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

Lovenheim, Michael E "How Far to the Border?: The Extent and Impact of Cross-Border Casual Cigarette Smuggling." Stanford Institute for Economic Policy Research Policy Paper No. 06--040. Stanford, CA: Stanford Institute for Economic Policy Research, 2007.

Nebraska Department of Revenue. Nebraska Taxation of Native American Reservation Indians. Lincoln: Nebraska Department of Revenue, 1996.

Orzechowski, William, and Robert Walker. The Tax Burden on Tobacco. Historical Compilation, Volume 41. Arlington: Orzechowski and Walker, 2006.

Slemrod, Joel. "The System--Dependent Tax Responsiveness of Cigarette Purchases: Evidence from Michigan." University of Michigan. Mimeo, 2007.

Stehr, Mark. "Cigarette Tax Avoidance and Evasion." Journal of Health Economics 24 No. 2 (March, 2005): 278-97.

Thursby, Jerry G., and Marie C. Thursby. "Interstate Cigarette Bootlegging: Extent, Revenue Losses, and Effects of Federal Intervention." National Tax Journal 53 No. 1 (March, 2000): 59-77.

Wansink, Brian. "Can Package Size Accelerate Usage Volume?" Journal of Marketing 60 No. 3 (July, 1996): 1-14.

Wansink, Brian, and S. B. Park. "At the Movies: How External Cues and Perceived Taste Impact Consumption Volume." Food Quality and Preference 12 No. 1 (January, 2001): 69-74.

Wong, Milly, Charles R. Ashcraft, and Charles W. Miller. "Disclosure of Idea to Prolong Shelf Life of Flavored Pellet in a Cigarette Filter." R. J. Reynolds Interoffice Memorandom to Grover M. Meyers, April 24, 1991. Bates: 511479824-511479825. http://tobacco documents.org/rjr/511479824-9825.html. Accessed January 23, 2008.

Yurekli, Ayda A., and Ping Zhang. "The Impact of Clean Indoor-Air Laws and Cigarette Smuggling on Demand for Cigarettes: An Empirical Model." Health Economics 9 No. 2 (March, 2000): 159-70.

Michael F. Lovenheim

SIEPR, Stanford

University, Stanford, CA 94305

(1) In most states, consumers can purchase legally a small quantity of cigarettes, usually no more than two or three cartons, from a lower--priced state. Purchasing more than that amount and avoiding local tax payments on the purchase is illegal.

(2) When taxed sales are used as the measure of consumption, smuggling will cause one to overstate the full price elasticity of cigarettes in absolute value. Conversely, when micro--level data on cigarette consumption are used as the measure of consumption, the bias in the elasticity due to smuggling will tend to understate the full price elasticity in absolute value

(3) There are two types of smuggling commonly discussed in the literature: organized smuggling and casual smuggling. The former type of smuggling typically involves illegally transporting large quantities of cigarettes from one of the tobacco producing states (such as North Carolina, Virginia, and Kentucky) for illegal resale in another state. Organized smuggling became a federal crime in 1978 with the Contraband Cigarette Act and was followed by a marked decrease in interstate bootlegging (ACIR, 1985). Thursby and Thursby (2000) estimate between three to seven percent of cigarette sales can be attributed to organized smuggling, which is lower than the estimates in Stehr (2005).

(4) See the sixth section on smoking increases, casual smuggling percentages, and net sales effects for a further discussion of this issue.

(5) I call this the "home state price bias."

(6) In the absence of smuggling, the full price elasticity is identical with respect to sales and consumption.

(7) I am unable to estimate the home state sales elasticity as I do not have geographically disaggregated sales data at below the state level. Coates (1995) estimates a home state sales elasticity of -0.81.

(8) This study focuses on casual smuggling, as the distance to a lower-price border state will most influence this type of behavior. However, to the extent this measure is correlated with organized smuggling, bootlegging activity will be included in the study as well.

(9) There are upwards of 40 MSAs that split state lines. However, for all but 11 cases, the CPS only identifies the more populous part of the state--MSA combination. Where these portions of the MSA are not identified, they are excluded from the analysis. A complete list of MSAs used in this study is available from the author upon request.

(10) There are a number of counties and cities that have local cigarette taxes. Unfortunately, no data exist on the history of these taxes back to 1992. I thus exclude these taxes from the analysis and only utilize state--level taxes. As a consequence, the cross-state price differences may be understated in some cases, causing an attenuation bias in the estimate of the effect of the price difference on cigarettes demanded.

(11) While MSA definitions were constant over the time period covered by this analysis and while CPS sampling is representative of the geographic distribution of the population, populations within MSAs might have shifted. I ignore such shifts due to lack of data on within--MSA population mobility.

(12) A major road is a census classification and contains most non-residential roads. The exclusion of residential roads is trivial as the vast majority of interstate travel does not occur on such roads.

(13) In many MSAs, there are farther lower-price jurisdictions with lower prices than the closest lower-price locality. Using the closest lower-price state will cause measurement error in the distance variable if people are willing to travel a little farther to obtain a slightly better price. The results from this paper suggest individuals are quite sensitive to the distance to a lower-price border but not the level of the price difference. Further, for most MSAs, the distance to a better price than the closest lower-price is quite substantial. Thus, the use of the closest lower-price border is consistent with the data and likely causes little measurement error.

(14) See Appendix A in Lovenheim (2007) for a discussion of Native American Reservation tax enforcement as well as information on the data and methodology used to calculate distance to Native American Reservations. Due to potential measurement error in this variable, I conduct the analysis below both including and excluding reservation smuggling incentives.

(15) See Gruber et al. (2003) for further discussion of the effect of this bias on elasticity estimates.

(16) The intensive margin is the number of cigarettes smoked by smokers, the extensive margin is the smoking participation rate, and the full margin is the number of cigarettes smoked by all consumers, including non-smokers.

(17) I also compare consumption responses to changes in home state and border state prices for those living on the high-price side and low-price side of the border in the 11 identified MSAs that split state lines. The results from this comparison are consistent with the existence of the home state price bias: those living on the high-price side of the border respond to changes in the border state price more than the home state price, and those living on the low-price side respond more to changes in the home state price than the border state price.

(18) More specifically, assume there is a random component to the cost of smuggling, [epsilon], which has a distribution F([epsilon]). Assuming a consumer will smuggle if the cost is greater than the benefit, P([S.sub.i] = 1) = P([alpha](ln([P.sub.h]) - ln([P.sub.])) > [delta] 1n(D) - [phi] + [epsilon]) = P([epsilon] > [phi] + [alpha](ln([P.sub.h] - ln([P.sub.b])) - [delta]ln(D)). This expression is identical to equation [4] under the assumption that [epsilon] is uniformly distributed on [0,1].

(19) This substitution implies [[PI].sub.1] [equivalent to] [[beta].sub.1], [[PI].sub.2] [equivalent to] - [[beta].sub.1] * [phi], [[PI].sub.3] [equivalent to] - [[beta].sub.1] * [alpha], and [[PI].sub.4] [equivalent to [[beta].sub.1] * [delta].

(20) Another way to proceed would be to relax the constraints imposed by a log distance measure and use a polynomial in distance or dummy variables for different ranges of distance. These specifications are attractive as they allow the relationship between demand and distance to be relatively flexible as distance changes. I estimate demand functions using such specifications, but the small sample sizes in the data do not allow meaningful statistical inferences to be drawn from the results. Taking the point estimates at face value yields results that are similar to the ones presented.

(21) The main advantage of using log distance rather than distance is when distance is used in the regression, the effect of distance on the responsiveness of consumption to the home state price is the same no matter how far the consumer is to a lower-price border. Using log distance, the impact of distance on consumption decreases with distance. Thus, a one-mile increase in distance to a lower-price state will impact the home state price elasticity more for a consumer living five miles from the border than for a consumer living 500 miles from the border.

(22) When I relax this restriction, the home state price elasticities become slightly more negative, but the substantive conclusions and findings reported do not change. I perform sensitivity tests by restricting the effect of distance on demand to be zero for those living far away from borders or for whom the savings per mile from smuggling is low. I find these models yield similar results to equation [6], and results are available upon request. Log distance is used in all regression for simplicity, but my results are robust to more complex relationships between smuggling and distance.