Don't Believe the Haiti/DR Graph

The graph is wrong.

It’s about a week after the assassination of Haiti’s President, Jovenel Moïse, and a graph is going viral. The virality is understandable because the graph defies conventional wisdom. Somehow, Haiti was richer than the Dominican Republic as recently as the 1950s.

Except, the graph is wrong. I know it, and it’s driving me crazy.

The virality began with Noah Smith. First with this tweet (4k+ likes, 1k RTs/QTs), then with this post. Here’s the graph.

If you only know one thing about Haiti, you know it is poor. You might not know it is the poorest country in the Americas (well, at least until Venezuela stole the title recently). But you know it’s poor and has been for a long time.

But this graph pushes back on the idea. Maybe poverty isn’t as old as we thought. It forces us to imagine a counterfactual Haiti. Maybe there’s a world where Haiti made a few decisions and looks like the Dominican Republic today.

If only the graph was right.

I’m going crazy because even though I know the graph is wrong, I don’t have direct evidence to refute it. I have other sources that show divergence began earlier, and I have hard-earned wisdom from graduate school. Starting with these two, I embarked on my journey to show the graph was wrong. Two months later, I can confidently say the graph is absolutely wrong.

(Note: even though I know Noah was not the one who created the graph, if you want to refer to this essay as “the brutal takedown of Noah’s Haiti take” when you share it, it would help this get attention.)


Sweet Alternative Evidence

Look, I don’t want to flex here, but I have a PhD in economics. And my dissertation focused on the economic history of Haiti. And just when I thought no one cared about that research, along comes this viral graph. From that paper (now published open-access in the Journal of Economic History) I have one graph that easily casts doubts on this idea that Haiti and the Dominican Republic didn’t begin to diverge until the 1960s. So if there is any time to flex, it’s right now.

The graph from my dissertation:

This graph shows that Haiti, once the global leader in sugar production, was far behind its Caribbean neighbors in the 20th century, and very behind the Dominican Republic. Instead of sugar, Haiti is producing coffee. But sugar was generating more revenue. If Haiti had produced the same amount of sugar as the Dominican Republic in 1949 (one year before the viral graph starts), then its export revenues would have increased by 130%.

There are a lot of reasons that direct comparison might not be convincing. But consider labor flows. Sugar production in the DR was so profitable that every year approximately 100,000 Haitians would travel to the DR to work on sugar plantations to earn wages about 2-3x what they could get in Haiti. Many even settled in the DR! (That is, until President Trujillo massacred thousands of Haitians, which is a story for another post.) The people knew there was a divergence as early as the 1920s.

As much as I’d like to say my paper has the final word, we can also quote Victor Bulmer-Thomas, from The Economic History of the Caribbean Since the Napoleonic Wars, on all the evidence that Haitian divergence began much earlier than the 1960s:

Yet Haiti did fall behind and by 1910 was lagging all the Americas — not just the Caribbean…Haiti by this time — only two decades after 1890, when its performance was still satisfactory — had the lowest foreign trade per head, the lowest budget figures, the smallest infrastructure per head and one of the highest rates of per capita indebtedness….The comparison with the Dominican Republic was particularly galling for the Haitians. As late as 1880, Haiti had outperformed its neighbour on all counts. furthermore, the Dominican government had defaulted on its first foreign loan in1874, but Haiti had punctually serviced her external debts. Yet, thirty years later, it was the other way around….We can therefore state with some accuracy that it was between 1890 and 1910 that Haiti fell behind. (pp. 187-189, emphasis added).

Clearly the divergence began before the 1960s. So why do the data say otherwise?

Unhappy Maddison Productions

If I’m right, then there must be something wrong with the data. Where do the data come from? If you check that note at the bottom of the graph, it says Maddison Project Database.

And that about explains everything you need to know.

Not making sense yet?

Well, during my graduate school years (did I mention I have a PhD?), I was raised by goodly advisors. One of the first pieces of advice they gave me was, “Never trust the Maddison data.”

As soon as I saw that source, I knew I had to investigate where the data came from.

Digging around, I found Maddison’s 2001 explanation of the data sources. Appendix A-2 says the data for Haiti and the DR came from Series Históricas del Crecimiento de America Latina. Going through every other Maddison Project publication I could find, I never discovered an updated data source for these two countries. As far as I can tell, the data never change across releases.

Which creates a puzzle. The divergence between Haiti and the DR is not consistent across releases. In fact, it is only in the most recent release that Haiti stays richer than the DR through the 1950s. If the data haven’t changed, then why does the gap shift?

From what I can tell, it has to deal with converting the raw data into constant 2011 dollars. This conversion has to account for inflation and purchasing power across countries. It’s a tough business! And, according to the Maddison Project, it’s even tougher in Latin America.

“Providing new estimates of income levels in Latin America has proven to be a particular challenge as the combination of relative price benchmarks with current National Accounts data often leads to notably lower income levels with especially implausible implications over the long run.” Rebasing "Maddison": New Income Comparisons and the Shape of Long-Run Economic Development p. 44

When the estimates are too low to be plausible, the Maddison Project adjusts the data to be more believable. And in that same report, it mentions that Haiti was too low to be plausible in 1960. So, they just adjusted it to look better.

Holy crap!

Did you catch that?

Haiti and the Dominican Republic look the same in the 1950s because the data collectors assumed they were the same! It has nothing to do with the original data.

As my advisors said, “Never trust the Maddison data.”


The OG Data

Are there any data we can trust? Well, why don’t we use the original data?

Through the miracle of the internet, I can look directly at the source, Series Históricas del Crecimiento de America Latina. The report uses real data reported by the countries, so we can trust it to an extent.

Maddison merged the raw data with later decades then converted everything to constant dollars. But, amazingly, the report already has a section that converts the GDP data into constant 1970 dollars. What do the comparisons look like in the original data?

As expected, Haiti was always poorer than the DR during this period. While every Maddison release says Haiti was richer than the DR in 1950, the source data says the DR was already twice as rich ($230/cap compared to Haiti’s $119).

Yes, the gap widens through this period, especially through the 1970s. But according to the source data (and multiple other sources) Haitian divergence began long before this period.

Who cares?

At first glance, this seems like a pedantic rant. Especially because the gap is still widening in the 60s and 70s. Isn’t the main point that something happened in those decades that we should care about?

We absolutely should investigate those decades more. But pinning down divergence helps with the hypotheses for that divide. For example, Noah inspects several hypotheses, and the first one is Haiti’s large (and unjust) debt to France. Here’s part of his dismissal:

But there are some big problems with this thesis. First of all, Haiti finished paying back this debt (which France reduced) in 1947. That’s at least a decade before Haiti and the D.R. started to diverge economically, and four decades before the divergence became pronounced.

Now that we know Haiti diverged about half a century before the Maddison data claims, that puts this hypothesis back in the game. (I have my own hot take on the debt-poverty hypothesis, but that will have to come in the future.)

Most importantly, it shows how tough economic history is. These comparisons are hard, but the questions are super important. If we want to understand them, we need to approach them with deeper work.

If you’re someone hoping to work on Haiti’s economic history, let me know!