YT YouTube2Text

YouTube2Text → Transcripts → Lecture 7: Agricultural productivity gap: measurement and explanations

Lecture 7: Agricultural productivity gap: measurement and explanations — Transcript

by CEPR & VideoVox Economics · 17,508 words · 2,874 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:03okay welcome
0:04back to steggs virtual course on key
0:07concepts in macro development
0:09uh which is sponsored by cepr's research
0:12program
0:13steg which is structural transformation
0:15and economic growth a new research
0:17program on macroeconomic development
0:21of low-income countries
0:25it is funded as part of the uk aid
0:27effort
0:28by the foreign commonwealth and
0:30development office
0:31uh we'd like to thank them for their
0:33funding
0:34uh today we're excited to have david
0:36lagakos who we'd also like to thank
0:39um giving a lecture on the agricultural
0:42productivity gap agricultural
0:43productivity
0:44more generally
0:48david has a uh
0:51teaching assistant martin shu
0:53who will be handling the q a
0:56for people that are um attending
0:59as through the web the webinar uh you
1:01can enter questions in the q a and
1:03martin will do his best to answer those
1:05uh i'll probably take a crack at some of
1:06them as well if it's a broader question
1:09we'll save it at the end for david uh
1:12david will be um leaving about 15
1:14minutes at the end for questions
1:16uh people that are in the front row we
1:18would ask you to mute your microphones
1:20but if you have a clarifying question uh
1:22it just interrupt with it and david said
1:24even if it's not clarifying uh just
1:27interrupt with them uh but there will
1:28also will be time at the end uh for
1:30questions from the front row as well
1:32with that oh the slides again as usual
1:35the slides are available
1:37um
1:38online
1:40uh at the stag website
1:42um
1:43i want to also thank uh mandy chan and
1:46kirsty mcneil and lauren waring at cepr
1:49for helping organize this behind the
1:51scenes and i'll just hand it off and
1:53give it to david go ahead david okay
1:56thanks joe for the
1:58introduction and for the opportunity to
2:00give this lecture so this lecture as joe
2:02said is going to be on the agricultural
2:04productivity gap a topic that i've been
2:06thinking about now for actually about a
2:09decade
2:10and um so i'm going to present
2:12um an overview of the paper we wrote uh
2:14with doug gollin and michael wah
2:18from 2014 but i'm going to spend only a
2:22little bit of the time talking about
2:23that paper i'm going to spend kind of
2:25the last hopefully two-thirds of the
2:26talk talking about
2:29progress that's been made since this
2:31paper and um
2:33and directions that i think are
2:35valuable for future uh for future
2:37research
2:39um so i'm gonna talk in the in the paper
2:41i we we
2:43we cover kind of every country in the
2:44world in this in these slides i'm only
2:46gonna focus in on developing countries
2:47just as an expositional thing
2:50and so one of the things that kind of
2:52you know everybody knows about the
2:54agricultural sector in developing
2:55countries that it's very large
2:58something that at least until this paper
3:00was you know maybe under appreciated was
3:02that the share of value-added
3:05originating in agriculture is in most
3:07countries you know pretty much every
3:09country lower than the share of
3:11employment
3:12in agriculture
3:14and so this is a graphic that shows this
3:17for
3:18around 100
3:20developing countries on the x-axis it's
3:22the employment share of agriculture
3:25the y-axis the value-added share of
3:26agriculture and it's easy to download
3:28these data these are
3:30united nations um who puts together the
3:33system of national accounts
3:35um is where we got the value-added share
3:37and the employment share we got from the
3:38world bank and
3:40there's not that much disagreement about
3:42about these these numbers and if you can
3:44see the red line is a 45 degree line
3:47there's kind of you know
3:49one country on the other side and n
3:51minus one countries on this side and on
3:53this side so if you take you know take
3:55these african countries over here rwanda
3:57burkina faso chad madagascar you know
4:00about 90 of their workers are in
4:02agriculture at least this was this is
4:05yeah this is now
4:0610 years old it's probably lower now but
4:08not that much lower
4:09um
4:11come over here to y-axis to say you know
4:13something like 40
4:15of value-added emanates from
4:18from
4:18uh from agriculture and so you kind of
4:21as you get richer and richer you kind of
4:22move down this this [snorts] range you
4:25know but even in this kind of these are
4:26again these are developing countries but
4:28even in kind of the richest of these
4:29developing countries
4:30uh you still see that more of the you
4:32know larger share of people are in
4:34agriculture than than
4:35than a share of value added coming from
4:37agriculture
4:38um
4:40and so kind of just like by arithmetic
4:41that has to mean that value added per
4:43worker is lower in agriculture than in
4:46non-agriculture in these countries
4:49and so in this paper we um
4:52you know we give this ratio a name we
4:54call it the agricultural productivity
4:55gap
4:56and so for any country you can you can
4:58take those statistics i just showed you
5:00and you can compute the ratio of value
5:02added per worker in non-agriculture to
5:04agriculture
5:06i'll show you a simple
5:07a very simple model but i think kind of
5:10a sensible
5:11uh benchmark that says this gap should
5:13be one
5:14you know so gap of one is like you know
5:16no gap
5:18a typical developing country has a gap
5:20of around four
5:21and it's easy to find countries that
5:23have a gap of eight or more and eight
5:25means again you know but take it if i
5:28take a random non-agricultural worker
5:30off the streets they're gonna produce
5:31eight times as much value added
5:33uh
5:35as a random agricultural worker um
5:40so why are these gaps kind of you know
5:42potentially important i'm sure you've if
5:45you've come to the other lectures you've
5:47you've seen um you've seen a version of
5:49this argument you know agriculture
5:51employs most of the people
5:53and um
5:54and then just on a mechanical level
5:55they've got you know most of the most of
5:57the workers and so these kind of seminal
6:00papers by francesco caselli diego
6:02restuccia
6:04and co-authors and dietz volroth
6:06kind of highlighted how just as an
6:08accounting sense is so important for
6:09understanding why foreign companies have
6:12such you know low income
6:14levels
6:15so if you kind of just take these gaps
6:16at face value we're going to show in a
6:18minute that you know it's
6:20it
6:21it suggests that there's lots of
6:23misallocation
6:24and that by taking workers out of this
6:26low productivity sector and moving into
6:28a more productive sector you could raise
6:29aggregate gdp
6:31um
6:32there's a policy debate that is um
6:35not any smaller than it was when we
6:37started um
6:39work on this project about trying to
6:41think about moving workers out of
6:42agriculture is there a way we can
6:44encourage that um
6:46are there bad side effects of that you
6:48know they crowd cities or something this
6:50paper is not really going to contribute
6:51to that policy date directly nor is it
6:53going to make a claim that there's
6:54misallocation it's going to ask a kind
6:57of a kind of a more basic question about
6:59measurement of these
7:01of these shares of value-added and
7:03shares of employment so there's a
7:05measurement paper that tries to say
7:07let's think real hard about how
7:09data and agriculture non-agriculture be
7:12measured how employment or the labor
7:14input to be measured in these two
7:16sectors and see whether these gaps
7:19survive
7:20uh
7:21kind of more serious um measurement
7:25and so what do we do in this paper well
7:26we just concretely kind of a real old
7:29story about why there could be gaps in
7:31our and uh
7:33value added per worker is that who
7:35people who are counted as workers end up
7:37not doing all that much work
7:39and so um so we're going to look at
7:41sector differences and hours worked per
7:43worker
7:44and um we're going to get that for 51
7:47developing countries
7:50another kind of basic story is that you
7:52know people who are in agriculture tend
7:53not to be the most educated workers it's
7:55hard to it's hard to find any evidence
7:58of that it's easy to find evidence of
7:59the opposite which is non-agricultural
8:01workers or where the more educated
8:03workers are gonna tend to go
8:04we're gonna try to measure human capital
8:06per worker at least using schooling and
8:09we're gonna do that in this paper for 98
8:12developing countries in the published
8:14version of the paper we get up to about
8:15150 countries total and for these slides
8:19as i said before i'm just going to show
8:20you the developing countries just um
8:22that's where the gaps are the biggest
8:24and it's a
8:25it's um
8:26it's just a choice and
8:28expositional choice really
8:30um
8:31and then finally the um
8:34you know the kind of the age-old issue
8:35is
8:36what about if the national income
8:37accounts are just
8:39agricultural output you know there's a
8:41lot of people you'd have to be
8:43you know um
8:45you're seriously demented to live in a
8:47developing country and not realize that
8:49people grow a lot of food in the
8:50backyard and eat it without actually
8:52selling it
8:53i've talked to national income accounts
8:55i've never met one who's so foolish as
8:57to not be aware of that
8:59but maybe still even if you're aware of
9:00this you just can't measure the
9:02subsistence farming and you just
9:03systematically undercut agriculture and
9:05um
9:06yeah that's kind of a lingering
9:08possibility so we're going to kind of
9:09try to recreate the national income
9:11accounts in this paper using our own
9:13estimates of value added using household
9:15survey data and that's going to be from
9:1710 countries um 10 low-income countries
9:21many in africa
9:23and um
9:25so that's what we do
9:27what do we conclude is that well if you
9:29do these various measurement you kind of
9:31let's call it better measurement
9:34you're going to be able to reduce the
9:35gap on average from four to around two
9:38and so that's kind of progress because
9:40about half of these gap we're going to
9:41argue is is
9:43is uh is miss measurement
9:45um but two is you know two is still not
9:48anywhere near one two still suggests the
9:50average worker's twice as productive
9:52even controlling for hours and human
9:54capital and mis-measurement output
9:57we're gonna look in the household survey
9:58data we're gonna find big gaps there too
10:00and so
10:01many micro development people that world
10:03bank's living measurement living
10:05standards measurement surveys have spent
10:07great
10:08you know great time and great effort to
10:10try to survey people about their
10:11consumption levels their output levels
10:13and their surveys don't really
10:15don't really show anything close to
10:17parity they also show large gaps
10:20and so in that
10:21this paper the paper concludes by saying
10:23not okay workers are misallocated now
10:26let's get them out some people kind of
10:28misinterpret our paper as saying that um
10:30we kind of are more agnostic in this
10:32paper and we say hey we need a better
10:34understanding of why these gaps are so
10:35large
10:36and so this kind of like i think maybe
10:39second let's say second half of the talk
10:41we'll uh we'll talk about where the
10:42literature's gone and what's left to do
10:46let me show you one slide
10:48to motivate uh the
10:50empirical work or just imagine the
10:51simplest of simple two sector models
10:54where there's a cop douglas production
10:56function in agriculture cobb douglas
10:58production function and non-agriculture
11:00imagine furthermore that there's the
11:02same power on labor
11:04um later in the paper we argue that
11:06that's actually not not such a bad
11:08assumption
11:09and imagine crucially that households
11:11can supply labor to either sector so
11:14there's just foot lose households they
11:15can do whatever they want they can pick
11:17agriculture they can pick
11:18non-agriculture
11:20and there's competitive labor markets so
11:22that workers are paid their marginal
11:24product
11:26well what happens in equilibrium
11:29in equilibrium workers are paid their
11:30marginal products and they're foot loose
11:32so you can't have a different wage in
11:33the agriculture sector in the
11:34non-agriculture it's got to be one wage
11:36and so the marginal product of labor is
11:38equal to this single wage and that means
11:40the marginal products of labor are
11:42equated in the two sectors
11:45and
11:46and
11:47by this cobb douglas assumption you get
11:49that because the marginal marginal
11:51products of labor are the same in the
11:53two sectors well then the average
11:54products are the same and so this
11:56agricultural productivity gap which is
11:57the average product in non-agriculture
12:00relative to the average product of labor
12:03in the agriculture sector is just one
12:07um
12:08so that's kind of the model you might
12:09say
12:10become a quick model yeah maybe so it's
12:13a useful benchmark
12:15so what about in the data where do we
12:17get these um
12:19agricultural productivity gaps
12:21um so you can get them from the well i
12:23say here the world bank there's many
12:25many ways you can get them they
12:26ultimately come from the countries who
12:28put their own national income accounts
12:29together
12:30and value added in in um in all the
12:33countries we're going to look at is
12:34defined as in the 1993 system of
12:37national accounts
12:39if you're really into national accounts
12:40you can go back and look this up but
12:41they define what's measured in output
12:43and so home product home production of
12:45goods that's in a system of national
12:48accounts that's output
12:49if you do services in your own house
12:51like you clean your own
12:53you clean your own
12:55house or take care of your own children
12:56that's home produced service that's not
12:58the national income accounts
13:01measures of la and ln the
13:04labor input and agriculture and
13:06non-agriculture those are the we're
13:07going to measure as the economically
13:09active population in each sector and
13:11that's just you know you can download
13:13data on economically active population
13:15but it basically amounts to employed or
13:18unemployed persons who are working
13:20or seeking work in the production of
13:22some good or service recognized by the
13:25system of national accounts and so this
13:27could be
13:28factory workers this could be people
13:30working in shops or restaurants this
13:32could be people growing food in the
13:34backyard because if you're growing food
13:36in the backyard your output counts in
13:38the system of national accounts
13:40and we get these from the world bank um
13:43you could get these from other places
13:45much of what we're going to do we're
13:46going to go to original population
13:48censuses or labor force surveys but
13:50these data are basically easy to you
13:52could go download these data so there's
13:53not a lot of contribution in this part
13:55of the paper
13:57here's 413 developing countries the same
14:00one i plotted on that graph here's the
14:02raw agricultural productivity gaps which
14:04are no adjustments for anything just um
14:07just taking the ratio of value added per
14:09worker and non-agriculture to
14:10agriculture so just you know you're
14:12going to show you every every
14:14statistic in this table but the mean
14:16countries where we weight
14:18uh by population of the country's four
14:21unweighted means 3.6
14:23um even if you go to the five fifth
14:25percentile you're still well above one
14:281.7 here if your weights and the 95th
14:31percentile by weights is five and
14:33unweighted is you know it's nine so
14:35these are very you know nowhere near the
14:37theory that i just showed you which is
14:38one
14:41um in the paper we talk about richer
14:42countries richer countries also have
14:44gaps they're um they're lower
14:46um they're lower than the poor country
14:48camps
14:50so one kind of kind of like basic
14:52concern is that um as i mentioned
14:55there's uh home production that's just
14:57um is just uh
14:59is just not counted
15:01and um in principle as i said it is
15:03counted but you know maybe in practice
15:05people have no idea they have no idea
15:07how to do this so in the few statistical
15:10agencies i've talked to the answer i
15:12usually get about how this is actually
15:13done in practice is
15:16crop by crop they estimate the output
15:19based on the product of two things which
15:20they can they can measure independently
15:22one is the area planted
15:25and so if you're in your country they
15:26grow maize
15:28um you can estimate through surveys or
15:30aerial footage or other information how
15:32much area is devoted to maize in the
15:34country
15:35um and then you know yields for most
15:39major crops are kind of well known in
15:41some places it's two tons per hectare
15:42and others it's 2.5 so people have ideas
15:45region by region what yields would be
15:48for these these crops
15:50and if you take a yield times an area
15:52planted output per land times the amount
15:54of land you're going to get output of
15:56that crop
15:57and so this is a way that you say well
15:59you know what if they sell it what if
16:00they eat it we don't know and you don't
16:01need to know
16:03it may be hard to estimate these things
16:04but this is this is actually a way that
16:06without getting into the nitty-gritty of
16:08whether you're gonna grow it or eat it
16:09you can get like you know kilograms of
16:11output and then you have to have some
16:13information on a price per kilogram and
16:15to turn um a quantity into a price
16:18and this is as far as i know what goes
16:20on in the major crops you know this uh
16:23this would cover the major grains wheat
16:25rice
16:26maize as far as i know you know for
16:28smaller tree crops uh pineapples things
16:30that are
16:31you know specialized those are harder to
16:33estimate and they may have some surveys
16:35but i actually don't think there's any
16:36i've never come across a reason to think
16:38that in even in practice we're
16:40systematically under counting these
16:42there's lots of reasons to think that in
16:44practice is hard to do
16:45um
16:47yeah
16:48what about agricultural employment one
16:50of one of the big worries we have is
16:51that
16:52you know maybe the productivity is low
16:54because they they go to villages and
16:56they just say hey how many people are in
16:57this village three thousand good call it
16:59three thousand you know agricultural
17:01workers then productivity would be low
17:03because you get a lot of people you know
17:04older people kids
17:06you know
17:07people who stay at home and don't work
17:09they'll be counted as agricultural
17:10workers and so like in principle um
17:14only economically active persons are
17:16included per the system of national
17:17income accounts
17:19basically we we use household surveys
17:21that i'll tell you about from a whole
17:22bunch of different countries and we
17:24don't find that they're they're really
17:25very different in terms of economically
17:27active people in agriculture
17:31um and so what we do
17:34concretely to to try to adjust the
17:36denominators here this or the so the
17:38workers is in the measure of labor input
17:41per sector we're going to show you two
17:42graphs one is sector differences in
17:44hours worked per worker
17:48so if you're working more hours you know
17:49you're going to be more productive but
17:51that doesn't mean that the hourly wage
17:53is different
17:54human capital per worker i'll show you
17:56that um that's another kind of concrete
17:58story about why the output would be so
18:01much less per worker the measures of
18:03value added per worker i'll show you in
18:04a few minutes we're going to reconstruct
18:06the national income accounts as i said
18:08from from household survey data
18:12so this is kind of the main we have a
18:13number of graphs and tables about hours
18:15but this is kind of the main summary one
18:17so this this again shows
18:19uh developing countries i think there
18:20are 50 of them here
18:22and one here means
18:24just the same number of hours on average
18:26among agricultural workers and
18:28non-agriculture workers 1.5 means 50
18:31more hours for aggregate for non excuse
18:33me non-agriculture workers two is twice
18:35as many hours for non-agriculture
18:37workers and you see kind of on average
18:39you know yeah there's some clumping
18:40around the the one line here but much of
18:43this is between 1 and 1.5 i think on
18:46average it's about 25 percent more hours
18:49if i'm not mistaken in this group of
18:50countries on average in the
18:52non-non-agricultural
18:54sector
18:55i think that's kind of interesting in
18:57its own right um you got real poor
18:59people aren't working at many hours and
19:01and i think there's kind of
19:02you know
19:03very legitimate constraints on the
19:05number of hours you can work if you're
19:07productively anyways if you're a
19:08subsistence farmer but for the purpose
19:10of this paper let's just take this as
19:12given let's just say okay it's not
19:14output per worker we want we want output
19:16per hour
19:17um and you can see a clear reason here
19:19higher hours in the non-agriculture
19:21sector that explains at least some chunk
19:23of this gap
19:27to measure sector differences in human
19:28capital
19:31by the way that for the frontier of this
19:32i there's these great papers by porzio
19:35rossi and santangelo and hobain shulman
19:38and vindis that try to think about
19:40um how kind of fundamentally
19:43you know to get out of the agriculture
19:44sector you need human capital um so
19:46these are these post art paper i just
19:48wanted to plug these these papers here
19:51if you want to get to the frontier these
19:54are the ones to look at
19:55what we did i thought was the frontier
19:57at the time it's um so we basically we
20:00go to
20:0297 different household surveys where
20:05individual by individual we know what
20:07their educational attainment is
20:09and using a measure of educational
20:11attainment like i finished high school
20:13or i didn't finish high school i only
20:14finished primary or i did some primary
20:17you can impute the um years of
20:21schooling that's what we do for almost
20:23all the countries
20:24in a handful of countries they actually
20:26directly say how many years of schooling
20:27did you complete
20:29and we do in the paper a number of more
20:31sophisticated things but but kind of a
20:34real basic
20:35um
20:36basic way to go from years of schooling
20:39to human capital is to just assume a
20:40constant rate of return
20:42per year of schooling
20:44so the most basic thing we do is we
20:45assume a ten percent uh return on a year
20:48of schooling so for every country you
20:50take their schooling multiply by 10 take
20:53e raise it raise it to that power and
20:55you're gonna get uh the human capital
20:57where does this come from well this is
21:00like motivated by a mints or regression
21:02if you take logs of both sides of this
21:04equation you get the log of human
21:06capital which we're going to proxy by
21:08wage is equal to schooling times 0.1 so
21:11that's like you know that mincer
21:13regression is saying every year of
21:15schooling gives you 10 percent more
21:16wages we're going to kind of just impute
21:18human capital in that way
21:20so this is what caselli does i think
21:22this is a nice a nice
21:25a nice way of practicing human capital
21:29and here's what the
21:31ratios of human capital look like again
21:33one is just as much human capital in the
21:35agriculture sector as non-agriculture
21:381.5 is 50 more
21:402 is twice as much
21:42and
21:43you just kind of go back to the formula
21:45suppose i plug in zero years of
21:47schooling
21:48you know zero e to the zero was um is
21:51one so one is like a blank slate you
21:53know one is like zero years of schooling
21:56some of these um agricultural countries
21:59that molly is in here i think guinea was
22:01in here
22:02are close to you know are close to this
22:03blank slate
22:05and um but overall the ratio here is i
22:07think 1.4
22:09meaning that all else equal we expect 40
22:12higher wages on average or 40 percent
22:14higher value added per worker in the
22:16non-agriculture sector based purely on
22:18these observed differences in schooling
22:22and so 1.4 is already kind of helping
22:24take a big chunk out of this gap of of
22:27four that we see
22:29and so what we do in the paper i'm
22:30giving you the kind of the
22:33um
22:34you know the quick version we we take
22:36these gaps that we originally showed you
22:38and we adjust them four hours worked so
22:40we make them rather than our value-added
22:42per worker we make them a value-added
22:43per hour
22:45and then we adjust them again by the
22:47human capital ratios i just showed you
22:49so now we get from value added per
22:51worker to evaluated per hour of human
22:53capital services if you will
22:56so um
22:58so what do you get with these adjusted
23:00gaps
23:01um
23:02mean now is about uh two
23:05um where we have um complete data
23:09if you take all the countries where we
23:10impute when we don't have it you're
23:12still kind of in the ballpark of two at
23:14the fifth percentile now you're actually
23:15less than one so um it kind of goes the
23:18opposite direction 95th percentile
23:20you're still around four so there's
23:22still uh still big gaps in many
23:24countries on average or about two so
23:26that's like when i said we make these
23:27adjustments to go from four to two this
23:30table is what i was what i was talking
23:32about
23:33um
23:35so that's what we do in the labor input
23:37for the next i guess two or three slides
23:39i want to talk about what about if we're
23:41systematically mismeasuring
23:43value added
23:45um
23:46and so what we're going to do here as i
23:47said before we're going to take the
23:49world bank's living standards
23:50measurement surveys which are kind of
23:52the gold standards of how you measure
23:54income and expenditure at the household
23:56level
23:57um
23:58in fact that's the explicit goal of the
24:00lsms
24:02and we're going to we're gonna try to
24:03compute value-added from the ground up
24:06and so i have one i think this is you
24:08know this is a very low-tech talk in
24:10general intentionally so it's a low-tech
24:13paper but this is the one you know the
24:15one side where you have to pay somewhat
24:16of attention if you don't want to it's
24:18it's no big deal but basically this is
24:20how we this is how we measure value
24:21added in agriculture so this is at the
24:23level of this the the sector so what we
24:26do using these household surveys is we
24:28add up all the self-employed
24:30agricultural income over all
24:32individuals eye i'll say in a moment
24:35what i mean by that you add up all the
24:37labor income and so many surveys ask do
24:39you work for a wage yeah okay how much
24:42did you earn last month
24:44and then all the capital income so this
24:46there's not a lot of this but this would
24:48be do you own some land yes do you rent
24:50it yeah how much do you rent it for that
24:52would be an example do you own an
24:53apartment building or an apartment do
24:55you own a housing unit and do you rent
24:57that that's unambiguous capital income
25:00and we add all the unambiguous capital
25:02income by agricultural households
25:04unambiguous labor income by agricultural
25:07households and the self-employed value
25:10added or self-employed output let's say
25:12and that's that's how we measure value
25:14added in agriculture
25:16and so this self-employed income here is
25:18where the advantages of the ls ms data
25:21show through and so now what we do here
25:24is we say take every agricultural crop
25:26in those data
25:28and um we're going to multiply the price
25:31of that crop
25:32times the sum of three quantities and so
25:35this would be like kilograms these are
25:36quantities kilograms of that crop as
25:39kilograms of maize that you have for
25:40home production
25:42how many kilograms of maize did you sell
25:45for profit or in the market we there
25:47that's recorded in the lsms how many did
25:49you hold back for investment that's
25:51recorded so we add up all the kilograms
25:53multiplied by the price
25:55and then do the same thing for all of
25:57the crops
25:58the lsms does a lot of this for you some
26:00of it we did by hand some of it was
26:02already done and then we subtract off
26:04all the costs that that households
26:08spent including labor costs or
26:10intermediate costs
26:12and so this basically explicitly and
26:14most of the action is here in this home
26:16you know a lot of this would be yeah i
26:17grew this and i and i um and i ate it it
26:19was home production this is trying to
26:21get at the value of all output whether
26:23or not it's um it's it's sold in the
26:26market or kept back for home home
26:28consumption or investment
26:31um in the non-agriculture sector we add
26:33up all self-employed business income
26:34which is just the stated revenues and
26:36self-employed businesses that's actually
26:38where i i worry the most
26:40uh minus the costs that they labor cost
26:43etc
26:44it plus the wage income for
26:46non-agricultural households and the
26:48capital income for non-agricultural
26:50households
26:51and so this is what we do it's it's
26:53painful it took a long time there's lots
26:55of there's lots of arbitrary decisions
26:57that the lsms people made and we had to
26:59mate about
27:00you know units sometimes the units were
27:02a bucket what's a bucket i have no idea
27:05how big that is you have to make an
27:06assumption
27:07mostly is kilograms or something but
27:09occasionally you get strange
27:11quantities
27:12so this is a table that that shows um
27:16i had a question uh
27:18no in the previous slide uh what do you
27:21do with people who are part-time as you
27:24understand
27:26you know um
27:28people are hired when they're like
27:29planting that rope or when they are like
27:31harvesting other things order and
27:34non-agricultural
27:35sector for the united states
27:41in this in this we made we made the we
27:44classified them by their primary uh
27:46sector and so they told us what their
27:49primary sector was but you're right
27:50there's an issue about
27:52splitting across the years so one of
27:54those that's actually a very good
27:56question one of the things that we do in
27:57the paper and i and i didn't show this
27:58in the interest of time is but we look
28:00at hours
28:02for households that say
28:04like i'm i'm an agricultural household
28:06that's my primary sector we looked at
28:08hours worked in agriculture and agro's
28:10work and non-agriculture
28:11and those households do work very little
28:13and at least in the week they asked them
28:15non-agricultural hours the households
28:17who said they were mostly
28:18non-agriculture ended up doing most of
28:20their hours in non-agriculture so i
28:22think that that on average and i think
28:24we looked about three or four countries
28:26for that on average the people who told
28:28us they were in this sector end up
28:29spending the vast majority of their
28:31hours in that sector
28:33thank you
28:35okay may i ask that one question
28:37uh yes
28:38uh about the previous slide
28:42so
28:43this uh data when you're gathered from
28:45households
28:47did they
28:48remember what amount they distributed
28:50among home market or invest or they had
28:53written down their
28:55amounts of kilograms or something okay
28:57so the so you're asking an important
28:59question about recall
29:01recalls are really important um a really
29:04important issue if i say how many hours
29:05have you worked last week maybe you
29:07could give me an answer if i see how
29:08many hours you work
29:10a year ago in a week you know you have
29:12no idea so so typically the way that the
29:15lsms people approach this is they say
29:18people can remember the last harvest
29:21because you know it's like a big event
29:23you know you you so so i don't know my
29:26cousins who are agricultural you say how
29:27many tons do you produce last year you
29:29can remember so this was asked about the
29:31last harvest and so the recall window
29:33there would be if you can remember the
29:35last harvest you can answer this
29:36reasonably but you were asked about the
29:38last harvest
29:39the self-employed income was asked over
29:42the last month
29:43wage income capital income i don't
29:45remember capital income there's not that
29:47much of it the wage income was also the
29:49last month
29:50and it's funny in the us when people
29:52talk about income they talk about annual
29:54income
29:55but in most countries the people are
29:57used to talking about monthly income so
29:59almost every survey i've come across
30:01when i ask people about wage income is
30:03over the last month
30:04yeah so that's the that's the
30:08yeah
30:09yeah so there's lots of measurement
30:10issues here you want my opinion on where
30:12the bodies are buried in these surveys
30:13this is not just my paper um this
30:16not just an answer that the question i i
30:18got
30:19um a second ago this is it's this one
30:21this is where i think the bodies are
30:22buried this is literally a question that
30:24says in your non-agricultural business
30:26what were your revenues last month
30:29and wouldn't you know it when you
30:30compute that and you subtract costs many
30:32people they they lose a lot of money you
30:34know like i forgot what the numbers are
30:36but some enormous fraction of the
30:37businesses on paper by these survey
30:39questions end up losing money
30:42and it's probably because they're under
30:43reporting revenues or maybe they're over
30:45reporting costs but um so i actually in
30:47the end of the day i came out of this
30:48paper
30:49very worried i came into the paper very
30:51worried about agriculture and came out
30:53very worried about non-agricultural
30:55uh outputs anyways i don't know if you
30:58guys are gonna solve that problem but
30:59that's um
31:01that's i think more of a problem than
31:02people realize and this is less of a
31:04problem than people realize
31:07okay so these are the ten countries at
31:09least i think it's ten countries that we
31:11did the
31:12um
31:13our own value-added calculations for and
31:15this this column right here all those
31:17equations on the previous slide
31:18correspond to this thing this is micro
31:20share of value added and in the end you
31:23know i you got value added in
31:25agriculture you got value added to
31:26non-agriculture
31:28the only thing i'm going to do is i'm
31:29going to give you the share of value
31:30added that comes out of this household
31:32survey from from agriculture
31:35and so like um
31:37so like armenia was one where we got it
31:39you know pretty close so we we estimated
31:42that 33 of value added comes from
31:44agriculture and if you just take that
31:46like graphic i showed you at the
31:47beginning of the
31:48the talk it's the 37
31:52we kind of just took one number for the
31:53employment share that this was like this
31:55part of the paper is saying all right
31:56we'll just take the labor input as given
31:59and see whether value added makes a
32:00difference
32:01and so you look down here there's some
32:03differences bulgaria it's in the macro
32:05data it's 11 we get 18. you know you
32:07could say that's very far off code of
32:09arts 32 we get 42.
32:12um
32:13i guess kyrgyz republic we got it kind
32:15of exactly right you know panama 7 it
32:17says 11. but if you look at the levels
32:19you know they're highly correlated
32:21across the 10 countries that we have and
32:23in magnitudes you know they're not you
32:25know they're not that far off
32:27here's the implied agricultural
32:29productivity that you get using this
32:32uh
32:33employment share what kind of the same
32:34for both and then alternatively using
32:36the macro or the micro that we just
32:38calculated value-added shares so like
32:40armenia the gap is 1.1 in our data and
32:430.9 in the macro data
32:45and it's pretty close to 1. bulgaria is
32:481.2 versus 0.7 yeah okay
32:51you know still kind of in the ballpark
32:52of 1.
32:53cote d'ivoire it's 4.7 according to our
32:56calculations it's it's four i mean you
32:58just kind of look through these gone
32:59that goes up a little bit
33:01and our read of this was that overall
33:03you know this is not really solving much
33:05of the problem
33:07um
33:08i guess it takes a little bit off the
33:10edge but you know kind of our our
33:11takeaway was that it's not that it's not
33:14that this is like singularly you see
33:15gaps here but then then once you do this
33:17this micro thing now you now you don't
33:20see the gaps anymore
33:22um in the paper we also look at
33:25consumption expenditures gaps you know
33:27across agriculture and non-agricultural
33:29household and measured income gaps those
33:31are kind of in the same ballpark they're
33:32a little bit lower
33:34you know they're not they're not saying
33:35parity
33:37so anyways that's this paper in a
33:39nutshell um we argued that you can
33:41reduce these gaps from around four to
33:43round two the gaps are kind of largely
33:45present household survey data
33:47and so what we'll turn to next is why
33:49these gaps are so large and whether
33:51there's um whether there's gains from uh
33:53from moving people out of out of
33:55agriculture i kind of say out of
33:57subsistence agriculture because that's
33:58where we
33:59think the you know the least productive
34:01agricultural workers
34:03are
34:04um okay so let's let's talk about some
34:06recent progress in this literature and
34:08i'll just say as a
34:09as a kind of a broader point i one of
34:11the things i'm proud about about this
34:12literature that um
34:14that is that there's it's not united in
34:17methodology it's united over questions
34:19and so there's been people who've who
34:22have people like ted miguel who are kind
34:23of the leaders of micro you know micro
34:25approaches to development and alwyn
34:27young who's um kind of a leader of macro
34:29approaches to development and they've
34:31joined to you know they've joined in
34:33this met in this literature if you will
34:36and and both made contributions i'm just
34:38singling them out as kind of as kind of
34:40some of the leaders of the field but
34:41people from all from all backgrounds
34:43have have contributed
34:45um
34:46so let me just talk i'll have one more
34:48slide on on measurement and then i'll go
34:50to kind of non-measurement issues
34:53one of the things that
34:54you know i think like todd shulman and
34:57bertold herrendorf have emphasized a lot
34:59is that you know you you know
35:02wages are just you know are much more
35:04basic descriptions of people's
35:06livelihoods it's you're asking two
35:08questions how much was your wage income
35:10how much was your hours worked so they
35:12say let's measure gaps in wages and they
35:14argue that they're smaller than gaps and
35:16value added
35:17um that's an eight very nice aj macro
35:19paper and they you know they take a
35:22particular interpretation related to
35:23sorting of workers and and i'll get into
35:25that in a second but they say the gaps
35:27are there in wages but they're uh but
35:28they're smaller the challenge with wages
35:31is you know you don't go to the poorest
35:33countries and find that many wage
35:34workers you know most workers in the
35:36poorest countries are not wage workers
35:37so you end up saying where we can look
35:40at people that have wages the gaps are
35:41smaller but then you you can't look at
35:44as many people and harrendors and
35:45shulman are not ignorant at that point
35:46they um they're probably sick of hearing
35:49it they have some you know they have
35:50some ways of addressing that
35:52they have another paper that's in red
35:54that's very nice that that looks at u.s
35:57agricultural gaps across states not
36:00because we think that you know we need
36:01to solve the problem of why arkansas has
36:03a big productivity gap but but because
36:05they say well look you know the u.s
36:07can't even get the measurement right how
36:08could sub-saharan africa get the
36:10measurement right and they kind of show
36:12substantial under counting of farm
36:14income even in the in the u.s
36:17i guess that's not a measurement point
36:19per se but i think of it as one pulido
36:21and swiki have this very nice paper that
36:23i think it's rnr at aj macro
36:25and they kind of argue you know for a
36:27real poor person in a subsistence
36:29setting it's not clear why we want to
36:31adjust for hours
36:32so they say think about somebody's in a
36:33village they can't work that many hours
36:35they just want the highest income they
36:37can get and you might say well more
36:39hours means foregone leisure and i think
36:41i kind of interpret them as saying
36:43you know how much vil
36:44how much leisure could there really be
36:46in a village so if if your utility is
36:49mostly driven by your consumption at
36:50that stage not your leisure maybe what
36:53we care about is just the pure income
36:55gaps and i thought that's kind of a nice
36:58nice point
37:00remember the early on i said the labor
37:02share is kind of the same in these two
37:04sectors and that model if you have a
37:05higher labor share in the
37:07non-agricultural sector suppose you
37:09think the labor share in non-agriculture
37:11is 0.8 and you think the labor share in
37:14agriculture is 0.4
37:16that would actually give you that could
37:17close the close the gap you know
37:20the problem is i i just you know as you
37:22move from totally agricultural economies
37:25to
37:26almost totally non-agricultural
37:28economies you don't see these massive
37:30movements in the labor share in the
37:31aggregate the labor share in the
37:32aggregate is
37:34you know is kind of two-thirds and yeah
37:35there's some ups and downs um
37:37i this
37:38you probably know gollum's jpe paper
37:40that looks at labor shares across
37:42countries finds they're not that
37:43different this humphreys and weisdorf is
37:45this excellent paper and economic
37:47journal that looks at england's labor
37:48share back to the year 1200 which they
37:51measure painfully from historic data
37:53from um
37:54from domestic servants where everything
37:56was written down and their labor share
37:58has some ups and downs but you could
37:59take a line of two-thirds and just draw
38:01it right through their you know their
38:02labor share
38:06um
38:07okay so
38:08you know i think one of the most
38:10productive
38:11turns this literature has taken is
38:13saying okay well you know we're
38:14interested in gaps but
38:16there's nothing better than looking at
38:17the same person
38:19and following that same person as they
38:21leave agricultural and move into
38:22non-agricultural work and so this
38:24hammery at all paper i i wish i'd
38:27written out all the names say hamrie
38:28hicks
38:29miguel and lee i believe is the they're
38:33author it's in the jea and it um it
38:36looks at some long panels and panel
38:38data's in kenya and indonesia
38:41and it finds there's kind of almost like
38:43a zero return
38:44to migration in those countries so if
38:46you look at wages for people who move
38:48from rural to urban or from
38:50agriculture to non-agriculture it's not
38:52exactly zero but in some cases it's
38:54small positive and can't you can't
38:56reject that it's zero
38:58and so they're kind of saying hey look
39:00there could be these big gaps in the
39:01cross section in their countries there
39:02are
39:03but if people who leave agriculture
39:05don't actually gain you know don't
39:06actually gain um
39:08doesn't doesn't really point to
39:09misallocation it suggests hey maybe this
39:11just were in this world where
39:13you know people who are you know more
39:15productive end up leaving agriculture
39:17and um
39:19and it's not that the average person
39:21could gain much by you know by moving
39:24um alvarez has a similar paper an aj
39:27macro that looks at brazil find
39:28something very similar once you put
39:30individual fixed effects and a
39:31regression of like wages on an urban
39:34dummy
39:35that individual fixed effect soaks up a
39:38lot you get a return close to
39:40close to zero so with um sam marshall
39:43who i think is on the call today uh
39:45mushvik mubarak corey verno and mike
39:48we did kind of a
39:50a look at we did five countries and one
39:52of them was indonesia
39:54um i thought one of them was china
39:56although that adds up to six maybe we
39:58were six countries
39:59and um
40:00we actually found higher returns than
40:03than um than hamri at all did but
40:05they're still nowhere near the
40:07cross-sectional gaps so so averaging
40:09across ghana malawi south africa tanzia
40:12we found i think is actually 23 percent
40:14is the average that we that we found and
40:16again this is not there's no causal you
40:18know there's no causal anything here
40:19this is just observing what happens to
40:21migrants when they move from uh from
40:23rural to urban or from
40:25agriculture to non-agriculture
40:28xiaodongju has this new paper with
40:30co-authors where they look at a policy
40:32reform that induced migration in china
40:35china is an interesting place you know
40:37based surely on its size i guess but
40:39they have very you know very stringent
40:41restrictions on migration and so if kind
40:43of even in a country with very stringent
40:46migration restrictions you still don't
40:48see gains from migration that that does
40:51not look promising for um
40:53you know for the rest of the world but
40:55they actually find huge uh gains for
40:57migration so there was a policy reform
40:58that got um
41:00it sent some larger transfers to rural
41:04um elderly households so there was like
41:06a whole generation of young people who
41:08were staying back in the village to take
41:09care of their elderly folks
41:12and once the elderly folks got these
41:13transfers it launched kind of a large
41:15wave of migration
41:17and there was i think a phased roll out
41:19of the roll into the program
41:21maybe it wasn't phased in maybe it was
41:23you know they were looking at cohorts
41:25right before and after i can't remember
41:26exactly but they estimate a 60 higher
41:29uh
41:30income level
41:32um
41:33in the in cities for the same person and
41:36in villages
41:38so i can kind of kind of obvious to say
41:40that we need more evidence here and uh
41:42and um you know a disadvantage of
41:44observational returns is they're hard to
41:46interpret um
41:48and so the more kind of
41:50efforts you have to have some kind of
41:51exogenous push factor that gets people
41:53out the more
41:55the more i think useful it would be and
41:58i just want to emphasize i think in this
41:59literature
42:00it's not just that we need more numbers
42:02here we need some more insights about
42:04the frictions that keep these gaps large
42:06or keep keep people back in rural areas
42:09so just to suggest if you're going to
42:10head down this path
42:12um try to think about insights about the
42:15the fractions that held them back in the
42:17first place like in chao dong jews paper
42:20this was you're held back by having to
42:21take care of your old folks in the in
42:24the village why don't you just bring the
42:26old folks to the city the answer in in
42:28is probably largely specific to china
42:30which is they don't have a huko in the
42:32city they're not they're not allowed to
42:33be there they have the village hoco they
42:35have to stay there
42:37um
42:41um
42:41so so all these kind of returns
42:46at least get me thinking about
42:48you know who's actually trying to move
42:50and to what extent workers are sorting
42:53into um
42:55sorting into
42:57the cities based on ability so i have
42:59this aer paper that thinks about that in
43:01a kind of a cross-country productivity
43:04and agriculture context i think kind of
43:06the paper that
43:07that carries the torch the most for this
43:09sorting view of the world is alwyn
43:11young's qj paper
43:13and he has these very nice facts that
43:15just say if you look at the people in
43:16the cities of the highest level of
43:18education they're the most likely to
43:20stay in the city the people with the
43:22least education are the most likely to
43:23go back to the village
43:25in rural areas it's the opposite the
43:27people with the highest education are
43:28the more likely to move to cities the
43:30least educated are likely to stay in the
43:31village so he he argues that most of
43:34this gap if not all of it can be
43:36accounted for by just sorting of workers
43:39um there's this nice paper by alvarez
43:41cuadrado a modio and pushki that's more
43:43of a cross-country paper um and um they
43:47argue that they're actually sorting
43:48doesn't work quite the way we think it
43:50does and so in like for example my aer
43:52paper with mike wahl
43:54the sorting works so that the people
43:56with the com the comparative advantage
43:58in agriculture also have an absolute
44:00advantage in agriculture making them
44:02actually good on average in agriculture
44:04alvarez quadrature and modi and porsche
44:06say the data actually support the
44:07opposite where people in agriculture are
44:09systematically
44:11systematically you know less productive
44:13in in agriculture
44:16adamopolis brant late and restuccio look
44:18at
44:19look at the barriers to
44:23well they look at misallocation let's
44:24say in farming but they have this nice
44:26interaction where one of the one of the
44:28things misallocation and farming does is
44:30keep people in the agriculture sector
44:32when they should otherwise
44:34leave
44:35one thing i think that's kind of a
44:36common theme in this literature just
44:38from a you know big picture perspective
44:40is that the more important sorting is
44:42the less room there is for misallocation
44:46and that's kind of at some level obvious
44:48but um
44:49but it's kind of
44:50you know it's like i really like this
44:52paper by garrett brian and melanie
44:53morton that that has a model that
44:55explicitly has sorting in it they
44:57calibrate it to match lots of features
44:59of who's moving and what weight gains
45:01they're getting after moving
45:03and cross-sectional gaps and they kind
45:05of explicitly have a role for sorting
45:07and explicitly a role for
45:09um misallocation and they quantify the
45:11gains from reducing the the frictions
45:14and it's kind of nice because they don't
45:15just say hey there's this gap and it's a
45:16friction they say well some of that
45:18sorting some of that's uh migration
45:20frictions and then they and they nicely
45:22quantify how much is
45:24each
45:26they'd be an example of a paper that
45:28theirs would be an example of a paper
45:29that finds i think a big aggregate
45:31effect of of reducing migration
45:33frictions i think they say gdp would be
45:3420 higher in indonesia if if people
45:37could move freely 20 is to me is huge
45:42um the challenge in this literature and
45:44this is going to be the topic of uh the
45:46ta lecture by martin shu which if i was
45:48taken as next monday
45:50um
45:51martin correct me if i'm wrong on that
45:56but um so that the topic is going to be
45:59how to actually
46:00model sorting
46:02and
46:03how to think about identifying a roy
46:06model and so roy model is kind of like
46:08the classic model of of sorting heckman
46:11an honorary of this paper arguing that
46:13you know if you don't want to make
46:14parametric assumptions about ability
46:16distributions it's
46:18it's really hard to identify a roy model
46:20and so the end most of the literature
46:22ends up making parametric assumptions
46:24and and since my paper which
46:26was on the frontier in 2013 now it's
46:28well inside the frontier there's been a
46:30lot of work and thinking about how you
46:32make kind of just enough structure
46:34uh you put just enough structure on the
46:36problem to be able to identify a kind of
46:39model of sorting
46:40but then not anymore than you need that
46:42imposes then kind of you know functional
46:44forms
46:45it kind of informs your answer more than
46:48the data
46:49it's always the risk when you make um
46:51parametric assumptions so that's the ta
46:53lecture this is i'll just say again my
46:55my lecture today is very low tech it's a
46:57big picture overview if you're at all
46:59interested in these questions i
47:00encourage you to see martin's lecture
47:02because it's going to help get you to
47:03the frontier in terms of tech you know
47:05technically how you'd actually get one
47:07of these models off the ground
47:12okay in the last um
47:15i guess 15 minutes or so of my talk i
47:18think that's how much i have i'm going
47:20to talk about some open research
47:22issues and then we can
47:24do some questions
47:27okay so i came up with six
47:29six topics i think are open issues this
47:31is not all this is not an exhaustive
47:33list some of them relate to stuff that's
47:34going to be covered in other stag
47:35lectures
47:36i and i thought i would spend a slide or
47:38two on on each of these so one of the
47:40spatial price differences
47:42seasonal migration
47:44which is kind of possibly a you know
47:47you know a gateway i guess to more
47:48permanent migration
47:50intergenerational effects of migration
47:53imperfect information about these gaps
47:55like i know about these gaps now you
47:57know about these gaps do villagers know
47:58about these guys if they don't know
47:59about the gaps there's no puzzle about
48:02why they're not you know moving to
48:03arbitrage them
48:05um something that's
48:07understudied here is household migration
48:09decisions rather than
48:10kind of individual decisions and then
48:12finally i'll talk about frictions and
48:14land markets and how that interacts with
48:16this literature on productivity gaps
48:20so let me talk about spatial price
48:22differences this always comes up when i
48:23think about
48:24these gaps and when i present work
48:26related to these gaps
48:28um you know maybe the city is a place
48:30where you're more productive and your
48:31wages are higher but that's like just
48:33more expensive to live there um
48:36so chen
48:38revelion and sangrelia have this paper
48:40where they look at data i'm pretty sure
48:42these are data collected by the world
48:43bank
48:44and um
48:45and they were i think i spelled
48:48revealing wrong whoops sorry about that
48:50but they they um they're trying to come
48:52up with an urban rural cost of living
48:54differential
48:56and so what they're trying to come up
48:57with is the cost of a basket of somebody
49:00earning a dollar a day and thankfully
49:03there's less of these people in the
49:04world i think they do two dollars a day
49:05too what would a basket look like if you
49:07earned a dollar a day a lot of food
49:10um you know i don't know what the
49:12weights would be must be you know 75
49:14percent food or more two dollars a day
49:16it's a little bit less food you get some
49:17housing you've got some other stuff so
49:19that's what they're that's what they're
49:20trying to to do
49:22and this is um
49:24uh for all the data they all all all of
49:26whom are developing countries this is
49:28the ratio of the urban
49:30to rural
49:31uh basket that they they compute so one
49:35is the basket's just as expensive in the
49:37in the urban as a rural area 1.38 i
49:40think is the median
49:41was the um
49:4338 higher cost of the same basket
49:46so if you're trying to explain a gap of
49:48two and you think hey it's you know you
49:49earn twice as much in a non-agricultural
49:51job but you then you have to live in a
49:53city and it's you know suppose you're
49:55over one of these countries it's 80
49:56percent more for the same basket
49:58that could be relevant for you that
50:00could kind of knock out that gap for you
50:01and you say i'm kind of a different
50:03between earning twice as much and
50:05and paying twice as much
50:07you know there's something puzzling in
50:08these data if you look and it's not you
50:10know
50:11every data have their issues there's a
50:13it's a whole lot of countries that have
50:14like exactly 1.38
50:17and when i was emailing with helen young
50:19about this at some earlier point he says
50:21you know you hate to be too critical but
50:23it seems like a real coincidence that
50:25half the you know half the countries or
50:26something have exactly 1.38 times as
50:29expensive urban as rural goods
50:31and so it's hard to know what went on
50:33behind the scenes but if you kind of
50:35just take the median you know the kind
50:36of the median country 1.38 that does get
50:39you that does get you in the direction
50:40of explaining a gap of four you know it
50:42gets you 40 percent of the way there
50:45um
50:45still given you know given the qualities
50:48data it's you know it's hard to know
50:49whether that's that's really our answer
50:53but actually there's kind of a bunch of
50:54open issues here um and so there's this
50:56great paper that's in re-stud by hanbury
50:59jesse hanbury and and uh david weinstein
51:02that that makes the point in the u.s
51:04where we have better data that is yeah
51:06there's more goods actually available in
51:08u.s cities
51:09it's not just a matter of
51:11um having the same basket cost more in a
51:14city you know there's things that you
51:15can get in a city services you can get
51:17that you just can't get in a rural area
51:20is this true in low-income countries it
51:22certainly seems likely but i don't know
51:24of any like concrete evidence to this
51:26effect to poke fun at the
51:29what i just showed you on the previous
51:30slide imagine you're gonna go to the
51:32movie theater in the city
51:33and then you say well in the village i
51:35go to the movie theater in the village
51:37and it's just
51:38it's uh cheaper
51:40you're not gonna have a movie theater in
51:41the village
51:42and so this actually raises a bunch of
51:44questions i find interesting anyways
51:46that that kind of touch on the urban
51:47literature about what is the appropriate
51:49spatial price index that we should use
51:51when there's some goods that just aren't
51:53available in the
51:55you know in the rural area
51:57um and so i can imagine a project or
51:59projects that think about taking this
52:00hanbury and weinstein availability of
52:03good not just prices of goods but
52:05availability of goods
52:07uh to a low income study and they
52:08actually have some nice progress they've
52:10made on how to think about spatial price
52:12indices in these
52:14on these settings
52:16another thing too i i think that's
52:18related here is that is that migration
52:20strategies are complicated
52:22you know i'm so like suppose your
52:24household is in a village and you're
52:25going to send a migrant so you sound
52:27like you know a head of household or an
52:29adult you know adult child
52:31well
52:32probably they're not going to move into
52:33the same house you'd have in the village
52:35they're going to kind of live on ramen
52:36if you will they're going to have a you
52:37know kind of a
52:40goal of not spending a lot of money and
52:41then maybe later you bring the family
52:44and so if that's true you know you don't
52:46want to just say hey what would my exact
52:48same basket cost in the city you'd want
52:50to think about exactly what am i going
52:52to spend if i if i
52:54you know move to the city i'm just
52:55thinking about it like just
52:57just about some more mate
52:59wait in the bones here i meet on the
53:00bones of this this argument i had like
53:02an
53:04an uncle i lived in west suburban boston
53:07he lost his job he
53:10took a temporary job in houston texas
53:13you know houston texas is cheaper than
53:15west suburban boston but that was kind
53:16of irrelevant for his joint because
53:18rather than living in a big house he
53:19lived in a small apartment and his
53:21consumption basket was totally different
53:23than it would have been if the whole
53:24family came so this to me seems like
53:27something as simple as what we did in
53:29the previous slide we're just going to
53:30take the same basket and say it's
53:31cheaper just seems like that's not you
53:33know that's not kind of the right way to
53:35think about
53:36price differences
53:40seasonal migration is a topic that there
53:43had definitely has been some work and
53:44there should be some more work so
53:47mushfix gonna talk mushrik mubarak in uh
53:50a
53:51couple of weeks i think if i'm not
53:52mistaken
53:54um he has this great paper in
53:55econometrica where he does a migration
53:57experiment and along with shaimal
53:59chowdhury and gareth bryan
54:01and they basically they have this
54:03headline finding that there's the 30
54:05consumption gain to migration during a
54:08lean season and so uh this paper is this
54:11paper set in bangladesh this experiment
54:14set in bangladesh and like many
54:15countries there's a lean season where
54:17kind of like the agricultural crop
54:19um you know has been planted but it
54:21hasn't been harvested yet and so yeah
54:24there's not that much to do in in a
54:26village and so by
54:28by giving people a small subsidy
54:31they uh like the equivalent of a couple
54:32weeks wages they got people to go to
54:34jobs largely in the city and then these
54:36people had 30 percent higher consumption
54:39um
54:40a couple years later most of them were
54:41kind of not sending migrants again which
54:44was like you know a little bit of a fly
54:46in the ointment hey 30 higher
54:47consumption it's like this
54:48transformative thing well but a few
54:50years later they're you know they're not
54:52migrating again
54:54um and so this paper that mike uh
54:57and mushvik and i wrote we
54:59have a structural model to interpret the
55:02experiment
55:03in the spirit of um joe's paper with rob
55:06townsend
55:07and we basically end up concluding that
55:10seasonal migration is valuable not
55:12because we want to get guys permanently
55:14out of the village you know forever to
55:16the non-agriculture sector but it's it's
55:18an insurance mechanism
55:21like in melanie morton's job market
55:23paper which is in the gpe and so the way
55:25to think about this is if you know if
55:26there's no opportunities in the village
55:28and your assets are really low migration
55:31is valuable and so temporarily you say
55:33let me go to the city now i'm in a tight
55:35spot let me go let me let me work and at
55:37this more productive area
55:39and then once your
55:40opportunities are good again in the
55:42village or your savings are up now
55:44you're you're kind of happy in the
55:45village you don't you don't need to
55:46migrate um and that's kind of in
55:49contrast to a different story originally
55:51pushed by uh brian at all after after
55:54they ran their experiment which is
55:56um that hey there's these workers are
55:57stuck in rural areas due to credit
55:59constraints
56:00and then if you give them a couple weeks
56:02wages then they go to the city they earn
56:03these much higher incomes
56:05and in the end that story just doesn't
56:07hold up a lot of um it just doesn't hold
56:10up um it's hard to kind of get that
56:12story to work in a structural model a
56:14big piece of the story is if if they're
56:16stuck how come they you know how come
56:18they kind of come back and are still in
56:19the village a few years later the data
56:21are much more consistent with this idea
56:22that it's valuable to be able to
56:24temporarily get to this more productive
56:26place
56:27um even if you're kind of not going to
56:28permanently go there
56:31of course a bunch of stuff happens when
56:32you have seasonal migration and this
56:34akram
56:35chow hurry and mubarak follow-up
56:37experiment did kind of larger the larger
56:41migration experiment i put it they did
56:43they randomized the number of people in
56:44in villages that would get um
56:47subsidies for migration and they and
56:49they actually found that
56:50if you had more out migration there was
56:52higher rural wages
56:55and so you just think about you know
56:56people talk about pecuniary
56:58externalities this is i i think of a
57:00pecuniary extraction so suppose we're in
57:02this village and i decide to leave
57:04that's actually going to help
57:06raise wages for the other people in the
57:08village so that actually you know that
57:10actually matters for other people it's
57:11not just that i do what i want on my own
57:14prices for for other people and if
57:17people are poor and um
57:19you know and and these are tough times
57:21raising their wages actually is actually
57:23good for them
57:24so there's a bunch of open just a
57:26question on that
57:28yeah felix um
57:30so how do you think about it um
57:33the fact that they're not always
57:35migrating so you said well it's valuable
57:37as temporary insurance it's not always
57:39valuable
57:40is it because like a migrant's lifestyle
57:42is not particularly appealing to them
57:45because if it's like like why don't they
57:46go more often of course like it's good
57:48to have insurance if it's like um i
57:51don't know i guess a bad time to be in
57:52the village but also it's like 30 more
57:55is like always pretty good
57:57yeah
57:58um so
57:59so it's a really good question the short
58:01answer is that
58:03you kind of have to think about who's
58:04going to get the 30 so it's 30 when
58:07you're kind of in a bad spot
58:10it isn't that in any given year like
58:12suppose it's a bumper harvest and you
58:13went to the city the our model tells you
58:15actually you could lose money by going
58:17to the city
58:18so in the end it's a combination of kind
58:20of you are permanently not that
58:22productive in the city that's why you
58:23don't want to kind of be based in the
58:25city
58:26but then you're migrating systematically
58:28when you're kind of in a tight spot
58:29where you'd have these big gains from
58:31moving
58:32so i hope hopefully that that
58:34quite that answers your your very good
58:36question
58:38okay thanks so just to clarify like the
58:41ideas that there's like some that
58:42basically the people who are going those
58:44are like the unproductive people
58:46so that's not the gain that's available
58:48to everyone as well yeah yeah in the
58:51model there's there's no arbitrage the
58:52model does have a big non-monetary
58:54disutility of moving which is like
58:57you know you got to find a house you
58:59could have a very you know very lousy
59:01housing opportunity you you know you
59:03have to find a job might take you a
59:04while to find one so there's like
59:06reasons why you don't want to go and and
59:08so you're going to wait until a time
59:09where where there's a big gain for you
59:11so that so that it's worth your while
59:13yeah in a model where it was costless to
59:15move you'd see lots of moves you you
59:17know you just see a lot of counter
59:18factual things that we kind of don't
59:20don't see
59:22all right thanks a lot yep thanks good
59:24good questions
59:26um
59:27let me try to wrap up um i let me just
59:30say like just close out seasonal
59:32migration
59:33there's a small
59:35set of papers of which this clemens and
59:36magruder is a good example thinking
59:38about negative effects on urban areas
59:40you know if you have a whole bunch of
59:41migrants flood an urban area
59:44does that lower their wages there and
59:46clemens and magruder use like weather
59:48shocks
59:49and they say okay well you know bad
59:51weather lots of migrants are going to go
59:52they actually find that in destinations
59:55um they get more migrants wages and
59:57unemployment's actually lower so this is
59:58some sign that at least in the short run
1:00:00there can be negative impacts on the on
1:00:02the kind of the destination areas
1:00:05you know in my head than
1:00:07you know thinking more about permanent
1:00:08migration city you know cities can
1:00:10expand um
1:00:13you know i know there's kind of only one
1:00:15you know central park or something um in
1:00:17some cities there's kind of one unique
1:00:19part but most of the developing world
1:00:21cities i i'm familiar with you know you
1:00:23there's lots of room to expand on the
1:00:25extensive margin and add kind of more of
1:00:27the same and so but i this is just my
1:00:29introspection i think that there'd be
1:00:32you know as african cities have expanded
1:00:34as indian cities have expanded over the
1:00:36last 50 years have they been you know
1:00:38driving down wages i'm skeptical that
1:00:40they have been i think that cities have
1:00:42just expanded and absorbed migrants and
1:00:44but i this is this is clearly an area we
1:00:47need more
1:00:48uh work
1:00:49um let me just talk about policy for a
1:00:51second here this is a policy issue
1:00:53politicians are definitely worried about
1:00:55the cities you know being flooded with
1:00:57migrants
1:00:58and so like this great paper by clemont
1:01:00and barrett and um co-author pap i i
1:01:03don't know the first name there they
1:01:05they show that india has this has this
1:01:07work fair program where um
1:01:10they um they kind of pay you to stay in
1:01:13a rural area and they give you some job
1:01:15which you know maybe you know helping
1:01:16build a road or something
1:01:18and um and it actually
1:01:20reduces migration substantially
1:01:23so this is like as a matter of fact
1:01:24policy does have an impact on seasonal
1:01:26migration and and if you want to slow it
1:01:29um subsidizing it can definitely speed
1:01:31it up and i kind of just think that
1:01:32there's almost like
1:01:34a shortage of papers thinking about
1:01:36social planner problems so if you want
1:01:38to take more of a quantitative
1:01:40modeling perspective thinking about what
1:01:42the planner would do
1:01:44and how the planner would deal with
1:01:45these various externalities these
1:01:47pecuniary externalities through wages or
1:01:49other channels you know crime or things
1:01:51like that um
1:01:53it's it's just to me in my head not not
1:01:55totally clear
1:01:57you know it's good for somebody if they
1:01:59move it could be bad for somebody else
1:02:01and so there's kind of once you think
1:02:02about externalities um you kind of think
1:02:05a planner's problem can be
1:02:07useful so this
1:02:08this is um
1:02:10something that from a quantitative
1:02:11modeling perspective i think could be
1:02:14valuable
1:02:15intergenerational effects um there's
1:02:18this great paper by nakamura sigurdsson
1:02:20of steinson that's um i think it's rnr
1:02:23or maybe forthcoming in review of
1:02:24economic studies where they look at this
1:02:26experiment and this natural experiment
1:02:28where a volcano erupted in iceland and
1:02:31melted kind of half a town and there's
1:02:34kind of these maps where like half the
1:02:36town was kind of wiped out the other
1:02:37wasn't
1:02:38and so there's this some people were
1:02:40kind of
1:02:42you know forced to leave and they ended
1:02:44up going to mostly reykjavik the bigger
1:02:46city
1:02:47and so they've got the data going back
1:02:49to the 70s when this was and they
1:02:51actually got data on not just the
1:02:52migrants but their kids
1:02:54and the migrant adults and the people
1:02:56who kind of had a house
1:02:58you know destroyed by lava they they
1:02:59didn't actually look that different than
1:03:00the house the adults that stayed back in
1:03:03the village
1:03:04it was the kids who were like very
1:03:06different so the kids
1:03:08who ended up being forced to migrate
1:03:10ended up with much more schooling i
1:03:12forgot how much it is like three years
1:03:13of school or something on average and
1:03:15much higher lifetime earnings partly due
1:03:18to the schooling and maybe partly due to
1:03:20other things and so i thought this was
1:03:22kind of you know this was
1:03:24you know who cares about iceland well
1:03:25iceland's a good place to learn about
1:03:27you know to learn about the world i
1:03:28think it's not that we care about
1:03:30iceland per se but i thought just the
1:03:32fact that you can track not only these
1:03:34people forced to migrants but their kids
1:03:36really kind of thinks about um
1:03:38you know the the intergenerational
1:03:40effects of migration and so like i think
1:03:43there's opportunities in this literature
1:03:44to think about interactions
1:03:46between the kind of migration and
1:03:48productivity gaps literature and the
1:03:49literature on investments and children
1:03:52um i do think that like i'm just going
1:03:54to say it here it's you know better
1:03:56opportunities for kids is a central
1:03:58rationale for migration
1:04:00i'm just asserting that without any
1:04:02evidence i think it would actually be
1:04:04nice to have some direct evidence on
1:04:05that like you could imagine surveying
1:04:07migrants of all the fossil you know let
1:04:09me let me list a bunch of
1:04:11factors that may have influenced your
1:04:13decision to migrate you tell me whether
1:04:15they're somewhat important very
1:04:16important not important i can imagine
1:04:19the survey like that i bet you better
1:04:21opportunities for kids would show up as
1:04:23important for a lot of migrants and so
1:04:25even if they're only marginally better
1:04:27they think hey now that my kids get
1:04:28schooling they get the opportunities the
1:04:30city can afford that you know that
1:04:33that's kind of absent from the you
1:04:35remember the simple model i showed you
1:04:36to motivate my paper hey there's just a
1:04:38wage here and a wage there you pick and
1:04:40pick the one that's higher
1:04:41it's a pretty style a stylized
1:04:43description of how you choose where you
1:04:45want to go
1:04:46and this kind of intergenerational
1:04:48aspect seems like it's something that
1:04:50could be explored more
1:04:53imperfect information is really
1:04:55important we're only kind of getting
1:04:56started
1:04:57in this literature that
1:04:59the brian chowdhury mubarak
1:05:02migration experiment in bangladesh they
1:05:04did this information treatment
1:05:06where they sat people down they told
1:05:08them about the jobs in the city how much
1:05:10they earned and it had a precise zero
1:05:12impact on migration
1:05:14mushvik kind of tells me he doesn't
1:05:16think that he's a place where even in
1:05:18the kind of the without any incentives a
1:05:20third of households are sending migrants
1:05:22he thinks this is
1:05:23you know these villages have lots of
1:05:25you have lots of migrants coming and
1:05:27going so there's there's good
1:05:28information
1:05:30travis basler looks at western kenya
1:05:33in a place that yeah there's migrants
1:05:34but i get the impression that's a lot
1:05:36less than in this bangladeshi setting
1:05:40and basically they you know like
1:05:42travis interviews these rural kenyans
1:05:44and says well what do you think they get
1:05:45in the city and they have they're way
1:05:47off they
1:05:49they systematically under report what
1:05:50people get in the city
1:05:52um
1:05:54and yeah it
1:05:55so he has this theory which he
1:05:57substantiates with it with an experiment
1:05:59which is that migrants go to the city
1:06:01and then they come back and the
1:06:02villagers say how are things going in
1:06:04the city they say ah
1:06:05not so good and they and they under they
1:06:07under report how much they've been
1:06:09earning and that's why people are
1:06:11uninformed
1:06:12and why would you want to under report
1:06:14well because once people find out you've
1:06:16got money then somebody knocks on your
1:06:17door they have a sick baby that you know
1:06:19they're looking to invest in something
1:06:21they need some money to pay back
1:06:22somebody they're they're hungry so once
1:06:24you've got money in a village there's
1:06:26these risk sharing kind of arrangements
1:06:27that are formal or informal i guess
1:06:29mostly informal but
1:06:31you don't want to be the one in the
1:06:32village who's got who's got money and so
1:06:34i thought that was kind of not only
1:06:36interesting that these guys
1:06:37systematically under
1:06:38predict the gap but um when you tell
1:06:41them about the gap they actually move
1:06:43more and so there's like experimental
1:06:45evidence that information can actually
1:06:47get them to move more
1:06:49and um and then there's kind of a nice
1:06:50theory about why there this there'd be
1:06:53this imperfect information to begin with
1:06:55and there's kind of like an obvious need
1:06:57for more evidence here
1:06:59um you know two countries that are very
1:07:01different the settings are very
1:07:02different it's hard to know whether this
1:07:04is a
1:07:05this is something that's systematic or
1:07:07maybe you know maybe this is the only
1:07:09place in the world where people are
1:07:10informed maybe this is the only place
1:07:11where people are uninformed we've kind
1:07:13of like no idea
1:07:16so i guess the last
1:07:18maybe this isn't the last
1:07:19household migration issues i'm running
1:07:21out of time let me let me let me kind of
1:07:23skip this one um i think we need better
1:07:26models of
1:07:27household migration decisions
1:07:29i'll just mention one last one which is
1:07:32frictions and land markets and then i'll
1:07:33stop
1:07:34um
1:07:36you know there's a kind of a growing
1:07:38literature on how the inability to
1:07:40actually sell land
1:07:42is important for understanding why
1:07:44agriculture is so unproductive and so
1:07:47diego restuccia and toso adamopoulos and
1:07:49then raoul santoleo lopez ciaran chan
1:07:52and diego stuccia
1:07:54are just two of the several other papers
1:07:56i could have cited here that that think
1:07:58about how if kind of you've got
1:07:59productive farmers that are stuck with
1:08:01small amounts of land
1:08:03and um unproductive farmers who've
1:08:05somehow inherited a lot of it you know
1:08:07this can be bad for productivity if
1:08:09these guys can't trade you know in a
1:08:12kind of a perfect world the unproductive
1:08:13guys would
1:08:14find it more profitable to
1:08:17rent off some of their land to the more
1:08:18productive person who can manage a
1:08:20bigger farm
1:08:21and um and the fact that they rentals or
1:08:25sales of land are not allowed and
1:08:26hinders productivity
1:08:29this uh gottlieb um
1:08:32grab of sac paper charles gottlieb and
1:08:34yen grab offset and charon chan have
1:08:36papers that talk about how if you if you
1:08:38were to have better functioning labor uh
1:08:42land markets
1:08:43you'd get large productivity gains in
1:08:46part from worker real locations so in
1:08:48part because workers would leave the
1:08:50leave the agricultural sector
1:08:53i'll just conclude by mentioning this
1:08:55one nice
1:08:56kind of i'd say quasi-experimental
1:08:58evidence from mexico there was this land
1:09:00reform where they gave titles to farmers
1:09:03who didn't have
1:09:05land titles before and so they were kind
1:09:07of holding back on their land because if
1:09:09you leave and you don't have a title
1:09:10like somebody else could take it and
1:09:11then you know that you're in trouble
1:09:14um so this is papers by alan de jean
1:09:15very kyle emmerich
1:09:17gonzalez navarro and satellite
1:09:20this is in the aer and it um it shows
1:09:22that once you give people titles they
1:09:24send a lot more migrants out
1:09:27and to me i just thought that's that's
1:09:28kind of exciting exciting as a
1:09:30possibility that hey maybe it's you know
1:09:32people are stuck in low productivity
1:09:34stuff because your main asset is this
1:09:35land but you don't have a title to it
1:09:37and it's kind of use it or lose it and
1:09:39so um so this mexico
1:09:42study shows that kind of
1:09:43a pretty concrete um barrier to
1:09:46migration out of agriculture
1:09:49so that's my talk there's some large
1:09:51productivity gaps in developing
1:09:53countries
1:09:54there's been a lot of progress
1:09:56in the last let's say decade
1:09:58many open questions i think kind of the
1:10:00big picture question that this
1:10:02literature's thinking about are what are
1:10:03the what are the frictions that keep
1:10:05people stuck in low
1:10:07productivity agriculture work or is it
1:10:09not frictions is it just sorting and you
1:10:12know there's some productive people
1:10:14and maybe you just make productive
1:10:16people by getting them educated and then
1:10:17once they're educated they leave maybe
1:10:20maybe other than education there's just
1:10:22sorting sufficient
1:10:24what if anything should policy makers do
1:10:26to get households out of subsistence
1:10:27farming should they just do laissez
1:10:29faire should they be titling the land um
1:10:32what should they be doing these are it's
1:10:34not that we've got no no ideas about any
1:10:36of these questions but i i think we're
1:10:38um there's lots of room for work
1:10:40research papers addressing these these
1:10:43questions
1:10:44so that's it so i guess i'll use my
1:10:46remaining um
1:10:49i forgot how much time i'm allowed but
1:10:51my reigning time to to answer questions
1:10:54how much time do i um do i have left
1:10:57how much time do we have questions
1:10:59what's that it's about 18 minutes okay
1:11:02perfect let me hand the moderating
1:11:04duties over to you since you're the
1:11:06most versed in this uh literature anyway
1:11:10it's me so i'm the moderator okay
1:11:13no i was saying to doug
1:11:15i'm happy
1:11:17to do it i don't know how much
1:11:18moderating
1:11:19there's gonna be but let's let's see
1:11:21what we've got for questions let's open
1:11:23it up first to the front row
1:11:28hi david sorry another question um on
1:11:31maybe the interaction between frictions
1:11:34and
1:11:35sorting so i guess if it
1:11:37i don't know if i'm thinking about a
1:11:39world where basically
1:11:40i'm a firm and i can't really expand
1:11:43because
1:11:44whatever like there are many stories for
1:11:46example i can't i can't hire people
1:11:48outside of the family could that also
1:11:50like influence the kind of sorting that
1:11:52you're observing but basically the most
1:11:54productive people end up leaving because
1:11:56maybe delegation is impossible
1:12:02yeah i mean
1:12:03yeah i'm
1:12:05so you you okay so let me give you this
1:12:07answer i you you're making me think
1:12:08about a reason why
1:12:11i might have a real productive
1:12:13non-agricultural family business
1:12:16and there could be some farmer in the
1:12:17neighboring village who's much less
1:12:19productive
1:12:20why at the margin wouldn't we send the
1:12:22you know villager say he's a you know
1:12:25productive village he's educated he just
1:12:26doesn't know anyone why doesn't he go
1:12:28and be the
1:12:30you know be the next employee at this
1:12:32the next manager at this you know this
1:12:34manufacturing establishment
1:12:35non-agriculture
1:12:37so you pointed to a friction which is
1:12:40the
1:12:41you know mo problems with monitoring
1:12:42problems with contract enforcement that
1:12:45makes it hard for people to
1:12:48you know add managers and add other
1:12:50workers that are outside of their family
1:12:52so maybe that's not exactly what you
1:12:54asked me but to me that i haven't seen
1:12:56anyone who's thought about that
1:12:57interaction between
1:12:59um
1:13:00between kind of let's just say between
1:13:03migration and
1:13:05barriers to hire people at formal firms
1:13:08um
1:13:09i don't know doug does that i'm gonna
1:13:11turn it over does that you've thought
1:13:13about these issues as much as i have is
1:13:14does that sound to you like something
1:13:16that's already been done i think that's
1:13:18kind of new
1:13:20i mean i think there's quite a lot of
1:13:21literature asking whether household
1:13:23labor family labor is substitutable
1:13:26with higher labor and agriculture and
1:13:28consistently you find that they're not
1:13:30very closely substitutable because of
1:13:32monitoring issues and
1:13:34my usual story there is that agriculture
1:13:36takes place in space unlike a factory
1:13:39where you you can watch the people
1:13:40working for you
1:13:42hover over them pretty closely you're
1:13:44sending people off to far corners of
1:13:47fields it's not so easy to see what
1:13:49they're doing
1:13:50what that then implies for how you
1:13:51allocate labor
1:13:53within the household
1:13:55how labor markets work i think i think
1:13:57we haven't seen much that tries to model
1:13:59that in general equilibrium
1:14:00and and so i i think that is a
1:14:03it's an interesting question
1:14:05so michael peters
1:14:08and harun alp i believe is the third
1:14:10author of a paper that think about
1:14:11delegating to managers managerial
1:14:13delegation frictions
1:14:15you know it's a
1:14:17it's a very good paper it's a bit of a
1:14:18stylized
1:14:20paper in some ways and um and they use
1:14:23this nick bloom
1:14:25aprijit mahajan i'm forgetting the other
1:14:27co-authors experiment in india where
1:14:30they do managerial incentives
1:14:33so it's it strikes me as somewhat
1:14:35related because it's a ge
1:14:37look at managerial frictions but it
1:14:39doesn't think about the interactions
1:14:40between
1:14:42you know getting people out of low
1:14:44productivity sectors and why they can't
1:14:46you know why they can't
1:14:47it's like a
1:14:49it's like a factor from the
1:14:51the non-agricultural side that keeps
1:14:53people out
1:14:54rather than uh you know something that
1:14:55holds them back in the village
1:14:58so yeah
1:15:00david i've got a couple of questions for
1:15:01you one
1:15:03um one in the
1:15:04q a from renee diaz de leon who says is
1:15:08there any evidence about the
1:15:09effectiveness of migration policies in
1:15:11developed countries in the early 20th
1:15:13century
1:15:14were there any large-scale successful
1:15:16policies
1:15:17um how about china so so that's
1:15:21that's one
1:15:22um let me give you just a couple and you
1:15:24can
1:15:25allocate your time how you want
1:15:27um corey verno has a question and again
1:15:30apologies if i'm mispronouncing people's
1:15:31names what do you think are the biggest
1:15:33data limitations holding back the
1:15:35literature right now
1:15:37so a question on data
1:15:39and then a question
1:15:41on
1:15:41on risk
1:15:42basically
1:15:44is there a role for food security and
1:15:46this is from john spray who i've given
1:15:49him an answer in the in the q a box but
1:15:52i thought you might have a different
1:15:53answer
1:15:54um
1:15:55our people may be willing to accept
1:15:56lower wages in agriculture
1:15:58if
1:16:00it means they have enough food in the
1:16:02event of a negative shock
1:16:03so that's three things one is migration
1:16:06policies one is data limitations
1:16:09and one is the role of risk
1:16:13so okay so i'll go and order it so the
1:16:15so renee's question um
1:16:19it's a really good question i don't know
1:16:20the answer i don't know of any studies
1:16:23that have thought about the
1:16:24effectiveness of migration policies um
1:16:28it's not completely obvious how you'd
1:16:29measure effectiveness one of them i
1:16:31guess would be you know if you um
1:16:34you know if you
1:16:36subsidize it do you get people to go if
1:16:38you
1:16:39if you um
1:16:42you know if you constrain it do you hold
1:16:44it back i i'm sure there's evidence that
1:16:46the huko policy is um
1:16:50is um as constrained migration in china
1:16:52i think there's no question about that
1:16:54yuan tien who's at nottingham has kind
1:16:57of an interesting paper that that talks
1:16:59about
1:17:00um
1:17:01like when there is a favorable terms of
1:17:04trade shock to a chinese region or city
1:17:07that
1:17:08um
1:17:10that that needs cheap migrant labor then
1:17:12they actually kind of
1:17:14retract their policies so she goes
1:17:17through like the text of all the laws
1:17:18and you can see that
1:17:20when they need more migrants they they
1:17:22kind of soften the laws and they get
1:17:25more migrants so i guess that's an
1:17:27example of a
1:17:29china
1:17:30policy that where you say hey in the
1:17:32event that we actually have a high
1:17:34demand for low-skilled labor from rural
1:17:36areas we change our policies to let our
1:17:38migrants in
1:17:39the history thing is really interesting
1:17:41i i don't know i think it's
1:17:43yeah i think it's a i think it's a nice
1:17:46a nice topic for future research
1:17:48potentially
1:17:49um
1:17:50the second
1:17:52question i boy i make i may have gotten
1:17:55the order wrong but there was uh there
1:17:56was one about risk maybe that was the
1:17:58third question let me handle that one so
1:17:59i intentionally did not talk about risk
1:18:01because mushfix giving a lecture if i'm
1:18:03not mistaken about risk and migration so
1:18:06it's not that i don't think that's an
1:18:07important topic it was more i just
1:18:08didn't want to
1:18:11you know double up on material there's
1:18:13this real old idea the harrison tadaro
1:18:15idea that you know it's you know
1:18:17villages may not be um real high paying
1:18:20places but you know at least you can you
1:18:22know at least you can eat and if you go
1:18:24to the city there's unemployment
1:18:26and so the the harrison tadaro story is
1:18:30which i think is a quite a sensible one
1:18:31is that there's risk in the city risk
1:18:33and migration
1:18:34and um
1:18:36and then you're kind of willing to
1:18:37accept this lower wage in the village in
1:18:39exchange for
1:18:42in exchange for less variance in your
1:18:44income i would say in recent years the
1:18:46evidence seems to be anything kind of
1:18:48almost the opposite of that and so if
1:18:50you take
1:18:51the munchie rosenzweig paper with which
1:18:53i'm sure
1:18:54um
1:18:55bushfig's gonna cover in his lecture
1:18:57they actually argue that because you
1:18:59have these village-rish
1:19:01uh you have these um
1:19:04yeah these
1:19:05village risk sharing arrangements kind
1:19:07of i think they they think it's like you
1:19:09know you need this because the village
1:19:11is actually a very risky place and so if
1:19:14you're in this village um
1:19:16you you can insure yourself when you're
1:19:18when you're kind of having a bad time
1:19:20but there's all this risk you have to
1:19:21worry about um if you get an aggregate
1:19:23shock to the village it's not sure it's
1:19:25not really clear how those
1:19:27villagers share risk like if there's a
1:19:30you know a terrible flood or a terrible
1:19:31drought or something you know all the
1:19:33all the good will between villagers in
1:19:35the world isn't going to help the
1:19:37you know people mature each other you're
1:19:39kind of it's hard to ensure against that
1:19:41and so that i think i think the view of
1:19:44munching roses drag is that the village
1:19:46is actually quite a risky place and that
1:19:48and and the city you know
1:19:50the city is actually a place that's
1:19:52that's not as risky in that experimental
1:19:54evidence of of brian choudhury and
1:19:56mubarak kind of an interesting thing is
1:19:58if you look at the variance of
1:20:00consumption
1:20:01so not income but consumption for people
1:20:04who migrate and people who don't
1:20:06it looks pretty much exactly the same
1:20:09um and so it kind of there's this
1:20:10consumption of the the non-migrants and
1:20:13there's this consumption of the migrants
1:20:14and the migrants are just shifted over
1:20:15by 30 percent basically and i thought
1:20:18that was pretty interesting because you
1:20:20just you know you you just didn't there
1:20:22was just we cut the data a bunch of ways
1:20:24and we just didn't see
1:20:26we didn't see that much evidence of um
1:20:29of higher risk for the people who had
1:20:31migrated um and so it's not completely
1:20:34clear what's driving that so i think
1:20:36that yeah it's not as open and shut that
1:20:38that that's a that that's a good story i
1:20:40think it's yeah empirically is the
1:20:42villages it's kind of a non-risky place
1:20:44i think that's that's kind of there's
1:20:46people who would dispute that i i do
1:20:48think that idea by harrison tadaras has
1:20:50a lot of a long
1:20:52intellectual history and
1:20:54mushvik i think will have more sensible
1:20:56things to me than say about that
1:20:59so corey um hi corey my co-author corey
1:21:02asked me about what data do we need um
1:21:05yeah i don't i don't
1:21:07i don't know um i don't know the answer
1:21:09that question i there's lots of
1:21:10different you know there's lots of
1:21:12different um types of data that we would
1:21:15need that would be not amenable to
1:21:17you know cross-country kind of data
1:21:19collection projects one of them would be
1:21:21like i gave the example of info you know
1:21:23information
1:21:24um
1:21:25you know
1:21:26so i think that that you know
1:21:29does um
1:21:32are people well informed about gaps
1:21:34around their country that that's
1:21:36something that you know you could do
1:21:37experiments but there may be a role for
1:21:39just obser you know like survey
1:21:41informations where we do surveys in a
1:21:42bunch of different countries and just do
1:21:45like the first part of travis basler's
1:21:46paper are people could actually people
1:21:48tell you on average
1:21:50whether there's
1:21:52twice as high wages in the cities and
1:21:54you know if you were to go to the city
1:21:55what how much higher would you be paying
1:21:57for the same basket of food let's say to
1:21:59make it concrete travis travis talks
1:22:01about that how you know if you go to um
1:22:03nairobi you know wages are twice as high
1:22:05but but
1:22:07um
1:22:08you know but baskets of food are i don't
1:22:10know 30 percent higher or something so
1:22:12that's that the kind of our our people
1:22:15that's something i think that um
1:22:17that we could use some more info on
1:22:19corey that's not a complete answer your
1:22:21question is just off the top of my head
1:22:23there's there's lots of answers to that
1:22:25there's lots of data that we need it's
1:22:26hard for me to it's hard for me to just
1:22:28say exactly
1:22:29you know exactly what um
1:22:32what the answer is
1:22:34i'm trying to think of doug you gave me
1:22:35a bunch of questions i think that was
1:22:36the full set but i've got another set
1:22:38for you if you're willing great
1:22:40so
1:22:41a question
1:22:43kind of a shared question from um maha
1:22:46al-qatar and rakish banerjee
1:22:49um which is about whether there's what
1:22:52else might be going into people's
1:22:54decision-making about
1:22:55whether or not to migrate rakish says
1:22:57is there work that thinks of
1:22:59non-monetary benefits or or broadly
1:23:02culture or values and and maha says um
1:23:06sort of asks the same question about
1:23:08decision-making processes that keep
1:23:10people in agriculture whether it's
1:23:12something about sentiments or behaviors
1:23:14or history or culture but sort of
1:23:17are we missing things in this very
1:23:18neoclassical framework i guess is a
1:23:21question for you
1:23:22yeah
1:23:23why don't you take that one that's a big
1:23:25question
1:23:26yeah i mean i think we are um
1:23:31i are i i we are i guess the the um
1:23:34the one about
1:23:36um
1:23:37yeah so
1:23:39so history and culture sentiments i mean
1:23:41i i guess
1:23:43you know the the sentiments come from
1:23:45somewhere um and so that could be good
1:23:48information about the lifestyle of a
1:23:50city and how you know we um
1:23:53you know in a village there's a
1:23:54different lifestyle i do think that
1:23:56there is a very real kind of city mouse
1:24:00country mouse thing which is not about
1:24:02your productivity in the cities you know
1:24:05you may be very productive in the city
1:24:07but you prefer the rural way of life i i
1:24:09know in my own
1:24:10family there's people who
1:24:13want you know want a rural lifestyle so
1:24:15i'm sure that there are taste
1:24:16differences across people
1:24:19um
1:24:19that's
1:24:21it'd be hard to quantify that i think
1:24:22but but i don't think it would be
1:24:24impossible i think one could think about
1:24:26that history in some broad sense must
1:24:28matter you know that's such a broad
1:24:30topic that i think that in many contexts
1:24:32you'd see that history does matter but
1:24:34you know it'd be
1:24:35it'd be nice to be more specific if one
1:24:37wanted to do a research um question on
1:24:40that
1:24:40um
1:24:42sentiments yeah sentiments could be
1:24:44information it could be fundamental
1:24:46tastes so i mean i think that like i i
1:24:48don't want to dismiss any of that
1:24:49because i think neoclassical
1:24:51approaches are good but then there's
1:24:52there's more to the world than that i
1:24:54just think you'd you have to be more
1:24:55concrete to turn it into a to turn it
1:24:58into a research project
1:25:02um
1:25:02great
1:25:03a couple of questions from damballa
1:25:05cotella
1:25:06who says what about credit market
1:25:08imperfections
1:25:10in relation to the migration choices
1:25:12um in [clears throat] general sort of
1:25:14uncertainty
1:25:16may be
1:25:17more than risk and and so
1:25:19um
1:25:20just wonder
1:25:21if you'd think a little bit about that
1:25:23and and
1:25:24the the question says
1:25:26uncertainty uncertainty seems relevant
1:25:28for migration decision more than risk
1:25:31objective assessment of the payoff
1:25:33um is really difficult for people and so
1:25:36i wonder if you'd comment just a little
1:25:37bit more on on that i think we all have
1:25:40that from harrison tadaro that intuitive
1:25:43sense that
1:25:44that there's [clears throat] something
1:25:46deeply risky about the migration process
1:25:49and certainly in fact we only see people
1:25:51doing this at certain points in the life
1:25:52cycle for the most part too
1:25:55yeah
1:25:56yeah no it's a great it's a great point
1:25:59um
1:26:00so not yet
1:26:02so uncertainty about the outcomes
1:26:05yeah it'd be hard to disagree with that
1:26:07um it'd be hard to disagree with that i
1:26:09i think that
1:26:10yeah i learned a lot from doing a like a
1:26:12little discreet choice experiment with
1:26:15mushvik mubarak where we
1:26:17we asked the same villagers in their
1:26:192014 econometrica paper we asked them a
1:26:22series of questions
1:26:25about hypothetical migration moves and
1:26:28like when i was thinking about how to
1:26:29design the survey i always had a
1:26:31different view and mushrik kind of said
1:26:32you have to ask them about a migration
1:26:34opportunity you can't just ask them
1:26:36about you know are they going to pack up
1:26:38and move to the city you have to kind of
1:26:39say an opportunity presents itself
1:26:42and then so we would we would ask them a
1:26:44whole bunch of hypothetical questions
1:26:45about um
1:26:47about migration opportunities and we
1:26:50would say like you're going to stay with
1:26:51a relative in the city or you're going
1:26:53to stay in kind of group quarters with
1:26:54no toilet
1:26:56and we would randomize all the
1:26:57characteristics we'd say you're going to
1:26:58earn this many taka a day but there's
1:27:00only a 50 chance you actually get the
1:27:02job in the end and you're gonna get um
1:27:06you know you're gonna you're gonna um
1:27:08have to be gone for six months or you're
1:27:10gonna have to be only gone for one month
1:27:12and we randomized all these
1:27:13characteristics and and just like the
1:27:15airport we have like the smiley face
1:27:17that's green that's you know the
1:27:18bathroom was clean and red is like
1:27:20bathroom was dirty these guys had to
1:27:22click the like the green button like yes
1:27:23i would take that migration stint or
1:27:25like red i would i would stay home
1:27:27and then we used that to estimate a
1:27:29little multinomial probit model of like
1:27:32all these different factors
1:27:34and um
1:27:35you know it's amazing that that your
1:27:38people yeah it's like the higher the
1:27:40probability that they it just it worked
1:27:42out just like you said the higher the
1:27:43probability that they'd actually get the
1:27:44offer the more likely they were to go
1:27:47and um and so but it just told us like
1:27:49that
1:27:50they're not thinking just about what the
1:27:51wage is but what's the probability that
1:27:53i can actually get you know and i can
1:27:54actually get this
1:27:56this job so this kind of like
1:27:57uncertainty about what's going to happen
1:27:58when you go there yeah for sure that for
1:28:00sure that matters
1:28:02um like let me let me um let me stop
1:28:06there not because i don't think that's
1:28:07an interesting question i just mushrik's
1:28:09gonna give a whole lecture right doug on
1:28:10risk and migration right somebody there
1:28:13was a related thing there about what
1:28:14role does insurance play i mean it
1:28:16definitely plays a role mushvik has
1:28:17several papers about that and so um
1:28:20yeah let me not say foolish things and
1:28:22then mushfig corrects me why don't we
1:28:23just let mushrik give his lecture
1:28:26fine so let me
1:28:27let me probably wrap up here with um a
1:28:30two-part question
1:28:32one part comes from radhika goyal um
1:28:36your student at times and mine
1:28:39is there a role for
1:28:40environment and climate change
1:28:42leading to sectoral misallocation
1:28:45and
1:28:47perhaps also just
1:28:49high temperatures reducing labor
1:28:50productivity in agriculture
1:28:53let me link that to a question that's a
1:28:55more general one and it would be a nice
1:28:57segue into
1:29:00into next week's lecture but
1:29:02we spend a lot of time talking about
1:29:04migration as a response to apgs these
1:29:07big productivity gaps across sectors
1:29:10we seem to spend less time
1:29:12talking about the reasons for low
1:29:14productivity in agriculture
1:29:16and i wonder
1:29:17if you have some thoughts on
1:29:19on that and
1:29:21whether we might need to think more
1:29:24about
1:29:25agricultural productivity
1:29:27and why it's so low
1:29:30rather than just
1:29:31focusing on the migration movement
1:29:33itself
1:29:34yeah
1:29:37so i couldn't agree more that we
1:29:39we probably should
1:29:41um as a profession think
1:29:44you know
1:29:45think as much or more about why the
1:29:47productivity is low to begin with
1:29:49um
1:29:51so this
1:29:52particular talk wasn't about that i do
1:29:54think there's a kind of a growing
1:29:57literature about that i mean i can cite
1:29:59your paper doug on the hybrid seeds and
1:30:02the green revolution you know i thought
1:30:04that kind of gives a very concrete you
1:30:07know description of what makes your
1:30:08productivity go up and how do you have a
1:30:10green revolution you adopt these hybrid
1:30:12seeds
1:30:13some of the places in africa you know
1:30:15don't grow the same crops that these
1:30:17hybrid seeds are available for if you
1:30:20grow wheat and corn i think those are
1:30:22good times for you if you grow
1:30:24tubers my understanding is that
1:30:26there weren't a lot of hybrid seeds that
1:30:27could really help you
1:30:29um and so that's a kind of a very
1:30:31concrete reason i think that the
1:30:33productivity is low in agriculture in
1:30:35africa it'd be hard not to talk about
1:30:37mechanization i mean i think that you
1:30:39just
1:30:40a farm in the midwest which would be
1:30:42very high output per worker would be
1:30:44very mechanized it'd have hardly any
1:30:46people in it a farm in sub-saharan
1:30:49africa would be hardly hardly mechanized
1:30:51at all and would have lots of um
1:30:54you know lots of workers per unit of
1:30:55land
1:30:56so that so kind of the combination of
1:30:58small scale
1:31:00lack of use of hybrid seeds probably
1:31:02lack of use of fertilizer
1:31:04and
1:31:06lack of
1:31:07lack of
1:31:08irrigation and being forced to do rain
1:31:11rain fed agriculture
1:31:13to me those would be the proximate
1:31:15causes of low productivity in
1:31:16agriculture i guess the question is why
1:31:18aren't people
1:31:20um
1:31:21i'm going to answer one question live
1:31:22are you talking about productivity as
1:31:23yield per hectare i'm actually not
1:31:25talking about output per hectare i'm i'm
1:31:28i have a bit of a pet peeve with that
1:31:30that's like um the plant the plant
1:31:32specialist can study output per acre i'm
1:31:34talking about the value added the value
1:31:37of agriculture output per worker so
1:31:39that's what kenneth's question was a
1:31:41clarifying question so value added per
1:31:44per worker um if we want consumption per
1:31:46worker to go up um because we think
1:31:48utility relates to consumption we want
1:31:51we want consumption per um
1:31:54you know consumption per worker to go up
1:31:55we need value of output to go up per
1:31:58worker
1:31:59so anyways that's what i think the
1:32:00proximate causes are that so back to
1:32:02roddick's question about environment i
1:32:04do think it's really important i think
1:32:06it's really important both for
1:32:08why holding fixed the people agriculture
1:32:10is
1:32:11um
1:32:12productivity is slow and um
1:32:16and thinking about why people aren't
1:32:18misallocating but i don't i don't you
1:32:20know i don't really have um
1:32:22a lot of information about that yet i
1:32:24think that's kind of the frontier of
1:32:26where this field is going a little bit
1:32:28in terms of climate and agriculture
1:32:30trying to think about is this going to
1:32:31get people to move is this um i mean i
1:32:34think climate climate change has
1:32:36definitely led to
1:32:37people moving across space and um i
1:32:40think if you're in a farming area that's
1:32:43becomes too dry to
1:32:45you know try to use or too wet to to use
1:32:48anymore your debt you know you're
1:32:49definitely going to see massive out
1:32:50migration and
1:32:52so that's a huge topic and it's in its
1:32:54own right and um yeah i i look forward
1:32:57to reading papers by some of the people
1:32:59in this in this session about just that
1:33:01very topic
1:33:03david i think we should wrap up here i
1:33:05think we've already probably um abused
1:33:08the the cepr staff and abused your
1:33:11kindness and generosity in
1:33:13in making your time available we're
1:33:15really grateful for you having taken the
1:33:17time
1:33:18um i would just encourage everyone who's
1:33:21interested to tune in
1:33:23again
1:33:24tomorrow
1:33:27make sure i've got the details
1:33:29the details right but tomorrow we have
1:33:33a lecture on technology diffusion and
1:33:35adoption from chris tinetti from
1:33:37stanford
1:33:39i hope that many or all of you will be
1:33:40able to join us there
1:33:42and david i just want to thank you so
1:33:43much for a really thoughtful lecture
1:33:47apart from that first part of the
1:33:49going back through this paper that um
1:33:52you know on productivity gaps which um
1:33:55as
1:33:55as joe says is just a terrible paper and
1:33:58brings down the quality here but thanks
1:33:59so much
1:34:00and uh a pleasure to have you joining us
1:34:03great thanks so much for having me it
1:34:04was a pleasure to be here thanks
1:34:06everybody for attending

About this transcript

This page contains the full transcript of Lecture 7: Agricultural productivity gap: measurement and explanations by CEPR & VideoVox Economics, generated from the public captions YouTube serves with the video. The transcript has 17,508 words across 2,874 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.