WEBVTT 00:00:00.000 --> 00:00:02.994 align:middle line:90% [SQUEAKING] 00:00:02.994 --> 00:00:03.992 align:middle line:90% [RUSTLING] 00:00:03.992 --> 00:00:06.986 align:middle line:90% [CLICKING] 00:00:06.986 --> 00:00:18.473 align:middle line:90% 00:00:18.473 --> 00:00:19.890 align:middle line:84% BENJAMIN OLKEN: So stepping back-- 00:00:19.890 --> 00:00:23.790 align:middle line:84% what we were talking about on the tax section was we were 00:00:23.790 --> 00:00:26.970 align:middle line:84% talking about two different approaches to how you can think 00:00:26.970 --> 00:00:29.820 align:middle line:84% about solving the challenge of the infor-- 00:00:29.820 --> 00:00:31.265 align:middle line:84% OK, of the information challenges 00:00:31.265 --> 00:00:33.390 align:middle line:84% in the developing-- in developing world and the tax 00:00:33.390 --> 00:00:34.440 align:middle line:90% structure. 00:00:34.440 --> 00:00:37.380 align:middle line:84% One of them was through thinking about how do you 00:00:37.380 --> 00:00:39.960 align:middle line:84% design a tax code that allows you 00:00:39.960 --> 00:00:41.280 align:middle line:90% to get information revelation. 00:00:41.280 --> 00:00:43.620 align:middle line:84% And the other is how do you actually 00:00:43.620 --> 00:00:46.860 align:middle line:84% maybe through the tax administration system just 00:00:46.860 --> 00:00:48.453 align:middle line:90% get better data on people. 00:00:48.453 --> 00:00:49.870 align:middle line:84% And those were the two approaches. 00:00:49.870 --> 00:00:52.290 align:middle line:84% So we talked last time in the examples about the VAT 00:00:52.290 --> 00:00:53.670 align:middle line:84% about how do you think about that through changing 00:00:53.670 --> 00:00:54.630 align:middle line:90% tax administration. 00:00:54.630 --> 00:00:56.610 align:middle line:84% The other thing I want to talk about now 00:00:56.610 --> 00:00:58.090 align:middle line:84% is, how do you think about that-- 00:00:58.090 --> 00:01:00.510 align:middle line:84% how do you think about whether the tax administration 00:01:00.510 --> 00:01:03.030 align:middle line:84% margin is kind of interesting or is relevant? 00:01:03.030 --> 00:01:05.470 align:middle line:84% And what are some of the challenges in doing that. 00:01:05.470 --> 00:01:08.035 align:middle line:84% So I want to talk about two papers on this, 00:01:08.035 --> 00:01:09.660 align:middle line:84% and then I'll more or less wrap this up 00:01:09.660 --> 00:01:11.650 align:middle line:90% and switch to the next topic. 00:01:11.650 --> 00:01:12.150 align:middle line:90% OK. 00:01:12.150 --> 00:01:14.240 align:middle line:84% So this is a paper actually-- this is also a paper 00:01:14.240 --> 00:01:18.650 align:middle line:84% that-- these are papers that I've worked on, so for better 00:01:18.650 --> 00:01:19.530 align:middle line:90% or for worse. 00:01:19.530 --> 00:01:21.050 align:middle line:84% So this actually is a recent paper-- 00:01:21.050 --> 00:01:23.420 align:middle line:90% it's actually now forthcoming-- 00:01:23.420 --> 00:01:25.310 align:middle line:84% on tax administration versus tax rates, 00:01:25.310 --> 00:01:28.880 align:middle line:84% to understand just at some very broad level 00:01:28.880 --> 00:01:36.080 align:middle line:84% is tax administration a really big deal or not? 00:01:36.080 --> 00:01:39.920 align:middle line:84% And so-- and we want to do that in a way in which we 00:01:39.920 --> 00:01:43.370 align:middle line:84% can benchmark it or compare it to a traditional tax rate 00:01:43.370 --> 00:01:44.340 align:middle line:90% change. 00:01:44.340 --> 00:01:45.920 align:middle line:84% So the typical thing you do if you're 00:01:45.920 --> 00:01:49.423 align:middle line:84% reading about debates about taxes, the typical thing we 00:01:49.423 --> 00:01:51.590 align:middle line:84% think about doing is raise the rate, lower the rate, 00:01:51.590 --> 00:01:53.215 align:middle line:84% change the base, those kinds of things. 00:01:53.215 --> 00:01:56.060 align:middle line:84% Can we compare just a straightforward change 00:01:56.060 --> 00:01:57.735 align:middle line:84% in improvement in tax administration 00:01:57.735 --> 00:01:59.360 align:middle line:84% to one of these more traditional things 00:01:59.360 --> 00:02:00.990 align:middle line:90% like changing the tax rates? 00:02:00.990 --> 00:02:04.940 align:middle line:84% And so we're going to do this in Indonesia studying two reforms. 00:02:04.940 --> 00:02:06.650 align:middle line:90% These are corporate taxpayers. 00:02:06.650 --> 00:02:10.340 align:middle line:84% One was a corporate-- a taxpayer administration reform. 00:02:10.340 --> 00:02:12.110 align:middle line:84% And here, the idea is that corporate taxes 00:02:12.110 --> 00:02:13.230 align:middle line:90% tend to be very skewed. 00:02:13.230 --> 00:02:16.790 align:middle line:84% So people on the-- basically, the firm distribution 00:02:16.790 --> 00:02:17.370 align:middle line:90% is skewed. 00:02:17.370 --> 00:02:19.328 align:middle line:84% So there's a small number of really large firms 00:02:19.328 --> 00:02:21.290 align:middle line:90% that pay most of the tax. 00:02:21.290 --> 00:02:24.320 align:middle line:84% And therefore, you may want to invest your tax administration 00:02:24.320 --> 00:02:26.180 align:middle line:90% kind of on those firms. 00:02:26.180 --> 00:02:28.610 align:middle line:84% Lots of countries do it for the very biggest taxpayers. 00:02:28.610 --> 00:02:30.110 align:middle line:84% Indonesia, when we're studying this, 00:02:30.110 --> 00:02:32.450 align:middle line:84% was kind of rolling out this same idea to medium size 00:02:32.450 --> 00:02:33.800 align:middle line:90% taxpayers. 00:02:33.800 --> 00:02:36.470 align:middle line:84% So, in particular, they rolled out 00:02:36.470 --> 00:02:40.280 align:middle line:84% these medium sized tax offices to serve the largest 00:02:40.280 --> 00:02:42.230 align:middle line:90% 330 taxpayers in each region. 00:02:42.230 --> 00:02:45.020 align:middle line:84% And we're going to study the impact on firms 00:02:45.020 --> 00:02:46.940 align:middle line:90% when those were created. 00:02:46.940 --> 00:02:50.120 align:middle line:84% And we're going to find really big impacts. 00:02:50.120 --> 00:02:52.970 align:middle line:84% The second thing is the tax rate reform. 00:02:52.970 --> 00:02:58.130 align:middle line:84% And basically, here, the identification 00:02:58.130 --> 00:03:00.275 align:middle line:84% is going to come from the fact that they-- 00:03:00.275 --> 00:03:03.680 align:middle line:84% I don't think I brought the graph actually. 00:03:03.680 --> 00:03:05.457 align:middle line:90% But they used to sort of-- 00:03:05.457 --> 00:03:07.040 align:middle line:84% actually, I may not focus on this one. 00:03:07.040 --> 00:03:08.660 align:middle line:84% I want to focus on the tax administration reform. 00:03:08.660 --> 00:03:10.430 align:middle line:84% But you can read the paper for interested. 00:03:10.430 --> 00:03:15.020 align:middle line:84% They had a switch from a tax rate 00:03:15.020 --> 00:03:17.087 align:middle line:84% that was based on your taxable income 00:03:17.087 --> 00:03:19.170 align:middle line:84% to a tax rate that was based on your gross income. 00:03:19.170 --> 00:03:20.670 align:middle line:84% And so that gives you identification 00:03:20.670 --> 00:03:22.160 align:middle line:90% for the change in tax rates. 00:03:22.160 --> 00:03:26.190 align:middle line:90% 00:03:26.190 --> 00:03:27.680 align:middle line:90% OK. 00:03:27.680 --> 00:03:30.660 align:middle line:84% So I want to focus on the tax administration reform. 00:03:30.660 --> 00:03:34.902 align:middle line:84% So the typical diffs-in-diffs assumption that you guys 00:03:34.902 --> 00:03:37.110 align:middle line:84% have seen in multiple papers throughout the semester. 00:03:37.110 --> 00:03:40.035 align:middle line:84% And I'm sure Ed talked about it in a diffs-in-diffs lecture. 00:03:40.035 --> 00:03:42.415 align:middle line:84% Did you-- how much did you go into matching-- propensity 00:03:42.415 --> 00:03:43.040 align:middle line:90% score matching? 00:03:43.040 --> 00:03:43.710 align:middle line:90% Is that-- not at all. 00:03:43.710 --> 00:03:44.210 align:middle line:90% Great. 00:03:44.210 --> 00:03:46.630 align:middle line:84% So I'm going to talk about that right now, is the control 00:03:46.630 --> 00:03:48.630 align:middle line:84% groups-- the control group is on the same trends 00:03:48.630 --> 00:03:50.672 align:middle line:84% as the treatment group, other than the treatment. 00:03:50.672 --> 00:03:51.700 align:middle line:90% That's the assumption. 00:03:51.700 --> 00:03:58.200 align:middle line:84% So that's-- in this context, we're going to study firms 00:03:58.200 --> 00:04:00.270 align:middle line:84% that-- so what's the identification? 00:04:00.270 --> 00:04:02.610 align:middle line:84% The identification is in each of these-- 00:04:02.610 --> 00:04:04.290 align:middle line:84% or the thing we're studying is in each 00:04:04.290 --> 00:04:06.960 align:middle line:84% of these regions they create a special office for large firms. 00:04:06.960 --> 00:04:10.830 align:middle line:90% 00:04:10.830 --> 00:04:13.320 align:middle line:84% And they move the kind of biggest firms 00:04:13.320 --> 00:04:14.610 align:middle line:90% into this special tax office. 00:04:14.610 --> 00:04:17.750 align:middle line:84% And we want to study the impact of that. 00:04:17.750 --> 00:04:21.019 align:middle line:84% The problem is that the firms that are moving in 00:04:21.019 --> 00:04:22.250 align:middle line:90% are large firms. 00:04:22.250 --> 00:04:24.740 align:middle line:84% They literally take the largest firms in each area 00:04:24.740 --> 00:04:26.010 align:middle line:90% and move them in. 00:04:26.010 --> 00:04:31.483 align:middle line:84% So if you saw this, and you said, what would be the-- 00:04:31.483 --> 00:04:32.900 align:middle line:84% they're going to treat the largest 00:04:32.900 --> 00:04:34.415 align:middle line:84% firms in each area, what would you 00:04:34.415 --> 00:04:36.290 align:middle line:84% think of in terms of identification strategy? 00:04:36.290 --> 00:04:39.670 align:middle line:90% 00:04:39.670 --> 00:04:42.610 align:middle line:84% If you were rummaging around like the Indonesian tax office 00:04:42.610 --> 00:04:44.860 align:middle line:84% and came about this thing where they're creating a tax 00:04:44.860 --> 00:04:46.570 align:middle line:84% office for the largest firms in each region, 00:04:46.570 --> 00:04:48.640 align:middle line:84% what would you be thinking about for identification? 00:04:48.640 --> 00:04:48.970 align:middle line:90% Yeah? 00:04:48.970 --> 00:04:49.870 align:middle line:84% AUDIENCE: Well, one thing you could do 00:04:49.870 --> 00:04:51.842 align:middle line:84% is if they had a particular cut-off 00:04:51.842 --> 00:04:53.970 align:middle line:84% for the size of the office to be [INAUDIBLE].. 00:04:53.970 --> 00:04:54.970 align:middle line:90% BENJAMIN OLKEN: Exactly. 00:04:54.970 --> 00:04:55.330 align:middle line:90% Right. 00:04:55.330 --> 00:04:56.788 align:middle line:84% You would think you would do an RD, 00:04:56.788 --> 00:04:58.450 align:middle line:90% and that's what we thought, too. 00:04:58.450 --> 00:05:03.700 align:middle line:84% Unfortunately-- so we thought we would do an RD. 00:05:03.700 --> 00:05:12.270 align:middle line:84% Unfortunately, we couldn't do that because, basically, the RD 00:05:12.270 --> 00:05:16.110 align:middle line:84% was done on some combination of different other variables-- 00:05:16.110 --> 00:05:18.540 align:middle line:84% gross income, which is your revenue, tax payments, maybe 00:05:18.540 --> 00:05:20.550 align:middle line:84% some other variables in some formula. 00:05:20.550 --> 00:05:21.930 align:middle line:84% They had some Excel sheet in each 00:05:21.930 --> 00:05:23.160 align:middle line:90% of the 19 different regions. 00:05:23.160 --> 00:05:27.450 align:middle line:84% And the Excel sheets were lost to the recycle bin of time. 00:05:27.450 --> 00:05:32.910 align:middle line:84% So unfortunately-- and I say, despite our many, many efforts, 00:05:32.910 --> 00:05:36.030 align:middle line:84% we got graphs where we could try to replicate the formula. 00:05:36.030 --> 00:05:39.000 align:middle line:84% We got graphs that look like this. 00:05:39.000 --> 00:05:43.020 align:middle line:84% Probability that you were assigned to the new thing 00:05:43.020 --> 00:05:43.860 align:middle line:90% looks like this. 00:05:43.860 --> 00:05:47.000 align:middle line:90% This is firm size. 00:05:47.000 --> 00:05:48.375 align:middle line:84% And it was beautifully continuous 00:05:48.375 --> 00:05:49.792 align:middle line:84% and there were no discontinuities. 00:05:49.792 --> 00:05:51.500 align:middle line:84% And I think that that's because-- 00:05:51.500 --> 00:05:53.648 align:middle line:84% so clearly, we had the right variables. 00:05:53.648 --> 00:05:55.940 align:middle line:84% They were definitely doing it based on this assumption. 00:05:55.940 --> 00:05:59.150 align:middle line:84% But because we couldn't reproduce the exact thing that 00:05:59.150 --> 00:06:00.470 align:middle line:90% was used, we couldn't do an RD. 00:06:00.470 --> 00:06:02.512 align:middle line:84% And I have to say this is not the first time this 00:06:02.512 --> 00:06:06.540 align:middle line:90% has happened to me that-- 00:06:06.540 --> 00:06:10.990 align:middle line:84% so we tried, but ultimately, it kind of looked like that. 00:06:10.990 --> 00:06:14.060 align:middle line:84% So what we get instead is we used-- yeah? 00:06:14.060 --> 00:06:17.888 align:middle line:84% AUDIENCE: [INAUDIBLE] cannot produce exactly [INAUDIBLE] RD. 00:06:17.888 --> 00:06:19.930 align:middle line:84% BENJAMIN OLKEN: But if the graph looks like this, 00:06:19.930 --> 00:06:22.610 align:middle line:90% there's no discontinuity. 00:06:22.610 --> 00:06:25.220 align:middle line:90% It's all continuous. 00:06:25.220 --> 00:06:26.533 align:middle line:90% AUDIENCE: I see. 00:06:26.533 --> 00:06:28.700 align:middle line:84% BENJAMIN OLKEN: I think that basically what happened 00:06:28.700 --> 00:06:30.800 align:middle line:84% is because we didn't know the formula well-- 00:06:30.800 --> 00:06:34.160 align:middle line:84% I think that the issue is we didn't know the underlying 00:06:34.160 --> 00:06:35.120 align:middle line:90% running variable. 00:06:35.120 --> 00:06:35.787 align:middle line:90% Nor do we know-- 00:06:35.787 --> 00:06:37.620 align:middle line:84% I think what happened is a couple of things. 00:06:37.620 --> 00:06:40.292 align:middle line:84% Number one, we couldn't reproduce the exact underlining 00:06:40.292 --> 00:06:41.000 align:middle line:90% running variable. 00:06:41.000 --> 00:06:43.970 align:middle line:84% For example, maybe they used an older version of the tax data 00:06:43.970 --> 00:06:45.140 align:middle line:90% that wasn't exactly right. 00:06:45.140 --> 00:06:47.670 align:middle line:84% Maybe they had other things going on. 00:06:47.670 --> 00:06:49.170 align:middle line:90% We couldn't find RD, basically. 00:06:49.170 --> 00:06:50.042 align:middle line:90% AUDIENCE: Propensity matching? 00:06:50.042 --> 00:06:50.480 align:middle line:90% BENJAMIN OLKEN: Huh? 00:06:50.480 --> 00:06:51.490 align:middle line:84% AUDIENCE: What about propensity matching? 00:06:51.490 --> 00:06:52.690 align:middle line:84% BENJAMIN OLKEN: That's what we're going to do. 00:06:52.690 --> 00:06:53.240 align:middle line:90% AUDIENCE: OK. 00:06:53.240 --> 00:06:54.050 align:middle line:90% BENJAMIN OLKEN: Yeah. 00:06:54.050 --> 00:06:54.620 align:middle line:90% That's the answer. 00:06:54.620 --> 00:06:54.830 align:middle line:90% Yes. 00:06:54.830 --> 00:06:55.980 align:middle line:90% That's where we're going. 00:06:55.980 --> 00:06:57.688 align:middle line:84% So instead, we're going to use propensity 00:06:57.688 --> 00:06:59.522 align:middle line:84% matching-- match differences-in-differences. 00:06:59.522 --> 00:07:01.363 align:middle line:84% And so this is what I wanted to talk about. 00:07:01.363 --> 00:07:02.780 align:middle line:84% So instead what we're going to do, 00:07:02.780 --> 00:07:04.155 align:middle line:84% is we're going to say, look, it's 00:07:04.155 --> 00:07:06.380 align:middle line:84% pretty obvious with this kind of design 00:07:06.380 --> 00:07:11.690 align:middle line:84% that if I look at, say, these for-- 00:07:11.690 --> 00:07:14.000 align:middle line:90% over here, everyone is treated. 00:07:14.000 --> 00:07:16.100 align:middle line:90% Over here, nobody is treated. 00:07:16.100 --> 00:07:18.050 align:middle line:84% So it's kind of hopeless there to find kind 00:07:18.050 --> 00:07:19.760 align:middle line:90% of reasonable comparison sets. 00:07:19.760 --> 00:07:24.170 align:middle line:84% But in this group kind of in the middle, 00:07:24.170 --> 00:07:25.760 align:middle line:84% your probability of being treated 00:07:25.760 --> 00:07:26.990 align:middle line:90% is somewhere in between-- 00:07:26.990 --> 00:07:28.940 align:middle line:90% is in some intermediate range. 00:07:28.940 --> 00:07:31.937 align:middle line:84% And therefore, we may be able to compare similar firms that 00:07:31.937 --> 00:07:33.770 align:middle line:84% just-- some of them happened to get treated, 00:07:33.770 --> 00:07:34.812 align:middle line:90% and some of them didn't-- 00:07:34.812 --> 00:07:36.890 align:middle line:90% happened not to get treated. 00:07:36.890 --> 00:07:40.850 align:middle line:84% And that's kind of the idea of propensity score matching. 00:07:40.850 --> 00:07:42.150 align:middle line:90% So what do you do? 00:07:42.150 --> 00:07:49.620 align:middle line:84% You basic-- so how do you do propensity score matching? 00:07:49.620 --> 00:07:50.370 align:middle line:90% OK? 00:07:50.370 --> 00:07:52.370 align:middle line:90% So-- or in general for matching. 00:07:52.370 --> 00:07:55.000 align:middle line:90% There's a couple of steps. 00:07:55.000 --> 00:07:56.710 align:middle line:90% The first step is exactly this. 00:07:56.710 --> 00:07:59.130 align:middle line:84% You have to restrict-- you take your sample, 00:07:59.130 --> 00:08:04.980 align:middle line:84% and you want to restrict to the area of which-- 00:08:04.980 --> 00:08:07.980 align:middle line:84% where there's called common support, where basically 00:08:07.980 --> 00:08:11.190 align:middle line:84% there's going to be reasonable fractions of people who are 00:08:11.190 --> 00:08:12.720 align:middle line:90% both treated and not treated. 00:08:12.720 --> 00:08:15.910 align:middle line:84% Or in this case, firms that are treated and not treated, 00:08:15.910 --> 00:08:18.510 align:middle line:84% so that you have reasonable matching sets. 00:08:18.510 --> 00:08:21.042 align:middle line:90% So otherwise, you end up-- 00:08:21.042 --> 00:08:22.500 align:middle line:84% there's nobody in the control group 00:08:22.500 --> 00:08:24.330 align:middle line:84% who's going to be a good comparison here, 00:08:24.330 --> 00:08:25.290 align:middle line:84% and no one in the treatment group 00:08:25.290 --> 00:08:26.920 align:middle line:84% who's going to be a comparison to this firm here. 00:08:26.920 --> 00:08:28.230 align:middle line:90% So you just cut that sample out. 00:08:28.230 --> 00:08:30.647 align:middle line:84% And you should recognize that the late you're estimating-- 00:08:30.647 --> 00:08:32.220 align:middle line:84% the [INAUDIBLE] you're estimating 00:08:32.220 --> 00:08:34.870 align:middle line:90% is for that kind of medium set. 00:08:34.870 --> 00:08:37.409 align:middle line:84% So in our case, for example, we're going to be estimating, 00:08:37.409 --> 00:08:41.312 align:middle line:84% an effect, for the smaller firms that go into the medium tax 00:08:41.312 --> 00:08:43.770 align:middle line:84% office because the larger firms that go into the medium tax 00:08:43.770 --> 00:08:45.990 align:middle line:84% office, they're just getting treated 00:08:45.990 --> 00:08:48.987 align:middle line:90% with close to probability 1. 00:08:48.987 --> 00:08:51.320 align:middle line:84% So in our case, actually, it's a little more complicated 00:08:51.320 --> 00:08:52.640 align:middle line:84% because we're actually doing this 00:08:52.640 --> 00:08:53.600 align:middle line:90% on two different variables. 00:08:53.600 --> 00:08:55.430 align:middle line:84% We have-- because we know that they matched 00:08:55.430 --> 00:08:58.370 align:middle line:84% on your revenue, which is your gross income 00:08:58.370 --> 00:08:59.857 align:middle line:90% and your taxes paid. 00:08:59.857 --> 00:09:01.190 align:middle line:90% So we're going to do it on both. 00:09:01.190 --> 00:09:02.482 align:middle line:90% But it looks kind of like this. 00:09:02.482 --> 00:09:05.510 align:middle line:90% So this is the distribution of-- 00:09:05.510 --> 00:09:10.060 align:middle line:84% this is the distribution of gross income in the baseline 00:09:10.060 --> 00:09:10.560 align:middle line:90% year. 00:09:10.560 --> 00:09:12.690 align:middle line:84% This is the year that you used for assignment. 00:09:12.690 --> 00:09:15.620 align:middle line:84% And you can see that the treated groups have much-- are much 00:09:15.620 --> 00:09:18.360 align:middle line:90% bigger than the control groups. 00:09:18.360 --> 00:09:22.070 align:middle line:84% And so what we're going to do is we're going to say, I think, 00:09:22.070 --> 00:09:26.460 align:middle line:84% do, say, 97th and 2.5th percentile. 00:09:26.460 --> 00:09:27.650 align:middle line:90% What does that mean? 00:09:27.650 --> 00:09:29.928 align:middle line:84% We'll say, well, when I get down over here 00:09:29.928 --> 00:09:31.970 align:middle line:84% where there's only 2.5% of the distribution here, 00:09:31.970 --> 00:09:35.290 align:middle line:84% we'll make a cut-off there in the treatment group. 00:09:35.290 --> 00:09:36.980 align:middle line:84% And then we'll go the other way around, 00:09:36.980 --> 00:09:38.563 align:middle line:84% and we'll get through the same cut-off 00:09:38.563 --> 00:09:41.620 align:middle line:84% here in the control-- in the 97th 00:09:41.620 --> 00:09:44.150 align:middle line:90% and a half in the control group. 00:09:44.150 --> 00:09:46.330 align:middle line:84% And so that gets us this intermediate set, 00:09:46.330 --> 00:09:49.240 align:middle line:90% where we have reasonably-- 00:09:49.240 --> 00:09:53.430 align:middle line:84% like a reasonable number of firms from both distributions. 00:09:53.430 --> 00:09:56.560 align:middle line:84% Now, one thing you'll notice if you look at this graph, 00:09:56.560 --> 00:09:58.890 align:middle line:84% of course, counterintuitively, is 00:09:58.890 --> 00:10:02.790 align:middle line:84% that over down on this part of the distribution 00:10:02.790 --> 00:10:06.270 align:middle line:84% you have way more controls than treatments. 00:10:06.270 --> 00:10:09.030 align:middle line:84% And over on this part of the distribution, 00:10:09.030 --> 00:10:13.030 align:middle line:84% you have way more treatments than controls. 00:10:13.030 --> 00:10:15.550 align:middle line:84% So the next step, after you made this common support 00:10:15.550 --> 00:10:18.130 align:middle line:84% restriction, is you just deal with that fact 00:10:18.130 --> 00:10:22.270 align:middle line:84% and make these two groups look similar by weighting-- 00:10:22.270 --> 00:10:24.040 align:middle line:90% by weighting. 00:10:24.040 --> 00:10:26.430 align:middle line:84% So as I mentioned, we actually do this in two dimensions. 00:10:26.430 --> 00:10:27.430 align:middle line:90% We do in this dimension. 00:10:27.430 --> 00:10:29.410 align:middle line:90% This is on taxes paid. 00:10:29.410 --> 00:10:30.280 align:middle line:90% It looks kind of-- 00:10:30.280 --> 00:10:31.632 align:middle line:90% similar idea. 00:10:31.632 --> 00:10:33.340 align:middle line:84% The firms assigned to the treatment group 00:10:33.340 --> 00:10:34.990 align:middle line:90% are paying a lot more taxes. 00:10:34.990 --> 00:10:38.320 align:middle line:90% So you do these restrictions. 00:10:38.320 --> 00:10:42.810 align:middle line:84% The next step is you use the preperiod data 00:10:42.810 --> 00:10:44.340 align:middle line:90% to reweight the treated-- 00:10:44.340 --> 00:10:46.590 align:middle line:84% in control groups-- so that the weighted distributions 00:10:46.590 --> 00:10:48.510 align:middle line:90% look the same. 00:10:48.510 --> 00:10:50.270 align:middle line:90% So instead of having this-- 00:10:50.270 --> 00:10:53.030 align:middle line:84% because if I just took this regression on this sample 00:10:53.030 --> 00:10:55.130 align:middle line:84% without doing any weights, it still 00:10:55.130 --> 00:10:58.863 align:middle line:84% would be mostly bigger firms in the treatment, 00:10:58.863 --> 00:11:00.530 align:middle line:84% and mostly smaller firms in the control. 00:11:00.530 --> 00:11:02.390 align:middle line:84% But you can easily see from this picture 00:11:02.390 --> 00:11:03.973 align:middle line:84% how I could just reweight them to make 00:11:03.973 --> 00:11:06.020 align:middle line:90% them look kind of the same. 00:11:06.020 --> 00:11:10.140 align:middle line:84% And that's really all propensity score matching is going to do. 00:11:10.140 --> 00:11:12.670 align:middle line:90% So how do you do that? 00:11:12.670 --> 00:11:14.420 align:middle line:84% So how do you actually compute the weight? 00:11:14.420 --> 00:11:16.910 align:middle line:84% So there's a couple of different options. 00:11:16.910 --> 00:11:19.220 align:middle line:84% So one is what's called a propensity score. 00:11:19.220 --> 00:11:24.010 align:middle line:84% So you know if the functional form of the assignment rule, 00:11:24.010 --> 00:11:26.350 align:middle line:84% you can just estimate it, and you can use 00:11:26.350 --> 00:11:27.710 align:middle line:90% it to calculate the weights. 00:11:27.710 --> 00:11:33.490 align:middle line:84% So for example, say, it was like a probit-- 00:11:33.490 --> 00:11:35.650 align:middle line:84% like probablity that you're in the treatment group 00:11:35.650 --> 00:11:38.740 align:middle line:84% is some X prime beta, where the X is, 00:11:38.740 --> 00:11:40.510 align:middle line:84% and you know this functional form, 00:11:40.510 --> 00:11:43.890 align:middle line:84% you can run this auxiliary regression, 00:11:43.890 --> 00:11:45.810 align:middle line:84% predict the probability you're in there 00:11:45.810 --> 00:11:48.390 align:middle line:84% as a function of these multiple different variables. 00:11:48.390 --> 00:11:52.860 align:middle line:84% And then you can use weights 1 over P for the treatment unit, 00:11:52.860 --> 00:11:55.590 align:middle line:84% and 1 over 1 minus P for the control group-- 00:11:55.590 --> 00:11:57.540 align:middle line:84% control units using the predicted values. 00:11:57.540 --> 00:11:59.850 align:middle line:84% And that will just make the weights balanced. 00:11:59.850 --> 00:12:02.710 align:middle line:90% 00:12:02.710 --> 00:12:05.410 align:middle line:90% So that's option one. 00:12:05.410 --> 00:12:09.050 align:middle line:84% The other option-- and this is kind of a newer approach-- 00:12:09.050 --> 00:12:11.460 align:middle line:84% is a more nonparametric approach, 00:12:11.460 --> 00:12:14.640 align:middle line:84% which just solves for the weights directly. 00:12:14.640 --> 00:12:16.380 align:middle line:84% And basically-- and this is, look, 00:12:16.380 --> 00:12:18.300 align:middle line:84% if I don't know the functional form, 00:12:18.300 --> 00:12:22.650 align:middle line:84% what I'm going to do is I'm going to take the-- 00:12:22.650 --> 00:12:27.120 align:middle line:84% I'm going to look for a set of weights in the control group, 00:12:27.120 --> 00:12:28.310 align:middle line:90% so that I balance-- 00:12:28.310 --> 00:12:29.870 align:middle line:84% so that I just match whatever moments 00:12:29.870 --> 00:12:36.180 align:middle line:84% I want to match in the treatment group 00:12:36.180 --> 00:12:38.720 align:middle line:84% on a variety of different characteristics. 00:12:38.720 --> 00:12:41.540 align:middle line:84% And so this-- there are a couple of different versions of this. 00:12:41.540 --> 00:12:42.620 align:middle line:84% This method by Hainmueller is the one 00:12:42.620 --> 00:12:43.850 align:middle line:90% that we use in the paper. 00:12:43.850 --> 00:12:45.870 align:middle line:84% The idea of it basically is to say, look, 00:12:45.870 --> 00:12:47.840 align:middle line:84% let me just solve for a set of weights that 00:12:47.840 --> 00:12:49.250 align:middle line:84% matches on whatever things I want 00:12:49.250 --> 00:12:53.540 align:middle line:84% it to match on that are going to minimally deviate 00:12:53.540 --> 00:12:55.518 align:middle line:90% from uniform weights. 00:12:55.518 --> 00:12:57.310 align:middle line:84% So, of course, many different combinations. 00:12:57.310 --> 00:12:59.060 align:middle line:84% If you have more observations than moments 00:12:59.060 --> 00:13:01.185 align:middle line:84% you want to match on, there are many different ways 00:13:01.185 --> 00:13:01.840 align:middle line:90% of doing it. 00:13:01.840 --> 00:13:04.270 align:middle line:84% It chooses the one that is like least-- 00:13:04.270 --> 00:13:06.190 align:middle line:84% has the least deviation from uniform weights. 00:13:06.190 --> 00:13:09.240 align:middle line:90% 00:13:09.240 --> 00:13:10.740 align:middle line:84% And there's some arguments actually 00:13:10.740 --> 00:13:13.290 align:middle line:84% that in cases where you don't actually know the propensity 00:13:13.290 --> 00:13:15.780 align:middle line:84% score, or the propensity score formula isn't inexact, 00:13:15.780 --> 00:13:18.260 align:middle line:84% these nonparametric methods can actually do a better job 00:13:18.260 --> 00:13:19.260 align:middle line:90% kind of matching things. 00:13:19.260 --> 00:13:21.090 align:middle line:84% And that was our experience, actually, 00:13:21.090 --> 00:13:23.088 align:middle line:84% in doing this, which is why we went for the-- 00:13:23.088 --> 00:13:24.130 align:middle line:90% It doesn't really matter. 00:13:24.130 --> 00:13:26.310 align:middle line:84% You can-- all the results in the paper we'll go through if we do 00:13:26.310 --> 00:13:27.900 align:middle line:84% the regular propensity score, but I just wanted to mention 00:13:27.900 --> 00:13:30.510 align:middle line:84% this, as well, because it's a newer-- and there's a nice-- 00:13:30.510 --> 00:13:33.930 align:middle line:84% it's a newer method and there's a nice canned data package 00:13:33.930 --> 00:13:37.340 align:middle line:90% that will do this for you. 00:13:37.340 --> 00:13:40.010 align:middle line:84% So in any case, first step-- common support; 00:13:40.010 --> 00:13:44.830 align:middle line:84% second step-- estimated propensity score and rebalance. 00:13:44.830 --> 00:13:47.260 align:middle line:84% Third thing is then you can just estimate 00:13:47.260 --> 00:13:51.440 align:middle line:84% your difference-in-difference model on the reweighted data. 00:13:51.440 --> 00:13:55.073 align:middle line:84% And that's what we're going to do. 00:13:55.073 --> 00:13:56.990 align:middle line:84% You might have said, well, why do I even care? 00:13:56.990 --> 00:14:00.320 align:middle line:84% Why even-- one question you may have is why even bother? 00:14:00.320 --> 00:14:02.112 align:middle line:84% Why even do a different-- why even 00:14:02.112 --> 00:14:03.320 align:middle line:90% bother with this reweighting? 00:14:03.320 --> 00:14:05.460 align:middle line:84% Why not just do a regular difference-in-difference? 00:14:05.460 --> 00:14:07.835 align:middle line:84% And I think the reason is that you might think that these 00:14:07.835 --> 00:14:08.990 align:middle line:90% different-- there's two-- 00:14:08.990 --> 00:14:11.840 align:middle line:84% these different groups might be on of different trends 00:14:11.840 --> 00:14:12.360 align:middle line:90% or whatever. 00:14:12.360 --> 00:14:14.652 align:middle line:84% And so you really want to have comparable to comparable 00:14:14.652 --> 00:14:17.610 align:middle line:84% to make sure you're balancing the trends properly. 00:14:17.610 --> 00:14:18.662 align:middle line:90% Any questions? 00:14:18.662 --> 00:14:20.422 align:middle line:90% 00:14:20.422 --> 00:14:22.880 align:middle line:84% That's mostly what I wanted to say about propensity scores, 00:14:22.880 --> 00:14:24.860 align:middle line:84% but I just wanted to make-- to throw it in there. 00:14:24.860 --> 00:14:25.360 align:middle line:90% Yeah? 00:14:25.360 --> 00:14:28.920 align:middle line:84% AUDIENCE: [INAUDIBLE] and weighting methods 00:14:28.920 --> 00:14:34.230 align:middle line:84% compared to, for example, a exact [INAUDIBLE] placement 00:14:34.230 --> 00:14:38.160 align:middle line:84% that match a treatment unit with one to three 00:14:38.160 --> 00:14:40.003 align:middle line:90% for every single control unit. 00:14:40.003 --> 00:14:40.795 align:middle line:90% BENJAMIN OLKEN: Oh. 00:14:40.795 --> 00:14:43.980 align:middle line:90% 00:14:43.980 --> 00:14:45.270 align:middle line:90% Oh, so like exact matching? 00:14:45.270 --> 00:14:46.307 align:middle line:90% AUDIENCE: Yeah. 00:14:46.307 --> 00:14:47.140 align:middle line:90% So it's [INAUDIBLE]. 00:14:47.140 --> 00:14:48.010 align:middle line:84% BENJAMIN OLKEN: So I don't have-- 00:14:48.010 --> 00:14:49.010 align:middle line:90% it's a good question. 00:14:49.010 --> 00:14:49.677 align:middle line:90% I don't have a-- 00:14:49.677 --> 00:14:51.260 align:middle line:84% let me think for a minute and see if I 00:14:51.260 --> 00:14:52.670 align:middle line:90% have a good answer for you. 00:14:52.670 --> 00:14:54.520 align:middle line:84% So another-- so you're saying another option 00:14:54.520 --> 00:14:56.020 align:middle line:84% is for every option in the treatment 00:14:56.020 --> 00:14:58.390 align:middle line:84% group I find the closest match in the control group, 00:14:58.390 --> 00:14:59.620 align:middle line:84% and I just put them together, and put it 00:14:59.620 --> 00:15:00.703 align:middle line:90% in a matched fixed effect. 00:15:00.703 --> 00:15:03.035 align:middle line:90% 00:15:03.035 --> 00:15:04.410 align:middle line:84% That's a good question, actually. 00:15:04.410 --> 00:15:07.270 align:middle line:84% I don't have a good answer for why-- 00:15:07.270 --> 00:15:08.260 align:middle line:90% whether that would be-- 00:15:08.260 --> 00:15:12.370 align:middle line:90% 00:15:12.370 --> 00:15:13.570 align:middle line:90% well, if you have more-- 00:15:13.570 --> 00:15:22.935 align:middle line:90% 00:15:22.935 --> 00:15:23.810 align:middle line:90% it's a good question. 00:15:23.810 --> 00:15:24.920 align:middle line:84% I'm not sure I have the right answer. 00:15:24.920 --> 00:15:27.128 align:middle line:84% I think if you have more treatment-- if you have more 00:15:27.128 --> 00:15:29.570 align:middle line:84% of one-- if you have unbalanced numbers of observations, 00:15:29.570 --> 00:15:32.070 align:middle line:84% and you end up having to reuse the same ones multiple times, 00:15:32.070 --> 00:15:33.885 align:middle line:84% it may be inefficient compared to this. 00:15:33.885 --> 00:15:34.760 align:middle line:90% That's my best guess. 00:15:34.760 --> 00:15:35.840 align:middle line:84% But I don't know the answer, actually. 00:15:35.840 --> 00:15:37.173 align:middle line:90% Does anyone else have any ideas? 00:15:37.173 --> 00:15:38.462 align:middle line:90% I don't know the answer. 00:15:38.462 --> 00:15:41.130 align:middle line:84% I do-- any-- it's a good question you-- 00:15:41.130 --> 00:15:42.558 align:middle line:90% I don't know the answer. 00:15:42.558 --> 00:15:46.020 align:middle line:90% 00:15:46.020 --> 00:15:49.480 align:middle line:84% So this just shows you the raw data. 00:15:49.480 --> 00:15:51.600 align:middle line:90% This is like total taxes paid. 00:15:51.600 --> 00:15:54.930 align:middle line:84% This is the year when the thing is partially-- 00:15:54.930 --> 00:15:57.090 align:middle line:84% the thing takes two years to fully turn on. 00:15:57.090 --> 00:16:01.200 align:middle line:84% It's partially turned on in 2006-- fully turned on in 2007. 00:16:01.200 --> 00:16:04.650 align:middle line:84% And what you can see is you can see that there-- this 00:16:04.650 --> 00:16:07.110 align:middle line:90% is just the raw data weighted. 00:16:07.110 --> 00:16:09.480 align:middle line:84% So you can see that the pretrends looked kind of nice 00:16:09.480 --> 00:16:12.180 align:middle line:84% and balanced between the-- once you-- 00:16:12.180 --> 00:16:13.590 align:middle line:90% in these two weighted samples. 00:16:13.590 --> 00:16:15.480 align:middle line:84% And then clearly, something pretty dramatic 00:16:15.480 --> 00:16:18.810 align:middle line:84% is happening when we turn on the medium tax office. 00:16:18.810 --> 00:16:20.880 align:middle line:84% They seem to be paying a lot more taxes. 00:16:20.880 --> 00:16:27.017 align:middle line:84% And this shows you too a normal diff-in-diff estimation form. 00:16:27.017 --> 00:16:29.100 align:middle line:84% So, in fact, actually, their taxes go up by a lot. 00:16:29.100 --> 00:16:31.500 align:middle line:90% They go up by more than 100%. 00:16:31.500 --> 00:16:33.510 align:middle line:84% And the government raised a lot of revenue. 00:16:33.510 --> 00:16:39.670 align:middle line:84% And comparing this to tax rate changes, 00:16:39.670 --> 00:16:44.500 align:middle line:84% it's staggeringly large and equivalent 00:16:44.500 --> 00:16:46.420 align:middle line:90% to raising taxes on-- 00:16:46.420 --> 00:16:49.450 align:middle line:84% raising the tax rate by, I think, 6% or 8% 00:16:49.450 --> 00:16:51.190 align:middle line:90% on all taxpayers. 00:16:51.190 --> 00:16:54.610 align:middle line:84% So the substantive point is that tax administration can 00:16:54.610 --> 00:16:55.935 align:middle line:90% make a really big difference. 00:16:55.935 --> 00:16:58.060 align:middle line:84% And the thing I want to talk about methodologically 00:16:58.060 --> 00:16:59.977 align:middle line:84% is, how do you think about doing these matched 00:16:59.977 --> 00:17:02.550 align:middle line:90% difference-in-differences? 00:17:02.550 --> 00:17:03.420 align:middle line:90% Any questions? 00:17:03.420 --> 00:17:06.079 align:middle line:90% 00:17:06.079 --> 00:17:06.579 align:middle line:90% OK. 00:17:06.579 --> 00:17:09.111 align:middle line:90% I'm going to skip this. 00:17:09.111 --> 00:17:09.819 align:middle line:90% Oh, actually, no. 00:17:09.819 --> 00:17:11.094 align:middle line:90% This I wanted to say, too. 00:17:11.094 --> 00:17:11.829 align:middle line:90% OK, sorry. 00:17:11.829 --> 00:17:15.829 align:middle line:84% So why might this make such a difference? 00:17:15.829 --> 00:17:18.400 align:middle line:84% So one thing that's kind of striking and surprising, 00:17:18.400 --> 00:17:24.829 align:middle line:84% I think, is that it continues to actually go up over time. 00:17:24.829 --> 00:17:29.580 align:middle line:84% So you might have thought that once tax-- 00:17:29.580 --> 00:17:33.000 align:middle line:84% once you put in a new tax enforcement regime 00:17:33.000 --> 00:17:34.560 align:middle line:84% there will be like a direct effect 00:17:34.560 --> 00:17:36.960 align:middle line:84% where you'll get more tax-- collect more taxes. 00:17:36.960 --> 00:17:39.510 align:middle line:84% And then firms will start to figure out how to adapt. 00:17:39.510 --> 00:17:44.580 align:middle line:84% So you might have predicted it would go kind of like up 00:17:44.580 --> 00:17:48.365 align:middle line:84% and then down over time as firms start to adapt. 00:17:48.365 --> 00:17:49.740 align:middle line:84% That doesn't seem to be the case. 00:17:49.740 --> 00:17:51.937 align:middle line:84% In other measures, it seems to always be going up. 00:17:51.937 --> 00:17:54.270 align:middle line:84% So one answer is it just takes some time to learn stuff. 00:17:54.270 --> 00:17:55.710 align:middle line:84% And the tax-- and these tax inspectors 00:17:55.710 --> 00:17:56.668 align:middle line:90% are figuring stuff out. 00:17:56.668 --> 00:17:57.790 align:middle line:90% That could be. 00:17:57.790 --> 00:18:00.720 align:middle line:84% But one thing that we think is kind of interesting 00:18:00.720 --> 00:18:04.230 align:middle line:84% is that there may be also a change in size-- 00:18:04.230 --> 00:18:06.460 align:middle line:84% a reduction in size-dependent taxation. 00:18:06.460 --> 00:18:08.500 align:middle line:90% So what do I mean by this? 00:18:08.500 --> 00:18:10.808 align:middle line:84% So imagine you have a simple model. 00:18:10.808 --> 00:18:13.350 align:middle line:84% This is actually not the model in the final version of paper. 00:18:13.350 --> 00:18:15.650 align:middle line:84% This is a simpler model that was in a previous version 00:18:15.650 --> 00:18:16.430 align:middle line:84% of the paper, but I think it just 00:18:16.430 --> 00:18:17.750 align:middle line:90% makes the point very clearly. 00:18:17.750 --> 00:18:21.066 align:middle line:84% So imagine you're a firm that you-- 00:18:21.066 --> 00:18:24.410 align:middle line:90% 00:18:24.410 --> 00:18:27.560 align:middle line:84% who can both make decisions about how much to pay, and they 00:18:27.560 --> 00:18:30.880 align:middle line:90% can maybe evade some taxes, e. 00:18:30.880 --> 00:18:38.110 align:middle line:84% They're taxed on-- this is a model where there's distortion. 00:18:38.110 --> 00:18:39.990 align:middle line:84% So they're supposed to pay taxes here. 00:18:39.990 --> 00:18:42.900 align:middle line:90% They pay taxes on this. 00:18:42.900 --> 00:18:46.670 align:middle line:84% But they can kind of evade some revenue here. 00:18:46.670 --> 00:18:52.910 align:middle line:84% So the firm's production will be distorted by the taxes. 00:18:52.910 --> 00:18:54.290 align:middle line:90% Because the tax is not-- 00:18:54.290 --> 00:18:56.930 align:middle line:84% if the tax is a profit tax, it doesn't distort your revenue. 00:18:56.930 --> 00:19:01.730 align:middle line:84% But if it's-- but since there's this wedge here it may be-- 00:19:01.730 --> 00:19:03.570 align:middle line:90% there may be some distortions. 00:19:03.570 --> 00:19:06.800 align:middle line:90% 00:19:06.800 --> 00:19:09.750 align:middle line:84% What I wanted to say is suppose that in addition to that, 00:19:09.750 --> 00:19:12.260 align:middle line:90% the cost of evasion now-- 00:19:12.260 --> 00:19:15.620 align:middle line:84% and you have some cost of tax evasion, 00:19:15.620 --> 00:19:19.320 align:middle line:84% where alpha is kind of the level of tax enforcement. 00:19:19.320 --> 00:19:22.860 align:middle line:84% Suppose you have the case where the level of tax enforcement 00:19:22.860 --> 00:19:25.520 align:middle line:90% is dependent on firm size. 00:19:25.520 --> 00:19:28.600 align:middle line:84% So it's pretty intuitive that kind of in a-- 00:19:28.600 --> 00:19:31.280 align:middle line:84% as I mentioned before, they're going after the largest firms 00:19:31.280 --> 00:19:33.220 align:middle line:90% because they're biggest. 00:19:33.220 --> 00:19:39.250 align:middle line:84% If that's true, then that's going to create an additional-- 00:19:39.250 --> 00:19:42.100 align:middle line:84% there'll be a general distortion of taxation. 00:19:42.100 --> 00:19:44.020 align:middle line:84% But there'll be this additional distortion 00:19:44.020 --> 00:19:46.870 align:middle line:84% from taxation coming from the fact that as you grow larger 00:19:46.870 --> 00:19:50.760 align:middle line:84% you're more likely to get taxed by the tax authority. 00:19:50.760 --> 00:19:53.330 align:middle line:90% So if that's true, once-- 00:19:53.330 --> 00:19:54.830 align:middle line:84% so then the question is, once you've 00:19:54.830 --> 00:19:57.980 align:middle line:84% been put into this medium-sized tax organization, 00:19:57.980 --> 00:20:00.502 align:middle line:84% does that change not just the level of enforcement, but also 00:20:00.502 --> 00:20:02.710 align:middle line:84% does it change this slope of enforcement with respect 00:20:02.710 --> 00:20:04.040 align:middle line:90% to firm size? 00:20:04.040 --> 00:20:06.110 align:middle line:84% And if so, that could potentially 00:20:06.110 --> 00:20:08.490 align:middle line:90% explain part of what's going on. 00:20:08.490 --> 00:20:11.870 align:middle line:84% And so we actually also looked at the relationship 00:20:11.870 --> 00:20:14.540 align:middle line:84% between the probability of enforcement 00:20:14.540 --> 00:20:16.080 align:middle line:90% as a function of firm size-- 00:20:16.080 --> 00:20:18.080 align:middle line:84% a function of firm size doesn't have that change 00:20:18.080 --> 00:20:20.330 align:middle line:90% when you move to the MTO. 00:20:20.330 --> 00:20:23.450 align:middle line:84% So for the regular people in the small tax office, 00:20:23.450 --> 00:20:25.680 align:middle line:84% the probability of being audited, for example, 00:20:25.680 --> 00:20:28.740 align:middle line:84% is very strongly increasing in firm size. 00:20:28.740 --> 00:20:32.155 align:middle line:84% Once you're moved to the medium tax office, it's flat. 00:20:32.155 --> 00:20:34.530 align:middle line:84% Basically, they have enough resources to audit everybody, 00:20:34.530 --> 00:20:35.070 align:middle line:90% essentially. 00:20:35.070 --> 00:20:37.320 align:middle line:84% Or to audit not everybody, but with high probability-- 00:20:37.320 --> 00:20:39.833 align:middle line:90% with much higher probability. 00:20:39.833 --> 00:20:42.000 align:middle line:84% And this is true on a couple of different dimensions 00:20:42.000 --> 00:20:43.107 align:middle line:90% that we can measure. 00:20:43.107 --> 00:20:44.940 align:middle line:84% And the reason I wanted to mention that is I 00:20:44.940 --> 00:20:46.890 align:middle line:90% think there's this issue where-- 00:20:46.890 --> 00:20:49.950 align:middle line:90% 00:20:49.950 --> 00:20:53.940 align:middle line:84% I think there could potentially be an issue where in a lot 00:20:53.940 --> 00:21:04.320 align:middle line:84% of these developing country contexts if the tax 00:21:04.320 --> 00:21:07.110 align:middle line:84% department is only starting to pay attention to people who 00:21:07.110 --> 00:21:10.588 align:middle line:84% are kind of larger, that that in itself can create distortions. 00:21:10.588 --> 00:21:12.130 align:middle line:84% And so I just wanted to mention that. 00:21:12.130 --> 00:21:14.190 align:middle line:84% So in this case, we actually have the opposite happen. 00:21:14.190 --> 00:21:15.870 align:middle line:84% Once they got sufficient-- once they were put in the MTO, 00:21:15.870 --> 00:21:17.303 align:middle line:84% at that point, that had already-- 00:21:17.303 --> 00:21:18.220 align:middle line:90% that was kind of sunk. 00:21:18.220 --> 00:21:21.390 align:middle line:84% They were already getting the high tax treatment regardless. 00:21:21.390 --> 00:21:24.380 align:middle line:84% So at that point, they may as well sort of-- they lose that-- 00:21:24.380 --> 00:21:26.790 align:middle line:84% that incentive not to grow too big kind of disappears. 00:21:26.790 --> 00:21:30.435 align:middle line:84% But more generally, that point about size-dependent taxation 00:21:30.435 --> 00:21:31.935 align:middle line:84% creating distortions for a firm size 00:21:31.935 --> 00:21:33.268 align:middle line:90% is one I just wanted to mention. 00:21:33.268 --> 00:21:35.660 align:middle line:90% 00:21:35.660 --> 00:21:37.910 align:middle line:84% That's all I want to say about this paper. 00:21:37.910 --> 00:21:38.600 align:middle line:90% Yeah, Aaron? 00:21:38.600 --> 00:21:41.060 align:middle line:84% AUDIENCE: I'm forgetting the paper, but it was there-- 00:21:41.060 --> 00:21:42.920 align:middle line:84% when these MTOs were created was there 00:21:42.920 --> 00:21:44.936 align:middle line:84% displacement of people who were previously 00:21:44.936 --> 00:21:47.650 align:middle line:84% working in the smaller taxpayer offices? 00:21:47.650 --> 00:21:50.452 align:middle line:84% And was it-- were there effects on revenue 00:21:50.452 --> 00:21:51.410 align:middle line:90% from that displacement? 00:21:51.410 --> 00:21:52.430 align:middle line:84% BENJAMIN OLKEN: I think it was small. 00:21:52.430 --> 00:21:54.290 align:middle line:84% Because the number of staff they needed 00:21:54.290 --> 00:21:56.010 align:middle line:90% for the MTOs was pretty small. 00:21:56.010 --> 00:21:57.530 align:middle line:84% And so I think that they didn't really have to-- 00:21:57.530 --> 00:21:59.070 align:middle line:84% and they could also promote other people 00:21:59.070 --> 00:21:59.900 align:middle line:90% and so on and so forth. 00:21:59.900 --> 00:22:00.400 align:middle line:90% Yeah. 00:22:00.400 --> 00:22:04.500 align:middle line:90% 00:22:04.500 --> 00:22:08.960 align:middle line:84% So if tax administration kind of matters, then 00:22:08.960 --> 00:22:10.760 align:middle line:84% there's now a whole series of questions 00:22:10.760 --> 00:22:12.830 align:middle line:90% for how do we actually do it? 00:22:12.830 --> 00:22:14.810 align:middle line:84% And what are the particular kind of like-- 00:22:14.810 --> 00:22:17.700 align:middle line:90% 00:22:17.700 --> 00:22:19.510 align:middle line:84% and that's true kind of everywhere. 00:22:19.510 --> 00:22:21.510 align:middle line:84% But I guess a question for a development classes 00:22:21.510 --> 00:22:24.150 align:middle line:84% is are there particular development-related issues 00:22:24.150 --> 00:22:26.820 align:middle line:84% in doing this-- in changing tax administration? 00:22:26.820 --> 00:22:30.000 align:middle line:84% So I wanted to mention one last paper, which is 00:22:30.000 --> 00:22:33.880 align:middle line:90% a paper of mine, on this issue. 00:22:33.880 --> 00:22:35.947 align:middle line:84% So one thing people have thought about doing 00:22:35.947 --> 00:22:38.280 align:middle line:84% to improve tax administration is to just give incentives 00:22:38.280 --> 00:22:39.780 align:middle line:90% to tax officers. 00:22:39.780 --> 00:22:41.490 align:middle line:84% The whole idea is we need these people 00:22:41.490 --> 00:22:42.840 align:middle line:90% to go out and collect taxes. 00:22:42.840 --> 00:22:45.300 align:middle line:84% Like in this low-information environment, 00:22:45.300 --> 00:22:48.030 align:middle line:90% we rely on the people. 00:22:48.030 --> 00:22:51.268 align:middle line:84% Maybe we need to give them incentives to do this. 00:22:51.268 --> 00:22:53.810 align:middle line:84% So, in fact, giving high-powered incentives to tax collectors 00:22:53.810 --> 00:22:56.790 align:middle line:90% is a very, very old idea. 00:22:56.790 --> 00:23:00.570 align:middle line:84% In fact, it goes back to the Roman Empire, 00:23:00.570 --> 00:23:03.150 align:middle line:90% or in the French-- 00:23:03.150 --> 00:23:04.980 align:middle line:84% prerevolution France, they had people 00:23:04.980 --> 00:23:06.720 align:middle line:84% who are called tax farmers who basically 00:23:06.720 --> 00:23:08.820 align:middle line:84% paid a fixed fee to the King, and they 00:23:08.820 --> 00:23:11.050 align:middle line:84% got to keep all the taxes on the margin. 00:23:11.050 --> 00:23:16.110 align:middle line:90% So it was like 100% incentive. 00:23:16.110 --> 00:23:17.970 align:middle line:90% This was very unpopular. 00:23:17.970 --> 00:23:19.980 align:middle line:84% So, for example, the tax farmers were beheaded 00:23:19.980 --> 00:23:21.188 align:middle line:90% during the French Revolution. 00:23:21.188 --> 00:23:22.590 align:middle line:90% So that was not-- 00:23:22.590 --> 00:23:24.663 align:middle line:84% people didn't like this very much. 00:23:24.663 --> 00:23:25.830 align:middle line:90% And why didn't they like it? 00:23:25.830 --> 00:23:27.705 align:middle line:84% Well, because these people were super zealous 00:23:27.705 --> 00:23:29.820 align:middle line:84% in terms of collecting taxes because they got 00:23:29.820 --> 00:23:32.220 align:middle line:90% to keep it all on the margin. 00:23:32.220 --> 00:23:34.390 align:middle line:90% So you don't want to go too far. 00:23:34.390 --> 00:23:37.350 align:middle line:90% And, in fact, if you were to-- 00:23:37.350 --> 00:23:40.170 align:middle line:90% in fact, if you look at the-- 00:23:40.170 --> 00:23:41.550 align:middle line:90% you all live in Massachusetts. 00:23:41.550 --> 00:23:43.560 align:middle line:84% If you look at the Massachusetts Department 00:23:43.560 --> 00:23:47.117 align:middle line:84% of Revenue's website, there's a taxpayer bill of rights. 00:23:47.117 --> 00:23:49.200 align:middle line:84% And one of the things that Massachusetts states is 00:23:49.200 --> 00:23:51.120 align:middle line:84% that they will not under any circumstances 00:23:51.120 --> 00:23:56.130 align:middle line:84% create any incentives or even informal targets for revenue 00:23:56.130 --> 00:23:59.810 align:middle line:84% for any of their tax staff, in part because tax payers 00:23:59.810 --> 00:24:02.060 align:middle line:90% don't like this. 00:24:02.060 --> 00:24:05.805 align:middle line:84% So I think-- so we wanted to understand 00:24:05.805 --> 00:24:07.680 align:middle line:84% in a more modern context-- on the other hand, 00:24:07.680 --> 00:24:09.260 align:middle line:84% though, if you to a lot of these developing country 00:24:09.260 --> 00:24:11.900 align:middle line:84% contexts where the information problems are really severe, 00:24:11.900 --> 00:24:14.930 align:middle line:84% and you have to motivate these people to go collect taxes, 00:24:14.930 --> 00:24:17.060 align:middle line:84% maybe the state should think about revisiting 00:24:17.060 --> 00:24:17.940 align:middle line:90% this trade-off. 00:24:17.940 --> 00:24:20.240 align:middle line:90% So how do we think about this? 00:24:20.240 --> 00:24:23.440 align:middle line:84% So we did this by doing a randomized experiment 00:24:23.440 --> 00:24:30.040 align:middle line:84% in Pakistan where we were working with property taxes. 00:24:30.040 --> 00:24:34.360 align:middle line:84% And property taxes are collected by a team of tax collectors. 00:24:34.360 --> 00:24:36.430 align:middle line:84% There's a team of three tax staff who-- 00:24:36.430 --> 00:24:38.680 align:middle line:84% and together, if they were put in the treatment group, 00:24:38.680 --> 00:24:41.860 align:middle line:84% they receive between 20% to 40% of all the revenue collected 00:24:41.860 --> 00:24:45.230 align:middle line:90% above some historical benchmark. 00:24:45.230 --> 00:24:48.340 align:middle line:84% So you can think of each of them as getting a 10% kind 00:24:48.340 --> 00:24:52.090 align:middle line:90% of payment on the margin. 00:24:52.090 --> 00:24:54.810 align:middle line:84% So any thoughts on what would likely to happen 00:24:54.810 --> 00:24:58.215 align:middle line:90% if you did this treatment? 00:24:58.215 --> 00:24:59.840 align:middle line:84% So you see why they would do it, right? 00:24:59.840 --> 00:25:02.040 align:middle line:84% They would do it because like-- so let me back up. 00:25:02.040 --> 00:25:04.610 align:middle line:84% Why do they need this incentive in Pakistan and not 00:25:04.610 --> 00:25:06.650 align:middle line:84% in Brookline, for example, where I live? 00:25:06.650 --> 00:25:10.010 align:middle line:84% So in Brookline, they have very good information 00:25:10.010 --> 00:25:12.650 align:middle line:90% about all the houses. 00:25:12.650 --> 00:25:17.750 align:middle line:84% And everyone's required to build a-- to have a building permit. 00:25:17.750 --> 00:25:19.250 align:middle line:84% The building permit is automatically 00:25:19.250 --> 00:25:20.360 align:middle line:90% linked to the tax people. 00:25:20.360 --> 00:25:22.130 align:middle line:84% If you change your house, the building people know. 00:25:22.130 --> 00:25:24.338 align:middle line:84% The town knows about it through the building permits. 00:25:24.338 --> 00:25:25.903 align:middle line:84% They can send the tax people out. 00:25:25.903 --> 00:25:28.070 align:middle line:84% They sort of-- and it's all kind of linked together. 00:25:28.070 --> 00:25:29.840 align:middle line:84% So they do rely on tax inspectors, 00:25:29.840 --> 00:25:32.750 align:middle line:84% but they have much more data underlying them. 00:25:32.750 --> 00:25:35.270 align:middle line:90% And also, the penal-- 00:25:35.270 --> 00:25:37.160 align:middle line:84% if the tax inspector were to come to my house 00:25:37.160 --> 00:25:40.820 align:middle line:84% and offer to adjust my taxes in return for a cash 00:25:40.820 --> 00:25:43.790 align:middle line:84% payment, the probability that would end up in detection, 00:25:43.790 --> 00:25:46.160 align:middle line:84% and they would end up going to jail 00:25:46.160 --> 00:25:47.510 align:middle line:90% is probably reasonably high. 00:25:47.510 --> 00:25:49.680 align:middle line:84% And so that probably doesn't happen very often. 00:25:49.680 --> 00:25:52.190 align:middle line:84% So that's-- but in Pakistan, that might not be true right. 00:25:52.190 --> 00:25:56.030 align:middle line:84% So what do you think is going to happen in this-- 00:25:56.030 --> 00:25:58.028 align:middle line:84% so that's why they might want to do incentives. 00:25:58.028 --> 00:26:00.320 align:middle line:84% So what do you think's going to happen with incentives? 00:26:00.320 --> 00:26:05.120 align:middle line:90% 00:26:05.120 --> 00:26:06.300 align:middle line:90% Yeah, Jenny. 00:26:06.300 --> 00:26:08.434 align:middle line:84% AUDIENCE: They exert more effort so there's 00:26:08.434 --> 00:26:09.648 align:middle line:90% more total tax revenue. 00:26:09.648 --> 00:26:10.440 align:middle line:90% BENJAMIN OLKEN: OK. 00:26:10.440 --> 00:26:10.940 align:middle line:90% Right. 00:26:10.940 --> 00:26:12.780 align:middle line:90% So incentives matter. 00:26:12.780 --> 00:26:16.460 align:middle line:84% They should exert more effort and collect more tax revenue. 00:26:16.460 --> 00:26:18.350 align:middle line:90% Paulo? 00:26:18.350 --> 00:26:21.230 align:middle line:84% AUDIENCE: If there is like a collusive evasion 00:26:21.230 --> 00:26:25.460 align:middle line:84% between the tax staff and the person being taxed, 00:26:25.460 --> 00:26:28.610 align:middle line:84% you might not observe any change in revenue, 00:26:28.610 --> 00:26:31.130 align:middle line:84% and you just have a change in how they bargain for, like-- 00:26:31.130 --> 00:26:31.520 align:middle line:90% BENJAMIN OLKEN: Exactly. 00:26:31.520 --> 00:26:33.410 align:middle line:84% You've seen this paper before in a previ-- yes. 00:26:33.410 --> 00:26:33.910 align:middle line:90% But yes. 00:26:33.910 --> 00:26:34.520 align:middle line:90% Exactly. 00:26:34.520 --> 00:26:38.030 align:middle line:90% So the other thing is-- 00:26:38.030 --> 00:26:39.140 align:middle line:90% so Jenny is totally right. 00:26:39.140 --> 00:26:40.830 align:middle line:84% They could collect more tax revenue. 00:26:40.830 --> 00:26:43.655 align:middle line:84% On the other hand, if they're bargaining 00:26:43.655 --> 00:26:45.380 align:middle line:84% with-- if they're potentially corrupt 00:26:45.380 --> 00:26:47.660 align:middle line:84% and kind of bargaining with the taxpayer, 00:26:47.660 --> 00:26:49.820 align:middle line:84% this could just-- all this could do is kind of 00:26:49.820 --> 00:26:52.650 align:middle line:84% change the underlying bargaining game. 00:26:52.650 --> 00:26:54.900 align:middle line:84% And so that's what we want to illustrate in this paper 00:26:54.900 --> 00:26:57.830 align:middle line:84% is how do we think about those two different things. 00:26:57.830 --> 00:27:02.930 align:middle line:84% So we set up in the paper a very simple kind of Nash bargaining 00:27:02.930 --> 00:27:07.930 align:middle line:84% model with equal bargaining weights between a taxpayer, P, 00:27:07.930 --> 00:27:10.960 align:middle line:84% and a tax collector, C, to collude and reduce 00:27:10.960 --> 00:27:12.010 align:middle line:90% their tax liabilities. 00:27:12.010 --> 00:27:13.010 align:middle line:90% What does this mean? 00:27:13.010 --> 00:27:15.040 align:middle line:90% This means I'm the taxpayer. 00:27:15.040 --> 00:27:17.820 align:middle line:84% I go to, say, Ashana, and I say-- 00:27:17.820 --> 00:27:18.760 align:middle line:90% I'm a tax collector. 00:27:18.760 --> 00:27:21.340 align:middle line:84% I go to you and you say, look, your house is worth $1,000-- 00:27:21.340 --> 00:27:23.020 align:middle line:90% your tax is $1,000. 00:27:23.020 --> 00:27:25.390 align:middle line:84% But maybe you and I can work something out. 00:27:25.390 --> 00:27:27.700 align:middle line:84% And in exchange for a payment, maybe I 00:27:27.700 --> 00:27:30.460 align:middle line:84% can reduce your official tax liability. 00:27:30.460 --> 00:27:33.540 align:middle line:84% So how does that kind of bargaining work? 00:27:33.540 --> 00:27:37.140 align:middle line:84% So imagine that tau star is the true amount of tax, 00:27:37.140 --> 00:27:39.040 align:middle line:90% which is the same for everyone. 00:27:39.040 --> 00:27:42.540 align:middle line:84% We can instead negotiate to pay a bribe, b, 00:27:42.540 --> 00:27:48.660 align:middle line:84% and report less tax, tau, to the government. 00:27:48.660 --> 00:27:51.720 align:middle line:90% Together we can do that. 00:27:51.720 --> 00:27:57.780 align:middle line:84% The taxpayer's utility is going to be minus tau-- 00:27:57.780 --> 00:27:59.610 align:middle line:90% they don't like paying taxes-- 00:27:59.610 --> 00:28:02.340 align:middle line:84% minus b-- they don't like paying bribes. 00:28:02.340 --> 00:28:05.100 align:middle line:84% Taxes and bribes are going to enter the same here. 00:28:05.100 --> 00:28:07.230 align:middle line:84% Those are just money kind of flowing out the door. 00:28:07.230 --> 00:28:09.570 align:middle line:84% And we're going to give them an additional cost 00:28:09.570 --> 00:28:13.110 align:middle line:84% like alpha times tau star minus tau, which is the-- like a disk 00:28:13.110 --> 00:28:17.430 align:middle line:84% utility cost from paying bribes, or from underpaying 00:28:17.430 --> 00:28:17.930 align:middle line:90% their taxes. 00:28:17.930 --> 00:28:19.830 align:middle line:90% Sorry. 00:28:19.830 --> 00:28:21.240 align:middle line:84% And maybe there's the different-- 00:28:21.240 --> 00:28:26.050 align:middle line:84% that alpha is heterogeneous among taxpayers. 00:28:26.050 --> 00:28:29.210 align:middle line:84% The tax collector's utility is going to look like this. 00:28:29.210 --> 00:28:31.390 align:middle line:90% It's going to have-- 00:28:31.390 --> 00:28:34.060 align:middle line:84% they're going to get an incentive payment, r, 00:28:34.060 --> 00:28:37.430 align:middle line:90% times the taxes they pay, r tau. 00:28:37.430 --> 00:28:38.838 align:middle line:90% There also may be some bribes. 00:28:38.838 --> 00:28:39.505 align:middle line:90% They like money. 00:28:39.505 --> 00:28:41.120 align:middle line:84% And they let-- money is treated equally here. 00:28:41.120 --> 00:28:41.995 align:middle line:90% They like incentives. 00:28:41.995 --> 00:28:42.950 align:middle line:90% They like bribes. 00:28:42.950 --> 00:28:44.660 align:middle line:84% But they also get a disutility cost 00:28:44.660 --> 00:28:47.360 align:middle line:84% from underpay-- underreporting their true taxes. 00:28:47.360 --> 00:28:56.660 align:middle line:84% And you can think of that as like some chance 00:28:56.660 --> 00:28:59.180 align:middle line:84% they get caught, or something-- maybe they want to be honest 00:28:59.180 --> 00:29:01.020 align:middle line:90% or whatever. 00:29:01.020 --> 00:29:02.880 align:middle line:84% And this simple model, everything is linear. 00:29:02.880 --> 00:29:05.240 align:middle line:84% There's a version in the paper in the appendix that 00:29:05.240 --> 00:29:07.936 align:middle line:84% makes this quadratic, and the linear version is easy. 00:29:07.936 --> 00:29:10.615 align:middle line:90% 00:29:10.615 --> 00:29:11.865 align:middle line:90% So what happens in this model? 00:29:11.865 --> 00:29:14.370 align:middle line:90% 00:29:14.370 --> 00:29:18.250 align:middle line:84% So the Nash bargaining solution says, 00:29:18.250 --> 00:29:21.395 align:middle line:84% look, we're going to come to an agreement that's 00:29:21.395 --> 00:29:22.770 align:middle line:84% going to max-- that's going to be 00:29:22.770 --> 00:29:25.907 align:middle line:84% efficient for the two of us-- for me and the taxpayer. 00:29:25.907 --> 00:29:26.740 align:middle line:90% What does that mean? 00:29:26.740 --> 00:29:29.340 align:middle line:84% That means we're going to maximize our joint surplus. 00:29:29.340 --> 00:29:32.050 align:middle line:84% And then we have to share the surplus somehow. 00:29:32.050 --> 00:29:34.220 align:middle line:84% And different bargaining models get 00:29:34.220 --> 00:29:36.220 align:middle line:84% predictions for how we should share the surplus. 00:29:36.220 --> 00:29:37.610 align:middle line:84% The Nash bargaining model says, look, 00:29:37.610 --> 00:29:39.402 align:middle line:84% we're each going to get our outside option. 00:29:39.402 --> 00:29:42.400 align:middle line:84% And then we're going to get some fixed share of the surplus. 00:29:42.400 --> 00:29:44.910 align:middle line:90% 00:29:44.910 --> 00:29:51.310 align:middle line:84% So what's-- so the joint surplus from agreement is we add up 00:29:51.310 --> 00:29:56.180 align:middle line:84% the taxpayer's utility and the tax collector's utility. 00:29:56.180 --> 00:29:58.600 align:middle line:90% We just add them up. 00:29:58.600 --> 00:30:00.850 align:middle line:90% And so I can rewrite terms. 00:30:00.850 --> 00:30:03.850 align:middle line:84% And I have this 1 minus r minus alpha minus beta times 00:30:03.850 --> 00:30:04.900 align:middle line:90% tau star minus tau. 00:30:04.900 --> 00:30:06.460 align:middle line:84% Tau star is the true tax liability. 00:30:06.460 --> 00:30:10.490 align:middle line:90% Tau is the amount we report. 00:30:10.490 --> 00:30:12.880 align:middle line:84% So what you can see is that as long as 1 00:30:12.880 --> 00:30:18.350 align:middle line:84% minus r minus alpha minus beta is greater than zero, 00:30:18.350 --> 00:30:20.390 align:middle line:84% then we're going to want to under-- we're 00:30:20.390 --> 00:30:23.030 align:middle line:84% going to want to not report any taxes. 00:30:23.030 --> 00:30:26.190 align:middle line:84% We'll maximize-- I mean, so you can't have negative taxes here, 00:30:26.190 --> 00:30:29.290 align:middle line:90% so we'll set tau equal to zero. 00:30:29.290 --> 00:30:31.040 align:middle line:84% And if this thing was less than zero, then 00:30:31.040 --> 00:30:33.710 align:middle line:84% we might as well tell the truth and pay tau equals tau star. 00:30:33.710 --> 00:30:37.350 align:middle line:90% 00:30:37.350 --> 00:30:40.260 align:middle line:84% So that tells us something interesting. 00:30:40.260 --> 00:30:45.450 align:middle line:84% That says, look, for some combinations 00:30:45.450 --> 00:30:47.610 align:middle line:90% of taxpayers and tax-- 00:30:47.610 --> 00:30:53.580 align:middle line:84% taxpayers and tax inspectors, they'd 00:30:53.580 --> 00:30:56.520 align:middle line:84% be in the collusive equilibrium if r is equal to 0. 00:30:56.520 --> 00:30:58.000 align:middle line:84% If alpha plus beta is less than 1, 00:30:58.000 --> 00:30:59.625 align:middle line:84% they'll be in the collusive equilibrium 00:30:59.625 --> 00:31:01.380 align:middle line:90% if r is less than 0. 00:31:01.380 --> 00:31:05.370 align:middle line:84% But as I start to increase r, then some of them 00:31:05.370 --> 00:31:06.870 align:middle line:84% are going to be like, you know what? 00:31:06.870 --> 00:31:07.953 align:middle line:90% It's not worth it anymore. 00:31:07.953 --> 00:31:10.120 align:middle line:84% We're going to switch from the collusive equilibrium 00:31:10.120 --> 00:31:12.450 align:middle line:84% to the noncollusive equilibrium and just pay taxes. 00:31:12.450 --> 00:31:14.580 align:middle line:84% And intuitively, what's going on here-- 00:31:14.580 --> 00:31:17.210 align:middle line:84% what's going on is I'm getting some utility losses 00:31:17.210 --> 00:31:21.060 align:middle line:90% from this collusion thing. 00:31:21.060 --> 00:31:22.860 align:middle line:84% And the government is sweetening the deal 00:31:22.860 --> 00:31:24.110 align:middle line:90% by throwing in this incentive. 00:31:24.110 --> 00:31:26.443 align:middle line:84% And it's just-- at some point, it gets to the case where 00:31:26.443 --> 00:31:28.740 align:middle line:84% the incentive plus the two utility losses it makes 00:31:28.740 --> 00:31:30.650 align:middle line:90% the whole thing not worth it-- 00:31:30.650 --> 00:31:31.910 align:middle line:90% the foregone incentive, right? 00:31:31.910 --> 00:31:33.950 align:middle line:84% So that says, look, so one thing that's 00:31:33.950 --> 00:31:38.090 align:middle line:84% going to happen in this model is I'm going to stop-- 00:31:38.090 --> 00:31:40.970 align:middle line:84% some people will stop paying-- stop colluding 00:31:40.970 --> 00:31:44.700 align:middle line:90% and pay more taxes. 00:31:44.700 --> 00:31:46.770 align:middle line:84% But the other thing that will happen 00:31:46.770 --> 00:31:50.600 align:middle line:84% is that-- that's this case over here. 00:31:50.600 --> 00:31:52.080 align:middle line:84% The other thing that will happen, 00:31:52.080 --> 00:31:59.190 align:middle line:84% though, is that if I stay colluding, 00:31:59.190 --> 00:32:02.470 align:middle line:84% the amount of bribes I have to pay is going to go up. 00:32:02.470 --> 00:32:03.308 align:middle line:90% Why is that? 00:32:03.308 --> 00:32:05.850 align:middle line:84% So that's what-- that's-- you can see that in this expression 00:32:05.850 --> 00:32:07.082 align:middle line:90% here, that's just-- 00:32:07.082 --> 00:32:08.790 align:middle line:84% so how do you solve this expression here? 00:32:08.790 --> 00:32:14.820 align:middle line:84% You basically solve for this by saying, what is the bribe? 00:32:14.820 --> 00:32:16.920 align:middle line:84% The bribe is a way of transferring 00:32:16.920 --> 00:32:19.650 align:middle line:84% some share of the surplus from the taxpayer 00:32:19.650 --> 00:32:24.100 align:middle line:84% to the tax inspector in this model. 00:32:24.100 --> 00:32:25.720 align:middle line:84% Because if we didn't have any payment, 00:32:25.720 --> 00:32:28.683 align:middle line:84% then the taxpayer would get all the surplus. 00:32:28.683 --> 00:32:30.100 align:middle line:84% But they had to share the surplus. 00:32:30.100 --> 00:32:33.860 align:middle line:84% So they share a surplus by paying through the bribe. 00:32:33.860 --> 00:32:35.600 align:middle line:90% And well, what happened? 00:32:35.600 --> 00:32:38.323 align:middle line:84% Well, remember that each person gets a share 00:32:38.323 --> 00:32:39.365 align:middle line:90% of their outside option-- 00:32:39.365 --> 00:32:41.880 align:middle line:90% 00:32:41.880 --> 00:32:44.950 align:middle line:84% they get their outside options, plus a share of the surplus. 00:32:44.950 --> 00:32:48.120 align:middle line:84% So now the tax inspector's outside option has gone up. 00:32:48.120 --> 00:32:49.890 align:middle line:84% Because if they don't collude, the tax 00:32:49.890 --> 00:32:52.120 align:middle line:84% inspectors getting the incentive from the government. 00:32:52.120 --> 00:32:55.050 align:middle line:84% So now the taxpayer has to compensate 00:32:55.050 --> 00:32:57.300 align:middle line:84% the tax inspector for their forgone incentive, 00:32:57.300 --> 00:32:58.723 align:middle line:90% plus you have a surplus. 00:32:58.723 --> 00:33:01.140 align:middle line:84% So that means that the amount they have to pay, the bribe, 00:33:01.140 --> 00:33:03.270 align:middle line:84% is going to be increasing in the incentive. 00:33:03.270 --> 00:33:05.400 align:middle line:84% And that-- if you just-- if you-- 00:33:05.400 --> 00:33:07.800 align:middle line:84% the math gets a little ugly with the alphas and betas. 00:33:07.800 --> 00:33:13.405 align:middle line:84% But if you just take the outside option and a share, a gamma, 00:33:13.405 --> 00:33:15.280 align:middle line:84% of the surplus, you get this expression here, 00:33:15.280 --> 00:33:18.603 align:middle line:84% which is increasing in r-- the incentive. 00:33:18.603 --> 00:33:21.990 align:middle line:90% 00:33:21.990 --> 00:33:23.490 align:middle line:90% So that's kind of the-- 00:33:23.490 --> 00:33:25.320 align:middle line:84% and you can do a simple version of this. 00:33:25.320 --> 00:33:27.480 align:middle line:84% Like, imagine if there was no incentive-- 00:33:27.480 --> 00:33:29.160 align:middle line:84% imagine there was no alpha, no beta. 00:33:29.160 --> 00:33:33.330 align:middle line:84% If it was a simple version, say, bargaining over $1, 00:33:33.330 --> 00:33:36.120 align:middle line:84% we split the dollar and pay $0.50. 00:33:36.120 --> 00:33:38.640 align:middle line:84% Now imagine the tax inspector was getting, say, 00:33:38.640 --> 00:33:43.050 align:middle line:84% $0.30 in incentive-- a 30% incentive. 00:33:43.050 --> 00:33:45.990 align:middle line:90% The surplus is now $0.70. 00:33:45.990 --> 00:33:48.065 align:middle line:90% Half the surplus is $0.35. 00:33:48.065 --> 00:33:49.440 align:middle line:84% What's the bribe you have to pay? 00:33:49.440 --> 00:33:52.207 align:middle line:84% It's the $0.30 of foregone incentive plus the $0.35 00:33:52.207 --> 00:33:53.040 align:middle line:90% of half the surplus. 00:33:53.040 --> 00:33:55.785 align:middle line:84% The bribe goes to $0.50 to $0.65. 00:33:55.785 --> 00:33:57.160 align:middle line:84% And that's kind of the intuition. 00:33:57.160 --> 00:33:59.950 align:middle line:90% 00:33:59.950 --> 00:34:02.330 align:middle line:84% Is that-- I'm going through this kind of fast, but is 00:34:02.330 --> 00:34:04.370 align:middle line:90% that clear? 00:34:04.370 --> 00:34:06.170 align:middle line:90% Yes? 00:34:06.170 --> 00:34:06.670 align:middle line:90% OK. 00:34:06.670 --> 00:34:09.969 align:middle line:84% So what that suggests is that as we-- 00:34:09.969 --> 00:34:11.710 align:middle line:90% I've basically said that. 00:34:11.710 --> 00:34:12.670 align:middle line:90% I said all this, right? 00:34:12.670 --> 00:34:13.909 align:middle line:84% So we have two different things going on. 00:34:13.909 --> 00:34:16.576 align:middle line:84% We have an equilibrium selection effect and an equilibrium bribe 00:34:16.576 --> 00:34:17.590 align:middle line:90% amount going up. 00:34:17.590 --> 00:34:19.449 align:middle line:84% And that suggests if we look in the data, 00:34:19.449 --> 00:34:20.949 align:middle line:84% we're going to want to look for both 00:34:20.949 --> 00:34:22.340 align:middle line:90% of those two different things. 00:34:22.340 --> 00:34:23.300 align:middle line:90% And that's what we do. 00:34:23.300 --> 00:34:24.790 align:middle line:90% So we chose property tax. 00:34:24.790 --> 00:34:27.527 align:middle line:84% So the first thing is we get tax revenue. 00:34:27.527 --> 00:34:29.110 align:middle line:84% That goes-- this is actual tax revenue 00:34:29.110 --> 00:34:30.280 align:middle line:90% from administrative data. 00:34:30.280 --> 00:34:31.150 align:middle line:90% This is in log. 00:34:31.150 --> 00:34:34.900 align:middle line:84% This goes up by about 9% in the first year, 00:34:34.900 --> 00:34:36.380 align:middle line:90% 9.4% in the second year. 00:34:36.380 --> 00:34:40.120 align:middle line:84% So the government really is collecting more tax revenue. 00:34:40.120 --> 00:34:42.730 align:middle line:84% But to look for this kind of other thing going on, 00:34:42.730 --> 00:34:49.560 align:middle line:84% we did a survey, and we looked both at the general population 00:34:49.560 --> 00:34:52.280 align:middle line:84% and at the small subset of people whose actual official 00:34:52.280 --> 00:34:53.239 align:middle line:90% taxes-- 00:34:53.239 --> 00:34:54.630 align:middle line:90% tax liabilities change. 00:34:54.630 --> 00:34:56.810 align:middle line:84% Let's just call this re-assessed sample. 00:34:56.810 --> 00:35:00.260 align:middle line:84% And for the general population, they report in treatment areas 00:35:00.260 --> 00:35:02.150 align:middle line:90% a higher going bribery. 00:35:02.150 --> 00:35:03.080 align:middle line:90% What does that mean? 00:35:03.080 --> 00:35:06.110 align:middle line:84% We ask the question, what is the typical rate 00:35:06.110 --> 00:35:08.000 align:middle line:84% people in your area might have to pay 00:35:08.000 --> 00:35:10.750 align:middle line:90% to get your taxes adjusted? 00:35:10.750 --> 00:35:13.130 align:middle line:84% That was-- we didn't ask what you did. 00:35:13.130 --> 00:35:15.200 align:middle line:84% But I said, what's the going rate in your area? 00:35:15.200 --> 00:35:17.460 align:middle line:84% And that's going up by about a third-- 00:35:17.460 --> 00:35:19.360 align:middle line:84% by about 30% roughly in the treatment area, 00:35:19.360 --> 00:35:21.070 align:middle line:90% as well as the control areas. 00:35:21.070 --> 00:35:24.220 align:middle line:84% How often people in your area tend to pay bribe payments also 00:35:24.220 --> 00:35:26.113 align:middle line:90% goes up. 00:35:26.113 --> 00:35:28.030 align:middle line:84% We have a subjective perception of corruption. 00:35:28.030 --> 00:35:31.530 align:middle line:84% That doesn't actually change, but both these things go up. 00:35:31.530 --> 00:35:36.990 align:middle line:84% We then also asked, how is this different for this group that 00:35:36.990 --> 00:35:41.440 align:middle line:84% was actually reassessed or seeing their taxes changed? 00:35:41.440 --> 00:35:45.520 align:middle line:84% So these people pay a lot more in tax. 00:35:45.520 --> 00:35:47.220 align:middle line:90% So their taxes are going way up. 00:35:47.220 --> 00:35:50.070 align:middle line:84% And they don't report any-- this increase in bribes. 00:35:50.070 --> 00:35:51.722 align:middle line:84% So this is like the reassessed people. 00:35:51.722 --> 00:35:53.430 align:middle line:84% And the reassessed in the treatment areas 00:35:53.430 --> 00:35:55.020 align:middle line:84% is almost like the opposite of this. 00:35:55.020 --> 00:35:57.228 align:middle line:84% So they're not actually seeing any changes in bribes, 00:35:57.228 --> 00:35:59.583 align:middle line:84% nor are they seeing any changes in bribe frequencies. 00:35:59.583 --> 00:36:01.500 align:middle line:84% They're not actually seeing it go down on net, 00:36:01.500 --> 00:36:03.493 align:middle line:84% but they're not seeing these big increases. 00:36:03.493 --> 00:36:04.910 align:middle line:84% So that's consistent with the idea 00:36:04.910 --> 00:36:06.950 align:middle line:84% that there's two different things going on. 00:36:06.950 --> 00:36:10.812 align:middle line:84% One group is seeing the high change in-- 00:36:10.812 --> 00:36:13.910 align:middle line:90% 00:36:13.910 --> 00:36:18.230 align:middle line:84% the high change in bribes, and the other one 00:36:18.230 --> 00:36:20.512 align:middle line:84% is seeing-- like, paying a lot more taxes. 00:36:20.512 --> 00:36:22.720 align:middle line:84% I think that's potentially consistent with the model. 00:36:22.720 --> 00:36:25.660 align:middle line:90% 00:36:25.660 --> 00:36:27.980 align:middle line:84% We also look at whether they're being overtaxed. 00:36:27.980 --> 00:36:28.480 align:middle line:90% They're not. 00:36:28.480 --> 00:36:30.520 align:middle line:84% They're basically moving from being undertaxed 00:36:30.520 --> 00:36:31.702 align:middle line:90% to being correctly taxed. 00:36:31.702 --> 00:36:33.160 align:middle line:84% And we'll just leave that at that-- 00:36:33.160 --> 00:36:34.243 align:middle line:90% more of that in the paper. 00:36:34.243 --> 00:36:37.530 align:middle line:90% 00:36:37.530 --> 00:36:39.450 align:middle line:84% So what's the point of going through this? 00:36:39.450 --> 00:36:44.538 align:middle line:84% This is one example, I think, of how we can think about the-- 00:36:44.538 --> 00:36:46.830 align:middle line:84% if you're moving to this question of tax administration 00:36:46.830 --> 00:36:49.470 align:middle line:84% as a solution to some of the tax problems in these developing 00:36:49.470 --> 00:36:51.720 align:middle line:84% country contexts, then you have to think about the realities 00:36:51.720 --> 00:36:52.450 align:middle line:90% of how you do this. 00:36:52.450 --> 00:36:53.640 align:middle line:84% And this is some of the challenges 00:36:53.640 --> 00:36:54.420 align:middle line:90% that you may actually face. 00:36:54.420 --> 00:36:56.610 align:middle line:84% And there's a lot more stuff you could do in this area. 00:36:56.610 --> 00:36:58.402 align:middle line:84% I just wanted to leave it leave it at that. 00:36:58.402 --> 00:37:01.460 align:middle line:90% 00:37:01.460 --> 00:37:05.030 align:middle line:84% I think I'll stop here and switch to the labor section, 00:37:05.030 --> 00:37:08.100 align:middle line:84% there's unless there's any other questions. 00:37:08.100 --> 00:37:09.380 align:middle line:90% OK. 00:37:09.380 --> 00:37:11.150 align:middle line:90% Actually, let me just give you-- 00:37:11.150 --> 00:37:16.160 align:middle line:84% let me just give you one slide on this, which is-- 00:37:16.160 --> 00:37:18.140 align:middle line:84% the last topic that I just wanted to mention 00:37:18.140 --> 00:37:20.430 align:middle line:84% is this idea of informal taxation. 00:37:20.430 --> 00:37:25.160 align:middle line:84% So in the same way that there are 00:37:25.160 --> 00:37:28.340 align:middle line:84% ideas about informal insurance or other informal mechanisms, 00:37:28.340 --> 00:37:32.630 align:middle line:84% like supplementing when the formal systems break down, 00:37:32.630 --> 00:37:36.030 align:middle line:84% that people are insuring each other or doing other things. 00:37:36.030 --> 00:37:39.450 align:middle line:84% I think the same ideas actually also apply to taxation. 00:37:39.450 --> 00:37:42.410 align:middle line:84% And we have a paper that discusses 00:37:42.410 --> 00:37:44.210 align:middle line:84% a few of these phenomenon and tries 00:37:44.210 --> 00:37:46.712 align:middle line:84% to think about how do we think about some of these-- 00:37:46.712 --> 00:37:48.920 align:middle line:84% what is this phenomenon, and how do we think about it 00:37:48.920 --> 00:37:49.820 align:middle line:90% as a tax perspective? 00:37:49.820 --> 00:37:53.150 align:middle line:84% And basically, there are a lot of cases where in local village 00:37:53.150 --> 00:37:55.880 align:middle line:84% context, for example, there are these phenomena where basically 00:37:55.880 --> 00:37:59.360 align:middle line:84% people are kind of contributing to local public goods rather-- 00:37:59.360 --> 00:38:01.290 align:middle line:84% for things that-- in the US, for example, 00:38:01.290 --> 00:38:03.060 align:middle line:84% we would do through the tax system. 00:38:03.060 --> 00:38:04.980 align:middle line:90% So, for example, one of-- 00:38:04.980 --> 00:38:08.180 align:middle line:84% I've been in these villages in Indonesia 00:38:08.180 --> 00:38:10.430 align:middle line:90% where we need to pave a road. 00:38:10.430 --> 00:38:13.227 align:middle line:84% In Brookline when we need to pave a road they set our taxes, 00:38:13.227 --> 00:38:14.810 align:middle line:84% and they raise revenue from the taxes, 00:38:14.810 --> 00:38:18.073 align:middle line:84% and they hire a road-paving company to pave the road. 00:38:18.073 --> 00:38:19.490 align:middle line:84% In an Indonesian village what they 00:38:19.490 --> 00:38:21.948 align:middle line:84% might do is they might say, look, we need to pave the road. 00:38:21.948 --> 00:38:24.770 align:middle line:84% So we're all going to get together and have a meeting. 00:38:24.770 --> 00:38:27.350 align:middle line:84% And we're going to decide how much we would like 00:38:27.350 --> 00:38:28.700 align:middle line:90% you to contribute to the road. 00:38:28.700 --> 00:38:31.370 align:middle line:84% And what are you all willing to contribute? 00:38:31.370 --> 00:38:34.610 align:middle line:84% And Sal is willing to contribute two days of labor. 00:38:34.610 --> 00:38:37.880 align:middle line:84% And Aaron is willing to contribute a gallon of-- 00:38:37.880 --> 00:38:40.092 align:middle line:84% a tank of asphalt, and so on and so forth. 00:38:40.092 --> 00:38:42.050 align:middle line:84% And we'll get all these voluntary contributions 00:38:42.050 --> 00:38:44.280 align:middle line:84% and use that together to build the road. 00:38:44.280 --> 00:38:46.500 align:middle line:84% At some level, these are like the same thing. 00:38:46.500 --> 00:38:49.590 align:middle line:84% We're both collect-- gathering a bunch of contributions 00:38:49.590 --> 00:38:53.280 align:middle line:84% from the population to fund some public good. 00:38:53.280 --> 00:38:55.890 align:middle line:84% But in one case, we're doing it through the tax system, which 00:38:55.890 --> 00:38:57.840 align:middle line:84% is both formal and kind of has these very 00:38:57.840 --> 00:39:01.782 align:middle line:84% strong formal punishments for refusal to comply. 00:39:01.782 --> 00:39:03.740 align:middle line:84% Or we could do it through this informal system. 00:39:03.740 --> 00:39:06.462 align:middle line:84% And so in this paper we think about, 00:39:06.462 --> 00:39:07.670 align:middle line:90% what is this informal system? 00:39:07.670 --> 00:39:08.795 align:middle line:90% And what does it look like? 00:39:08.795 --> 00:39:10.000 align:middle line:90% And how progressive is it? 00:39:10.000 --> 00:39:11.535 align:middle line:90% And what's going on? 00:39:11.535 --> 00:39:12.910 align:middle line:84% And this is an area where I think 00:39:12.910 --> 00:39:16.600 align:middle line:84% we've done a little work through this paper. 00:39:16.600 --> 00:39:18.608 align:middle line:84% But I think that understanding a little more 00:39:18.608 --> 00:39:20.650 align:middle line:84% of these voluntary systems providing public goods 00:39:20.650 --> 00:39:24.540 align:middle line:84% is an interesting area for future work and development. 00:39:24.540 --> 00:39:29.100 align:middle line:84% So I just wanted to leave it out there at that. 00:39:29.100 --> 00:39:30.202 align:middle line:90% OK? 00:39:30.202 --> 00:39:32.660 align:middle line:84% So we have some stylized facts in this paper, which I'm not 00:39:32.660 --> 00:39:33.660 align:middle line:90% going to go through now. 00:39:33.660 --> 00:39:37.430 align:middle line:84% But if you're interested, you could look at that. 00:39:37.430 --> 00:39:37.930 align:middle line:90% OK. 00:39:37.930 --> 00:39:40.150 align:middle line:90% Other questions? 00:39:40.150 --> 00:39:40.650 align:middle line:90% OK. 00:39:40.650 --> 00:39:42.150 align:middle line:84% Sorry if I was a little bit rushing, 00:39:42.150 --> 00:39:48.540 align:middle line:84% but I wanted to make sure we get on to the labor topics. 00:39:48.540 --> 00:39:49.770 align:middle line:90% OK. 00:39:49.770 --> 00:39:51.285 align:middle line:90% All right-- switching gears. 00:39:51.285 --> 00:39:54.630 align:middle line:90% 00:39:54.630 --> 00:39:58.173 align:middle line:84% So the next set of things I wanted to talk about 00:39:58.173 --> 00:39:59.590 align:middle line:84% are labor markets and development. 00:39:59.590 --> 00:40:00.850 align:middle line:84% And this is actually another area 00:40:00.850 --> 00:40:01.930 align:middle line:90% where there's been a lot of-- 00:40:01.930 --> 00:40:03.805 align:middle line:84% I think a lot of movement in this literature. 00:40:03.805 --> 00:40:06.878 align:middle line:84% Even though the paper I assigned you to read is an older one, 00:40:06.878 --> 00:40:08.920 align:middle line:84% there's been a lot of movement in this literature 00:40:08.920 --> 00:40:12.470 align:middle line:90% over the past five years or so. 00:40:12.470 --> 00:40:14.170 align:middle line:84% And so you'll see that even though some 00:40:14.170 --> 00:40:16.850 align:middle line:84% of this I'm going to start with today is a bit older, 00:40:16.850 --> 00:40:17.620 align:middle line:84% some of the stuff I'm going to talk 00:40:17.620 --> 00:40:19.620 align:middle line:84% about over the next-- in the next lecture or two 00:40:19.620 --> 00:40:22.330 align:middle line:90% is pretty recent. 00:40:22.330 --> 00:40:29.640 align:middle line:90% So here's the overall plan. 00:40:29.640 --> 00:40:32.010 align:middle line:84% So a theme that we're going to look at is, 00:40:32.010 --> 00:40:34.620 align:middle line:84% how do we think about efficiency in labor markets 00:40:34.620 --> 00:40:36.060 align:middle line:90% in developing contexts? 00:40:36.060 --> 00:40:37.620 align:middle line:84% I'm going to start with motivation 00:40:37.620 --> 00:40:39.817 align:middle line:84% from this idea of the, quote, "surplus labor 00:40:39.817 --> 00:40:41.400 align:middle line:84% hypothesis," which is a very old idea, 00:40:41.400 --> 00:40:43.860 align:middle line:84% and how do we think about that relating to the separation 00:40:43.860 --> 00:40:45.638 align:middle line:90% paper. 00:40:45.638 --> 00:40:47.180 align:middle line:84% I'm going to talk about labor supply. 00:40:47.180 --> 00:40:49.550 align:middle line:84% We're going to talk about the separation test 00:40:49.550 --> 00:40:52.460 align:middle line:84% as a test for frictionless labor markets. 00:40:52.460 --> 00:40:56.385 align:middle line:84% Then next class we're going to talk about nominal rigidities, 00:40:56.385 --> 00:40:58.010 align:middle line:84% behavioral issues, and other challenges 00:40:58.010 --> 00:41:02.013 align:middle line:84% in labor supply and some frictions in labor demand. 00:41:02.013 --> 00:41:04.430 align:middle line:84% And then I'm going to talk a bit about urban labor markets 00:41:04.430 --> 00:41:06.722 align:middle line:84% and how do we think about some of those issues as well. 00:41:06.722 --> 00:41:08.410 align:middle line:84% So that's kind of the overall plan. 00:41:08.410 --> 00:41:11.110 align:middle line:90% 00:41:11.110 --> 00:41:12.590 align:middle line:90% All right. 00:41:12.590 --> 00:41:14.990 align:middle line:84% So going even further back than the reading 00:41:14.990 --> 00:41:18.320 align:middle line:84% that I gave you for this, in some sense the-- 00:41:18.320 --> 00:41:21.104 align:middle line:90% 00:41:21.104 --> 00:41:25.410 align:middle line:84% a classic reading in development is this paper 00:41:25.410 --> 00:41:28.310 align:middle line:90% by Arthur Lewis from 1954. 00:41:28.310 --> 00:41:34.660 align:middle line:84% And Arthur Lewis basically argued that there was, quote, 00:41:34.660 --> 00:41:38.370 align:middle line:84% "surplus labor" in the countryside, 00:41:38.370 --> 00:41:40.620 align:middle line:90% in developing countries. 00:41:40.620 --> 00:41:45.750 align:middle line:84% And, in particular, he said, roughly speaking, "about 25% 00:41:45.750 --> 00:41:49.990 align:middle line:84% of the labor has zero marginal value." 00:41:49.990 --> 00:41:54.563 align:middle line:84% There's way more people than we need for the tasks that 00:41:54.563 --> 00:41:55.480 align:middle line:90% are needed to be done. 00:41:55.480 --> 00:41:57.010 align:middle line:84% And they're just kind of like-- just 00:41:57.010 --> 00:41:59.975 align:middle line:84% really on the kind of the flat part of the production 00:41:59.975 --> 00:42:00.475 align:middle line:90% function. 00:42:00.475 --> 00:42:04.308 align:middle line:90% 00:42:04.308 --> 00:42:06.100 align:middle line:84% And his claim, therefore-- this is actually 00:42:06.100 --> 00:42:08.240 align:middle line:84% an important development claim-- was he said, 00:42:08.240 --> 00:42:12.820 align:middle line:84% look, we can increase aggregate productivity if we can just 00:42:12.820 --> 00:42:14.410 align:middle line:84% move some of that surplus labor that's 00:42:14.410 --> 00:42:18.730 align:middle line:84% not doing anything useful from the countryside to the cities. 00:42:18.730 --> 00:42:19.370 align:middle line:90% We can do that. 00:42:19.370 --> 00:42:21.412 align:middle line:84% And he said, we can do it without even decreasing 00:42:21.412 --> 00:42:26.600 align:middle line:84% agricultural output because they have a zero marginal product. 00:42:26.600 --> 00:42:27.780 align:middle line:90% OK. 00:42:27.780 --> 00:42:31.510 align:middle line:84% So that would mean that either the marginal product of labor 00:42:31.510 --> 00:42:35.140 align:middle line:84% is zero maybe because the agriculture production is 00:42:35.140 --> 00:42:35.800 align:middle line:90% Leontief. 00:42:35.800 --> 00:42:38.710 align:middle line:84% In general, we tend to think-- we don't tend to draw 00:42:38.710 --> 00:42:39.850 align:middle line:90% a production functions-- 00:42:39.850 --> 00:42:42.640 align:middle line:90% 00:42:42.640 --> 00:42:45.010 align:middle line:84% we tend to draw production functions as concave, 00:42:45.010 --> 00:42:47.530 align:middle line:90% not actually flat. 00:42:47.530 --> 00:42:53.380 align:middle line:90% But maybe it's flat. 00:42:53.380 --> 00:42:57.700 align:middle line:90% 00:42:57.700 --> 00:42:59.500 align:middle line:84% And why would it actually be flat? 00:42:59.500 --> 00:43:01.540 align:middle line:84% Because at some point it ends up being 00:43:01.540 --> 00:43:03.710 align:middle line:90% that you just need more land. 00:43:03.710 --> 00:43:05.620 align:middle line:84% So when I say it's Leontief, what do I mean? 00:43:05.620 --> 00:43:09.012 align:middle line:90% At some point you run out-- 00:43:09.012 --> 00:43:11.470 align:middle line:84% there's only so much we can do with a given amount of land. 00:43:11.470 --> 00:43:13.690 align:middle line:84% No matter how intensively we farm it or weed it 00:43:13.690 --> 00:43:16.840 align:middle line:84% or whatever, at some point you're running out of-- 00:43:16.840 --> 00:43:18.290 align:middle line:90% you run out of land. 00:43:18.290 --> 00:43:21.310 align:middle line:90% 00:43:21.310 --> 00:43:24.690 align:middle line:84% Or the other reason that this could be 00:43:24.690 --> 00:43:26.250 align:middle line:84% is actually maybe labor supply is 00:43:26.250 --> 00:43:28.410 align:middle line:84% totally elastic at some reservation wage rate. 00:43:28.410 --> 00:43:31.740 align:middle line:84% And at some point, maybe the marginal product 00:43:31.740 --> 00:43:34.600 align:middle line:84% is still kind of like positive over here. 00:43:34.600 --> 00:43:36.360 align:middle line:84% But people will just stop working. 00:43:36.360 --> 00:43:38.730 align:middle line:84% Because if people are being paid their marginal product, 00:43:38.730 --> 00:43:42.600 align:middle line:84% then at some point if there's some reservation wage rate-- 00:43:42.600 --> 00:43:45.787 align:middle line:84% if no one's going to work beyond-- 00:43:45.787 --> 00:43:48.840 align:middle line:90% 00:43:48.840 --> 00:43:50.580 align:middle line:84% say that the slope of this thing is 00:43:50.580 --> 00:43:52.890 align:middle line:84% equal to the marginal product is equal to the wage. 00:43:52.890 --> 00:43:56.670 align:middle line:84% If at this point-- if that wage no one's going to work any-- 00:43:56.670 --> 00:43:58.800 align:middle line:84% below that wage no one will work anymore, 00:43:58.800 --> 00:44:00.300 align:middle line:84% that's how much labor we would get. 00:44:00.300 --> 00:44:01.830 align:middle line:84% And all the people kind of beyond that 00:44:01.830 --> 00:44:02.913 align:middle line:90% would just be not working. 00:44:02.913 --> 00:44:05.580 align:middle line:90% 00:44:05.580 --> 00:44:07.760 align:middle line:90% OK? 00:44:07.760 --> 00:44:10.238 align:middle line:84% And so the rationale here was that there 00:44:10.238 --> 00:44:11.780 align:middle line:84% was an important goal for development 00:44:11.780 --> 00:44:14.720 align:middle line:84% was to bring this surplus labor kind of into the cities. 00:44:14.720 --> 00:44:19.240 align:middle line:90% 00:44:19.240 --> 00:44:23.170 align:middle line:84% This question of should we have more people in the cities 00:44:23.170 --> 00:44:26.935 align:middle line:84% and less in the countryside is a very broad question 00:44:26.935 --> 00:44:27.560 align:middle line:90% of development. 00:44:27.560 --> 00:44:28.090 align:middle line:90% I'm not going to talk-- 00:44:28.090 --> 00:44:30.257 align:middle line:84% I'm going to talk a little bit about that right now, 00:44:30.257 --> 00:44:32.865 align:middle line:84% but we're going to talk a lot more about that in 14772 when 00:44:32.865 --> 00:44:34.990 align:middle line:84% we talk-- where we're going to talk about migration 00:44:34.990 --> 00:44:36.940 align:middle line:84% and the structural transformation from 00:44:36.940 --> 00:44:40.083 align:middle line:90% the countryside to the-- 00:44:40.083 --> 00:44:41.500 align:middle line:84% and moving people from agriculture 00:44:41.500 --> 00:44:43.350 align:middle line:90% into the urban areas. 00:44:43.350 --> 00:44:45.100 align:middle line:84% But I just wanted to say it's kind of-- it 00:44:45.100 --> 00:44:47.740 align:middle line:84% is related to this idea that people are not 00:44:47.740 --> 00:44:50.620 align:middle line:90% doing very much proactive. 00:44:50.620 --> 00:44:51.120 align:middle line:90% OK. 00:44:51.120 --> 00:44:56.870 align:middle line:90% 00:44:56.870 --> 00:45:00.600 align:middle line:84% So in the spirit of some old papers, 00:45:00.600 --> 00:45:04.590 align:middle line:84% this is a paper-- there was a paper by Schultz in 1964 00:45:04.590 --> 00:45:07.280 align:middle line:84% which was one of the earliest natural experiment papers 00:45:07.280 --> 00:45:12.300 align:middle line:84% that I can find, which was about the surplus labor hypothesis. 00:45:12.300 --> 00:45:20.410 align:middle line:84% So Schultz studies the 1917, 1918 flu pandemic, 00:45:20.410 --> 00:45:23.610 align:middle line:84% which killed 6% of the population in India 00:45:23.610 --> 00:45:27.850 align:middle line:84% and reduced the workforce by about 8%. 00:45:27.850 --> 00:45:31.630 align:middle line:84% So you can imagine-- given that we were just 00:45:31.630 --> 00:45:36.073 align:middle line:84% going through a disease that has a 1% or 2% mortality rate, 00:45:36.073 --> 00:45:37.990 align:middle line:84% you can imagine what this must have been like. 00:45:37.990 --> 00:45:41.560 align:middle line:90% 00:45:41.560 --> 00:45:45.650 align:middle line:84% So Schultz says, well, look, if there was really-- 00:45:45.650 --> 00:45:50.000 align:middle line:84% if Lewis is right, and there's really 25% surplus labor, 00:45:50.000 --> 00:45:52.522 align:middle line:90% then we shouldn't have any-- 00:45:52.522 --> 00:45:54.605 align:middle line:84% this should have no impact on agricultural output. 00:45:54.605 --> 00:45:57.420 align:middle line:90% 00:45:57.420 --> 00:45:59.185 align:middle line:84% So what do you think of this as a test-- 00:45:59.185 --> 00:45:59.685 align:middle line:90% as an idea? 00:45:59.685 --> 00:46:09.920 align:middle line:90% 00:46:09.920 --> 00:46:12.150 align:middle line:84% This plausible, not-- is it a good test? 00:46:12.150 --> 00:46:12.650 align:middle line:90% Bad test? 00:46:12.650 --> 00:46:13.280 align:middle line:90% What you think? 00:46:13.280 --> 00:46:15.820 align:middle line:90% 00:46:15.820 --> 00:46:16.320 align:middle line:90% Yeah, Ahmed. 00:46:16.320 --> 00:46:18.850 align:middle line:90% 00:46:18.850 --> 00:46:22.420 align:middle line:84% AUDIENCE: Call it bias, but I kind of 00:46:22.420 --> 00:46:28.420 align:middle line:84% find it hard to believe a pandemic can have 00:46:28.420 --> 00:46:31.170 align:middle line:90% no impact on capital markets. 00:46:31.170 --> 00:46:32.852 align:middle line:84% BENJAMIN OLKEN: Oh, on capital markets. 00:46:32.852 --> 00:46:33.560 align:middle line:90% What do you mean? 00:46:33.560 --> 00:46:36.260 align:middle line:84% AUDIENCE: So the test here assumes that, OK, 00:46:36.260 --> 00:46:37.950 align:middle line:90% there's a change in labor. 00:46:37.950 --> 00:46:40.250 align:middle line:84% And if nothing else changes, this 00:46:40.250 --> 00:46:44.920 align:middle line:84% will be a test of that [INAUDIBLE] on the margin, 00:46:44.920 --> 00:46:48.180 align:middle line:84% whether labor is needed for these markets. 00:46:48.180 --> 00:46:50.720 align:middle line:84% But if they are getting their inputs from like, say, 00:46:50.720 --> 00:46:54.170 align:middle line:84% from the countryside, from other places, 00:46:54.170 --> 00:46:56.870 align:middle line:84% chances are this is like epidemic. 00:46:56.870 --> 00:47:00.050 align:middle line:84% Things were not going good from all places the other way 00:47:00.050 --> 00:47:04.400 align:middle line:84% around, as in the [INAUDIBLE] supply crisis now. 00:47:04.400 --> 00:47:08.750 align:middle line:84% So there could still be a reduction in production, 00:47:08.750 --> 00:47:09.840 align:middle line:90% even though-- 00:47:09.840 --> 00:47:11.340 align:middle line:84% even in the case where [INAUDIBLE].. 00:47:11.340 --> 00:47:11.760 align:middle line:90% BENJAMIN OLKEN: Yes, yes, yes. 00:47:11.760 --> 00:47:12.300 align:middle line:90% Absolutely. 00:47:12.300 --> 00:47:12.960 align:middle line:90% Yes, yes, yes. 00:47:12.960 --> 00:47:15.898 align:middle line:84% So let me just also say that I have taught 00:47:15.898 --> 00:47:17.940 align:middle line:84% this paper many times, and this is the first time 00:47:17.940 --> 00:47:21.640 align:middle line:84% someone has brought up supply chain disruptions in my-- 00:47:21.640 --> 00:47:22.140 align:middle line:90% in this. 00:47:22.140 --> 00:47:24.015 align:middle line:84% But I think we've all had a lot of experience 00:47:24.015 --> 00:47:25.530 align:middle line:84% as to the many different channels 00:47:25.530 --> 00:47:27.400 align:middle line:84% that pandemics can screw up the economy. 00:47:27.400 --> 00:47:27.900 align:middle line:90% So-- 00:47:27.900 --> 00:47:28.357 align:middle line:90% AUDIENCE: I'm just saying. 00:47:28.357 --> 00:47:29.730 align:middle line:90% I'm still waiting for my PS5. 00:47:29.730 --> 00:47:29.830 align:middle line:90% BENJAMIN OLKEN: What? 00:47:29.830 --> 00:47:30.537 align:middle line:90% Sorry? 00:47:30.537 --> 00:47:32.370 align:middle line:84% AUDIENCE: I'm just still waiting for my PS5. 00:47:32.370 --> 00:47:33.412 align:middle line:90% That's still [INAUDIBLE]. 00:47:33.412 --> 00:47:35.520 align:middle line:90% BENJAMIN OLKEN: [LAUGHS]. 00:47:35.520 --> 00:47:36.570 align:middle line:90% But yes, yes, yes. 00:47:36.570 --> 00:47:37.070 align:middle line:90% Good point. 00:47:37.070 --> 00:47:37.290 align:middle line:90% OK. 00:47:37.290 --> 00:47:37.860 align:middle line:90% What else? 00:47:37.860 --> 00:47:41.467 align:middle line:90% 00:47:41.467 --> 00:47:42.550 align:middle line:90% That's a new one for this. 00:47:42.550 --> 00:47:43.870 align:middle line:84% I've taught this many times-- never heard that. 00:47:43.870 --> 00:47:44.380 align:middle line:90% OK. 00:47:44.380 --> 00:47:45.952 align:middle line:84% What's another-- yeah, Christine? 00:47:45.952 --> 00:47:48.160 align:middle line:84% AUDIENCE: The people that were killed by the pandemic 00:47:48.160 --> 00:47:51.970 align:middle line:84% are not necessarily equal to the ones that didn't die. 00:47:51.970 --> 00:47:53.500 align:middle line:84% So it could have been like people 00:47:53.500 --> 00:47:57.340 align:middle line:84% that were more weak or less productive physically, 00:47:57.340 --> 00:47:58.790 align:middle line:90% or the ones that were killed. 00:47:58.790 --> 00:48:00.420 align:middle line:90% So I don't know. 00:48:00.420 --> 00:48:01.580 align:middle line:90% Or the other way around. 00:48:01.580 --> 00:48:02.890 align:middle line:90% It's not necessarily random. 00:48:02.890 --> 00:48:03.290 align:middle line:90% BENJAMIN OLKEN: Yeah. 00:48:03.290 --> 00:48:04.457 align:middle line:90% It's not necessarily random. 00:48:04.457 --> 00:48:07.060 align:middle line:84% Again, we're having so much better insight on how 00:48:07.060 --> 00:48:10.090 align:middle line:84% to-- on understanding pandemics as an identification strategy. 00:48:10.090 --> 00:48:11.360 align:middle line:90% Isn't it-- yes, exactly. 00:48:11.360 --> 00:48:11.860 align:middle line:90% Right? 00:48:11.860 --> 00:48:12.670 align:middle line:90% It's not random. 00:48:12.670 --> 00:48:13.270 align:middle line:90% Yeah. 00:48:13.270 --> 00:48:14.380 align:middle line:90% Totally right. 00:48:14.380 --> 00:48:16.660 align:middle line:84% And, in particular, it would be bad if it was-- 00:48:16.660 --> 00:48:19.630 align:middle line:90% 00:48:19.630 --> 00:48:23.540 align:middle line:84% he's actually going to reject this. 00:48:23.540 --> 00:48:25.690 align:middle line:84% And so if you thought the pandemic was 00:48:25.690 --> 00:48:29.380 align:middle line:84% killing people who were more productive than average, right? 00:48:29.380 --> 00:48:30.960 align:middle line:90% Yeah. 00:48:30.960 --> 00:48:33.183 align:middle line:90% Other thoughts? 00:48:33.183 --> 00:48:34.480 align:middle line:90% AUDIENCE: [INAUDIBLE] assume-- 00:48:34.480 --> 00:48:35.600 align:middle line:90% BENJAMIN OLKEN: What? 00:48:35.600 --> 00:48:36.940 align:middle line:90% AUDIENCE: You would assume-- 00:48:36.940 --> 00:48:39.010 align:middle line:84% I would have assumed [INAUDIBLE] to it 00:48:39.010 --> 00:48:41.620 align:middle line:84% for the people who died to go the other way. 00:48:41.620 --> 00:48:43.420 align:middle line:90% Like, it's the-- 00:48:43.420 --> 00:48:46.150 align:middle line:84% I mean, I would assume younger people would have a higher 00:48:46.150 --> 00:48:47.950 align:middle line:90% probability of survival. 00:48:47.950 --> 00:48:50.350 align:middle line:84% Chances are in a [INAUDIBLE] culture, [INAUDIBLE] 00:48:50.350 --> 00:48:53.412 align:middle line:84% it takes a lot of manual labor to do the tasks. 00:48:53.412 --> 00:48:55.870 align:middle line:84% I would assume people [? are ?] [INAUDIBLE] [? Australia ?] 00:48:55.870 --> 00:48:58.150 align:middle line:84% would be more productive than they just-- 00:48:58.150 --> 00:48:59.050 align:middle line:90% or [INAUDIBLE]. 00:48:59.050 --> 00:48:59.380 align:middle line:90% BENJAMIN OLKEN: Yes. 00:48:59.380 --> 00:49:00.940 align:middle line:84% That would go-- that will go-- if that was true, 00:49:00.940 --> 00:49:02.610 align:middle line:84% that would go against the direction of finding, 00:49:02.610 --> 00:49:03.190 align:middle line:90% in effect. 00:49:03.190 --> 00:49:04.040 align:middle line:90% Yeah. 00:49:04.040 --> 00:49:05.050 align:middle line:90% Yeah. 00:49:05.050 --> 00:49:06.540 align:middle line:90% What else? 00:49:06.540 --> 00:49:09.390 align:middle line:84% AUDIENCE: Maybe people that don't die but get sick 00:49:09.390 --> 00:49:11.737 align:middle line:84% have lower marginal product after. 00:49:11.737 --> 00:49:12.570 align:middle line:90% BENJAMIN OLKEN: Yep. 00:49:12.570 --> 00:49:12.870 align:middle line:90% Yep. 00:49:12.870 --> 00:49:14.160 align:middle line:84% So maybe there's an additional thing 00:49:14.160 --> 00:49:15.960 align:middle line:84% beyond the death of people who are like-- 00:49:15.960 --> 00:49:16.670 align:middle line:90% who are sick. 00:49:16.670 --> 00:49:17.170 align:middle line:90% OK. 00:49:17.170 --> 00:49:18.760 align:middle line:90% Any other ones? 00:49:18.760 --> 00:49:19.260 align:middle line:90% Yeah, Hazel. 00:49:19.260 --> 00:49:20.820 align:middle line:84% AUDIENCE: There's also like a demand shock, right? 00:49:20.820 --> 00:49:22.950 align:middle line:84% Because people who are dead are no longer demanding-- 00:49:22.950 --> 00:49:23.250 align:middle line:90% BENJAMIN OLKEN: What? 00:49:23.250 --> 00:49:23.580 align:middle line:90% Sorry? 00:49:23.580 --> 00:49:25.122 align:middle line:84% AUDIENCE: There's also a demand shock 00:49:25.122 --> 00:49:27.445 align:middle line:84% because the people who died no longer demand food. 00:49:27.445 --> 00:49:28.320 align:middle line:90% BENJAMIN OLKEN: Yeah. 00:49:28.320 --> 00:49:29.010 align:middle line:90% It's also a demand shock. 00:49:29.010 --> 00:49:29.510 align:middle line:90% Right. 00:49:29.510 --> 00:49:30.120 align:middle line:90% Totally agree. 00:49:30.120 --> 00:49:32.400 align:middle line:84% So-- and that would depend on whether or not thought 00:49:32.400 --> 00:49:33.575 align:middle line:90% the products-- 00:49:33.575 --> 00:49:35.200 align:middle line:84% whether it was world trade or whatever. 00:49:35.200 --> 00:49:37.075 align:middle line:84% Although, I guess there's probably a global-- 00:49:37.075 --> 00:49:37.938 align:middle line:90% probably demand. 00:49:37.938 --> 00:49:40.230 align:middle line:84% This is a global pandemic, so there was a global demand 00:49:40.230 --> 00:49:41.410 align:middle line:90% shock for food. 00:49:41.410 --> 00:49:41.940 align:middle line:90% So yeah. 00:49:41.940 --> 00:49:42.440 align:middle line:90% OK. 00:49:42.440 --> 00:49:43.770 align:middle line:90% So-- all right. 00:49:43.770 --> 00:49:46.000 align:middle line:90% So you may not love this. 00:49:46.000 --> 00:49:47.670 align:middle line:84% And the other thing I would also-- 00:49:47.670 --> 00:49:49.843 align:middle line:90% so I would also-- 00:49:49.843 --> 00:49:51.510 align:middle line:84% the other thing I would also think about 00:49:51.510 --> 00:49:54.558 align:middle line:84% is the land reallocation process. 00:49:54.558 --> 00:49:56.100 align:middle line:84% Would it may have been particularly-- 00:49:56.100 --> 00:49:58.110 align:middle line:84% that's the one I've always most worried about, 00:49:58.110 --> 00:50:01.560 align:middle line:84% is that basically if someone died-- 00:50:01.560 --> 00:50:04.960 align:middle line:84% land markets are not super efficient and fast. 00:50:04.960 --> 00:50:10.560 align:middle line:84% And so if I own my farm, and I'm a victim of the flu pandemic, 00:50:10.560 --> 00:50:13.688 align:middle line:84% then it's not clear who's going to take over my farm. 00:50:13.688 --> 00:50:16.230 align:middle line:84% And that process of reallocating that land to more productive 00:50:16.230 --> 00:50:17.670 align:middle line:90% uses may take a while too. 00:50:17.670 --> 00:50:21.600 align:middle line:84% So that's all good reasons why this is not 00:50:21.600 --> 00:50:23.310 align:middle line:90% the last word on this subject. 00:50:23.310 --> 00:50:24.100 align:middle line:90% What did he do? 00:50:24.100 --> 00:50:26.648 align:middle line:84% He looks at the output and the post-year 00:50:26.648 --> 00:50:27.690 align:middle line:90% relevant to the pre-year. 00:50:27.690 --> 00:50:29.100 align:middle line:90% We had similar weather. 00:50:29.100 --> 00:50:31.708 align:middle line:84% And he looks at provinces that had greater influenza deaths, 00:50:31.708 --> 00:50:33.000 align:middle line:90% had greater declines in output. 00:50:33.000 --> 00:50:35.430 align:middle line:90% That's his test. 00:50:35.430 --> 00:50:38.202 align:middle line:84% And he looks at actually eight total acres sown. 00:50:38.202 --> 00:50:39.910 align:middle line:84% Actually, there's enough data and output. 00:50:39.910 --> 00:50:41.730 align:middle line:84% So they-- are they like plotting less-- 00:50:41.730 --> 00:50:42.750 align:middle line:90% doing less stuff. 00:50:42.750 --> 00:50:47.010 align:middle line:84% And here's his-- he basically has the measure of deaths 00:50:47.010 --> 00:50:49.843 align:middle line:84% and has predicted reduction in agricultural output 00:50:49.843 --> 00:50:51.760 align:middle line:84% and observed reduction in agricultural output. 00:50:51.760 --> 00:50:53.460 align:middle line:84% So he's running a regression of this 00:50:53.460 --> 00:50:56.950 align:middle line:90% against this or against that. 00:50:56.950 --> 00:51:02.180 align:middle line:84% And I actually had to-- ran the regression on my computer. 00:51:02.180 --> 00:51:04.170 align:middle line:84% And you basically get-- with only 10 states, 00:51:04.170 --> 00:51:06.045 align:middle line:84% he gets the elasticity of output with respect 00:51:06.045 --> 00:51:08.000 align:middle line:90% to population of about 0.4. 00:51:08.000 --> 00:51:10.820 align:middle line:84% So he is actually finding a pretty substantial reduction 00:51:10.820 --> 00:51:13.112 align:middle line:90% in output from the flu pandemic. 00:51:13.112 --> 00:51:14.570 align:middle line:84% Although, as we've discussed, there 00:51:14.570 --> 00:51:18.390 align:middle line:84% are other reasons why this could be an issue as well. 00:51:18.390 --> 00:51:18.890 align:middle line:90% OK. 00:51:18.890 --> 00:51:21.460 align:middle line:90% 00:51:21.460 --> 00:51:23.160 align:middle line:90% That's all by way of background. 00:51:23.160 --> 00:51:25.800 align:middle line:90% 00:51:25.800 --> 00:51:26.670 align:middle line:90% But I guess-- 00:51:26.670 --> 00:51:29.670 align:middle line:84% I do-- I want to mention this question of are labor markets 00:51:29.670 --> 00:51:32.802 align:middle line:84% efficient, and are people operating-- 00:51:32.802 --> 00:51:35.010 align:middle line:84% where are people relative to the production function. 00:51:35.010 --> 00:51:37.530 align:middle line:84% And are maybe they way past the point which 00:51:37.530 --> 00:51:39.990 align:middle line:84% they have low marginal products is an issue 00:51:39.990 --> 00:51:43.950 align:middle line:84% that people have been thinking about for a very long time. 00:51:43.950 --> 00:51:44.700 align:middle line:90% OK. 00:51:44.700 --> 00:51:46.680 align:middle line:84% I think it's-- sometimes I feel like we only teach really 00:51:46.680 --> 00:51:47.263 align:middle line:90% modern papers. 00:51:47.263 --> 00:51:48.960 align:middle line:84% And it's sometimes nice to realize 00:51:48.960 --> 00:51:51.760 align:middle line:84% people have been thinking about these issues for a while. 00:51:51.760 --> 00:51:52.315 align:middle line:90% OK. 00:51:52.315 --> 00:51:54.940 align:middle line:84% So now onto this Benjamin paper, which you guys read for today. 00:51:54.940 --> 00:51:57.490 align:middle line:84% And some of you asked why did I choose this particular paper 00:51:57.490 --> 00:51:58.810 align:middle line:90% to read? 00:51:58.810 --> 00:52:02.800 align:middle line:84% It was not-- as you picked up, it was not for the empirics. 00:52:02.800 --> 00:52:04.300 align:middle line:90% But I think that the-- 00:52:04.300 --> 00:52:06.160 align:middle line:84% and actually, I'm going to show you a more-- 00:52:06.160 --> 00:52:07.940 align:middle line:90% up until all of this-- 00:52:07.940 --> 00:52:09.940 align:middle line:84% I'm going to show you a more modern paper that 00:52:09.940 --> 00:52:12.040 align:middle line:84% is a better empirical test, actually, of exactly 00:52:12.040 --> 00:52:13.658 align:middle line:90% these issues in a sec. 00:52:13.658 --> 00:52:15.700 align:middle line:84% But I didn't want you to have to read two papers. 00:52:15.700 --> 00:52:17.158 align:middle line:84% But what I like about this paper is 00:52:17.158 --> 00:52:19.300 align:middle line:84% I feel like it really cleanly illustrates 00:52:19.300 --> 00:52:22.728 align:middle line:84% this idea of separation failures in going 00:52:22.728 --> 00:52:24.770 align:middle line:84% through the various cases and the various models, 00:52:24.770 --> 00:52:28.420 align:middle line:84% so I want to talk about that a little bit here. 00:52:28.420 --> 00:52:35.560 align:middle line:84% And I think this idea of separation between-- 00:52:35.560 --> 00:52:38.880 align:middle line:84% why is it an important development model? 00:52:38.880 --> 00:52:40.620 align:middle line:84% I think that one thing that you see 00:52:40.620 --> 00:52:43.350 align:middle line:84% in a lot of developing contexts is you 00:52:43.350 --> 00:52:46.200 align:middle line:90% see a lot more self-employment. 00:52:46.200 --> 00:52:50.000 align:middle line:84% That basically, as we talked about, actually, 00:52:50.000 --> 00:52:51.900 align:middle line:84% in the last lecture the employee share 00:52:51.900 --> 00:52:55.770 align:middle line:84% is a lot lower in a lot of developing country comments-- 00:52:55.770 --> 00:52:56.320 align:middle line:90% context. 00:52:56.320 --> 00:53:00.270 align:middle line:84% So here, in developed country most people are-- 00:53:00.270 --> 00:53:03.065 align:middle line:84% not everyone-- most people are employees, and-- 00:53:03.065 --> 00:53:04.440 align:middle line:84% whereas, in developing countries, 00:53:04.440 --> 00:53:09.210 align:middle line:84% I think most people are self-employed in some context, 00:53:09.210 --> 00:53:11.790 align:middle line:84% or working in their-- or production. 00:53:11.790 --> 00:53:15.670 align:middle line:84% And that means we want to think about 00:53:15.670 --> 00:53:18.200 align:middle line:84% both for understanding their own household decision-making, 00:53:18.200 --> 00:53:19.867 align:middle line:84% for understanding consumption decisions, 00:53:19.867 --> 00:53:22.640 align:middle line:84% but also for understanding production. 00:53:22.640 --> 00:53:27.380 align:middle line:84% If we have separation failures, then understanding 00:53:27.380 --> 00:53:28.970 align:middle line:84% what's going on on the household side 00:53:28.970 --> 00:53:31.100 align:middle line:84% is going to be important for the productive side of the economy, 00:53:31.100 --> 00:53:31.310 align:middle line:90% too. 00:53:31.310 --> 00:53:33.110 align:middle line:84% And that's going to be really important in a developing 00:53:33.110 --> 00:53:35.840 align:middle line:84% country context if most of the people are kind of doing both. 00:53:35.840 --> 00:53:38.450 align:middle line:84% Whereas, if we've separated things out, 00:53:38.450 --> 00:53:40.707 align:middle line:84% and we are in much larger context 00:53:40.707 --> 00:53:42.290 align:middle line:84% where people are employees, that issue 00:53:42.290 --> 00:53:44.878 align:middle line:90% is kind of less relevant for-- 00:53:44.878 --> 00:53:46.170 align:middle line:90% in a developed country context. 00:53:46.170 --> 00:53:49.910 align:middle line:84% So I think of this as like one of these important concepts 00:53:49.910 --> 00:53:52.070 align:middle line:84% that comes up a lot because it describes kind a lot 00:53:52.070 --> 00:53:55.035 align:middle line:84% of what people are doing in developing country contexts. 00:53:55.035 --> 00:53:57.160 align:middle line:84% So the question he's trying to answer at some level 00:53:57.160 --> 00:53:59.740 align:middle line:84% is how efficient are rural labor markets? 00:53:59.740 --> 00:54:02.740 align:middle line:84% And, in particular, he has this test 00:54:02.740 --> 00:54:05.860 align:middle line:90% of the separation hypothesis. 00:54:05.860 --> 00:54:08.530 align:middle line:84% And broadly speaking, the way to think 00:54:08.530 --> 00:54:10.460 align:middle line:84% about the test of separation hypothesis 00:54:10.460 --> 00:54:12.307 align:middle line:84% is like is my production decision 00:54:12.307 --> 00:54:14.890 align:middle line:84% and is my consumption decision-- are those separate decisions, 00:54:14.890 --> 00:54:18.730 align:middle line:90% or are they conjoined decisions? 00:54:18.730 --> 00:54:24.190 align:middle line:84% And the basic theoretical idea is 00:54:24.190 --> 00:54:26.710 align:middle line:84% if I have fully functioning efficient markets, 00:54:26.710 --> 00:54:31.990 align:middle line:84% then households can freely buy or sell their labor at wage w. 00:54:31.990 --> 00:54:35.760 align:middle line:84% And therefore, they're going to make two separate choices. 00:54:35.760 --> 00:54:39.090 align:middle line:84% They're first going to choose the labor input for their farms 00:54:39.090 --> 00:54:43.098 align:middle line:84% to maximize profits, given the prevailing wage w. 00:54:43.098 --> 00:54:44.890 align:middle line:84% And then they're going to separately choose 00:54:44.890 --> 00:54:47.265 align:middle line:84% the optimal labor leisure trade-off for the family, given 00:54:47.265 --> 00:54:49.340 align:middle line:90% wage w. 00:54:49.340 --> 00:54:53.030 align:middle line:84% And with full ability to buy and sell w, 00:54:53.030 --> 00:54:55.970 align:middle line:84% there is no reason those decisions should be linked. 00:54:55.970 --> 00:54:59.478 align:middle line:90% So here's the key graph. 00:54:59.478 --> 00:55:00.520 align:middle line:90% So this is the key graph. 00:55:00.520 --> 00:55:04.242 align:middle line:84% This basically says, look, here's a production. 00:55:04.242 --> 00:55:05.200 align:middle line:90% This is the production. 00:55:05.200 --> 00:55:09.190 align:middle line:84% And by the way, I found these graphs a little confusing. 00:55:09.190 --> 00:55:10.690 align:middle line:84% I don't if you-- did you find them-- 00:55:10.690 --> 00:55:14.390 align:middle line:90% were they clear or confusing? 00:55:14.390 --> 00:55:14.890 align:middle line:90% [INAUDIBLE] 00:55:14.890 --> 00:55:15.380 align:middle line:90% OK. 00:55:15.380 --> 00:55:17.230 align:middle line:84% I'm going to try to go-- let me say I've taught them 00:55:17.230 --> 00:55:18.700 align:middle line:84% a couple-- it's took me a couple of times 00:55:18.700 --> 00:55:20.390 align:middle line:84% to teach them before they became less confusing to me. 00:55:20.390 --> 00:55:22.930 align:middle line:84% So I'm going to try-- let me see if I can elucidate them. 00:55:22.930 --> 00:55:24.972 align:middle line:90% So what is going on? 00:55:24.972 --> 00:55:26.305 align:middle line:90% This is the production function. 00:55:26.305 --> 00:55:29.330 align:middle line:90% 00:55:29.330 --> 00:55:35.170 align:middle line:84% So how do you choose how much labor to use? 00:55:35.170 --> 00:55:39.088 align:middle line:84% We're going to look for the tangency of-- 00:55:39.088 --> 00:55:40.630 align:middle line:84% you're going to go to the point where 00:55:40.630 --> 00:55:44.090 align:middle line:84% the marginal product of labor is equal to the prevailing wage. 00:55:44.090 --> 00:55:46.240 align:middle line:84% So that's this-- that's given by the slope w. 00:55:46.240 --> 00:55:47.890 align:middle line:90% That's the prevailing wage. 00:55:47.890 --> 00:55:50.110 align:middle line:84% The amount of labor you're going to want on your farm 00:55:50.110 --> 00:55:55.030 align:middle line:84% is given by the tangency of a line with slope w, the wage 00:55:55.030 --> 00:55:58.460 align:middle line:84% rate, to the production function of the farm. 00:55:58.460 --> 00:55:58.960 align:middle line:90% OK. 00:55:58.960 --> 00:56:01.510 align:middle line:90% That's this. 00:56:01.510 --> 00:56:05.060 align:middle line:84% So that's the labor used on the farm. 00:56:05.060 --> 00:56:08.935 align:middle line:84% And then how much labor am I going to supply as a household? 00:56:08.935 --> 00:56:10.310 align:middle line:84% Well, I'm going to look for the-- 00:56:10.310 --> 00:56:13.040 align:middle line:84% I'm going to have some utility function. 00:56:13.040 --> 00:56:18.110 align:middle line:84% And I'm going to look for the tangency between my utility 00:56:18.110 --> 00:56:20.390 align:middle line:90% function and the wage rate. 00:56:20.390 --> 00:56:20.890 align:middle line:90% All right. 00:56:20.890 --> 00:56:22.540 align:middle line:84% So I have some labor leisure trade-off. 00:56:22.540 --> 00:56:25.690 align:middle line:84% This is the utility function over here. 00:56:25.690 --> 00:56:29.830 align:middle line:90% The bliss point is over here. 00:56:29.830 --> 00:56:32.890 align:middle line:84% High consumption, low labor-- that's the bliss point. 00:56:32.890 --> 00:56:35.380 align:middle line:84% So the utility function is facing this way. 00:56:35.380 --> 00:56:38.915 align:middle line:84% And I'm going to go to the point where it's tangent to the wage 00:56:38.915 --> 00:56:41.290 align:middle line:84% rate, and that's going to decide how much labor I'm going 00:56:41.290 --> 00:56:41.800 align:middle line:90% to choose. 00:56:41.800 --> 00:56:46.550 align:middle line:90% 00:56:46.550 --> 00:56:53.260 align:middle line:84% So what you can see here is these are set-- 00:56:53.260 --> 00:56:56.410 align:middle line:90% these are separate decisions. 00:56:56.410 --> 00:56:58.807 align:middle line:84% This is choosing-- this is going to be driven-- 00:56:58.807 --> 00:57:00.640 align:middle line:84% different people with different preferences. 00:57:00.640 --> 00:57:02.967 align:middle line:84% They'll move their labor/leisure trade-off. 00:57:02.967 --> 00:57:04.300 align:middle line:90% Some people really like leisure. 00:57:04.300 --> 00:57:05.650 align:middle line:90% They're going to be over here. 00:57:05.650 --> 00:57:06.490 align:middle line:84% Some people really like labor-- they're 00:57:06.490 --> 00:57:07.270 align:middle line:90% going to, like, over here. 00:57:07.270 --> 00:57:08.440 align:middle line:84% Like consumption-- they're going to be over here. 00:57:08.440 --> 00:57:10.857 align:middle line:84% Whatever-- people will choose different amounts over here. 00:57:10.857 --> 00:57:12.100 align:middle line:90% And then the farms-- 00:57:12.100 --> 00:57:14.590 align:middle line:84% different farms might have-- be steeper or flatter 00:57:14.590 --> 00:57:18.430 align:middle line:84% with different marginal products or different amounts of labor 00:57:18.430 --> 00:57:20.650 align:middle line:84% to sort of-- different marginal products for a given 00:57:20.650 --> 00:57:21.250 align:middle line:90% amount of labor. 00:57:21.250 --> 00:57:23.208 align:middle line:84% And therefore, different amounts optimal labor. 00:57:23.208 --> 00:57:25.580 align:middle line:84% And that's going to be a separate decision over here. 00:57:25.580 --> 00:57:28.880 align:middle line:84% So the baseline model has two separate decisions. 00:57:28.880 --> 00:57:29.380 align:middle line:90% Sorry, Chan. 00:57:29.380 --> 00:57:29.920 align:middle line:90% Do you have a question? 00:57:29.920 --> 00:57:30.190 align:middle line:90% Yeah. 00:57:30.190 --> 00:57:30.815 align:middle line:90% AUDIENCE: Yeah. 00:57:30.815 --> 00:57:32.447 align:middle line:84% AUDIENCE: Those wages are [INAUDIBLE].. 00:57:32.447 --> 00:57:33.280 align:middle line:90% BENJAMIN OLKEN: Yes. 00:57:33.280 --> 00:57:34.572 align:middle line:90% AUDIENCE: So this is not like-- 00:57:34.572 --> 00:57:37.510 align:middle line:84% I was just wondering why the utility curve is not 00:57:37.510 --> 00:57:41.135 align:middle line:84% tangent to the production curve the same as [INAUDIBLE].. 00:57:41.135 --> 00:57:42.260 align:middle line:90% BENJAMIN OLKEN: Wait, what? 00:57:42.260 --> 00:57:43.843 align:middle line:84% AUDIENCE: Like, I was wondering like-- 00:57:43.843 --> 00:57:44.718 align:middle line:90% BENJAMIN OLKEN: Yeah. 00:57:44.718 --> 00:57:45.720 align:middle line:90% This is market wage. 00:57:45.720 --> 00:57:47.625 align:middle line:90% AUDIENCE: Which is [INAUDIBLE]. 00:57:47.625 --> 00:57:48.500 align:middle line:90% BENJAMIN OLKEN: Yeah. 00:57:48.500 --> 00:57:49.430 align:middle line:90% So exactly. 00:57:49.430 --> 00:57:51.860 align:middle line:84% So-- and I think the idea-- the way to think about it is 00:57:51.860 --> 00:57:53.210 align:middle line:84% there's lots of different people with lots 00:57:53.210 --> 00:57:54.170 align:middle line:84% of different [? use ?] around there. 00:57:54.170 --> 00:57:55.430 align:middle line:84% And some of them are down here, and some of them 00:57:55.430 --> 00:57:56.280 align:middle line:90% are over here or whatever. 00:57:56.280 --> 00:57:58.280 align:middle line:84% And there's some equilibrium in the labor market 00:57:58.280 --> 00:57:59.300 align:middle line:90% that's determining w. 00:57:59.300 --> 00:57:59.800 align:middle line:90% Yeah. 00:57:59.800 --> 00:58:02.130 align:middle line:90% 00:58:02.130 --> 00:58:04.745 align:middle line:90% OK, Other questions? 00:58:04.745 --> 00:58:06.120 align:middle line:84% AUDIENCE: [INAUDIBLE] the labor-- 00:58:06.120 --> 00:58:09.698 align:middle line:84% [INAUDIBLE] labor demand not being called labor supply? 00:58:09.698 --> 00:58:10.490 align:middle line:90% BENJAMIN OLKEN: No. 00:58:10.490 --> 00:58:12.730 align:middle line:84% That was related to Vashan's question. 00:58:12.730 --> 00:58:14.590 align:middle line:90% No. 00:58:14.590 --> 00:58:17.340 align:middle line:84% For an individual household it does not. 00:58:17.340 --> 00:58:19.680 align:middle line:90% In the market, sure. 00:58:19.680 --> 00:58:22.595 align:middle line:84% But this is an individual household's decision. 00:58:22.595 --> 00:58:24.470 align:middle line:84% AUDIENCE: If households are identical, then-- 00:58:24.470 --> 00:58:24.820 align:middle line:90% BENJAMIN OLKEN: Sorry. 00:58:24.820 --> 00:58:25.095 align:middle line:90% What? 00:58:25.095 --> 00:58:25.370 align:middle line:90% Sorry. 00:58:25.370 --> 00:58:27.410 align:middle line:84% AUDIENCE: If households are identical, then it should. 00:58:27.410 --> 00:58:27.920 align:middle line:90% BENJAMIN OLKEN: Yes. 00:58:27.920 --> 00:58:29.960 align:middle line:84% If households are all identical, then it should, yes. 00:58:29.960 --> 00:58:32.252 align:middle line:84% But if there's heterogeneity, then there's no reason it 00:58:32.252 --> 00:58:34.310 align:middle line:84% should in the-- for any given-- like some-- 00:58:34.310 --> 00:58:34.700 align:middle line:90% that's what I'm saying. 00:58:34.700 --> 00:58:36.200 align:middle line:84% Some people really like consumption. 00:58:36.200 --> 00:58:37.590 align:middle line:90% Some people really like leisure. 00:58:37.590 --> 00:58:39.860 align:middle line:84% Some people have big farms, and some have small farms. 00:58:39.860 --> 00:58:41.270 align:middle line:90% It's all going to kind of-- 00:58:41.270 --> 00:58:42.980 align:middle line:84% for any-- there's no reason those 00:58:42.980 --> 00:58:46.270 align:middle line:84% should be the same in any particular household. 00:58:46.270 --> 00:58:48.880 align:middle line:84% AUDIENCE: But a conceptual framework is like I-- 00:58:48.880 --> 00:58:51.005 align:middle line:84% so then the production is my own production, right? 00:58:51.005 --> 00:58:51.880 align:middle line:90% BENJAMIN OLKEN: Yeah. 00:58:51.880 --> 00:58:53.567 align:middle line:84% This is my-- this is like my farm. 00:58:53.567 --> 00:58:55.150 align:middle line:84% AUDIENCE: I can work in my farm, and I 00:58:55.150 --> 00:58:57.010 align:middle line:84% can-- say, if I have three hours-- 00:58:57.010 --> 00:58:58.472 align:middle line:84% if I want to work three more hours, 00:58:58.472 --> 00:59:00.430 align:middle line:84% I can go work for three hours and [INAUDIBLE].. 00:59:00.430 --> 00:59:01.430 align:middle line:90% BENJAMIN OLKEN: Correct. 00:59:01.430 --> 00:59:02.560 align:middle line:90% That is the model. 00:59:02.560 --> 00:59:03.700 align:middle line:90% AUDIENCE: All right then. 00:59:03.700 --> 00:59:04.575 align:middle line:90% BENJAMIN OLKEN: Yeah. 00:59:04.575 --> 00:59:06.110 align:middle line:90% Exactly. 00:59:06.110 --> 00:59:08.440 align:middle line:90% OK? 00:59:08.440 --> 00:59:09.850 align:middle line:90% All right. 00:59:09.850 --> 00:59:20.400 align:middle line:84% So now, what happens if there's rationing 00:59:20.400 --> 00:59:24.700 align:middle line:84% in the amount of off-farm work I can get? 00:59:24.700 --> 00:59:26.457 align:middle line:84% So suppose that, for example, like-- 00:59:26.457 --> 00:59:28.540 align:middle line:84% so, for example, suppose there was a minimum wage. 00:59:28.540 --> 00:59:31.410 align:middle line:90% 00:59:31.410 --> 00:59:34.200 align:middle line:84% So the minimum wage says that basically people 00:59:34.200 --> 00:59:39.430 align:middle line:84% don't want to hire that much labor up above some wage, 00:59:39.430 --> 00:59:42.430 align:middle line:90% so it's rationed. 00:59:42.430 --> 00:59:45.130 align:middle line:90% So now what's going to happen? 00:59:45.130 --> 00:59:47.780 align:middle line:90% Or what might happen? 00:59:47.780 --> 00:59:55.170 align:middle line:84% So at this high wage, suppose it's rationed to H. 00:59:55.170 --> 00:59:58.170 align:middle line:84% So the farmer-- what is the farmer going to do? 00:59:58.170 --> 01:00:02.610 align:middle line:84% They're going to first work at the outside high wage 01:00:02.610 --> 01:00:04.950 align:middle line:84% as much as they can up to this cap 01:00:04.950 --> 01:00:07.560 align:middle line:84% H. That's why this graph for the farmer 01:00:07.560 --> 01:00:10.930 align:middle line:84% is shifted to the left of the farm. 01:00:10.930 --> 01:00:12.980 align:middle line:90% They're going to work at H. 01:00:12.980 --> 01:00:16.010 align:middle line:84% And now after that, then they're going to say, well, I still-- 01:00:16.010 --> 01:00:18.453 align:middle line:90% I'm not done consuming. 01:00:18.453 --> 01:00:20.120 align:middle line:84% Suppose they're not done consuming at H. 01:00:20.120 --> 01:00:23.330 align:middle line:84% They would like to work more hours than that. 01:00:23.330 --> 01:00:27.690 align:middle line:84% Then they're going to start working up their own farm. 01:00:27.690 --> 01:00:28.590 align:middle line:90% OK? 01:00:28.590 --> 01:00:30.040 align:middle line:84% And how much are they going to do? 01:00:30.040 --> 01:00:32.190 align:middle line:84% Well, now they're just going to look to the point 01:00:32.190 --> 01:00:34.880 align:middle line:90% where the tangency of their-- 01:00:34.880 --> 01:00:36.630 align:middle line:84% they're going to-- how much are they going 01:00:36.630 --> 01:00:38.058 align:middle line:90% to work on their own farm? 01:00:38.058 --> 01:00:39.600 align:middle line:84% They're going to work up to the point 01:00:39.600 --> 01:00:43.020 align:middle line:84% where their utility function is tangent to the production 01:00:43.020 --> 01:00:46.920 align:middle line:90% function from the farm. 01:00:46.920 --> 01:00:48.810 align:middle line:84% Because the more they work, as they work-- 01:00:48.810 --> 01:00:50.610 align:middle line:84% right over here, they're earning a lot, 01:00:50.610 --> 01:00:52.470 align:middle line:84% in some sense, because they're on the steep part. 01:00:52.470 --> 01:00:53.730 align:middle line:84% As they work, work, work, work, work, 01:00:53.730 --> 01:00:55.605 align:middle line:84% they'll start earning less and less and less. 01:00:55.605 --> 01:01:00.010 align:middle line:84% And they'll work to the point where they're indifferent. 01:01:00.010 --> 01:01:02.500 align:middle line:90% Or sorry-- where it's tangent. 01:01:02.500 --> 01:01:08.278 align:middle line:84% And the tangency here is going to be at a-- 01:01:08.278 --> 01:01:16.590 align:middle line:84% at this high-- at this higher point here. 01:01:16.590 --> 01:01:18.200 align:middle line:90% And so-- yeah. 01:01:18.200 --> 01:01:22.730 align:middle line:84% So in this case, this is a case where they want more hours 01:01:22.730 --> 01:01:25.750 align:middle line:90% than they can get. 01:01:25.750 --> 01:01:34.340 align:middle line:84% And what's critical here is that now the amount of work that's 01:01:34.340 --> 01:01:36.402 align:middle line:90% going to happen on the farm-- 01:01:36.402 --> 01:01:38.360 align:middle line:84% the amount of labor that's going to be employed 01:01:38.360 --> 01:01:40.527 align:middle line:84% on this farm with this production function over here 01:01:40.527 --> 01:01:44.360 align:middle line:84% is going to be driven not by the market wage, 01:01:44.360 --> 01:01:46.753 align:middle line:84% but by how much labor this particular household 01:01:46.753 --> 01:01:47.420 align:middle line:90% wants to supply. 01:01:47.420 --> 01:01:50.140 align:middle line:90% 01:01:50.140 --> 01:01:50.800 align:middle line:90% Yeah? 01:01:50.800 --> 01:01:54.850 align:middle line:84% AUDIENCE: If the slope and the utility production dependency 01:01:54.850 --> 01:01:56.315 align:middle line:84% was higher than the off-farm work, 01:01:56.315 --> 01:01:57.940 align:middle line:84% they wouldn't work off the farm at all. 01:01:57.940 --> 01:01:58.330 align:middle line:90% Is that right? 01:01:58.330 --> 01:01:59.413 align:middle line:90% BENJAMIN OLKEN: I'm sorry. 01:01:59.413 --> 01:02:00.310 align:middle line:90% Say it again. 01:02:00.310 --> 01:02:03.040 align:middle line:84% AUDIENCE: If the slope of that dotted line 01:02:03.040 --> 01:02:05.995 align:middle line:84% was steeper than the slope of the off-farm wage, 01:02:05.995 --> 01:02:07.870 align:middle line:84% then would they not work off the farm at all? 01:02:07.870 --> 01:02:10.420 align:middle line:90% 01:02:10.420 --> 01:02:12.470 align:middle line:84% BENJAMIN OLKEN: If it was down here? 01:02:12.470 --> 01:02:12.970 align:middle line:90% Yeah. 01:02:12.970 --> 01:02:14.110 align:middle line:84% That's going to be coming in the next case. 01:02:14.110 --> 01:02:14.860 align:middle line:90% AUDIENCE: Oh, OK. 01:02:14.860 --> 01:02:15.360 align:middle line:90% Sorry. 01:02:15.360 --> 01:02:17.060 align:middle line:90% BENJAMIN OLKEN: Yeah. 01:02:17.060 --> 01:02:24.120 align:middle line:84% If they wanted to work less than that, then they would-- 01:02:24.120 --> 01:02:25.690 align:middle line:84% then we'd be in a separation case. 01:02:25.690 --> 01:02:27.315 align:middle line:84% What they would do is they would work-- 01:02:27.315 --> 01:02:31.890 align:middle line:90% 01:02:31.890 --> 01:02:33.670 align:middle line:84% I think what would happen is on this farm, 01:02:33.670 --> 01:02:36.100 align:middle line:84% actually, what they would do is they would work, say, 01:02:36.100 --> 01:02:37.960 align:middle line:84% this amount down here, and then they'd hire some extra labor 01:02:37.960 --> 01:02:38.530 align:middle line:90% on the farm. 01:02:38.530 --> 01:02:40.540 align:middle line:84% So it would be a separation case because they 01:02:40.540 --> 01:02:44.400 align:middle line:90% would be choosing the-- 01:02:44.400 --> 01:02:45.900 align:middle line:84% the amount of production on the farm 01:02:45.900 --> 01:02:49.400 align:middle line:84% would be driven by the market wage. 01:02:49.400 --> 01:02:50.000 align:middle line:90% So yeah. 01:02:50.000 --> 01:02:51.200 align:middle line:84% So it was only kind of an interesting case 01:02:51.200 --> 01:02:52.910 align:middle line:84% if the farm wants more labor-- if they want 01:02:52.910 --> 01:02:54.202 align:middle line:90% to supply more labor than that. 01:02:54.202 --> 01:02:57.986 align:middle line:90% 01:02:57.986 --> 01:03:00.860 align:middle line:90% OK. 01:03:00.860 --> 01:03:06.140 align:middle line:84% The next case is kind of the opposite, which is to say, 01:03:06.140 --> 01:03:09.560 align:middle line:84% suppose there's some rationing on the amount of labor 01:03:09.560 --> 01:03:13.280 align:middle line:84% I can hire because the market wage is too low. 01:03:13.280 --> 01:03:15.650 align:middle line:84% Suppose the market wage doesn't go above-- 01:03:15.650 --> 01:03:17.210 align:middle line:84% it would clear at some higher level, 01:03:17.210 --> 01:03:18.355 align:middle line:90% but it's not going above. 01:03:18.355 --> 01:03:20.730 align:middle line:84% And, by the way, I'm going to talk-- in the next lecture, 01:03:20.730 --> 01:03:23.105 align:middle line:84% we're going to talk about some evidence that these market 01:03:23.105 --> 01:03:23.888 align:middle line:90% wages do or not-- 01:03:23.888 --> 01:03:25.430 align:middle line:84% looking at wages directly, on do they 01:03:25.430 --> 01:03:27.770 align:middle line:84% or do they not clear and adjust and why or whatever. 01:03:27.770 --> 01:03:29.330 align:middle line:90% So that's another topic. 01:03:29.330 --> 01:03:32.760 align:middle line:84% But what would happen-- what does this look like over here? 01:03:32.760 --> 01:03:36.610 align:middle line:84% So now, it's kind of the opposite graph. 01:03:36.610 --> 01:03:45.580 align:middle line:84% So in this case, the farm would like 01:03:45.580 --> 01:03:51.250 align:middle line:84% to hire this amount of the labor at the market wage, 01:03:51.250 --> 01:03:54.860 align:middle line:90% but it can't get it. 01:03:54.860 --> 01:03:57.350 align:middle line:84% So how much-- so in that case, what's going to happen is 01:03:57.350 --> 01:03:59.150 align:middle line:90% it's going to get-- 01:03:59.150 --> 01:04:05.720 align:middle line:84% it'll hire-- it can only get this much L at the market wage. 01:04:05.720 --> 01:04:10.980 align:middle line:84% And then after that, then the household fills in the margin. 01:04:10.980 --> 01:04:13.267 align:middle line:84% And if the household is in this range over here-- 01:04:13.267 --> 01:04:15.600 align:middle line:84% like if the household kind of wanted to work more, fine. 01:04:15.600 --> 01:04:16.820 align:middle line:84% But if the household is in this range over here, 01:04:16.820 --> 01:04:19.460 align:middle line:84% then, again, the total amount that's produced in the farm 01:04:19.460 --> 01:04:22.340 align:middle line:84% is driven by the household utility function. 01:04:22.340 --> 01:04:26.450 align:middle line:84% The final graph they have in the paper on this is 01:04:26.450 --> 01:04:29.570 align:middle line:84% imagine that higher labor costs more than the farmers 01:04:29.570 --> 01:04:31.340 align:middle line:90% return to off-farm employment. 01:04:31.340 --> 01:04:32.330 align:middle line:90% What's that? 01:04:32.330 --> 01:04:36.500 align:middle line:84% Imagine that basically it is cheaper for-- 01:04:36.500 --> 01:04:39.060 align:middle line:84% it's cheaper and more efficient for me to do the work 01:04:39.060 --> 01:04:40.310 align:middle line:90% than for me to hire a laborer. 01:04:40.310 --> 01:04:43.050 align:middle line:84% So you might think there's agency problems-- moral hazard. 01:04:43.050 --> 01:04:45.050 align:middle line:84% So I'm going to work really hard on my own farm, 01:04:45.050 --> 01:04:47.330 align:middle line:84% and kind of slack off and work on someone else's farm. 01:04:47.330 --> 01:04:52.340 align:middle line:84% So in that case, if this is the market 01:04:52.340 --> 01:04:55.880 align:middle line:84% wage-- the wage at the-- the market wage, 01:04:55.880 --> 01:04:59.540 align:middle line:84% if I want to supply less labor in the market wage, then fine. 01:04:59.540 --> 01:05:02.360 align:middle line:84% I can supply that, and I can hire some extra at the market. 01:05:02.360 --> 01:05:04.940 align:middle line:90% That's separation. 01:05:04.940 --> 01:05:08.272 align:middle line:84% If I want to supply tons at outside wage, 01:05:08.272 --> 01:05:09.230 align:middle line:90% I can supply over here. 01:05:09.230 --> 01:05:09.867 align:middle line:90% That's fine. 01:05:09.867 --> 01:05:11.450 align:middle line:84% But in this intermediate range, again, 01:05:11.450 --> 01:05:15.890 align:middle line:84% will be driven by these tangencies. 01:05:15.890 --> 01:05:16.390 align:middle line:90% OK. 01:05:16.390 --> 01:05:18.700 align:middle line:84% So what's the point of going through that? 01:05:18.700 --> 01:05:20.650 align:middle line:84% If we think that the market is fully-- 01:05:20.650 --> 01:05:21.820 align:middle line:90% is going back to this point. 01:05:21.820 --> 01:05:24.300 align:middle line:84% If the market is fully efficient, 01:05:24.300 --> 01:05:28.350 align:middle line:84% there should be no relationship between how much labor I want 01:05:28.350 --> 01:05:34.840 align:middle line:84% to supply and how much labor is used on the farm. 01:05:34.840 --> 01:05:37.060 align:middle line:84% But with these different kinds of restrictions, 01:05:37.060 --> 01:05:39.940 align:middle line:84% if I can't hire enough labor, or I can't sell enough labor, 01:05:39.940 --> 01:05:42.150 align:middle line:90% I may end up either wanting to-- 01:05:42.150 --> 01:05:44.643 align:middle line:84% how much-- it may be that how much labor I can provide 01:05:44.643 --> 01:05:46.060 align:middle line:84% on my own farm determines how much 01:05:46.060 --> 01:05:47.470 align:middle line:90% labor is available on the farm. 01:05:47.470 --> 01:05:51.610 align:middle line:84% Or conversely, it may be that how much labor I can just 01:05:51.610 --> 01:05:52.660 align:middle line:90% supply-- 01:05:52.660 --> 01:05:54.220 align:middle line:84% how much work I can get just depends 01:05:54.220 --> 01:05:56.178 align:middle line:84% on how much work there is to do on my own farm. 01:05:56.178 --> 01:05:58.840 align:middle line:84% And in those cases, the labor supply decisions 01:05:58.840 --> 01:06:01.990 align:middle line:84% and the production decisions are interlinked. 01:06:01.990 --> 01:06:04.967 align:middle line:90% That's the basic idea. 01:06:04.967 --> 01:06:05.550 align:middle line:90% Is that clear? 01:06:05.550 --> 01:06:08.670 align:middle line:90% 01:06:08.670 --> 01:06:09.740 align:middle line:90% OK. 01:06:09.740 --> 01:06:14.010 align:middle line:84% So then, the idea of the paper is 01:06:14.010 --> 01:06:23.180 align:middle line:84% to say, well, if that's true, then 01:06:23.180 --> 01:06:28.480 align:middle line:84% do household demographic characteristics, which 01:06:28.480 --> 01:06:32.450 align:middle line:84% should affect labor supply, do those affect 01:06:32.450 --> 01:06:35.780 align:middle line:90% labor demand for the firm-- 01:06:35.780 --> 01:06:40.050 align:middle line:84% or for the farm-- for the farm, in this case. 01:06:40.050 --> 01:06:41.600 align:middle line:90% So that's the idea. 01:06:41.600 --> 01:06:46.500 align:middle line:84% You say, look, we have shifters for household labor supply. 01:06:46.500 --> 01:06:48.330 align:middle line:84% If we're in the separation world those 01:06:48.330 --> 01:06:52.090 align:middle line:84% should not affect labor demand on the farm. 01:06:52.090 --> 01:06:56.140 align:middle line:84% And if they do, then we're rejecting separation. 01:06:56.140 --> 01:06:57.630 align:middle line:90% Is that clear? 01:06:57.630 --> 01:06:58.130 align:middle line:90% OK. 01:06:58.130 --> 01:07:00.720 align:middle line:90% 01:07:00.720 --> 01:07:03.370 align:middle line:90% And I think this is a bit-- 01:07:03.370 --> 01:07:06.220 align:middle line:84% it's kind of related to a milder view of the surplus labor idea. 01:07:06.220 --> 01:07:07.780 align:middle line:84% If there are labor market frictions, 01:07:07.780 --> 01:07:10.570 align:middle line:84% you may employ labor on the farm even if the marginal product 01:07:10.570 --> 01:07:12.040 align:middle line:90% is below the outside wage. 01:07:12.040 --> 01:07:16.850 align:middle line:84% But we'll talk about this in a bit. 01:07:16.850 --> 01:07:18.680 align:middle line:84% But then this becomes the empirical test. 01:07:18.680 --> 01:07:22.780 align:middle line:90% 01:07:22.780 --> 01:07:29.710 align:middle line:84% So why might there be the misseparation? 01:07:29.710 --> 01:07:30.940 align:middle line:90% So here are a few. 01:07:30.940 --> 01:07:34.630 align:middle line:84% So one could be literally minimum wages. 01:07:34.630 --> 01:07:40.720 align:middle line:84% That would create minimum wages if the market wage is too high. 01:07:40.720 --> 01:07:42.640 align:middle line:84% And therefore, I may not be able to-- the farm 01:07:42.640 --> 01:07:44.470 align:middle line:84% may not be able to get as many workers 01:07:44.470 --> 01:07:47.150 align:middle line:90% as it would like at that wage. 01:07:47.150 --> 01:07:49.765 align:middle line:84% And the minimum wage could be a statutory minimum wage, 01:07:49.765 --> 01:07:51.890 align:middle line:84% or it could just be kind of a prevailing wage norm. 01:07:51.890 --> 01:07:54.200 align:middle line:84% And we'll talk about that in some of the papers 01:07:54.200 --> 01:07:56.280 align:middle line:90% we'll look at next time. 01:07:56.280 --> 01:08:00.360 align:middle line:84% It could be imperfect labor markets or agency problems 01:08:00.360 --> 01:08:02.230 align:middle line:90% or other market failures. 01:08:02.230 --> 01:08:05.310 align:middle line:84% So there could things going on in the labor market, 01:08:05.310 --> 01:08:10.450 align:middle line:84% so the thing isn't totally [? clearing. ?] OK. 01:08:10.450 --> 01:08:15.750 align:middle line:84% So any questions on that basic theoretical setup? 01:08:15.750 --> 01:08:16.250 align:middle line:90% Yes? 01:08:16.250 --> 01:08:16.750 align:middle line:90% No? 01:08:16.750 --> 01:08:18.810 align:middle line:90% Clear? 01:08:18.810 --> 01:08:21.600 align:middle line:90% OK. 01:08:21.600 --> 01:08:22.770 align:middle line:90% So as you guys noted-- 01:08:22.770 --> 01:08:26.062 align:middle line:90% 01:08:26.062 --> 01:08:28.229 align:middle line:84% so what did you think of the empirics in this paper? 01:08:28.229 --> 01:08:31.489 align:middle line:90% 01:08:31.489 --> 01:08:33.406 align:middle line:84% There were various concerns about the empirics 01:08:33.406 --> 01:08:36.380 align:middle line:84% in this paper that I read in your comments. 01:08:36.380 --> 01:08:37.408 align:middle line:90% So I want to-- 01:08:37.408 --> 01:08:38.200 align:middle line:90% what did you think? 01:08:38.200 --> 01:08:42.200 align:middle line:90% 01:08:42.200 --> 01:08:42.800 align:middle line:90% Somebody. 01:08:42.800 --> 01:08:48.240 align:middle line:90% 01:08:48.240 --> 01:08:49.735 align:middle line:90% Yeah, Ahmed. 01:08:49.735 --> 01:08:53.015 align:middle line:84% AUDIENCE: Wages can impact population density [INAUDIBLE].. 01:08:53.015 --> 01:08:53.890 align:middle line:90% BENJAMIN OLKEN: What? 01:08:53.890 --> 01:08:55.109 align:middle line:90% Sorry? 01:08:55.109 --> 01:08:58.630 align:middle line:84% AUDIENCE: Population density can impact wages, yes. 01:08:58.630 --> 01:09:01.350 align:middle line:84% But can it go the other way around? 01:09:01.350 --> 01:09:03.450 align:middle line:84% [INAUDIBLE] using population density 01:09:03.450 --> 01:09:10.978 align:middle line:84% as an instrument for labor supply, like [INAUDIBLE]?? 01:09:10.978 --> 01:09:13.800 align:middle line:84% I don't think [INAUDIBLE] any consumption would be 01:09:13.800 --> 01:09:15.569 align:middle line:90% satisfied for [? industry. ?] 01:09:15.569 --> 01:09:17.350 align:middle line:84% BENJAMIN OLKEN: For using popula-- yes. 01:09:17.350 --> 01:09:17.850 align:middle line:90% Right. 01:09:17.850 --> 01:09:18.885 align:middle line:90% So-- and why? 01:09:18.885 --> 01:09:24.350 align:middle line:90% 01:09:24.350 --> 01:09:27.800 align:middle line:84% AUDIENCE: So you are interested in the effect on labor 01:09:27.800 --> 01:09:31.149 align:middle line:90% supply on wages, right? 01:09:31.149 --> 01:09:38.050 align:middle line:84% Say places that have higher wages can attract people. 01:09:38.050 --> 01:09:40.110 align:middle line:84% [INAUDIBLE] it can increase population density. 01:09:40.110 --> 01:09:42.950 align:middle line:90% 01:09:42.950 --> 01:09:45.433 align:middle line:90% Did that cor-- one second. 01:09:45.433 --> 01:09:48.260 align:middle line:90% 01:09:48.260 --> 01:09:50.973 align:middle line:90% I just don't see any way of-- 01:09:50.973 --> 01:09:51.890 align:middle line:90% BENJAMIN OLKEN: So OK. 01:09:51.890 --> 01:09:53.080 align:middle line:90% So because I'm on-- so OK. 01:09:53.080 --> 01:09:55.270 align:middle line:84% So, as you saw, one of the things he's doing 01:09:55.270 --> 01:09:58.420 align:middle line:84% is using population density as an instrument for wages-- 01:09:58.420 --> 01:10:00.332 align:middle line:84% as an instrument for labor supply. 01:10:00.332 --> 01:10:02.290 align:middle line:84% His idea is that if there's more people around, 01:10:02.290 --> 01:10:05.740 align:middle line:84% that's going to lead to lower wages. 01:10:05.740 --> 01:10:07.990 align:middle line:84% So why is that not a good instrument for labor supply? 01:10:07.990 --> 01:10:09.370 align:middle line:90% Someone know? 01:10:09.370 --> 01:10:10.960 align:middle line:90% What are the concerns there? 01:10:10.960 --> 01:10:14.530 align:middle line:90% 01:10:14.530 --> 01:10:15.870 align:middle line:90% Yeah? 01:10:15.870 --> 01:10:19.167 align:middle line:84% AUDIENCE: Population density [INAUDIBLE].. 01:10:19.167 --> 01:10:20.000 align:middle line:90% BENJAMIN OLKEN: How? 01:10:20.000 --> 01:10:23.416 align:middle line:84% AUDIENCE: [INAUDIBLE] or [INAUDIBLE] 01:10:23.416 --> 01:10:27.645 align:middle line:90% in cities like [INAUDIBLE]. 01:10:27.645 --> 01:10:29.020 align:middle line:84% BENJAMIN OLKEN: So one is there's 01:10:29.020 --> 01:10:30.690 align:middle line:90% demand for nontradeables. 01:10:30.690 --> 01:10:33.107 align:middle line:84% So like if we have-- if labor supply-- 01:10:33.107 --> 01:10:35.440 align:middle line:84% if pop-- in people-- where there are more people around, 01:10:35.440 --> 01:10:36.630 align:middle line:90% we need more haircuts. 01:10:36.630 --> 01:10:38.580 align:middle line:84% So it could affect it through nontradeables. 01:10:38.580 --> 01:10:39.240 align:middle line:90% Any other? 01:10:39.240 --> 01:10:41.840 align:middle line:90% 01:10:41.840 --> 01:10:44.230 align:middle line:84% I'll just give you another one, which is that-- 01:10:44.230 --> 01:10:47.428 align:middle line:84% so why are all those people there? 01:10:47.428 --> 01:10:49.720 align:middle line:84% Maybe there's something productive about the production 01:10:49.720 --> 01:10:51.970 align:middle line:84% side, which gave all those people there to begin with. 01:10:51.970 --> 01:10:54.430 align:middle line:84% So, in particular, this one takes place in-- this study 01:10:54.430 --> 01:10:56.390 align:middle line:84% takes place in Java in Indonesia, 01:10:56.390 --> 01:10:59.885 align:middle line:84% which has a lot of volcanic soil which is super productive. 01:10:59.885 --> 01:11:02.260 align:middle line:84% And that's kind of why all the people are there, I think, 01:11:02.260 --> 01:11:04.280 align:middle line:84% is because the soil is productive. 01:11:04.280 --> 01:11:07.800 align:middle line:84% So it's not necessarily a [? suitable ?] instrument. 01:11:07.800 --> 01:11:10.213 align:middle line:90% So just as you guys-- 01:11:10.213 --> 01:11:11.880 align:middle line:84% I think you guys already saw this, but-- 01:11:11.880 --> 01:11:12.380 align:middle line:90% I'm sure. 01:11:12.380 --> 01:11:15.390 align:middle line:84% So-- in fact, I'm sure you guys have already seen this, right? 01:11:15.390 --> 01:11:18.202 align:middle line:84% So this is from an earlier version of the class 01:11:18.202 --> 01:11:19.410 align:middle line:90% where we taught this earlier. 01:11:19.410 --> 01:11:21.090 align:middle line:90% But obviously, you need-- 01:11:21.090 --> 01:11:26.750 align:middle line:84% the [INAUDIBLE] this is probably not going to hold in this case. 01:11:26.750 --> 01:11:29.110 align:middle line:84% So another issue that comes up is this question 01:11:29.110 --> 01:11:31.240 align:middle line:90% of endogenous household size. 01:11:31.240 --> 01:11:32.275 align:middle line:90% So what's the issue? 01:11:32.275 --> 01:11:34.872 align:middle line:90% 01:11:34.872 --> 01:11:36.747 align:middle line:84% Was there [INAUDIBLE] going to say something? 01:11:36.747 --> 01:11:38.445 align:middle line:84% AUDIENCE: If I want more labor, I'll have more kids. 01:11:38.445 --> 01:11:38.910 align:middle line:90% BENJAMIN OLKEN: Huh? 01:11:38.910 --> 01:11:40.660 align:middle line:84% AUDIENCE: If I want more labor, I'll have more kids. 01:11:40.660 --> 01:11:41.493 align:middle line:90% BENJAMIN OLKEN: Yes. 01:11:41.493 --> 01:11:43.370 align:middle line:84% If I want more labor, I'll have more kids. 01:11:43.370 --> 01:11:45.370 align:middle line:84% So now I want to ask you guys a harder question. 01:11:45.370 --> 01:11:46.885 align:middle line:90% Is that a problem? 01:11:46.885 --> 01:11:50.310 align:middle line:90% 01:11:50.310 --> 01:11:55.740 align:middle line:84% And what I mean by is that a problem is can that generate-- 01:11:55.740 --> 01:11:57.710 align:middle line:90% suppose that was true. 01:11:57.710 --> 01:11:59.460 align:middle line:84% So let's put us-- so suppose we're 01:11:59.460 --> 01:12:07.190 align:middle line:84% under the null hypothesis of delta equals 0. 01:12:07.190 --> 01:12:10.160 align:middle line:84% Can endogenous household size generate a false rejection 01:12:10.160 --> 01:12:10.730 align:middle line:90% of the null? 01:12:10.730 --> 01:12:27.845 align:middle line:90% 01:12:27.845 --> 01:12:30.215 align:middle line:84% Do you see what-- do you see the question I'm asking? 01:12:30.215 --> 01:12:30.965 align:middle line:90% What do you think? 01:12:30.965 --> 01:12:41.110 align:middle line:90% 01:12:41.110 --> 01:12:42.010 align:middle line:90% Yeah, Vashan? 01:12:42.010 --> 01:12:44.552 align:middle line:84% AUDIENCE: Does endogenous [INAUDIBLE] to-- 01:12:44.552 --> 01:12:47.010 align:middle line:84% that the labor [? method ?] should have the [? left-hand ?] 01:12:47.010 --> 01:12:47.970 align:middle line:90% side? 01:12:47.970 --> 01:12:50.310 align:middle line:90% Then you'll get a correlation. 01:12:50.310 --> 01:12:51.810 align:middle line:84% But in terms of [INAUDIBLE],, that's 01:12:51.810 --> 01:12:55.350 align:middle line:84% fine because you're trying to see if there's 01:12:55.350 --> 01:12:56.430 align:middle line:90% a correlation anyway. 01:12:56.430 --> 01:12:57.160 align:middle line:90% I don't know. 01:12:57.160 --> 01:12:59.160 align:middle line:84% This is only a [? distant ?] correlation, right? 01:12:59.160 --> 01:12:59.760 align:middle line:90% BENJAMIN OLKEN: Yeah. 01:12:59.760 --> 01:13:01.620 align:middle line:84% AUDIENCE: So in that sense, it doesn't matter 01:13:01.620 --> 01:13:02.520 align:middle line:90% whether you're [INAUDIBLE]. 01:13:02.520 --> 01:13:03.210 align:middle line:90% BENJAMIN OLKEN: Exactly. 01:13:03.210 --> 01:13:03.710 align:middle line:90% Right? 01:13:03.710 --> 01:13:05.200 align:middle line:90% So this is the idea. 01:13:05.200 --> 01:13:13.070 align:middle line:84% So in Wesley's story, I have more kids 01:13:13.070 --> 01:13:15.750 align:middle line:84% because I want more labor in the household. 01:13:15.750 --> 01:13:21.240 align:middle line:84% But the only reason I do that is because of separation failure. 01:13:21.240 --> 01:13:26.460 align:middle line:84% So if, in fact, there was no separation failure, 01:13:26.460 --> 01:13:28.380 align:middle line:84% the delta was equal to 0, then I wouldn't 01:13:28.380 --> 01:13:30.870 align:middle line:90% be having more kids to-- 01:13:30.870 --> 01:13:32.240 align:middle line:90% right? 01:13:32.240 --> 01:13:34.563 align:middle line:84% So I think-- I do think this is a bit of an issue, 01:13:34.563 --> 01:13:36.230 align:middle line:84% but I think it's worth thinking through, 01:13:36.230 --> 01:13:40.195 align:middle line:84% is this actually a problem in there? 01:13:40.195 --> 01:13:41.570 align:middle line:84% What I do think could be an issue 01:13:41.570 --> 01:13:42.680 align:middle line:90% is if we had omitted variables. 01:13:42.680 --> 01:13:45.055 align:middle line:84% Like, for example, if we have better land quality leading 01:13:45.055 --> 01:13:48.350 align:middle line:84% to more income, and more income leading 01:13:48.350 --> 01:13:53.240 align:middle line:84% to that kids are a normal good, then that could-- 01:13:53.240 --> 01:13:58.380 align:middle line:90% that channel could be a problem. 01:13:58.380 --> 01:13:59.840 align:middle line:84% But that's a more complicated story 01:13:59.840 --> 01:14:01.100 align:middle line:84% than I think the one you were thinking of. 01:14:01.100 --> 01:14:03.170 align:middle line:84% Because your story is only a problem under-- 01:14:03.170 --> 01:14:04.170 align:middle line:90% if the null is false. 01:14:04.170 --> 01:14:06.340 align:middle line:84% But, in fact, under the null, it's not a problem. 01:14:06.340 --> 01:14:08.257 align:middle line:84% So I just think it's worth being a little more 01:14:08.257 --> 01:14:12.142 align:middle line:84% precise about the concerns that we have sometimes. 01:14:12.142 --> 01:14:14.100 align:middle line:84% Does everyone understand this alternative story 01:14:14.100 --> 01:14:16.267 align:middle line:84% and why this-- this could generate a false reduction 01:14:16.267 --> 01:14:17.800 align:middle line:90% in the null? 01:14:17.800 --> 01:14:18.860 align:middle line:90% Yes? 01:14:18.860 --> 01:14:19.360 align:middle line:90% OK. 01:14:19.360 --> 01:14:22.832 align:middle line:90% 01:14:22.832 --> 01:14:25.290 align:middle line:84% What other-- any other comments or thoughts from this paper 01:14:25.290 --> 01:14:27.413 align:middle line:90% from your-- 01:14:27.413 --> 01:14:29.580 align:middle line:84% what else do you guys think of this from reading it? 01:14:29.580 --> 01:14:34.370 align:middle line:90% 01:14:34.370 --> 01:14:36.095 align:middle line:90% Any other questions or thoughts? 01:14:36.095 --> 01:14:39.580 align:middle line:90% 01:14:39.580 --> 01:14:40.550 align:middle line:90% No? 01:14:40.550 --> 01:14:41.360 align:middle line:90% OK. 01:14:41.360 --> 01:14:50.460 align:middle line:84% So as many of you noted, the empirics of this paper 01:14:50.460 --> 01:14:52.210 align:middle line:90% are a little bit dated. 01:14:52.210 --> 01:14:56.780 align:middle line:90% And so one natural solution-- 01:14:56.780 --> 01:14:59.290 align:middle line:84% so LaFave and Thomas is kind of literally just updating 01:14:59.290 --> 01:15:01.780 align:middle line:90% this paper with better empirics. 01:15:01.780 --> 01:15:05.000 align:middle line:84% And, in particular, they have panel data. 01:15:05.000 --> 01:15:06.940 align:middle line:84% And so why is panel data going to help? 01:15:06.940 --> 01:15:10.750 align:middle line:90% 01:15:10.750 --> 01:15:11.260 align:middle line:90% Yeah? 01:15:11.260 --> 01:15:12.840 align:middle line:90% AUDIENCE: Land fixed effects. 01:15:12.840 --> 01:15:14.700 align:middle line:84% BENJAMIN OLKEN: So you're going to have land fixed effects. 01:15:14.700 --> 01:15:15.825 align:middle line:90% So what's going to be left? 01:15:15.825 --> 01:15:16.775 align:middle line:90% So what else-- right. 01:15:16.775 --> 01:15:18.900 align:middle line:84% So now we're going to have land fixed effects so we 01:15:18.900 --> 01:15:21.210 align:middle line:84% can control for all the heterogeneity of land 01:15:21.210 --> 01:15:22.350 align:middle line:90% quality and [INAUDIBLE]. 01:15:22.350 --> 01:15:22.710 align:middle line:90% That's good. 01:15:22.710 --> 01:15:23.210 align:middle line:90% What else? 01:15:23.210 --> 01:15:29.848 align:middle line:90% 01:15:29.848 --> 01:15:32.390 align:middle line:84% AUDIENCE: Suppose we can find good variation in labor supply. 01:15:32.390 --> 01:15:33.080 align:middle line:90% BENJAMIN OLKEN: Huh? 01:15:33.080 --> 01:15:34.650 align:middle line:84% AUDIENCE: We can find good variation-- 01:15:34.650 --> 01:15:35.442 align:middle line:90% BENJAMIN OLKEN: No. 01:15:35.442 --> 01:15:36.660 align:middle line:90% What's the variation? 01:15:36.660 --> 01:15:40.470 align:middle line:84% AUDIENCE: Over time, even if a place has, like you say, 01:15:40.470 --> 01:15:41.340 align:middle line:90% productive land-- 01:15:41.340 --> 01:15:42.132 align:middle line:90% BENJAMIN OLKEN: No. 01:15:42.132 --> 01:15:44.118 align:middle line:90% It's individuals, households. 01:15:44.118 --> 01:15:44.910 align:middle line:90% AUDIENCE: There's-- 01:15:44.910 --> 01:15:45.840 align:middle line:90% BENJAMIN OLKEN: Huh? 01:15:45.840 --> 01:15:47.530 align:middle line:90% AUDIENCE: Migration [INAUDIBLE]. 01:15:47.530 --> 01:15:53.822 align:middle line:90% 01:15:53.822 --> 01:15:55.030 align:middle line:90% BENJAMIN OLKEN: Someone else? 01:15:55.030 --> 01:15:57.730 align:middle line:84% So why would panel data help solve some of the problems 01:15:57.730 --> 01:15:58.665 align:middle line:90% from before? 01:15:58.665 --> 01:16:03.220 align:middle line:90% 01:16:03.220 --> 01:16:03.760 align:middle line:90% OK. 01:16:03.760 --> 01:16:04.730 align:middle line:90% We're out of time. 01:16:04.730 --> 01:16:06.280 align:middle line:84% I will leave you to think-- ponder this question. 01:16:06.280 --> 01:16:08.060 align:middle line:84% We'll come back to it at the beginning of next class. 01:16:08.060 --> 01:16:09.190 align:middle line:84% So you can have a look at this paper, 01:16:09.190 --> 01:16:11.560 align:middle line:84% and then I think whatever the next couple of starred papers 01:16:11.560 --> 01:16:12.393 align:middle line:90% on the syllabus are. 01:16:12.393 --> 01:16:13.743 align:middle line:90% We'll come back to this on-- 01:16:13.743 --> 01:16:14.410 align:middle line:90% actually, sorry. 01:16:14.410 --> 01:16:17.470 align:middle line:90% There's no class on Wednesday. 01:16:17.470 --> 01:16:20.460 align:middle line:84% And we're going to come back to this-- so a week from today. 01:16:20.460 --> 01:16:35.000 align:middle line:90%