Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. =Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable. = # Sowing Seeds of Mobility: The Gendered Impact of Land Reform Authors: Ting Chen, Jiajia Gu, L. Rachel Ngai, Jin Wang Discussant: None Video: https://www.youtube.com/watch?v=0mE9GNhwdY8&t=21500s ## Talk (05:58:20 – 06:42:29) [05:58:20] Hello receipt. So we are in the last part of the program. So we're going to have two [05:58:31] papers. The first one um on uh on gender reform in China, the gener. [05:58:40] >> Okay. So thank you everyone for coming back and thank you for the organizers for including our paper in the program. [05:58:45] So the paper is joined with Ting Chung from uh Hong Kong Baptist University, Rachel from LSC and Jing Wang from uh Hong Kong University of Science and Technology who is also here in the audience today. Because of my employment [05:58:58] with the IMF, the usual disclaimer applies. Everything I say in this presentation are those of my own opinions and has nothing to do with the IMF. So we start with the motivation. [05:59:10] Yeah. [laughter] So due to the work of some of the people who are here in this room, we actually know that the process of structural transformation is not genderneutral, right? And in particular, women transit [05:59:23] into non-aggricultural employment at a slower rate than men. There are many explanations offered in the literature related to that including norms and also stigma associated with early you know manufacturing work in early development [05:59:37] stage. And also there are other reasons looking at you know access to child care or safety transportation. So in this paper we argue um as a complement to the literature that actually land reform the lack of land rights in security play an [05:59:51] important role on the structural transformation by gender. So what do we mean by land rights in security and why is it particularly relevant for women. [06:00:00] So the type of light runs land rights insecurity we discussed in this paper is actually very common across developing countries. It's typically the case that farmers they have the right to farm the land but they don't have ownership. So [06:00:13] ownership remains communally owned. So then the right to use the land is contingent on continued farming. So if you don't farm it then you might end up losing the land. So the so-called use it [06:00:27] or lose it policy. So in response to that when deciding whether household members are going to migrate from agriculture to non-aggricultural sector they may leave someone behind as a god labor on the farm just to farm the land [06:00:41] to keep the rental to keep the usage rights. And why are women more likely to serve this role? Uh for various reasons. [06:00:48] Um the literature pointed to some reasons including so the task and the type of environment in early development stage is more male intensive right so women are um there is stigmas for example associated women doing those [06:01:03] type of work another reason is that uh farm work is actually easier to combine with home production when production and market work takes at the same place women actually do a lot of work right so that's like an early form of working [06:01:17] from home so for those reasons when you need a god labor then women are more likely to serve that role and that's our central argument in the paper today. So because of that then genderneutral land insecurity can act as a gender specific [06:01:31] mobility barrier and that's the thing I hope to convince you with. So we study this idea in the context of China because even though that this type of insecurity is very common across developing countries. China is unique in that its hook system that land [06:01:45] institutional arrangement is embedded in its hook system making this uh barrier which is implicit in many other countries explicit and measurable. And then there are two major land reforms which I will tell you more about which [06:02:00] rolled out across China. And the third reason is that we have access to a large panel data set tracking rural individuals and households. I thought that would be a plus. But seeing some of the earlier presentations, maybe this is not something I should brag about. But [06:02:15] [laughter] so for those of you who are less familiar with the institutional setting in China, um so the system uh was introduced in 1958 to curb migration. So it acts as an internal passport to [06:02:30] regulate migration and it's usually assigned to you at birth. Changing that hookco type is very difficult. So one piece of information on the hookco is that uh this one here you see that that that cell is saying whether you have [06:02:44] agricultural hook or non-aggricultural hook. So if you have a non-aggricultural hook you almost always work in the urban non-manufacturing sector. But if you have a agricultural hook then you're entitled to farm the government [06:02:58] allocated land rent free but you don't own it. So this is how the usage right look like in in rural China in terms of the land rights. So there was a land administration law which gave rural farmers the right to farm their land. [06:03:12] But before the two reform that I will talk about they they can farm the land but they don't have the right to rent out their land. Um so the reform exactly strengthened rural farmers right to rent in and out of their land and in the [06:03:25] constit in the setting of China um the right to sell your land is still a no. [06:03:30] So mostly when I talking about when I talk about strengthening land rental rights I mean the right to rent out their land. So it has nothing to do about the right to sell. So uh some suggestive evidence on why why land rights insecurity actually [06:03:44] disproportionately affect women. So uh before as I was saying so the land was allocated to household based on household size for a fixed tenure and rural village leaders they typically [06:03:56] prioritize active farmers. So if you leave the agricultural sector leaving the land unfarmed then there's this risk of reallocation before tenure ends. So there are some ev so there's not a lot of evidence out there directly measuring [06:04:10] this but there are some survey showing that since the first round of uh contracting since the land administration law then by around 2000 actually about 80 villages have experienced land reallocation. So then [06:04:24] in response to that household leave someone behind as got labor and then we have some using our own data we have access to the rural fix point survey and in 2006 the survey asked a one-off [06:04:36] question to the village leaders asking the village leader to recall the major incidences of land reallocation since 1970. So using that information we show that in villages with more past land [06:04:51] redistribution there's actually lower current uh in the early 2000s lower non-aggricultural employment and migration and that is particularly uh more significant for women. So I'm not showing you today because I have a lot [06:05:04] of things to show. So that was the you know some micro suggestive evidence. On the macro side we observe trends like that as well. So on the black line you see so we we take all the rural married couples and we group them into four [06:05:17] types. So the black is those who both uh with both couples working in non-aggriculture and red is the the type of household where only the female works in non-aggriculture. The blue refers to the type of households where the husband [06:05:31] works in non-aggriculture and green is both in agriculture. So you see before the reform the reform started in 2003 right? So before that actually the dominating migration or employment type is husband working in non-aggriculture [06:05:44] but as reform rolled out over time actually both couples working in non-aggriculture take over. So the timing is also suggestive that the land reform played a role but to formally [06:05:57] test whether it's actually due to land reform uh we first need to construct some measure of the land reform. The two land reforms that we talk about um one is the land contracting reform rolled [06:06:09] out across China from 2003 to 2014. So this is the first reform implemented. [06:06:14] The reform basically established the legal framework to pro protect rural farmers rental rights and it has a two-stage roll out. So usually the provincial government released some u documents um about the roll out of the [06:06:29] suggesting that reform starts in this province and then followed by more detailed procedures of implementation by pre prefecture and county level governments. So but that was only the first state the the first it was only the first reform because it it [06:06:44] established the legal framework but still there are two things that they didn't do which is individual farmers still didn't have a piece of certificate stating their plot of land area location and also there was no national [06:06:58] registration system for for the first reform. So then even if you have disputes there's not a you know um objective measure to look at. So that was exactly what the second reform did, the land titling reform. This reform was [06:07:12] directly rolled out at the county level and then on average for each county it took about two years to complete the reform because you have to you know certifying each line parcel with a detailed GIS information. So we want to [06:07:26] construct a comprehensive measure basically capturing both reforms. So what we do in this case is we use all the policy documents more than three million from this website called pkulaw.com which has the most comprehensive [06:07:41] collection of all the documents released in China throughout that that period. [06:07:45] Then the first thing we do is we apply these if you read Chinese these are the words that's related to the land reform. [06:07:51] So that gives us about 4,500 documents but not all the document are actually about the reform because some of them mentioned reform in passing. So then that's where we used the lang large language model to help. So we used this [06:08:05] DOVA which is a China based um language model and then we asked it to actually identify whether the document is indeed about the implementation of the reform. [06:08:14] So with that we reduce ourselves to about 4,400 documents and then we extract the administrative level of the reform and its location and the year of completion. So with that we construct [06:08:26] this four valued reform index because land title reform is like the you know throughout my description you probably all realize that it's the strongest reform. So when you complete the land titling reform we give this county a [06:08:40] score of three and if it only has provincial land contracting reform you receive a score of one and if you have strengthened contracting reform at the county or prefecture level you get a score of two. So then as you can see [06:08:54] there's quite some spatial and timing variation across the across China in the roll out of this reform. So one natural question to ask in this case even though both reforms are top down from central government then there is wide variation [06:09:08] in the timing of implementation. So what drives the implementation timing? So we run a regression for each of the contracting reform and titling reform. [06:09:18] We run a regression of the timing of the reform on a bunch of economic control variables. So two things stands out is that uh real estate from the urban side and land disputes. So if there's more real estate real estate investment [06:09:32] demand for land then there the counties are more likely to implement the reform early and if there's more land disputes then have an urgency to resolve these issues. So they also implement it early. [06:09:41] So in our main specification of the regression we control for these two factors but we also do a event study just to show um whether just to check whether is any pre-trend and um on so we did this we do this for both contracting [06:09:56] reform and title reform at the local level. So uh we don't see any obvious pre-trend in the implementation um on on the reform on non-farm employment. So that was the index part. So the key um [06:10:11] thing that I want to convince you today is basically this slides the impact of reform on rural individuals. So we look at two out for each individual from individual I from origin county O in [06:10:24] year T we look at two outcomes. One is their non-aggricultural employment um decision and the other one is migration. [06:10:32] They're usually very correlated but we look at both. [06:10:37] Sorry. Is land reform measured such that I need to interpret the effect of contracting as three times the effect of just the land reform? [06:10:48] >> So um in the main specification that's our uh that's what we assume because we want a comprehensive measure uh which we I will show you today. that in the paper we also have we include three reform margins basically um provincial [06:11:02] contracting pro uh local contracting and local titling we include three reform margins separately so and then you can see that the local reform has much stronger impact and even controlling for uh contracting the titling still play an [06:11:16] important role so it's roughly consistent with our um even though we are lumping them together in what I show you today >> okay because you could also imagine that the relative value of each of those [06:11:30] reforms is quite different for women compared to men. So, so that kind of parameterization kind just using the value of that categorical variable seems seems [06:11:43] >> uh restricting. >> Yeah. So, um let me finish this slide and I show you the individual and I'll tell you what exactly we did. So um so so yeah as as as um Jessica already [06:11:57] pointed out in this main specification we actually just use the land reform index ranging from zero to three and then uh we control for individual level observables and village level economic controls as well as individual and year [06:12:10] fixed effect and the key uh coefficient of interest is the beta 2 here which is the interaction term of land reform and female. So um now so a few observations so the column one to four are the [06:12:25] non-farm employment outcome and column five I have the for the migration outcome so if you look at column one so land reform increases non-farm employment for everyone but then um in addition to that the impact on female is [06:12:39] larger so which is consistent with our story of when land rights are insecure when you have someone staying behind guarding the land it's more likely that women take on this role and then as we include more controls from column one to [06:12:53] column 3 the coefficient and our the size of the coefficient and significance level didn't change much so precisely because of the reason Jesse had mentioned we also have some regressions where uh we include three reform margins [06:13:08] instead of just um lumping everything together. So here you see that both the contracting level of reform and titling uh both the local level of contracting reform and titling reform has a you know [06:13:23] in addition to the main effect on male it it has additional uh in addition to you know the average effect they both have additional impact on female especially when the reform is implemented um in the local level. So [06:13:36] relative to the >> how do I interpret those coefficients? [06:13:40] >> H >> how do I interpret those coefficients? [06:13:44] >> So uh so this is percent. So these can be percent. So because the it's a dummy variable. So it's percentage increase. [06:13:51] So it's tiny >> uh uh not necessarily which so if you aggregate these things up actually the you know weighted by the reform implemented and actually increased female non-aggricultural employment by [06:14:05] eight percentage points and men by five around that yeah when you aggregate them multiply by the corresponding reforms which is something actually I will use to calibrate our macro model to make sure um it's tightly linked you know the empirical part and the macro model part [06:14:20] um And and this project is not thinking at all about the efficiency of the use of wind. [06:14:27] >> No no no no. >> So the results I previously showed was including everyone. So we have additional um regressions where we include um you know [06:14:42] a dummy for unmarried male, unmarried female and married female. And then you see that the additional impact of land reform is actually happening through the married female suggesting our story of [06:14:55] God labor. Um it's consistent with our story of gut labor and then these are including every individual and we also look at a sap sample where we can identify the husband and wife. So using [06:15:09] this subsample then we can test basically against you know the four type of household we have. So column one and two I show you um the group of household by non-farm employment and three and four are for migration. So basically [06:15:22] what column one is testing is a joint non-farm both couples working in non-aggriculture versus only husband working in agriculture. So we see that reform actually increase the share of joint [06:15:35] couples working in non-aggriculture. And then the second column where I show which is both couples working in non-aggriculture versus no one working in non-aggriculture. And you see again here that reform increases the share of [06:15:48] this type of household. So these are the key results um from the rural side. Uh I won't talk too much about the urban side but in the paper we also show you know because you have this inflow of workers especially female workers from rural to [06:16:02] the urban region there might be spillover impact on the urban side. So >> consistent with the guard labor story, did you observe any change in the adolescent like daughter or like children's employment or education [06:16:16] outcomes since they can be guard labor as well? Um so we so in the majority of the household we're in this paper we're looking at the adult but then I think there are some [06:16:29] papers out there who specifically looking at the children's outcome but it's not so um as far as the story goes it's because you don't just bring your children to your urban when you move. So that's a different barrier that we're not touching in this paper. So I mean [06:16:44] like a daughter or like the son is like old enough but like still >> uh so they can >> they can be can they be a guard labor if they can be like a guard labor is there like now they're more likely to get educated or they're less likely to do this in that >> uh we didn't [06:16:58] >> we didn't check whether the younger like adults in the household can be using that that that's something we can check we checked against the older you know par older working age parents whether they can but good point yeah we should [06:17:12] do Um how how are you can you go back to non uh farm employment is that non-farm employment in the rural area or is that nonfarm employment in the urban area [06:17:25] >> so um most of them are in urban area so in China I think in the '9s you have a lot of non-farm employment in the rural area because of the village township uh enterprises but those things are not so relevant for the period that we look at [06:17:39] which is post 2000 so we're not So, so most of these employment actually you do have to leave. [06:17:46] >> Yeah. So, that's why you see the coefficient on the migration is quite similar to the non-farm employment. [06:17:51] >> So, what I'm worried about is you said a bit earlier that the places that were doing the land reform um were those places that had high demand for land in the urban areas for you know enterprise [06:18:04] growth or whatever and so how are you taking care of that? So uh we include that as a control and we also have this control if you notice the weighted employment growth using sort of a shift [06:18:17] share measure linking the origin county to the urban potential destinations and changing the employment growth in those urban areas which takes care of the demand side. [06:18:28] >> Okay. You might you might want to if you can see where they're migrating to, you might also want to try and see, you know, is the land reform triggering migration further away to kind of get around that concern that there's some interaction between what's going on with [06:18:43] urban urban like labor growth close by and area. [06:18:48] >> Yeah, we're not checking that in this paper, but one feature of the migration pattern in China is actually in the early days, you have to really move out to the coastal area to get a job, right? [06:18:58] with development uh throughout the period that land reform is taking place actually uh because the other provinces inland are also developing. So you don't even need to migrate so far away to get a job because you can just move to your closest capital city but we control for [06:19:13] those in the regression as well. Yeah. So on the urban side uh we have some regressions ship share designed to show [06:19:27] that actually uh the inflow of female workers from rural to urban area led to lower employment rate and wage rate relative to the uh urban male counterpart. But that's not the focus of today. So I want to focus on the rural [06:19:40] side. So now the the results I showed you on the empirical side are comparisons cross region right um you know very depending on how the reform intensity so it doesn't directly tell [06:19:53] you what is the magnitude aggregate magnitude of the impact of reform so for that we need a model so um we want a model with minimum infrastructure so that we can do some serious calibration so uh what the model we have here today [06:20:06] is basically we have one urban region that hosts the non-aggricultural sector and then we have a continuum of rural regions. Um there's reform regions and non-reform regions in the rural area and [06:20:20] the the rural region only has agricultural sector. So the key thing that we include the reform region and non-reform region is because when you have these two regions coexisting then the difference between the two types of [06:20:32] regions kind of is similar to the empirical um impact that we can detect which is also cross region. So can you just tell me what the objective function is because I'm a whole so it's often the [06:20:46] case and it was in China that these rural farms are way under sized and the land is very close to cities. [06:20:59] It's very valuable. And the question is how to encourage individuals to actually move out. And so the idea in certain [06:21:10] places was to have a means by which land was treated for that or they were given some future promise. So the question is well in in this current model what is [06:21:24] the objective of of the state? um we do not directly model the role of state in >> this model what what what is the question >> of this model? [06:21:36] >> Yes. So we want to use the model to answer the question which is how much of the structural transformation we observe can be explained by reform and whether reform play a bigger role >> transformation here is the shift of [06:21:50] individuals from rural areas to >> to non-aggricultural sector >> non agricultural sector but is the land itself switching >> uh land itself is not switching so [06:22:03] because we have a representative which I will show you. We have a representative production. [06:22:08] >> So, so then there's an inefficiency in the use of land. [06:22:14] >> So, we may have a greater efficiency in the use of people but we have an inefficiency in the use of land. [06:22:20] >> Yeah. So, in the sense that then fewer people is working with larger plot of land in that sense. Yeah. [06:22:26] >> Yeah. >> Um so you what one thing one can do is by modeling this at the household level. [06:22:33] >> I don't care about the model. I care about the real world. [06:22:37] >> Okay. And I I I'm trying to think about everything that I know about what went on in China in the 2000s. And I know just snippets about how individual [06:22:51] mayors of certain cities managed to encourage individuals to leave the land and to get the land over to the city boundaries. so that you know so that [06:23:06] modernization can actually proceed. So it's not just the shift of people but the shift of the land. [06:23:13] >> Um that we I agree that is a potentially important thing that our paper because we want to focus on the difference between gender. So then we're not looking at you know the heterogeneity across. So what some some people when [06:23:27] they respond to that is they model heterogeneity at the household level with different abilities. So when you have the less efficient farmer leaving then the more efficient ones are plotting relating to the efficiency story. So because we really want to [06:23:40] focus on the gender differential side so we try to minimize the rest of the okay so um and then so uh so that was um [06:23:55] we have a representative production function in the uh agricultural sector as I will show you. So um and then we each household in this case consists of a female and a male member and we have an exogenously share of urban household in this case and we just assume there's [06:24:09] frictionless goods market. Um so this is the the production side we try to keep it minimum. So uh urban sector um um the only uses labor the non-aggricultural sector residing in urban sector only [06:24:23] uses labor and it's a CES aggregate of female labor and male labor and the female labor would include the local workers as well as migrant workers you know moving from rural to urban area and [06:24:35] then uh on the agricultural side um two things matter uh so n is the labor and k is the total capital in uh which is land in our case and then so because we allow [06:24:50] reform regions and non-reform regions to coexist at the same time so R here is for the reform index so we have a five share of regions that actually experienced reform and then five one minus five share of regions that are not [06:25:02] reformed so on the rural side basically these rural household will have to make the decisions of whether they want to have one member moving to the at least one members moving to work in the [06:25:16] non-aggricultural sector and then or having everyone staying behind. So that's the first stage of their choice and we will assume that it's captured by uh fresh a um distribution with kapa parameter cap one and then on the second [06:25:28] stage conditional on choosing to migrate then they also have the option of whether they want the female member to migrate the male member to migrate or both migrate into the non-aggricultural sector. So uh in the second stage we [06:25:42] will assume that they have another independent fresh a distribution with kapa 2. Uh these two will play these two parameters will play an important role for our cross region results to match our um empirical regression coefficient. [06:25:56] And then here's the land policy that we want to emphasize which is uh so lambda I here I refers to the four types of household that I've just talked about whether you have both working in agriculture and one member or female or male working in agriculture or both [06:26:10] working in non-aggriculture. So in a reform region for example uh when I is equal to one then this is like the region has um reformed so you're not subject to land incoming security so that is equal to one throughout but in a [06:26:25] non-reform region for example when both members leave agriculture uh then your lambda is going to be zero right because you've left the agriculture and before the reform then you basically don't get any land income so that's the key mechanism that we're building into the [06:26:39] difference between reform and non-reform region So seeing the land reallocation rule depending on which region they live in and also knowing their preference which uh the oopsilon term captures the two migration preference. Household [06:26:53] basically decide whether they choose their time allocation as well as employment allocation. So uh consumption is relatively straightforward. You have um agricultural consumption with the subsistence term and non-aggricultural [06:27:06] consumption. So two terms here we build in is delta f. So by calibration it will be smaller than one one. We're not implying that. What does it means is that um your when women leave [06:27:19] agriculture because we know that women tend to do more home production right when they separate from the agricultural sector home production efficiency suffers. So that's what our delta f captures and we also allow delta s to [06:27:32] vary just so that we match data. So that captures the utility when couples split. [06:27:38] So H basically captures the whole um nonhome production aggregate and it's efficiency the at a household level the efficiency of home production will depend on their migration type as well [06:27:50] basically and then this is the the how these barriers enter at the household level right household um the choice is standard they have one unit of time to allocate between market work and home [06:28:04] production and then they um they use they have one unit of time where they uh sorry their income is spent either on agricultural goods or non-aggricultural goods and then they have two income right one is the land income and how much land income they will earn is [06:28:18] exactly where the land policy distortion come in because it's directly entering the resource constraint and then they can also and then second income they have is their wage income which will depend on whether they work in agricultural sector or non-aggricultural [06:28:32] sector so we model an exogenous squish mu capturing that even when migrant workers rural workers migrate to work in the city um they don't necessarily receive the same subsidy as urban uh [06:28:47] individuals and in the context of China this could capture um misallocation and also the fact that when rural people move to urban to work their children and their own you know health care they don't receive the same amount of subsidy [06:29:00] as urban resident so even though they pay the same tax so that's all um captured by the reduced foam mu in a And so on the urban side uh because we want to emphasize everything on the road on the urban side we have a [06:29:14] representative household very minimum infrastructure choosing the employment rate as well as the choosing the employment rate for both female and male and consumption. So there's not much [06:29:24] going on on the urban side. um the the key of this paper is how and I won't talk too much about the calibration but there are three things that I do want to [06:29:38] talk about but because it's how we match the data so as I was saying so lambda represent the land reallocation policy so when in the reform region because reform has already taken place whether [06:29:53] you have someone guarding staying behind guarding the land is irrelevant. So in that case lambda is equal to one but in the in the non-reform regions then your migration choices will matter for your land income. So if both couples leave [06:30:08] agricultural sector so B for both so you receive land income of zero and then the god labor mechanism basically means that uh for the household where one member staying behind so lambda f and [06:30:23] female member staying behind and lambda m means male members staying behind right and o is for no uh both in agriculture. So what this means is that when you have someone staying behind guarding the land, you actually get to [06:30:36] keep the land income. So uh I will show you a counterfactual exercise where we we are we um we don't have the guard labor mechanism but for now this is our baseline calibration. So that's how we set the land reform uh parameters. So [06:30:50] when the region reforms basically they move from uh this arrangement to this arrangement on the right hand side. So another thing we take from our data is the share of the reform regions. So [06:31:03] remember we constructed the index of 0 to three when with three denoting reform is completed in that region. So we use that divided by three to mean what is the share of reform regions in the [06:31:17] model. So in that sense in the model in 2000 there's no reform regions and then in 2020 which corresponds to the map I show almost all counties have completed reform. Yes. Just [06:31:30] >> can you help me understand the intuition for how this would play out in a dynamic model. So what you know guard labor not only changes lambda in a non-reform [06:31:44] region but it also changes the option value of having agricultural income in subsequent periods. Um actually we have a static model where we just calibrate [06:31:58] it to much different time periods >> right but I guess what I'm trying to understand is intuitively right if we had a dynamic model where is stuff that really is kind of capturing the option [06:32:09] value of future agricultural income loading in in the model in the static model that you're estimating. [06:32:23] So if I have a okay if I have >> if if part of your choice to choose guard labor is not about the income that you receive the share of the of the income that you can preserve this period [06:32:38] but it's actually because in some period in the future you want to be able to have that income right then how how is that reflected in in this static model that that you right like where's the [06:32:52] distortion coming from assuming that away >> so um I think now we we don't directly model that uncertainty right but when you uh ask people I think it it is indeed the case that they take into [06:33:05] their future uh land in like probability of land reallocation as well so um so our model will not be able to directly speak to that but one thing we do in the model is that so despite I didn't [06:33:19] despite all of that we actually we will make sure that you know relative to empirics our model exactly is capturing what is our empirical is telling us in terms of migration so we're not overstating or understating at least [06:33:32] relative to our empirical benchmark the impact of reform too >> but I think that the question was since you know the parameters in these models right are going to be pinned down by equity conditions about relative [06:33:46] you know your your main distortionary parameter in the end to match everything is going to be I believe mu >> uh mu will help >> and then yeah and then the other ones so I think the question was where are this [06:34:01] distortion loading in terms of your wedges in this current model even if you don't have a dynamic model if you're trying to match the data you know [06:34:13] this this distortion has to go somewhere for you to match um the equity room you know since the equity condition in your model that >> that's true. So what you mean is basically we are loading these other [06:34:27] parameters but we so yeah so uh but what we do is basically we will let these parameters vary as well to match the time varying targets in that sense and then we will then isolate the impact of [06:34:40] reform alone. So then in that sense I think we're still um you know quantitatively fine with yeah >> and then so uh uh so so to match as I [06:34:55] was saying because we want our model to match our empirical coefficient right so um the three reform uh the three reform margins that I showed you basically imply a certain amount of female [06:35:08] migration and male migration and we use the kapa 1 and kapa 2 in the model to exactly match those two parameters such that the model's cross region comparison is the same as what the empirical [06:35:20] predicts. So um so in the aggregate trend our reform and so the we so here what I'm showing you is basically what happens in the reform region and non-reform region because the weighted [06:35:33] average of these two gives you the aggregate picture which is what our model uh match but this one is a prediction. So one obvious thing you see is that uh in the reform region you have much higher share of the black line [06:35:46] which represents both couples working in non-aggriculture and the the household pattern implied individual outcomes are the ones that you see in column one. So basically these are also what we [06:35:58] computed from regressions. So uh in the reform region female employment rate in non-aggricultural share is higher by 8.4 four percentage points relative to the non-reform region. But this is the baseline where we in we use the guard [06:36:13] labor mechanism to match that. Right? So you would want to ask maybe what if the guard labor doesn't exist. So we perform a counterfactual exercise where we you know hold all the other parameters the same but we change the amount of land. [06:36:26] Um having one member stay behind can help you uh secure. So when it's 0.5 it's basically saying you can secure the land for yourself because you were there but you cannot secure the land for the partner who migrated to the non-aggricultural sector. So that's what [06:36:40] we mean by setting lambda to equal to 0.5. So in this case what you see is actually um in the reform region and non-reform region the difference across those two region is actually symmetric across gender. So you see about 7.5 [06:36:55] percentage increase in both female and non male non-aggricultural employment share. So that's basically now in both cases the baseline and the no guarding case women disproportionately work in [06:37:08] agriculture because for two reasons one is that uh it's still the case that agricultural sector is more female intensive and when women leave agriculture uh home production suffer but in the baseline case in addition to [06:37:22] that having some women stay having a household member staying behind can help you gut the land which means women take on that role disproportionately. So when you reform removing that you actually [06:37:35] get a larger share of women moving out of agriculture because of the interaction. [06:37:40] >> It's a small question. Isn't it also the case that their children can only go to school in that that area. So they have to stay >> in that area if they have children of a certain age. [06:37:53] >> Yes. Yes. >> But when when their children age out then it doesn't matter. then it's not going to >> uh so you exactly for that reason that's why the home production term is particularly relevant. Yeah, because children are only allowed to go to [06:38:07] school in Yohooko registration place. It's not that you move to Urban and you can go to urban to school. So yeah, so but that we so we're not directly modeling that we have lambda capturing that as a residual time to match the data. Yeah. So, so the so basically with [06:38:22] this I hope you I hope I convince you you know the interaction term between the gut labor mechanism and women's role in home production and their market you know work actually interacts such that land reform disproportionately affect [06:38:37] women. So then that remains the aggregate question right how much of the observed structural transformation by gender can be explained by land reform. [06:38:47] So um responding to Jessica's question we do have parameters matching varying over time to match the model right so there are four set of parameters that we vary over time one is the share of reform region and one set of parameters is the productivity growth and [06:39:01] urbanization process and then mu and delta s reflecting genderneutral change and then sui h delta f these are women's role in market and home production right so um so the real change because we [06:39:15] matched our aggregate model uh to data on that respect. So we have female for a rural household female non-aggricultural employment increase by about 34 percentage points throughout the period [06:39:28] and male increase by 45 percentage point roughly. So the question is we do a sharply open decomposition basically to isolate the impact of each of the four terms that I set of parameters that I [06:39:41] talked about. So a few observations of course productivity growth and urbanization process um quantity wise plays the biggest role in terms of structural transformation for rural women and men but the impact is actually [06:39:54] quite genderneutral and then um the share of reform regions in um share of the the increase in the share of reform regions over time explains about 11% out of uh 34% for women. So that's a much [06:40:09] larger share about 33% of women's structural transformation whereas it only explains about 10% of men's structural transformation. So reform is the only and relative to the other terms. So the mu and delta s is these [06:40:23] are gender neutral but because delta s capturing disutility or attitudes towards couple splitting over time it's becoming more acceptable. So that's why you observe more male only moving to non-aggriculture. And then the gender uh [06:40:37] specific terms SUI H and delta F representing women's role in market and home these things are actually quantitatively not so important. So reform is actually not just quantitatively important for women but it's only it's the only force at least [06:40:52] according to the model based decomposition that we have that explains more women's structural transformation than men. So um on the urban side uh because our model has a small urban side as well. So it predicts that female [06:41:07] employment um the the flow of labor especially the relative increase of female migrants uh will decrease uh urban females employment rate and uh wage rate relative to urban man which are consistent with our empirical [06:41:20] predictions. So I'm early on time actually. So um so what we do in this paper is based on a quasi natural experiment in China and then also a quantitative model. We show that land reform has important implications for [06:41:34] structural transformation by gender and then the way it works in our the story that we're proposing is the interactions of the god labor mechanism and women's role in home and market production. And [06:41:47] now the type of policy that that we focus in the context of China. But the type of land in security that we talk about are actually quite common in developing countries especially in Africa. So we think this has wide [06:42:00] implications beyond China's context. Thank you. We still have some time for questions. [06:42:07] [applause] Yeah. If anybody has a question as we switch presenter.