Demand for cigarettes in the United States: effects of prices in bordering states and contiguity with Mexico and Canada

Abstract

Based on a cross-section of US states for 2004, this article estimates a demand function for cigarettes after including a proxy for prices in the bordering states and simple proxies for contiguity with Mexico and Canada and for being a major tobacco producer. One major point seems to be that the negative elasticity for within-state price is similar in magnitude to the positive elasticity for the (lowest) price in bordering states. Several additional points also seem noteworthy. First, having a border with Mexico lowers sales in the state sizably. Second, the share of Hispanic/Latino population in the state also lowers sales significantly. Third, contiguity with Canada appears to have no significant effect. Fourth, partial impact of the state being a major tobacco producer appears minor even though consumption in these states is considerably higher. Fifth, education shows the expected negative association with cigarette consumption, but its statistical significance is low. Last, income carries a weak negative parameter, perhaps reflecting the lower prevalence of smoking in higher-income households.

Acknowledgement

Many insightful suggestions from an anonymous referee on an earlier version are gratefully acknowledged. The usual disclaimer, however, applies.

Notes

¹Following most literature, we treat recorded sales as reflecting consumption (‘demand’) in the state. Note also that our measure of border-state price is crude and does not distinguish between areas that are adjacent to the border from those that are more distant. Similarly, our proxies for contiguity with Mexico and Canada are rather simple.

²CDC (2004, p. 9) reports that state-level median prevalence of smoking among adults drops from 29.1% for households with annual income below $35 000 to 19.7% among households whose income is at or above $35 000. In our own sample, simple correlation between income and sales is –0.34 for which SAS reports a p-value of 0.014. These are, of course, ‘gross’ relations and do not account for other differences.

³Smoking and Health Action Foundation web site indicates October 2006 prices for 200 cigarettes to vary from US $66.53 in Quebec to $78.17 in Manitoba. Similarly, among the bordering US states, prices per pack in 2004 ranged from $3.43 in Minnesota to $4.66 in Washington.

⁴Guindon et al. (2002, p. 37) indicate that, as of March 2001, average per-pack price in Mexico for Marlboro was US $1.55 against the US average of $3.71. Based on Orzechowski and Walker (2005) data, simple average of 2001 prices in the four states bordering Mexico was $3.28.

⁵If the border-price variable is not included, the own-price elasticity estimate of column F in Table 2 ‘drops’ from −1.67 to −1.24.

⁶Sweanor (2003) explains the ‘very complicated’ structure of Canadian tobacco taxes and the changes in taxes and prices since February 1994. In short, while there was a substantial reduction in taxes and prices in 1994, both have tended to increase since then. For the US, Orzechowski and Walker (2005) indicate that average (real) retail price fell from 1993 to 1998 and rose from 1998 to 2003, falling again in 2004. Taxes as percent of retail price generally fell from 1994 (31.4%) to 2004 (30.1%), and the drop was quite sharp during 2000-2002.

⁷Full regressions for 1993 and 1999 are available from the authors. The broad pattern of estimates for all variables is very similar for 1993, 1999 and 2004.

⁸It seems that a major dimension of the smuggling problem in Canada is largely internal. A news release of 18 December 2006 by Physicians for a Smoke-Free Canada indicates the major problem to be illegal movement of tax-free tobacco from Canadian reservations to main consuming areas of Ontario and Quebec.

⁹For 2004, simple averages of prices in producer and nonproducer states are $3.18 and $3.91, respectively. The simple means for education are 83.8% and 86.8%, and average incomes are $26 803 and $29 516, respectively.

¹⁰Relevant t-statistic produced by SAS for the difference-in-means test is –2.80 with a p-value of 0.0073. This is based on equal variances since the null of variance equality is not rejected at any meaningful level.

¹¹As indicated in footnote 2, CDC (2004, pp. 8 and 9) shows higher household income to be pervasively associated with lower smoking prevalence in all states, and the simple correlation between income and sales in our sample is also significantly negative.

¹²Taking some account of the possible spatial correlation might improve the estimates. However, as Wooldridge (2002, p. 134) notes, inclusion of a variable like our border-price term can capture the spillover effects. Moreover, our crude test indicates that spatial correlation may not be significant. Based on the estimates of model F of Table 2, when we relate each state's residual with the median residual of bordering states, neither the simple correlation nor the ‘slope’ coefficient in the regression of own residual on the median residual for bordering states reaches significance at the conventional 5% level.

¹³The test has been explained by Ramsey and Schmidt (1976). Following the observation by Wooldridge (2006, p. 308), squared and cubed terms of the fitted values were added to the model of column F in Table 2. The F-statistic (2, 38) for the joint significance of the fitted-value terms is 1.66 with a p-value of 0.2043.

¹⁴It might be interesting to see how US–Mexico cigarette smuggling may be affected after the proposed strengthening of this border takes place. It should also be useful to undertake similar studies of border effects in Europe and other regions.