align:middle line:84% they spend that money on A, paying back credit and loans, 00:32:54.860 --> 00:32:58.970 align:middle line:84% and B, household expenditure, household necessities 00:32:58.970 --> 00:32:59.630 align:middle line:90% if you want. 00:32:59.630 --> 00:33:00.950 align:middle line:84% So that is essentially saying that people 00:33:00.950 --> 00:33:03.380 align:middle line:84% are credit constrained and actually having cash on hand 00:33:03.380 --> 00:33:06.860 align:middle line:84% makes a big difference in people's behavior. 00:33:06.860 --> 00:33:10.400 align:middle line:84% So and that's-- then you know at least reason to believe that 00:33:10.400 --> 00:33:13.490 align:middle line:84% the cash infusion in fact meaningful and reduced 00:33:13.490 --> 00:33:17.300 align:middle line:84% financial stress because people are now able to make these 00:33:17.300 --> 00:33:19.200 align:middle line:90% types of expenses. 00:33:19.200 --> 00:33:22.610 align:middle line:84% Second, we find relatively large effects 00:33:22.610 --> 00:33:23.750 align:middle line:90% on people's productivity. 00:33:23.750 --> 00:33:26.000 align:middle line:84% So now these are difference in differences regression. 00:33:26.000 --> 00:33:27.860 align:middle line:84% The previous one was a cross-sectional regression, 00:33:27.860 --> 00:33:29.510 align:middle line:84% because we have only one point in time. 00:33:29.510 --> 00:33:31.850 align:middle line:84% This one is now cash, which is whether we 00:33:31.850 --> 00:33:35.270 align:middle line:84% received cash and post pay, whether you're in the post pay 00:33:35.270 --> 00:33:36.050 align:middle line:90% period. 00:33:36.050 --> 00:33:41.120 align:middle line:84% And we find that people are more productive about 0.1, 00:33:41.120 --> 00:33:44.960 align:middle line:84% 0.12 standard deviations, which is about 6 to 13%, 00:33:44.960 --> 00:33:48.120 align:middle line:90% a 6% increases in productivity. 00:33:48.120 --> 00:33:53.030 align:middle line:84% These effects are almost entirely 00:33:53.030 --> 00:33:57.080 align:middle line:84% confined or concentrated among people 00:33:57.080 --> 00:33:58.940 align:middle line:84% with high financial strain, so the people 00:33:58.940 --> 00:34:01.490 align:middle line:84% who exhibit the highest financial strain 00:34:01.490 --> 00:34:04.250 align:middle line:84% at the beginning before the treatment at baseline, those 00:34:04.250 --> 00:34:07.340 align:middle line:84% have the highest treatment effects. 00:34:07.340 --> 00:34:09.685 align:middle line:84% The effects don't seem to be going away immediately, 00:34:09.685 --> 00:34:11.060 align:middle line:84% so this is what you can see here. 00:34:11.060 --> 00:34:14.540 align:middle line:84% We split the sample into lower and higher financial strain, 00:34:14.540 --> 00:34:16.130 align:middle line:84% and what you see in blue is the people 00:34:16.130 --> 00:34:17.929 align:middle line:84% with higher financial strain who become 00:34:17.929 --> 00:34:20.480 align:middle line:84% more productive upon receiving money. 00:34:20.480 --> 00:34:23.300 align:middle line:84% These effects don't go away after like a day or two, 00:34:23.300 --> 00:34:27.330 align:middle line:84% in fact, if anything, they get a little bit larger over time. 00:34:27.330 --> 00:34:30.440 align:middle line:84% And that's to say that what you would expect in like gift 00:34:30.440 --> 00:34:35.870 align:middle line:84% exchange or other types of models of reciprocity, 00:34:35.870 --> 00:34:37.250 align:middle line:90% you might think that people-- 00:34:37.250 --> 00:34:39.500 align:middle line:84% what have people found in these types of experiments, 00:34:39.500 --> 00:34:41.167 align:middle line:84% that there's a short run effect but that 00:34:41.167 --> 00:34:42.270 align:middle line:90% goes away very quickly. 00:34:42.270 --> 00:34:46.305 align:middle line:84% But in fact, these effects seem to be persistent over time. 00:34:46.305 --> 00:34:46.805 align:middle line:90% Yeah? 00:34:46.805 --> 00:34:48.430 align:middle line:84% AUDIENCE: Wouldn't the treatment group, 00:34:48.430 --> 00:34:50.873 align:middle line:84% [? part ?] of the productivity after the [INAUDIBLE] 00:34:50.873 --> 00:34:55.235 align:middle line:84% compare to productivity a day after the first payment? 00:34:55.235 --> 00:34:56.860 align:middle line:84% FRANK SCHILBACH: Sorry, say that again. 00:34:56.860 --> 00:34:58.402 align:middle line:84% AUDIENCE: Within the treatment group, 00:34:58.402 --> 00:35:00.810 align:middle line:84% how does productivity a day after the treatment compared 00:35:00.810 --> 00:35:05.470 align:middle line:84% to a day after the first payment? 00:35:05.470 --> 00:35:06.580 align:middle line:90% FRANK SCHILBACH: I see. 00:35:06.580 --> 00:35:10.150 align:middle line:84% It's hard to make that comparison, 00:35:10.150 --> 00:35:12.170 align:middle line:84% because that's very strong learning. 00:35:12.170 --> 00:35:15.665 align:middle line:84% So the first payment essentially means day one? 00:35:15.665 --> 00:35:16.290 align:middle line:90% AUDIENCE: Yeah. 00:35:16.290 --> 00:35:18.248 align:middle line:84% FRANK SCHILBACH: So day one, essentially people 00:35:18.248 --> 00:35:19.770 align:middle line:90% are essentially learning-- 00:35:19.770 --> 00:35:21.630 align:middle line:84% people on day one-- and they want 00:35:21.630 --> 00:35:24.270 align:middle line:84% to actually not incentivize, because we're just saying like, 00:35:24.270 --> 00:35:26.645 align:middle line:84% this is a training period galore and you get a fixed wage 00:35:26.645 --> 00:35:28.230 align:middle line:90% of like 250 rupees. 00:35:28.230 --> 00:35:30.605 align:middle line:84% So it's very, very strong learning trends 00:35:30.605 --> 00:35:32.730 align:middle line:84% during the first three days, which is partially why 00:35:32.730 --> 00:35:34.335 align:middle line:90% we have this baseline period. 00:35:34.335 --> 00:35:36.960 align:middle line:84% So it's hard to, and of course, the treatment and control group 00:35:36.960 --> 00:35:37.560 align:middle line:90% gets money. 00:35:37.560 --> 00:35:39.435 align:middle line:84% I don't actually know what the difference is, 00:35:39.435 --> 00:35:42.840 align:middle line:84% but the trouble is that you have to disentangle learning 00:35:42.840 --> 00:35:45.120 align:middle line:90% from treatment effects. 00:35:45.120 --> 00:35:47.250 align:middle line:84% And the learning has flattened off 00:35:47.250 --> 00:35:50.430 align:middle line:84% at the time when we're giving people's cash by design. 00:35:50.430 --> 00:35:54.960 align:middle line:90% 00:35:54.960 --> 00:35:55.620 align:middle line:90% OK. 00:35:55.620 --> 00:35:59.610 align:middle line:84% So then, as I told you, so here are some productivity effects. 00:35:59.610 --> 00:36:01.800 align:middle line:84% Now, what we also do in the paper is 00:36:01.800 --> 00:36:04.830 align:middle line:84% we try to shed some light on the underlying mechanism trying 00:36:04.830 --> 00:36:07.020 align:middle line:84% to say, OK, is this really about attention 00:36:07.020 --> 00:36:09.570 align:middle line:84% or is it about something else, perhaps effort people just 00:36:09.570 --> 00:36:11.070 align:middle line:84% try harder because they have cash, 00:36:11.070 --> 00:36:12.528 align:middle line:84% and maybe they're happier, and that 00:36:12.528 --> 00:36:14.950 align:middle line:90% leads them to work really hard. 00:36:14.950 --> 00:36:18.030 align:middle line:84% So this is what the leaf plates look like. 00:36:18.030 --> 00:36:20.580 align:middle line:84% And they're actually very hard to make well. 00:36:20.580 --> 00:36:27.128 align:middle line:84% You have to try to make a round shape. 00:36:27.128 --> 00:36:28.920 align:middle line:84% There shouldn't be any holes because people 00:36:28.920 --> 00:36:30.810 align:middle line:84% eat curry and having holes in your plate 00:36:30.810 --> 00:36:32.980 align:middle line:90% is a really bad idea. 00:36:32.980 --> 00:36:34.470 align:middle line:90% And so this is not. 00:36:34.470 --> 00:36:35.530 align:middle line:90% I tried this myself. 00:36:35.530 --> 00:36:37.072 align:middle line:84% This is not what my plates look like. 00:36:37.072 --> 00:36:39.640 align:middle line:84% This is actually hard to do well and hard to do fast. 00:36:39.640 --> 00:36:42.210 align:middle line:84% And so in particular, you need some planning 00:36:42.210 --> 00:36:44.850 align:middle line:84% and strategic placement to reduce 00:36:44.850 --> 00:36:47.140 align:middle line:84% the amount of work or the effort you have to put in. 00:36:47.140 --> 00:36:48.780 align:middle line:84% So when you plan well how to do this, 00:36:48.780 --> 00:36:51.810 align:middle line:84% if you go in the wrong direction and you make extra work 00:36:51.810 --> 00:36:57.390 align:middle line:84% and you have to undo your work if you do it really badly, 00:36:57.390 --> 00:36:59.820 align:middle line:84% you use like 20 leaves instead of 12 leaves, 00:36:59.820 --> 00:37:02.132 align:middle line:84% and that costs you a lot of time and effort 00:37:02.132 --> 00:37:04.590 align:middle line:84% because you have to actually put all these leaves together. 00:37:04.590 --> 00:37:07.620 align:middle line:84% If you place them well and do it carefully, 00:37:07.620 --> 00:37:10.410 align:middle line:84% then you have essentially less physical effort and time 00:37:10.410 --> 00:37:12.430 align:middle line:90% for the same piece weight. 00:37:12.430 --> 00:37:15.990 align:middle line:84% And so we measured essentially three markers of attentiveness 00:37:15.990 --> 00:37:18.780 align:middle line:84% and production, which is the number of double holes, which 00:37:18.780 --> 00:37:20.430 align:middle line:84% is you can see here these little sticks 00:37:20.430 --> 00:37:24.090 align:middle line:84% where people make these sticks through the leaves 00:37:24.090 --> 00:37:25.480 align:middle line:90% to put them together. 00:37:25.480 --> 00:37:29.802 align:middle line:84% And if you make a wrong stitch, then you have to undo it, 00:37:29.802 --> 00:37:31.260 align:middle line:84% and then there's like a double hole 00:37:31.260 --> 00:37:33.990 align:middle line:84% that you can identify in the leaves. 00:37:33.990 --> 00:37:39.150 align:middle line:84% We measure these pieces, the measures of attentiveness. 00:37:39.150 --> 00:37:40.950 align:middle line:84% We take people's production and then 00:37:40.950 --> 00:37:44.448 align:middle line:84% measure it unbeknownst to workers, 00:37:44.448 --> 00:37:46.740 align:middle line:84% so workers don't even know that we're collecting these. 00:37:46.740 --> 00:37:48.750 align:middle line:90% So we measure the double holes. 00:37:48.750 --> 00:37:50.640 align:middle line:84% We measure the number of leaves, the number 00:37:50.640 --> 00:37:52.290 align:middle line:90% of stitches per plate. 00:37:52.290 --> 00:37:54.462 align:middle line:84% So you have to have a minimum size of the plate, 00:37:54.462 --> 00:37:56.670 align:middle line:84% but conditional on that, essentially, you want to use 00:37:56.670 --> 00:37:59.290 align:middle line:90% as few leaves as possible. 00:37:59.290 --> 00:38:01.945 align:middle line:90% And so these measures-- 00:38:01.945 --> 00:38:03.570 align:middle line:84% these measures are essentially measures 00:38:03.570 --> 00:38:12.510 align:middle line:84% of attentiveness that tell us how attentively where people 00:38:12.510 --> 00:38:14.920 align:middle line:90% are working during that time. 00:38:14.920 --> 00:38:18.840 align:middle line:84% And so when you then collect these measures, what we find 00:38:18.840 --> 00:38:21.870 align:middle line:84% is that not only does the treatment 00:38:21.870 --> 00:38:23.910 align:middle line:84% affect people's productivity, how fast 00:38:23.910 --> 00:38:26.070 align:middle line:84% or how much people produce overall. 00:38:26.070 --> 00:38:28.140 align:middle line:84% But they also affect these measures 00:38:28.140 --> 00:38:30.090 align:middle line:84% of attentiveness, this attentiveness index 00:38:30.090 --> 00:38:33.000 align:middle line:84% or like high attentiveness which tells us 00:38:33.000 --> 00:38:37.470 align:middle line:84% that people work faster, but also with fewer mistakes. 00:38:37.470 --> 00:38:40.260 align:middle line:84% So essentially, higher attentiveness, 00:38:40.260 --> 00:38:45.060 align:middle line:84% lower number of mistakes is at least one channel that 00:38:45.060 --> 00:38:46.870 align:middle line:90% leads to higher productivity. 00:38:46.870 --> 00:38:48.420 align:middle line:84% Now, this is consistent as I said, 00:38:48.420 --> 00:38:51.660 align:middle line:84% with a range of potential psychological mechanisms, 00:38:51.660 --> 00:38:53.880 align:middle line:84% stress, worries, et cetera that we can't necessarily 00:38:53.880 --> 00:38:54.810 align:middle line:90% disentangle. 00:38:54.810 --> 00:38:57.690 align:middle line:84% But we know that at least attention or attentiveness 00:38:57.690 --> 00:39:00.900 align:middle line:84% plays some role in what's going on. 00:39:00.900 --> 00:39:04.557 align:middle line:84% These effects also persist until the end of the contract period. 00:39:04.557 --> 00:39:07.140 align:middle line:84% Now, in the paper, and I don't want to get too much into this, 00:39:07.140 --> 00:39:09.780 align:middle line:84% we spend a lot of time trying to disentangle 00:39:09.780 --> 00:39:12.060 align:middle line:84% a test for other mechanisms, in particular, 00:39:12.060 --> 00:39:14.370 align:middle line:84% things like fairness and gift exchange. 00:39:14.370 --> 00:39:15.960 align:middle line:84% This is why we have some variation 00:39:15.960 --> 00:39:18.030 align:middle line:84% on like when exactly people are paid. 00:39:18.030 --> 00:39:21.900 align:middle line:84% We have also some variation the announcement periods 00:39:21.900 --> 00:39:22.558 align:middle line:90% and the like. 00:39:22.558 --> 00:39:24.600 align:middle line:84% There's a lot of discussion of that in the paper, 00:39:24.600 --> 00:39:26.430 align:middle line:84% specifically that we trying to deal with. 00:39:26.430 --> 00:39:28.350 align:middle line:90% That's like a bit of-- 00:39:28.350 --> 00:39:32.014 align:middle line:90% 00:39:32.014 --> 00:39:37.520 align:middle line:84% there's a careful test for these types of effects 00:39:37.520 --> 00:39:39.620 align:middle line:84% that we want to make sure they're not going on. 00:39:39.620 --> 00:39:42.980 align:middle line:84% In addition, we also try to rule out investment channels 00:39:42.980 --> 00:39:44.090 align:middle line:90% such as nutrition. 00:39:44.090 --> 00:39:46.580 align:middle line:84% So while nutrition effects would be quite interesting, 00:39:46.580 --> 00:39:48.350 align:middle line:84% that's not the focus of this paper. 00:39:48.350 --> 00:39:50.820 align:middle line:84% We find some effects on like when people are paid. 00:39:50.820 --> 00:39:53.385 align:middle line:84% In fact, they do purchase more food. 00:39:53.385 --> 00:39:55.010 align:middle line:84% But that's mostly like, I think, things 00:39:55.010 --> 00:39:56.720 align:middle line:84% like food for the family or the like. 00:39:56.720 --> 00:39:59.480 align:middle line:84% We don't see, for example, that breakfast consumption actually 00:39:59.480 --> 00:40:01.400 align:middle line:84% changes across treatment and control 00:40:01.400 --> 00:40:03.930 align:middle line:90% group following the treatment. 00:40:03.930 --> 00:40:06.200 align:middle line:90% So what do we learn from this? 00:40:06.200 --> 00:40:09.800 align:middle line:84% So we essentially learned that using the payment timing 00:40:09.800 --> 00:40:14.300 align:middle line:84% within context of liquidity constrained workers 00:40:14.300 --> 00:40:16.280 align:middle line:84% can help us construct a clean test 00:40:16.280 --> 00:40:19.880 align:middle line:84% of whether financial strain affects people's productivity. 00:40:19.880 --> 00:40:21.680 align:middle line:84% And we find essentially, that the ability 00:40:21.680 --> 00:40:25.310 align:middle line:84% to pay for expenses at home boosts productivity 00:40:25.310 --> 00:40:27.200 align:middle line:90% and attentiveness at work. 00:40:27.200 --> 00:40:30.770 align:middle line:84% So financial strain itself then can be detrimental for earnings 00:40:30.770 --> 00:40:32.550 align:middle line:84% exactly when money is most needed. 00:40:32.550 --> 00:40:34.490 align:middle line:84% So exactly when people don't have money, 00:40:34.490 --> 00:40:36.230 align:middle line:84% they might be not able to or that 00:40:36.230 --> 00:40:40.460 align:middle line:84% might interfere with their ability to work productively. 00:40:40.460 --> 00:40:42.680 align:middle line:84% And that in some ways, that's to be very clear, 00:40:42.680 --> 00:40:44.510 align:middle line:84% this is not a policy evaluation, this 00:40:44.510 --> 00:40:46.520 align:middle line:84% is not a cash transfer program or the like. 00:40:46.520 --> 00:40:49.700 align:middle line:84% It's rather a proof of concept that productivity effects can 00:40:49.700 --> 00:40:53.750 align:middle line:84% occur in high stakes setting such as for labor income, 00:40:53.750 --> 00:40:56.907 align:middle line:84% affecting labor income, and some potential implication 00:40:56.907 --> 00:40:59.240 align:middle line:84% is there might be additional channels through its policy 00:40:59.240 --> 00:41:01.850 align:middle line:84% level such as pay structure or social programs could 00:41:01.850 --> 00:41:03.600 align:middle line:90% affect productivity. 00:41:03.600 --> 00:41:06.410 align:middle line:84% So and in particular, cash transfer programs 00:41:06.410 --> 00:41:09.500 align:middle line:84% could also have productivity benefits. 00:41:09.500 --> 00:41:11.270 align:middle line:84% Now there's all sorts of open questions 00:41:11.270 --> 00:41:13.220 align:middle line:84% here then, in particular, you want to know, 00:41:13.220 --> 00:41:15.080 align:middle line:84% are such effects present in other settings, 00:41:15.080 --> 00:41:18.350 align:middle line:84% for example, cash and other transfers programs? 00:41:18.350 --> 00:41:21.590 align:middle line:84% Abhijit Banerjee and coauthors have some evidence of this, 00:41:21.590 --> 00:41:22.610 align:middle line:90% I believe, in Ghana. 00:41:22.610 --> 00:41:25.040 align:middle line:84% But they find some effects on labor supply 00:41:25.040 --> 00:41:28.195 align:middle line:84% and some effects on productivity as well. 00:41:28.195 --> 00:41:29.570 align:middle line:84% But I think we need more evidence 00:41:29.570 --> 00:41:32.000 align:middle line:84% to test for these types of effects. 00:41:32.000 --> 00:41:35.370 align:middle line:84% Then a natural question to say is, what is the optimal pay 00:41:35.370 --> 00:41:35.870 align:middle line:90% structure? 00:41:35.870 --> 00:41:38.540 align:middle line:84% If you're an employer, how should you pay workers? 00:41:38.540 --> 00:41:41.090 align:middle line:84% Should you pay them weekly or daily or monthly? 00:41:41.090 --> 00:41:43.550 align:middle line:84% We can't really test for this because we have only one 00:41:43.550 --> 00:41:45.987 align:middle line:84% payment that's being done, so we can't say, 00:41:45.987 --> 00:41:47.570 align:middle line:84% now you should pay people differently, 00:41:47.570 --> 00:41:50.030 align:middle line:84% because of course, the control group gets paid more 00:41:50.030 --> 00:41:51.540 align:middle line:90% at the end of the experiment. 00:41:51.540 --> 00:41:54.230 align:middle line:84% So we can't really say anything about optimal policy, 00:41:54.230 --> 00:41:56.720 align:middle line:84% but trying to understand whether certain programs 00:41:56.720 --> 00:42:00.650 align:middle line:84% or interventions to improve employees' financial well-being 00:42:00.650 --> 00:42:03.140 align:middle line:84% has productivity or other types of effects 00:42:03.140 --> 00:42:05.900 align:middle line:84% for worker well-being and their retention 00:42:05.900 --> 00:42:11.630 align:middle line:84% et cetera seems really important and a natural thing to do next. 00:42:11.630 --> 00:42:12.820 align:middle line:90% Any questions on this paper? 00:42:12.820 --> 00:42:22.530 align:middle line:90% 00:42:22.530 --> 00:42:23.220 align:middle line:90% OK. 00:42:23.220 --> 00:42:25.990 align:middle line:84% So then let me tell you about sleep. 00:42:25.990 --> 00:42:28.405 align:middle line:84% I know, I assume that you have all read this paper, 00:42:28.405 --> 00:42:30.030 align:middle line:84% so I'm not going to spend a lot of time 00:42:30.030 --> 00:42:35.230 align:middle line:84% talking about the design and the specific of the experiment. 00:42:35.230 --> 00:42:38.580 align:middle line:84% But rather about the motivation of why is this a-- 00:42:38.580 --> 00:42:40.900 align:middle line:84% why did we start to study this in the first place 00:42:40.900 --> 00:42:44.500 align:middle line:84% and why is it an important perhaps thing to engage in? 00:42:44.500 --> 00:42:48.780 align:middle line:84% So first-- at a very basic level of sleep, 00:42:48.780 --> 00:42:50.970 align:middle line:84% we spend a lot of time sleeping in our lives. 00:42:50.970 --> 00:42:54.870 align:middle line:84% It's about a quarter to a third of our lives, we're asleep. 00:42:54.870 --> 00:42:58.770 align:middle line:84% And there's a huge body of sleep medicine 00:42:58.770 --> 00:43:01.193 align:middle line:84% that believes that sleep is really important 00:43:01.193 --> 00:43:02.610 align:middle line:84% or that shows that sleep is really 00:43:02.610 --> 00:43:06.210 align:middle line:84% important for a range of outcomes, including 00:43:06.210 --> 00:43:09.720 align:middle line:84% cognition, including happiness, perhaps decision making, health 00:43:09.720 --> 00:43:12.870 align:middle line:90% wages, et cetera. 00:43:12.870 --> 00:43:16.470 align:middle line:84% Now, a lot of this work is essentially 00:43:16.470 --> 00:43:18.960 align:middle line:84% done in the US and rich countries, and a lot of it 00:43:18.960 --> 00:43:20.530 align:middle line:90% is coming from lab setting. 00:43:20.530 --> 00:43:24.090 align:middle line:84% So in particular, most of the evidence is here on cognition. 00:43:24.090 --> 00:43:26.970 align:middle line:84% In particular, there's thousands of experiments 00:43:26.970 --> 00:43:29.190 align:middle line:84% that look at what happens to cognitive function 00:43:29.190 --> 00:43:31.140 align:middle line:90% when you sleep well versus not. 00:43:31.140 --> 00:43:32.730 align:middle line:84% Almost all of that is essentially 00:43:32.730 --> 00:43:36.100 align:middle line:84% in rich country settings and coming from the lab. 00:43:36.100 --> 00:43:38.160 align:middle line:84% And then there's a smaller literature 00:43:38.160 --> 00:43:40.950 align:middle line:84% on things like happiness and subjective well-being, decision 00:43:40.950 --> 00:43:43.230 align:middle line:84% making, health, and a little bit of work 00:43:43.230 --> 00:43:46.380 align:middle line:90% recently on wages, et cetera. 00:43:46.380 --> 00:43:50.520 align:middle line:84% Now, when you read books such as the book on Walker 00:43:50.520 --> 00:43:53.340 align:middle line:84% on Why We Sleep, you get the impression 00:43:53.340 --> 00:43:55.530 align:middle line:84% that if only we slept-- everybody slept an hour more 00:43:55.530 --> 00:43:58.080 align:middle line:84% in our lives, that would have transformative effects 00:43:58.080 --> 00:44:00.700 align:middle line:84% on society, people would be happier, healthier, 00:44:00.700 --> 00:44:03.150 align:middle line:84% safer, smarter, and so on and so forth. 00:44:03.150 --> 00:44:09.300 align:middle line:84% And investing in sleep is the biggest thing we could just do. 00:44:09.300 --> 00:44:11.950 align:middle line:84% So now, a natural thing is to say, well, 00:44:11.950 --> 00:44:13.450 align:middle line:84% can we test this in field setting, 00:44:13.450 --> 00:44:16.620 align:middle line:84% not just in India, but in any type of settings? 00:44:16.620 --> 00:44:19.500 align:middle line:84% And there's very little field evidence overall, 00:44:19.500 --> 00:44:21.870 align:middle line:84% and we should just try to understand this is actually 00:44:21.870 --> 00:44:24.750 align:middle line:84% true in a sense of a lot of the claims, 00:44:24.750 --> 00:44:27.210 align:middle line:84% in particular when it comes to productivity, et cetera, 00:44:27.210 --> 00:44:29.650 align:middle line:84% are based on correlational evidence. 00:44:29.650 --> 00:44:33.130 align:middle line:84% So now can we test for causal evidence? 00:44:33.130 --> 00:44:36.840 align:middle line:84% Now, if you then look at sleep settings in rich versus 00:44:36.840 --> 00:44:39.420 align:middle line:84% poor countries and there Gapminder Dollar Street 00:44:39.420 --> 00:44:42.060 align:middle line:84% is a beautiful website, but you can have a window 00:44:42.060 --> 00:44:44.010 align:middle line:90% into people's lives. 00:44:44.010 --> 00:44:46.680 align:middle line:84% If you look at, compare people's sleep conditions, 00:44:46.680 --> 00:44:49.560 align:middle line:84% such as here the Howard family or the Shah family. 00:44:49.560 --> 00:44:51.960 align:middle line:84% Howard family in Colorado, and the Shah family 00:44:51.960 --> 00:44:56.010 align:middle line:84% in Calcutta, you'll easily see and if you visit people's homes 00:44:56.010 --> 00:44:59.280 align:middle line:84% and in particular in urban developing countries 00:44:59.280 --> 00:45:01.860 align:middle line:84% settings that sleeping conditions are dramatically 00:45:01.860 --> 00:45:04.230 align:middle line:84% different for low income populations 00:45:04.230 --> 00:45:09.360 align:middle line:84% compared to the US or other rich settings. 00:45:09.360 --> 00:45:11.272 align:middle line:84% And we know essentially next to nothing 00:45:11.272 --> 00:45:13.230 align:middle line:84% about sleep in developing country environments. 00:45:13.230 --> 00:45:17.243 align:middle line:84% We have essentially no idea how people sleep in most countries, 00:45:17.243 --> 00:45:18.660 align:middle line:84% how much they sleep, how well they 00:45:18.660 --> 00:45:22.140 align:middle line:84% sleep, why they don't sleep well, et cetera. 00:45:22.140 --> 00:45:26.160 align:middle line:84% Now, one perhaps obvious observation 00:45:26.160 --> 00:45:28.620 align:middle line:84% is while sleep might be worse given the difficult sleeping 00:45:28.620 --> 00:45:32.490 align:middle line:84% conditions, the two potentially competing forces. 00:45:32.490 --> 00:45:35.070 align:middle line:84% One is people might be sleeping less and the effects of sleep 00:45:35.070 --> 00:45:37.575 align:middle line:84% might be concave, so and there's quite a bit of evidence 00:45:37.575 --> 00:45:38.700 align:middle line:90% that seems to suggest that. 00:45:38.700 --> 00:45:41.108 align:middle line:84% If you sleep only three hours, the marginal benefit 00:45:41.108 --> 00:45:42.900 align:middle line:84% of another hour is larger than if you sleep 00:45:42.900 --> 00:45:44.850 align:middle line:90% nine hours or eight hours. 00:45:44.850 --> 00:45:48.420 align:middle line:84% But on the other hand, if sleep is so disrupted, 00:45:48.420 --> 00:45:51.030 align:middle line:84% the marginal effects or any effects of sleep 00:45:51.030 --> 00:45:52.950 align:middle line:84% might be just not very beneficial 00:45:52.950 --> 00:45:56.640 align:middle line:84% because you're just not getting deep sleep or any full sleep 00:45:56.640 --> 00:45:58.080 align:middle line:90% cycles. 00:45:58.080 --> 00:46:00.960 align:middle line:84% And so, but that's an entirely open question. 00:46:00.960 --> 00:46:04.080 align:middle line:84% Our thought initially was to say, 00:46:04.080 --> 00:46:06.630 align:middle line:84% and a lot of the consideration about sleep quality 00:46:06.630 --> 00:46:09.930 align:middle line:84% that you may have read in the paper are more exposed. 00:46:09.930 --> 00:46:12.150 align:middle line:84% We came into this and saying, look people 00:46:12.150 --> 00:46:14.280 align:middle line:84% seem to be sleeping only five hours. 00:46:14.280 --> 00:46:16.500 align:middle line:84% Couldn't this be one of those factors 00:46:16.500 --> 00:46:18.823 align:middle line:84% that I talked about on Monday, some of the high returns 00:46:18.823 --> 00:46:20.490 align:middle line:84% that people are not taking advantage of? 00:46:20.490 --> 00:46:24.690 align:middle line:84% It looks like this is really like potentially a very cheap 00:46:24.690 --> 00:46:26.863 align:middle line:84% policy to try to increase people's sleep, 00:46:26.863 --> 00:46:28.530 align:middle line:84% and that could have really large effects 00:46:28.530 --> 00:46:30.150 align:middle line:84% because there's all this evidence from the sleep 00:46:30.150 --> 00:46:31.740 align:middle line:90% literature that suggests that. 00:46:31.740 --> 00:46:34.680 align:middle line:84% That was the starting point of this paper. 00:46:34.680 --> 00:46:37.650 align:middle line:90% Now, what did we actually do? 00:46:37.650 --> 00:46:40.110 align:middle line:84% A good share of this paper is, in fact, descriptive. 00:46:40.110 --> 00:46:42.120 align:middle line:84% It's essentially just saying like we're 00:46:42.120 --> 00:46:43.260 align:middle line:90% providing some evidence. 00:46:43.260 --> 00:46:45.180 align:middle line:84% Among the first evidence from India 00:46:45.180 --> 00:46:47.640 align:middle line:84% or other similar countries, how much people 00:46:47.640 --> 00:46:48.850 align:middle line:90% are objectively sleeping. 00:46:48.850 --> 00:46:53.170 align:middle line:84% This is because sleep self-reports are pretty bad. 00:46:53.170 --> 00:46:56.760 align:middle line:84% And then we have 450 data entry workers 00:46:56.760 --> 00:46:59.190 align:middle line:84% or people who are hired to do data entry work, 00:46:59.190 --> 00:47:01.020 align:middle line:84% and randomized them into interventions 00:47:01.020 --> 00:47:04.770 align:middle line:84% to increase their sleep and measure 00:47:04.770 --> 00:47:07.480 align:middle line:90% impact on a range of outcomes. 00:47:07.480 --> 00:47:08.820 align:middle line:90% Now, why is this possible? 00:47:08.820 --> 00:47:11.820 align:middle line:84% There's lots of progress in the last decades on sleep 00:47:11.820 --> 00:47:14.770 align:middle line:84% measurement technologies and those have only gotten better. 00:47:14.770 --> 00:47:16.235 align:middle line:84% So these are Actigraphs, and what 00:47:16.235 --> 00:47:17.610 align:middle line:84% they look like, which essentially 00:47:17.610 --> 00:47:19.985 align:middle line:84% is like a wristwatch, you can think of this like an Apple 00:47:19.985 --> 00:47:23.220 align:middle line:84% Watch or like a Fitbit or the like, 00:47:23.220 --> 00:47:25.860 align:middle line:84% that measures sleep through motion. 00:47:25.860 --> 00:47:27.630 align:middle line:84% These devices have gotten really good. 00:47:27.630 --> 00:47:29.903 align:middle line:84% So they can actually measure sleep pretty well. 00:47:29.903 --> 00:47:32.070 align:middle line:84% There are some question of how well do they actually 00:47:32.070 --> 00:47:34.103 align:middle line:84% work in settings such as ours, but I 00:47:34.103 --> 00:47:35.520 align:middle line:84% think overall, essentially, that's 00:47:35.520 --> 00:47:38.310 align:middle line:84% to say we can run field experiments that meaningfully 00:47:38.310 --> 00:47:42.240 align:middle line:84% measure sleep and we can measure like first stage on sleep 00:47:42.240 --> 00:47:43.175 align:middle line:90% pretty well. 00:47:43.175 --> 00:47:44.550 align:middle line:84% And given that so many people now 00:47:44.550 --> 00:47:46.080 align:middle line:84% have these types of devices, it's 00:47:46.080 --> 00:47:48.090 align:middle line:84% relatively easy to actually enroll people now 00:47:48.090 --> 00:47:48.923 align:middle line:90% in any experiment. 00:47:48.923 --> 00:47:50.340 align:middle line:84% Even if you did that on mTurk, you 00:47:50.340 --> 00:47:52.080 align:middle line:84% could do that probably and find people 00:47:52.080 --> 00:47:55.110 align:middle line:84% that have Apple Watches or Fitbits or the like 00:47:55.110 --> 00:47:58.412 align:middle line:90% and run experiments with that. 00:47:58.412 --> 00:48:00.870 align:middle line:84% Importantly, they can measure a sleep quantity pretty well. 00:48:00.870 --> 00:48:02.510 align:middle line:84% They also can measure sleep efficiency 00:48:02.510 --> 00:48:04.260 align:middle line:84% pretty well, which is the fraction of time 00:48:04.260 --> 00:48:06.220 align:middle line:84% that you're asleep when you're in bed. 00:48:06.220 --> 00:48:09.060 align:middle line:84% They don't measure sleep quality that well 00:48:09.060 --> 00:48:11.580 align:middle line:84% in terms of deep sleep, et cetera. 00:48:11.580 --> 00:48:13.770 align:middle line:84% I think there's progress has been made recently, 00:48:13.770 --> 00:48:17.460 align:middle line:84% but it's not great, but for our purposes, I guess, 00:48:17.460 --> 00:48:18.870 align:middle line:90% that's good enough. 00:48:18.870 --> 00:48:22.140 align:middle line:84% Now, and as I said, a lot of what we're doing 00:48:22.140 --> 00:48:24.220 align:middle line:90% is descriptive. 00:48:24.220 --> 00:48:26.748 align:middle line:84% The first finding in fact is just how much do people sleep. 00:48:26.748 --> 00:48:28.290 align:middle line:84% And what we find is that well, if you 00:48:28.290 --> 00:48:30.580 align:middle line:84% look at self-reported sleep, things 00:48:30.580 --> 00:48:31.630 align:middle line:90% look actually not so bad. 00:48:31.630 --> 00:48:33.370 align:middle line:90% People sleep about 7.2 hours. 00:48:33.370 --> 00:48:36.400 align:middle line:84% That's actually quite similar to the US. 00:48:36.400 --> 00:48:37.870 align:middle line:90% But once you look at-- 00:48:37.870 --> 00:48:39.340 align:middle line:84% objectively measure sleep, people 00:48:39.340 --> 00:48:42.730 align:middle line:84% sleep about 5.5 hours in our study. 00:48:42.730 --> 00:48:44.650 align:middle line:84% Lots of people, and these are averages 00:48:44.650 --> 00:48:48.550 align:middle line:84% over a week at baseline, lots of people 00:48:48.550 --> 00:48:50.518 align:middle line:90% see 5 or 6 hours on average. 00:48:50.518 --> 00:48:52.810 align:middle line:84% So it's not like they have a bad night once in a while. 00:48:52.810 --> 00:48:55.090 align:middle line:84% It's like, on average, the typical night of sleep 00:48:55.090 --> 00:48:58.580 align:middle line:90% is about 5 or 6 hours. 00:48:58.580 --> 00:49:00.900 align:middle line:84% And this seems to be also broadly representative. 00:49:00.900 --> 00:49:03.350 align:middle line:84% So you have a little bit of a broader study with a broader 00:49:03.350 --> 00:49:06.390 align:middle line:84% sample, but we find similar results. 00:49:06.390 --> 00:49:07.640 align:middle line:90% Now, it's not-- one second-- 00:49:07.640 --> 00:49:10.375 align:middle line:84% that's not because people don't try to sleep. 00:49:10.375 --> 00:49:11.750 align:middle line:84% People spend lots of time in bed, 00:49:11.750 --> 00:49:13.958 align:middle line:84% about eight hours per night, which is, in fact, quite 00:49:13.958 --> 00:49:15.620 align:middle line:90% similar to the US. 00:49:15.620 --> 00:49:17.690 align:middle line:84% But what seems to be the case is that people 00:49:17.690 --> 00:49:19.730 align:middle line:90% sleep efficiency is really low. 00:49:19.730 --> 00:49:22.610 align:middle line:84% That's to say, in the US, when you measure 00:49:22.610 --> 00:49:25.640 align:middle line:84% what fraction of time are people asleep when they're in bed, 00:49:25.640 --> 00:49:28.490 align:middle line:84% well usually it's about 85%, 90%, 95%. 00:49:28.490 --> 00:49:30.320 align:middle line:84% So if you are below 85%, people would 00:49:30.320 --> 00:49:32.090 align:middle line:84% say you're a problematic sleeper and you 00:49:32.090 --> 00:49:33.930 align:middle line:90% should get some treatment. 00:49:33.930 --> 00:49:36.390 align:middle line:90% Here we have 70% on average. 00:49:36.390 --> 00:49:38.870 align:middle line:84% And this is as bad as people with sleep apnea 00:49:38.870 --> 00:49:41.330 align:middle line:84% or elderly populations who have really bad sleep. 00:49:41.330 --> 00:49:43.175 align:middle line:84% This is the average sleep that we find 00:49:43.175 --> 00:49:44.300 align:middle line:90% in these types of settings. 00:49:44.300 --> 00:49:46.730 align:middle line:84% So sleep is really, really interrupted, lots 00:49:46.730 --> 00:49:47.570 align:middle line:90% of awakenings. 00:49:47.570 --> 00:49:49.580 align:middle line:84% And the longest sleep episode that people get 00:49:49.580 --> 00:49:50.940 align:middle line:90% is really, really short. 00:49:50.940 --> 00:49:51.440 align:middle line:90% Yes? 00:49:51.440 --> 00:49:52.232 align:middle line:90% You had a question. 00:49:52.232 --> 00:49:54.470 align:middle line:84% AUDIENCE: What is the variation across days 00:49:54.470 --> 00:49:56.620 align:middle line:84% in sleep [? efficiency ?] or hours of sleep. 00:49:56.620 --> 00:49:59.620 align:middle line:84% Like are people sleeping five hours every single day, 00:49:59.620 --> 00:50:00.250 align:middle line:90% or is it? 00:50:00.250 --> 00:50:01.250 align:middle line:90% FRANK SCHILBACH: I think so. 00:50:01.250 --> 00:50:02.958 align:middle line:84% I actually don't-- it's a great question. 00:50:02.958 --> 00:50:04.870 align:middle line:84% I don't know the answer that well. 00:50:04.870 --> 00:50:08.290 align:middle line:84% I think there's quite a bit of variation still within person 00:50:08.290 --> 00:50:08.900 align:middle line:90% as well. 00:50:08.900 --> 00:50:11.830 align:middle line:84% So people sometimes get really terrible sleep. 00:50:11.830 --> 00:50:14.530 align:middle line:84% I don't know whether that's different from how 00:50:14.530 --> 00:50:17.800 align:middle line:84% it would be in the US, like the women there. 00:50:17.800 --> 00:50:19.450 align:middle line:84% All of our data is actually online now, 00:50:19.450 --> 00:50:22.550 align:middle line:90% I think so you can check it out. 00:50:22.550 --> 00:50:23.270 align:middle line:90% Yes? 00:50:23.270 --> 00:50:25.187 align:middle line:84% AUDIENCE: [INAUDIBLE] sleep actually recommend 00:50:25.187 --> 00:50:26.950 align:middle line:90% 7 to 9 hours per night. 00:50:26.950 --> 00:50:29.550 align:middle line:90% Is that based on self-reported? 00:50:29.550 --> 00:50:33.480 align:middle line:84% FRANK SCHILBACH: That should be actual sleep. 00:50:33.480 --> 00:50:35.910 align:middle line:84% Now, it's a little bit hard to disentangle that. 00:50:35.910 --> 00:50:38.490 align:middle line:84% In a sense, like the recommendation in the US 00:50:38.490 --> 00:50:41.150 align:middle line:84% are mostly for the US in which countries 00:50:41.150 --> 00:50:43.283 align:middle line:84% where people have high sleep efficiency. 00:50:43.283 --> 00:50:44.700 align:middle line:84% So then it actually doesn't matter 00:50:44.700 --> 00:50:51.840 align:middle line:84% so much whether it's actual sleep or self-reported sleep, 00:50:51.840 --> 00:50:53.930 align:middle line:90% because that's quite similar. 00:50:53.930 --> 00:50:56.650 align:middle line:84% The discrepancy between self-reported and actual sleep 00:50:56.650 --> 00:50:58.840 align:middle line:84% comes mostly from the time that's awake, 00:50:58.840 --> 00:51:02.017 align:middle line:84% because people at nights-- because when people wake up, 00:51:02.017 --> 00:51:03.850 align:middle line:84% they often don't remember that they woke up, 00:51:03.850 --> 00:51:06.183 align:middle line:84% but then remember like how long they woke up, and so on. 00:51:06.183 --> 00:51:09.340 align:middle line:84% And they often also confuse time in bed and time asleep. 00:51:09.340 --> 00:51:11.110 align:middle line:84% But since in the US, that's quite similar, 00:51:11.110 --> 00:51:14.980 align:middle line:84% because 95% of the time in bed, people are actually sleeping, 00:51:14.980 --> 00:51:16.660 align:middle line:84% it doesn't actually matter so much which 00:51:16.660 --> 00:51:18.198 align:middle line:90% of the two statistics you mean. 00:51:18.198 --> 00:51:19.990 align:middle line:84% But I think they mean-- their main object-- 00:51:19.990 --> 00:51:24.000 align:middle line:90% they mean actually time asleep. 00:51:24.000 --> 00:51:28.110 align:middle line:84% And so I often show the statistic and people 00:51:28.110 --> 00:51:31.590 align:middle line:84% kind of look at it and say, well 70%, that seems kind of low, 00:51:31.590 --> 00:51:33.550 align:middle line:90% but what's the big deal? 00:51:33.550 --> 00:51:36.210 align:middle line:84% And at some point we created this graph, which I think 00:51:36.210 --> 00:51:38.670 align:middle line:84% makes things much clearer and makes things more 00:51:38.670 --> 00:51:41.580 align:middle line:84% show how dramatically bad people sleep. 00:51:41.580 --> 00:51:44.640 align:middle line:84% This is what a typical night in Chennai looks like. 00:51:44.640 --> 00:51:47.010 align:middle line:84% In red, you see times when people are awake. 00:51:47.010 --> 00:51:48.720 align:middle line:84% In gray, you see people are sleeping. 00:51:48.720 --> 00:51:50.595 align:middle line:84% And you see, essentially, it's riddled with-- 00:51:50.595 --> 00:51:52.870 align:middle line:84% and this is like a median person in our sample. 00:51:52.870 --> 00:51:56.590 align:middle line:84% This is not a bad night that one person had one time. 00:51:56.590 --> 00:51:59.430 align:middle line:84% These are like median values of awakenings, et cetera, 00:51:59.430 --> 00:52:00.690 align:middle line:90% and sleep efficiency. 00:52:00.690 --> 00:52:03.570 align:middle line:84% This night is riddled with interruptions. 00:52:03.570 --> 00:52:06.480 align:middle line:84% People-- the night is highly fragmented. 00:52:06.480 --> 00:52:09.600 align:middle line:84% There's 31 awakenings of at least one minute, but at least 00:52:09.600 --> 00:52:11.460 align:middle line:90% 10 awakenings that are longer. 00:52:11.460 --> 00:52:14.370 align:middle line:84% And the longest sleep episode, uninterrupted sleep episode, 00:52:14.370 --> 00:52:16.260 align:middle line:90% is about 40 to 45 minutes. 00:52:16.260 --> 00:52:20.040 align:middle line:84% So essentially, people never get any full sleep cycles. 00:52:20.040 --> 00:52:23.190 align:middle line:84% And for comparison, this is what a good night looks 00:52:23.190 --> 00:52:25.140 align:middle line:84% like in our sample and what maybe 00:52:25.140 --> 00:52:27.930 align:middle line:84% sleep looks like in the US or other types of-- your sleep 00:52:27.930 --> 00:52:28.855 align:middle line:90% might look like. 00:52:28.855 --> 00:52:30.480 align:middle line:84% Notice that there is also interruption, 00:52:30.480 --> 00:52:33.240 align:middle line:84% people actually wake up often or at least wake up a little bit, 00:52:33.240 --> 00:52:37.150 align:middle line:84% but there's also long periods of like uninterrupted sleep 00:52:37.150 --> 00:52:39.090 align:middle line:84% where people really get to enjoy their night. 00:52:39.090 --> 00:52:41.110 align:middle line:84% So I think this is like, in fact, 00:52:41.110 --> 00:52:42.840 align:middle line:84% if I had to show one graph of this paper, 00:52:42.840 --> 00:52:45.540 align:middle line:84% was in fact, this one, is to say sleep is dramatically 00:52:45.540 --> 00:52:47.940 align:middle line:90% different in people's lives. 00:52:47.940 --> 00:52:51.420 align:middle line:84% And that warrants further investigation. 00:52:51.420 --> 00:52:54.480 align:middle line:84% Now, this also happens to be the case, probably 00:52:54.480 --> 00:52:57.830 align:middle line:84% my hypothesis, why we find no effects of increasing night 00:52:57.830 --> 00:53:00.840 align:middle line:84% sleep, because precisely, it's so interrupted. 00:53:00.840 --> 00:53:04.110 align:middle line:84% We didn't realize that until much later, so this 00:53:04.110 --> 00:53:07.042 align:middle line:90% is like an exposed hypothesis. 00:53:07.042 --> 00:53:08.250 align:middle line:90% So here's the study timeline. 00:53:08.250 --> 00:53:09.330 align:middle line:90% I think you have all seen this. 00:53:09.330 --> 00:53:11.413 align:middle line:84% I think one-- the only thing I'm going to say here 00:53:11.413 --> 00:53:13.350 align:middle line:84% is that when you randomize sleep, 00:53:13.350 --> 00:53:15.720 align:middle line:84% you have to think about what the outside option is. 00:53:15.720 --> 00:53:17.370 align:middle line:84% What do people do in that time when 00:53:17.370 --> 00:53:20.667 align:middle line:84% they're sleeping or not sleeping, 00:53:20.667 --> 00:53:21.750 align:middle line:90% when they're not sleeping? 00:53:21.750 --> 00:53:24.000 align:middle line:84% And so that's why we have in the no-nap condition 00:53:24.000 --> 00:53:26.700 align:middle line:84% we randomized work versus break, because that 00:53:26.700 --> 00:53:29.850 align:middle line:84% allows us to look at the opportunity costs of sleep. 00:53:29.850 --> 00:53:31.645 align:middle line:84% And that's in fact, quite important. 00:53:31.645 --> 00:53:33.270 align:middle line:84% And the sleep literature has not really 00:53:33.270 --> 00:53:34.398 align:middle line:90% focused on that very much. 00:53:34.398 --> 00:53:36.690 align:middle line:84% When you try to think about why are people not sleeping 00:53:36.690 --> 00:53:38.700 align:middle line:84% as much as perhaps we think they should, 00:53:38.700 --> 00:53:43.060 align:middle line:84% well, often it's because they have high opportunity costs. 00:53:43.060 --> 00:53:46.080 align:middle line:84% So then second, night sleep treatments 00:53:46.080 --> 00:53:49.140 align:middle line:84% increased sleep duration, but not sleep efficiency. 00:53:49.140 --> 00:53:52.710 align:middle line:84% We had set things up such that we are hoping to increase both. 00:53:52.710 --> 00:53:55.700 align:middle line:84% We thought we know it would increase sleep duration, 00:53:55.700 --> 00:53:58.200 align:middle line:84% but we thought it would in fact come through increased sleep 00:53:58.200 --> 00:54:01.440 align:middle line:84% efficiency, through earplugs, eyeshades, et cetera. 00:54:01.440 --> 00:54:03.840 align:middle line:84% It just turns out that that's not really the case. 00:54:03.840 --> 00:54:06.570 align:middle line:84% In some way, that makes things simpler for our analysis, 00:54:06.570 --> 00:54:08.730 align:middle line:84% because we can say, look, we really now looking 00:54:08.730 --> 00:54:11.830 align:middle line:84% at like sleep duration and holding efficiency constant. 00:54:11.830 --> 00:54:13.830 align:middle line:84% And if both of them had actually been increased, 00:54:13.830 --> 00:54:16.440 align:middle line:84% it would have been like some way messier. 00:54:16.440 --> 00:54:20.850 align:middle line:84% But in terms of policy or in terms of like, 00:54:20.850 --> 00:54:23.910 align:middle line:84% as a policy intervention, that's not what we had intended. 00:54:23.910 --> 00:54:26.520 align:middle line:84% Because of course, once you increase people's sleep 00:54:26.520 --> 00:54:29.432 align:middle line:84% duration through having people spend more time in bed, 00:54:29.432 --> 00:54:31.890 align:middle line:84% now you have opportunity because now people are essentially 00:54:31.890 --> 00:54:33.720 align:middle line:84% have less time available for work, 00:54:33.720 --> 00:54:36.150 align:middle line:84% et cetera, which we in fact find because people 00:54:36.150 --> 00:54:38.940 align:middle line:84% lower their labor supply or time at work. 00:54:38.940 --> 00:54:41.070 align:middle line:84% Now, these effects are sizable compared 00:54:41.070 --> 00:54:42.930 align:middle line:84% to other types of intervention, for example, 00:54:42.930 --> 00:54:44.260 align:middle line:90% sleeping pills and the like. 00:54:44.260 --> 00:54:46.960 align:middle line:84% So this is a meaningful increase in sleep. 00:54:46.960 --> 00:54:48.840 align:middle line:84% But as I said, it's entirely coming 00:54:48.840 --> 00:54:51.990 align:middle line:90% from increased time in bed. 00:54:51.990 --> 00:54:55.510 align:middle line:84% Sleep efficiency essentially stays the same. 00:54:55.510 --> 00:54:56.762 align:middle line:90% Yeah? 00:54:56.762 --> 00:54:58.470 align:middle line:84% AUDIENCE: Do you have a sense of if there 00:54:58.470 --> 00:55:02.360 align:middle line:84% are kinds of interventions that could increase the efficiency, 00:55:02.360 --> 00:55:04.110 align:middle line:84% like sleeping just might be better, or not 00:55:04.110 --> 00:55:05.640 align:middle line:90% especially in the setting or? 00:55:05.640 --> 00:55:06.030 align:middle line:90% FRANK SCHILBACH: Yeah? 00:55:06.030 --> 00:55:06.988 align:middle line:90% That's a good question. 00:55:06.988 --> 00:55:09.090 align:middle line:84% So broadly what our intervention was 00:55:09.090 --> 00:55:12.180 align:middle line:84% doing was trying to address environmental factors, 00:55:12.180 --> 00:55:16.860 align:middle line:84% you know, earplugs, eyeshades, curtain, table fan, et cetera. 00:55:16.860 --> 00:55:20.530 align:middle line:84% Now, you could address a broader set of environmental factors. 00:55:20.530 --> 00:55:22.710 align:middle line:84% For example, we didn't do mosquito repellent 00:55:22.710 --> 00:55:24.210 align:middle line:84% because we were a little bit worried 00:55:24.210 --> 00:55:27.810 align:middle line:84% about exclusion restriction issues with health problems 00:55:27.810 --> 00:55:31.380 align:middle line:84% and so on associated with mosquitoes. 00:55:31.380 --> 00:55:35.340 align:middle line:84% You could have, in principle, provided 00:55:35.340 --> 00:55:38.280 align:middle line:84% better sleep environments in the sense of like there's still 00:55:38.280 --> 00:55:41.680 align:middle line:84% five people in the same room in a very crowded space and so on. 00:55:41.680 --> 00:55:44.490 align:middle line:84% So you could have done more to effect environmental factors. 00:55:44.490 --> 00:55:46.830 align:middle line:84% And we don't quite know what that would have done. 00:55:46.830 --> 00:55:51.242 align:middle line:84% Second, you could have tried to address psychological factors 00:55:51.242 --> 00:55:53.700 align:middle line:84% or internal factors if you want that affect people's sleep. 00:55:53.700 --> 00:55:57.420 align:middle line:84% And there's, I'll tell you about this in a few slides. 00:55:57.420 --> 00:55:59.490 align:middle line:84% There's quite a bit of evidence that 00:55:59.490 --> 00:56:01.350 align:middle line:84% cognitive behavioral therapy for insomnia 00:56:01.350 --> 00:56:03.110 align:middle line:90% is, in fact, highly effective. 00:56:03.110 --> 00:56:05.280 align:middle line:84% It's essentially-- it's some form of CBT that 00:56:05.280 --> 00:56:10.530 align:middle line:84% target directly to try to help people deal with anxiety 00:56:10.530 --> 00:56:13.020 align:middle line:84% of falling asleep and worrying too much about not falling 00:56:13.020 --> 00:56:15.228 align:middle line:84% asleep and therefore not falling asleep and therefore 00:56:15.228 --> 00:56:17.970 align:middle line:90% not worrying about sleep again. 00:56:17.970 --> 00:56:20.017 align:middle line:84% And there's quite a bit of work on that, 00:56:20.017 --> 00:56:21.850 align:middle line:84% that shows in fact, that's highly effective. 00:56:21.850 --> 00:56:22.530 align:middle line:90% So any of you. 00:56:22.530 --> 00:56:25.340 align:middle line:84% I have sleeping problems, I would look into CBTI as it's 00:56:25.340 --> 00:56:28.160 align:middle line:90% called, which is-- 00:56:28.160 --> 00:56:30.500 align:middle line:84% there's lots of studies and like meta studies 00:56:30.500 --> 00:56:33.270 align:middle line:84% that show this is very effective in doing so. 00:56:33.270 --> 00:56:36.020 align:middle line:84% And so that would be a natural thing to try and do. 00:56:36.020 --> 00:56:39.080 align:middle line:84% It's not entirely clear that this is helping in the setting 00:56:39.080 --> 00:56:42.810 align:middle line:84% because there's also lots of environmental factors, 00:56:42.810 --> 00:56:44.220 align:middle line:90% so maybe you have to do both. 00:56:44.220 --> 00:56:45.590 align:middle line:90% We're not entirely sure. 00:56:45.590 --> 00:56:50.810 align:middle line:84% But that would be the next thing I would like to try. 00:56:50.810 --> 00:56:52.210 align:middle line:90% So then people also like to nap. 00:56:52.210 --> 00:56:56.680 align:middle line:84% Lots of people nap, and they nap about 15 minutes at the office. 00:56:56.680 --> 00:56:58.540 align:middle line:90% Now here's the overview. 00:56:58.540 --> 00:57:01.660 align:middle line:84% And I think it's quite helpful to summarize 00:57:01.660 --> 00:57:03.400 align:middle line:90% the effects that way. 00:57:03.400 --> 00:57:05.290 align:middle line:84% Essentially, there's no positive effect 00:57:05.290 --> 00:57:07.810 align:middle line:84% at all of increasing night sleep on any of the outcomes 00:57:07.810 --> 00:57:10.360 align:middle line:84% that we find, nor on the overall index. 00:57:10.360 --> 00:57:13.720 align:middle line:84% And we were like shocked about these effects, and a sense, 00:57:13.720 --> 00:57:16.240 align:middle line:84% like we had pretty strong priors that there 00:57:16.240 --> 00:57:18.310 align:middle line:84% would be some positive things going on, based on, 00:57:18.310 --> 00:57:20.050 align:middle line:84% we had like spend lots and lots of time 00:57:20.050 --> 00:57:21.470 align:middle line:90% reading the sleep literature. 00:57:21.470 --> 00:57:23.640 align:middle line:84% So this was really surprising at the time 00:57:23.640 --> 00:57:27.280 align:middle line:84% and took us quite a while to understand and digest 00:57:27.280 --> 00:57:27.970 align:middle line:90% if you want. 00:57:27.970 --> 00:57:31.030 align:middle line:84% And ironically, there's one significant effect 00:57:31.030 --> 00:57:35.885 align:middle line:84% that's negative, which is time spent working, 00:57:35.885 --> 00:57:38.260 align:middle line:84% which is coming from precisely because of the opportunity 00:57:38.260 --> 00:57:38.920 align:middle line:90% cost of sleep. 00:57:38.920 --> 00:57:41.050 align:middle line:84% People spent more time in bed, therefore, 00:57:41.050 --> 00:57:43.240 align:middle line:84% have less time for other things available. 00:57:43.240 --> 00:57:48.520 align:middle line:84% And therefore, part of that goes into reduced work time. 00:57:48.520 --> 00:57:49.090 align:middle line:90% Yeah? 00:57:49.090 --> 00:57:51.345 align:middle line:84% AUDIENCE: Do they work at night-- 00:57:51.345 --> 00:57:52.553 align:middle line:90% FRANK SCHILBACH: What's that? 00:57:52.553 --> 00:57:54.790 align:middle line:84% AUDIENCE: --asleep, so the type of time-- 00:57:54.790 --> 00:57:56.020 align:middle line:90% This is [? night ?] people. 00:57:56.020 --> 00:57:57.437 align:middle line:84% FRANK SCHILBACH: Yeah, but this is 00:57:57.437 --> 00:58:00.153 align:middle line:84% so people spend like an hour or more in bed, 00:58:00.153 --> 00:58:01.820 align:middle line:84% either in the evening or in the morning. 00:58:01.820 --> 00:58:03.580 align:middle line:84% So now, they have to do all sorts of other things, 00:58:03.580 --> 00:58:05.980 align:middle line:84% like, for example, a lot of the majority of our sample 00:58:05.980 --> 00:58:06.800 align:middle line:90% are women. 00:58:06.800 --> 00:58:11.110 align:middle line:84% So now they have lots of other households and other chores 00:58:11.110 --> 00:58:15.040 align:middle line:84% that they have to do, take care of children, et cetera. 00:58:15.040 --> 00:58:18.230 align:middle line:84% And so then they arrive later at work because of that. 00:58:18.230 --> 00:58:19.490 align:middle line:90% Does that make sense? 00:58:19.490 --> 00:58:20.930 align:middle line:90% Yeah. 00:58:20.930 --> 00:58:21.560 align:middle line:90% OK. 00:58:21.560 --> 00:58:25.910 align:middle line:84% Now, in contrast, we find a pretty robust evidence 00:58:25.910 --> 00:58:29.510 align:middle line:84% that, in fact, naps have positive effects 00:58:29.510 --> 00:58:32.420 align:middle line:84% on a range of outcomes, including well being, including 00:58:32.420 --> 00:58:34.783 align:middle line:84% cognition, and including work productivity. 00:58:34.783 --> 00:58:36.200 align:middle line:84% So the effect on work productivity 00:58:36.200 --> 00:58:37.730 align:middle line:90% are not particularly large. 00:58:37.730 --> 00:58:39.680 align:middle line:84% This is normal in these types of setting, 00:58:39.680 --> 00:58:43.460 align:middle line:84% in a sense of affecting work productivity is quite 00:58:43.460 --> 00:58:45.740 align:middle line:84% difficult, or like often, productivity 00:58:45.740 --> 00:58:48.290 align:middle line:90% is not particularly elastic. 00:58:48.290 --> 00:58:50.300 align:middle line:84% But again, there's opportunity costs of time. 00:58:50.300 --> 00:58:52.548 align:middle line:84% When you nap for half an hour in the afternoon, 00:58:52.548 --> 00:58:54.590 align:middle line:84% that takes quite a bit of time of your work time. 00:58:54.590 --> 00:58:56.210 align:middle line:84% And therefore, since people don't 00:58:56.210 --> 00:58:58.910 align:middle line:84% seem to shift their work hours, they 00:58:58.910 --> 00:59:00.500 align:middle line:84% spend the same amount of time at work. 00:59:00.500 --> 00:59:02.042 align:middle line:84% And if the nap for half an hour, now, 00:59:02.042 --> 00:59:04.577 align:middle line:84% they spend less time working, and therefore it 00:59:04.577 --> 00:59:05.660 align:middle line:90% has no effect on earnings. 00:59:05.660 --> 00:59:09.960 align:middle line:84% If anything, earnings I think go down a little bit. 00:59:09.960 --> 00:59:12.920 align:middle line:84% So it's not like naps pay for themselves in any way 00:59:12.920 --> 00:59:15.650 align:middle line:84% or that's a way to increase your earnings. 00:59:15.650 --> 00:59:18.740 align:middle line:84% Now naps look, and this is important to look 00:59:18.740 --> 00:59:20.520 align:middle line:90% at the opportunity cost. 00:59:20.520 --> 00:59:21.770 align:middle line:90% What are people doing instead? 00:59:21.770 --> 00:59:25.340 align:middle line:84% If you think about naps versus breaks, naps look great, 00:59:25.340 --> 00:59:27.560 align:middle line:84% because people earn more money, people are happier, 00:59:27.560 --> 00:59:32.570 align:middle line:84% people have improved cognition and save more. 00:59:32.570 --> 00:59:35.840 align:middle line:84% But instead, if you look at naps versus taking a break, 00:59:35.840 --> 00:59:42.065 align:middle line:84% the labor supply effects are, of course, negative. 00:59:42.065 --> 00:59:43.940 align:middle line:84% And the earnings a factor, are also negative. 00:59:43.940 --> 00:59:45.890 align:middle line:84% And there's a question of whether that's 00:59:45.890 --> 00:59:49.170 align:middle line:84% significant or not, but at the end of the day, 00:59:49.170 --> 00:59:50.810 align:middle line:84% it's not-- it's not a thing that you 00:59:50.810 --> 00:59:52.970 align:middle line:84% want to do to improve your earnings unless you 00:59:52.970 --> 00:59:56.510 align:middle line:84% have some slack time that you could use anyway. 00:59:56.510 --> 00:59:59.090 align:middle line:84% Now, if you're an employer, you want to think about, 00:59:59.090 --> 00:59:59.782 align:middle line:90% is it worth? 00:59:59.782 --> 01:00:02.240 align:middle line:84% Some of you were asking, is it worth doing for an employer, 01:00:02.240 --> 01:00:04.410 align:middle line:90% what's that rationale? 01:00:04.410 --> 01:00:05.960 align:middle line:84% Well, on the one hand, it's not clear 01:00:05.960 --> 01:00:08.000 align:middle line:84% that you want to do this because it's not 01:00:08.000 --> 01:00:09.540 align:middle line:90% like overall output goes up. 01:00:09.540 --> 01:00:12.150 align:middle line:84% And that's probably what you care about as an employer. 01:00:12.150 --> 01:00:13.850 align:middle line:84% But perhaps you also care about things 01:00:13.850 --> 01:00:16.520 align:middle line:84% like well-being of your employees, 01:00:16.520 --> 01:00:18.380 align:middle line:84% you care about retention, potentially. 01:00:18.380 --> 01:00:20.900 align:middle line:84% You care about potentially attention or other types 01:00:20.900 --> 01:00:24.750 align:middle line:84% of measures that maybe prevent accidents or the like at work. 01:00:24.750 --> 01:00:27.110 align:middle line:84% So it might be worth offering, in particular, 01:00:27.110 --> 01:00:29.960 align:middle line:84% if they're like breaks that people take anyway that you 01:00:29.960 --> 01:00:31.700 align:middle line:90% could essentially crowd out. 01:00:31.700 --> 01:00:34.740 align:middle line:90% 01:00:34.740 --> 01:00:36.700 align:middle line:90% So a quick discussion. 01:00:36.700 --> 01:00:40.320 align:middle line:84% So as I said, there's no effects on sleep, 01:00:40.320 --> 01:00:41.430 align:middle line:90% on night's sleep overall. 01:00:41.430 --> 01:00:43.620 align:middle line:84% This is in contrast to expert predictions 01:00:43.620 --> 01:00:45.000 align:middle line:90% and our own predictions. 01:00:45.000 --> 01:00:48.340 align:middle line:84% The expert predictions turned out to be quite useful for us. 01:00:48.340 --> 01:00:50.670 align:middle line:84% And the reason for that is that-- 01:00:50.670 --> 01:00:54.460 align:middle line:84% and some of you were saying the same, people were saying like, 01:00:54.460 --> 01:00:56.550 align:middle line:84% well, isn't that obvious that there's no effect? 01:00:56.550 --> 01:00:58.710 align:middle line:90% People should adjust. 01:00:58.710 --> 01:01:01.260 align:middle line:84% Why do you think that people are making mistakes, et cetera 01:01:01.260 --> 01:01:01.760 align:middle line:90% and so on? 01:01:01.760 --> 01:01:04.718 align:middle line:84% That's a reasonable perspective, in particular, exposed. 01:01:04.718 --> 01:01:06.510 align:middle line:84% It's not like ex-ante, people were thinking 01:01:06.510 --> 01:01:07.740 align:middle line:90% that, including ourselves. 01:01:07.740 --> 01:01:10.620 align:middle line:84% So there are, the expert surveys protects us 01:01:10.620 --> 01:01:14.430 align:middle line:84% from these kinds of comments in some ways. 01:01:14.430 --> 01:01:16.170 align:middle line:84% We even had some referees who are saying, 01:01:16.170 --> 01:01:18.420 align:middle line:90% well, that seems obvious. 01:01:18.420 --> 01:01:21.000 align:middle line:84% And then, we could say, well, look in fact, the experts 01:01:21.000 --> 01:01:23.220 align:middle line:84% survey, the experts didn't seem to think that 01:01:23.220 --> 01:01:26.430 align:middle line:90% including sleep researchers. 01:01:26.430 --> 01:01:28.650 align:middle line:84% And there's a large literature that 01:01:28.650 --> 01:01:31.230 align:middle line:84% seems to suggest there should be these types of effects. 01:01:31.230 --> 01:01:32.970 align:middle line:90% Now, why is there no effect? 01:01:32.970 --> 01:01:35.310 align:middle line:84% And we're a little bit unequipped 01:01:35.310 --> 01:01:38.820 align:middle line:84% to answer this question because we're really not testing 01:01:38.820 --> 01:01:40.260 align:middle line:90% the effects of sleep quality. 01:01:40.260 --> 01:01:42.690 align:middle line:84% What we can test is see the effect of sleep quantity. 01:01:42.690 --> 01:01:44.832 align:middle line:84% We don't have any variation, for example, exposed 01:01:44.832 --> 01:01:46.290 align:middle line:84% what we should have done perhaps is 01:01:46.290 --> 01:01:49.740 align:middle line:84% to say have some variation and good nap quality versus bad nap 01:01:49.740 --> 01:01:50.280 align:middle line:90% quality. 01:01:50.280 --> 01:01:52.740 align:middle line:84% You nap in a quiet space versus a loud space, 01:01:52.740 --> 01:01:55.020 align:middle line:84% and then we could look at the effect of sleep quality 01:01:55.020 --> 01:01:56.050 align:middle line:90% on outcomes. 01:01:56.050 --> 01:01:57.270 align:middle line:90% We don't quite have that. 01:01:57.270 --> 01:02:00.540 align:middle line:84% So we can't really-- we suspect it's to do with sleep quality, 01:02:00.540 --> 01:02:03.750 align:middle line:84% but it could also be that some of the findings from the sleep 01:02:03.750 --> 01:02:06.330 align:middle line:84% labs don't generalize to the field in the sense 01:02:06.330 --> 01:02:09.540 align:middle line:84% that small increases in sleep might just 01:02:09.540 --> 01:02:11.310 align:middle line:84% not have the effects of the effects 01:02:11.310 --> 01:02:13.470 align:middle line:84% and sleep labs that often come from estimating 01:02:13.470 --> 01:02:17.810 align:middle line:84% the effect of all-nighters on people's performance. 01:02:17.810 --> 01:02:20.265 align:middle line:84% And consistent with that, and we have 01:02:20.265 --> 01:02:21.890 align:middle line:84% new evidence that people in our setting 01:02:21.890 --> 01:02:23.340 align:middle line:90% under invest in night sleep. 01:02:23.340 --> 01:02:27.020 align:middle line:84% So as I said, this is now not another example 01:02:27.020 --> 01:02:30.260 align:middle line:84% of people like underinvesting in high return opportunity, 01:02:30.260 --> 01:02:35.330 align:middle line:84% at least not in the type of way we were trying to affect it. 01:02:35.330 --> 01:02:38.270 align:middle line:84% That's to say that, given people's environment, 01:02:38.270 --> 01:02:40.850 align:middle line:84% sleeping more actually, on average, at least, 01:02:40.850 --> 01:02:42.710 align:middle line:90% doesn't seem to be a good idea. 01:02:42.710 --> 01:02:44.600 align:middle line:84% Now importantly, that doesn't mean 01:02:44.600 --> 01:02:47.540 align:middle line:84% that if you relocated people from like their homes 01:02:47.540 --> 01:02:49.040 align:middle line:84% to a really nice home, you couldn't 01:02:49.040 --> 01:02:50.720 align:middle line:90% see really large effects. 01:02:50.720 --> 01:02:53.090 align:middle line:84% This is to say, if you make marginal adjustments given 01:02:53.090 --> 01:02:54.465 align:middle line:84% the condition you're in, it's not 01:02:54.465 --> 01:02:58.010 align:middle line:84% worth doing this type of investment or sleeping more. 01:02:58.010 --> 01:03:01.303 align:middle line:84% But if you instead moved everybody to a nice hotel, 01:03:01.303 --> 01:03:02.720 align:middle line:84% maybe there could be large effects 01:03:02.720 --> 01:03:04.580 align:middle line:84% because now people get much better sleep 01:03:04.580 --> 01:03:07.560 align:middle line:90% quality and duration and so on. 01:03:07.560 --> 01:03:10.220 align:middle line:84% So we didn't really test that, but within there, 01:03:10.220 --> 01:03:11.840 align:middle line:84% what's available in that choice set, 01:03:11.840 --> 01:03:14.540 align:middle line:84% we can reject the broad hypothesis 01:03:14.540 --> 01:03:16.370 align:middle line:84% that people under invest in nights 01:03:16.370 --> 01:03:18.410 align:middle line:84% sleep in these types of settings. 01:03:18.410 --> 01:03:20.540 align:middle line:90% Now, in contrast-- Yeah? 01:03:20.540 --> 01:03:23.492 align:middle line:84% AUDIENCE: I'm just kind of curious because you mentioned 01:03:23.492 --> 01:03:25.850 align:middle line:84% there was some ex post reactions of, 01:03:25.850 --> 01:03:28.740 align:middle line:84% oh, maybe these findings aren't so surprising because we 01:03:28.740 --> 01:03:30.330 align:middle line:84% at least, from referees and stuff, 01:03:30.330 --> 01:03:34.340 align:middle line:84% but I'm curious among sleep researchers, how 01:03:34.340 --> 01:03:38.910 align:middle line:84% these results were perceived and especially given 01:03:38.910 --> 01:03:40.470 align:middle line:84% this implication that potentially 01:03:40.470 --> 01:03:43.260 align:middle line:84% some of these findings are not quite 01:03:43.260 --> 01:03:44.468 align:middle line:90% robust to actual [INAUDIBLE]. 01:03:44.468 --> 01:03:46.385 align:middle line:84% FRANK SCHILBACH: Yeah, that's a good question. 01:03:46.385 --> 01:03:48.270 align:middle line:84% I think it's really important with engage 01:03:48.270 --> 01:03:53.250 align:middle line:84% with these types of audiences because there's 01:03:53.250 --> 01:03:56.790 align:middle line:84% the number of people who work on sleep and sleep research, 01:03:56.790 --> 01:03:58.980 align:middle line:84% of course, is way, way larger than in economics. 01:03:58.980 --> 01:04:02.250 align:middle line:84% And so we, actually writing a separate small piece 01:04:02.250 --> 01:04:04.980 align:middle line:84% that I put in the syllabus and that can be almost done, 01:04:04.980 --> 01:04:08.220 align:middle line:84% I can send it to Ed to put on the website 01:04:08.220 --> 01:04:12.690 align:middle line:84% where we try to encourage sleep researchers to-- 01:04:12.690 --> 01:04:16.590 align:middle line:84% or more broadly, try to encourage field research 01:04:16.590 --> 01:04:18.180 align:middle line:90% on sleep. 01:04:18.180 --> 01:04:20.380 align:middle line:84% And in particular, we try to encourage two things. 01:04:20.380 --> 01:04:23.610 align:middle line:84% One is field experiments, improve sleep 01:04:23.610 --> 01:04:25.810 align:middle line:84% in people's natural environments, 01:04:25.810 --> 01:04:28.710 align:middle line:84% including duration but also things like cognitive behavior 01:04:28.710 --> 01:04:31.920 align:middle line:84% therapy for insomnia or the like to improve sleep quality. 01:04:31.920 --> 01:04:33.870 align:middle line:84% And in particular, then in settings 01:04:33.870 --> 01:04:38.280 align:middle line:84% such as among the urban poor or in the US, 01:04:38.280 --> 01:04:40.470 align:middle line:84% but also say in India, in settings 01:04:40.470 --> 01:04:44.860 align:middle line:84% where sleep quality is much lower than it is in sleep labs. 01:04:44.860 --> 01:04:48.530 align:middle line:84% And so I think another way to frame 01:04:48.530 --> 01:04:50.960 align:middle line:84% this in the way we trying to put this when it comes 01:04:50.960 --> 01:04:53.580 align:middle line:84% to the discussion with the sleep researchers is to say, 01:04:53.580 --> 01:04:56.520 align:middle line:84% well, they seem to be large benefits in the lab. 01:04:56.520 --> 01:04:58.130 align:middle line:84% And now why are people not sleeping 01:04:58.130 --> 01:05:01.490 align:middle line:84% as much as perhaps sleep researchers think they should? 01:05:01.490 --> 01:05:04.310 align:middle line:84% Is it the case that either there are some barriers that 01:05:04.310 --> 01:05:04.957 align:middle line:90% prevent them? 01:05:04.957 --> 01:05:07.290 align:middle line:84% They could be behavioral barriers, could be information, 01:05:07.290 --> 01:05:08.960 align:middle line:84% it could be self control or the like. 01:05:08.960 --> 01:05:11.300 align:middle line:84% Or is it that for some reason these benefits in fact, 01:05:11.300 --> 01:05:12.980 align:middle line:84% don't manifest and perhaps that could 01:05:12.980 --> 01:05:14.480 align:middle line:90% be because of sleep quality. 01:05:14.480 --> 01:05:16.490 align:middle line:84% But regardless, we kind want to know, 01:05:16.490 --> 01:05:19.190 align:middle line:90% does that affect people's-- 01:05:19.190 --> 01:05:21.470 align:middle line:84% how does increasing sleep or affecting sleep 01:05:21.470 --> 01:05:23.060 align:middle line:90% affect people's lives? 01:05:23.060 --> 01:05:25.790 align:middle line:84% And in particular, and this is what sleep research does 01:05:25.790 --> 01:05:28.910 align:middle line:84% tend to not do, we want to measure things like earnings. 01:05:28.910 --> 01:05:31.880 align:middle line:84% We want to measure things like time at work, 01:05:31.880 --> 01:05:34.880 align:middle line:84% financial well-being, et cetera, economic outcomes, 01:05:34.880 --> 01:05:38.240 align:middle line:84% adding these types of outcomes to what sleep 01:05:38.240 --> 01:05:40.850 align:middle line:84% researchers would call pragmatic trials that 01:05:40.850 --> 01:05:43.350 align:middle line:84% try to improve people's sleep in field settings 01:05:43.350 --> 01:05:44.460 align:middle line:90% seems really important. 01:05:44.460 --> 01:05:46.460 align:middle line:84% To be very clear, there's some field experiments 01:05:46.460 --> 01:05:47.900 align:middle line:84% that sleep researchers are doing, 01:05:47.900 --> 01:05:50.630 align:middle line:84% but usually they measure sleep as the primary outcome as 01:05:50.630 --> 01:05:52.410 align:middle line:90% opposed to other stuff. 01:05:52.410 --> 01:05:54.380 align:middle line:84% So I think, we very much trying to engage 01:05:54.380 --> 01:05:56.940 align:middle line:84% and trying to constructively, and we 01:05:56.940 --> 01:05:58.580 align:middle line:90% have a wonderful co-author. 01:05:58.580 --> 01:05:59.850 align:middle line:90% Susan Redline is her name. 01:05:59.850 --> 01:06:02.540 align:middle line:84% She's a very senior sleep researcher 01:06:02.540 --> 01:06:05.060 align:middle line:84% who is very supportive and interested 01:06:05.060 --> 01:06:06.725 align:middle line:90% in trying to figure stuff out. 01:06:06.725 --> 01:06:09.675 align:middle line:90% 01:06:09.675 --> 01:06:10.175 align:middle line:90% Yeah? 01:06:10.175 --> 01:06:11.450 align:middle line:90% AUDIENCE: I'm curious about-- 01:06:11.450 --> 01:06:16.618 align:middle line:84% [INAUDIBLE] measure that, we measure 01:06:16.618 --> 01:06:19.238 align:middle line:90% the productivity [INAUDIBLE]? 01:06:19.238 --> 01:06:20.780 align:middle line:84% FRANK SCHILBACH: Yeah, we do in fact, 01:06:20.780 --> 01:06:23.013 align:middle line:90% we measure that quite precisely. 01:06:23.013 --> 01:06:24.430 align:middle line:84% And we find pretty similar effects 01:06:24.430 --> 01:06:27.370 align:middle line:84% in the morning and the afternoon. 01:06:27.370 --> 01:06:30.610 align:middle line:84% You might have expected the nap effects to be in the afternoon 01:06:30.610 --> 01:06:32.570 align:middle line:84% and they're slightly larger in the afternoon. 01:06:32.570 --> 01:06:35.028 align:middle line:84% So you can look at productivity over the course of the day, 01:06:35.028 --> 01:06:37.300 align:middle line:84% you see, productivity goes down a little bit right 01:06:37.300 --> 01:06:39.400 align:middle line:84% after the nap, which is people are kind of like, they call it, 01:06:39.400 --> 01:06:41.500 align:middle line:84% sleep inertia, people are kind of like a little bit out of it 01:06:41.500 --> 01:06:43.030 align:middle line:90% after just having taken a nap. 01:06:43.030 --> 01:06:45.230 align:middle line:90% And then productivity goes up. 01:06:45.230 --> 01:06:48.520 align:middle line:84% But in fact, we cannot reject that the effects in the morning 01:06:48.520 --> 01:06:51.020 align:middle line:84% are the same as the effects in the afternoon. 01:06:51.020 --> 01:06:51.760 align:middle line:90% Now why is that? 01:06:51.760 --> 01:06:56.500 align:middle line:84% It could be effects from cumulative effects of naps, 01:06:56.500 --> 01:06:59.590 align:middle line:84% like you napped yesterday, and therefore today you're 01:06:59.590 --> 01:07:00.760 align:middle line:90% more productive. 01:07:00.760 --> 01:07:02.830 align:middle line:84% Or it could be like anticipatory effects. 01:07:02.830 --> 01:07:05.200 align:middle line:84% You know at 11:00 AM, that at 1:00 PM 01:07:05.200 --> 01:07:07.080 align:middle line:84% you're going to nap in the afternoon, 01:07:07.080 --> 01:07:08.830 align:middle line:84% and therefore, you might be working harder 01:07:08.830 --> 01:07:10.040 align:middle line:90% than you would otherwise. 01:07:10.040 --> 01:07:11.913 align:middle line:84% So you find effects in both of those. 01:07:11.913 --> 01:07:13.330 align:middle line:84% For night's sleep, you essentially 01:07:13.330 --> 01:07:16.120 align:middle line:84% find no effects at any point in time, 01:07:16.120 --> 01:07:18.700 align:middle line:90% even early in the morning. 01:07:18.700 --> 01:07:20.890 align:middle line:84% So it seems to be-- it's not the hypothesis 01:07:20.890 --> 01:07:23.922 align:middle line:84% that right after sleeping more in the morning, maybe 01:07:23.922 --> 01:07:24.880 align:middle line:90% you're more productive. 01:07:24.880 --> 01:07:29.790 align:middle line:84% That doesn't seem to be the case for night's sleep either. 01:07:29.790 --> 01:07:31.350 align:middle line:84% There's lots of open questions here, 01:07:31.350 --> 01:07:32.820 align:middle line:84% and I'm going to just leave you with those. 01:07:32.820 --> 01:07:34.830 align:middle line:84% In particular, as I said, doing field research 01:07:34.830 --> 01:07:37.080 align:middle line:84% on these types of questions seems 01:07:37.080 --> 01:07:39.040 align:middle line:84% really important in rich settings, 01:07:39.040 --> 01:07:44.910 align:middle line:84% but also in low income settings, in rural settings, 01:07:44.910 --> 01:07:46.980 align:middle line:84% and people's natural sleep environments. 01:07:46.980 --> 01:07:49.530 align:middle line:84% In particular, sort of the key question that we can't answer 01:07:49.530 --> 01:07:52.410 align:middle line:84% is what's the impact of improving sleep quality? 01:07:52.410 --> 01:07:54.810 align:middle line:84% As I said, cognitive behavior therapy for insomnia 01:07:54.810 --> 01:07:56.970 align:middle line:90% seems to be really promising. 01:07:56.970 --> 01:07:59.975 align:middle line:84% Looking at longer run effects could be important. 01:07:59.975 --> 01:08:02.100 align:middle line:84% We look at one month, which is pretty long compared 01:08:02.100 --> 01:08:04.860 align:middle line:84% to the typical sleep lab literature. 01:08:04.860 --> 01:08:09.130 align:middle line:84% But maybe, the effects of the year, et cetera are different. 01:08:09.130 --> 01:08:11.910 align:middle line:84% One particular interesting topic might be sleep and learning 01:08:11.910 --> 01:08:13.290 align:middle line:90% for children. 01:08:13.290 --> 01:08:15.120 align:middle line:84% In education, there's lots of evidence 01:08:15.120 --> 01:08:18.100 align:middle line:84% that sleep helps with attention, memory consolidation, et 01:08:18.100 --> 01:08:18.600 align:middle line:90% cetera. 01:08:18.600 --> 01:08:20.767 align:middle line:84% And so that could be a really high return investment 01:08:20.767 --> 01:08:23.310 align:middle line:84% in education that's really cheap to do if you got children 01:08:23.310 --> 01:08:25.500 align:middle line:84% to sleep more or better if they're in fact 01:08:25.500 --> 01:08:27.782 align:middle line:90% are effects on learning. 01:08:27.782 --> 01:08:29.490 align:middle line:84% And then more broadly, there's a question 01:08:29.490 --> 01:08:31.470 align:middle line:84% about how do people decide how much to sleep 01:08:31.470 --> 01:08:33.450 align:middle line:90% and what are barriers to sleep? 01:08:33.450 --> 01:08:38.319 align:middle line:84% There's a nice paper by Avery et al. that tries to get at that. 01:08:38.319 --> 01:08:39.287 align:middle line:90% Let me skip this. 01:08:39.287 --> 01:08:40.870 align:middle line:84% You can look at a little bit about why 01:08:40.870 --> 01:08:44.393 align:middle line:90% this paper is working or not. 01:08:44.393 --> 01:08:46.810 align:middle line:84% Let me tell you a little bit about depression and anxiety, 01:08:46.810 --> 01:08:48.935 align:middle line:84% in part, because I think this is, in fact, the most 01:08:48.935 --> 01:08:51.910 align:middle line:84% interesting and exciting area that people are now currently 01:08:51.910 --> 01:08:53.540 align:middle line:90% are working on. 01:08:53.540 --> 01:08:57.130 align:middle line:84% So we have an overview of paper with Matthew Ridley who 01:08:57.130 --> 01:08:59.770 align:middle line:84% is on the job market this year from MIT, 01:08:59.770 --> 01:09:01.479 align:middle line:84% as well as Gautham Rao and Vikram 01:09:01.479 --> 01:09:06.760 align:middle line:84% Patel, who's a very senior global health and mental health 01:09:06.760 --> 01:09:10.810 align:middle line:84% researcher on the relationship between poverty, depression, 01:09:10.810 --> 01:09:12.850 align:middle line:90% and anxiety. 01:09:12.850 --> 01:09:16.840 align:middle line:84% And you can read that paper in detail. 01:09:16.840 --> 01:09:19.450 align:middle line:84% And I think we try to summarize broadly what we know 01:09:19.450 --> 01:09:24.310 align:middle line:84% and what our open research questions. 01:09:24.310 --> 01:09:26.740 align:middle line:84% And the first thing I want to emphasize 01:09:26.740 --> 01:09:28.810 align:middle line:84% is that when I started working on this topic, 01:09:28.810 --> 01:09:29.770 align:middle line:84% I was kind of thinking, like well, 01:09:29.770 --> 01:09:31.562 align:middle line:84% can we even measure depression and anxiety? 01:09:31.562 --> 01:09:33.250 align:middle line:90% Can we measure these factors? 01:09:33.250 --> 01:09:36.229 align:middle line:84% And it turns out we can actually do that quite well. 01:09:36.229 --> 01:09:38.290 align:middle line:84% And there's short screening surveys 01:09:38.290 --> 01:09:41.770 align:middle line:84% such as PHQ-9 or GAD-7 for depression and anxiety 01:09:41.770 --> 01:09:43.720 align:middle line:84% that are, in fact, highly predictive 01:09:43.720 --> 01:09:46.810 align:middle line:84% of the in-depth diagnosis by a trained psychiatrist. 01:09:46.810 --> 01:09:49.359 align:middle line:84% So we can actually measure these concepts pretty well 01:09:49.359 --> 01:09:50.482 align:middle line:90% in the field. 01:09:50.482 --> 01:09:52.149 align:middle line:84% Now, these are all kind of self-reports, 01:09:52.149 --> 01:09:54.490 align:middle line:84% right, so they are inherently self-reports. 01:09:54.490 --> 01:09:56.770 align:middle line:84% But I think people come to terms with that this 01:09:56.770 --> 01:10:00.423 align:middle line:84% is, in fact, these are pretty good 01:10:00.423 --> 01:10:01.840 align:middle line:84% measures, in part, because they're 01:10:01.840 --> 01:10:07.570 align:middle line:84% pretty highly correlated with the in-depth diagnosis. 01:10:07.570 --> 01:10:09.250 align:middle line:84% And now more and more papers actually 01:10:09.250 --> 01:10:11.410 align:middle line:84% have on the left hand side, depression and anxiety, 01:10:11.410 --> 01:10:14.070 align:middle line:90% like GAD-7 or PHQ-9. 01:10:14.070 --> 01:10:16.390 align:middle line:84% And economists, including economists 01:10:16.390 --> 01:10:19.250 align:middle line:90% seem to be fine with that. 01:10:19.250 --> 01:10:20.717 align:middle line:90% Now, how common are these? 01:10:20.717 --> 01:10:23.050 align:middle line:84% Actually in some sense, if you look at point prevalence, 01:10:23.050 --> 01:10:25.630 align:middle line:84% they're not that common in terms of if you look at anxiety 01:10:25.630 --> 01:10:27.160 align:middle line:90% or depression disorder. 01:10:27.160 --> 01:10:29.040 align:middle line:84% But when you look at lifetime prevalence, 01:10:29.040 --> 01:10:30.290 align:middle line:90% they're in fact, quite common. 01:10:30.290 --> 01:10:32.450 align:middle line:84% And we look at the number of families for example, 01:10:32.450 --> 01:10:34.867 align:middle line:84% who have at least one person who is at least at some point 01:10:34.867 --> 01:10:36.730 align:middle line:84% depressed, they are in fact, quite high. 01:10:36.730 --> 01:10:41.740 align:middle line:84% Prevalence is a lot higher among women and a lot higher 01:10:41.740 --> 01:10:44.025 align:middle line:84% among the low income populations. 01:10:44.025 --> 01:10:46.150 align:middle line:84% We don't actually necessarily understand the gender 01:10:46.150 --> 01:10:48.730 align:middle line:84% difference that well, and that could be in part some of this 01:10:48.730 --> 01:10:53.140 align:middle line:90% could be perhaps reporting. 01:10:53.140 --> 01:10:55.595 align:middle line:84% One thing I want to emphasize is you might say, well, 01:10:55.595 --> 01:10:57.970 align:middle line:84% what I'm telling you here is the prevalence of depression 01:10:57.970 --> 01:11:01.030 align:middle line:84% disorder and anxiety disorder as a clinical definition 01:11:01.030 --> 01:11:01.960 align:middle line:90% in some ways. 01:11:01.960 --> 01:11:05.050 align:middle line:84% But of course, they might also affect 01:11:05.050 --> 01:11:06.850 align:middle line:84% these types of effects or mental health 01:11:06.850 --> 01:11:10.090 align:middle line:84% might affect much broader share of the population. 01:11:10.090 --> 01:11:13.180 align:middle line:84% You might have some depressive symptoms in some ways, 01:11:13.180 --> 01:11:16.840 align:middle line:84% but you might not be diagnosed with depression and anxiety, 01:11:16.840 --> 01:11:19.630 align:middle line:84% and that might affect you well-being and your decision 01:11:19.630 --> 01:11:23.260 align:middle line:84% making or your behavior in various ways as well. 01:11:23.260 --> 01:11:25.990 align:middle line:84% So things like, [INAUDIBLE] who was also in the job market 01:11:25.990 --> 01:11:28.630 align:middle line:84% this year, working on mindfulness, meditation, 01:11:28.630 --> 01:11:31.960 align:middle line:84% might apply to a much larger population that 01:11:31.960 --> 01:11:35.530 align:middle line:84% are not necessarily have anxiety disorder, but are 01:11:35.530 --> 01:11:38.680 align:middle line:84% just are anxious or worried about their lives 01:11:38.680 --> 01:11:40.810 align:middle line:84% or about COVID or about unemployment et cetera 01:11:40.810 --> 01:11:41.380 align:middle line:90% and so on. 01:11:41.380 --> 01:11:43.630 align:middle line:84% So you might say, well, this number looks kind of low. 01:11:43.630 --> 01:11:46.540 align:middle line:84% But in fact, I think a much larger share of the population 01:11:46.540 --> 01:11:50.690 align:middle line:84% is affected by mental health one way or the other. 01:11:50.690 --> 01:11:53.890 align:middle line:84% Now importantly, there's large treatment gaps, especially 01:11:53.890 --> 01:11:55.140 align:middle line:90% in low-income populations. 01:11:55.140 --> 01:11:57.760 align:middle line:84% So something like 80%, 90% of people who are-- 01:11:57.760 --> 01:12:00.070 align:middle line:84% If you diagnose people with depression and anxiety, 01:12:00.070 --> 01:12:02.950 align:middle line:84% we'll essentially see no treatment whatsoever 01:12:02.950 --> 01:12:03.860 align:middle line:90% in their lives. 01:12:03.860 --> 01:12:06.020 align:middle line:84% And that's a huge number, and that's of course, 01:12:06.020 --> 01:12:10.630 align:middle line:84% is also a huge opportunity to try and make progress. 01:12:10.630 --> 01:12:13.270 align:middle line:84% Now, why should you study mental health as an economist? 01:12:13.270 --> 01:12:15.230 align:middle line:84% Usually I would ask this and wait for a second. 01:12:15.230 --> 01:12:18.820 align:middle line:90% But let me just tell you. 01:12:18.820 --> 01:12:21.250 align:middle line:84% The first order answer is like mental ill health 01:12:21.250 --> 01:12:24.670 align:middle line:84% makes people profoundly unhappy, like depression, anxiety, et 01:12:24.670 --> 01:12:27.340 align:middle line:84% cetera it's like extremely debilitating in their effects 01:12:27.340 --> 01:12:29.260 align:middle line:90% on people's lives. 01:12:29.260 --> 01:12:31.720 align:middle line:84% And as economists, we're trying to maximize 01:12:31.720 --> 01:12:33.860 align:middle line:90% utility, well-being of people. 01:12:33.860 --> 01:12:37.060 align:middle line:84% And so, if you find some interventions that 01:12:37.060 --> 01:12:41.020 align:middle line:84% make people happier or less depressed or less anxious, 01:12:41.020 --> 01:12:42.880 align:middle line:84% that's like in terms of welfare, that 01:12:42.880 --> 01:12:45.070 align:middle line:84% swamps any effects on like productivity 01:12:45.070 --> 01:12:47.170 align:middle line:84% or whatever that you might find of like 5% 01:12:47.170 --> 01:12:49.630 align:middle line:84% here and there, if you can affect the fraction of people 01:12:49.630 --> 01:12:51.760 align:middle line:84% who are depressed for the rest of their lives 01:12:51.760 --> 01:12:54.370 align:middle line:84% or at least for a part of their lives. 01:12:54.370 --> 01:12:56.620 align:middle line:84% That's got to be welfare relevant and that's something 01:12:56.620 --> 01:12:58.880 align:middle line:84% that we should study because it's important. 01:12:58.880 --> 01:13:00.910 align:middle line:84% Now, of course, mental health can also 01:13:00.910 --> 01:13:02.140 align:middle line:90% affect economic behavior. 01:13:02.140 --> 01:13:04.690 align:middle line:84% And economic forces might affect mental health. 01:13:04.690 --> 01:13:06.550 align:middle line:84% And so in that sense, you can try 01:13:06.550 --> 01:13:08.650 align:middle line:84% to understand what economic policies might 01:13:08.650 --> 01:13:09.730 align:middle line:90% affect mental health. 01:13:09.730 --> 01:13:12.020 align:middle line:84% Or we might try to understand well, 01:13:12.020 --> 01:13:15.070 align:middle line:84% how does-- how can we understand economic behavior 01:13:15.070 --> 01:13:17.950 align:middle line:84% through better understanding mental health? 01:13:17.950 --> 01:13:20.740 align:middle line:84% Now, how might mental health affect economic behavior 01:13:20.740 --> 01:13:21.850 align:middle line:90% and outcomes? 01:13:21.850 --> 01:13:25.840 align:middle line:84% Broadly speaking, you can think about work outcomes, 01:13:25.840 --> 01:13:31.090 align:middle line:84% performance at work, dealing with failure, job surge, labor 01:13:31.090 --> 01:13:34.870 align:middle line:84% supply, earnings and the like, classical labor economics 01:13:34.870 --> 01:13:36.170 align:middle line:90% topics. 01:13:36.170 --> 01:13:38.800 align:middle line:84% Second, you might kind of study economic permitives 01:13:38.800 --> 01:13:41.140 align:middle line:84% about beliefs, about confidence in yourself, 01:13:41.140 --> 01:13:44.380 align:middle line:84% about beliefs about whether you find a job or not, 01:13:44.380 --> 01:13:47.360 align:middle line:84% updating based on information that you receive, 01:13:47.360 --> 01:13:49.750 align:middle line:84% whether you just got rejected from a job application. 01:13:49.750 --> 01:13:51.250 align:middle line:84% If you're depressed, you might think 01:13:51.250 --> 01:13:54.490 align:middle line:84% that that's because of you as opposed to were just unlucky 01:13:54.490 --> 01:13:55.930 align:middle line:90% and you might give up. 01:13:55.930 --> 01:13:58.960 align:middle line:84% Then preferences, time, risk and social preferences. 01:13:58.960 --> 01:14:01.180 align:middle line:84% This could be things like patient's risk 01:14:01.180 --> 01:14:02.230 align:middle line:90% aversion for example. 01:14:02.230 --> 01:14:04.023 align:middle line:84% Think about if you have anxiety disorder 01:14:04.023 --> 01:14:05.440 align:middle line:84% and you think about risky choices. 01:14:05.440 --> 01:14:09.250 align:middle line:84% Well, presumably that will affect how risk averse you are 01:14:09.250 --> 01:14:11.470 align:middle line:84% and understanding these relationships better 01:14:11.470 --> 01:14:12.530 align:middle line:90% seem important. 01:14:12.530 --> 01:14:14.910 align:middle line:84% And then other things like decision making, default 01:14:14.910 --> 01:14:16.680 align:middle line:90% effects, choice overload, etc. 01:14:16.680 --> 01:14:18.220 align:middle line:90% could also be affected. 01:14:18.220 --> 01:14:19.645 align:middle line:84% There's a broader set of outcomes 01:14:19.645 --> 01:14:22.270 align:middle line:84% that you might be interested in like stigma and discrimination. 01:14:22.270 --> 01:14:25.140 align:middle line:84% This is partially what Matthew Ridley is working on, 01:14:25.140 --> 01:14:27.930 align:middle line:84% health behaviors, medical adherence, 01:14:27.930 --> 01:14:29.850 align:middle line:90% exercising, expenditures. 01:14:29.850 --> 01:14:32.100 align:middle line:84% But other topics such as female empowerment. 01:14:32.100 --> 01:14:33.990 align:middle line:84% There's a very nice paper by Baranov et al. 01:14:33.990 --> 01:14:39.810 align:middle line:84% that shows effects on spending expenditures 01:14:39.810 --> 01:14:42.360 align:middle line:84% and who makes decisions in households 01:14:42.360 --> 01:14:44.130 align:middle line:84% as well as other things like human capital 01:14:44.130 --> 01:14:47.010 align:middle line:84% accumulation, schooling, intergenerational effects, et 01:14:47.010 --> 01:14:49.050 align:middle line:84% cetera, so essentially children's well-being, 01:14:49.050 --> 01:14:52.320 align:middle line:84% but also the education and so on might be affected. 01:14:52.320 --> 01:14:54.600 align:middle line:84% At the same time, as I said, economic forces 01:14:54.600 --> 01:14:55.930 align:middle line:90% might affect mental health. 01:14:55.930 --> 01:14:59.910 align:middle line:84% And so there's lots-- we talk about this in a lot of detail 01:14:59.910 --> 01:15:00.600 align:middle line:90% in this paper. 01:15:00.600 --> 01:15:04.710 align:middle line:84% There's lots of evidence of things like pollution, 01:15:04.710 --> 01:15:10.290 align:middle line:84% losing your job, early life conditions, exposure to trauma, 01:15:10.290 --> 01:15:12.360 align:middle line:84% et cetera, all affect people's mental health. 01:15:12.360 --> 01:15:16.190 align:middle line:84% So to try and understand these forces better 01:15:16.190 --> 01:15:18.780 align:middle line:84% or perhaps ways in which we can avoid such effects 01:15:18.780 --> 01:15:22.130 align:middle line:90% seems important. 01:15:22.130 --> 01:15:25.700 align:middle line:84% Now, you might say, well, we have all these treatment gaps. 01:15:25.700 --> 01:15:27.950 align:middle line:84% There's potentially these effects. 01:15:27.950 --> 01:15:29.730 align:middle line:90% Now what can we actually do? 01:15:29.730 --> 01:15:32.553 align:middle line:84% And so in development in particular, so 01:15:32.553 --> 01:15:34.220 align:middle line:84% we know treatments are highly effective. 01:15:34.220 --> 01:15:36.530 align:middle line:84% There's lots of evidence of CBT and other interventions 01:15:36.530 --> 01:15:38.120 align:middle line:90% working. 01:15:38.120 --> 01:15:40.220 align:middle line:84% But there's only very few trained psychiatrists 01:15:40.220 --> 01:15:41.250 align:middle line:90% in many settings. 01:15:41.250 --> 01:15:46.140 align:middle line:84% And so drugs in particular, often not really available. 01:15:46.140 --> 01:15:47.190 align:middle line:90% So what can we do? 01:15:47.190 --> 01:15:49.940 align:middle line:84% So one alternative is inexpensive, scalable 01:15:49.940 --> 01:15:52.280 align:middle line:84% psychotherapy interventions that can 01:15:52.280 --> 01:15:53.780 align:middle line:90% treat depression and anxiety. 01:15:53.780 --> 01:15:55.160 align:middle line:84% And there's actually quite a bit of evidence 01:15:55.160 --> 01:15:57.360 align:middle line:84% that these types of interventions work really well. 01:15:57.360 --> 01:16:00.300 align:middle line:84% And one of those people is our co-author Vikram Patel, 01:16:00.300 --> 01:16:03.770 align:middle line:84% who is dedicated essentially decades of his life 01:16:03.770 --> 01:16:06.290 align:middle line:84% to developing these types of interventions 01:16:06.290 --> 01:16:09.030 align:middle line:84% and they happen to work really well. 01:16:09.030 --> 01:16:11.060 align:middle line:84% And so then, that's not a scalable thing 01:16:11.060 --> 01:16:14.040 align:middle line:84% that we can essentially introduce in many places. 01:16:14.040 --> 01:16:15.510 align:middle line:84% And one of the studies for example, 01:16:15.510 --> 01:16:18.380 align:middle line:84% that we follow up with, there's like persistent effects 01:16:18.380 --> 01:16:20.810 align:middle line:84% on depression, about 10, 15 percentage 01:16:20.810 --> 01:16:23.480 align:middle line:84% points five years later after the treatment 01:16:23.480 --> 01:16:25.580 align:middle line:84% for an intervention that takes six 01:16:25.580 --> 01:16:28.203 align:middle line:90% sessions for $50 per person. 01:16:28.203 --> 01:16:29.870 align:middle line:84% And that's like-- if you think about it, 01:16:29.870 --> 01:16:32.570 align:middle line:84% that's like hugely cost effective in terms 01:16:32.570 --> 01:16:37.290 align:middle line:84% of just reducing years lived with depression. 01:16:37.290 --> 01:16:39.380 align:middle line:84% And so kind of trying to figure out 01:16:39.380 --> 01:16:41.090 align:middle line:84% how to scale these types of interventions 01:16:41.090 --> 01:16:43.010 align:middle line:84% and then look at potential other benefits 01:16:43.010 --> 01:16:44.430 align:middle line:90% seems really important. 01:16:44.430 --> 01:16:47.270 align:middle line:84% The second potential alternative is new technologies 01:16:47.270 --> 01:16:48.860 align:middle line:90% using the internet or apps. 01:16:48.860 --> 01:16:51.560 align:middle line:84% There's some promising results, but mostly 01:16:51.560 --> 01:16:53.425 align:middle line:84% what I would call efficacy trials. 01:16:53.425 --> 01:16:54.800 align:middle line:84% So these are kind of trials where 01:16:54.800 --> 01:16:56.360 align:middle line:84% people are highly motivated and want 01:16:56.360 --> 01:17:00.680 align:middle line:84% to try to engage with their apps, et cetera. 01:17:00.680 --> 01:17:04.430 align:middle line:84% Therapy can be quite effective for those types of population. 01:17:04.430 --> 01:17:07.040 align:middle line:84% Key issues here, take up an adherence, 01:17:07.040 --> 01:17:09.720 align:middle line:84% how can we get people to actually do this and follow 01:17:09.720 --> 01:17:10.220 align:middle line:90% through? 01:17:10.220 --> 01:17:12.290 align:middle line:84% There's very little work in developing countries, 01:17:12.290 --> 01:17:15.380 align:middle line:84% but it seems clear that in the last next 5 , 01:17:15.380 --> 01:17:18.343 align:middle line:84% 10 years this market will explode. 01:17:18.343 --> 01:17:19.760 align:middle line:84% And there will be lots of activity 01:17:19.760 --> 01:17:22.670 align:middle line:84% and trying to figure out the right apps for people 01:17:22.670 --> 01:17:24.783 align:middle line:84% and trying to figure out how to improve 01:17:24.783 --> 01:17:27.200 align:middle line:84% people's mental well-being through that, just because it's 01:17:27.200 --> 01:17:30.110 align:middle line:84% so scalable and so broadly available potentially, 01:17:30.110 --> 01:17:32.310 align:middle line:90% and everybody has phones. 01:17:32.310 --> 01:17:33.580 align:middle line:90% OK. 01:17:33.580 --> 01:17:34.580 align:middle line:90% Let me skip all of that. 01:17:34.580 --> 01:17:36.747 align:middle line:84% There's some evidence that essentially mental health 01:17:36.747 --> 01:17:37.930 align:middle line:90% can affect economic outcome. 01:17:37.930 --> 01:17:41.830 align:middle line:84% Let me just leave you with some open questions in this area. 01:17:41.830 --> 01:17:44.480 align:middle line:84% There's lots of exciting and important questions. 01:17:44.480 --> 01:17:48.250 align:middle line:84% And one broad question is, how exactly does mental health 01:17:48.250 --> 01:17:49.900 align:middle line:90% affect economic outcomes? 01:17:49.900 --> 01:17:53.830 align:middle line:84% We have overall measures of like, when you provide therapy 01:17:53.830 --> 01:17:57.070 align:middle line:84% to people, it seems to be that labor supply a few years later 01:17:57.070 --> 01:17:58.370 align:middle line:90% seem to go up. 01:17:58.370 --> 01:18:00.670 align:middle line:84% But we don't quite know what's going on. 01:18:00.670 --> 01:18:03.190 align:middle line:84% Is this about income, productivity, labor supply, 01:18:03.190 --> 01:18:03.820 align:middle line:90% earnings? 01:18:03.820 --> 01:18:05.650 align:middle line:84% Is it coming through economic preferences, 01:18:05.650 --> 01:18:07.280 align:middle line:90% beliefs, decision making? 01:18:07.280 --> 01:18:08.050 align:middle line:90% Is it depression? 01:18:08.050 --> 01:18:08.707 align:middle line:90% Is it anxiety? 01:18:08.707 --> 01:18:10.790 align:middle line:84% It's really hard to tell what exactly is going on. 01:18:10.790 --> 01:18:13.720 align:middle line:84% And so understanding that better really seems important. 01:18:13.720 --> 01:18:16.750 align:middle line:84% Second, can we use online tools as a social media tools 01:18:16.750 --> 01:18:21.190 align:middle line:84% in some ways or phone tools, apps, to improve mental health 01:18:21.190 --> 01:18:22.330 align:middle line:90% and economic well-being? 01:18:22.330 --> 01:18:26.458 align:middle line:84% Can we use them to make progress in helping people, 01:18:26.458 --> 01:18:28.000 align:middle line:84% not just improve their mental health, 01:18:28.000 --> 01:18:31.460 align:middle line:84% but perhaps there are also economic benefits of those? 01:18:31.460 --> 01:18:33.680 align:middle line:84% Third, how can we increase take up. 01:18:33.680 --> 01:18:36.290 align:middle line:84% So like once we make these types of interventions available, 01:18:36.290 --> 01:18:38.600 align:middle line:84% how can we actually help people make use of them, 01:18:38.600 --> 01:18:42.360 align:middle line:84% take them up in the first place, but also adhere to them? 01:18:42.360 --> 01:18:44.360 align:middle line:84% There's different interventions that some people 01:18:44.360 --> 01:18:45.235 align:middle line:90% have been working on. 01:18:45.235 --> 01:18:46.670 align:middle line:90% You can try to reduce stigma. 01:18:46.670 --> 01:18:48.980 align:middle line:84% You can increase perceived efficacy, 01:18:48.980 --> 01:18:51.868 align:middle line:90% provide incentives, et cetera. 01:18:51.868 --> 01:18:54.410 align:middle line:84% Next, and I think this is one of the most important questions 01:18:54.410 --> 01:18:57.710 align:middle line:84% of our time is to say, what are the impacts 01:18:57.710 --> 01:18:59.390 align:middle line:84% of different economic interventions 01:18:59.390 --> 01:19:03.020 align:middle line:84% relative to cash, other social welfare programs or safety 01:19:03.020 --> 01:19:03.920 align:middle line:90% programs? 01:19:03.920 --> 01:19:07.040 align:middle line:84% But in particular, what's the psychosocial value of labor? 01:19:07.040 --> 01:19:10.645 align:middle line:84% There's a really nice paper by Reshmaan Hussam and colleagues 01:19:10.645 --> 01:19:12.770 align:middle line:84% that essentially looks at the effect of giving cash 01:19:12.770 --> 01:19:14.090 align:middle line:90% to people versus jobs. 01:19:14.090 --> 01:19:17.210 align:middle line:84% And they find that essentially having jobs people value, 01:19:17.210 --> 01:19:19.520 align:middle line:84% having jobs much more than just cash, 01:19:19.520 --> 01:19:28.790 align:middle line:84% this is in a refugee setting, I think in Myanmar, I think. 01:19:28.790 --> 01:19:32.220 align:middle line:84% And more broadly, what is the effect of employment programs, 01:19:32.220 --> 01:19:33.890 align:middle line:90% so having work versus UBI? 01:19:33.890 --> 01:19:36.050 align:middle line:84% And given that there's lots of automation, 01:19:36.050 --> 01:19:40.220 align:middle line:84% et cetera, taking over and various places. 01:19:40.220 --> 01:19:42.380 align:middle line:84% And people argue, well, let's just give people cash 01:19:42.380 --> 01:19:43.310 align:middle line:90% and they'll be happy. 01:19:43.310 --> 01:19:45.492 align:middle line:84% Well, that might not be as simple 01:19:45.492 --> 01:19:47.450 align:middle line:84% because there's lots of identity in other types 01:19:47.450 --> 01:19:49.400 align:middle line:84% of considerations that might affect 01:19:49.400 --> 01:19:52.580 align:middle line:84% people's well-being and their mental health 01:19:52.580 --> 01:19:55.040 align:middle line:84% and understanding these types of trade offs 01:19:55.040 --> 01:19:56.720 align:middle line:90% seems really important. 01:19:56.720 --> 01:19:58.580 align:middle line:84% Related to that is the effect of relative 01:19:58.580 --> 01:19:59.780 align:middle line:90% versus absolute poverty. 01:19:59.780 --> 01:20:01.550 align:middle line:84% Does inequality harm mental health? 01:20:01.550 --> 01:20:03.170 align:middle line:84% We actually don't very much about that 01:20:03.170 --> 01:20:06.440 align:middle line:84% because it's notoriously hard to identify. 01:20:06.440 --> 01:20:09.080 align:middle line:84% And finally, what is the optimal mix of economic and mental 01:20:09.080 --> 01:20:10.340 align:middle line:90% health interventions? 01:20:10.340 --> 01:20:12.737 align:middle line:84% You might say, well, to some degree as-- 01:20:12.737 --> 01:20:15.320 align:middle line:84% you might just say, well, let's just do the ultra-poor program 01:20:15.320 --> 01:20:18.350 align:middle line:84% and that seems to be working and improving also mental health. 01:20:18.350 --> 01:20:20.250 align:middle line:84% But then, there's a fraction of people 01:20:20.250 --> 01:20:22.348 align:middle line:84% who have severe mental health conditions, perhaps 01:20:22.348 --> 01:20:24.140 align:middle line:84% for that type of population, we should also 01:20:24.140 --> 01:20:26.300 align:middle line:84% provide mental health support or we should 01:20:26.300 --> 01:20:27.920 align:middle line:90% do combinations of those two. 01:20:27.920 --> 01:20:29.690 align:middle line:84% There are some studies that have started 01:20:29.690 --> 01:20:33.350 align:middle line:84% to look at the effect of cash versus therapy, which 01:20:33.350 --> 01:20:35.120 align:middle line:84% is Johannes Haushofer and colleagues, 01:20:35.120 --> 01:20:36.578 align:middle line:84% and Dean Karlan and colleagues that 01:20:36.578 --> 01:20:38.220 align:middle line:90% have started to look at that. 01:20:38.220 --> 01:20:39.930 align:middle line:84% But I think at the end of the day, 01:20:39.930 --> 01:20:42.530 align:middle line:84% it'll probably be the case that doing combinations 01:20:42.530 --> 01:20:45.750 align:middle line:84% of economic and psychological interventions will be optimal. 01:20:45.750 --> 01:20:48.440 align:middle line:84% And depending on what the population looks like, 01:20:48.440 --> 01:20:50.857 align:middle line:84% depending what people are most in need. 01:20:50.857 --> 01:20:52.440 align:middle line:84% And my guess is at the end of the day, 01:20:52.440 --> 01:20:55.670 align:middle line:84% you'll probably want to offer both to some degree. 01:20:55.670 --> 01:20:58.220 align:middle line:84% And some of the intervention that Esther showed you already 01:20:58.220 --> 01:20:59.960 align:middle line:84% last week, which is the interventions 01:20:59.960 --> 01:21:03.620 align:middle line:84% of the ultra poor programs, are in fact, in some sense, 01:21:03.620 --> 01:21:06.320 align:middle line:84% providing already a lot of hand-holding and support. 01:21:06.320 --> 01:21:07.910 align:middle line:84% And in some ways, you can almost think 01:21:07.910 --> 01:21:09.030 align:middle line:90% of this as like some form. 01:21:09.030 --> 01:21:10.360 align:middle line:84% It's not therapy, but it's some form 01:21:10.360 --> 01:21:12.620 align:middle line:84% of that gets pretty close, that somebody's providing 01:21:12.620 --> 01:21:15.560 align:middle line:84% essentially psychological support for people, 01:21:15.560 --> 01:21:18.020 align:middle line:84% and perhaps directly integrating some forms 01:21:18.020 --> 01:21:22.820 align:middle line:84% of therapy or techniques would be the most effective tool 01:21:22.820 --> 01:21:24.418 align:middle line:90% to do. 01:21:24.418 --> 01:21:25.460 align:middle line:90% That's all I have to say. 01:21:25.460 --> 01:21:28.070 align:middle line:84% I'm happy to talk to you afterwards. 01:21:28.070 --> 01:21:30.550 align:middle line:84% You are very welcome to come to my office hours 01:21:30.550 --> 01:21:32.550 align:middle line:84% to talk about your research or any of the things 01:21:32.550 --> 01:21:34.720 align:middle line:90% I have discussed. 01:21:34.720 --> 01:21:46.000 align:middle line:90%