Aishwarya Kekre on Land Fragmentation and Its Impact on Structural Transformation

SHRUTI RAJAGOPALAN: Welcome to Ideas of India, a podcast where we examine academic ideas that can propel India forward. My name is Shruti Rajagopalan, and this is the 2026 job market series, where I speak with young scholars entering the academic job market about their latest research on India. 

We are launching our series with Aishwarya Kekre, who is a PhD Candidate in Economics at the University of Virginia. Her research focus is at the intersection development, trade and spatial economics. We discussed her job market paper, “Barriers to Structural Transformation: The Role of Agricultural Land Fragmentation.”

We talked about the reasons for land fragmentation in India, its impact on manufacturing enterprises, on employment in both agriculture and manufacturing, profitability of small farms, how it impacts overall structural transformation, and much more. 

For a full transcript of this conversation, including helpful links of all the references mentioned, click the link in the show notes or visit mercatus.org/podcasts.

Hi, Aishwarya. Welcome to the show. It’s a pleasure to have you here.

AISHWARYA KEKRE: Hi, Shruti. I am very excited to be here. Thank you for having me on the Job Market series. This is my first time ever on a podcast, so it’s also exciting and a little nervous, [laughs] 

RAJAGOPALAN: Well, it’s all downhill from my podcast after this. [laughter]

Land Fragmentation and Structural Transformation: Surprising Trends

RAJAGOPALAN: I wanted to kick off our Job Market series with you because the paper is on land and land fragmentation and structural transformation. These are two themes that have preoccupied my imagination and my work for a long time. So, when I saw this, I was like, “OK, this is it. We got to start with this paper,” [laughter] and I’m really excited. What you do in your job market paper is you look at the impact of land fragmentation, specifically on India’s structural transformation.

And because of a combination of land redistribution laws and land ceiling laws, and India’s inheritance laws and custom, the average operational landholding size is halved in India since the ’70s, down from 5.6 acres to 2.8 acres. You find that this greater fragmentation actually reallocates labor away from manufacturing to agriculture, thereby reducing manufacturing activity, which is exactly the opposite of what the policy goals and the development goals are in India. This is, of course, what you find is more acute in the case of land-intensive industries.

More specifically, in terms of the results, you find that on average, district-level farm sizes fell by 12.45 percent between 2001 and 2011, implying a 3.2 percentage point increase in agricultural employment share and a 2.2 percentage point decline in manufacturing employment share. That’s the first big result, and again, exactly the opposite of what the policy goals and intentions are. 

The second major finding is that between 2005 and 2013, the average decline in farm size implies a 30 percent decline in manufacturing employment, a 20 percent decline in the number of firms, and a 15 percent decline in the average firm size. These are very large effects. And when I started reading the paper, I didn’t expect the effects to be this large. And then, again, the negative effects are larger in more land-intensive industries. 

And one surprising element—at least it was surprising to me—what you find is that while this is the news when it comes to manufacturing, smaller farms may often end up being more profitable per acre. There are various reasons for that. You talk about how they mostly rely on family labor, there is less supervision, and so on, but that is a slightly counterintuitive result. Everything is not going in one direction. There is this tension in your land fragmentation story.

This small-farm profitability increase maybe partially offsets some of the other problems caused because it hinders structural transformation, especially industrial land assembly and so on. 

But then, there is the broader question of, overall, is this good? Does it leave farm workers and manufacturing workers better off or worse off? Then eventually, you create a general-equilibrium model to walk us through those effects. Is this a good summary of your paper? Have I largely gotten this right?

KEKRE: This is absolutely a very good summary of my paper. What I would just like to add to it is that, as you said, what basically steered me towards this question was this divergent observation on farm sizes. What we see is that since [the] 1970s onwards, you would see this pattern that in low- and middle-income countries, the average farm size has been declining over time. This is completely opposite to the experience of high-income countries, where farm sizes have increased along the process of development.

If you look at this, along with the fact that what we know, that in developing countries, the largest share of their employment is still in agriculture, which also happens to be the sector with the worst labor productivity. It made me connect the dots that, does this phenomenon of declining farm size have any role to play in explaining why such a large share of employment still remains in agriculture in developing markets.

So, that’s where the whole question started. And then, because of my familiarity with India, and as you very rightly discussed, for several reasons, India is a good setting to study this question. We have the problem of declining farm sizes. It has almost halved in the last 50 years. We do see a stagnant manufacturing sector, and still agriculture remaining the largest employing sector. It was a very good setting to actually study this question.

Basically, what I do in this paper is, I start by doing some empirical investigation first. I just wanted to see in the data what is really happening. Is land fragmentation actually affecting structural transformation? For that, I look at Indian districts, and I look at the period from 2001 to 2011. As you rightly said, what I find is that declining farm size actually hinders structural transformation. I find that people are moving out of manufacturing and into agriculture during this period.

One of the mechanisms that I document in the paper is through this manufacturing site friction, that as land sizes decrease, manufacturing activity also goes down in terms of number of firms and employment and average firm size. So, given this finding, the goal with the model then was that I wanted to quantify what are really the underlying mechanisms driving it, and then how do we interpret this finding in terms of welfare. 

Then comes the model. As you mentioned, one of the key, surprising findings, even for me, was that although land fragmentation hinders structural transformation, but if I look at districts’ welfare—which is, say, a district’s real income—it actually improves with land fragmentation. There’s a negative relationship between the two things.

That was a very surprising finding that, “OK, then what really is driving this welfare effect?” So, I decomposed this to actually see who is taking away all these gains and who is losing from land fragmentation. That’s where the model, as we’ll discuss this further as well, reveals that it’s the workers that actually bear the entire cost of land fragmentation. The real wages in districts go down, and all of these gains essentially are just coming from increases in the rents that landlords—landowning individuals—can make from land fragmentation. Basically, we are looking at this very big distributional gap, and so we should be careful about how we interpret these welfare gains in the model.

Understanding Land Fragmentation

RAJAGOPALAN: Yes.Before we get into the precise mechanism and then the welfare effects, I want to actually start with what is land fragmentation for those who are not land nerds like you and me. [laughter] This is not a phrase we hear often. I think most people are aware of the fact that land ends up being quite a roadblock. It’s very difficult to buy and sell land in India, very difficult to convert land use. It’s very difficult to assemble parcels for an industrial project. Those are the themes that get picked up in the headlines. 

But can you first just set up, for those who are unfamiliar, what is land fragmentation? How does one think about it? And how does one go about measuring land fragmentation in the Indian context?

KEKRE: What is land fragmentation? I’ll just speak through the lens of India now. Broadly, very simply, what I’m trying to measure by land fragmentation is the decline in farm sizes. But where does this all come from? Where does this start? There are several factors in India that have contributed to this decline in average farm sizes. It goes all the way back to the policies that were implemented right after independence from the British rule.

So, after 1947, one of the policies that was implemented by the government was the Abolition of Intermediaries Act. What this act did was that it transferred land from large landowners to landless tenant farmers. In doing so, it actually created almost 20 million new landowning small farmers. What it did was that it immediately set us up with smaller farm sizes to begin with. And then, over time, there’s been a combination of factors that have just led to this small farm size even further decline with time.

For example, in India, as most people must be aware, the inheritance laws essentially lead to a parcel of land being split among all heirs of an individual. So, now, if you look at that law, and then you think about the growth rate of population in India, what I see in the data is—and there’s this figure in my paper, which shows that the number of farms has almost increased at the same rate of rural population growth. So, with a fixed supply of land, these inheritance policies and population growth have implied that, on average, farm sizes have continued to decline over time. 

Then what happens is that there are also a lot of other things that have, over time, prevented the consolidation of land. For example, in India, there are a lot of social norms regarding land ownership. People attach a lot of higher status in the society with respect to land ownership, which makes people reluctant to part with even a small piece of land, regardless of how viable it is for production. 

And then, even if people are willing to—then, as you said, there are so many regulatory hurdles. There are poor property rights, and then there are these multiple processes that you need to navigate in order to either sell your land within agriculture or even to just convert it to non-agricultural use that make the process of consolidation even harder. All of these factors together have basically led to these farm sizes declining over time. 

In terms of measurement, there’s the Agriculture Census of India, which is done almost every five years. What this agriculture census does is that it records the number and the area of operational holdings in India—at the district level, by different size classes—which I use for measurement. 

What we really mean by an operational holding here is that it is a piece of land that is operated by either one person individually or jointly with somebody else. So, it is not really land ownership, but it is at the level of who is really operating a land parcel. 

It gives me a sense of how many operating land parcels are there in India over time. Now, ideally, what I would like for my context—as we go further, it might become more clear—is I would like to study land fragmentation based on land ownership. It might be a case that I am the owner of the land, but I lease it out to multiple small farmers, which means that I would see in the data that there are a lot of small landholdings available, whereas the reality is that I’m just leasing it out to others. So, ownership and operational holdings in that sense might be different in data. 

But what I find is through the Indian Human Development Survey, it reveals that there are very few farming households that actually engage in leasing. It’s roughly 11 percent of the entire sample, which gives me some reassurance that I can still use these operational holdings as a proxy for actual land ownership, because most of the people do not engage in land leasing.

Essentially, what then I do with this data is, I treat this as my proxy for land ownership. Using these 10 different size classes that are available in the agriculture census, I use the number of farms in each size category as a weight and take the midpoint of every size bin as a representative size to create a weighted average farm size of a district. And I’m able to track that from 2000 until 2015, over four rounds of agriculture census.

RAJAGOPALAN: Yes, and we have such terrible land laws when it also comes to leasing that it completely disincentivizes leasing, right?

KEKRE: Yes.

RAJAGOPALAN: So, I completely understand why your survey results with the other survey also suggest that, actually, leasing is a very small number overall.

KEKRE: Right. There are states like Kerala which explicitly make it very, very hard to use land, which prohibited it effectively. Yes, that’s true.

RAJAGOPALAN: But there’s another interesting question. The way you set up the paper is that it’s very hard to consolidate land for manufacturing. But what I find is it’s also very hard to consolidate land for agriculture, right? So, we manage to kill both ends. If someone managed to have a really large farm, then they could just get the Change of Land Use certificate and they can manage to sell it. But we don’t allow farmers to sell land to non-farmers. Because of urban land ceiling laws, there are caps on what individuals and households can own. So, you’re prevented from selling to a richer farmer, right?

I know that part of the story is social norms, and people like to hold on to ancestral property. But I think people, at the end of the day, are rational when it is so hard to actually transact in land and sell to a farmer. The prices get so depressed because we’ve distorted the agricultural land market that it functionally becomes impossible. Then, of course, we say, “Oh, we’re very attached to our property,” or something like that. [laughter]

But what it really comes down to—for everything else, we somehow manage to sell. We also manage to offer very close possessions and ancestral property as collateral, but somehow, we can’t manage to sell it. So, I’m not so sure about the social norms story. But for your paper, that hardly matters. 

Measuring the Impact of Land Fragmentation

RAJAGOPALAN: So, now that we know that the operational size, which is a good proxy for overall landholding size by ownership, tends to be extremely small, what is a good way for us to think about the impact? Because this could have so many different kinds of impact, like on agricultural productivity, on our ability to transform to manufacturing, agricultural labor and their wages, industrial labor and their wages. Can you walk us through, when you study a big picture question like structural transformation, how do you go about measuring the impact of something like land, which touches everything?

KEKRE: I think basically there are some bits and parts of this in the literature, which got me interested in this question. For example, I can think about two separate channels through which this land fragmentation phenomenon actually affects structural change. These two separate phenomena operate through the two separate sectors. One operates through the agriculture sector; one operates through the manufacturing sector.

Let’s just start with the agriculture sector. There is this vast literature that basically says that the yields, or the output per unit of land on a farm, are related to the farm size. Basically, what I’m saying is that with fragmentation, as farm sizes change, it can alter my returns from cultivation. It can alter how much I can produce on my land and how much profit I can make from that land. That can change my incentives of remaining in agriculture versus leaving agriculture. When I say “leaving,” I’m talking through the lens of land, like keeping my land under agriculture or removing it from agriculture.

In terms of incentives, that is one channel that exists in the literature. There are various arguments for it. It doesn’t necessarily go in one way or the other. So, for example, if you think about smaller farms, there’s this whole literature on inverse farm size productivity, which essentially says that the yields are larger on smaller farms. Often it is attributed to things like, “Smaller farms use family labor more intensively, and family labor requires less supervision.” It can also be this way—if you think about the principal-agent problem, they’ll shirk less. Because of these factors, using more family labor on smaller farms, you can raise the yields on smaller farms. 

But then there’s the other side of the literature as well, which says that mechanization is important. There are economies of scale in mechanization, which can prevent smaller farms from adopting these technologies that improve your productivity, enhance your productivity. With fragmentation, as farm sizes decline, farms can, on the other hand, also suffer from a lack of mechanization, which can reduce their yields.

So, overall, we don’t know in which direction fragmentation can affect your returns from cultivation and therefore your incentives to remain in agriculture. That’s the agriculture side of the story, which is there in the literature. 

RAJAGOPALAN: So, just to clarify one part, do we not know which direction it’ll go in, or do we not know what the thresholds are? Like with economies of scale and mechanization, it’s pretty obvious what direction it’ll go in. It’s odd to think that land would be an exception to this.

So is this just a question of thresholds that up to a certain threshold, where mechanization is pretty much impossible or unfeasible, a slightly smaller holding with family labor ends up doing a little bit better than a slightly larger holding with outside wage labor? But past a certain threshold, of course, mechanization and economies of scale are going to kick in, right?

KEKRE: That’s right. That’s 100 percent right. 

And that’s where the seminal paper of Foster and Rosenzweig from 2022 comes in, that in the Indian context with a more representative data, they are able to observe both small farms and large farms in the same dataset. That’s where they show that this is a U-shaped relationship, that there is a farm size threshold below which the returns from using family labor are higher, and above which, as the share of family labor, the share of that keeps on falling, then there are greater returns from mechanizing. So, there’s definitely the threshold, and then it depends on which side of that threshold you are when fragmentation hits you in terms of changing your incentives to remain in agriculture.

RAJAGOPALAN: Yes.If I remember that paper correctly, most Indian farms are so small that they are all on one end of that U-curve.

KEKRE: Exactly.

RAJAGOPALAN: Basically mechanization is simply not even in the conversation because they are so far from that threshold that it might just be easier, instead of hunting for seasonal labor or outside labor every farming season, that it’s just easier to actually have a smaller lot size and use your family as workers.

KEKRE: Exactly. That’s something that I will eventually find in the paper as well, that given that their threshold, if I correctly remember, is somewhere around 10 acres.That’s a very big farm size. [chuckles] If I look at the district average in my dataset, the highest is nine acres by the end of the time. It’s a very high threshold. So, that’s why most of the farms actually end up being on the downward-sloping side of that U shape.

RAJAGOPALAN: Yes.The reason I’m belaboring this point is someone who reads only your headline result shouldn’t walk away with the feeling and romanticize, “Oh, smaller agricultural holdings are slightly more profitable.” They are still in a very low-productivity environment, completely trapped, and we do need exit from agriculture for the larger structural transformation story. Does that make sense?

KEKRE: Exactly. It does. That’s something I will circle back to in one of my counterfactuals as well, how important it is to ease the frictions in manufacturing and what are the gains that we receive from it.

The Effect of Declining Farm Size on Manufacturing

RAJAGOPALAN: You said this is one mechanism, which is through agriculture. What is the other mechanism?

KEKRE: I think I found that second mechanism to be super interesting because objectively, when you think about it, it’s not very obvious that farm sizes declining can directly affect manufacturing. Because it’s not happening in the manufacturing sector. But there is this very interesting finding by Deichmann et al., and even Ghani and coauthors, which shows that in the context of developing countries—and I think these papers look at Indonesia and India—manufacturing activity is actually moving steadily towards peri-urban and rural areas.

These are the areas where agriculture is the predominant land use. So, what this automatically does is it sets manufacturing at competition with agriculture for this limited resource, which is land. Then, for manufacturing activity to expand at all, land actually needs to move out of agriculture and into manufacturing.

RAJAGOPALAN: Yes.

KEKRE: This is where this problem of declining farm size becomes salient. Because if you think about a firm which is trying to allocate, say, a given size of land, in the presence of fragmentation, it will have to negotiate with multiple landowners, coordinate, bargain, everything with multiple landowners. We know that this process of converting land in India from agriculture to non-agriculture is costly. It’s tedious. It has multiple stages. So, now, when I have to do it with one versus five farmers, it is naturally going to raise my land acquisition costs as a firm.

That can basically dampen the expansion of manufacturing activity. It can dampen the reallocation of land from agriculture to manufacturing. That can have implications for structural transformation as well, in terms of some of the pull forces.

RAJAGOPALAN: Yes. Now when you lay it out like that, it makes perfect sense, right? If we had to parse it out, there’s a set of transaction costs, which is what you point out. You need to negotiate with five people or 50 people instead of one, right? There is a set of regulatory costs, which is the cost of conversion and things like that. 

KEKRE: Yes.

RAJAGOPALAN: But there’s also a layer of uncertainty, which is the more number of people you negotiate with, there’s a greater chance that the provenance for that land is really off. It gets stuck in litigation. There is some random first cousin who comes and claims that this is an illegal sale because it was a Hindu Undivided Family, and they did not get their share in land. So, each family, or each landowner who’s trying to sell, needs to get no-objection certificates from 15 different cousins, right?

KEKRE: Exactly.

RAJAGOPALAN: So, it’s layer upon layer of transactions costs and added uncertainty on land. And now, when you set it up like that, you’re like, “Oh, now we get why land is the binding constraint,” right?

KEKRE: Right. And I think one of the most popular examples in the Indian context, and that is also the failed land acquisition of the Tata Nano automobile plant in the Singur district, where essentially for a thousand acres, they had to consolidate land from 12,000 landowners.

RAJAGOPALAN: Yes.

KEKRE: Because of all of these frictions and people not agreeing to the rehabilitation prices and bargaining and coordination, it eventually failed. And there are many more such examples. 

RAJAGOPALAN: Yes, there are so many examples like that. But what I find interesting in your paper is, Nano is a phenomenal example for a particular type of scale. You show that this is failing at the five-acre, 10-acre, 20-acre level, not the 1,000-acre level, right?

KEKRE: Yes.

RAJAGOPALAN: So, if your average size of holding is two acres, and now you need to negotiate with five or six people, even a really relatively small manufacturing plant is dead on arrival.

KEKRE: Exactly,.I would also like to make one mention here of a paper by Aradhya Sood, which also documents, in the Indian context, how private firms find it difficult to assemble lands that they acquired and build much more slowly in states which have smaller farm sizes, and their lifetime profits also decrease. It’s quite a salient problem in the Indian context, irrespective of how large is the scale of acquisition, as you said.

RAJAGOPALAN: Yes, and the reason I like these two mechanisms, one, they’re very nicely done in your paper, but overall, if I were to zoom out, there are many, many barriers to structural transformation. There are many developing countries where the barrier might be capital, right? For some, it might be that they’re landlocked. That is a barrier to structural transformation. For some, it might be the lack of labor, especially countries that had a big shock after a war or civil war or genocide or something like that.

In India, as of today, it seems like the binding constraint seems to be land. It’s hard to call one thing the binding constraint because it’s not easy to find strong empirical support for that. But work like yours is very much pointing in that direction, that land and agricultural land being regulated or governed the way it is, is what is trapping all people and land, both the factor of production of land and labor. Therefore, it impacts everything else, [chuckles] which is, in this case, manufacturing, which is in peri-urban areas.

Measuring Fragmentation

Now that we’ve got that figured out, how would you go about measuring something like this? Now, if I were to wear my economist hat, if you have places that have a lot of land fragmentation and manufacturing is hard to do in those same places, is there some third thing about that place which is making both land more fragmented as well as manufacturing harder to pick up? Or can we attribute the decline in manufacturing activity, or the difficulty in starting new manufacturing activity, to the land fragmentation? How does one think about that?

KEKRE: I think that’s bang-on where the empirical challenge lies, that there could be these local economic shocks that can simultaneously affect a district’s average farm size and structural transformation in the district. I think one simple, intuitive example that I could think about, I suppose, through eminent domain or something like that, there is an industrial park that comes up in our district, or there’s a construction of a highway or a special economic zone.

Suppose that project entails acquisition in a systematic way, that it takes away most of the larger farms to avoid all of these contracting frictions. So, the third shock is declining your average farm size in the district, and it is propelling structural transformation as well. It will reduce agricultural farming share, increase it in manufacturing or services, depending on the kind of project, the kind of shock that we think about. So, there’s definitely this endogeneity problem that really needs to be addressed in this empirical setting.

That’s another reason where I like India as a setting, because we have this inheritance law that was passed in 2005, which gives me some variation in how land fragmentation changes across districts. So, to give all the listeners a background on that, inheritance in India is governed differently for different religious groups. But for the Hindus, the Sikhs, the Buddhists, and the Jains—and I just together call them the Hindu-plus community, which comprises more than 80 percent of India’s population—it is governed by the Hindu Succession Act of 1956.

Now, prior to 1956, ancestral property, often individual in a family will just go down the male lineage. But after 2005, this law was nationally amended, and what it did was it granted inheritance right over ancestral property to women as well, so women also became equal coparceners when it comes to ancestral property. What it does for my setting is, by increasing the number of potential legal claimants to a land parcel, it actually induces land fragmentation.

That is the source of variation that I’m using and, because my analysis is at the district level, and this reform was passed nationally at the same time, I use this as an instrument by measuring the exposure of a district to this reform. The exposure is measured by the share of rural Hindu-plus women in a district’s population before the reform.

In 2001, what is the share of rural Hindu-plus women in a district’s total population? So, the districts with this higher share are more exposed to the reform. They’re likely to undergo more land-splitting and therefore larger declines in farm size. This instrumental variable strategy somehow helps me get some variation from the land fragmentation side, which I can then use to study how it impacts structural transformation. This was the broad idea.

RAJAGOPALAN: Once again, this is based off the share of potential claimants, is how you look at it?

Because the actual data that we have, you have the same issue with land fragmentation, which is we don’t have the legal ownership data and we don’t know if actually the legal ownership got more fragmented or less fragmented. So, the theory is that because you have census data and economic census data and agricultural census data, you are able to triangulate at a district level if both those things are happening simultaneously. Is that a good way to think about it?

KEKRE: Yes. I, basically, using the population census from which I construct these exposure measures, I—it’s not in my paper, but when I started the project, I started with an event study to just see, “OK, does this reform actually change the female ownership share in agricultural land?” When I say “ownership share,” it’s through operational holdings, definitely because it’s the agriculture census. I actually see that it genuinely increases after this reform is passed. So, that becomes my first stage then in this empirical setting.

The Magnitude of the Distortions

RAJAGOPALAN: OK. So, now that you have your empirical setting in place, can you walk us through the results of what you find? I think this is really important because I expected these results, but even I was genuinely surprised by how big the results are.

KEKRE: Right. So, the timeline I would just repeat is, I’m looking at the period from 2001 to 2011. I’m using two rounds of the population census. From the population census, I’m able to construct measures of what is the share of employment in agriculture versus manufacturing versus services at the district level. And then, from the agriculture census, I have these changes in district-level farm size from 2001 to 2011.

And then, I’m basically using this inheritance reform as an instrument. When I do this analysis, what I find is that land fragmentation actually hinders structural transformation. What I see is that the employment share of agriculture increases and that of manufacturing declines. 

I, in my analysis, do not find any significant effect on services, but I think I have some potential reasoning behind it, because services might not be as land-reliant as manufacturing. It doesn’t rely on too many big-size plots. So, maybe these frictions are not very salient for services, and also services might just be a very urban consumption phenomenon; land fragmentation is more towards peri-urban and rural areas. Maybe that’s the reason. 

And thinking about the size of these effects, what I find is that, between 2001 and 2011, farm sizes on average fell by 12.45 percent, and that implies a 3.2 percentage point increase in the agriculture employment share. Which is roughly—if I look at the baseline employment share, that’s about 5.5 percent decline relative to the baseline. I think, given this big share of people that are in agriculture, it is pretty sizable, is what my understanding was.

Similarly, for manufacturing, I find that there’s a 2.2 percentage decline due to this average decline in farm size, which is about 18 percent relative to the baseline employment share in 2001. These are pretty sizable is what my takeaway was from the structural transformation analysis.

RAJAGOPALAN: Yes.Are you able to figure out or separate in any way the employment share difference between men and women? Because that is also impacted by the inheritance law, right? Part of that is it changes how women view themselves, their pool of assets, how much one can educate a woman. Does she get married earlier or later, depending on what the pool of assets is that she will inherit and so on and so forth?

KEKRE: That’s absolutely right. That’s one of the biggest concerns with using this instrument, because there is literature—by the way, this reform was passed in five states in India during the ’70s and the ’90s. So, there is a ton of literature by Heath and Tan, by Calvi and by Roy, which shows that, as a result of this reform, they find that educational attainment of women increases. They increase their labor-force participation and they have greater autonomy within the household.

All of these effects basically point towards this concern that there can be direct effects of these reforms which do not go through land fragmentation, and they can be driven through supply-side changes in female labor-force participation. To address these concerns, what I do is—one thing is that I first restrict the sample to just males. These results that I’ve been telling you come from just a male sample, where these direct effects are less likely to be prevalent.

And there can still be some residual concerns that if women have higher bargaining power within the household through some intra-household bargaining, there might be some second-order effects on even male labor-supply decisions, right?

So, to address that, what I do is that the IHDS survey has this very nice questionnaire on gender relations, where it asks women questions about their autonomy within the household when it comes to taking decisions for the household, their financial autonomy, their physical mobility.

And so, I follow Heath and Tan’s paper, and I use the responses to these three questions to actually construct a district-level measure of female autonomy and how it changes in the IHDS survey from round one to round two. I control for that to see if these results still survive? What I find is that the results, the magnitudes, are almost the same even after controlling for this. So, this indirect channel might be less prevalent when we’re thinking about structural transformation in the male sample.

RAJAGOPALAN: Yes, and female labor-force participation in India, first of all, tends to be low. And a lot of it tends to be in exactly what you would expect driving the big part of the result, which is supporting the family farm and things like that, right?

KEKRE: Right.

RAJAGOPALAN: So now, the story just hangs together very cohesively based on everything we understand in the Indian context. 

Impacts on Manufacturing Employment

RAJAGOPALAN: You also have a couple of other results. This is between 2005 and 2013, how the decline in farm size changes manufacturing employment, the number of firms, and the decline in average firm size. Now, this is the part which is both quite surprising to me and very, very large. Can you walk us through how you went about measuring and finding that result?

KEKRE: Right, so, for this result, this was essentially my attempt at trying to test for one of those mechanisms that we were discussing through which structural transformation can be affected. I use the Economic Census of India, which records at the establishment level. It gives me information on how much labor is employed, the industry of the establishment, and . . . Yes, it’s not very detailed, but it gives me the important things that I would require for this analysis.

I aggregate this establishment-level data at the district level. What I see is the three measures that I track, as you said, is the total employment in the manufacturing sector in a district, the number of firms—which is basically establishments in the manufacturing sector—and the average firm size, which is the number of employees per firm.

RAJAGOPALAN: Yes.

KEKRE: And I run this through the same empirical strategy with two periods and the inheritance reform. What I actually find is—and even I was surprised to see how large these effects were. I was expecting them because of the presence of other literature on India that there should be some friction to the manufacturing sector, but it was surprising to see that land is such a big constraint. It translates into almost a 20 percent decline in the number of firms and a 30 percent decline in employment, and on average firms becoming smaller. 

Then, just to tie this all together in terms of, is it still picking up something else? Why are these effects so large? Are they picking some other effect? I did this heterogeneity exercise where I divide my establishments based on their industry affiliation as more land-intensive versus less land-intensive. Basically, depending on the median, the high-above-median land-reliant industries versus small. And I again aggregate my sample to the district level, and what I see is that these negative effects are larger for the high land-intensive industry. That reaffirmed this mechanism, that it is operating through the land-related frictions somehow.

RAJAGOPALAN: Well, not somehow. You’ve told us how.

KEKRE: [laughs] Yes.

RAJAGOPALAN: Land fragmentation is not exactly an unintended consequence in India, right? It was the intended consequence. Right? After the zamindari experience, for 50-plus years, we have tried our very best to make landholding smaller. And oftentimes, landholding got smaller because the governments were successful, and sometimes landholding size got smaller because of population growth and not having primogeniture and having the Hindu undivided family set up and so on and so forth.

But it’s not like this is a surprising, weird unintended consequence of a weird regulation. This was very much intended government policy. And now, we see the terrible effects of what that policy does. It’s a little bit hard to now undo time and expect that everything will consolidate overnight. The reason I say is you tell us exactly how this is happening.

Modeling this Problem

So, now, to step away from the empirical results you find, you then develop a general-equilibrium model, because now you’re finding these really large effects. The large effects are both about the share of agriculture versus manufacturing in the economy, but also about the wages, the differential in wages between agricultural wages and manufacturing wages, agricultural profitability, and overall welfare effects. 

So, first, can you walk us through how does someone model something like this for such a complex topic? And then, what sort of welfare effects do you find when you look at both transmission mechanisms?

KEKRE: Right. So, I tried to get to the welfare results using this two-sector, general-equilibrium model. I omit services from my model because I don’t find significant effects there. This model is essentially based on models by Fajgelbaum and Redding in the trade literature, and also by Heblich and Redding. These people, recently, have these papers in the trade literature, which looks at structural change, and the models are based on land allocation. That’s where I got my idea that, because I’m precisely trying to study this, this might be a good framework to begin with.

What is happening in this model is—and I’ll try to give a very intuitive picture of the model—within every district, there is a fixed supply of land. This land is held by some n number of landowners in fixed parcel size. I specifically in the model don’t account for any transactions of land between landholders because, as we see, that is not very prevalent in India. It’s not very easy for landowners to transact among each other.

So, I treat this parcel of land that’s an endowment to a landlord as an exogenous endowment. What I do with this is, essentially, the whole point of structural transformation in this model comes from the land allocation decision of the landlords. Each landlord has this land parcel, which has a distinct suitability for agriculture. You can think about it as differences in soil quality, fertility, which makes some parcels better suited for agriculture as compared to others, but each plot is equally productive in manufacturing.

From a landlord’s perspective, the landlord first observes the suitability of their land plot, and then they decide whether they want to put it to agriculture or whether they want to lease their land plot to a manufacturing firm. Because I modeled there’s this one representative manufacturing firm in each district which wants to take land from landlords for manufacturing production. So, basically, there is this land allocation decision that landlords need to navigate.

In the model, it’s a very simplified framework because I wanted to get to its larger implications for structural transformation. I just have two factors of production in this model. It’s just land and labor for now. In this model, I essentially don’t treat separately agriculture versus manufacturing wages, but it’s more of—the labor in this model is willing to supply its labor inelastically to whichever sector has demand. There’s going to be one common wage in the district, in the equilibrium. 

So, this is how land fragmentation enters the model. There’s a fixed stock of supply. You think about any shock or even the general process of land fragmentation. What it is going to do is it’s going to increase the number of landlords. Given the fixed supply of land, it basically changes the land endowment of each landlord. On average, the land fragmentation shock decreases the size of the land owned by a landlord.

And then, it alters the land allocation decision through precisely the two channels that we’ve been discussing so far. In the agriculture sector, as that land size becomes smaller, it changes the per-acre profits from that particular land plot. It alters the rents or the profits that a landlord can get from putting his own land into cultivation. And in the manufacturing sector, because of small farm sizes, these land acquisition costs are going to increase, which are going to reduce firm profits, as a result of which, the rents that landlords can receive from firms are going to decline.

This friction, I basically model this like a reduced firm wedge on the rental rate that firms can offer to landlords. This wedge, I model as a function of the district’s farm size such that when the average farm size declines, this wedge increases and it reduces the rent that landlords can get from the manufacturing sector. In a very reduced-form way, I’m trying to capture this entire story into the land allocation decision.

As the shock hits, what it does is, through both these sectors, it basically alters the relative rent a landlord can get from putting the land in manufacturing versus agriculture, and that shapes, or that drives, the land allocation process. Consequently, it shapes the labor allocation as well. Depending on which sector expands, there’s more labor in that sector, and that drives structural transformation as well in the model. That’s a big-picture overview of how the model is working in the setup.

One of the central exercises in this paper was essentially to estimate the elasticity with which, as farm sizes change, these two frictions in the manufacturing and agriculture sector shape up. What I did in this paper was, for the manufacturing sector, I used firm-level data from the Annual Survey of Industries. What’s good about this data is that it actually reports the value of new land purchases made by a firm.

Using that data, I’m able to estimate the elasticity with which these land acquisition costs change as districts’ farm sizes change. What I find is that between 2001 and 2011, on average, this land-assembly friction that is there in my model increases by about 35 percent.

RAJAGOPALAN: Wow!

KEKRE: It’s a very large increase. Similarly, then, using microdata from farms—from IHDS, which reports in detail the output and input expenditures at the farm level—using the panel structure of that data, I’m able to estimate that within the sample of households where the land size changes due to, say, inheritance. It’s more of an exogenous change in the farm size. In this subsample, I find that the profits per acre increase as farm sizes fall.

But this profitability advantage in my sample roughly, on average, increases by 1.1 percent, is what I find. So, then, when I take these elasticities, I use them to actually quantify the model.

What I still find is that when it comes to welfare—or maybe worse, the model does a good job of replicating the structural transformation result that I find in the empirics, which was reassuring, that it is very closely able to match the empirical estimates. It does show that structural transformation is hindered. But then when you look at welfare, which is the total real income in a district, which is the income of workers and landlords together, that actually goes up.

There is substantial variation. There are some districts that even lose. In fact, 50 percent of the districts in my sample experience a welfare loss, but 50 percent also gain. It’s about—a variation is from minus 0.8 percent to about 1.85 percent. There’s huge variation. But overall, the relationship is negative; districts that experience more fragmentation are experiencing overall welfare gains. I was overall a bit puzzled to think about, “OK, what is going on here?”

I then look at the distributional implications. What I find is that the two components that make up welfare are wages and land rents. What I find is that real wages are actually falling by about 1.92 percent on average. All of these gains are coming from land rents increasing, which increased by almost 5 percent. So, even whatever we are seeing as a positive effect is actually not overall should be construed as a good thing, because we are creating a very large distributional gap.

The workers, who in the model are landless, are essentially the ones who are bearing the cost of all of this land fragmentation, because manufacturing frictions prevent the expansion of manufacturing activity. It’s a fact that manufacturing is relatively more labor-intensive than agriculture. When the more labor-intensive sector shrinks, it reduces the aggregate labor demand. It reduces the wages in the model. That was the key takeaway for me from this model, that, yes, we do find some positive effects, but there are these very sharp distributional implications that we should be careful about.

RAJAGOPALAN: Yes. Just common sense tells us, even without taking sides, there are more workers than there are landowners. [chuckles] 

KEKRE: Yes.

RAJAGOPALAN: Right? Just simply put. What you’re now trying to parse through is there’s going to be a small number of people who see these huge gains, and there’s a large number of people who are actually trying to exit agriculture and move into slightly more productive sectors of the employment who simply can’t find employment in those sectors.

KEKRE: Right.

Potential Policy Implications

RAJAGOPALAN: Which also tells you why there are so many people vaguely hanging around agriculture, underemployed, sort of in there, but don’t end up finding manufacturing jobs or moving into anything more productive. So, overall, again, the story hangs together. I think it’s a cautionary tale, the way you’ve walked us through the model and these distribution effects, that we should not look at headline results. [chuckles] 

KEKRE: Yes.

RAJAGOPALAN: We should really parse through what’s happening. So, overall, I find that your structural transformation story hangs together, both empirically and in the model. Now, what does this mean for policy reforms, or at least the most immediate policy reforms that you can think of when it comes to land?

KEKRE: Right, so, I think to get to that, one exercise that I do in the model is I start by switching off all land-assembly frictions and manufacturing in the model. This is the benchmark case where there are just no frictions for assembling land for manufacturing. When I do that, what I find is that these welfare gains that were coming from agriculture due to land fragmentation, they almost increase by 50 percent. The average effect on the welfare increases by 50 percent. 

The better result is that even the distributional gap closes. Because now what is happening is that you’re allowing [the] manufacturing sector to relatively not contract by as much as it was in the baseline. It does not shrink as much, which means that the demand for labor and the wages are also not going to fall by as much. Now, in this counterfactual, where there are no land-assembly frictions, real wages just declined by about 0.17 percent on average, which was 1.9 percent in the baseline.

Even the real rents don’t go up by as much. Basically, we are closing the gap between the owners of labor and the owners of land. That was a very interesting finding. That’s when I would like to circle back the discussion to even your [Rajagopalan et al] policy paper on the ease of converting agricultural land to nonagricultural use, that it points towards working in that direction. I’m not saying that we completely eliminate fragmentation because that’s a bigger political economy issue to deal with. You can’t so easily change.

RAJAGOPALAN: Yes, but you’re hinting towards the need to reduce frictions. Some of those frictions come from changing land use. Some of those frictions come from assembling parcels. Some of those frictions come from inability to sell to a farmer or a nonfarmer and so on and so forth. So, then all your policy recommendations would be targeted towards reducing frictions in converting land use, either by changing the user or the use.

KEKRE: Correct. As your paper mentions, there are six to eight policy hurdles on average to convert your land to nonagricultural use. Just a very utopian world where it’s a one-stop destination to converting your land. Then even if I have, as a firm, I’m supposed to do that process with 10 landholders, it will significantly reduce the cost of doing that rather than doing these eight processes eight to 10 times with 10 landowners. That was my first policy takeaway from this paper, that any efforts that can reduce these land-assembly frictions are very important in this context, both distributionally as well as when we think about the aggregate district-level welfare.

RAJAGOPALAN: Yes.That sounds simple, but that’s actually pretty big, right? Because we don’t have well-functioning land markets. Part of that is the frictions. The consequence of that is also the frictions. So, the big takeaway is we need to systematically move towards better-functioning land markets, because then consolidation can happen through the market with greater ease. We don’t need to rely on eminent domain or political brokers—we are basically pointing a gun at people [chuckles]—

KEKRE: Right. Exactly.

RAJAGOPALAN: —and forcing them to consolidate their land. A lot of that can be avoided. I think we’re on the same page when it comes to some of the policy recommendations.

KEKRE: Right, right. Exactly. Another thing that I wanted to pinpoint was that there is this discussion on what really the small-farm profitability advantage means. There’s a lot of discussion in the literature that it could also just reflect some of these size-dependent distortions in markets, which enable small farms to have larger profits per acre. If these are [the] result of some market distortions, that also points us towards how carefully or how cautiously you should interpret the welfare results. Because what we’re essentially talking about is that in my model, when I shut down this profitability advantage, I see the results flip completely.

With greater land fragmentation, welfare actually goes down. Basically, what this exercise shows is that this profitability advantage is an important countervailing force against this very large manufacturing friction. To the extent that this profitability advantage is a real resource gain or is it just reflective of some distortions in the input markets, we again need to be cautious about promoting small farms as a policy. Although my paper is not taking a stance on decomposing this, and whether this is a resource gain or whether these are distortions, but that’s something that should be interpreted with caution, is definitely something I would say.

RAJAGOPALAN: Yes, and also interpreted with common sense, wouldn’t you say? [laughter] There is this story across the world that things get more productive as fewer people work in agriculture, and there’s more mechanization, and there is larger holding size.

KEKRE: Right.

RAJAGOPALAN: I see no reason to think of India as some unique outlier in this story. I literally have no reason to believe that. It may be true for very specific things, particular kind of organic farm that produces a particular kind of mango or something like that, but largely, when we’re talking about regular agriculture, scale effects are everything. Ambedkar was saying this [chuckles] 120 years ago.

People are still saying this now. But, yes, we do have a very romantic notion attached to the small farmer and the small landholding, a family farm. I think some of this comes from the colonial experience and otherwise, but I’m pretty glad that you’re helping us overall dispel these crazy myths. 

Thank you so much for doing this. This was such a pleasure. The paper’s fantastic. It’s really, really rigorous and very important. Because we just see what a big barrier land has become to India’s structural transformation, which is really the only ladder we have towards greater economic growth and development in India. Thank you so much for doing this.

KEKRE: And a very long ladder to climb, but yes.

RAJAGOPALAN: Yes, for a very large number of people.

KEKRE: Thank you so much. It was really fun engaging with you on the job market paper. You and I learned a lot from this conversation.