On demography and living standards in the colonial era

This is a topic that has been bugging me. Very often, historians will (accurately) point out mortality statistics in the United States, Canada (Quebec) and the Latin America during the colonial era were better than in the comparable Old World (comparing French with French, British with British, Spanish with Spanish). However, they will argue that this is evidence that living standards were higher. This is where I wish to make an important nuance.

Settlement colonies (so, here there is a bigger focus on North America, but it applies to smaller extent to Latin America which I am more tempt to label as extractive – see here) are generally frontier economies. This means that they are small economies because of small populations.  This means that labor and capital are scarce relative to land. All outputs that come from the relatively abundant factor will thus tend to be cheaper if there is little international trade for the goods that they are best at producing. The colonial period pretty much fits that bill. The American and Canadian colonies were basically agricultural colonies, but very few of those agricultural outputs actually crossed the Atlantic. As such, agricultural produces were cheap. This is akin to saying that nutrition was cheap.

This, by definition, will give settlement colonies an advantage in terms of biological living standards. As they are not international price takers, wheat is cheaper than in the old world. This is why James Lemon spoke of the New World as the “Best poor man’s country” (I love that expression) : it was easy to earn subsistence. However, beyond that it is very hard to go beyond. For example, in my dissertation (articles still in consideration at Cliometrica and Canadian Journal of Economics) I found that when wages were deflated by a subsistence basket containing very few services and manufactured goods and which relied heavily on untransformed foods, Canada was richer than the richest city of France. Once you shifted to a basket that marginally increased transformed goods and manufactured goods, the advantage was wiped away.

Yet, everything indicates that mortality rates were greater in Paris and France and than in Quebec City and Quebec as a whole (but not by a lot) (see images below).  Similar gaps seem to exist for the United States relative to Britain, but the data is not as rich as for Quebec. However, the data that exists for New England suggests that death rates were lower than in England but the “bare bones” real incomes measured by Lindert and Williamson show that New England may have been poorer than Great Britain (not by much though).

Crude Death Rates


I am not saying that demographic and biological data is worthless. Quite the contrary (even I wanted to, I could not since I have a paper on the heights of French-Canadians from 1780 to 1830)! The point is that data matters in context.  The world is full of small non-linearities between variables. While “good” demographic outcomes are generally tracking “good” economic outcomes, there are contexts where this may be a weaker relation (curvilinear relations between variables). I think that this is a good example of that point.


On the reversal of fortune, urbanization and Canada

One of the more famous articles of economist Daron Acemoglu is his 2002 article on the reversal of fortunes where he points out that countries colonized by Europeans in 1500 that were relatively rich then are relatively poor now. In the paper, they use urban density as a proxy for economic development at that point in time.

I was not particularly convinced by this because of the issue of ruralization in colonial economies. I am still not convinced in fact. As many scholars interested in American colonial history point out, the country de-urbanized (ruralized) during the colonial era as cities grew at a slower pace than the general population. As such, the share of the US population in rural areas increased. But Jeffrey Williamson and Peter Lindert documented that in 1774, the United States were the richest place in the world (beating England on top of being more egalitarian). 

This is normal. Economies on the frontier had land to labor ratios that were the exact opposite of those in Europe. The opportunity cost of congregating in one area was high given the abundance of land that could be brought under cultivation. This is why the Americas (North America at least) was the Best Poor Man’s Country. As such, areas with low population density are not necessarily poor (even if urbanization is a pretty strong predictor of wealth).

This is where Canada comes in. Today, the country easily fits in the “relatively rich” group. According to the figures 1 and 2 in the work of Acemoglu, Johnson and Robinson, it would have been in the “relatively poor” group well behind countries in Latin America. However, I recently finished compiling the Canadian GDP figures between 1688 and 1790 which I can now compare with those of Arroyo Abad and Van Zanden for Peru and Mexico. With my Canadian data (see the figure below), we can see that Canada was as poor as Latin America around 1680 (the start date of my data).


So, Canada was a relatively poor country back which was equally poor (or moderately richer) than Latin American countries. Why does that matter to the reversal of fortune story? Well, with the urbanization data, one shows that the non-urbanized of 1500 are the rich of the today. With the GDP data for the 1680s, we see that the more urbanized countries were also poorer than the less urbanized countries.

Now, my argument is limited by the fact that I am using 1680s GDP rather than 1500 GDP. But, one should simply extend the urbanization series to circa 1700 and the issue is resolved.  In any case, this should fuel the skepticism towards the strength of the reversal of fortune argument.

On getting the data right : price disparities before 1914

I am a weird bird. I get excited at weird things. I get excited at reading economics and history papers (and books). I get particularly excited when I read papers and books that “get the data right”. This is because I believe that most theoretical debates in economics stem from poor data forcing us to develop grandiose theories or very advanced models to explain simple things. One example of that is the work of Joshua Hendrickson who argued that monetary aggregates (M1, M2 etc.) are not necessarily perfect indicators of money. However, these aggregates were used in statistical tests and generated strange results inconsistent with theory. This issue has been the cause of many debates. Josh stepped in and said that we just had a variable that was not created to measure what the theory said. Using broader measures of money, he found the results consistent with theory. The debates were driven by poor data (as I think is the case in issues over fiscal multipliers, crowding-out and business cycles).

Thus, I am always excited to see data work that “get things right”. One recent example that adds to cases like that of Hendrickson is Peter Lindert’s working paper at the National Bureau of Economic Researcher. Now, before I proceed, I must state that I am very partial to Lindert as he has been a big supporter of my own research and has volunteered important quantities of his time to helping me move forward. Thus, I have a favorable bias towards Lindert (and his partner in crime, Jeffrey Williamson).  Nonetheless, his working paper requires a discussion because it “gets prices right”.

The essence of his new working paper is that our GDP per capita estimates prior to 1914 may overestimate divergence between countries over time.

Generally, when we measure GDP, we try to derive “volume indexes” that measure quantities produced at a fixed vector of prices. For example, when I measured Canadian economic growth from 1688 to 1790 (I am submitting it in a few weeks), I took the quantities of grain reported in censuses and weighed them by prices for a fixed year. This is a good approach for measuring productivity (changes in quantities). Nonetheless, there are issues when you try to move this method over a very long period in time. The prices may become unrepresentative.  So you get time-related distortions. Add to this that all the time-related distortions may be different over space. After all, should we believe the relative price of wheat to oats in 1910 was the same in Canada as it was in Russia?  Variations in relative prices over space will affect this issue. Basically, you juxtapose these two types of distortions when trying to measure GDP per capita over centuries and you may end up so far in the left field that you’re in fact in the right field.

In his working paper, Lindert tried to adjust for those problems by moving to prices that were more representative. The approach he used is basically the one used by Robert Allen in his work on the Great Divergence. You create a bundle of goods that capture the cost of living in different regions – a basic bundle of goods. This generates purchasing power parities. From there, he recomputed incomes per capita with these measures prior to 1914. The results are striking: there is much more divergence between Europe and Asia that commonly proposed and the United States are much richer than otherwise believed (and were more richer very early on – as far back as the colonial era).

Now, why does this matter?

Well, consider the debate on convergence. Many scholars have been unimpressed by the level of income convergence across countries (at least until the 1980s). However, Lindert’s estimates suggest that the starting point was well below what we think it was. In a way, what this is telling us is that many puzzles regarding the “catching-up” of poor countries may be simply related to poor data. Imagine, for a second, that we could redo what Lindert did with many more countries at a higher time frequency. What would this tell us? Imagine also that this new data would confirm Lindert’s point, what would that entail for those entangled in debates over development?

Basically, what I am saying is this: most of our debates often stem from poor data. If a simple (and theoretically sound) correction can eliminate the puzzles, maybe our task as economists should be to stop bickering over advanced theory and make sure the data is actually geared towards testing our theories!

A depressing take on inequality

Recently, I reviewed Unequal Gains (Princeton University Press) which is basically the magnum opus of economic historians Peter Lindert and Jeffrey Williamson. In the pages of Essays in Economic and Business HistoryI survey the history of growth and inequality in the United States since 1700 that they portrayed in their book.

Coming out of their book, I could not help feel depressed and simultaneously vindicated in my classical liberal outlook of the world. While they avoid the Pikettyesque tendency to create “general laws” of inequality, their results suggest that inequality has risen in spite of massive government intervention since the 1920s.

To be clear, Unequal Gains is probably the best book you can get on understanding the dynamic of inequality. Although I am biased in their favor since both authors have given me great help in my academic career, the book should overthrow Capital in the 21st century as the reference work on inequality. Throughout the book, they use normal economic theory to explain why inequality increased or decreased (discrimination, capital flows, immigration, changes in labor force participation, urbanization, relative factor scarcities, uneven supply shocks, changes in returns to human capital, regional income differences). They constantly eschew general laws. From the book, we should understand that inequality is context-specific. Like a recipe, difference mixes of the ingredients of inequality will yield different courses. This is the main strength of the book (plus the tons of data).

And this is also why it is depressing. The vast majority of inequality before 1910 in the United States would have been the result of market forces (immigration, urbanization, capital flows, relative factor scarcities, regional income differences) and not of governmental decisions. I believe that the pre-1910 level of inequality is sensibly overestimated and that, while not gigantic, government policies did have a non-negligible role in raising inequality. Nonetheless, most of these inequalities are hard to judge negatively. More immigrants from poor Italy may depress (I do not agree with that claim, but people like G.Borjas of Harvard could make this claim) wages in the United States in 1900 and increase inequality, but the migration of the Italian to America leaves no one worse off while improving the living standard of the Italian migrant. Urbanization, as part of the industrialization, is a hard process to fault and criticize. So, inequalities before 1910 are simply an issue of explaining their levels and trends.

After 1910 however, there is what Lindert and Williamson call the “great leveling” where there is an important decrease in inequality which ends in 1970. This is where I become depressed. In my paper, I highlighted that most of the fall in inequality between 1910 and 1970 occurs because or regional convergence, gender wage convergence and racial wage convergence. Between the 1910s and 1970s, differences in per capita state-level incomes narrowed dramatically (and they have since slightly widened). Between 1910 and 1970, thanks to the migration of blacks to the north, wages between whites and blacks grew closer together. Between 1910 and 1970, thanks to the arrival of household amenities like running water, appliances and electricity, women joined the labor force and the gender wage gap narrowed. None of these factors have anything to do with redistributive policy. Now, I am not claiming that redistributive policy had no impact on inequality measures (that would be empirically false). What I am claiming is that numerous forces were at play – some of which were related to non-governmental factors. Between 1910 and 1970, if one looks at ratios of government spending to GDP, there is a massive increase in the size of government. And yet, many factors of convergence had little to do with government.


Since the 1970s, inequality has surged again – and this is in spite of the fact that governments are growing larger in many respects. While spending is at all levels seems to be either stable or growing, regulatory barriers like licensing regulations and rent-seeking arrangements in the form of corporate bailouts have multiplied. Thus, the rise of inequality occurs in spite of a very active state. Not only that, but I am working on papers with John Moore of Northwood University to study inequality from 1890 to 1940 because we believe that the level is overestimated and misunderstood and (by definition) that this affects the trendline of inequality in the 20th century. If inequality in the 1920s falls slightly, the U-shaped curve of inequality (very high before 1910 falling to 1970 and increasing thereafter) described by Piketty and others becomes a flatter upward slopping curve (maybe more like a J-shaped curve). If me and John are correct (we are still crunching numbers and collecting data) inequality increased with state intervention.

And that is highly depressing. Now, I am a classical liberal who believes that state intervention should be limited. But it is not beyond to recognize that when the state throws tons of money of something, it might get a few things the way it wants (a broken clock is still right twice a day). Thus, I expected some social programs to have an impact (and I still believe that on a case-by-case basis, some social programs do reduce inequality) but I did not expect such a disappointing performance. One could even say “depressing” performance.

Nonetheless, I would suggest to everyone to read Unequal Gains and throw out Capital in the 21st century. 

Note: To be clear, Lindert and Williamson are not making the claim I am making here. While their book is predominantly a “positive economics” work, they do propose some policy courses to reduce inequality and argue favorably for redistributive policy. This is merely my “positive take” on their book.