Rethink Priorities has been piloting expanding into human-focused neartermist global priorities research. This post is one of three outputs from the pilot program. Open Philanthropy provided funding for this project and we use their general frameworks for evaluating cause areas, but they do not necessarily endorse its conclusions. We are open to revising our views as more information is uncovered.

Key Takeaways

• Agricultural land redistribution is a type of agrarian reform in which large farms are broken up and distributed to tenants or landless peasants.
• Land redistribution typically requires exceptional circumstances to succeed. Past redistributive efforts have been most successful in the aftermath of revolution, war, or independence.
• When redistribution has succeeded, it has been accompanied by extensive agricultural support, such as rural infrastructure development, subsidies for fertilizers and high-yield seeds, agronomic training, and cheap credit.
• The main value of redistribution appears to be improved agricultural yields, but redistribution is neither a necessary nor sufficient condition for improved yields.
• The exact mechanism by which redistribution improves yields is unclear. Redistribution creates smaller farms, and there’s some evidence that smaller farms have higher yields. (A common theory to explain this is that small landowners face better incentives than landless tenants, so they provide more labor and invest more in improving yield.) However, it’s also possible that redistribution in isolation provides little benefit to yields. Recent evidence suggests either a U-shaped or flat relationship between yields and farm size, suggesting that consolidation could be as attractive an option (for improving yields) as redistribution.
• There appear to be no large and reputable NGOs that advocate for radical land redistribution. NGOs that promote tenure reform (another type of land reform) might be convinced to advocate for redistribution.
• Our weakly held belief is that South Asia is the best region to pursue land redistribution.
• We believe that advocating for agricultural land redistribution is unlikely to be a cost-effective intervention and very intractable.

Executive Summary

Land reform encompasses two distinct topics: tenure security and land redistribution. The former mostly involves formalizing a preexisting (hitherto informal) ownership structure. The latter is about modifying who owns the land itself. This report concerns the latter.

Our view is that agricultural land redistribution is not a cost-effective way to alleviate poverty. We think that if land redistribution were done well, it could be a high impact intervention for kickstarting growth. Our pessimism mainly comes from our belief that redistribution of the kind described in Joe Studwell’s How Asia Works is intractable.

The thesis of How Asia Works is that differences in agricultural, industrial, and financial policies across countries in East Asia after the Second World War explain the developmental differences between the fast-growing northeast (South Korea, Japan, China, and Taiwan) and the slow-growing southeast (especially the Philippines, Indonesia, Thailand, and Malaysia). In the northeast, wealthy landowners were dispossessed and their land was redistributed across equally-sized smallholder farms. Combined with rural infrastructure investment, extension services, and generous subsidies, the northeast saw crop yields spike. In the southeast, where land redistribution was either not attempted or not underpinned with adequate agricultural support, yields grew much more slowly or stagnated. The agricultural surplus in the northeast boosted savings and investment, raised net exports (increasing foreign exchange available for importing capital for manufacturing), and created a positive rural consumption shock that powered demand for domestically produced manufactured goods. This economic momentum was sustained with industrial policies that nurtured local firms but forced them to compete abroad and financial policies that kept lending focused on long-term growth.

Although we believe Studwell overstates the importance of small farm sizes for producing higher yields, his overall characterization of the Asian growth miracle seems to match the perspectives of other scholars in the area we spoke to and read. Replicating this success in other regions, however, appears unlikely.

For redistribution to make sense as an intervention, there must be an initial unequal distribution of land. But wherever there exists a concentration of land ownership, there also exists a concentration of political power opposed to land redistribution. Land redistribution was possible in northeast Asia after World War Two because the political groups that controlled the land had been violently overthrown. This sort of dynamic does not exist at the moment as far as we can tell.

To overcome entrenched political opposition, land redistribution needs fierce champions. Unfortunately, there don’t appear to be any major, reputable NGOs that promote radical redistribution. For reasons that are not entirely clear to us, the political headwinds against redistribution have grown significantly since northeast Asian land redistribution. Because redistribution is a highly charged issue, advocacy requires an understanding of the local context. Hence, although land redistribution appears to be very neglected, creating an advocacy ecosystem from scratch looks difficult.

Furthermore, we’re very uncertain how much value redistribution provides in isolation. The Asian growth miracle was not a product of land redistribution alone. Studwell makes clear that smart industrial and financial planning was essential to success in Japan, Taiwan, South Korea, and China. Moreover, the redistribution effort was not merely a matter of handing out free land to peasants. Governments in northeast Asia also provided agronomic training, rural infrastructure development, crop storage, price floors, fertilizer and seed subsidies, cheap credit, and marketing support. Backing out the role that redistribution itself played is thus difficult, and implementing additional supports as part of the land reform package further decreases the tractability and increases the cost.

A key argument for redistribution is that smaller farms allegedly produce higher per-acre yields resulting in a higher total output. Although this appears to have been the consensus in development economics for several decades, the past decade has cast considerable doubt on the claim. The most recent empirical evidence appears to suggest that there is either a U-shaped relationship between farm size and yields with both large and small farms outperforming medium-sized farms, or there is no negative relationship between farm size and yields. This suggests that consolidation of land could be as good a strategy for improving yields as redistribution to smallholders.

Our weakly held view is that South Asia is the best region to attempt land redistribution today. South Asia exhibits a high degree of land inequality, a large (and in some areas politically influential) class of poor farmers, and a modest history of land reform.

Ontology of Land Reform Concepts

The term ‘land reform’ means different things to different people. In this report, we have adopted the terminology of the World Bank as exemplified in their 2009 report Agricultural Land Redistribution: Toward Greater Consensus. According to this usage, ‘land reform’ is an umbrella term that encompasses two main components: land redistribution and land tenure reform. In most cases, the land in question is agricultural, situating land reform beneath the broader category of agrarian reform.^[1] Land reform of this kind is typically pursued in low- and middle-income countries.

In this context, land reform ought to be clearly distinguished from land use reform. ‘Land use’ often refers to the way local governments zone (or otherwise restrict) the way that land is used. ‘Land use reform’ generally refers to a loosening of those restrictions, often with the goal of increasing housing density in high-wage metropolitan areas.^[2] Land use reform is typically pursued in high-income countries.

Land Redistribution

According to the nomenclature we are following, “Land redistribution is an effort by governments to modify the distribution of land ownership. It is often an attempt to transform an agrarian structure composed mainly of large-scale farms into one where family farms are predominant by taking land away from large landowners, or the state, and redistributing it to tenants and landless peasants” (Binswanger-Mkhize et al. 2009: 4).

Land Tenure Reform

According to the nomenclature we are following, “Land tenure reform is a program designed to change the legal and institutional framework for land administration. Other common changes attempted by land tenure reform programs include modification of the land tenure system and decentralization of the land administration and management function” (Binswanger-Mkhize et al. 2009: 4). Many NGOs refer to land tenure as ‘land rights’. Tenure reform is a much more prominent aspect of the current land reform ecosystem—redistribution has mostly fallen off the political agenda—and for that reason some NGOs appear to use ‘land reform’ as synonymous with ‘tenure reform.’

Scope of this Report

In this report we investigate the story in Joe Studwell’s 2013 book How Asia Work, assess its plausibility, and evaluate the potential for replicating the program outlined in the book in other parts of the world. The central mechanism and focus of How Asia Works is land redistribution. Tenure reform plays a smaller role in the book.

It’s important to recognize that we have only investigated one type of land reform. Although we briefly discuss tenure reform, our main focus is land redistribution. Tenure reform may merit its own investigation. Gignoux et al. (2014) and Lawry et al. (2017) are systematic reviews of the literature on land tenure interventions, but we haven’t looked into them.

How Asia Works

According to Joe Studwell’s 2013 book How Asia Works, in the period immediately following World War Two, many Asian countries had an unequal distribution of land, with a handful of wealthy landowners and a vast underclass of tenant farmers. This system was both unfair and inefficient. Countries in northeast Asia (Japan, South Korea, Taiwan and China) embarked on a program of radical land redistribution. Large landholders were forced to sell their plots, often below market rates, and a (quality-adjusted) 3 hectare limit on agricultural land was introduced. In addition, governments provided a host of subsidies and support, from agronomic training to price floors to rural infrastructure development.

Empowered to (literally) reap what they sow, the new smallholders worked their own land with much higher labour intensity, resulting in much improved yields ($yields=output/land$), but at the cost of lower productivity ($productivity=output/labour$). Redistribution also corrected misaligned incentives that had discouraged large landholders from investing in yield-boosting land improvements. Together, these changes generated an agricultural surplus that, along with smart industrial and financial policy, catalyzed broader economic growth and helped lift millions of individuals out of poverty in the area during the second half of the twentieth century.

Throughout, Studwell contrasts the rapid growth in northeast Asia with the much slower growth in southeast Asia (especially the Philippines, Indonesia, Thailand, and Malaysia).^[3] (See Figure 1.) According to Studwell, the different outcomes can be traced to differences in agricultural, industrial, and financial policy.

Figure 1: Northeast and southeast Asian economies were once in a similar position but have since diverged considerably

How Asia Works is divided into three^[4] main sections. The first section is devoted to agriculture. The second section is devoted to manufacturing. The third section is devoted to finance. The three sections detail the reforms in each area that northeast Asia got right and that southeast Asia got wrong.

In our report, we focus on the first section. However, it’s important to note that Studwell makes clear that the three areas are interrelated. In northeast Asia effective agricultural land redistribution provided the agricultural output surpluses necessary to power an export-oriented manufacturing boom. The manufacturing sector was successful because governments in northeast Asia adopted the right kind of trade and financial protectionism that sheltered domestic industries from the whiplash of short-term returns, encouraged exports of high quality goods and put those industries on a path to long-term growth through technological learning. In conversation, Studwell stressed that there are no developmental magic bullets. Agricultural land redistribution in isolation is not going to unlock the sort of economic growth that northeast Asia experienced in the second half of the twentieth century.

The Problem

According to Studwell, a relatively small number of elites owned the majority of productive agricultural land in pre-War northeast Asia.^[5] The problematic effects of inequality were exacerbated by historically unprecedented population growth.^[6] Because Japan, Korea, and Taiwan are mountainous regions with relatively little arable land,^[7] the supply of cultivated land could not keep up with demand. The mismatch between supply and demand created opportunities for landowners to increase profits without increasing agricultural yields. Rather than farm the entirety of their land themselves, landowners leased large portions to take advantage of rising rents.^[8] The landowners also served as moneylenders to their tenants, often at unsustainable interest rates.^[9] Saddled with high rents and expensive debt, tenants were forced to sell their crop as soon as it was harvested, while their landlords stored their harvest and sold it later when prices were higher. When tenants inevitably defaulted on their loans, the landowners recovered the leased land and rented it to others. With little tenure security and no spare capital, tenants were unwilling and unable to make improvements to the land that could increase yields. Landowners had the capital to make such improvements but chose not to because they had found a business model that was more profitable.

As Studwell puts it (2013: 3),

The problem with agriculture in pre-industrial states with rising populations, however, is that when market forces are left to themselves agricultural yields tend to stagnate or even fall. This happens because demand for land increases faster than supply, and so landlords lease out land at increasing rents. They also act as money lenders at very high rates of interest. Tenants, facing stiff rents and expensive debts and with little security of tenure, are unable to make the investments ‒ for instance, in improving irrigation or buying fertiliser ‒ that will increase yields on the land they farm. Landlords could make the investments to increase yields, but they make money more easily by exacting the highest possible rents and by usury, which adds to their land holding when debt cannot be paid and they take over plots that have been pledged as collateral. A situation arises where ‘the market’ fails to maximise output (3).

We believe this narrative, but we don’t fully understand what landlords being able to make money ‘more easily’ means in economic terms. We would be keen to see if this narrative (where landlords profit maximise by following a rent-exacting business model instead of a yield-improving business model) could be replicated in a formal theoretical economic model where the landlord is a profit maximiser with time-consistent discounting (e.g. not hyperbolic discounting) and how the results change depending on the population growth rate.

The Solution

The market failure described above demanded bold and comprehensive reform. The turmoil of the Second World War and subsequent civil strife afforded the political opportunity to enact such reform. The solution that both Communist and Western-backed governments landed on was radical land redistribution coupled with extensive support for smallholder farming.^[10]

The first step was to acquire agricultural land from large landowners. In Communist China, this process was often bloody, with Studwell reporting that “estimates of the death toll related to land reform in China range from hundreds of thousands of people to several million” (20). In Japan, Taiwan, and South Korea, the process was much less violent^[11] in large part (according to Studwell) because the process was imposed by the United States.^[12] Studwell quotes agricultural historian Ronald Dore (Dore 2013: 172-3): “‘The very fact that it [land reform] was imposed from outside was a powerful factor in making the land reform a peaceful and orderly one. Tenants could take over the land, not with the light of revolution in their eyes but half-apologetically, as if it hurt them more than it hurt their landlords, for the cause was not in themselves but in a law for which they bore no responsibility either personally or collectively’” (25-26).

Although the U.S. administration mandated the redistribution, they sensibly devolved control over much of the process to local stakeholders. In Japan, “The critical mechanism to implement this was the creation of land committees on which local tenants and owner-farmers outnumbered landlords. These committees had the adjudicating power in what for landlords was a distinctly painful process” (25). In theory, landowners were compensated for their appropriated land, but in practice, the bonds with which they were paid amounted to a price far below market value.^[13] The acquired land was then equally (on a quality-adjusted basis) divided amongst the farming population. To prevent the reaccumulation of land, strict size limits, generally around a maximum of 3 hectares, were enacted and the land committees were required to review and approve transactions that would shift land from the former tenants back to the old landowners.^[14]

Land redistribution wasn’t the whole story, however. Studwell makes clear that “in order to thrive, smallholders require not only their fields, but also the extension, marketing and credit infrastructure that allows them to compete” (56). He emphasizes that “in every instance where it succeeds, land reform was coupled with state investments in rural infrastructure, agricultural extension services and marketing support” (32). In northeast Asia, governments provided agronomic training, subsidies for fertilizer and high-yield seeds, access to cheap credit, and price floors to protect farmers against market fluctuations.^[15]

This combination of reforms had several effects. It destroyed the usury and rent-exacting business model that had discouraged landowners from maximizing yields. Price floors ensured a minimum income, thereby improving tenure security. Cheap credit reduced the risk associated with capital improvements. Durable ownership incentivized long-term investment in the land, such as irrigation. Ownership also incentivized farmers to saturate their land with labor that could not be productively applied elsewhere. (Northeast Asia had no shortage of unskilled labourers at the time.^[16]) Finally, the smallholder system reduced labor costs by lowering monitoring and transaction barriers.^[17] (See below for more discussion of the advantages that small farms might enjoy over medium-sized farms).

The package of reforms didn’t maximize productivity or return on investment. Instead, they achieved something Studwell considers much more important for countries in the early stages of development: the reforms maximized yields.^[18] As Studwell puts it,

the question of efficiency depends on what outcome you are looking for. Big capitalist farms may produce the highest return on cash invested. But that is not the agricultural ‘efficiency’ that is appropriate to a developing state. At an early stage, a poor country with a surfeit of labour is better served by maximising its crop production until the return on any more labour falls to zero. Put another way, you might as well use the labour you have ‒ even if the return per man hour looks terribly low on paper ‒ because that is the only use you have for your workers (4).^[19]

The Result

According to Studwell, radical land redistribution and support for smallholder farming “produced two enormously beneficial effects that could not have been achieved through other policies” (65). First, the reforms produced an impressive increase in agricultural yields. Studwell writes, “In the first ten to fifteen years following the shift to small-scale household agriculture in successful east Asian states, gross output of foodstuffs increased by somewhere between half (in Japan, which was already the most productive country) and three-quarters (Taiwan)” (8).^[20] This jump in output increased gross savings and made more money available for industrial investment. The higher yields also increased net exports by reducing food imports or increasing food exports, which in turn increased the amount of foreign exchange available for purchasing capital goods needed for manufacturing. Successful agriculture also created a social safety net of sorts: during periods of economic downturn, rural-to-urban migrants could return to their family farms if they were laid off from factory jobs.

Perhaps most importantly, the agricultural surplus that the reforms generated “created a kind of ‘consumption shock’ as waves of spending power for basic, domestically manufactured consumer goods spread through the economy” (65). As Studwell observes in the subsequent section of the book (on manufacturing), governments in northeast Asia had adopted industrial policies to nurture their nations’ nascent manufacturing sectors. Because of this protection, local businesses were able to benefit from increased rural purchasing power. Studwell notes that:

Famous east Asian corporations from Meiji Japan to post-war Korea and contemporary China made their first millions adapting products to the exigencies of extensive but cash-limited rural markets. Local firms learned critical lessons about marketing from rural populations with whom they had a natural cultural affinity. Examples from Japan include Toyota and Nissan building robust cars for unpaved roads on small truck chassis after the Second World War, or Honda’s early 50cc engines being used to convert cycles into motorcycles (8).

According to Studwell, the second major benefit of land redistribution “was separate from the output and consumption shock effect, but combined with it to produce yet more economic virtue. It was the creation of a high level of social mobility as the result of an equal distribution of society’s most basic non-human asset – land. People not only competed on equal terms but they could realistically believe they had a chance of success” (65). Throughout How Asia Works, Studwell observes that although U.S. commanders on the ground in northeast Asia pushed for land redistribution, leadership in Washington remained skeptical, frequently complaining that redistribution smacked of communism. Studwell argues this concern was misplaced. On his view, land redistribution created a capitalist paradise: an almost perfectly level playing field in which profits were determined solely by one’s efforts and ingenuity. The newly-leveled playing field increased economic efficiency.

The level playing field may have had another effect. At various points, Studwell gestures at the potential for effective land redistribution to reduce violent conflict. This claim has also been made elsewhere, e.g., Binswanger-Mkhize et al. (2009: 7). In the book Studwell cites the Philippines as an example where landlessness exacerbated unrest (pp. 35-37). In conversation, Studwell also mentioned contemporary Pakistan and Kenya circa 2007 as examples where an unequal distribution of land has contributed to violence. Interestingly, there is some recent empirical evidence (Albertus 2019) that greater redistribution was associated with lower conflict intensity in Peru during the Shining Path insurgency (1980-2000).^[21] We haven’t vetted this study or otherwise attempted to quantify the extent to which redistribution mitigates violent conflict.

Other Aspects of How Asia Works

Studwell is clear that land redistribution alone, even when successfully implemented and supported by the right ancillary services, cannot transform a low-income country into a high-income country. In his words, “There are no one-stop solutions to economic progress” (219). At best, redistribution enables a country to transition to the next phase of development, which brings its own challenges and opportunities.^[22]

Historically, the next phase of development has relied on manufacturing and trade.^[23] According to Studwell, state support (through protection and subsidies) is crucial to success in this stage. While governments in both northeast and southeast Asia attempted to nurture their domestic industries, the countries differed in three key respects. First, governments in the northeast encouraged firms to export goods and governments measured firms’ success by their ability to export. Studwell calls this “export discipline.”^[24] Encouraging firms to compete on the global market kept the firms innovative in a way that the protected domestic market never could. Second, governments in the northeast were more interested in culling losers (by letting them go bankrupt or forcing them to merge with other firms) than they were in picking winners.^[25] Identifying underperforming firms is easier, Studwell argues, than identifying standout firms. Third, governments in the northeast helped firms acquire foreign technology. Often this meant denying market access to a foreign firm unless they partnered with a local firm and shared knowledge with it, which might be more difficult today under World Trade Organization rules.^[26] Of these policies, Studwell claims that export discipline was “by far the most important.”^[27]

In the third major section of the book, Studwell contends that agricultural and manufacturing development require financial policies tailored to promote long-term growth,^[28] even at the expense of short-term returns.^[29] Studwell argues that northeast Asian economies were successful because, in his words, they kept financial institutions “on a short leash.” The central mechanism through which governments exercised this control was through a central bank controlled by the government and charged with preferential lending “for exports and for industrial upgrading, with the system policed through requirements for export letters of credit from the ultimate borrowers” (218-219). Capital controls were also needed because in their absence, international investors will seek the most immediately profitable investments, which often don’t align with a nation’s long-term developmental objectives.^[30]

We haven’t closely vetted the story Studwell tells in the sections on manufacturing and finance. It’s clear, however, that today’s developmental differences between northeast and southeast Asia cannot be traced wholly to differences in land allocation or agricultural policy. While we are fairly confident that the counterfactual impact of land redistribution in northeast Asia is positive, we are very uncertain about the magnitude of the impact. We are even less sure about the counterfactual impact of land redistribution if it were implemented in a new region today. Land equality is certainly not a sufficient condition for transformational economic growth; we suspect it is not a necessary condition either. (An agricultural surplus may be necessary for transformational growth, but land equality does not seem to be a necessary condition for generating an agricultural surplus.)

In our opinion, there are two other broad factors that might play an important role that Studwell neglects. The first is state capacity. The sort of policy reforms Studwell recommends in agriculture, industrial policy and financing seem to require a large number of competent bureaucrats to implement them. Perhaps the biggest barrier holding southeast Asian and other countries from implementing successful land reform was the lack of state capacity, rather than a lack of political will.

The second is education. In the introduction Studwell rejects education playing a major role, saying, “the evidence of a positive correlation between total years of education and GDP growth is much weaker than most people imagine” (xxii-xxiii) and points to South Korea and Taiwan having lower educational capital than average, whereas the Philippines and Cuba have high literacy and education levels but haven’t experienced huge growth.^[31] However, our impression is that the received wisdom amongst economists is that the Asian Tigers (South Korea, Taiwan, Hong Kong, & Singapore) had education and human capital levels higher than would be predicted given their income levels at the time and high levels of human capital played a significant role in their growth (see figure 2 below from Page 1994: 227). Education could play a role in improving state capacity discussed above and making both agricultural and manufacturing workers more productive.

We haven’t looked deeply into the case for these alternate factors being more important than Studwell discusses, but they are often considered important amongst development economists and would be worth considering in a deeper investigation.

Figure 2: Education against GDP per capita with East Asian states labelled from Page (1994: 227)

Where Else Would Land Redistribution Be Desirable?

We think there are four main reasons for land redistribution being successful in northeast Asia:

1. Initial unequal distribution of agricultural land
2. Excess low-skilled labour with limited opportunities in sectors outside of agriculture
3. Governments willing and able to provide the necessary ancillary agricultural services
4. Political will to push through land redistribution unpopular with elites

Contemporary land redistribution is more likely to be impactful and tractable in other countries if these four factors are also present. We briefly discuss each of them in turn.

Unequal land distribution

Studwell says that the initial distribution of land in northeast Asia was very unequal, but doesn’t provide precise statistics on this. We were able to find some statistics in Walinsky (1977) (a collection of Wolf Ladejinsky’s essays and reports; Ladejinsky is one of the ‘heroes’ of How Asia Works for his role in advocating for land reform) but the sources of these statistics are unclear and we haven’t attempted to confirm their credibility.

In 1937 Japan (p. 71), the top 3% of the population owned 29% of the land. The bottom 28% of farmers owned no land. The bottom 50% of farmers owned 18% of the land, the bottom 75% owned 34% and the bottom 97% owned 71%. In 1938 Korea^[32] (p. 50), the top (less than) 4% of households were landlords who owned 54.5% of the arable land. The bottom 46% of farm households owned no land at all and had to rent. In 1949 Taiwan (p. 96), the top 3% of households owned 36% of all the land, while the bottom (almost) 60% owned 14%. We plot these figures in Lorenz curves and compare them to modern day distributions in Figure 4 below.

So how do these figures compare to land distribution in the world today? Bauluz et al. (2020) study the distribution of agricultural land across LMICs, focusing on reporting top 10% and bottom 50% shares.^[33] They compare the results from agricultural census and household surveys and show that the results from different data sources are highly correlated. In figure 3 (below), they show that land inequality is greatest in South Asia and Latin America “with top 10% landowners capturing up to 75% of agricultural land and bottom 50% owning less than 2%”. African countries have less inequality, with their figures being more comparable to current day China and Vietnam. Studwell counts China as having a successful land reform but doesn’t discuss Vietnam in the book. This weakly suggests that successful land redistribution in South Asian or Latin American countries may have more impact than in African countries.

Figure 3: current regional land inequality

A couple of caveats are worth noting. Firstly and most importantly, these averages hide heterogeneity; see table 1 below for the country level figures, where we can see that Ethiopia has inequality levels closer to South Asia and Latin America. Secondly, they only have a few countries per region. Given the heterogeneity within regions and the fact that any land redistribution intervention will almost certainly be limited to one country at most, using the country as a unit of analysis rather than the region is more informative. Thirdly, they don’t currently include communal land which isn’t held as private property but could play a similar role as redistributed land, and our impression is that this is more common in African countries. Fourthly, northeast Asian countries typically placed 3 hectare limits on the size of farms so it would be good to see the proportion of farms that are above and below this threshold but we can’t back this out from the figures in the paper.

Table 1: current country-level land inequality

We take the statistics from Walinsky (1977) and Bauluz et al. (2020) and plot them as a set of Lorenz curves in figure 4 below (data here). The further away the slope is from the dashed 45° line, the more inequality there is. As the calculation of land distributions uses different methods in each paper and we have different sets of percentiles for different countries, we should be cautious when comparing inequality across countries. However, if we do naively compare them, it appears that many countries today have land inequality equal to or higher than the northeast Asian countries historically, which suggests that successful land reform in these countries could have a large impact.

Figure 4: Lorenz curves of land inequality for northeast Asian countries pre land reform and other countries recently

Excess low skilled labour

Studwell focuses on yields (output/land area) instead of productivity (output/labour). He writes (p. 4), “At an early stage, a poor country with a surfeit of labour is better served by maximising its crop production until the return on any more labour falls to zero. Put another way, you might as well use the labour you have ‒ even if the return per man hour looks terribly low on paper ‒ because that is the only use you have for your workers.” Once other sectors such as manufacturing have developed and workers have good outside options where their marginal product of labour is higher, focusing on increasing agricultural yields becomes less attractive. This means that when choosing countries or regions to support with agricultural land reform, you want to pick places that are still early in their developmental path and still have high levels of labour market inactivity, whether through unemployment, underemployment, or exiting the labour market.

Excess labour supply in rural areas of low income countries is one of the most notable stylised facts in development economics. Multiple theoretical models central to development economics, such as the Lewis dual-sector model and the Harris-Todaro model start with the assumption that there is a surplus of labour in the agricultural/rural/subsistence sector with low or zero marginal product of labour. Economists typically use the share of labour in agriculture as a proxy for whether there is excess labour supply. This correlates strongly with GDP per capita. Our World in Data has good visualizations of the current share of the labor force employed in agriculture across the world. Sub-Saharan Africa and South Asian countries typically have the highest shares.

Figure 5: Current shares of labour force employed in agriculture

Studwell also emphasises the fact that northeast Asian countries had rapidly growing populations at the time. This demographic dividend results in a larger share of the population being of working age and therefore having more excess labour to boost agricultural yields (and later move into manufacturing). As such, ceteris paribus, it would also be better to target agricultural land reforms in countries with high population growth rates. Our World in Data show that sub-Saharan Africa is the region with the current highest population growth rates and it is projected to have higher growth rates for the rest of the century.

Figure 6: Current population growth rates

Ancillary agricultural services

As Studwell makes clear in the book and reiterated during our conversation with him, land redistribution by itself is not a magic bullet. The reforms in Japan, China, South Korea, and Taiwan were accompanied by many forms of agricultural aid. In Japan, “The government spent heavily on rural infrastructure, offering scores of different subsidies and grants-in-aid to farmers and providing an average of one agricultural extension worker per village. Relatively low-interest credit was also offered through village co-operatives” (26). For South Korea, Studwell writes (63-4):

South Korea moved to increasing subsidy and protection of domestic agriculture in the 1970s and 1980s. The government bought larger and larger quantities of rice at above-market prices, which it then sold on to consumers at a discount of up to 50 per cent. A state monopoly on fertiliser supply, which early in the country’s development was used to squeeze money out of farmers by charging them above-world-market rates, was redeployed to further subsidise input costs by providing fertiliser cheap… almost every crop benefited from import protection. Of the 547 standard product categories that faced quantitative import restrictions in Korea by the late 1980s, all but a few dozen were agricultural products.

For Taiwan, Studwell writes (32):

As in every instance where it succeeds, land reform was coupled with state investments in rural infrastructure, agricultural extension services and marketing support. The JCRR [Joint Commission on Rural Reconstruction] … was hugely important in supporting these initiatives in Taiwan. By one estimate, the agency handled one-third of US aid to the island between 1951 and 1965, running 6,000 projects and accounting for a little over half of net investment in Taiwanese farming. Through it, Taiwan acquired the world’s proportionately largest cohort of agricultural research and extension workers.

During the period covered in How Asia Works, agricultural support in these countries amounted to about 20% of public expenditure. Today, the average level of agricultural spending in African countries is 6% of public expenditure (Lipton 2020: 29). Where land redistribution has been attempted without accompanying agricultural support, it has generally failed. Studwell writes (7):

The arguments about the efficiency of small-scale farms are not without their complexities. The very high yields achieved in Japan, Korea, Taiwan and China are not simply the outcome of farm size, but of farm size combined with complicated infrastructures that have been set up to deliver inputs like fertilisers and seeds, and to facilitate storage, marketing and sales. Without adequate supporting infrastructure, small farms struggle anywhere, as has been the case after failed land reform attempts in places like the Philippines.

A quantitative study of land reform in the Philippines supports that it was a failure, finding that land redistribution actually decreased productivity by ~17% (Adamopoulos & Restuccia 2020). Note that this is referring to productivity (output/labour) rather than what Studwell typically focuses on, yields (output/land area). The causal identification in this paper comes from an assumption-heavy structural model so it is complicated to assess the credibility, but nonetheless, the descriptive statistics (see figures 7 and 8 below) make a compelling case that productivity decreased as a result of the reforms. In their table 10, they implement a basic difference-in-differences approach and find that yield (land productivity) fell in the farms subject to redistribution but it isn’t statistically significant.^[34] The authors don’t include agricultural supports in the model so we can’t get estimates of productivity with and without supports. They also don’t claim that the reform would have been successful had these supports been included in the reform package, so while they show that land reform in the Philippines was indeed not particularly successful, they don’t show that it would have been more successful with these additional supports.

Figure 7: Agricultural labour productivity in the Philippines

Figure 8: Changes in farm-size distribution from the Philippines census

Keswell & Carter (2014) study the income impacts of the South African Land Redistribution for Agricultural Development (LRAD) program. “LRAD makes land purchase grants to landless farm workers and labor tenants. The program does not mandate redistribution of land from rich to poor, but rather operates through markets on a willing buyer–willing seller basis” (p. 251) and the authors don’t discuss any accompanying agricultural support policies. They find that three years after receiving the LRAD treatment, household consumption increases by 50%, from $2 a day to $3. They highlight a comparison to the results of Behrman et al. (2005) who supposedly find that Mexico’s conditional cash transfer program increased living standards by 5% (but we haven’t looked into this other paper at all).

Keswell & Carter (2014) claim that the LRAD application process creates exogenous variation in how long an individual receives treatment for and that they can therefore compare people who received land a while ago with those who just received land.^[35] This is a relatively strong identification assumption which depends a lot on the details of the LRAD process which we haven’t looked into in detail. At the moment we would discount the results by 50-90% to account for potential residual selection bias, but this is not a robust belief. There is also a broader literature on the LRAD program that we haven’t had time to look into, but several of the titles include something along the lines of, ‘a critical review’, which makes us suspect that the overall picture is not so rosy. It’s also important to note that this paper is measuring the effect on a household of getting land earlier versus sooner, and not looking at the overall impact on yields. In other words, it doesn’t account for what happens to those on the other side of the transaction, people who sell their land.

Finally, we note that although the government provided the ancillary services in northeast Asia, many of these could, in principle, be outsourced to other actors and indeed might be the most fruitful part of the package for non-governmental actors to focus on. There may be some concern that countries that would most benefit from providing these services do not have the capacity to do so, so outsourcing them in these contexts may be particularly effective.

Political will

The political climate in northeast Asia in the middle of the twentieth century is very different from the political climate today. The 1930s, 40s, and 50s saw northeast Asia gripped by violence and conflict.^[36] The turmoil may have played a role in enabling significant change. In defeated and occupied Japan, the United States played a role in forcing land reform on the country. In Taiwan, the United States drew up the cadastral maps which were the basis of the land reform program. In China, South Korea, and Taiwan, large landowners were associated with Japanese occupiers. The looming ideological struggle between communists and capitalists made radical redistribution a relevant topic of discussion. Today, radical redistribution is mostly off the political agenda.

During our conversation, Studwell said that many African governments are run by elites who are divorced from smallholder farming. Many government officials view smallholder farming as backward, and strongly support large-scale farming. This is in stark contrast to the mid-twentieth-century leaders of China, Japan, South Korea, and Taiwan—many of whom came from farming families—who actively supported and encouraged smallholder farming. This perhaps suggests that you would need to focus on electing politicians who were more sympathetic to land reform before you could make progress, but these politicians would likely have impacts in many other areas which you would also want to assess.

Dietrich Vollrath also raised the concern of a trade-off between impact and tractability when we spoke with him. Regions with the most unequal land redistribution which might benefit the most from redistribution, e.g. Latin America, are also those where large landholders are politically powerful and will be able to forcefully oppose any major redistribution of land. You would need to find a country in the sweet spot (if one exists) where there is enough inequality such that redistribution would still be beneficial but not so much that large landholders can strangle any attempt at being forced to relinquish their assets. We suspect this is a particularly thorny issue to solve.

We don’t have a particularly coherent story for why land redistribution has become less politically attractive or how to make it more attractive. Studwell would likely point to the rise of neoliberal economics and a kneejerk opposition to redistribution. Neoliberal economists would likely point to failures of land redistribution across the world. In their 2020 report on land inequality Uneven Ground, the International Land Coalition writes, “While at specific moments in history redistributive agrarian reforms have played a decisive role, they require exceptional social and political conditions to succeed. For instance, they have been effective during revolutions: in Mexico in 1910, in Bolivia in 1953, and in Cuba in 1959. The same happened in China and Vietnam a few decades later. They have also been successful in the aftermath of wars and during a country’s occupation. Examples include Japan, Taiwan, and Korea after World War II” (50).

Potential regions

Our main takeaway from this analysis is that land redistribution will be very difficult anywhere in the world because nowhere seems to have the right combination of factors that were present in northeast Asia. Our impression is that South Asia is the best place to start looking for potential opportunities, but we would again emphasise that there is potentially lots of heterogeneity within each region. There is a fairly large degree of land inequality in South Asia and more of the population currently works in agriculture than any other region apart from sub-Saharan Africa. There is some history of land reform attempts in the region (e.g. the Bhoodan movement, a voluntary land reform movement in India) and it seems like farmers are a relatively powerful political unit given the recent protests in India, so the political will constraint may be less binding, but we discuss below how one NGO’s attempt was politically thwarted in India.

Latin America, although having high levels of land inequality, has a much smaller share of its population working in agriculture and it seems like there will be significant political opposition to aggressive land redistribution. Sub-Saharan Africa has less land inequality but a higher share of the population in agriculture and looks set to benefit from a demographic dividend so initially looks promising. However, more common communal land ownership in the region complicates the picture somewhat as it may substitute for redistribution, so we’re less sure that land redistribution is desirable. Southeast Asia is also potentially worth looking at, but given the previous attempts and failures at land reform described by Studwell, tractability again seems like a key issue.

Farm Size and Yields

A central component of How Asia Works is the idea that smaller farms, when properly supported, outperform larger farms in terms of yields. Studwell makes some strong claims on this front. For example, he writes, “Almost anything – and perhaps everything – that grows is able to benefit from increased human attention. This is why there is so much evidence from so many countries that farm yields per hectare are in inverse proportion to farm size” (68). Granting Studwell’s assessment of the literature on the question, it nonetheless appears that some cracks have appeared in the consensus since How Asia Works was researched and written. In this section, we first detail the mechanisms by which smaller farms might be more productive than larger farmers, then we summarize the current state of knowledge regarding the relationship between farm size and yields.

Mechanisms

This section borrows heavily from Vollrath (2018), a blog post literature review of the farm size yield relationship.

In the 1960s and 70s multiple researchers noted the inverse farm-size yield relationship (IFSY) in a number of contexts. But if there are perfect land and labour markets and people act as profit maximisers, the IFSY shouldn’t exist. If a large farm has a lower yield than a small farm, then the profit maximising approach is for the large farmer to rent the small farmer some land, thus increasing yield and decreasing size on the large farm and decreasing yield and increasing size on the small farm with the net effect of increasing total output. This should occur until yield is identical on both farms and the total output of the two farms is maximised. The IFSY therefore became a stylised fact that needed to be explained by a more complicated theory.

A first common explanation claims that the effective wage for labour within the family is lower than that for labour hired outside the family. This could be due to a number of reasons: surplus labour within the family, a low opportunity cost for family labour, or a transaction or monitoring cost for hired workers. The lower effective wage would mean that a small farm would hire more family labour, increasing the labour intensity per hectare and raising yields. Large farms need to hire outside workers since the supply of family labour is not sufficient and since outside workers have a higher effective wage, large farmers would hire fewer workers per hectare resulting in lower yields.

A second explanation invokes uncertainty in how much output is produced by the land each year and the need to produce a minimum subsistence amount of output in order to survive (which doesn’t depend on farm size). Small farmers will have to raise the expected yield of their land to eliminate the risk that they don’t produce enough to survive which will give them high average yields. Large farmers get large output from their larger supply of land even in bad years so don’t need to increase their expected yield in order to survive, so their average yield is low.

A third explanation is decreasing returns to scale; large farms just are less productive i.e. have a lower total factor productivity, than small farms. Vollrath (2018) says that this just assumes the answer and would still require a market failure of some sort to generate the IFSY, but doesn’t discuss it further. It’s very rare to see estimates of total factor productivity for farms so we’re not sure why he dismisses it so quickly; we think this would be an interesting area for further research.

Another set of explanations posit that the IFSY doesn’t actually exist and is just a statistical or data artifact. The main explanation along these lines is that the difference in yields is driven by differences in land quality; small farmers could have higher quality land which is higher yielding for a fixed set of inputs. Alternatively, random measurement error in farm size and/or non-random measurement error in yields or farm size could generate a spurious relationship between yields and farm size.

Empirical Evidence

Michael Lipton said in conversation that when he reviewed pre-2015 literature, 90% of it found an inverse relationship between farm size and yields. He didn’t provide us with a meta-analysis or literature review here but we trust his claim. His preferred explanation for the IFSY is the first one we offered; non-family labour has a higher effective wage as there are monitoring and transaction costs for hired labour while family labour is more motivated than non-hired labour.

More recent, and in our estimation, higher quality evidence does not support small farms being more productive than large farms though. However, we note that there is likely to be publication bias in this recent literature, with results disproving or explaining away the IFSY being more novel and more likely to be published than results confirming the IFSY, so we interpret it with care.

Deininger et al. (2018) find that the inverse relationship significantly weakened over time in India (see figure 9 below). They use an IV strategy whereby the size of inherited land is an instrument for actual land size to try to induce exogenous variation in farm size. We don’t think this is a particularly credible instrument as size of inherited land is likely correlated with other factors that can affect yield such as education and access to credit to purchase inputs. Nonetheless, the descriptive relationship is still useful.

They argue that the reason the relationship has declined over time is because labour market failures imperfections have been reduced. In other words, the difference between the effective wage for family labor and hired labor (the first explanation for the ISFY described above) has shrunk over time. They support this explanation by showing that labor supply and demand decisions are separable in the later period but not in the initial period, but we haven’t investigated this mechanism section in detail nor thought through possible alternative explanations.

Figure 9: IFSY relationship in India over time from Deininger et al. (2018)

Foster and Rosenzweig (2017) find a U-shaped relationship between farm size and yields in India from 2009-2014, with very small farms highly productive (due to family labour) and very large farms highly productive (due to expensive capital investments), but with medium size farms (~10 acres) the least productive because they are too big for family labour but too small for capital investment. A lot of existing literature doesn’t sample farms greater than 10 acres so misses the upward curve in the relationship. Vollrath (2018) points out an implication of a U-shaped relationship:

If you took all the land and redistributed it out in 10-hectare [sic]^[37] parcels, so that there was no inequality in land holdings, you’d get the lowest possible yield from the land. Your best bet, given their data, is to redistribute such that you maximized inequality, with a bunch of very small farms (high yields due to low labour costs) and very large farms (big enough to use capital)

Figure 10: relationship between real average profits per acre (a proxy for yield) and farm size (acres) from Foster and Rosenzweig (2017)

A second strand of recent literature suggests that the IFSY is due to measurement error. The IFSY is typically estimated with regressions of the following specification:

$$ln{Y}_i=\alpha +\beta X_i+\gamma ln{A}_i+{\epsilon }_i,$$

Where $Y_i$ is the output, $A_i$ is land area, and $X_i$ is a set of controls (typically including measures of land quality amongst other things). If $\gamma =1$, then a plot $c\%$ larger would have $c\%$ higher output so there would be no relationship between farm size and yield. The elasticity of yield with respect to area $\left(\gamma \right)$ is typically found to be less than one which could imply the IFSY.

An alternate explanation, initially proposed by Lamb (2003), says that random measurement error in the independent variable $A_i$, land area, could be to blame. If there is random measurement error in the independent variable then the coefficient on that variable suffers from attenuation bias and is attenuated towards 0. Lamb (2003) finds some empirical support for this explanation by comparing the results of fixed effects models to random effects models. He claims that if there was no measurement error, the results of the fixed effects estimation should be closer to one than the random effects estimation, and therefore since the fixed effects estimation is further away from one, there is empirical support of the measurement error explanation. We haven’t checked the econometrics here but we assume it is correct given the paper is published in the Journal of Development Economics, a top field journal.

On the other hand, Carletto et al. (2013) use GPS measurements of land area which should suffer less from measurement error relative to the previously used self-reported land size. They find that the IFSY relationship is actually stronger with more accurate land measures. This is because when households self-report their plot size they typically round to the nearest acre or half acre, and this matters proportionally more for smaller plots as the same amount of rounding represents a larger percentage of the actual plot size. Essentially, farmers are more likely to overreport the size of their small plots and underreport the size of their large plots.

Figure 11: bias in land measurement: rounding problems from Carletto et al. (2013)

An even more recent literature looks into measurement error in the dependent variable, output. If there was just random/classical measurement error in the dependent variable then this wouldn’t cause a bias; it would instead just make the results noisier. We need non-random measurement error in the dependent variable to cause a bias.

Desire & Jollife (2018) compare farmer self-reported outputs to crop-cuts in Ethiopia, where trained enumerators measure, harvest and weigh randomly selected subplots, to reduce measurement error. They find the IFSY relationship exists when self-reported outputs are used as the dependent variable but is non-existent when crop-cut outputs are used. This is because farmers systematically overreport production on small plots and underreport it on larger plots, as seen in figure 12 below, and similarly to the misreporting of area discussed above. This means that the measurement error is negatively correlated with plot size. They propose a potential explanation for this but they can’t test it; farmers round up their output to the nearest multiple of 5/10/25 kg, which leads to overreporting by a constant positive quantity regardless of plot size. This means there is proportionally more overreporting in small plots. This result is replicated in Uganda in Gourlay et al. (2019) and in Mali in Yacoubou Djima & Kilic (2020); both papers similarly finding that farmers with smaller plots overreport their production relative to more objective crop cuts and the IFSY disappears when using the more objective measure.^[38] Note that these three papers use GPS measures for land area rather than self reports so concerns about independent variable measurement error are minimised.

Figure 12: Maize yield estimates, self-reports and crop-cuts for same plots by wave. Vertical axis is yields in kilograms per hectare, horizontal axis is plot-size quartiles. From Desire & Jollife (2018)

Finally, Abay et al. (2019) combine these two literatures and analyse the case when you can have non-classical measurement error in the dependent and independent variable. They empirically show that when using more objective yield and area measures with reduced risk of measurement error, there is no IFSY in Ethiopia. They also show theoretically and empirically, that if you only correct the measurement error on one side of the equation, you actually make the bias worse relative to a model where you have variables with measurement error on both sides. This provides a convincing explanation for the Carletto et al. (2013) result which used objective area measurements but subjective output measurements and found the IFSY.

Our key takeaways on the farm size-yields relationship are:

• Heterogeneity across countries (especially with respect to different farm sizes in different countries) and crops is important and understudied. A lot of the older data comes from India whereas a lot of the newer data comes from East Africa, and we would be keen to see the replication of these new results in multiple contexts.
• If there is either no relationship or a U-shaped relationship, then the empirical support for land redistribution from large landholders to smallholders on the efficiency basis of higher yields is much reduced. With a U-shaped relationship, you could even argue for redistribution towards larger landholders to increase their yields. Note that you still might want to redistribute for equity reasons though.
• Our current best guess is that Studwell overstates the importance of IFSY; large farms are more productive than small farms but below a certain threshold it seems like there is either no relationship or a weak negative one. However, there will likely be further twists and turns in this still active literature so we don’t want to overconfidently claim that there is no IFSY.

Funding Opportunities

Unfortunately, there appear to be few good funding options in this area. Essentially no large and reputable NGOs work on land redistribution.^[39] Multiple experts cautioned that getting land redistribution right requires deep understanding of the local context. This suggests that building an advocacy network for redistribution from scratch would be difficult.

One could support political candidates in countries of interest that support land redistribution, but this strategy comes with obvious risks, and thus we have not seriously investigated it. Alternatively, one could fund a think tank in a country of interest that has the capacity to investigate redistribution. This strategy is much less risky, but the theory of change is also less clear. Finally, one could offer funding to an NGO that works in a related area (like tenure reform or agricultural support) conditional on their advocating for redistribution. This strategy’s theory of change is clearer, but there may not be an organization well-placed to pivot in this direction. Landesa is probably the most promising candidate.

Landesa

Landesa, founded as the Rural Development Institute (RDI) in 1981, is one of the world’s foremost NGOs focused on land issues. Its founder, Roy Prosterman, worked on the U.S. government’s “Land to the Tiller” program in South Vietnam from 1970 to 1973, which redistributed land to peasant farmers in the South in an attempt to reduce support for the Viet Cong. For the first two decades of its existence, a large part of Landesa’s advocacy centered on redistribution. In large part, this decision was based on the political opportunities that existed at the time. According to one account, “RDI’s work up until this point [2004] had been largely opportunistic, working in countries [China and Russia] with explicit invitations for its research and legal expertise” (Thorogood et al. 2017: 8). However, as the new millennium wore on, Landesa found fewer opportunities to promote redistribution. The organization began advocacy efforts in India because of South Asia’s high rate of land inequality and because, on paper at least, the Indian government had expressed support for redistribution. In practice, however, they encountered a different situation: “Virtually every [Indian] state adopted land redistribution laws, but implementation was poor. The proposed two-acre plot distributions by taking land from large landowners were largely seen as unfeasible, as this would have required redistributing approximately 25% of India’s agricultural land, primarily through taking it from politically powerful interests” (Thorogood et al. 2017: 8). In the end, the RDI advocated for redistributing 0.1 acre “micro-plots” because that was what was politically feasible. This appears to be the last time the group seriously attempted redistribution. In 2009 RDI underwent a strategic overhaul, rebranding as “Landesa” and switching emphasis almost entirely to tenure reform.

It’s possible that Landesa could be convinced to pivot back to redistribution, but given that the pivot away from redistribution appears to have been deliberate, it seems likely that they don’t consider redistribution a particularly tractable intervention.^[40]

Other NGOs

There are many organizations that work on tenure reform, and even more that work on agrarian reform more generally, but we haven’t systematically evaluated these types of interventions or vetted any organizations pursuing them. (The agricultural support orgs listed below are just those that were recommended to us.)

Land

Cadasta - tenure reform

Global Land Alliance - tenure reform

Lincoln Institute for Land Policy - land stewardship

Centre for Social Innovation - land rights in India

International Land Coalition - tenure reform and smallholder support; in their 2020 report on land inequality Uneven Ground, they write, “redistributive land reforms should not be dismissed as an answer to land inequality” but that such reforms “require exceptional social and political conditions to succeed” (50).

Agricultural Support^[41]

Altertrade - agricultural support in SE Asia (based in the Philippines); too small to be useful

Proximity - agricultural support in Myanmar

One Acre Fund - agricultural support in eastern and southern Africa.^[42] One Acre Fund were recommended to us twice independently

International Development Enterprises - microirrigation and pedal pump irrigation in Asia, Africa, and Central America

Farm Africa - agricultural support in eastern Africa

Precision Agriculture for Development - smallholder farming support in Africa and South Asia (already a GiveWell standout charity)

RTI - smallholder farming support in Latin America, Africa, and South Asia

Think Tanks with a Possible Interest in Land Reform

PLAAS - think tank in South Africa (that maybe did some land reform work in the past)

Makerere Institute for Social Research - think tank in Uganda (that maybe did some land reform work in the past)

Foundation for Economic Freedom - thank tank in the Philippines (that has done work on land reform in the past) but they seem to be against redistribution

Schumacher Center for New Economics - US-based think tank; promotes community land trusts

Research

One might also want to fund further research in this area. This could include more in depth research on questions specific to land redistribution or a shallow review of interventions in related areas such as land tenure reform, or agricultural support policies, or other ways of boosting agricultural yields. In terms of redistribution options could include:

• Funding more studies across a variety of contexts and crops of the relationship between farm size and yield, using the best techniques for measuring both farm size and yield.
• Commissioning more rigorous impact evaluations of historic land redistribution programs. The reforms in northeast Asia specified a maximum farm size and farms above that size were redistributed. In theory this seems to provide a nice setting for well identified regression discontinuity studies of the partial impacts of the redistribution on yields and incomes so we were surprised not to see studies of this sort. Lack of data seems to be the main issue, so funding archival work to collect data might be beneficial here.
• Looking more closely at opportunities in South Asia, the top region we identified. This report was necessarily shallow and had a global focus, whereas a geographically limited study might have a better chance of understanding the important local contexts and finding good opportunities.

What We Would Do With More Time

1. Speak to Landesa
2. Speak to the International Land Coalition
3. Speak to Klaus Deininger
4. Speak to Andrew Foster or Mark Rosenzweig
5. Investigate more closely the role measurement error may play in the farm size - yields literature
6. Investigate more closely the interventions and policies that tenure reform NGOs pursue
7. Read Boestel et al. 1999, Lipton 2009, Binswanger-Mkhize et al. 2009, and ADB 2020 more closely

Acknowledgements

David Rhys Bernard and Jason Schukraft jointly researched and composed this report. Thanks to Marcus A. Davis, Neil Dullaghan, Peter Hurford, and Linch Zhang for helpful comments on earlier drafts. Further thanks to Yajna Govind, Michael Lipton, Joe Studwell, Dietrich Vollrath and Ismaël Yacoubou Djima for taking the time to speak with us. Open Philanthropy provided funding for this project and we use their general frameworks for evaluating cause areas, but they do not necessarily endorse its conclusions.

References

Abay, K. A., Abate, G. T., Barrett, C. B., & Bernard, T. (2019). Correlated non-classical measurement errors, ‘Second best’ policy inference, and the inverse size-productivity relationship in agriculture. Journal of Development Economics, 139, 171-184.

Adamopoulos, T., & Restuccia, D. (2020). Land reform and productivity: A quantitative analysis with micro data. American Economic Journal: Macroeconomics, 12(3), 1-39.

Albertus, M. (2020). Land reform and civil conflict: Theory and evidence from Peru. American Journal of Political Science, 64(2), 256-274.

Anseeuw, W. & Baldinelli, G.M. (2020). Uneven Ground: Land Inequality at the Heart of Unequal Societies. International Land Coalition.

Bauluz, L., Govind, Y., & Novokmet, F. (2020). Global Land Inequality. World Inequality Database working paper.

Behrman, J. R., Sengupta, P., & Todd, P. (2005). Progressing through PROGRESA: An impact assessment of a school subsidy experiment in rural Mexico. Economic Development and Cultural Change, 54(1), 237-275.

Binswanger-Mkhize, H. P., Bourguignon, C., & van den Brink, R. J. E. (eds.). (2009). Agricultural Land Redistribution: Toward Greater Consensus. World Bank Publications.

Boestel, J., Francks, P., & Kim, C. H. (2013). Agriculture and Economic Development in East Asia: from Growth to Protectionism in Japan, Korea and Taiwan. Routledge.

Carletto, C., Savastano, S., & Zezza, A. (2013). Fact or artifact: The impact of measurement errors on the farm size–productivity relationship. Journal of Development Economics, 103, 254-261.

Deininger, K., Jin, S., Liu, Y., & Singh, S. K. (2018). Can labor-market imperfections explain changes in the inverse farm size–productivity relationship? Longitudinal evidence from rural India. Land Economics, 94(2), 239-258.

Desiere, S., & Jolliffe, D. (2018). Land productivity and plot size: Is measurement error driving the inverse relationship?. Journal of Development Economics, 130, 84-98.

Dore, R. (2013). Land reform in Japan. A&C Black.

Foster, A. D., & Rosenzweig, M. R. (2017). Are there too many farms in the world? Labor-market transaction costs, machine capacities and optimal farm size (No. w23909). National Bureau of Economic Research working paper.

Gignoux, J., Macours, K., & Wren-Lewis, L. (2015). Impact of land administration programs on agricultural productivity and rural development: existing evidence, challenges and new approaches. Revue d'Etudes en Agriculture et Environnement-Review of agricultural and environmental studies, 96(3), 467-498.

Gourlay, S., Kilic, T., & Lobell, D. B. (2019). A new spin on an old debate: Errors in farmer-reported production and their implications for inverse scale-Productivity relationship in Uganda. Journal of Development Economics, 141, 102376.

Keswell, M., & Carter, M. R. (2014). Poverty and land redistribution. Journal of Development Economics, 110, 250-261.

Lamb, R. L. (2003). Inverse productivity: Land quality, labor markets, and measurement error. Journal of Development Economics, 71(1), 71-95.

Lawry, S., Samii, C., Hall, R., Leopold, A., Hornby, D., & Mtero, F. (2017). The impact of land property rights interventions on investment and agricultural productivity in developing countries: a systematic review. Journal of Development Effectiveness, 9(1), 61-81.

Lipton, M. (2020.) Land reform contexts: demography/employment, farms, soil-water resources/authority. Rethinking Land Reform in Africa: New Ideas, Opportunities and Challenges, 20-33.

Studwell, J. (2013). How Asia Works: Success and Failure in the World's Most Dynamic Region. Open Road+ Grove/Atlantic.

Walinsky, L. (ed.) (1977). Agrarian Reform as Unfinished Business: the Selected Papers of Wolf Ladejinsky. World Bank Publications.

Yacoubou Djima, I. & Kilic, T (2020). Survey Measurement Errors and the Assessment of the Relationship between Yields and Inputs for Smallholder Farmers: Evidence from Mali. Restricted access working paper

Notes

---

1. “Land reform is a broader term comprising both land redistribution and land tenure reform. Land reform often takes place within an even broader strategy of agrarian reform: a collection of activities and changes designed to alter the agrarian structure of a country” (Binswanger-Mkhize et al. 2009: 4). ↩︎

2. See, for example, Open Philanthropy’s 2015 shallow investigation of land use reform. ↩︎

3. In each of the 3 main sections of the book, Studwell goes through each of the southeast Asia countries in turn and says how they failed at reform, implying that southeast Asia should be seen as a counterfactual for northeast Asia had it not reformed successfully. ↩︎

4. There’s a fourth major section dedicated specifically to China. There’s also an introduction and an epilogue. ↩︎

5. See below for the best data we could find on the extent of this inequality. ↩︎

6. For example, Japan’s population doubled from around 35 million in the 1870s to over 70 million in the 1930s (Boestel et al. 1999: 22). ↩︎

7. Around 25% of Taiwan’s total area is arable. In Korea the figure is 20%, and in Japan the figure is a scant 14% (Boestel et al. 1999: 21). By way of comparison, in Germany and France the comparable figure is around 35% (World Bank). ↩︎

8. Lack of arable land coupled with population pressure pushed land prices to unsustainable heights during the inter-war period in Japan. In 1938, arable land in Japan sold for, on average, $479-748 per acre. In the land rich United States, the comparable figure was $80-130 per acre of arable land (Walinsky, 1977: 572). ↩︎

9. “Prior to the First World War the total farm debt in Japan was estimated at 750 million yen, or 135 yen per farming household. After that time the increasing disparity between farm income and expenditure led to a rapidly mounting indebtedness so that in 1937 the debt totaled more than 6 billion yen, or over 1,000 yen per household. [...] Although interest rates are high for all types of borrowers in Japan because of chronic shortage of capital, they have been highest by an appreciable margin in agriculture, averaging between 12 and 15 percent per year and frequently totaling as much as 20 or 30 percent. The annual charge on the farm debt consumed almost a third of the net yearly farm income during the period 1931-1935” (Walinsky, 1977: 572). ↩︎

10. “The vehicle for the change was… to take available agricultural land and to divide it up on an equal basis (once variation in land quality was allowed for) among the farming population. This, backed by government support for rural credit and marketing institutions, agronomic training and other support services, created a new type of market. It was a market in which owners of small household farms were incentivised to invest their labour and the surplus they generated towards maximising production. The result was a hugely increased yields” (4). ↩︎

11. In Japan, there were “only 110 incidents of violence between landlords and tenants [...] reported in the reform years 1947-8 and not one life was lost” (25). ↩︎

12. Local U.S. commanders believed that land redistribution would act as a bulwark against communism, though this view was not always shared in Washington. ↩︎

13. In Japan, landholders were given “30-year bonds paying 3.6 per cent interest on below-market valuations, despite an inflation rate so high that it would render the payment almost worthless” (25). In South Korea, “Remuneration to landlords was particularly ungenerous, with some losing 90 per cent of the value of their assets” (28). ↩︎

14. In Japan, “the Ministry of Agriculture estimated there were 250,000 cases of landlords attempting to retain their land by taking it back from their tenants. But the land committees, which were required to review all transfers that attempted to circumvent land reform, managed to reverse almost all of them” (25). ↩︎

15. In conversation, we asked Studwell how much of northeast Asia’s success could be attributed to land redistribution alone and how much was the product of agricultural support. He declined to offer an estimate, saying that all parts of the package are needed for success. ↩︎

16. Even in industrialized Japan, around half of the labor force was employed in agriculture in the late 1930s (Boestel et al. 1999: 3). ↩︎

17. “Farm production and labour normally occupy a lot of space and are therefore not easily watched. They also normally need time between applying labour or other inputs (water, fertiliser, cattle medicines) and outputs. Supervising and managing workers, and teaching them on the job, usually involves higher ‘transaction costs’, per unit of output, when land is used for farming rather than for factory or service production. [...] Family farming means that the household decision-taker and her family provide over half the farm’s labour. Such workers are normally more motivated than hired labour. The head of the farm household is likelier to know their availability, abilities and weaknesses. That reduces supervision cost” (Lipton 2020: 23-24). ↩︎

18. Studwell stresses the importance of “the fullest possible use of labour in rural economies in order to maximise output. As Michael Lipton put it: ‘In early development, with labour plentiful and the ability to save scarce, small farming is especially promising, because it is the part of the economy in which a given amount of scarce investible resources will be supported by the most human effort’” (65). ↩︎

19. Later, Studwell writes, “In agriculture, the initial developmental challenge is to maximise yields and output by utilising all the labour in an economy. Gardening-style cultivation achieves this. However, as industry takes off and rural dwellers begin to move into better-paid industrial and service jobs, farming needs to rebalance towards a greater emphasis on productivity and profit. Gradually, this requires a shift towards larger, more mechanized farms, allowing the incomes of the remaining farmers to rise beyond what is permitted by the cultivation of equalised small plots” (61). ↩︎

20. This coincides with the Green Revolution. Studwell doesn’t discuss the Green Revolution in detail so seems to implicitly assume that it affected northeast and southeast Asian states similarly. This seems like an alternative story worth investigating but we haven’t looked into (1) the differences between the crops grown in northeast and southeast Asian countries (2) whether the Green Revolution impacted those crops differentially. ↩︎

21. From the abstract: “Using a geographic regression discontinuity design that takes advantage of Peru's regional approach to land reform through zones that did not entirely map onto major preexisting administrative boundaries, I find that greater land reform dampened subsequent conflict. Districts in core areas of land reform zones that received intense land reform witnessed less conflict relative to comparable districts in adjacent peripheral areas where less land reform occurred. Further tests suggest that land reform mitigated conflict by facilitating counterinsurgency and intelligence gathering, building local organizational capacity later used to deter violence, undercutting the Marxist left, and increasing opportunity costs to supporting armed groups” (Albertus 2019: 256). ↩︎

22. “Returns from land reform and other agricultural improvements begin to taper off after only a decade or so, and the emerging economies have to transition into another phase of development” (73). ↩︎

23. “Manufacturing and trade drive the second phase of rapid economic development. The challenge to policymakers is therefore to direct entrepreneurial talent towards manufacturing rather than services, and in particular towards large-scale manufacturing with the heft to compete globally. Manufacturing firms are nurtured by the state in two ways: through protection and through subsidy. These interventions create breathing space for entrepreneurs while they learn to manufacture competitively” (75). ↩︎

24. “This term refers to a policy of continually testing and benchmarking domestic manufacturers that are given subsidies and market protection by forcing them to export their goods and hence face global competition. It is their level of exports that reveals whether they merit state support or not” (76). ↩︎

25. “North-east Asian politicians then improved their industrial policy returns through a second intervention – culling those firms which did not measure up. This might have meant a forced merger with a more successful firm, the withdrawal of capital by a state-directed financial system, withholding – or threatening to withhold – production licenses, or even the ultimate capitalist sanction, bankruptcy. Since the 1970s, there has been much talk about state industrial policy in western countries being an attempt to ‘pick winners’ among firms, something that most people would agree is extremely difficult. But this term does not describe what happened in successful developing states in east Asia. In Japan, Korea, Taiwan and China, the state did not so much pick winners as weed out losers” (77). ↩︎

26. “A third intervention in north-east Asia was to provide a great deal of bureaucratic support to manufacturers which exported successfully. In addition to domestic market protection and a supply of credit, states provided important assistance in the field of technology acquisition. Governments in Japan, Korea, Taiwan and China have variously undertaken collective bargaining operations to buy foreign technology – often forcing foreign firms to hand over know-how or to lower their price for it in return for local market access – and organised public sector or joint public-private research initiatives where individual firms were unable to undertake research and development investments alone” (78). ↩︎

27. “Governments in all the major economies of east Asia tried at some stage to nurture domestic manufacturers. That those in north-east Asia succeeded, while those in south-east Asia failed miserably, turned on a small number of policy differences. By far the most important of these was the presence – or absence – of what I call ‘export discipline’” (76). ↩︎

28. “financial control that keeps money aligned with agricultural and industrial policies is essential in the formative stages of development” (219). ↩︎

29. In northeast Asia, “financial policy frequently accepted low near-term returns on industrial investment in order to build industries capable of producing higher returns in the future” (169). ↩︎

30. “Japan had faced formal, repeated demands from the IMF and the secretariat of the General Agreement on Tariffs and Trade (the GATT, the forerunner of the World Trade Organization) to move to currency convertibility from 1959, but hung on to its highly restrictive Foreign Capital Law until 1980. Korea and Taiwan were similarly recalcitrant in resisting demands to free up capital flows until the 1990s. In south-east Asia, states liberalised far earlier in the development process. The result was escalating short-term foreign debt flows which governments were unable to control. Rich-country money, also with short-term gains in mind, poured into the region chasing relatively high interest rates. Moreover, currencies in south-east Asia were either pegged to the US dollar or appreciating with it, making short-term investment doubly attractive to foreigners – so long as the pegs held” (173-174). ↩︎

31. "South Korea and Taiwan took off economically with educational capital that was well below average. Fifty-five percent of Taiwanese were illiterate at the end of the Second World War; the figure was still 45 percent in 1960. Literacy in South Korea in 1950 was lower than in contemporary Ethiopia" (xxiii). ↩︎

32. The report refers to ‘Chosen’ which was the Japanese name for Korea at the time. ↩︎

33. The countries they include in their analysis are: Ethiopia, Gambia, Malawi, Niger, Nigeria, Tanzania, Bangladesh, China, India, Pakistan, Vietnam, Ecuador and Guatemala. They also mention data from Burkina Faso, Indonesia, Brazil, Mexico and Peru but these countries don’t appear in their results table. ↩︎

34. The parallel trends assumption for difference-in-difference to be valid would require that if the reform had not taken place, productivity and/or yields growth should have been the same in small and large farms. The authors don’t present any evidence that these were on the same trend before the reform, and it seems plausible that they could have had different growth rates. Given that the land redistribution in this context sets a maximum farm size, with farms greater than the 5 hectare cut-off being forced to sell their land, this seems like a good setting to run a regression discontinuity design study instead, although sample size (334) might be a limiting factor. ↩︎

35. By comparing individuals who made it through the application process, you remove the larger concerns about selection in terms of who applies versus who doesn’t, and who completes the process versus who doesn’t, but there may still be residual selection due to the quality of the application. ↩︎

36. “The [Second World War] itself devastated large sections of Japanese industry and infrastructure and left the nation close to starvation in 1945. Meanwhile, as Japan began the process of recovery in the late 1940s and early 1950s, its former colonies, Taiwan and Korea, were engulfed in civil strife which destroyed much of the investment in rural and urban infrastructure and industry which had been built up during the colonial period. South Korea after the end of the Korean War in 1953 and Taiwan after the Nationalist take-over in 1949 were amongst the poorest countries in the world at the time, and heavily dependent on American aid” (Boestel et al. 1999). ↩︎

37. The minimum yield is achieved at 10 acres, not 10 hectares. 10 hectares = 0.1 sq kilometres, 10 acres = 0.04 sq kilometres ↩︎

38. There is actually an upwards slope in crop cut yields as plot size increases, not an inverse relationship, as shown in figure 12, but the relationship is not statistically significant. ↩︎

39. That’s not to say that no groups work on redistribution. For example, there’s an organization in Brazil called MST (Movimento dos Trabalhadores Rurais Sem Terra – roughly “Landless Workers’ Movement”). MST is a Marxist organization that aims to increase land ownership among the rural poor in Latin America. The group’s main tactic appears to be to occupy large tracts of land until the government appropriates the land and redistributes it. These occupations have, in the past, turned violent, and the group seems to veer uncomfortably close to militantism. Hence, not a good funding option. ↩︎

40. Landesa is not alone in this regard. Keswell & Carter 2014 note (without explanation) “the shift of land policy over the last 25 years away from redistribution and towards land titling and tenancy reform programs” (251). ↩︎

41. Examples of agricultural support opportunities aimed at improving yields include (but are by no means limited to): input subsidies (fertiliser, seeds, tools and other forms of capital), extension services (training on new farming techniques), market facilitation (both for purchasing inputs and selling outputs), subsidised (micro)credit, subsidised insurance (rainfall insurance especially), crop storage, infrastructure (irrigation, roads to improve market access), and direct cash support. ↩︎

42. See GiveWell conversation notes here. ↩︎