Working Paper: NBER ID: w18122
Authors: Fernando Borraz; Alberto Cavallo; Roberto Rigobon; Leandro Zipitra
Abstract: The "border effect" literature finds that political borders have a very large impact on relative prices, implicitly adding several thousands of miles to trade. In this paper we show that the standard empirical specification suffers from selection bias, and propose a new methodology based on quantile regressions. Using a novel data set from Uruguay, we apply our procedure to measure the segmentation introduced by city borders. City borders should matter little for trade. We find that when the standard methodology is used, two supermarkets separated by 10 kilometers across two different cities have the same price dispersion as two supermarkets separated by 30 kilometers within the same city; so the city border triples the distance. When our methodology is used, the city border effect becomes insignificant. We further test our methodology using online prices for the largest supermarket chain in the country, and show that the "online border" is equivalent to the average distance from the online warehouse to each of the offline stores.
Keywords: Political Borders; Price Segmentation; Quantile Regressions; Inequality Constraints
JEL Codes: F40; F41
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Political borders (F55) | relative prices (P22) |
| Selection bias in price observations (P22) | overestimation of border effect (F14) |
| Standard methods (C90) | same price dispersion for supermarkets separated by 10 km and 30 km (R12) |
| City border effect (F55) | statistically insignificant (C29) |
| Online border effect (F55) | average distance from online warehouse to offline stores (L81) |
| Methodological flaws (C90) | overstated previous estimates of border effects (F16) |