How far to the border? The extent and impact of cross - border casual cigarette smuggling.

The remainder of this paper is organized as follows. The second section provides a description of the data used throughout the analysis. The third section presents evidence on the home state price bias, and the fourth section derives the demand model used throughout this study and discusses its implications. The estimation strategy is described in the fifth section, and all results are presented in the sixth section. The seventh section concludes.

DATA

The individual-level data in this analysis come from the CPS Tobacco Supplements: September 1992, 1995, and 1998; January 1993, 1996, and 1999; March 1993, 1996, and 1999; June and November 2001; and February 2002. These surveys span nine years in four waves given approximately every two years. Because I am interested in combining these data with a measure of smuggling distance, I restrict the sample to those living in an identified metropolitan statisical area (MSA); this is the most specific level of geographic identification available in the CPS. As there are MSAs that split state lines, each identifiable state--MSA combination is taken as a separate MSA. (9) I will use state--MSA and MSA interchangeably.

I combine these data with state average price and tax data from The Tax Burden on Tobacco compilation (Orzechowski and Walker, 2006). All prices are inflated to real 2004 dollars using the gross domestic product (GDP) implicit price deflator. Prices listed in this compilation are spot prices as of November of that year. To construct a more accurate price series, I subtract the November excise tax in each state from the listed price and smooth the pre-tax price changes evenly over the entire year. I then add in the appropriate excise and sales taxes for each state and month in the Tobacco Supplement. (10)

The central variable in the analysis is the distance to a lower-price locality. I use 2000 Census geographic data to estimate a population--weighted average distance from each state--MSA combination to the closest lower-price border. (11) This calculation is done by finding the "crow-flies" distance from each census block point in a state--MSA to each intersection between a state border and "major road." (12) Once I calculate the distance from each block point to each road crossing, I take the closest crossing from each block point to a given border state and calculate a population--weighted average across block points for each border state. By measuring distance from the population center rather than the geographic center of a given MSA, I am able to more accurately characterize the distance an average individual must travel to smuggle cigarettes. In the tables that follow, the distance measure is the distance to the closest lower-price border, which is often, but not always, a border state. (13)

In addition to neighboring states, many individuals can obtain lower-price cigarettes from Native American Reservations. Native American Reservations are considered separate legal entities from the United States and are thus not subject to sales and excise taxes. In 1976, the U.S. Supreme Court ruled in Moe v. Confederated Salish and Kootenai that states have the right to impose sales and excise taxes on cigarette sales occurring on reservations to non-tribal members. Although evidence suggests a substantial amount of sales occur on reservations to non-tribal members (ACIR, 1985; FACT Alliance, 2005), only 12 states have passed legislation that allows taxation of these sales. Table 1 contains information on which states tax non-tribal reservation sales and the case law or regulation that legitimates these taxes. I collected these data using Cigarette Tax Evasion: A Second Look (ACIR, 1985), which documents much of the case law and state legislation through 1985 on Native American cigarette sales. I augmented and updated this information using state taxation statutes found through LexisNexis. Reservations in the states listed in Table 1 are excluded from the analysis. (14)

Table 2 presents means of distance, price differences, and tax differences for all identified MSAs by state. The table also lists the number of tax changes observed in the data as well as all of the closest lower-price localities for each state. Table 2 illustrates the heterogeneity across states in smuggling incentives. For example, consumers in Massachusetts, New York, Illinois, and Wisconsin live close to areas in which cigarettes are substantially less expensive. However, in states such as Delaware, Nevada, and Oregon, consumers likely live too far away from the lower-priced jurisdictions to realize the savings from purchasing cigarettes there.

Because my empirical models all include MSA fixed effects (see the fifth section on estimation strategy), I will be restricted to using within--MSA variation in distance over time. Cross-time variation in distance within a state--MSA is driven by price changes; when a home or border state changes its cigarette price, the closest lower-price border can change, thereby generating variation in distance. Table 3 contains the number of distance changes, the average change in distance, and the standard deviation of the distance changes between each CPS survey. While the majority of MSAs experience no distance change between each period, there is a substantial amount of variation in the distance measure of varying sign and magnitudes.

HOME STATE PRICE BIAS

When the opportunity to purchase cigarettes in lower-price localities exists, demand models that utilize the home state price as the measure of the true price paid by consumers can generate biased estimates of the average partial effect of price on consumption if there are unobserved differences in how individuals respond to home state price changes. The heterogeneity in demand response is a function of smuggling incentives that typically are not included in models of cigarette demand using micro-data. This problem essentially equates to an omitted variables bias as the propensity to smuggle is likely correlated with home state cigarette prices. I term this source of bias the "home state price bias" because it stems from an inability of the home state price to correctly measure the true price paid by consumers. (15)

While many studies using individual cigarette data assert the existence of this bias (Lewit, Coate, and Grossman, 1981; Lewit and Coate, 1982; Chaloupka, 1991; Gruber et al., 2003), there has been no documentation of how the responsiveness of consumption to the home state price varies with smuggling incentives. Table 4 contains mean residuals by distance quartile from a regression of log mean MSA cigarette consumption on log home state cigarette prices, MSA demographic characteristics, and MSA fixed effects using the CPS data described in the previous section and in the fifth section. The residuals from this regression represent the within--MSA variation in cigarette consumption that is unexplained by demographics and home state prices. I calculate mean log cigarette residuals by quartile of distance to the nearest lower-price border state for three margins of demand: intensive, extensive, and full. (16) As Table 4 illustrates, the residuals are positive in MSAs that are closer to the border and negative for those farther away from the border. These signs are consistent with a home state price bias because consumers who live closer to the border smoke more than suggested by the home state price. (17) In order to obtain parameters of the cigarette demand function that are less prone to this source of bias, I explicitly model the heterogeneity in home state price effects due to varying smuggling incentives. In lieu of directly observing smuggling activity (which is unobservable in the data), I construct a model of cigarette demand that incorporates the decision of whether to smuggle based on observable consumer characteristics.

A MODEL OF CIGARETTE DEMAND WITH CROSS-BORDER PURCHASES

Assume each consumer faces two prices: the price of cigarettes in the home state ([P.sub.h]) and the price of cigarettes in the closest lower-price locality ([P.sub.b]). Additionally, assume the parameters of the demand function are the same regardless of the place of purchase. In other words, consumers differ solely by the price they pay for cigarettes. Let demand of consumer i be given by

[1] E[ln([Q.sub.i])| [P.sub.j], [X.sub.i] = [[beta].sub.0] + [[beta].sub.1] ln([P.sub.j]) + [gamma] [X.sub.i], j = b, h

where X is a vector of individual characteristics. Demand can then be written as

[2] [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where [S.sub.i] is an indicator function that equals one if an individual smuggles and zero otherwise. One can see from equation [2] the biases associated with treating the home state cigarette price as the actual price paid by all consumers. The elasticity with respect to the home state price (hereafter the "home state price elasticity') is given by

[3] [[epsilon].sub.H] = [[beta].sub.1](1 - [S.sub.i]) - [DELTA][S.sub.i] / [DELTA] ln([P.sub.h]) [[beta].sub.1] ln ([P.sub.h] / [P.sub.b])

Note that unless [S.sub.i] = 0 and the price change does not induce consumer i to smuggle, the home state price elasticity will be less than [[beta].sub.1] in absolute value as the home state price is higher than the border price by construction.

The other elasticity of interest is the "full price elasticity," which yields the percent change in cigarette demand when the full price of cigarettes changes by one percent. In other words, the full price elasticity measures the responsiveness of demand when all prices change such that the smuggling decision is unaltered. This elasticity is given by [[beta].sub.1] in equation [2].