Endogenous Borders? The Effects of New Borders on Trade in Central Europe 1885-1933

A large literature on "border effects" in the wake of McCallum (1995) documents the massive impact of borders on trade. However, all these studies suffer from an identification problem. "Border effects" are usually identified from cross-sectional variation alone. We do not know how trade would change in response to a change in borders - the "treatment effect" of borders on trade - simply because trade flows across "future" borders are typically not documented. Nor can we rule out that there is "reverse causation": that borders run along pre-existing trade patterns rather than shape trade flows. We exploit a natural experiment from history to explore this issue: the many dramatic border changes that were imposed and codified by the peace treaties in 1919 across Europe. We follow Ritschl and Wolf (2008) and implement Ashenfelter's difference-in-difference estimator in levels on a large, new data set on sub-national trade flows. This allows us to trace the effects of changing borders over time and produces two key results: first, new borders have a large effect on trade. However second, the "treatment effects" of borders tend to be significantly smaller than the pure cross-sectional effects. This is so, because most of the 1919 border changes followed a pattern of trade relations across the region that was clearly visible already before 1914. Borders shape trade, and trade shapes borders