How bad was land inequality in Haiti?

Evidence from new data

These results are from a book chapter I’m preparing on the role of land and property rights in Haiti’s economic development. I’ll release the working manuscript in the near future, but I wanted to highlight some results before its release.

---

There is a little debate about how bad land inequality was in Haiti. According to some, land was available to all, with few large latifundia-style holdings. But some dispute this view. They argue that Haiti was characterized by significant inequality in landholdings. The heart of the dispute is the scarcity of data and the insufficiency of one particular source of data.

But now we have new data to assess the claims.

The Data Debate

The earliest source of land data has been the 1950 census. But there are some problems with it. As with most censuses around the world, the census reported landholdings in bins. The smallest bin is land smaller than 0.32 hectare (ha). The largest bin is anything bigger than 25.8 ha. This bin accounts for 0.2% of properties.

While this small share seems to confirm that the large farms were limited, there’s a lot of room for interpretation. Because the census reports only the number of properties in that bin. It does not tell us the amount of land. All we know about the largest landholdings is that they are larger than 25.8 ha. This means that our views on inequality, when using the census data, depend entirely on how much land we think is in that top bin.

The amount of land in the top bin is exactly the source of debate. Mats Lundahl presents data from two people looking at the same data. The two people use different aggregations, so it makes the comparisons hard. But in one case, Locher assumes that about 30% of land area is on plots greater than 6.5 ha. In the other, Brisson assumes that plots that are 12.9 ha or greater account for 67% of land area. While the Locher assumptions are not explored, Lundahl shows that Brisson uses an estimate of Haiti’s arable land, subtracts the amount of land in the lower bins, and assigns the residual to the higher bins.

The difference in assumptions leads to different Gini coefficients. Locher’s assumption yields a Gini of 0.73 while Brisson’s coefficient is 22% higher at 0.89. While this might not look too different, Locher’s estimate would put Haiti just below its neighbors while Brisson’s would make it one of the most unequal countries in the region. The difference is substantial.

Unfortunately, the researchers on Haiti understand the same data so differently as to destroy all confidence in settling the question by an appeal to the 1950 census. We must either remain in darkness and confusion, or else we can do what economic historians do best. That is, find new data.

The New Data

I have collected over 60 irrigation schedules for canals administered by Haiti’s Department of Public Works. Below is an example. Each schedule gives the name of the property owner and the size of the irrigated plot. The schedules span from 1920 to 1943 and cover over 3,400 plots. This is unprecedented data for this period.

Let me identify the data’s strengths before acknowledging the weaknesses. The clearest strength is that these are real plots with names attached and exact sizes reported. They are not tallies in a bin that someone might have fabricated. There’s no top coding. And, as opposed to the census where some worried they would be taxed if they reported honestly, these plots have an economic incentive to be included in the list. If you’re off the list, you don’t get the water your land needs.

The clearest weakness, on the other hand, is that this is a non-representative sample of Haitian farms. These are the fortunate few that have access to irrigation. Since this should make the land more productive, we might see larger plots than average. Consistent with this hypothesis, Mats Lundahl reported that irrigated land was becoming more concentrated. An estimate of inequality using this sample is likely biased.

But sometimes our weaknesses can be our strengths. This sample should provide an upper-bound estimate of inequality. That is, inequality in irrigated land should give us a worst-case scenario for inequality across Haiti.

The Results

In the table below, I show the estimates of Haiti’s land Gini coefficient. First, I compare the distribution of plots in the census and the irrigation schedules. They are pretty similar - the 90th percentile plot is almost exactly the same, and the 50th is a lot closer than it looks since the 1.93 figure is actually the midpoint of a bin that ranges from 1.29 to 2.58.

First, I do an exercise where I get the lower-bound for inequality from the 1950 census. Instead of assuming how much land is in the top bin, I take the amount of land we know is in there, assuming all plots are just 25.8 ha. This produces a Gini of 0.49. Then I apply the same censoring to the irrigation data, and I find a Gini of 0.54. This confirms that the irrigation data are biased upward and the Gini on the full data should be an upper bound.

Using the full data, I find that the Gini is 0.61. This is significantly smaller than even Locher’s 0.73 and nowhere near Brisson’s 0.89. Compared to other Caribbean countries, Haiti’s land Gini was much, much lower.

Implications

There are a lot of theories of how land inequality led to limited growth in Latin America and the Caribbean. Haiti is a counterexample. Much lower land inequality, yet still long-run economic struggles.

Of course, land inequality is just one dimension of inequality. There are other aspects in which Haiti had significant inequality. But inasmuch as most theories come from land inequality, Haiti shows that it was not a necessary condition for achieving the types of inequality that derail economic growth.