The Effects of Land Redistribution

1Argument

The Coase theorem says that when property rights are clear and bargaining is cheap, land ends up with whoever can use it best, whatever the starting allocation. When bargaining is expensive the starting allocation matters, because the transactions that would move land to better uses do not take place. The paper puts it this way. In the presence of large transaction costs, the initial allocation of land mattered a great deal for achieving efficiency in agriculture.

The French Revolution supplies the test. In August 1789 the Constituent Assembly abolished the feudal system, and every region of France received clearer property rights at the same moment. Transacting in land stayed expensive. Overlapping feudal claims took decades to settle, the Civil Code was enforced unevenly, and many land transactions took place only in the second half of the 19th century, once the cadastre had mapped each plot. In November 1789 the Assembly also confiscated the property of the Catholic Church and sold it at auction to the highest bidder. Where the Church had owned more land, more land changed hands at once.

The paper compares 194 districts. Those where a larger share of land had belonged to the Church had higher wheat yields in 1841, less fallow land and more land sown with nitrogen-fixing crops in 1852, and more drainage pipe manufacturers nearby. Their largest farms were larger and fewer of their tenant farmers were sharecroppers. The advantage was gone by 1875, as other districts reallocated their land by slower means.

2The sale of Church land

Before 1789 the Church owned between 6 and 10 percent of the land of France. The share varied widely. In towns it ranged from a third in Toulouse to 3 percent in Limoges. Church land was often leased in small, scattered plots, and under the rule of mortmain it could not be sold without a lengthy legal procedure that ended with the king's consent. When the archbishop of Aix sought to sell half his properties in 1782 he needed special permission from the courts.

The auctions moved this land quickly. Local governments ran them under rules set by the Assembly, and assessors surveyed each property to set an opening bid. About 700,000 Church properties, roughly 6.5 percent of French territory, were sold from 1790. A further 400,000 properties of émigrés, roughly 3.5 percent, were sold from 1793 under more complicated rules, and the paper focuses on the Church sales. The proceeds backed the assignats, which became a fiat currency and lost their value, so the state recovered little from the sales. In the district of Bernay 27 members of the grand bourgeoisie bought 39 percent of the Church land auctioned, while 350 peasants bought 33.2 percent.

3Where Church land was sold

The share of a district's land confiscated from the Church ranges from 0 to 40 percent across the 194 districts with data, with a mean of 5.9 percent. Bodinier and Teyssier (2000) compiled these figures from regional studies, and the districts they cover held about two-thirds of Church land before the Revolution. Choose a variable to map it, and add the drainage pipe manufacturers of 1862 or the seats of bishoprics.

4What followed the auctions

The paper regresses each outcome on the percentage of a district's land confiscated from the Church. Controls add the district's soil suitability for wheat from the FAO, or its elevation range for pipe manufacturers, and its market access in 1789. Market access sums the population of every district and of nearby foreign cities, each divided by the lowest-cost travel route to it over the roads, canals, rivers, and seas of 1789. The preferred specifications add 12 region fixed effects, so they compare districts within the same region.

With all controls and region fixed effects, a district with 10 percentage points more of its land confiscated had wheat yields about 10 percent higher in 1841. In 1852 it left 3.8 percentage points less of its arable land fallow, against an average of 20, and sowed 1.6 percentage points more with artificial prairie, against an average of 10. It had 1.6 more pipe manufacturers within 25 kilometers, against an average of 5.4. Ten percentage points is about 1.7 standard deviations of the share confiscated.

5Bishoprics as an instrument

The Church may have owned better land to begin with, which would explain the results without any effect of the sales. No agricultural survey before 1841 can rule this out, so the paper uses an instrument. Between the 4th and 6th centuries the Church acquired vast amounts of land in Gaul, and between the 6th and 11th centuries one of a bishop's main tasks was to keep that land out of lay hands. Almost all French bishoprics existed by 1200. Districts closer to a bishopric therefore held more Church land in 1789, and the distance is plausibly unrelated to where farming became profitable after the Black Death and the opening of the Atlantic.

One threat is that bishops managed nearby land better. The paper tests this with the assessed value per hectare of 812 Church farms and vineyards, recorded by the revolutionary assessors before the auctions. Plots closer to a bishopric were worth no more.

The instrumental-variable estimates are two to six times the ordinary least squares estimates. The paper offers two readings. Ordinary least squares may understate the effect, or the instrument may weight districts near seats of Church administration where the gains from redistribution were larger. The first-stage F statistics run from 10.3 to 11.8, below the Stock and Yogo critical value of 16.4 for a test of 10 percent maximal size.

6Larger farms

The paper traces the gains to consolidation. Galor and Moav (2004) argue that inequality favors growth when growth depends on physical capital, and drainage, irrigation, and new rotations were investments of that kind. The paper argues that the auctions most likely preserved the unequal holdings of the Old Regime. In the district of Limoux 41 members of the grand bourgeoisie bought 52.6 percent of the Church land auctioned, while 170 peasants bought 15.4 percent.

Sharecropping points the same way. The 1852 survey counts tenants who paid a fixed rent and tenants who paid a share of the crop. Where more Church land had been sold, sharecroppers made up a smaller share of tenants, by about 10 percentage points for 10 points more land confiscated with all controls and region fixed effects, against an average of 34 percent. Choose that outcome in Figure 3 to see it.

7The head start fades

If the auctions gave some districts a head start, the other districts should have caught up as they reallocated land by slower means. The paper estimates the effect on wheat yields in 1841, 1852, 1875, 1892, 1912, and 1929. After 1852 yields are recorded only for departments, so each district takes its department's yield and standard errors are clustered by department.

The effect falls by about a sixth from 1841 to 1852 and is indistinguishable from zero in 1875 and 1892. It reappears in 1912 and 1929 at about half its 1841 size. The paper shows that two districts produce this later effect, Cambrai, where 40 percent of the land was confiscated, and Arles, where 25 percent was. A robust regression that downweights outliers removes it.

8What it means for land reform

A reform that clarifies property rights everywhere at once does not by itself move land to its best use. Where transacting stays expensive, the allocation the reform inherits persists, and the paper finds that the gains of the Revolution's reforms arrived first where the state itself moved land to new owners. Their head start is visible in 1841 and 1852 and gone by 1875, as the transaction costs that held back the rest of France came down.

The French sales also differ from most land reforms. They allocated land by auction to the highest bidder rather than equally, and they preserved large holdings. The paper sets this against studies of other reforms. Besley and Burgess (2000) find that Indian states with more land reform saw poverty fall and agricultural wages rise. Bandiera (2003) on 19th-century Sicily, Dell (2012) on Mexico in the 1910s, and Domenech and Herreros (2017) on Spain in the 1930s find reforms that raised conflict and corruption. In the French case productivity rose where, the paper argues, the redistribution kept land concentrated, at a time when agricultural growth depended on investment in drainage, irrigation, and new crops. The paper concludes that the benefits of institutional reform depend a great deal on the transaction costs faced by economic agents.

9Limits

The confiscation data cover 194 of the 534 districts of 1790. They come from regional studies that favored well-known and economically central places, and the sampled districts have higher market access than the others. The comparisons within regions and the controls for market access address part of this.

No agricultural survey exists before 1841, so the paper cannot show that districts with more Church land were no more productive before the Revolution. The controls for soil suitability, the instrument, and the test of plot values bear on this without settling it.

Farm size is observed only for departments and only from 1862. The mechanism rests on that measure, on sharecropping in 1852, and on case studies of who bought at auction.

The instrument is moderately strong and its estimates are much larger than ordinary least squares. The effects in 1912 and 1929 rest on two districts.