WEBVTT 00:00:00.000 --> 00:00:04.960 align:middle line:84% [SQUEAKING] [RUSTLING] [CLICKING] 00:00:04.960 --> 00:00:14.446 align:middle line:90% 00:00:14.446 --> 00:00:16.890 align:middle line:84% ESTHER DUFLO: So I hope you enjoyed Frank's lecture 00:00:16.890 --> 00:00:19.950 align:middle line:90% on behavioral economics. 00:00:19.950 --> 00:00:23.370 align:middle line:84% We have put them early in the semester this year 00:00:23.370 --> 00:00:26.340 align:middle line:84% because the insight are going to keep 00:00:26.340 --> 00:00:31.320 align:middle line:84% coming back because, of course, they apply to any topic. 00:00:31.320 --> 00:00:33.300 align:middle line:84% But in some sense, today is going 00:00:33.300 --> 00:00:37.110 align:middle line:84% to be a bit of a fall back to much 00:00:37.110 --> 00:00:40.440 align:middle line:84% more neoclassical territory because I 00:00:40.440 --> 00:00:44.620 align:middle line:84% do want some time talking about the model that, in a sense, 00:00:44.620 --> 00:00:47.250 align:middle line:84% was the workhorse model, still is the workhorse 00:00:47.250 --> 00:00:50.700 align:middle line:84% model of how to think about human capital investment 00:00:50.700 --> 00:00:51.300 align:middle line:90% decision. 00:00:51.300 --> 00:00:53.640 align:middle line:84% And then once we know that model, 00:00:53.640 --> 00:00:57.910 align:middle line:84% we can think about what everything is wrong about it. 00:00:57.910 --> 00:01:00.940 align:middle line:90% But it's a good place to start. 00:01:00.940 --> 00:01:06.630 align:middle line:84% So I'm going to take one embodiment of that model, which 00:01:06.630 --> 00:01:08.790 align:middle line:84% is from a paper by Abhijit, which 00:01:08.790 --> 00:01:10.830 align:middle line:90% is sort of a review paper. 00:01:10.830 --> 00:01:17.500 align:middle line:84% But it's a very classical and commonly-used model. 00:01:17.500 --> 00:01:21.360 align:middle line:84% So you could call it the Barro-Becker human capital 00:01:21.360 --> 00:01:22.990 align:middle line:90% investment model. 00:01:22.990 --> 00:01:25.890 align:middle line:84% And the question it's trying to ask 00:01:25.890 --> 00:01:30.990 align:middle line:84% is how human capital investment decisions are made. 00:01:30.990 --> 00:01:34.620 align:middle line:84% Does the fact that the decisions are made within families, 00:01:34.620 --> 00:01:36.990 align:middle line:84% and therefore, by parents, potentially on behalf 00:01:36.990 --> 00:01:38.910 align:middle line:84% of their children, who might potentially be 00:01:38.910 --> 00:01:40.680 align:middle line:90% credit-constrained or not-- 00:01:40.680 --> 00:01:43.830 align:middle line:84% might that contribute to the persistence 00:01:43.830 --> 00:01:46.970 align:middle line:90% of poverty and inequality? 00:01:46.970 --> 00:01:50.210 align:middle line:84% What's the role for public policy in this world? 00:01:50.210 --> 00:01:52.550 align:middle line:84% And how it depends on various assumption? 00:01:52.550 --> 00:01:54.770 align:middle line:84% So most development economics have 00:01:54.770 --> 00:01:57.860 align:middle line:84% used, implicitly or explicitly, that framework 00:01:57.860 --> 00:02:00.560 align:middle line:84% in the back of their mind, the framework 00:02:00.560 --> 00:02:04.340 align:middle line:84% of the poor but efficient of Schultz, which is let's 00:02:04.340 --> 00:02:06.380 align:middle line:84% think about people, they have no money. 00:02:06.380 --> 00:02:09.050 align:middle line:84% But otherwise, they are making the best decision 00:02:09.050 --> 00:02:11.900 align:middle line:84% they can within their limited resources. 00:02:11.900 --> 00:02:14.360 align:middle line:84% To the Schultz framework, we're thinking 00:02:14.360 --> 00:02:19.430 align:middle line:84% about the farmer maximizing the use of inputs on their plot. 00:02:19.430 --> 00:02:21.770 align:middle line:84% But that framework in particular, 00:02:21.770 --> 00:02:24.230 align:middle line:84% in the human capital model, has been 00:02:24.230 --> 00:02:28.400 align:middle line:84% the way that people have been framing the decision, people 00:02:28.400 --> 00:02:31.100 align:middle line:84% like Duncan Thomas, like John Strauss, 00:02:31.100 --> 00:02:34.650 align:middle line:84% and have used as a framework to look at the empirical evidence. 00:02:34.650 --> 00:02:37.100 align:middle line:84% So if you read papers by Thomas, by Strauss, 00:02:37.100 --> 00:02:42.020 align:middle line:84% by Foster or Rosensweig, they will all have some version 00:02:42.020 --> 00:02:43.910 align:middle line:84% of that model, either on the page 00:02:43.910 --> 00:02:45.290 align:middle line:90% or on the back of their mind. 00:02:45.290 --> 00:02:47.860 align:middle line:90% 00:02:47.860 --> 00:02:51.420 align:middle line:84% So the idea there is that parents treat investment 00:02:51.420 --> 00:02:54.480 align:middle line:84% in human capital of themselves and their children 00:02:54.480 --> 00:02:57.300 align:middle line:84% as economic decisions exclusively, 00:02:57.300 --> 00:02:59.940 align:middle line:90% which have costs and returns. 00:02:59.940 --> 00:03:02.220 align:middle line:84% We'll write down a version of this model. 00:03:02.220 --> 00:03:05.490 align:middle line:84% We'll write it down at the beginning in a very general 00:03:05.490 --> 00:03:08.910 align:middle line:84% way, so where we'll allow for the education decision 00:03:08.910 --> 00:03:11.610 align:middle line:84% to respond to other thing then cost in return, 00:03:11.610 --> 00:03:14.760 align:middle line:84% and see how we need to narrow it and specialize it 00:03:14.760 --> 00:03:23.520 align:middle line:84% in order to get to the classical Barro-Becker framework. 00:03:23.520 --> 00:03:26.630 align:middle line:84% So let's start with a production technology, which 00:03:26.630 --> 00:03:29.600 align:middle line:84% is very standard, so none of the shenanigans 00:03:29.600 --> 00:03:33.140 align:middle line:84% we had in our [INAUDIBLE] scale relationship. 00:03:33.140 --> 00:03:37.040 align:middle line:84% There is a final good that's produced 00:03:37.040 --> 00:03:39.890 align:middle line:90% with two kind of human inputs-- 00:03:39.890 --> 00:03:43.160 align:middle line:90% skills and unskilled labor. 00:03:43.160 --> 00:03:45.980 align:middle line:84% So it's a very standard production function. 00:03:45.980 --> 00:03:48.470 align:middle line:90% You need these two inputs. 00:03:48.470 --> 00:03:52.730 align:middle line:84% The more inputs, more skills is better, 00:03:52.730 --> 00:03:57.660 align:middle line:84% more labor is better, but in a concave way. 00:03:57.660 --> 00:04:01.430 align:middle line:84% So there is no non-convexity in here. 00:04:01.430 --> 00:04:04.310 align:middle line:84% The labor supply is such that each person 00:04:04.310 --> 00:04:08.620 align:middle line:84% has one unit of unskilled labor and some skills. 00:04:08.620 --> 00:04:10.130 align:middle line:90% They supply them. 00:04:10.130 --> 00:04:11.500 align:middle line:90% Labor has no disutility. 00:04:11.500 --> 00:04:14.670 align:middle line:90% So we don't enter into that. 00:04:14.670 --> 00:04:17.820 align:middle line:84% And people leave for three periods. 00:04:17.820 --> 00:04:21.370 align:middle line:84% In the first period, they acquire skill. 00:04:21.370 --> 00:04:22.740 align:middle line:90% Their parents invest in them. 00:04:22.740 --> 00:04:24.730 align:middle line:90% They acquire skill. 00:04:24.730 --> 00:04:26.560 align:middle line:90% They don't need to work. 00:04:26.560 --> 00:04:29.920 align:middle line:84% In the second period is when they work. 00:04:29.920 --> 00:04:31.210 align:middle line:90% They earn an income. 00:04:31.210 --> 00:04:32.530 align:middle line:90% They have one child. 00:04:32.530 --> 00:04:35.785 align:middle line:84% And in the third period, they are retired, but they consume. 00:04:35.785 --> 00:04:38.520 align:middle line:90% 00:04:38.520 --> 00:04:39.330 align:middle line:90% There are markets. 00:04:39.330 --> 00:04:40.920 align:middle line:90% So there is a labor market. 00:04:40.920 --> 00:04:43.390 align:middle line:90% The price of labor is wtL. 00:04:43.390 --> 00:04:44.640 align:middle line:90% And the price of skill is wtH. 00:04:44.640 --> 00:04:47.980 align:middle line:90% 00:04:47.980 --> 00:04:50.680 align:middle line:84% Now, how is human capital produced? 00:04:50.680 --> 00:04:55.600 align:middle line:84% Human capital is produced using a combination of human capital. 00:04:55.600 --> 00:04:58.750 align:middle line:84% You need teachers, who are themselves educated, 00:04:58.750 --> 00:04:59.950 align:middle line:90% and unskilled labor. 00:04:59.950 --> 00:05:02.510 align:middle line:90% So they need some time. 00:05:02.510 --> 00:05:06.040 align:middle line:84% So to get each unit of human capital, 00:05:06.040 --> 00:05:12.430 align:middle line:84% you need gamma s h of s of h and h 00:05:12.430 --> 00:05:18.950 align:middle line:84% minus, where h minus is the human capital of your parents. 00:05:18.950 --> 00:05:22.730 align:middle line:84% So the idea is that if your parents are themselves 00:05:22.730 --> 00:05:27.830 align:middle line:84% educated, then it becomes cheaper to educate the child. 00:05:27.830 --> 00:05:29.660 align:middle line:84% If some parents are themselves educated, 00:05:29.660 --> 00:05:32.660 align:middle line:84% it becomes cheaper for them to educate their child 00:05:32.660 --> 00:05:35.120 align:middle line:84% because they can help out or something, 00:05:35.120 --> 00:05:38.720 align:middle line:84% although that's at a declining rate. 00:05:38.720 --> 00:05:40.830 align:middle line:90% So the family environment helps. 00:05:40.830 --> 00:05:45.710 align:middle line:84% And then in order to avoid create non-convexity there, 00:05:45.710 --> 00:05:48.860 align:middle line:84% this is the return to the human capital 00:05:48.860 --> 00:05:50.630 align:middle line:84% of the parents in the human capital 00:05:50.630 --> 00:05:56.210 align:middle line:84% production of the child is also positive, 00:05:56.210 --> 00:05:57.590 align:middle line:90% but decreasing over time. 00:05:57.590 --> 00:06:02.560 align:middle line:90% 00:06:02.560 --> 00:06:04.510 align:middle line:90% So this is our functions. 00:06:04.510 --> 00:06:07.395 align:middle line:90% 00:06:07.395 --> 00:06:11.860 align:middle line:84% It costs money to acquire education. 00:06:11.860 --> 00:06:15.010 align:middle line:90% 00:06:15.010 --> 00:06:17.860 align:middle line:84% What does that translate as in plain language? 00:06:17.860 --> 00:06:30.340 align:middle line:90% 00:06:30.340 --> 00:06:31.790 align:middle line:90% It's d2s as d2. 00:06:31.790 --> 00:06:45.630 align:middle line:90% 00:06:45.630 --> 00:06:46.130 align:middle line:90% Yeah? 00:06:46.130 --> 00:06:48.410 align:middle line:84% STUDENT: That the cost of education [INAUDIBLE].. 00:06:48.410 --> 00:06:50.790 align:middle line:84% ESTHER DUFLO: Yeah, the cost of education is convex. 00:06:50.790 --> 00:06:53.660 align:middle line:84% So it becomes more and more expensive to acquire 00:06:53.660 --> 00:06:58.440 align:middle line:84% subsequent skill, a subsequent unit as it goes. 00:06:58.440 --> 00:07:01.740 align:middle line:84% So that's avoid the source of non-convexity as well. 00:07:01.740 --> 00:07:05.870 align:middle line:90% It's probably realistic also. 00:07:05.870 --> 00:07:07.820 align:middle line:84% If your parents are more educated, 00:07:07.820 --> 00:07:10.100 align:middle line:84% it's cheaper to educate your kid, 00:07:10.100 --> 00:07:12.800 align:middle line:84% although that's at a declining rate. 00:07:12.800 --> 00:07:17.840 align:middle line:84% And then in steady state, so it's 00:07:17.840 --> 00:07:22.640 align:middle line:84% when all of the generations have the same human capital, 00:07:22.640 --> 00:07:26.450 align:middle line:84% it still costs money to educate people. 00:07:26.450 --> 00:07:29.900 align:middle line:90% It doesn't go to 0. 00:07:29.900 --> 00:07:33.860 align:middle line:84% So what this already rules out, and we are not 00:07:33.860 --> 00:07:36.200 align:middle line:84% going to play with the tool in this, 00:07:36.200 --> 00:07:41.990 align:middle line:84% is sources of non-convexity similar to the one that we saw 00:07:41.990 --> 00:07:45.290 align:middle line:90% in the Dasgupta and Ray model. 00:07:45.290 --> 00:07:53.390 align:middle line:84% We have seen that if fHH is less than 0, 00:07:53.390 --> 00:07:58.310 align:middle line:84% so there is no non-convexity in skill production at any point. 00:07:58.310 --> 00:08:04.520 align:middle line:84% And the cost of acquiring education 00:08:04.520 --> 00:08:06.600 align:middle line:84% is increasing with further units. 00:08:06.600 --> 00:08:10.250 align:middle line:90% So there is no gains here from-- 00:08:10.250 --> 00:08:11.750 align:middle line:90% so this is a very standard. 00:08:11.750 --> 00:08:13.830 align:middle line:84% So what this holds out, for example, 00:08:13.830 --> 00:08:18.260 align:middle line:84% is the Galor-Zeira model, which is 00:08:18.260 --> 00:08:23.900 align:middle line:84% one of the early, mid-1990 model introducing dynamic 00:08:23.900 --> 00:08:27.890 align:middle line:84% of inequality and growth in a non-representative agent 00:08:27.890 --> 00:08:30.710 align:middle line:84% model, where the fundamental assumption that 00:08:30.710 --> 00:08:35.549 align:middle line:84% makes everything tick is that there are non-convexity. 00:08:35.549 --> 00:08:38.780 align:middle line:84% So for example, the return to primary education is 0. 00:08:38.780 --> 00:08:40.280 align:middle line:90% And then there is a fixed cost. 00:08:40.280 --> 00:08:42.260 align:middle line:84% And you start getting a return once you 00:08:42.260 --> 00:08:45.570 align:middle line:84% get at least to middle school or at least to college. 00:08:45.570 --> 00:08:48.848 align:middle line:84% So that's something that is happening here, 00:08:48.848 --> 00:08:49.890 align:middle line:90% which we don't have here. 00:08:49.890 --> 00:08:51.690 align:middle line:84% We are just going to not have that. 00:08:51.690 --> 00:08:53.895 align:middle line:84% One reason why we don't have that is we 00:08:53.895 --> 00:08:55.175 align:middle line:90% already know how that works. 00:08:55.175 --> 00:08:58.943 align:middle line:84% And we know that works if you have a shape somewhere 00:08:58.943 --> 00:09:01.443 align:middle line:84% in a production function, you can generate a bow at the top. 00:09:01.443 --> 00:09:02.170 align:middle line:90% OK. 00:09:02.170 --> 00:09:05.160 align:middle line:84% So we're going to rule that out and start 00:09:05.160 --> 00:09:07.680 align:middle line:84% from a very well-behaved production function. 00:09:07.680 --> 00:09:11.940 align:middle line:84% What we also going to rule out is s12, 00:09:11.940 --> 00:09:15.940 align:middle line:90% the cost partial being negative. 00:09:15.940 --> 00:09:17.670 align:middle line:84% So the cost partial being negative 00:09:17.670 --> 00:09:21.750 align:middle line:84% means that if you are yourself more educated, 00:09:21.750 --> 00:09:25.425 align:middle line:84% it is cheaper to acquire a higher unit of human capital. 00:09:25.425 --> 00:09:27.980 align:middle line:90% 00:09:27.980 --> 00:09:32.050 align:middle line:84% So there, again, that might generate complementarity. 00:09:32.050 --> 00:09:35.350 align:middle line:84% Like if you have sufficiently strong complementarity 00:09:35.350 --> 00:09:39.130 align:middle line:84% between the human capital of the parents 00:09:39.130 --> 00:09:41.590 align:middle line:84% and educated the kid at higher level, 00:09:41.590 --> 00:09:45.910 align:middle line:84% you can also get spread out into two. 00:09:45.910 --> 00:09:48.580 align:middle line:84% You can get to two steady states, 00:09:48.580 --> 00:09:51.640 align:middle line:84% whereas the dynasties which start 00:09:51.640 --> 00:09:53.650 align:middle line:84% with a lot of human capital find it cheaper 00:09:53.650 --> 00:09:55.780 align:middle line:90% to move above a certain level. 00:09:55.780 --> 00:09:59.710 align:middle line:84% And the people who start with little human capital don't. 00:09:59.710 --> 00:10:03.140 align:middle line:84% So that's also a gain of people that get [INAUDIBLE].. 00:10:03.140 --> 00:10:07.510 align:middle line:84% So we are also ruling that out for the moment. 00:10:07.510 --> 00:10:11.530 align:middle line:84% We basically do nothing funny with the production function, 00:10:11.530 --> 00:10:14.140 align:middle line:84% either in the economy or human capital, 00:10:14.140 --> 00:10:15.325 align:middle line:90% because we already know. 00:10:15.325 --> 00:10:19.570 align:middle line:84% Not to say that this can't be realistic, but we've already-- 00:10:19.570 --> 00:10:20.410 align:middle line:90% this is behind us. 00:10:20.410 --> 00:10:22.580 align:middle line:90% We already played with them. 00:10:22.580 --> 00:10:26.800 align:middle line:84% So given that the cost of education is that S function, 00:10:26.800 --> 00:10:31.630 align:middle line:84% so if you yourself have Ht unit of human capital 00:10:31.630 --> 00:10:35.650 align:middle line:84% and you would like your kid to get Ht plus 1, 00:10:35.650 --> 00:10:39.400 align:middle line:84% you need to pay this is H function multiplied 00:10:39.400 --> 00:10:40.660 align:middle line:90% by gamma wH. 00:10:40.660 --> 00:10:46.900 align:middle line:84% That's your unit of skills plus 1 minus gamma wT. 00:10:46.900 --> 00:10:49.660 align:middle line:84% That's your units of non-skilled. 00:10:49.660 --> 00:10:51.960 align:middle line:90% OK? 00:10:51.960 --> 00:10:55.950 align:middle line:84% So that's a functional form we'll use. 00:10:55.950 --> 00:10:58.500 align:middle line:90% 00:10:58.500 --> 00:11:01.500 align:middle line:84% Now, we are going to start with very, very general utility 00:11:01.500 --> 00:11:08.340 align:middle line:84% functions, both in the static and in the link 00:11:08.340 --> 00:11:09.570 align:middle line:90% across generation. 00:11:09.570 --> 00:11:13.200 align:middle line:84% And then we are going to simplify them and see 00:11:13.200 --> 00:11:15.610 align:middle line:90% how Barro-Becker simplified. 00:11:15.610 --> 00:11:19.960 align:middle line:84% So starting with a very general utility function 00:11:19.960 --> 00:11:22.780 align:middle line:84% of what people might care about, they 00:11:22.780 --> 00:11:25.090 align:middle line:84% might care about what they consume, obviously. 00:11:25.090 --> 00:11:27.220 align:middle line:84% So this is your material consumption. 00:11:27.220 --> 00:11:31.180 align:middle line:84% And they might care about symbolic consumption, 00:11:31.180 --> 00:11:35.440 align:middle line:84% deriving for example, from how much education they have, 00:11:35.440 --> 00:11:39.760 align:middle line:84% or in fact, how much consumption they have, their kids have, 00:11:39.760 --> 00:11:42.590 align:middle line:84% how much money they've spent on their education, 00:11:42.590 --> 00:11:44.460 align:middle line:90% and so on, and so forth. 00:11:44.460 --> 00:11:46.190 align:middle line:90% So we'll start with that. 00:11:46.190 --> 00:11:48.320 align:middle line:84% Then the material consumption is pretty simple. 00:11:48.320 --> 00:11:49.880 align:middle line:90% You get money. 00:11:49.880 --> 00:11:52.760 align:middle line:84% While you're working, you will consume ct. 00:11:52.760 --> 00:11:55.730 align:middle line:84% In retirement, you're going to consume some amount. 00:11:55.730 --> 00:12:01.880 align:middle line:84% We are calling it p, both in terms of the final good. 00:12:01.880 --> 00:12:04.790 align:middle line:84% And let's start with writing the symbolic consumption 00:12:04.790 --> 00:12:09.170 align:middle line:90% with a very generic term. 00:12:09.170 --> 00:12:16.560 align:middle line:84% So you might be very proud that you go to MIT, 00:12:16.560 --> 00:12:23.391 align:middle line:84% just independently of any value that there is in going to MIT. 00:12:23.391 --> 00:12:25.320 align:middle line:84% You might very well be very proud 00:12:25.320 --> 00:12:28.530 align:middle line:90% if your child goes to MIT. 00:12:28.530 --> 00:12:31.560 align:middle line:90% So that's ht plus 1. 00:12:31.560 --> 00:12:33.210 align:middle line:90% So you like to consume yourself. 00:12:33.210 --> 00:12:38.520 align:middle line:84% But you might also be proud if your children have a Ferrari. 00:12:38.520 --> 00:12:42.480 align:middle line:84% So that's another part, potentially, 00:12:42.480 --> 00:12:44.740 align:middle line:84% of your symbolic utility function. 00:12:44.740 --> 00:12:47.730 align:middle line:84% So this is separate from the fact 00:12:47.730 --> 00:12:53.550 align:middle line:84% that you might just derive utility from their utility. 00:12:53.550 --> 00:12:56.310 align:middle line:84% This is just your direct utility from their consumption, 00:12:56.310 --> 00:12:59.280 align:middle line:84% so from the warm glow of their consumption. 00:12:59.280 --> 00:13:01.380 align:middle line:84% And irrespective of where you go, 00:13:01.380 --> 00:13:05.370 align:middle line:84% you might be happy to brag to your friends 00:13:05.370 --> 00:13:09.660 align:middle line:84% that you're going to this super expensive preschool. 00:13:09.660 --> 00:13:12.660 align:middle line:84% And there, it costs you $50,000 a year. 00:13:12.660 --> 00:13:15.820 align:middle line:84% And you are very proud to say that. 00:13:15.820 --> 00:13:20.490 align:middle line:84% So this could be all sorts of things 00:13:20.490 --> 00:13:25.455 align:middle line:84% that might enter in the symbolic consumption of someone. 00:13:25.455 --> 00:13:25.955 align:middle line:90% OK? 00:13:25.955 --> 00:13:28.860 align:middle line:90% 00:13:28.860 --> 00:13:32.370 align:middle line:84% So the reason why that's going to turn out to be important 00:13:32.370 --> 00:13:38.100 align:middle line:84% is that that creates a motive for educating 00:13:38.100 --> 00:13:45.570 align:middle line:84% your children, which goes over and above the return from doing 00:13:45.570 --> 00:13:50.970 align:middle line:84% so in terms of either their own earnings or their ability 00:13:50.970 --> 00:13:54.150 align:middle line:84% to educate their kids in the future. 00:13:54.150 --> 00:13:57.170 align:middle line:84% And if you did derive utility from it, 00:13:57.170 --> 00:14:00.320 align:middle line:84% then assuming that this is a normal good to some extent, 00:14:00.320 --> 00:14:02.150 align:middle line:84% or it might even be a luxury good, 00:14:02.150 --> 00:14:05.530 align:middle line:84% then richer people are going to want more of it-- 00:14:05.530 --> 00:14:07.363 align:middle line:84% not for its functional value, just 00:14:07.363 --> 00:14:08.530 align:middle line:90% because they like it better. 00:14:08.530 --> 00:14:12.960 align:middle line:90% 00:14:12.960 --> 00:14:18.150 align:middle line:84% Likewise, we can start from a very generic utility function 00:14:18.150 --> 00:14:23.355 align:middle line:90% for how we link the generation. 00:14:23.355 --> 00:14:30.700 align:middle line:84% So the family-level outcome is going 00:14:30.700 --> 00:14:44.200 align:middle line:84% to be the utility of all of the future dynasties, discounted 00:14:44.200 --> 00:14:47.050 align:middle line:90% at some rate. 00:14:47.050 --> 00:14:50.230 align:middle line:84% And it's defined recursively as well, and then 00:14:50.230 --> 00:14:57.370 align:middle line:84% the utility, your own valuation of the family value of all 00:14:57.370 --> 00:14:59.640 align:middle line:90% of the future generations. 00:14:59.640 --> 00:15:05.650 align:middle line:84% So in any period, someone is looking about. 00:15:05.650 --> 00:15:09.280 align:middle line:84% Someone is thinking, what's the what's the family-level utility 00:15:09.280 --> 00:15:10.600 align:middle line:90% that I'm observing today? 00:15:10.600 --> 00:15:15.640 align:middle line:84% Well, it's the discounted sum of all the older utility 00:15:15.640 --> 00:15:19.600 align:middle line:84% in the future, all of the instantaneous utility 00:15:19.600 --> 00:15:22.210 align:middle line:84% in the future, plus the fact that they 00:15:22.210 --> 00:15:25.330 align:middle line:84% take into account the utility of the entire family 00:15:25.330 --> 00:15:26.650 align:middle line:90% in the future. 00:15:26.650 --> 00:15:28.720 align:middle line:90% So that's recursive. 00:15:28.720 --> 00:15:30.850 align:middle line:90% So that's a complicated object. 00:15:30.850 --> 00:15:37.330 align:middle line:84% And then your own utility is the sum of your personal utility 00:15:37.330 --> 00:15:43.660 align:middle line:84% plus however you end up valuing the utility 00:15:43.660 --> 00:15:47.130 align:middle line:84% of your entire family at this point. 00:15:47.130 --> 00:15:50.000 align:middle line:84% So this seems like much more heavy notation 00:15:50.000 --> 00:15:52.650 align:middle line:90% that you are used to. 00:15:52.650 --> 00:15:54.130 align:middle line:90% And why is that the case? 00:15:54.130 --> 00:15:55.800 align:middle line:84% Because we are very used to work with 00:15:55.800 --> 00:15:57.300 align:middle line:90% the Barro-Becker preferences. 00:15:57.300 --> 00:16:02.586 align:middle line:84% And how does Barro-Becker simplify it? 00:16:02.586 --> 00:16:03.880 align:middle line:90% Do you know that? 00:16:03.880 --> 00:16:06.195 align:middle line:90% How do we go from this? 00:16:06.195 --> 00:16:12.310 align:middle line:84% You might value the future utility of the others. 00:16:12.310 --> 00:16:16.380 align:middle line:84% Then you might value the utility of the family in the future, 00:16:16.380 --> 00:16:20.980 align:middle line:84% with some different discount rate, and so on, and so forth. 00:16:20.980 --> 00:16:25.600 align:middle line:84% How do we simplify these intergenerational things? 00:16:25.600 --> 00:16:32.070 align:middle line:84% So this says, when you thinking about the utility 00:16:32.070 --> 00:16:36.360 align:middle line:84% to value your own utility, you value 00:16:36.360 --> 00:16:38.760 align:middle line:84% your own utility plus some discounted 00:16:38.760 --> 00:16:40.680 align:middle line:90% value of your family. 00:16:40.680 --> 00:16:43.050 align:middle line:84% And the family, in turn, is the discounted sum 00:16:43.050 --> 00:16:45.240 align:middle line:84% of their own utility plus the discounted sum 00:16:45.240 --> 00:16:50.240 align:middle line:84% of the utility values of the family in the future. 00:16:50.240 --> 00:16:53.800 align:middle line:84% So from that potentially very generic things, 00:16:53.800 --> 00:16:57.180 align:middle line:90% how does Barro-Becker think? 00:16:57.180 --> 00:17:02.110 align:middle line:84% What are the Barro preferences, if anyone knows? 00:17:02.110 --> 00:17:05.790 align:middle line:84% You would have seen them maybe in second-year macro 00:17:05.790 --> 00:17:09.970 align:middle line:84% or first-year macro at some point maybe. 00:17:09.970 --> 00:17:13.520 align:middle line:84% But then I haven't taken macro in a long time. 00:17:13.520 --> 00:17:14.095 align:middle line:90% So maybe. 00:17:14.095 --> 00:17:23.369 align:middle line:90% 00:17:23.369 --> 00:17:25.900 align:middle line:90% I'm sure you noticed. 00:17:25.900 --> 00:17:26.832 align:middle line:90% Yeah. 00:17:26.832 --> 00:17:30.696 align:middle line:84% STUDENT: Is it when you want to take [INAUDIBLE] solutions, 00:17:30.696 --> 00:17:35.103 align:middle line:84% so you would just have the sum of [INAUDIBLE]?? 00:17:35.103 --> 00:17:36.020 align:middle line:90% ESTHER DUFLO: Exactly. 00:17:36.020 --> 00:17:40.290 align:middle line:84% So the first thing is that you only care one generation above. 00:17:40.290 --> 00:17:47.060 align:middle line:84% So delta 1 is some number delta, which is the same as this one. 00:17:47.060 --> 00:17:51.980 align:middle line:84% You value the utility of the next generation 00:17:51.980 --> 00:17:54.080 align:middle line:84% and the family in the next generation 00:17:54.080 --> 00:17:57.730 align:middle line:90% 1 delta, the same delta. 00:17:57.730 --> 00:18:01.610 align:middle line:90% So it's your same. 00:18:01.610 --> 00:18:06.690 align:middle line:84% And then that delta also happens to be your discount rate. 00:18:06.690 --> 00:18:09.890 align:middle line:84% So you value the utility, the consumption of your child 00:18:09.890 --> 00:18:13.340 align:middle line:84% tomorrow at some discounted rate, 00:18:13.340 --> 00:18:15.090 align:middle line:84% which is the same as you apply to yourself 00:18:15.090 --> 00:18:16.806 align:middle line:90% for your retirement. 00:18:16.806 --> 00:18:19.380 align:middle line:84% So there is only one discount rate. 00:18:19.380 --> 00:18:22.470 align:middle line:84% There is no extra thing that comes from the fact 00:18:22.470 --> 00:18:24.360 align:middle line:84% that this is your children, this is not 00:18:24.360 --> 00:18:27.430 align:middle line:84% you, which would introduce some kind of beta delta 00:18:27.430 --> 00:18:31.450 align:middle line:84% if you care differently about other generation than yourself. 00:18:31.450 --> 00:18:33.960 align:middle line:84% You just have one discount rate, which applies 00:18:33.960 --> 00:18:37.590 align:middle line:90% to anything in the future. 00:18:37.590 --> 00:18:40.680 align:middle line:84% And you care only one generation below. 00:18:40.680 --> 00:18:42.780 align:middle line:84% So you care about the utility, not 00:18:42.780 --> 00:18:47.220 align:middle line:84% the consumption, the utility of the future generation. 00:18:47.220 --> 00:18:49.170 align:middle line:84% In turn, they care about the utility 00:18:49.170 --> 00:18:51.810 align:middle line:84% of the future generation, discounted delta. 00:18:51.810 --> 00:18:53.850 align:middle line:84% And in turn, the future generation 00:18:53.850 --> 00:18:55.900 align:middle line:84% care about the utility of the future generation, 00:18:55.900 --> 00:18:57.000 align:middle line:90% and so on, and so forth. 00:18:57.000 --> 00:18:59.730 align:middle line:90% 00:18:59.730 --> 00:19:01.470 align:middle line:84% And then there is another thing that 00:19:01.470 --> 00:19:04.095 align:middle line:84% the Barro-Becker preferences does, 00:19:04.095 --> 00:19:05.595 align:middle line:84% which is going to simplify our life. 00:19:05.595 --> 00:19:12.490 align:middle line:90% 00:19:12.490 --> 00:19:14.930 align:middle line:84% They remove symbolic consumption, 00:19:14.930 --> 00:19:18.560 align:middle line:84% so utilities is material consumption. 00:19:18.560 --> 00:19:27.130 align:middle line:84% So all of that very kindly simplifies the preferences 00:19:27.130 --> 00:19:29.260 align:middle line:90% in this recursive way. 00:19:29.260 --> 00:19:33.330 align:middle line:84% From here, you can just rewrite with-- 00:19:33.330 --> 00:19:41.260 align:middle line:84% you now have just UCt as delta Ut plus 1, p plus delta 00:19:41.260 --> 00:19:42.700 align:middle line:90% Ut plus 1C. 00:19:42.700 --> 00:19:49.600 align:middle line:84% And then you replace the plus 1C by t plus 2C, et cetera. 00:19:49.600 --> 00:19:58.630 align:middle line:84% And you have a sum of discounted utility with the discounted-- 00:19:58.630 --> 00:20:00.670 align:middle line:84% it should be a superscript, this guy. 00:20:00.670 --> 00:20:07.150 align:middle line:90% 00:20:07.150 --> 00:20:10.820 align:middle line:84% And then if we got rid of material consumption as well, 00:20:10.820 --> 00:20:13.040 align:middle line:90% we are just discounted. 00:20:13.040 --> 00:20:16.390 align:middle line:84% We are just getting a discounted sum of material consumption 00:20:16.390 --> 00:20:18.650 align:middle line:90% over time. 00:20:18.650 --> 00:20:21.760 align:middle line:84% So that makes an object that we are very used 00:20:21.760 --> 00:20:27.010 align:middle line:84% to play with because now, it is like one 00:20:27.010 --> 00:20:28.960 align:middle line:84% individual, the same type of things 00:20:28.960 --> 00:20:32.020 align:middle line:84% that we play with in the [? Ransom ?] model of one 00:20:32.020 --> 00:20:36.220 align:middle line:84% individual investing, and deciding how much to save 00:20:36.220 --> 00:20:42.400 align:middle line:84% and consume, and maximizing a discounted sum of utility 00:20:42.400 --> 00:20:44.380 align:middle line:90% to infinity. 00:20:44.380 --> 00:20:47.070 align:middle line:84% But that comes from drastically simplifying 00:20:47.070 --> 00:20:50.850 align:middle line:84% the model of how you care about the other generation. 00:20:50.850 --> 00:20:52.440 align:middle line:84% You just care about your children. 00:20:52.440 --> 00:20:54.373 align:middle line:90% You care one. 00:20:54.373 --> 00:20:56.040 align:middle line:84% The other thing that we are going to add 00:20:56.040 --> 00:21:03.360 align:middle line:84% is that anybody can borrow or lend at growth rate rt. 00:21:03.360 --> 00:21:06.600 align:middle line:84% There is no imperfect credit market. 00:21:06.600 --> 00:21:10.010 align:middle line:84% And there is no contracting between generations. 00:21:10.010 --> 00:21:12.362 align:middle line:84% So the parent can choose their own consumption, 00:21:12.362 --> 00:21:13.570 align:middle line:90% the middle age, and the old-- 00:21:13.570 --> 00:21:16.728 align:middle line:90% 00:21:16.728 --> 00:21:19.270 align:middle line:84% their own consumption in middle age and while they are older, 00:21:19.270 --> 00:21:21.310 align:middle line:84% the investment in their children, and that's it. 00:21:21.310 --> 00:21:26.260 align:middle line:84% They cannot dictate to them how they should invest in their own 00:21:26.260 --> 00:21:26.873 align:middle line:90% grandchildren. 00:21:26.873 --> 00:21:28.540 align:middle line:84% All they can do is doing the investment. 00:21:28.540 --> 00:21:29.980 align:middle line:84% And the child, in turn, will make 00:21:29.980 --> 00:21:31.550 align:middle line:90% the investment in the future. 00:21:31.550 --> 00:21:33.130 align:middle line:84% So there is no commitment problem 00:21:33.130 --> 00:21:34.570 align:middle line:84% because there is no decision that 00:21:34.570 --> 00:21:36.810 align:middle line:84% is made over the long run, no contracting possible. 00:21:36.810 --> 00:21:43.460 align:middle line:90% 00:21:43.460 --> 00:21:47.120 align:middle line:84% With Frank, you haven't done hyperbolic discounting 00:21:47.120 --> 00:21:51.260 align:middle line:84% and inconsistent time preferences 00:21:51.260 --> 00:21:55.280 align:middle line:84% for the individual, the very famous Beta Delta model. 00:21:55.280 --> 00:21:58.820 align:middle line:84% But we're going to see it when we study savings. 00:21:58.820 --> 00:22:01.340 align:middle line:84% But once we have studied it, you can 00:22:01.340 --> 00:22:02.990 align:middle line:84% see that none of these complexities 00:22:02.990 --> 00:22:05.690 align:middle line:84% are there in the Barro-Becker model because 00:22:05.690 --> 00:22:06.920 align:middle line:90% of this simplification. 00:22:06.920 --> 00:22:08.720 align:middle line:84% And if we didn't have this simplification, 00:22:08.720 --> 00:22:10.700 align:middle line:84% we would find ourselves in a world, 00:22:10.700 --> 00:22:13.730 align:middle line:84% in a sense, of commitment problem, 00:22:13.730 --> 00:22:15.800 align:middle line:84% contracting problem, which are very 00:22:15.800 --> 00:22:20.360 align:middle line:84% similar to the type of issues that we think about when 00:22:20.360 --> 00:22:25.760 align:middle line:84% an individual is trying to deal with a lot of impatience 00:22:25.760 --> 00:22:29.510 align:middle line:84% in the short run, for example, or an attempt 00:22:29.510 --> 00:22:32.880 align:middle line:84% to commit their future selves to behave in such and such way. 00:22:32.880 --> 00:22:36.650 align:middle line:84% So the same issues will appear within the family 00:22:36.650 --> 00:22:38.550 align:middle line:90% as soon as you have. 00:22:38.550 --> 00:22:40.640 align:middle line:84% And you can see even more easily, in a sense, 00:22:40.640 --> 00:22:44.990 align:middle line:84% that you don't have to have any funky preferences like we 00:22:44.990 --> 00:22:46.920 align:middle line:84% need to have at the individual level. 00:22:46.920 --> 00:22:49.370 align:middle line:84% It is just like if you care about your grandchildren, 00:22:49.370 --> 00:22:50.990 align:middle line:84% everything's messed up in this model. 00:22:50.990 --> 00:22:54.340 align:middle line:90% 00:22:54.340 --> 00:22:57.370 align:middle line:84% By the way, similarly in political economy, 00:22:57.370 --> 00:23:02.410 align:middle line:84% we have these models of how a country might decide 00:23:02.410 --> 00:23:06.250 align:middle line:84% to take a debt that will apply to the future citizens 00:23:06.250 --> 00:23:10.120 align:middle line:84% of that country and how you make the commitment in the policy 00:23:10.120 --> 00:23:10.780 align:middle line:90% space. 00:23:10.780 --> 00:23:13.750 align:middle line:84% It's, again, the same issues of what contracting 00:23:13.750 --> 00:23:15.910 align:middle line:90% is possible between generations. 00:23:15.910 --> 00:23:21.430 align:middle line:84% And how do we take into account our caring for the future? 00:23:21.430 --> 00:23:23.500 align:middle line:84% And it's only in this very simplified 00:23:23.500 --> 00:23:28.950 align:middle line:84% model where that things work out analytically. 00:23:28.950 --> 00:23:31.530 align:middle line:90% So you can borrow. 00:23:31.530 --> 00:23:34.770 align:middle line:84% Credit and loan can be bequeathed. 00:23:34.770 --> 00:23:37.530 align:middle line:84% So that simplifies our life a lot. 00:23:37.530 --> 00:23:42.060 align:middle line:84% So we start that set that gives us a very standard credit 00:23:42.060 --> 00:23:45.210 align:middle line:84% constraint, where defined in terms of wealth, 00:23:45.210 --> 00:23:48.570 align:middle line:84% how much financial assets someone has, 00:23:48.570 --> 00:23:52.260 align:middle line:84% which could be positive assets or negative assets. 00:23:52.260 --> 00:23:55.930 align:middle line:84% So the wealth that you're going to have in the future 00:23:55.930 --> 00:23:58.610 align:middle line:84% is r, which is the gross interest rate. 00:23:58.610 --> 00:24:03.850 align:middle line:84% So suppose if you have any assets 00:24:03.850 --> 00:24:06.555 align:middle line:84% after the end of this operation, you put it in the bank. 00:24:06.555 --> 00:24:07.930 align:middle line:84% And next period, they'll give you 00:24:07.930 --> 00:24:11.720 align:middle line:84% that multiplied by the gross interest rate, so 1 00:24:11.720 --> 00:24:13.870 align:middle line:90% plus something. 00:24:13.870 --> 00:24:15.240 align:middle line:90% And what do you put in the bank? 00:24:15.240 --> 00:24:16.470 align:middle line:90% Or what do you borrow? 00:24:16.470 --> 00:24:20.380 align:middle line:84% Well, you start with some wealth. 00:24:20.380 --> 00:24:20.890 align:middle line:90% You consume. 00:24:20.890 --> 00:24:23.510 align:middle line:90% 00:24:23.510 --> 00:24:27.050 align:middle line:90% You feed the generation before. 00:24:27.050 --> 00:24:30.720 align:middle line:90% They've made a commitment. 00:24:30.720 --> 00:24:32.100 align:middle line:90% You earn some money. 00:24:32.100 --> 00:24:36.870 align:middle line:84% That is your unit for your unit of unskilled 00:24:36.870 --> 00:24:42.270 align:middle line:84% and for however you need of skills you happen to have, 00:24:42.270 --> 00:24:46.230 align:middle line:84% ht multiplied by the rate of return to education. 00:24:46.230 --> 00:24:50.310 align:middle line:84% And then you spend some money to educate your kid. 00:24:50.310 --> 00:24:52.560 align:middle line:84% So if you want your kid to have ht plus 1, 00:24:52.560 --> 00:24:56.460 align:middle line:84% you're going to have to spend phi of ht plus 1ht, 00:24:56.460 --> 00:25:04.980 align:middle line:84% which, well, this phi is just a shortcut for the equation we 00:25:04.980 --> 00:25:05.670 align:middle line:90% wrote before. 00:25:05.670 --> 00:25:09.130 align:middle line:90% 00:25:09.130 --> 00:25:14.280 align:middle line:84% So that's how much you're going to have in the next period. 00:25:14.280 --> 00:25:16.340 align:middle line:84% So now, we have our utility function. 00:25:16.340 --> 00:25:17.720 align:middle line:90% We have our budget constraint. 00:25:17.720 --> 00:25:22.110 align:middle line:84% How do we go about solving this problem? 00:25:22.110 --> 00:25:25.260 align:middle line:84% What are the three equations that 00:25:25.260 --> 00:25:29.955 align:middle line:84% are going to help us solve this issue, this maximization 00:25:29.955 --> 00:25:30.455 align:middle line:90% problem? 00:25:30.455 --> 00:25:42.890 align:middle line:90% 00:25:42.890 --> 00:25:46.400 align:middle line:84% So this, I know you have seen in macro, even in today's times. 00:25:46.400 --> 00:25:52.920 align:middle line:90% 00:25:52.920 --> 00:25:53.420 align:middle line:90% Yeah? 00:25:53.420 --> 00:25:55.293 align:middle line:84% STUDENT: Maybe it's the [INAUDIBLE].. 00:25:55.293 --> 00:25:56.210 align:middle line:90% ESTHER DUFLO: Exactly. 00:25:56.210 --> 00:26:02.390 align:middle line:84% So we get to Euler equation, one with respect to ct, 00:26:02.390 --> 00:26:05.015 align:middle line:90% one with respect to pt plus 1. 00:26:05.015 --> 00:26:07.970 align:middle line:90% 00:26:07.970 --> 00:26:11.720 align:middle line:84% And then we are going to take the first-order condition 00:26:11.720 --> 00:26:14.570 align:middle line:90% on the budget constraint. 00:26:14.570 --> 00:26:17.870 align:middle line:84% I'm not going to write the ct and the Euler equation 00:26:17.870 --> 00:26:19.742 align:middle line:90% for the two consumption levels. 00:26:19.742 --> 00:26:20.700 align:middle line:90% They are what they are. 00:26:20.700 --> 00:26:22.470 align:middle line:90% And they are not interesting. 00:26:22.470 --> 00:26:23.600 align:middle line:90% I mean, they are. 00:26:23.600 --> 00:26:25.625 align:middle line:84% They would be interesting in another world. 00:26:25.625 --> 00:26:27.500 align:middle line:84% But the reason I'm not going to write them up 00:26:27.500 --> 00:26:29.180 align:middle line:84% is that if you did write them up, 00:26:29.180 --> 00:26:32.300 align:middle line:84% you would notice that the human capital doesn't 00:26:32.300 --> 00:26:35.210 align:middle line:90% appear in them anywhere. 00:26:35.210 --> 00:26:38.780 align:middle line:84% The Euler equation is defined in terms of consumption, 00:26:38.780 --> 00:26:42.130 align:middle line:84% and the interest rate, and your discount rate. 00:26:42.130 --> 00:26:44.050 align:middle line:84% So depending how impatient you are, 00:26:44.050 --> 00:26:48.550 align:middle line:84% you're going to spend more on consumption 00:26:48.550 --> 00:26:54.490 align:middle line:84% or less for this, while you're working, 00:26:54.490 --> 00:26:56.990 align:middle line:90% and also for next year. 00:26:56.990 --> 00:27:02.770 align:middle line:84% Meanwhile, the only place where the human capital decision 00:27:02.770 --> 00:27:09.780 align:middle line:84% is going to enter is in the budget constraint. 00:27:09.780 --> 00:27:16.550 align:middle line:84% So to figure out the level of the human capital investment, 00:27:16.550 --> 00:27:19.460 align:middle line:84% it's ht plus 1 we are trying to determine because ht 00:27:19.460 --> 00:27:21.740 align:middle line:90% is what we already have. 00:27:21.740 --> 00:27:22.700 align:middle line:90% What do we do? 00:27:22.700 --> 00:27:25.990 align:middle line:90% 00:27:25.990 --> 00:27:27.460 align:middle line:90% Sure. 00:27:27.460 --> 00:27:30.773 align:middle line:90% STUDENT: Is it just 00:27:30.773 --> 00:27:32.190 align:middle line:84% ESTHER DUFLO: Yeah, so we're going 00:27:32.190 --> 00:27:36.130 align:middle line:84% to take the derivative on both sides. 00:27:36.130 --> 00:27:38.400 align:middle line:84% And basically, what it says in plane 1 00:27:38.400 --> 00:27:43.500 align:middle line:84% is that we want to equalize the marginal return of investing 00:27:43.500 --> 00:27:49.500 align:middle line:84% in one more unit of ht plus 1 to its cost. 00:27:49.500 --> 00:27:57.090 align:middle line:84% And just before we even do the calculation, why 00:27:57.090 --> 00:27:58.650 align:middle line:90% is it interesting to invest? 00:27:58.650 --> 00:28:01.560 align:middle line:84% What is the value to invest in ht plus 1? 00:28:01.560 --> 00:28:04.750 align:middle line:84% Where are we going to get bang for our buck 00:28:04.750 --> 00:28:06.670 align:middle line:90% when we invest in ht plus 1? 00:28:06.670 --> 00:28:10.891 align:middle line:90% 00:28:10.891 --> 00:28:11.830 align:middle line:90% Tyler. 00:28:11.830 --> 00:28:13.360 align:middle line:90% STUDENT: Child's consumption. 00:28:13.360 --> 00:28:14.110 align:middle line:90% ESTHER DUFLO: Yes. 00:28:14.110 --> 00:28:20.200 align:middle line:84% The child's wage is going to be higher, and therefore, later, 00:28:20.200 --> 00:28:20.950 align:middle line:90% their consumption. 00:28:20.950 --> 00:28:22.408 align:middle line:84% But in the budget constraints, it's 00:28:22.408 --> 00:28:23.890 align:middle line:90% going to show up at their wage. 00:28:23.890 --> 00:28:25.670 align:middle line:90% And then what else? 00:28:25.670 --> 00:28:26.170 align:middle line:90% Yeah. 00:28:26.170 --> 00:28:27.500 align:middle line:90% STUDENT: Ht plus 1. 00:28:27.500 --> 00:28:28.250 align:middle line:90% ESTHER DUFLO: Yes. 00:28:28.250 --> 00:28:31.760 align:middle line:84% So again, so to their wage, then that's going to give them money 00:28:31.760 --> 00:28:32.260 align:middle line:90% too. 00:28:32.260 --> 00:28:34.840 align:middle line:84% But first, the value of investing in human capital 00:28:34.840 --> 00:28:37.110 align:middle line:90% is that they'll earn more. 00:28:37.110 --> 00:28:40.440 align:middle line:84% And what's the second value investing 00:28:40.440 --> 00:28:43.500 align:middle line:84% human capital, coming straight out of the budget constraint? 00:28:43.500 --> 00:28:46.362 align:middle line:90% 00:28:46.362 --> 00:28:47.977 align:middle line:90% STUDENT: Would it be cheaper? 00:28:47.977 --> 00:28:48.810 align:middle line:90% ESTHER DUFLO: Sorry? 00:28:48.810 --> 00:28:50.130 align:middle line:84% STUDENT: Would it be cheaper for their children? 00:28:50.130 --> 00:28:51.047 align:middle line:90% ESTHER DUFLO: Exactly. 00:28:51.047 --> 00:28:53.680 align:middle line:84% It will be cheaper for them to educate their own children. 00:28:53.680 --> 00:28:55.170 align:middle line:84% So these are the two places we are 00:28:55.170 --> 00:28:58.200 align:middle line:84% going to see a value investing human capital. 00:28:58.200 --> 00:29:02.730 align:middle line:84% And that, in turn, is going to translate into their higher 00:29:02.730 --> 00:29:06.750 align:middle line:84% wage, going to translate into more consumption for them 00:29:06.750 --> 00:29:08.340 align:middle line:90% and more return. 00:29:08.340 --> 00:29:10.200 align:middle line:84% But we are not going to see it later. 00:29:10.200 --> 00:29:16.120 align:middle line:90% 00:29:16.120 --> 00:29:21.150 align:middle line:84% So what you can do is to take the derivative of this, 00:29:21.150 --> 00:29:27.480 align:middle line:84% and then the derivative of this, and then rearrange the terms 00:29:27.480 --> 00:29:28.420 align:middle line:90% a little bit. 00:29:28.420 --> 00:29:31.810 align:middle line:84% And you get exactly what we were discussing. 00:29:31.810 --> 00:29:44.890 align:middle line:84% So you get-- it costs some amount of money to get. 00:29:44.890 --> 00:29:47.680 align:middle line:90% 00:29:47.680 --> 00:29:53.040 align:middle line:84% This is the marginal cost of getting ht plus 1. 00:29:53.040 --> 00:29:57.735 align:middle line:84% This is the first s1, meaning d of the first derivative, 00:29:57.735 --> 00:30:00.520 align:middle line:84% the derivative with respect to this first term. 00:30:00.520 --> 00:30:05.220 align:middle line:84% So that's the marginal cost of one more unit of human capital 00:30:05.220 --> 00:30:11.580 align:middle line:84% for your kid is equal to the marginal return, which you're 00:30:11.580 --> 00:30:12.880 align:middle line:90% getting in the next period. 00:30:12.880 --> 00:30:15.530 align:middle line:84% So it's divided by the interest rate. 00:30:15.530 --> 00:30:23.460 align:middle line:84% Extra wage for them, plus reduction and the minus 00:30:23.460 --> 00:30:26.920 align:middle line:84% in the marginal cost of educating their own children. 00:30:26.920 --> 00:30:30.390 align:middle line:90% So it's now s2 of ht plus 1. 00:30:30.390 --> 00:30:35.940 align:middle line:84% So you can, straight from here, take the derivative 00:30:35.940 --> 00:30:39.990 align:middle line:84% of this bit and that bit with respect to ht plus 1 00:30:39.990 --> 00:30:41.070 align:middle line:90% and rearrange the term. 00:30:41.070 --> 00:30:42.090 align:middle line:90% And you'll get this. 00:30:42.090 --> 00:30:44.670 align:middle line:84% But this is the intuition that you 00:30:44.670 --> 00:30:47.840 align:middle line:84% want to keep in mind is that that's 00:30:47.840 --> 00:30:49.190 align:middle line:90% how you're going to define. 00:30:49.190 --> 00:30:51.920 align:middle line:84% That's how you're going to decide how much to spend 00:30:51.920 --> 00:30:54.200 align:middle line:90% in your own human capital. 00:30:54.200 --> 00:30:59.270 align:middle line:84% So what is interesting is what's not in there at all. 00:30:59.270 --> 00:31:04.120 align:middle line:90% 00:31:04.120 --> 00:31:09.240 align:middle line:84% So this equation, we can rearrange a little bit. 00:31:09.240 --> 00:31:13.320 align:middle line:84% If we are in steady state we can replace 1 over r by delta 00:31:13.320 --> 00:31:15.960 align:middle line:84% because that's what's going to determine the interest 00:31:15.960 --> 00:31:17.220 align:middle line:90% rate in steady state. 00:31:17.220 --> 00:31:21.180 align:middle line:84% We can replace the ht plus 1, ht plus 2 by hs. 00:31:21.180 --> 00:31:26.820 align:middle line:84% But even in this particular equation, what is not there? 00:31:26.820 --> 00:31:29.450 align:middle line:90% 00:31:29.450 --> 00:31:29.950 align:middle line:90% Yeah. 00:31:29.950 --> 00:31:31.280 align:middle line:90% STUDENT: [INAUDIBLE] 00:31:31.280 --> 00:31:33.720 align:middle line:84% ESTHER DUFLO: Yeah, or what else? 00:31:33.720 --> 00:31:34.290 align:middle line:90% Yes? 00:31:34.290 --> 00:31:36.170 align:middle line:84% STUDENT: And I don't know, like, all the other things 00:31:36.170 --> 00:31:37.045 align:middle line:90% that were in utility. 00:31:37.045 --> 00:31:41.160 align:middle line:84% ESTHER DUFLO: It is nothing about the utility function. 00:31:41.160 --> 00:31:42.840 align:middle line:84% There's nothing about the utility 00:31:42.840 --> 00:31:44.910 align:middle line:90% function of the parents. 00:31:44.910 --> 00:31:48.030 align:middle line:84% It's got nothing to do with their taste. 00:31:48.030 --> 00:31:50.790 align:middle line:84% It's got nothing to do even with how patient or impatient 00:31:50.790 --> 00:31:51.780 align:middle line:90% they are. 00:31:51.780 --> 00:31:53.400 align:middle line:90% There is a 1 minus there. 00:31:53.400 --> 00:31:56.040 align:middle line:84% But from the individual point of view of the parent, 00:31:56.040 --> 00:32:01.020 align:middle line:84% this is given by whatever happens to be the interest 00:32:01.020 --> 00:32:02.740 align:middle line:90% rate in society. 00:32:02.740 --> 00:32:07.050 align:middle line:84% So any individual parent, however impatient, 00:32:07.050 --> 00:32:11.970 align:middle line:84% or nasty, or whatever, is going to spend the same amount 00:32:11.970 --> 00:32:14.590 align:middle line:90% in educating their children. 00:32:14.590 --> 00:32:17.030 align:middle line:90% And where does this come from? 00:32:17.030 --> 00:32:18.220 align:middle line:90% Why is that result? 00:32:18.220 --> 00:32:20.980 align:middle line:84% Why is it 2, irrespective of the math? 00:32:20.980 --> 00:32:25.160 align:middle line:84% What's the intuition that makes this true? 00:32:25.160 --> 00:32:26.390 align:middle line:90% This is quite powerful. 00:32:26.390 --> 00:32:29.540 align:middle line:84% You would think that doesn't really 00:32:29.540 --> 00:32:30.740 align:middle line:90% completely fit with our-- 00:32:30.740 --> 00:32:33.904 align:middle line:90% 00:32:33.904 --> 00:32:35.590 align:middle line:84% STUDENT: I think you were assuming 00:32:35.590 --> 00:32:37.840 align:middle line:84% that it's a steady state so with that it's short. 00:32:37.840 --> 00:32:38.680 align:middle line:90% ESTHER DUFLO: Right. 00:32:38.680 --> 00:32:39.790 align:middle line:90% This is not steady state. 00:32:39.790 --> 00:32:41.050 align:middle line:90% It's full of h plus 1. 00:32:41.050 --> 00:32:42.950 align:middle line:84% It's the individual-level decision today. 00:32:42.950 --> 00:32:44.488 align:middle line:90% There is no steady state yet. 00:32:44.488 --> 00:32:45.530 align:middle line:90% We're going to get there. 00:32:45.530 --> 00:32:53.000 align:middle line:84% But in this model, in the first instance, another thing that's 00:32:53.000 --> 00:32:54.690 align:middle line:90% striking, everyone will-- 00:32:54.690 --> 00:32:57.190 align:middle line:90% yeah. 00:32:57.190 --> 00:32:58.090 align:middle line:90% Yes? 00:32:58.090 --> 00:33:01.064 align:middle line:84% STUDENT: My guess is that it's because the returns are all 00:33:01.064 --> 00:33:03.238 align:middle line:90% the same across all people? 00:33:03.238 --> 00:33:06.040 align:middle line:90% 00:33:06.040 --> 00:33:07.165 align:middle line:90% ESTHER DUFLO: The returns-- 00:33:07.165 --> 00:33:11.660 align:middle line:90% 00:33:11.660 --> 00:33:12.920 align:middle line:90% the function is the same. 00:33:12.920 --> 00:33:15.140 align:middle line:84% But potentially, they have a different level 00:33:15.140 --> 00:33:16.730 align:middle line:90% of human capital themselves. 00:33:16.730 --> 00:33:19.270 align:middle line:90% 00:33:19.270 --> 00:33:19.770 align:middle line:90% Yeah? 00:33:19.770 --> 00:33:21.520 align:middle line:84% STUDENT: Is it because [? the insurance ?] 00:33:21.520 --> 00:33:23.853 align:middle line:90% on savings and [INAUDIBLE]? 00:33:23.853 --> 00:33:24.770 align:middle line:90% ESTHER DUFLO: Exactly. 00:33:24.770 --> 00:33:28.100 align:middle line:84% There is this other thing they could do at any point. 00:33:28.100 --> 00:33:29.090 align:middle line:90% They can save. 00:33:29.090 --> 00:33:30.530 align:middle line:90% And they can borrow. 00:33:30.530 --> 00:33:33.560 align:middle line:84% They can borrow against the earning of their children. 00:33:33.560 --> 00:33:36.170 align:middle line:84% Or they can save the earning of their children. 00:33:36.170 --> 00:33:39.950 align:middle line:84% Therefore, the utility function is 00:33:39.950 --> 00:33:43.780 align:middle line:84% going to determine how much they want to save. 00:33:43.780 --> 00:33:50.500 align:middle line:84% But then how this is separated between human capital and debt 00:33:50.500 --> 00:33:55.900 align:middle line:84% or assets, there is no choice involved. 00:33:55.900 --> 00:33:59.020 align:middle line:84% Therefore, the only thing that matters 00:33:59.020 --> 00:34:00.340 align:middle line:90% is how productive it is. 00:34:00.340 --> 00:34:03.010 align:middle line:84% So you have the return to education 00:34:03.010 --> 00:34:07.270 align:middle line:84% in terms of their earning of their kid 00:34:07.270 --> 00:34:11.770 align:middle line:84% as well as in terms of how cheaper it 00:34:11.770 --> 00:34:13.870 align:middle line:90% is to invest your future kids. 00:34:13.870 --> 00:34:17.590 align:middle line:84% But the only thing that matters is the returns to education, 00:34:17.590 --> 00:34:21.659 align:middle line:84% and how that compares to the cost, and nothing else 00:34:21.659 --> 00:34:25.389 align:middle line:84% because a more patient parent will invest less. 00:34:25.389 --> 00:34:31.600 align:middle line:84% But that will all go through how much they put in assets. 00:34:31.600 --> 00:34:36.420 align:middle line:84% There is no reason to not do the efficient thing 00:34:36.420 --> 00:34:38.050 align:middle line:90% with respect to education. 00:34:38.050 --> 00:34:42.540 align:middle line:84% So that's a pretty striking result. 00:34:42.540 --> 00:34:44.290 align:middle line:84% There is no parental preference effect. 00:34:44.290 --> 00:34:47.050 align:middle line:90% There is also no income effect. 00:34:47.050 --> 00:34:50.130 align:middle line:84% In fact, in this particular model, in one shot, 00:34:50.130 --> 00:34:53.060 align:middle line:90% everybody invest the same thing. 00:34:53.060 --> 00:34:56.860 align:middle line:84% And in steady state, r is equal to the inverted 00:34:56.860 --> 00:35:00.550 align:middle line:84% of the discount rate and the two human capital levels. 00:35:00.550 --> 00:35:03.430 align:middle line:84% All the human capital become constant. 00:35:03.430 --> 00:35:08.080 align:middle line:84% And that gives you an expression for the human capital that's 00:35:08.080 --> 00:35:09.790 align:middle line:90% chosen in a generation. 00:35:09.790 --> 00:35:11.350 align:middle line:90% And there, you have delta. 00:35:11.350 --> 00:35:14.300 align:middle line:84% So you have a preference effect in steady state. 00:35:14.300 --> 00:35:16.570 align:middle line:84% But the delta comes from because it's 00:35:16.570 --> 00:35:20.300 align:middle line:84% saying how patient is society is, and therefore, 00:35:20.300 --> 00:35:22.780 align:middle line:84% how much they want to save overall 00:35:22.780 --> 00:35:26.480 align:middle line:84% in the form of either human capital or physical capital. 00:35:26.480 --> 00:35:31.000 align:middle line:84% So at the level of society, more patient society 00:35:31.000 --> 00:35:33.340 align:middle line:84% will save more, both in physical form 00:35:33.340 --> 00:35:35.950 align:middle line:84% and in the form of human capital. 00:35:35.950 --> 00:35:38.200 align:middle line:84% And otherwise, all you have in here 00:35:38.200 --> 00:35:44.100 align:middle line:84% is how expensive it is to educate someone 00:35:44.100 --> 00:35:50.290 align:middle line:84% in terms of educated and uneducated labor 00:35:50.290 --> 00:35:56.140 align:middle line:84% and what are the returns to education, which enters here. 00:35:56.140 --> 00:35:58.077 align:middle line:84% So if the returns to education are higher, 00:35:58.077 --> 00:35:59.410 align:middle line:90% people are going to invest more. 00:35:59.410 --> 00:36:04.910 align:middle line:84% If it's going to become more expensive, they'll invest less. 00:36:04.910 --> 00:36:06.950 align:middle line:84% There's also no inequality in this model. 00:36:06.950 --> 00:36:09.750 align:middle line:84% Everyone will converge to the same thing, 00:36:09.750 --> 00:36:12.050 align:middle line:84% in fact, in one generation, because they're all going 00:36:12.050 --> 00:36:15.140 align:middle line:90% to pick the same ht plus 1. 00:36:15.140 --> 00:36:16.790 align:middle line:90% And that breaks down. 00:36:16.790 --> 00:36:18.590 align:middle line:84% One way in which this breaks down 00:36:18.590 --> 00:36:23.260 align:middle line:84% is if s12 is very negative, where, 00:36:23.260 --> 00:36:25.000 align:middle line:84% if you start from a very different level 00:36:25.000 --> 00:36:27.310 align:middle line:84% of human capital, your cost of education 00:36:27.310 --> 00:36:28.810 align:middle line:90% might be quite different. 00:36:28.810 --> 00:36:31.200 align:middle line:84% And then you might end up with two different statistics. 00:36:31.200 --> 00:36:36.560 align:middle line:84% So that's, again, just to repeat the Galor-Zeira model. 00:36:36.560 --> 00:36:42.710 align:middle line:84% But otherwise, if you don't have this extreme complementarity, 00:36:42.710 --> 00:36:44.080 align:middle line:90% you're good. 00:36:44.080 --> 00:36:46.600 align:middle line:90% There is unique steady state. 00:36:46.600 --> 00:36:49.840 align:middle line:84% So there is no longer an effect of a large shock 00:36:49.840 --> 00:36:52.200 align:middle line:90% to human capital. 00:36:52.200 --> 00:36:58.560 align:middle line:84% So we are discussing with David Atkin about COVID 00:36:58.560 --> 00:37:00.330 align:middle line:84% and saying, oh, all these kids who are not 00:37:00.330 --> 00:37:01.680 align:middle line:90% going to school this year. 00:37:01.680 --> 00:37:04.050 align:middle line:90% Oh, it's a disaster. 00:37:04.050 --> 00:37:08.010 align:middle line:84% In this model, it's just going to be sorted very quickly 00:37:08.010 --> 00:37:10.320 align:middle line:84% because the return to human capital assets-- 00:37:10.320 --> 00:37:14.353 align:middle line:84% that investment will take place when schools become open again, 00:37:14.353 --> 00:37:15.645 align:middle line:90% at least as a level of society. 00:37:15.645 --> 00:37:18.210 align:middle line:90% 00:37:18.210 --> 00:37:20.325 align:middle line:84% And interestingly, that was his immediate-- 00:37:20.325 --> 00:37:23.220 align:middle line:84% I guess that part was kind of his instinctive reaction, 00:37:23.220 --> 00:37:27.470 align:middle line:84% showing how powerful this model is for us. 00:37:27.470 --> 00:37:30.460 align:middle line:84% So in this case, you can think very quickly 00:37:30.460 --> 00:37:35.980 align:middle line:84% in the framework of this very simple model about your policy 00:37:35.980 --> 00:37:37.870 align:middle line:90% instrument. 00:37:37.870 --> 00:37:41.560 align:middle line:84% So you could have an education subsidy. 00:37:41.560 --> 00:37:45.370 align:middle line:84% For example, if you spend E, you get some amount. 00:37:45.370 --> 00:37:48.550 align:middle line:84% Say, primary education is free, so even if you spend 0, 00:37:48.550 --> 00:37:49.780 align:middle line:90% you'll get e0. 00:37:49.780 --> 00:37:53.980 align:middle line:84% And then you get tuition subsidies, tuition discount, 00:37:53.980 --> 00:37:55.630 align:middle line:90% depending where you go. 00:37:55.630 --> 00:37:59.260 align:middle line:84% And then you might have a tax, which it could be 00:37:59.260 --> 00:38:01.010 align:middle line:90% a progressive income taxation. 00:38:01.010 --> 00:38:06.270 align:middle line:90% So it's a tax on human capital. 00:38:06.270 --> 00:38:10.540 align:middle line:84% If we take this, it's very easy to put in the steady state. 00:38:10.540 --> 00:38:13.710 align:middle line:84% The steady state become this one. 00:38:13.710 --> 00:38:18.240 align:middle line:84% So your return to education is now smaller. 00:38:18.240 --> 00:38:21.000 align:middle line:84% The direct return to education is smaller because 00:38:21.000 --> 00:38:22.800 align:middle line:90% of the tax on human capital. 00:38:22.800 --> 00:38:25.260 align:middle line:84% But the cost of education is also smaller 00:38:25.260 --> 00:38:27.630 align:middle line:90% because you now don't have to-- 00:38:27.630 --> 00:38:31.920 align:middle line:84% you're getting this subsidy every time you educate more. 00:38:31.920 --> 00:38:34.820 align:middle line:84% And then this stays the same as before. 00:38:34.820 --> 00:38:37.180 align:middle line:90% So what's the first observation? 00:38:37.180 --> 00:38:47.010 align:middle line:84% Where do you see an e0 here and tau 0 in this formula? 00:38:47.010 --> 00:38:50.020 align:middle line:90% 00:38:50.020 --> 00:38:50.520 align:middle line:90% Yeah? 00:38:50.520 --> 00:38:52.628 align:middle line:84% STUDENT: These are a lump sum as taxes. 00:38:52.628 --> 00:38:53.420 align:middle line:90% ESTHER DUFLO: Yeah. 00:38:53.420 --> 00:38:54.253 align:middle line:90% They are a lump sum. 00:38:54.253 --> 00:38:55.800 align:middle line:90% They don't bear out there. 00:38:55.800 --> 00:38:59.390 align:middle line:84% So in this model, subsidizing primary of education 00:38:59.390 --> 00:39:00.260 align:middle line:90% is irrelevant. 00:39:00.260 --> 00:39:03.140 align:middle line:90% 00:39:03.140 --> 00:39:09.090 align:middle line:84% This is lump sum, doesn't matter to the choice. 00:39:09.090 --> 00:39:11.760 align:middle line:84% And then a second thing is this one. 00:39:11.760 --> 00:39:25.740 align:middle line:90% 00:39:25.740 --> 00:39:31.460 align:middle line:84% What does enter is the marginal subsidy and the marginal cost. 00:39:31.460 --> 00:39:34.300 align:middle line:84% So it would not make sense in this model 00:39:34.300 --> 00:39:39.860 align:middle line:84% to pay a tax in order to finance education. 00:39:39.860 --> 00:39:44.320 align:middle line:84% That's just going to be distorting people. 00:39:44.320 --> 00:39:46.870 align:middle line:84% But you might want to have a subsidy if you need 00:39:46.870 --> 00:39:49.270 align:middle line:90% a tax for some other reason. 00:39:49.270 --> 00:39:54.870 align:middle line:84% So the idea is that if you want a tax on education because you 00:39:54.870 --> 00:39:57.240 align:middle line:84% want to have progressive taxation to do something 00:39:57.240 --> 00:40:05.220 align:middle line:84% on the other, then it might make sense 00:40:05.220 --> 00:40:08.270 align:middle line:84% to have a subsidy for education that is going to, 00:40:08.270 --> 00:40:17.260 align:middle line:84% to some extent, outweigh the cost of education. 00:40:17.260 --> 00:40:19.030 align:middle line:84% So the proportional tax and subsidies 00:40:19.030 --> 00:40:20.170 align:middle line:90% have the predicted effect. 00:40:20.170 --> 00:40:22.870 align:middle line:84% Ideally, you are going to put the optimal. 00:40:22.870 --> 00:40:25.690 align:middle line:84% You go back to optimal if you have the subsidies 00:40:25.690 --> 00:40:27.250 align:middle line:90% the same as the tax. 00:40:27.250 --> 00:40:29.680 align:middle line:84% So in the absence of any other reason to tax, 00:40:29.680 --> 00:40:31.660 align:middle line:90% you want laissez faire. 00:40:31.660 --> 00:40:33.520 align:middle line:84% So education subsidy might be a way 00:40:33.520 --> 00:40:35.830 align:middle line:84% to contract on tax on human capital. 00:40:35.830 --> 00:40:41.690 align:middle line:84% That's there because of the necessity to have progressive. 00:40:41.690 --> 00:40:45.200 align:middle line:84% Now, something that is true here as well 00:40:45.200 --> 00:40:48.140 align:middle line:84% is that these are the return to education in a way 00:40:48.140 --> 00:40:52.970 align:middle line:84% because that's the ratio of the unskilled to the skilled labor. 00:40:52.970 --> 00:40:57.700 align:middle line:84% The higher return to human capital means more investment. 00:40:57.700 --> 00:41:01.320 align:middle line:84% So that's the best education policy. 00:41:01.320 --> 00:41:04.320 align:middle line:84% And if you listen to Mark Rosensweig, 00:41:04.320 --> 00:41:06.150 align:middle line:90% he says that very explicitly. 00:41:06.150 --> 00:41:09.330 align:middle line:84% The best education policy is no education policy. 00:41:09.330 --> 00:41:13.230 align:middle line:84% The best education policy is one which 00:41:13.230 --> 00:41:17.130 align:middle line:84% is in the labor market, where it becomes more 00:41:17.130 --> 00:41:20.930 align:middle line:90% valuable to get an education. 00:41:20.930 --> 00:41:25.070 align:middle line:84% So in terms of things that we want to look at empirically, 00:41:25.070 --> 00:41:27.290 align:middle line:84% one, we want to look at whether people 00:41:27.290 --> 00:41:31.400 align:middle line:84% are sensitive to the cost of education. 00:41:31.400 --> 00:41:34.950 align:middle line:84% Direct or other subsidies to education. 00:41:34.950 --> 00:41:36.630 align:middle line:84% One is we want to see whether people 00:41:36.630 --> 00:41:39.610 align:middle line:84% are sensitive to the returns to education. 00:41:39.610 --> 00:41:42.700 align:middle line:84% Is there, as they see it, the real return or the passive 00:41:42.700 --> 00:41:45.250 align:middle line:90% return to education? 00:41:45.250 --> 00:41:48.440 align:middle line:84% Those are kind of the main prediction 00:41:48.440 --> 00:41:50.960 align:middle line:84% of this model in term of how should 00:41:50.960 --> 00:41:57.230 align:middle line:84% decision-- that we should see how so being responsive to. 00:41:57.230 --> 00:41:58.820 align:middle line:84% So to summarize this model, and then 00:41:58.820 --> 00:42:03.240 align:middle line:84% we can start discussing it, this Barro-Becker world is great. 00:42:03.240 --> 00:42:05.930 align:middle line:84% There is no inequality in the steady state. 00:42:05.930 --> 00:42:10.933 align:middle line:84% Wealth, income, even preferences don't matter in the short run. 00:42:10.933 --> 00:42:12.350 align:middle line:84% And the only thing that determines 00:42:12.350 --> 00:42:16.250 align:middle line:84% education in this world is the return on the cost. 00:42:16.250 --> 00:42:19.780 align:middle line:90% 00:42:19.780 --> 00:42:21.850 align:middle line:84% So why might education be different 00:42:21.850 --> 00:42:22.930 align:middle line:90% in different countries? 00:42:22.930 --> 00:42:32.190 align:middle line:90% 00:42:32.190 --> 00:42:32.690 align:middle line:90% Yeah? 00:42:32.690 --> 00:42:34.808 align:middle line:84% STUDENT: Returns can be very different. 00:42:34.808 --> 00:42:35.600 align:middle line:90% ESTHER DUFLO: Yeah. 00:42:35.600 --> 00:42:38.297 align:middle line:84% The return could be very different. 00:42:38.297 --> 00:42:38.880 align:middle line:90% So that's one. 00:42:38.880 --> 00:42:41.490 align:middle line:90% 00:42:41.490 --> 00:42:42.450 align:middle line:90% What's another one? 00:42:42.450 --> 00:42:50.950 align:middle line:90% 00:42:50.950 --> 00:42:51.850 align:middle line:90% Yeah? 00:42:51.850 --> 00:42:53.530 align:middle line:84% STUDENT: So with this, it seems like you 00:42:53.530 --> 00:42:55.450 align:middle line:84% might prefer your education split there, 00:42:55.450 --> 00:42:57.610 align:middle line:90% but there might be [INAUDIBLE]. 00:42:57.610 --> 00:43:01.825 align:middle line:90% 00:43:01.825 --> 00:43:03.450 align:middle line:84% ESTHER DUFLO: So I heard the beginning, 00:43:03.450 --> 00:43:06.035 align:middle line:84% this assume that the market for education is perfect. 00:43:06.035 --> 00:43:06.660 align:middle line:90% STUDENT: Right. 00:43:06.660 --> 00:43:08.440 align:middle line:90% That there might be frictions. 00:43:08.440 --> 00:43:10.090 align:middle line:84% ESTHER DUFLO: Yes, that there might be, 00:43:10.090 --> 00:43:13.800 align:middle line:84% in the delivery of education itself. 00:43:13.800 --> 00:43:15.570 align:middle line:84% So we'll go back to that in a minute. 00:43:15.570 --> 00:43:21.013 align:middle line:84% And then within this model, with the context of this assumption, 00:43:21.013 --> 00:43:23.430 align:middle line:84% the other thing that's going to determine education level, 00:43:23.430 --> 00:43:26.040 align:middle line:84% potentially, is the discount rate. 00:43:26.040 --> 00:43:29.570 align:middle line:84% So the technologies is going to determine the education 00:43:29.570 --> 00:43:32.030 align:middle line:84% level, both the technology in the labor market 00:43:32.030 --> 00:43:35.430 align:middle line:84% and the technology of providing education, and then 00:43:35.430 --> 00:43:38.963 align:middle line:90% the discount rate. 00:43:38.963 --> 00:43:40.880 align:middle line:84% And those are the things that should determine 00:43:40.880 --> 00:43:44.850 align:middle line:84% education and not education subsidies and education 00:43:44.850 --> 00:43:45.350 align:middle line:90% policies. 00:43:45.350 --> 00:43:48.910 align:middle line:90% 00:43:48.910 --> 00:43:53.560 align:middle line:84% However, there is scope to improve the production function 00:43:53.560 --> 00:43:56.340 align:middle line:90% of education potentially. 00:43:56.340 --> 00:43:59.140 align:middle line:84% So if you make education more productive, cheaper, that is, 00:43:59.140 --> 00:44:01.390 align:middle line:84% for example, you increase the quality of education 00:44:01.390 --> 00:44:06.480 align:middle line:84% so you get more unit of capital per amount of money 00:44:06.480 --> 00:44:09.990 align:middle line:84% that you're spending, you might get more education 00:44:09.990 --> 00:44:11.610 align:middle line:90% as a result of it. 00:44:11.610 --> 00:44:14.690 align:middle line:84% So even though the Rosensweig conclusion 00:44:14.690 --> 00:44:22.820 align:middle line:84% is don't bother with education, just improve the technology 00:44:22.820 --> 00:44:24.920 align:middle line:84% outside of the school system such 00:44:24.920 --> 00:44:27.230 align:middle line:84% that people find it valuable to be educated, 00:44:27.230 --> 00:44:29.750 align:middle line:84% one could also say, well, even in your model, 00:44:29.750 --> 00:44:31.790 align:middle line:84% I still have a reason to make education 00:44:31.790 --> 00:44:34.310 align:middle line:84% cheaper by making it more effective, more productive, 00:44:34.310 --> 00:44:36.900 align:middle line:90% et cetera. 00:44:36.900 --> 00:44:39.570 align:middle line:84% So for a long time, development economists 00:44:39.570 --> 00:44:42.310 align:middle line:84% had no interest in the economics of education, 00:44:42.310 --> 00:44:45.600 align:middle line:84% unlike the World Bank, who would spend a lot of time and effort 00:44:45.600 --> 00:44:49.110 align:middle line:84% trying to push investment in education. 00:44:49.110 --> 00:44:52.290 align:middle line:84% And to some extent, many people still hold this view 00:44:52.290 --> 00:44:55.920 align:middle line:84% and don't have much time for any of the work that 00:44:55.920 --> 00:44:59.395 align:middle line:84% exists on the economics of education 00:44:59.395 --> 00:45:01.020 align:middle line:84% and the technology of education that we 00:45:01.020 --> 00:45:02.850 align:middle line:84% are going to spend some time discussing 00:45:02.850 --> 00:45:03.900 align:middle line:90% in the next few weeks. 00:45:03.900 --> 00:45:06.840 align:middle line:84% And to some extent, many still hold this view that all of that 00:45:06.840 --> 00:45:17.410 align:middle line:84% is secondary relative to the return to education. 00:45:17.410 --> 00:45:19.950 align:middle line:84% So now, what I want to do for the rest of today 00:45:19.950 --> 00:45:24.180 align:middle line:84% is to discuss the different assumption that 00:45:24.180 --> 00:45:29.400 align:middle line:84% live inside this model and try to have you think, too, 00:45:29.400 --> 00:45:32.490 align:middle line:84% whether or not they break the fundamental logic of the model 00:45:32.490 --> 00:45:33.410 align:middle line:90% or not. 00:45:33.410 --> 00:45:34.320 align:middle line:90% Yeah? 00:45:34.320 --> 00:45:36.780 align:middle line:84% STUDENT: I guess, maybe, this is a big assumption. 00:45:36.780 --> 00:45:39.564 align:middle line:84% But is the core intuition in this 00:45:39.564 --> 00:45:42.150 align:middle line:84% that we should think of education 00:45:42.150 --> 00:45:44.670 align:middle line:84% as an asset just like the many other assets 00:45:44.670 --> 00:45:46.700 align:middle line:84% that people decide to invest in or not? 00:45:46.700 --> 00:45:47.617 align:middle line:90% ESTHER DUFLO: Exactly. 00:45:47.617 --> 00:45:49.180 align:middle line:84% STUDENT: So then in portfolio theory, 00:45:49.180 --> 00:45:50.560 align:middle line:84% we should think of it in that lens? 00:45:50.560 --> 00:45:51.070 align:middle line:90% ESTHER DUFLO: Exactly. 00:45:51.070 --> 00:45:52.880 align:middle line:84% So this is exactly what it's doing. 00:45:52.880 --> 00:45:55.870 align:middle line:84% So where this is buried, or maybe not buried, 00:45:55.870 --> 00:45:58.900 align:middle line:84% but we are where this took place is 00:45:58.900 --> 00:46:05.320 align:middle line:84% when we remove the symbolic consumption from the utility 00:46:05.320 --> 00:46:11.460 align:middle line:84% function so that there is no consumption value in education, 00:46:11.460 --> 00:46:13.913 align:middle line:84% either direct, via the fact that people 00:46:13.913 --> 00:46:15.330 align:middle line:84% like to spend on education or like 00:46:15.330 --> 00:46:17.940 align:middle line:84% to be educated, or indirect, via the fact 00:46:17.940 --> 00:46:24.030 align:middle line:84% that you enjoy the fact that your child is educated 00:46:24.030 --> 00:46:27.410 align:middle line:90% or consume more. 00:46:27.410 --> 00:46:31.295 align:middle line:84% STUDENT: So education has big externalities in terms of it 00:46:31.295 --> 00:46:35.940 align:middle line:90% could even help shift the curve. 00:46:35.940 --> 00:46:39.730 align:middle line:84% Is that a big part of the literature of the discussion? 00:46:39.730 --> 00:46:40.910 align:middle line:90% ESTHER DUFLO: Yes. 00:46:40.910 --> 00:46:44.350 align:middle line:84% So here, education has no externality. 00:46:44.350 --> 00:46:48.340 align:middle line:84% And you're absolutely right that, for example, the World 00:46:48.340 --> 00:46:51.310 align:middle line:84% Bank argument of subsidizing education 00:46:51.310 --> 00:46:54.430 align:middle line:84% has long been around this idea that there 00:46:54.430 --> 00:46:58.420 align:middle line:84% are huge externalities to subsidize education, so 00:46:58.420 --> 00:47:00.470 align:middle line:90% to have more educated people. 00:47:00.470 --> 00:47:07.180 align:middle line:84% So here, if you change this world, so where that is buried, 00:47:07.180 --> 00:47:09.930 align:middle line:84% where is that buried no externality? 00:47:09.930 --> 00:47:12.360 align:middle line:90% Where does it live? 00:47:12.360 --> 00:47:14.075 align:middle line:84% In which assumption does it live? 00:47:14.075 --> 00:47:22.810 align:middle line:90% 00:47:22.810 --> 00:47:25.330 align:middle line:84% So early, quite early in the production 00:47:25.330 --> 00:47:31.880 align:middle line:84% function, so there is no externality in the production 00:47:31.880 --> 00:47:32.380 align:middle line:90% function. 00:47:32.380 --> 00:47:37.960 align:middle line:84% Just it doesn't have an A in front of it, which 00:47:37.960 --> 00:47:41.700 align:middle line:90% could be a product of h. 00:47:41.700 --> 00:47:44.540 align:middle line:84% So the way that externalities show up 00:47:44.540 --> 00:47:48.290 align:middle line:84% in models like the Lucas model is 00:47:48.290 --> 00:47:52.130 align:middle line:84% by having the entire production function 00:47:52.130 --> 00:47:54.140 align:middle line:84% multiplied by a term, which is itself 00:47:54.140 --> 00:47:56.130 align:middle line:90% a function of human capital. 00:47:56.130 --> 00:47:58.610 align:middle line:84% So that's pulled out by the production function 00:47:58.610 --> 00:48:00.810 align:middle line:90% that we wrote. 00:48:00.810 --> 00:48:02.920 align:middle line:84% Another place where we have ruled out, 00:48:02.920 --> 00:48:06.030 align:middle line:84% we have actually limited the externalities, 00:48:06.030 --> 00:48:14.040 align:middle line:84% is by saying that educating your children 00:48:14.040 --> 00:48:16.590 align:middle line:84% is cheaper when you are yourself educated. 00:48:16.590 --> 00:48:20.430 align:middle line:90% But this is decreasing. 00:48:20.430 --> 00:48:23.760 align:middle line:84% If, instead, we had-- and that's another way to write the Lucas 00:48:23.760 --> 00:48:24.720 align:middle line:90% model-- 00:48:24.720 --> 00:48:27.900 align:middle line:84% that education became cheaper as you 00:48:27.900 --> 00:48:30.750 align:middle line:84% yourself are more educated in a way that is linear, 00:48:30.750 --> 00:48:34.810 align:middle line:84% for example, then it would again give us 00:48:34.810 --> 00:48:39.120 align:middle line:84% externality of education, and potentially, very different 00:48:39.120 --> 00:48:44.280 align:middle line:84% steady state level of growth from that externalities. 00:48:44.280 --> 00:48:46.440 align:middle line:84% And that could happen either within family 00:48:46.440 --> 00:48:49.780 align:middle line:84% or via the education production function. 00:48:49.780 --> 00:48:52.290 align:middle line:84% So another place where the education production function 00:48:52.290 --> 00:48:59.130 align:middle line:84% is located is we don't have an A in front of it, which again, we 00:48:59.130 --> 00:49:02.340 align:middle line:84% have the unit of human capital times 00:49:02.340 --> 00:49:06.510 align:middle line:84% the little s plus unit of regular labor 00:49:06.510 --> 00:49:09.300 align:middle line:84% times the little s, which is all individual. 00:49:09.300 --> 00:49:11.490 align:middle line:84% We don't have an A that could be collective, which 00:49:11.490 --> 00:49:13.110 align:middle line:84% is if everybody is more educated, 00:49:13.110 --> 00:49:16.170 align:middle line:84% that's going to make the production function more 00:49:16.170 --> 00:49:16.720 align:middle line:90% educated. 00:49:16.720 --> 00:49:20.910 align:middle line:84% So that's the two places where the externalities are missing. 00:49:20.910 --> 00:49:24.090 align:middle line:84% And this idea that there are these huge externalities 00:49:24.090 --> 00:49:29.340 align:middle line:84% of education was kind of a big World Bank justification 00:49:29.340 --> 00:49:33.090 align:middle line:84% for investing in education, with or without 00:49:33.090 --> 00:49:35.620 align:middle line:84% much empirical evidence behind it. 00:49:35.620 --> 00:49:41.310 align:middle line:90% But we'll discuss it later. 00:49:41.310 --> 00:49:42.910 align:middle line:90% Yeah, [? Shawn? ?] 00:49:42.910 --> 00:49:46.900 align:middle line:84% STUDENT: This one also seems that parents are freely 00:49:46.900 --> 00:49:49.030 align:middle line:84% normalizing the [INAUDIBLE] they have 00:49:49.030 --> 00:49:50.820 align:middle line:84% on their children's education by-- 00:49:50.820 --> 00:49:52.987 align:middle line:84% ESTHER DUFLO: Do you mind holding your [INAUDIBLE]?? 00:49:52.987 --> 00:49:56.250 align:middle line:84% STUDENT: So the second, also, it seems that parents are freely 00:49:56.250 --> 00:49:59.490 align:middle line:84% normalizing the [INAUDIBLE] that their [INAUDIBLE] has 00:49:59.490 --> 00:50:00.690 align:middle line:90% on children's education. 00:50:00.690 --> 00:50:02.499 align:middle line:84% It's like that [INAUDIBLE] function 00:50:02.499 --> 00:50:04.560 align:middle line:90% is different from theirs. 00:50:04.560 --> 00:50:07.730 align:middle line:84% Parents value their children's education less than the 00:50:07.730 --> 00:50:11.400 align:middle line:84% [INAUDIBLE],, then the results may spill forward. 00:50:11.400 --> 00:50:12.150 align:middle line:90% ESTHER DUFLO: Yes. 00:50:12.150 --> 00:50:15.340 align:middle line:90% So that's very important. 00:50:15.340 --> 00:50:19.170 align:middle line:84% So this is, in a sense, what we are discussing now 00:50:19.170 --> 00:50:22.440 align:middle line:84% is the form of the utility function. 00:50:22.440 --> 00:50:24.740 align:middle line:84% So the form of the utility function 00:50:24.740 --> 00:50:28.070 align:middle line:90% is restrictive in two places. 00:50:28.070 --> 00:50:30.110 align:middle line:84% And the one you call attention to 00:50:30.110 --> 00:50:35.330 align:middle line:84% is the idea that the discount rate for future generation 00:50:35.330 --> 00:50:39.430 align:middle line:84% is the same discount rate that applies from period to period. 00:50:39.430 --> 00:50:45.480 align:middle line:84% So this beautiful trick of the Barro-Becker preferences, 00:50:45.480 --> 00:50:48.390 align:middle line:84% where I care about the next generation 00:50:48.390 --> 00:50:51.630 align:middle line:84% only, and with effect on delta, which 00:50:51.630 --> 00:50:56.350 align:middle line:84% is similar to my own discount rate over the period. 00:50:56.350 --> 00:50:59.680 align:middle line:84% And that absolutely means that there 00:50:59.680 --> 00:51:03.340 align:middle line:84% is no contracting difficulty between parents and children 00:51:03.340 --> 00:51:05.350 align:middle line:84% and means that the planner is going 00:51:05.350 --> 00:51:10.400 align:middle line:84% to have the same objective as the family. 00:51:10.400 --> 00:51:15.740 align:middle line:84% If the parents don't care about the children at all 00:51:15.740 --> 00:51:19.780 align:middle line:84% or they care about them less, then we 00:51:19.780 --> 00:51:22.300 align:middle line:84% find ourselves in a world that is 00:51:22.300 --> 00:51:26.080 align:middle line:84% going to have a time inconsistent utility 00:51:26.080 --> 00:51:28.100 align:middle line:90% preferences. 00:51:28.100 --> 00:51:29.720 align:middle line:84% So we are going to find ourselves 00:51:29.720 --> 00:51:31.380 align:middle line:90% where the entire future-- 00:51:31.380 --> 00:51:34.250 align:middle line:84% so suppose that this is the same model, but I modify it 00:51:34.250 --> 00:51:38.120 align:middle line:84% just a little bit, which is I like the consumption 00:51:38.120 --> 00:51:41.580 align:middle line:84% of my child just a little less than mine. 00:51:41.580 --> 00:51:43.740 align:middle line:90% Then we have two discount rates. 00:51:43.740 --> 00:51:47.190 align:middle line:84% Just this change leaves us with two discount rate 00:51:47.190 --> 00:51:49.050 align:middle line:90% that we have to play with. 00:51:49.050 --> 00:51:52.290 align:middle line:84% On the one hand, we have the over time, 00:51:52.290 --> 00:51:53.940 align:middle line:90% my own consumption over time. 00:51:53.940 --> 00:51:55.650 align:middle line:90% Let's call that delta. 00:51:55.650 --> 00:51:58.020 align:middle line:84% And on the other hand, we have the future generation. 00:51:58.020 --> 00:52:00.470 align:middle line:90% Let's call that beta. 00:52:00.470 --> 00:52:05.180 align:middle line:84% And rewriting now the utility function, 00:52:05.180 --> 00:52:07.370 align:middle line:84% we have one discount rate beta, which 00:52:07.370 --> 00:52:09.290 align:middle line:84% applies to the entire future, and then 00:52:09.290 --> 00:52:10.340 align:middle line:90% within that, the delta. 00:52:10.340 --> 00:52:13.430 align:middle line:84% So we are in a beta delta world, time inconsistent. 00:52:13.430 --> 00:52:16.370 align:middle line:84% Therefore, the decision of the individual today 00:52:16.370 --> 00:52:19.670 align:middle line:84% in the absence of commitment is going to look different 00:52:19.670 --> 00:52:22.760 align:middle line:84% than the decision of someone who cares 00:52:22.760 --> 00:52:25.280 align:middle line:84% not about any individual generation, 00:52:25.280 --> 00:52:28.050 align:middle line:84% but about all the generation in the future. 00:52:28.050 --> 00:52:30.830 align:middle line:84% So that's what you were saying about the planners potentially 00:52:30.830 --> 00:52:32.493 align:middle line:90% having a different objective. 00:52:32.493 --> 00:52:34.160 align:middle line:84% So here, we could say, well, the planner 00:52:34.160 --> 00:52:39.860 align:middle line:84% just is going to maximize the welfare of all 00:52:39.860 --> 00:52:41.220 align:middle line:90% of the future generation. 00:52:41.220 --> 00:52:43.340 align:middle line:84% So he is going to have exactly those preferences 00:52:43.340 --> 00:52:46.820 align:middle line:84% without the beta, whereas the individual is going to look 00:52:46.820 --> 00:52:48.780 align:middle line:90% like they are more impatient. 00:52:48.780 --> 00:52:51.380 align:middle line:84% So they are going to favor their own consumption 00:52:51.380 --> 00:52:54.860 align:middle line:84% to the expense of the investment in future 00:52:54.860 --> 00:52:57.380 align:middle line:84% of their children, educational investment 00:52:57.380 --> 00:52:59.750 align:middle line:90% or savings investment. 00:52:59.750 --> 00:53:03.510 align:middle line:90% And that creates a wedge there. 00:53:03.510 --> 00:53:07.300 align:middle line:84% But not that it creates a wedge in term of all the investment. 00:53:07.300 --> 00:53:08.550 align:middle line:90% They are going to overconsume. 00:53:08.550 --> 00:53:11.310 align:middle line:84% It's not that they are going to invest less 00:53:11.310 --> 00:53:14.430 align:middle line:84% in education relative to other financial assets. 00:53:14.430 --> 00:53:17.530 align:middle line:90% So that's one thing. 00:53:17.530 --> 00:53:22.380 align:middle line:84% And then once we have that, then we enter this delicate things 00:53:22.380 --> 00:53:25.320 align:middle line:90% about what can you commit? 00:53:25.320 --> 00:53:29.160 align:middle line:84% What can you commit your future generation to do? 00:53:29.160 --> 00:53:35.280 align:middle line:84% Because the person who cares about their children, 00:53:35.280 --> 00:53:40.670 align:middle line:84% but a little less than they care about themselves, 00:53:40.670 --> 00:53:43.070 align:middle line:84% they would like to commit their children 00:53:43.070 --> 00:53:45.950 align:middle line:84% to invest a lot in their own children's education 00:53:45.950 --> 00:53:50.168 align:middle line:84% because viewed from this generation perspective, 00:53:50.168 --> 00:53:51.960 align:middle line:84% the children, the grandchildren, et cetera, 00:53:51.960 --> 00:53:54.590 align:middle line:84% they are all equally valuable because we all 00:53:54.590 --> 00:53:56.030 align:middle line:90% value them at beta. 00:53:56.030 --> 00:53:59.210 align:middle line:84% So they would like to convince their children to invest a lot, 00:53:59.210 --> 00:54:01.490 align:middle line:84% to invest more in their own children's education. 00:54:01.490 --> 00:54:04.190 align:middle line:84% But of course, when the time comes for the children 00:54:04.190 --> 00:54:09.270 align:middle line:84% to take the education decision, they are not going to do that. 00:54:09.270 --> 00:54:12.860 align:middle line:84% So that's when we are going to have a wedge. 00:54:12.860 --> 00:54:15.080 align:middle line:84% And that's when we can have a planner, who is going 00:54:15.080 --> 00:54:16.550 align:middle line:90% to take a different decision. 00:54:16.550 --> 00:54:18.110 align:middle line:90% And he might, for example-- 00:54:18.110 --> 00:54:21.050 align:middle line:84% or she might-- decide to subsidize education 00:54:21.050 --> 00:54:26.600 align:middle line:84% to counteract this tendency overspend today. 00:54:26.600 --> 00:54:27.565 align:middle line:90% Yeah? 00:54:27.565 --> 00:54:38.520 align:middle line:84% STUDENT: [INAUDIBLE] But in realistic settings, 00:54:38.520 --> 00:54:40.887 align:middle line:84% in settings where the governments do not 00:54:40.887 --> 00:54:42.470 align:middle line:84% provide institutional or good support, 00:54:42.470 --> 00:54:45.540 align:middle line:84% parents probably also invest in children because children 00:54:45.540 --> 00:54:48.920 align:middle line:84% are going to earn money and then give them back 00:54:48.920 --> 00:54:50.480 align:middle line:90% in order to support them. 00:54:50.480 --> 00:54:51.230 align:middle line:90% ESTHER DUFLO: Yes. 00:54:51.230 --> 00:54:54.090 align:middle line:84% STUDENT: So I wonder if that's useful to capture 00:54:54.090 --> 00:54:58.632 align:middle line:84% that investment model too or [INAUDIBLE].. 00:54:58.632 --> 00:54:59.340 align:middle line:90% ESTHER DUFLO: No. 00:54:59.340 --> 00:55:01.620 align:middle line:84% So you're thinking of another place 00:55:01.620 --> 00:55:03.540 align:middle line:84% where this might break down, which 00:55:03.540 --> 00:55:11.150 align:middle line:84% I was hoping we would get, is that 00:55:11.150 --> 00:55:15.280 align:middle line:84% what's the assumption that you are taking away from me here. 00:55:15.280 --> 00:55:18.200 align:middle line:90% 00:55:18.200 --> 00:55:19.040 align:middle line:90% Not the contract. 00:55:19.040 --> 00:55:22.640 align:middle line:84% Here, let's stay to this world where there is no contract. 00:55:22.640 --> 00:55:24.710 align:middle line:84% All you can do is invest in your kid's education. 00:55:24.710 --> 00:55:26.530 align:middle line:90% That's it. 00:55:26.530 --> 00:55:32.120 align:middle line:84% But I have one assumption, one assumption here, 00:55:32.120 --> 00:55:35.620 align:middle line:84% which you removed when you were talking about pension and stuff 00:55:35.620 --> 00:55:36.950 align:middle line:90% like that. 00:55:36.950 --> 00:55:38.572 align:middle line:90% So what did you have in mind? 00:55:38.572 --> 00:55:40.030 align:middle line:84% Actually, don't have pensions here. 00:55:40.030 --> 00:55:41.840 align:middle line:90% There is no pension system. 00:55:41.840 --> 00:55:43.120 align:middle line:90% But what can people do? 00:55:43.120 --> 00:55:49.697 align:middle line:90% 00:55:49.697 --> 00:55:51.530 align:middle line:84% Usually, it's kind of the first thing people 00:55:51.530 --> 00:55:53.420 align:middle line:90% note along with the externality. 00:55:53.420 --> 00:56:00.130 align:middle line:90% 00:56:00.130 --> 00:56:01.030 align:middle line:90% I'll look for it. 00:56:01.030 --> 00:56:11.033 align:middle line:90% 00:56:11.033 --> 00:56:12.075 align:middle line:90% Now, it's pretty obvious. 00:56:12.075 --> 00:56:15.270 align:middle line:90% 00:56:15.270 --> 00:56:17.730 align:middle line:84% Well, the assumption you removed from me 00:56:17.730 --> 00:56:21.810 align:middle line:84% is that anyone can borrow or lend that growth rate r. 00:56:21.810 --> 00:56:24.670 align:middle line:84% And in fact, credit and loans can be requested. 00:56:24.670 --> 00:56:26.520 align:middle line:84% So you had in mind a pension system. 00:56:26.520 --> 00:56:28.020 align:middle line:84% We don't need a pension system here 00:56:28.020 --> 00:56:32.200 align:middle line:90% because people are able to save. 00:56:32.200 --> 00:56:36.540 align:middle line:84% So what the pension system does is that it helps us save. 00:56:36.540 --> 00:56:38.400 align:middle line:90% So here, there is no-- 00:56:38.400 --> 00:56:41.087 align:middle line:84% people are able to make their decisions perfectly. 00:56:41.087 --> 00:56:43.170 align:middle line:84% And they are able to put money in the bank account 00:56:43.170 --> 00:56:44.190 align:middle line:90% or to borrow. 00:56:44.190 --> 00:56:46.625 align:middle line:84% So they are able to transfer money in the future. 00:56:46.625 --> 00:56:48.000 align:middle line:84% So they don't need their children 00:56:48.000 --> 00:56:50.710 align:middle line:90% as a vessel to transfer money. 00:56:50.710 --> 00:56:53.710 align:middle line:84% So that's why that motive that we 00:56:53.710 --> 00:56:56.860 align:middle line:84% could find in the real world of investing in your children 00:56:56.860 --> 00:56:59.860 align:middle line:84% human capital in order for your children 00:56:59.860 --> 00:57:04.370 align:middle line:84% to pay you in the future doesn't have to be there. 00:57:04.370 --> 00:57:08.200 align:middle line:84% So the only reason why you invest in your children 00:57:08.200 --> 00:57:11.230 align:middle line:90% is because you care about them. 00:57:11.230 --> 00:57:15.170 align:middle line:84% Otherwise, you would have no reason to do it. 00:57:15.170 --> 00:57:21.560 align:middle line:84% So now, since we are on that slide, what happens if I say, 00:57:21.560 --> 00:57:26.300 align:middle line:84% well, you cannot borrow, or save, 00:57:26.300 --> 00:57:31.488 align:middle line:84% to remove whichever way you want, or both of them. 00:57:31.488 --> 00:57:33.655 align:middle line:84% Maybe let's start with borrowing in the first place. 00:57:33.655 --> 00:57:37.220 align:middle line:90% 00:57:37.220 --> 00:57:39.930 align:middle line:84% What implication is going to have in the short run 00:57:39.930 --> 00:57:43.577 align:middle line:90% and in steady state? 00:57:43.577 --> 00:57:44.847 align:middle line:90% STUDENT: [INAUDIBLE] 00:57:44.847 --> 00:57:45.680 align:middle line:90% ESTHER DUFLO: Sorry? 00:57:45.680 --> 00:57:48.310 align:middle line:84% STUDENT: There are going to be no hikes in the wages. 00:57:48.310 --> 00:57:49.060 align:middle line:90% ESTHER DUFLO: Yes. 00:57:49.060 --> 00:57:50.435 align:middle line:84% So in the short run, you're going 00:57:50.435 --> 00:57:53.550 align:middle line:84% to have an income effect because in the short run, 00:57:53.550 --> 00:57:57.360 align:middle line:84% you are going to see that some people are 00:57:57.360 --> 00:57:59.250 align:middle line:90% going to be constrained. 00:57:59.250 --> 00:58:02.400 align:middle line:84% Again, that's like the standard [INAUDIBLE] consumption level. 00:58:02.400 --> 00:58:04.620 align:middle line:84% Some people would like to consume more today 00:58:04.620 --> 00:58:06.150 align:middle line:90% than they get to. 00:58:06.150 --> 00:58:08.550 align:middle line:84% Here, some people are going to want to consume today 00:58:08.550 --> 00:58:11.470 align:middle line:90% more than they would like to. 00:58:11.470 --> 00:58:14.470 align:middle line:84% And then if those constraints are strong enough, 00:58:14.470 --> 00:58:17.250 align:middle line:90% then they might not-- 00:58:17.250 --> 00:58:19.860 align:middle line:84% they might not be able to pay for the level of education 00:58:19.860 --> 00:58:21.690 align:middle line:90% that they would like to get. 00:58:21.690 --> 00:58:25.680 align:middle line:90% Now, so that's going to change. 00:58:25.680 --> 00:58:32.868 align:middle line:84% That's potentially going to introduce income effect 00:58:32.868 --> 00:58:34.480 align:middle line:90% in this equation. 00:58:34.480 --> 00:58:37.560 align:middle line:84% This equation might not be satisfied anymore. 00:58:37.560 --> 00:58:43.500 align:middle line:84% We might be able to find cases where poor people invest less. 00:58:43.500 --> 00:58:48.060 align:middle line:84% Is that going to lead to multiple steady state? 00:58:48.060 --> 00:58:50.140 align:middle line:90% So two questions for you. 00:58:50.140 --> 00:58:53.000 align:middle line:84% The first one is that is it a big deal or not? 00:58:53.000 --> 00:58:55.310 align:middle line:84% Is that going to lead to multiple steady state, 00:58:55.310 --> 00:58:59.920 align:middle line:84% to persistence of inequality, and so on, and so forth 00:58:59.920 --> 00:59:01.830 align:middle line:84% if I leave everything else the same? 00:59:01.830 --> 00:59:08.270 align:middle line:90% 00:59:08.270 --> 00:59:10.130 align:middle line:84% I'm making you do a lot of gymnastics 00:59:10.130 --> 00:59:11.240 align:middle line:90% with standard models. 00:59:11.240 --> 00:59:16.950 align:middle line:90% 00:59:16.950 --> 00:59:18.450 align:middle line:90% I'm going to try someone else. 00:59:18.450 --> 00:59:21.140 align:middle line:90% 00:59:21.140 --> 00:59:22.970 align:middle line:90% I trust you know the answer. 00:59:22.970 --> 00:59:32.630 align:middle line:90% 00:59:32.630 --> 00:59:33.586 align:middle line:90% Heather. 00:59:33.586 --> 00:59:34.360 align:middle line:90% STUDENT: Hazel. 00:59:34.360 --> 00:59:35.485 align:middle line:90% ESTHER DUFLO: Hazel, sorry. 00:59:35.485 --> 00:59:37.460 align:middle line:84% STUDENT: And I don't think you would lead off 00:59:37.460 --> 00:59:38.390 align:middle line:90% the steady states. 00:59:38.390 --> 00:59:40.445 align:middle line:90% This is still no long capacity. 00:59:40.445 --> 00:59:41.570 align:middle line:90% ESTHER DUFLO: Yes, exactly. 00:59:41.570 --> 00:59:47.990 align:middle line:84% So what's going to happen if we have this credit constraint? 00:59:47.990 --> 00:59:52.970 align:middle line:84% What's going to happen over multiple generations? 00:59:52.970 --> 00:59:55.368 align:middle line:84% STUDENT: So we'll just take on to get the steady state 00:59:55.368 --> 00:59:57.035 align:middle line:84% because you can't built up human capital 00:59:57.035 --> 00:59:58.520 align:middle line:84% realistically because you're going to try-- 00:59:58.520 --> 00:59:59.437 align:middle line:90% ESTHER DUFLO: Exactly. 00:59:59.437 --> 01:00:02.570 align:middle line:84% So in the simple model, it's in one period, boom. 01:00:02.570 --> 01:00:04.320 align:middle line:90% Everyone invest the same thing. 01:00:04.320 --> 01:00:05.660 align:middle line:90% We're all in the same place. 01:00:05.660 --> 01:00:07.920 align:middle line:84% In this model, it might not be possible. 01:00:07.920 --> 01:00:09.650 align:middle line:84% So people are going to invest a little. 01:00:09.650 --> 01:00:17.060 align:middle line:84% But those who are poor, what are they going to enjoy that 01:00:17.060 --> 01:00:17.930 align:middle line:90% is going to help? 01:00:17.930 --> 01:00:21.520 align:middle line:90% 01:00:21.520 --> 01:00:24.160 align:middle line:84% Since they invest a little, their return to human capital 01:00:24.160 --> 01:00:26.710 align:middle line:90% is very high. 01:00:26.710 --> 01:00:28.780 align:middle line:84% So their return to human capital is very high. 01:00:28.780 --> 01:00:30.688 align:middle line:84% So that gives them some money that their kids 01:00:30.688 --> 01:00:31.980 align:middle line:90% are going to be able to invest. 01:00:31.980 --> 01:00:34.230 align:middle line:84% Their kids are still going to feel constrained, maybe. 01:00:34.230 --> 01:00:38.160 align:middle line:84% But over a few generations, it's going to all work out. 01:00:38.160 --> 01:00:41.800 align:middle line:84% So there is still only one steady state. 01:00:41.800 --> 01:00:45.660 align:middle line:84% That's another thing that is important to note is. 01:00:45.660 --> 01:00:49.200 align:middle line:84% In and of themselves, credit constraints 01:00:49.200 --> 01:00:53.948 align:middle line:84% are not a big deal in this model over time. 01:00:53.948 --> 01:00:56.240 align:middle line:84% Now, in the long run, as Keynes said, we'd all be dead. 01:00:56.240 --> 01:00:58.910 align:middle line:84% So this still sucks for the generation that doesn't 01:00:58.910 --> 01:01:00.080 align:middle line:90% get their own education. 01:01:00.080 --> 01:01:02.610 align:middle line:90% But that's relevant. 01:01:02.610 --> 01:01:06.230 align:middle line:84% So the way, for example, that Galor-Zeira 01:01:06.230 --> 01:01:10.250 align:middle line:84% make multiple steady state and inequality emerging equilibrium 01:01:10.250 --> 01:01:14.120 align:middle line:84% is to combine credit constraint with the non-convexity 01:01:14.120 --> 01:01:15.620 align:middle line:84% that they have, which is the return. 01:01:15.620 --> 01:01:18.920 align:middle line:84% There is some jump return to education at 0 01:01:18.920 --> 01:01:21.330 align:middle line:84% and then become higher after some level. 01:01:21.330 --> 01:01:24.980 align:middle line:84% And then you could get stuck in the ways 01:01:24.980 --> 01:01:27.200 align:middle line:84% that we discussed in our first lecture. 01:01:27.200 --> 01:01:29.818 align:middle line:90% Very good. 01:01:29.818 --> 01:01:31.610 align:middle line:84% So that's the first thing I wanted to note. 01:01:31.610 --> 01:01:35.180 align:middle line:84% So that in and of themselves, credit constraint 01:01:35.180 --> 01:01:37.250 align:middle line:84% is not really the thing that is going 01:01:37.250 --> 01:01:39.500 align:middle line:84% to make the massive difference in this model, 01:01:39.500 --> 01:01:43.320 align:middle line:84% unless it's combined with non-convexity. 01:01:43.320 --> 01:01:45.385 align:middle line:84% So that is something that you get. 01:01:45.385 --> 01:01:46.760 align:middle line:84% That's one of the tropes that you 01:01:46.760 --> 01:01:52.843 align:middle line:84% keep hearing in policy circles, which 01:01:52.843 --> 01:01:55.010 align:middle line:84% are good to keep in mind to re-examine a little bit. 01:01:55.010 --> 01:01:58.460 align:middle line:84% The other one is that it's just in terms 01:01:58.460 --> 01:02:01.350 align:middle line:90% of empirical observation. 01:02:01.350 --> 01:02:06.460 align:middle line:84% So you rightly observed that in the short run, 01:02:06.460 --> 01:02:08.960 align:middle line:84% there might be income effect that would be absent otherwise. 01:02:08.960 --> 01:02:12.210 align:middle line:84% So people who are richer will invest more. 01:02:12.210 --> 01:02:14.090 align:middle line:84% So going the other way, in general, 01:02:14.090 --> 01:02:19.170 align:middle line:84% that is shown as a proof or as a sign 01:02:19.170 --> 01:02:22.170 align:middle line:84% that there might be credit constraint. 01:02:22.170 --> 01:02:25.620 align:middle line:84% But what else would make income effect appear? 01:02:25.620 --> 01:02:31.130 align:middle line:90% 01:02:31.130 --> 01:02:36.770 align:middle line:84% What other assumption we could relax 01:02:36.770 --> 01:02:38.780 align:middle line:84% that would make income effect appear, 01:02:38.780 --> 01:02:40.850 align:middle line:84% even in the absence of credit constraint? 01:02:40.850 --> 01:02:49.590 align:middle line:90% 01:02:49.590 --> 01:02:50.310 align:middle line:90% I'll find it. 01:02:50.310 --> 01:03:01.160 align:middle line:90% 01:03:01.160 --> 01:03:03.410 align:middle line:90% It's that one. 01:03:03.410 --> 01:03:05.900 align:middle line:84% As soon as we introduce another motive 01:03:05.900 --> 01:03:11.290 align:middle line:84% to educate your kid, other than the altruism via material 01:03:11.290 --> 01:03:18.210 align:middle line:84% consumption, then you're going to have a potential income 01:03:18.210 --> 01:03:19.470 align:middle line:90% effect on consumption. 01:03:19.470 --> 01:03:23.630 align:middle line:90% 01:03:23.630 --> 01:03:27.848 align:middle line:84% So the fact that we find the income effect on education-- 01:03:27.848 --> 01:03:29.390 align:middle line:84% sorry, on education, the fact that we 01:03:29.390 --> 01:03:34.560 align:middle line:84% find income effect on education is not a proof 01:03:34.560 --> 01:03:36.602 align:middle line:84% that there are credit constraints on education. 01:03:36.602 --> 01:03:39.060 align:middle line:84% And the fact that there are credit constraints on education 01:03:39.060 --> 01:03:41.490 align:middle line:84% is not in itself an indication that there 01:03:41.490 --> 01:03:43.050 align:middle line:84% might be differentiable statistics, 01:03:43.050 --> 01:03:48.450 align:middle line:84% and therefore, the model will break down. 01:03:48.450 --> 01:03:49.760 align:middle line:90% Yeah, OK. 01:03:49.760 --> 01:03:51.960 align:middle line:84% STUDENT: So I just want to know why 01:03:51.960 --> 01:03:55.613 align:middle line:84% we're having the extra debate on the utility function. 01:03:55.613 --> 01:03:56.530 align:middle line:90% ESTHER DUFLO: Oh, why? 01:03:56.530 --> 01:03:59.560 align:middle line:90% So think on it. 01:03:59.560 --> 01:04:05.440 align:middle line:84% We now have a term of the utility function 01:04:05.440 --> 01:04:09.850 align:middle line:84% which has a human capital directly of your kids, 01:04:09.850 --> 01:04:12.280 align:middle line:84% as well as potentially even what you spend on it 01:04:12.280 --> 01:04:16.740 align:middle line:84% as being directly some argument of the utility function. 01:04:16.740 --> 01:04:18.900 align:middle line:90% So just think about that. 01:04:18.900 --> 01:04:22.820 align:middle line:84% Your parents are probably delighted that you're here. 01:04:22.820 --> 01:04:27.590 align:middle line:84% And therefore, they are deriving some utility out of it. 01:04:27.590 --> 01:04:31.550 align:middle line:84% So potentially, they gave you the school that you needed, 01:04:31.550 --> 01:04:34.050 align:middle line:84% and the tuition that you needed, and so on, and so forth. 01:04:34.050 --> 01:04:41.910 align:middle line:84% But if they had been starving, then they wouldn't have, 01:04:41.910 --> 01:04:44.580 align:middle line:84% not just because they couldn't borrow for it, 01:04:44.580 --> 01:04:46.080 align:middle line:84% but they wouldn't even have borrowed 01:04:46.080 --> 01:04:47.880 align:middle line:84% for it because therefore, this is a luxury 01:04:47.880 --> 01:04:50.250 align:middle line:90% that we can't afford. 01:04:50.250 --> 01:04:54.180 align:middle line:84% So education becomes a thing that 01:04:54.180 --> 01:04:55.740 align:middle line:90% is valued for the sake of it. 01:04:55.740 --> 01:04:58.680 align:middle line:84% But then it's valued like a normal good 01:04:58.680 --> 01:05:01.357 align:middle line:84% or even a luxury good, that the richer we are, 01:05:01.357 --> 01:05:02.190 align:middle line:90% the more we like it. 01:05:02.190 --> 01:05:06.510 align:middle line:90% 01:05:06.510 --> 01:05:16.050 align:middle line:84% So then if that's the case, then people spend on education, both 01:05:16.050 --> 01:05:21.930 align:middle line:84% for investment purposes and because they enjoy it. 01:05:21.930 --> 01:05:27.660 align:middle line:84% And then now, you have an asset that's 01:05:27.660 --> 01:05:30.220 align:middle line:90% not your regular asset anymore. 01:05:30.220 --> 01:05:32.380 align:middle line:84% You have an asset that's at the same time 01:05:32.380 --> 01:05:34.720 align:middle line:90% an asset and a consumption good. 01:05:34.720 --> 01:05:41.500 align:middle line:84% And that creates havoc because rich people invest more. 01:05:41.500 --> 01:05:43.810 align:middle line:84% And therefore, their kids are more productive. 01:05:43.810 --> 01:05:46.990 align:middle line:84% And that makes them richer, which therefore, 01:05:46.990 --> 01:05:48.610 align:middle line:90% wants them to invest even more. 01:05:48.610 --> 01:05:51.840 align:middle line:84% And that makes them productive, and so on, and so forth. 01:05:51.840 --> 01:05:55.730 align:middle line:84% So you can see how you could have, 01:05:55.730 --> 01:06:03.230 align:middle line:84% without even credit constraint, with the education or education 01:06:03.230 --> 01:06:06.500 align:middle line:84% expenses entering in the utility function, 01:06:06.500 --> 01:06:12.190 align:middle line:84% you can generate multiple statistics on inequality. 01:06:12.190 --> 01:06:12.690 align:middle line:90% Yeah? 01:06:12.690 --> 01:06:17.550 align:middle line:84% STUDENT: So on a similar line of analysis, 01:06:17.550 --> 01:06:20.250 align:middle line:84% I noticed one thing that we haven't touched on 01:06:20.250 --> 01:06:23.280 align:middle line:84% in this model is it's kind of assumed that there's maybe 01:06:23.280 --> 01:06:26.390 align:middle line:90% just one child in a household. 01:06:26.390 --> 01:06:29.720 align:middle line:84% So could, in this same symbolic consumption framework, 01:06:29.720 --> 01:06:33.232 align:middle line:90% let's say you have two children. 01:06:33.232 --> 01:06:34.940 align:middle line:84% And maybe there's birth order preference. 01:06:34.940 --> 01:06:38.120 align:middle line:84% Or you just invest more in quality between children 01:06:38.120 --> 01:06:40.050 align:middle line:84% as some form of smaller consumption. 01:06:40.050 --> 01:06:42.860 align:middle line:84% So there could also be effects in terms 01:06:42.860 --> 01:06:47.035 align:middle line:84% of equalizing the level of education between children. 01:06:47.035 --> 01:06:47.785 align:middle line:90% ESTHER DUFLO: Yes. 01:06:47.785 --> 01:06:50.680 align:middle line:90% 01:06:50.680 --> 01:06:54.350 align:middle line:84% So we haven't introduced family dynamics. 01:06:54.350 --> 01:06:56.440 align:middle line:90% But we'll get there in a while. 01:06:56.440 --> 01:06:58.540 align:middle line:84% Well, we have a little bit introduced in term 01:06:58.540 --> 01:07:01.330 align:middle line:90% of intergenerational conflict. 01:07:01.330 --> 01:07:08.980 align:middle line:84% And then as soon as you introduce that people just 01:07:08.980 --> 01:07:11.590 align:middle line:84% don't care about the consumption of their kids, 01:07:11.590 --> 01:07:16.180 align:middle line:84% the generation later, they have to start thinking about how 01:07:16.180 --> 01:07:17.590 align:middle line:90% is it possible to contract? 01:07:17.590 --> 01:07:19.690 align:middle line:90% What can you contract on? 01:07:19.690 --> 01:07:21.730 align:middle line:84% Can you force them to do everything? 01:07:21.730 --> 01:07:24.700 align:middle line:84% And can you force them to do nothing? 01:07:24.700 --> 01:07:26.420 align:middle line:84% And there is something in between? 01:07:26.420 --> 01:07:29.920 align:middle line:84% So that's the first set of family issues. 01:07:29.920 --> 01:07:31.690 align:middle line:84% And then the second set of family issue 01:07:31.690 --> 01:07:37.120 align:middle line:84% is even within a generation, different parents 01:07:37.120 --> 01:07:40.450 align:middle line:84% might have different views about what to do with their children. 01:07:40.450 --> 01:07:45.070 align:middle line:84% And the same parents might have different views about their two 01:07:45.070 --> 01:07:46.590 align:middle line:90% children. 01:07:46.590 --> 01:07:52.920 align:middle line:84% So they might value the consumption of sons 01:07:52.920 --> 01:07:55.230 align:middle line:84% differently than the consumption of daughters. 01:07:55.230 --> 01:08:02.490 align:middle line:84% They might derive different symbolic joys 01:08:02.490 --> 01:08:04.860 align:middle line:84% from an educated son versus an educated daughter, 01:08:04.860 --> 01:08:06.310 align:middle line:90% and so on, and so forth. 01:08:06.310 --> 01:08:10.890 align:middle line:84% So that's also going to generate differences within the family, 01:08:10.890 --> 01:08:14.790 align:middle line:84% and potentially, gender dynamics as well. 01:08:14.790 --> 01:08:19.520 align:middle line:84% And then once the two are integrated-- 01:08:19.520 --> 01:08:22.310 align:middle line:84% and I want you to hold this thought because 01:08:22.310 --> 01:08:26.060 align:middle line:84% in a couple of lectures, we'll look about our paper 01:08:26.060 --> 01:08:28.670 align:middle line:84% on Ghana, where we have boys and girls. 01:08:28.670 --> 01:08:34.550 align:middle line:84% And the girls are much less educated than boys. 01:08:34.550 --> 01:08:39.130 align:middle line:84% But it looked at the return to educating girls is much higher. 01:08:39.130 --> 01:08:41.080 align:middle line:84% Or at least, the marginal girl, the return 01:08:41.080 --> 01:08:43.689 align:middle line:84% to educate the marginal girl is higher than the return 01:08:43.689 --> 01:08:45.520 align:middle line:90% to educate the marginal boy. 01:08:45.520 --> 01:08:49.149 align:middle line:84% And that wouldn't exist in that model. 01:08:49.149 --> 01:08:50.118 align:middle line:90% That was not possible. 01:08:50.118 --> 01:08:51.910 align:middle line:84% You could have two different types of agent 01:08:51.910 --> 01:08:52.827 align:middle line:90% with different return. 01:08:52.827 --> 01:08:56.189 align:middle line:84% But they should both be equalized. 01:08:56.189 --> 01:08:58.260 align:middle line:84% They could have different level if girls happens 01:08:58.260 --> 01:09:01.319 align:middle line:84% to be less productive than boys for whatever reason, 01:09:01.319 --> 01:09:03.750 align:middle line:84% for example, because they don't get to work or something 01:09:03.750 --> 01:09:04.479 align:middle line:90% like that. 01:09:04.479 --> 01:09:07.500 align:middle line:84% But in this setting, we returned to educating 01:09:07.500 --> 01:09:09.479 align:middle line:84% the marginal girls is higher than the return 01:09:09.479 --> 01:09:10.771 align:middle line:90% to educating the marginal boys. 01:09:10.771 --> 01:09:13.888 align:middle line:84% That doesn't fit with that model. 01:09:13.888 --> 01:09:15.430 align:middle line:84% Because they need to all be equalized 01:09:15.430 --> 01:09:19.069 align:middle line:84% and be equalize to the interest rate. 01:09:19.069 --> 01:09:24.609 align:middle line:84% So that's something like a utility would have to enter. 01:09:24.609 --> 01:09:28.649 align:middle line:90% 01:09:28.649 --> 01:09:35.250 align:middle line:84% So now, I think we have kind of gotten over quite 01:09:35.250 --> 01:09:39.600 align:middle line:84% systematically over each of the different departures 01:09:39.600 --> 01:09:42.149 align:middle line:90% of the simple model. 01:09:42.149 --> 01:09:46.200 align:middle line:84% The last exercise I would like to do before we break 01:09:46.200 --> 01:09:50.979 align:middle line:84% is to think about the return to education. 01:09:50.979 --> 01:09:53.560 align:middle line:84% So remember, in the basic model, I 01:09:53.560 --> 01:09:58.240 align:middle line:84% told you that the return to education in steady state 01:09:58.240 --> 01:10:02.590 align:middle line:84% have to be positively related to how much education 01:10:02.590 --> 01:10:05.430 align:middle line:90% is being provided. 01:10:05.430 --> 01:10:06.060 align:middle line:90% All right. 01:10:06.060 --> 01:10:09.820 align:middle line:84% We can see it here, if the return to education is higher-- 01:10:09.820 --> 01:10:12.240 align:middle line:84% this is the return to education, that's 01:10:12.240 --> 01:10:14.250 align:middle line:84% the inverse of the return to education. 01:10:14.250 --> 01:10:16.230 align:middle line:84% If the return to education is higher, 01:10:16.230 --> 01:10:22.180 align:middle line:84% then people are going to invest in more schooling. 01:10:22.180 --> 01:10:30.620 align:middle line:84% Now, in a world where education potentially has these income 01:10:30.620 --> 01:10:33.440 align:middle line:84% effects, there is this income effect 01:10:33.440 --> 01:10:35.810 align:middle line:84% in education coming from preferences 01:10:35.810 --> 01:10:38.930 align:middle line:84% for more educated kids that behave like a normal 01:10:38.930 --> 01:10:44.900 align:middle line:84% or like a luxury good, is it always 01:10:44.900 --> 01:10:49.010 align:middle line:84% going to be true that the return to education 01:10:49.010 --> 01:10:55.450 align:middle line:84% are positively correlated with education in a country? 01:10:55.450 --> 01:11:19.240 align:middle line:90% 01:11:19.240 --> 01:11:23.570 align:middle line:84% So think about the two forces that are at play. 01:11:23.570 --> 01:11:26.170 align:middle line:84% So when I increase the return to education, I increase this. 01:11:26.170 --> 01:11:38.090 align:middle line:90% 01:11:38.090 --> 01:11:39.110 align:middle line:90% Yeah? 01:11:39.110 --> 01:11:41.465 align:middle line:84% You look like you're ready to speak. 01:11:41.465 --> 01:11:43.820 align:middle line:84% STUDENT: No, I'm just still thinking now. 01:11:43.820 --> 01:11:44.820 align:middle line:90% ESTHER DUFLO: Go for it. 01:11:44.820 --> 01:11:45.900 align:middle line:90% It is not obvious. 01:11:45.900 --> 01:11:52.290 align:middle line:84% So I don't expect you to deliver the answer. 01:11:52.290 --> 01:11:56.910 align:middle line:84% But I would like to just think through the pieces 01:11:56.910 --> 01:12:05.440 align:middle line:84% for why that's kind of what seems 01:12:05.440 --> 01:12:09.310 align:middle line:84% to be an automatic recipe of oh, just 01:12:09.310 --> 01:12:11.440 align:middle line:90% increase the rate to education. 01:12:11.440 --> 01:12:14.240 align:middle line:84% You're going to increase education might or might not 01:12:14.240 --> 01:12:14.740 align:middle line:90% apply. 01:12:14.740 --> 01:12:37.275 align:middle line:90% 01:12:37.275 --> 01:12:38.150 align:middle line:90% I'll give you a hint. 01:12:38.150 --> 01:12:40.940 align:middle line:84% So what is education provided with? 01:12:40.940 --> 01:12:42.080 align:middle line:90% By who? 01:12:42.080 --> 01:12:42.770 align:middle line:90% By whom? 01:12:42.770 --> 01:12:46.782 align:middle line:90% Who is education provided by? 01:12:46.782 --> 01:12:47.490 align:middle line:90% STUDENT: Schools? 01:12:47.490 --> 01:12:48.025 align:middle line:90% Governments? 01:12:48.025 --> 01:12:48.525 align:middle line:90% Parents? 01:12:48.525 --> 01:12:50.150 align:middle line:84% ESTHER DUFLO: Yeah, school, government. 01:12:50.150 --> 01:12:51.556 align:middle line:90% But who is in the school? 01:12:51.556 --> 01:12:53.020 align:middle line:90% STUDENT: Are workers. 01:12:53.020 --> 01:12:53.770 align:middle line:90% ESTHER DUFLO: Yes. 01:12:53.770 --> 01:12:56.240 align:middle line:90% Who have what characteristics? 01:12:56.240 --> 01:12:57.323 align:middle line:90% STUDENT: They're educated. 01:12:57.323 --> 01:12:59.073 align:middle line:84% ESTHER DUFLO: They're educated themselves. 01:12:59.073 --> 01:13:00.460 align:middle line:84% So if you remember, if we go back 01:13:00.460 --> 01:13:04.220 align:middle line:84% to the education production function, 01:13:04.220 --> 01:13:07.460 align:middle line:84% the education is being provided by a combination 01:13:07.460 --> 01:13:12.440 align:middle line:84% of uneducated workers and educated workers. 01:13:12.440 --> 01:13:16.330 align:middle line:84% So when we return to education increase, what happens 01:13:16.330 --> 01:13:17.590 align:middle line:90% to the cost of education? 01:13:17.590 --> 01:13:24.730 align:middle line:90% 01:13:24.730 --> 01:13:26.160 align:middle line:90% STUDENT: It also increases? 01:13:26.160 --> 01:13:28.880 align:middle line:84% ESTHER DUFLO: It also increases because now, you 01:13:28.880 --> 01:13:30.990 align:middle line:84% need to pay more those educated people. 01:13:30.990 --> 01:13:32.930 align:middle line:84% So unless you manage to change the technology 01:13:32.930 --> 01:13:37.130 align:middle line:84% at the same time, you need more educated people. 01:13:37.130 --> 01:13:39.080 align:middle line:84% So think about who is going to teach them 01:13:39.080 --> 01:13:42.120 align:middle line:90% while they're in school? 01:13:42.120 --> 01:13:45.360 align:middle line:84% So now, you have two forces that can potentially 01:13:45.360 --> 01:13:47.710 align:middle line:90% compete with each other. 01:13:47.710 --> 01:13:50.505 align:middle line:84% On the one hand, you have the return to education 01:13:50.505 --> 01:13:51.130 align:middle line:90% have increased. 01:13:51.130 --> 01:13:55.270 align:middle line:84% So people find it more valuable to educate their own kids. 01:13:55.270 --> 01:13:58.360 align:middle line:84% But remember, they wanted to educate their kids anyways, 01:13:58.360 --> 01:14:01.930 align:middle line:84% that they like to educate their kids irrespective 01:14:01.930 --> 01:14:06.020 align:middle line:84% of the return, just because it's a consumption good. 01:14:06.020 --> 01:14:08.510 align:middle line:84% At the same time, the cost of education 01:14:08.510 --> 01:14:13.160 align:middle line:84% is now going up because the education system is competing 01:14:13.160 --> 01:14:16.180 align:middle line:90% with the rest of the economy. 01:14:16.180 --> 01:14:18.600 align:middle line:84% So it could well be it doesn't have to happen. 01:14:18.600 --> 01:14:22.980 align:middle line:84% And it depends on exactly how things shake out 01:14:22.980 --> 01:14:24.600 align:middle line:90% with the parameters. 01:14:24.600 --> 01:14:28.590 align:middle line:84% But it could well be that the effect 01:14:28.590 --> 01:14:33.090 align:middle line:84% of the increase in the cost of education 01:14:33.090 --> 01:14:37.215 align:middle line:84% more than outweigh the effect of the increase in the return. 01:14:37.215 --> 01:14:40.160 align:middle line:90% 01:14:40.160 --> 01:14:42.200 align:middle line:84% For example, take the mother in mind, 01:14:42.200 --> 01:14:44.930 align:middle line:84% where a big driver of education was just 01:14:44.930 --> 01:14:47.480 align:middle line:84% that you like to have your educated kids. 01:14:47.480 --> 01:14:50.060 align:middle line:84% Then as it becomes more expensive, 01:14:50.060 --> 01:14:52.530 align:middle line:90% then you like it a little less. 01:14:52.530 --> 01:14:57.680 align:middle line:84% You're going to buy other things for them that 01:14:57.680 --> 01:15:00.760 align:middle line:90% will also make you happy. 01:15:00.760 --> 01:15:02.920 align:middle line:84% So you might find yourself in a situation 01:15:02.920 --> 01:15:13.660 align:middle line:84% where if a country starts going really fast, they actually, 01:15:13.660 --> 01:15:15.760 align:middle line:84% even though the return to education increases, 01:15:15.760 --> 01:15:19.870 align:middle line:84% and they find themselves having difficulty staffing 01:15:19.870 --> 01:15:24.350 align:middle line:84% the schools, and therefore, educating 01:15:24.350 --> 01:15:30.920 align:middle line:84% kids because of the wages they have to pay to the teachers. 01:15:30.920 --> 01:15:33.830 align:middle line:84% So you might find yourself where, in fact, 01:15:33.830 --> 01:15:35.470 align:middle line:84% the return to education increases 01:15:35.470 --> 01:15:39.310 align:middle line:84% decreases education level that a country attains. 01:15:39.310 --> 01:15:43.090 align:middle line:84% And again, you don't need any credit constraint for that. 01:15:43.090 --> 01:15:44.410 align:middle line:90% You just need this. 01:15:44.410 --> 01:15:51.160 align:middle line:84% This is just a consequence of introducing a utility value 01:15:51.160 --> 01:15:55.280 align:middle line:84% to the education expenditure or the education level themselves. 01:15:55.280 --> 01:15:57.913 align:middle line:90% So that looks like a much more-- 01:15:57.913 --> 01:15:59.330 align:middle line:84% I don't know, maybe it looks to me 01:15:59.330 --> 01:16:04.152 align:middle line:84% like a smaller departure from our traditional model. 01:16:04.152 --> 01:16:05.610 align:middle line:84% But in fact, it kind of breaks down 01:16:05.610 --> 01:16:07.155 align:middle line:84% a lot of things as soon as-- and I 01:16:07.155 --> 01:16:09.030 align:middle line:84% think [INAUDIBLE] summarized it the best-- is 01:16:09.030 --> 01:16:13.320 align:middle line:84% that if education is not like your regular old assets, 01:16:13.320 --> 01:16:17.500 align:middle line:90% everything can go south. 01:16:17.500 --> 01:16:26.470 align:middle line:84% So for example, in India, during the growth of the '80s and '90s 01:16:26.470 --> 01:16:34.190 align:middle line:84% led to an increase in the demand for educated workers, 01:16:34.190 --> 01:16:37.870 align:middle line:84% including in the software industry 01:16:37.870 --> 01:16:40.120 align:middle line:84% and then in the call centers that 01:16:40.120 --> 01:16:42.910 align:middle line:84% could take people with secondary education 01:16:42.910 --> 01:16:44.080 align:middle line:90% or the beginning of college. 01:16:44.080 --> 01:16:50.750 align:middle line:84% And then it becomes very, very difficult to staff the schools. 01:16:50.750 --> 01:16:53.710 align:middle line:84% So here, it doesn't necessarily translate into increasing costs 01:16:53.710 --> 01:16:56.560 align:middle line:84% for families, which in principle, can 01:16:56.560 --> 01:16:57.530 align:middle line:90% go to secondary school. 01:16:57.530 --> 01:17:01.540 align:middle line:84% But it leads, in a way, to increasing the effective cost, 01:17:01.540 --> 01:17:05.260 align:middle line:84% because that leads to a decrease in the education and quality. 01:17:05.260 --> 01:17:08.830 align:middle line:84% And you're basically getting less you need of human capital 01:17:08.830 --> 01:17:11.080 align:middle line:90% for time spent in the school. 01:17:11.080 --> 01:17:15.810 align:middle line:84% Hence the tension between the two. 01:17:15.810 --> 01:17:18.920 align:middle line:84% So that's what I wanted to do for today. 01:17:18.920 --> 01:17:24.590 align:middle line:84% Unless you have any more questions or remark, that's it. 01:17:24.590 --> 01:17:25.730 align:middle line:90% Yeah. 01:17:25.730 --> 01:17:28.970 align:middle line:84% STUDENT: I was wondering what would happen if you introduced 01:17:28.970 --> 01:17:32.300 align:middle line:84% risky returns into this model, if things would break down 01:17:32.300 --> 01:17:35.440 align:middle line:90% as a result of that as well? 01:17:35.440 --> 01:17:37.310 align:middle line:84% ESTHER DUFLO: Yeah, that's a great point. 01:17:37.310 --> 01:17:40.960 align:middle line:84% So you mean on education, for example? 01:17:40.960 --> 01:17:44.170 align:middle line:84% STUDENT: Yeah, like if I invest in my child's education, maybe 01:17:44.170 --> 01:17:47.160 align:middle line:84% there's like some distribution of wages that they can get. 01:17:47.160 --> 01:17:49.220 align:middle line:84% And I don't necessarily know if they're actually 01:17:49.220 --> 01:17:51.800 align:middle line:84% going to get a really good job out of it. 01:17:51.800 --> 01:17:52.550 align:middle line:90% ESTHER DUFLO: Yes. 01:17:52.550 --> 01:17:55.700 align:middle line:90% So I think it's a great point. 01:17:55.700 --> 01:17:59.045 align:middle line:84% So again, everything else constant, it's going to depend. 01:17:59.045 --> 01:18:01.760 align:middle line:90% 01:18:01.760 --> 01:18:03.980 align:middle line:84% So the net return to education is 01:18:03.980 --> 01:18:09.510 align:middle line:84% going to be weighted down by your risk aversion. 01:18:09.510 --> 01:18:13.310 align:middle line:84% But the other assets might themselves be risky as well. 01:18:13.310 --> 01:18:16.280 align:middle line:90% 01:18:16.280 --> 01:18:19.435 align:middle line:84% So I don't know how to think about how the two compare when 01:18:19.435 --> 01:18:20.560 align:middle line:90% we need to write them down. 01:18:20.560 --> 01:18:24.290 align:middle line:84% But basically, again, you're going to do the same thing. 01:18:24.290 --> 01:18:27.220 align:middle line:84% But you're going to-- so I don't think it fundamentally changes 01:18:27.220 --> 01:18:33.220 align:middle line:84% the logic, except that instead of having a simple budget 01:18:33.220 --> 01:18:34.870 align:middle line:84% constraint where everything is certain, 01:18:34.870 --> 01:18:37.670 align:middle line:84% you're going to have an expected budget constraint. 01:18:37.670 --> 01:18:40.540 align:middle line:84% And then it's going to depend whether the expected budget 01:18:40.540 --> 01:18:43.220 align:middle line:84% constraint needs to be satisfied with equality every time. 01:18:43.220 --> 01:18:49.870 align:middle line:84% And so that might generate a little wiggle here and there, 01:18:49.870 --> 01:18:53.200 align:middle line:84% but probably doesn't change the fundamental logic 01:18:53.200 --> 01:18:58.090 align:middle line:84% of the problem in particular, since if you introduce 01:18:58.090 --> 01:18:59.710 align:middle line:84% uncertainty in return distribution, 01:18:59.710 --> 01:19:02.200 align:middle line:84% probably, you need to introduce uncertainty 01:19:02.200 --> 01:19:07.430 align:middle line:84% in the return to other kind of investments as well. 01:19:07.430 --> 01:19:08.900 align:middle line:90% STUDENT: Can I [INAUDIBLE]? 01:19:08.900 --> 01:19:10.970 align:middle line:84% ESTHER DUFLO: Oh, we didn't finish your point, 01:19:10.970 --> 01:19:15.660 align:middle line:84% because we move the to borrowing constraint. 01:19:15.660 --> 01:19:17.102 align:middle line:84% But with saving constraint, you're 01:19:17.102 --> 01:19:18.310 align:middle line:90% going to have the same thing. 01:19:18.310 --> 01:19:21.910 align:middle line:84% And again, Hazel's point continues to apply, 01:19:21.910 --> 01:19:24.900 align:middle line:84% which is it might lead to the opposite of people 01:19:24.900 --> 01:19:29.490 align:middle line:84% overinvesting in their kids because they are trying 01:19:29.490 --> 01:19:34.770 align:middle line:84% to get money back later in the form of the kids paying 01:19:34.770 --> 01:19:36.510 align:middle line:90% for them. 01:19:36.510 --> 01:19:38.670 align:middle line:84% But in the absence of other things, 01:19:38.670 --> 01:19:41.670 align:middle line:84% that also should eventually even itself out. 01:19:41.670 --> 01:19:45.670 align:middle line:90% 01:19:45.670 --> 01:19:46.390 align:middle line:90% Great, thank you. 01:19:46.390 --> 01:19:48.370 align:middle line:84% So next time, we are going to start-- 01:19:48.370 --> 01:19:50.050 align:middle line:84% basically, we're going to go back 01:19:50.050 --> 01:19:52.540 align:middle line:84% with this model in the back of our mind, or sometimes 01:19:52.540 --> 01:19:54.783 align:middle line:90% explicitly, on the slides. 01:19:54.783 --> 01:19:56.200 align:middle line:84% We're going to go back to thinking 01:19:56.200 --> 01:19:57.280 align:middle line:90% about all these things. 01:19:57.280 --> 01:19:58.880 align:middle line:90% Do people pay attention to it? 01:19:58.880 --> 01:20:02.260 align:middle line:84% That's going to be our framework to think about do people pay 01:20:02.260 --> 01:20:03.760 align:middle line:84% attention to the cost of education? 01:20:03.760 --> 01:20:06.820 align:middle line:84% Do people pay attention to the return to education? 01:20:06.820 --> 01:20:09.580 align:middle line:84% How do we even think about the return to education? 01:20:09.580 --> 01:20:13.960 align:middle line:84% And then so next week, on Wednesday, 01:20:13.960 --> 01:20:16.770 align:middle line:84% it's still going to be within that kind 01:20:16.770 --> 01:20:18.150 align:middle line:90% of neoclassical world. 01:20:18.150 --> 01:20:22.170 align:middle line:84% And then we can start breaking the things one by one, 01:20:22.170 --> 01:20:25.880 align:middle line:84% adding externalities, adding credit constraint. 01:20:25.880 --> 01:20:28.920 align:middle line:90% And Thanks. 01:20:28.920 --> 01:20:36.000 align:middle line:90%