Health, Arbitrage, and the Velocity of Information

Because of their fiscal and health implications, cigarette taxes have gained in importance over the last few decades. Governments levy taxes to increase revenue and health officials argue for higher cigarette taxes because they believe higher taxes will reduce rates of smoking prevalence. Decision-makers at the state level understand the importance of cross-border purchasing for fiscal and health policy. Cigarette tax differentials create incentives for consumers to incur the costs of gas and time to purchase cigarettes in bordering states. This arbitrage behavior reduces tax receipts from a given tax increase and implies that cigarette consumption falls by less than measured sales imply. Fiscal authorities have longed concerned themselves with this type of tax leakage (FTA Studies, NYT 1964).

While a healthy literature studies cross-border shopping, we contribute in two important ways. First, while most existing studies use annual sales data, we use finer-grained monthly data. Second, because we have a longer time-series (starting in the 1930s), we directly model and measure how the introduction of the internet affected cross-border sales.

With these data we describe a more nuanced dynamic in the temporal pattern of cross-border shopping. In doing so, we reveal the extent of the bias in estimated price elasticities that ignore these sales (Baltagi and Goel 1987; Stehr 2005; Lovenheim 2008). Moreover, we quantify the temporal variation in the bias that occurs as people initially cross borders to arbitrage prices and as such behavior settles down to a new equilibrium level.

We also exploit our long time-series to explore whether and how these patterns have changed as information became cheaper and as the velocity of information diffusion increased because of internet penetration.

Examining annual and monthly sales data illustrates these points. We examine annual and monthly sales data before and after cigarette tax increases in 1967, 1989, and 1999. If we compare the price responsiveness, it is less in annual than in monthly data. It is clear that the annual data mask stockpiling of cigarettes in the few months just before the tax increases and the lower than equilibrium consumption in the months immediately following the increase. And, if one examines responses in the monthly data over these three years, the data hint that this stockpiling behavior grew over time. In our paper we move beyond this description to test empirically whether the internet contributed to that change in behavior. For this exercise we exploit data collected by the Federal Trade Commission that track the number of internet service providers in each US ZIP code area over time. We create a population weighted measure of ISP providers in each state and the population weighted distance to the next lower tax jurisdiction. These data provide temporal and geographic variation in the cost of internet service access that we exploit to test whether that variation explains temporal differences in the pattern of cross-border sales.