Via a rather circuitous route, I’ve become extremely interested in domestic market integration: basically, I was thinking about the way we measure agricultural output in the past. Often, it’s a version of what we might call the “Malanima shortcut”; though he wasn’t the first to use the idea of a demand function to estimate agricultural output, his application to Italy is probably one of the best-known uses of the trick. The way it works it basically just to assume a demand function for food: y = f (food prices, other prices, income). This function is often abbreviated to simply y = f (income). If you have a time series of real wages—serving here as a proxy for income— then you can back out the demand for food.

What does this have to do with market integration? Well, consider a country with a rainforest region and a savanna region: in the rainforest region, they grow maize, and in the savanna region, they grow millet. These regions form part of the same country, but they’re not integrated at all: there’s no grain trade between them, so the Law of One Price doesn’t hold. And let’s assume for the sake of argument that the nominal wage in both regions is fixed at $1000. In the first year, there are good harvests of maize and millet, and the price of each is, say, $1/kg. The real wage, in terms of kilograms of grain, is 1000.

Now let’s assume that in year two there’s a terrible harvest of millet and a good harvest of maize. The price of millet shoots up to $2/kg, while maize prices stay the same. The real wage in terms of grain is now 1000 in the rainforest region and only 500 in the savanna! As historical GDP reconstructors, this decline in the real wage (caused by a spike in millet prices) is a very strong indication of a bad harvest. But if all we have access to are maize prices, and hence real wages from the rainforest region, we will have no way of knowing about the terrible millet harvest. If there were free and costless trade between the two regions (and only these two regions), this wouldn’t be a problem: the price of maize in the rainforest region would also rise, because of an increase in demand caused by the price increase for its substitute in consumption, millet, and our demand function would account for it. But if markets aren’t well-integrated, then there’s no reason to suppose that the maize-denominated real wage in Rainforest can tell us anything about the agricultural output of Savanna.

So the integration of markets matters! And it’s not an understudied topic in economic history. A lot of work has been done on this question in Europe and North America, obviously, but also in India and China. Interestingly, one bit of low hanging fruit hasn’t been plucked, as far as I can tell, so I went ahead and plucked it: British Burma, for which grain prices are given the standard Prices and Wages in British India compendia. Because most previous discussions of market integration in British India have focused on the railway system, to which British Burma was not connected, they seem to have omitted Burma. But Burma is well-worth studying on its own, because and not just because it became one of the most important ‘rice-bowls’ of Asia in the colonial period. So I did some transcribing.

Obviously, there are a lot of fancy econometrics on market integration one can perform that I haven’t yet, but here is the coefficient of variation (the standard deviation divided by the mean. I only used towns with data for all years from 1861-1917, meaning that coverage is limited to the Megui Islands, Dawei, Mawlamyine, Rangoon, Pathein, Pyay, Taungoo, Thayet, and Kyaukpyu. So what happened in the mid 1870s? The Rangoon–Pyay railway was opened in 1877, and at first blush seems to have had a major influence in driving price convergence across Lower Burma.