Cigarette taxes and the master settlement agreement.

Trogdon, Justin G. ; Sloan, Frank A.

I. INTRODUCTION

In 1998, 46 states and the four major tobacco companies entered into the Master Settlement Agreement (MSA), which settled litigation brought on behalf of the states to recover medical expenses paid by government insurance agencies for illness brought on by consumption of tobacco products (National Association of Attorneys General 2003). The four remaining states (Florida, Minnesota, Mississippi, and Texas) settled separately with the companies. The MSA stipulated that the tobacco companies pay the states an estimated $206 billion over the next several years. The payments are made annually and are adjusted for the number of cigarettes sold in each state in each year. In this sense, the MSA payment structure resembles an excise tax.

The strategy of using litigation as an instrument for discouraging consumption of a commodity deemed to be harmful to consumers is becoming more common (Parmet and Daynard 2000). The impact that such litigation has on other tobacco control policies, such as excise taxes, depends on the extent to which litigation and such policies are substitutable.

The MSA payments could substitute for excise taxes at the state level. This may be so to the extent that state legislatures have succeeded in levying socially optimal excise tax rates. (1) If so, states would be expected to have reduced excise taxes on cigarettes after the settlements were reached. Similarly, one would expect that various tobacco control policies, such as workplace smoking bans, would be substitutes, albeit imperfect ones, for state cigarette excise taxes and for penalties resulting from litigation that function as excise taxes. Alternatively, litigation and the resulting settlement may have changed the balance of power between tobacco control advocates and the tobacco manufacturers with the consequence that the settlements and state excise taxes are complements. The tobacco industry's influence on federal and state legislatures, historically, has been an impediment to enacting tobacco control legislation at either federal or state levels (Kelder and Daynard 1997; Parmet and Daynard 2000). Rather than crowd out state excise taxes, the MSA could have led to crowding in.

A cursory glance at the data supports the view that litigation changes the balance of power. Since the MSA and four individual state settlements were reached, the former in November 1998 and the individual settlements somewhat earlier, state legislatures have increasingly looked to state excise taxes on cigarettes as a source of revenue, relative to both excise taxes imposed on alcohol and state taxes more generally (Figure 1). In particular, the largest jumps in the mean real state excise tax on cigarettes occurred in 1997 and 2002. In fiscal year 2003, for example, state excise tax increases were larger than any other single type of tax including major sources of revenue such as income taxes (National Governors Association and National Association of State Budget Officers 2002). These increases did not coincide with increases in the mean real state excise tax on beer; the only year in which real state excise taxes on beer rose was 1998.

[FIGURE 1 OMITTED]

Such trends could have been due to factors other than the settlement. The goal of our empirical analysis is to assess whether the changes in cigarette excise taxes can be attributed to the settlement when other factors are held constant. Using pre-post as well as state excise taxes on beer as controls in a panel data difference-in-difference approach, the evidence on balance provides support for the view that litigation and cigarette excise taxes are complements, which is consistent with changes in the political equilibrium.

II. DATA AND EMPIRICAL SPECIFICATION

Overview

We use a panel data difference-in-difference (D-D) regression design to test whether the litigation substituted for (crowd out) or complemented (crowd in) state excise taxes. One test of the competing hypotheses would be a simple significance test of indicator variables for pre- and post-MSA after controlling for other political and economic factors that determine cigarette excise tax levels.

However, evidence of structural shifts in a pre-post analysis does not rule out causes of the structural shifts other than litigation and the resulting settlements. For this reason, we used D-D analysis with state excise taxes on beer as the control group. Beer provides an interesting comparison because, like cigarette tax, beer excise taxes are regressive and are known to affect harmful consumption, such as heavy drinking among youths (Coate and Grossman 1988; Cook and Moore 1994; Grossman et al. 1987, 1994; Kenkel 1993). (2) The crucial difference between excise taxes on cigarettes and beer is that major litigation was not pursued against beer companies as it was against cigarette companies. Thus, we do not expect to see significant changes in beer taxes due to the tobacco settlements.

Analysis of Excise Taxes

To be more specific, consider a linear model for cigarette excise taxes in state s at time t, [y.sup.c.sub.st], of the following form:

(1) [y.sup.c.sub.st] = [[gamma].sup.c] + [[gamma].sup.c.sub.s] + [[gamma].sup.c.sub.t] + [[gamma].sup.s] + [[gamma].sup.t] + [[gamma].sup.st] + [X.sup.c.sub.st][[beta].sup.c] + [[delta].sup.c.1]1(1993 [is less than or equal to] t [is less than or equal to] 1997) + [[delta].sup.c.2](t = 1998) + [[delta].sup.c.3](1999 [is less than or equal to] t [is less than or equal to] 2002) + [[epsilon].sup.c.sub.st],

where the first three parameters represent cigarette specific effects (overall, state- and time-specific, respectively) and the second three parameters capture general state, time, and state-time effects. (3) The variables in [X.sup.c.sub.st] capture the important features from the theoretical models of taxation reviewed shortly. The parameters of interest are those on the time period indicators. The pre- and post-MSA periods are further subdivided into two subperiods: 1990 to 1992 (omitted reference category), 1993 to 1997, 1998, and 1999 to 2002. The parameter for 1993 to 1997 captures any effect of the individual state suits against the tobacco industry. The parameter for 1998 captures any immediate shift in policy at the time of the MSA, and the parameter for 1999 to 2002 captures any long-term structural change in excise taxes. The model also includes a mean zero error component ([[epsilon].sup.c.sub.st]).

A similar model for state excise taxes on beer, [y.sup.b.sub.st], is

(2) [y.sup.b.sub.st] = [[gamma].sup.b] + [[gamma].sup.b.sub.s] + [[gamma].sup.b.sub.t] + [[gamma].sub.s] + [[gamma].sub.t] + [[gamma].sub.st] + [X.sup.b.sub.st][[beta].sup.b] + [[epsilon].sup.b.sub.st],

where the first three parameters are beer-specific effects, and [X.sup.b.sub.st] contain some of the same elements as [X.sup.c.sub.st] in addition to the additional variables for beer detailed shortly. We assume that the time periods dummied out in the cigarette tax equation do not have an impact on beer taxes (conditional on the included variables and state and time effects).

From these specifications, estimates of the difference in cigarette excise taxes pre- and post-MSA will be biased if we do not control for other state and time specific trends. Therefore, we subtract (2) from (1) to form a D-D estimator of the effect of the MSA on cigarette excise taxes. Our D-D estimation equation is

(3) [y.sup.c.sub.st] - [y.sup.b.sub.st] = [[beta].sub.1] + [[beta].sub.s] + ([X.sup.c.sub.st][[beta].sup.c] - [X.sup.b.sub.st][[beta].sup.b]) + [[delta].sup.c.sub.1]1(1993 [is less than or equal to] t [is less than or equal to] 1997) + [[delta].sup.c.sub.2](t = 1998) + [[delta].sup.c.sub.3](1999 [is less than or equal to] t [is less than or equal to] 2002) + [u.sub.st],

where

[[beta].sub.1] = ([[gamma].sup.c] - [[gamma].sup.b]), [[beta].sub.2] = ([[gamma].sup.c.sub.s] - [[gamma].sup.b.sub.s]), [u.sub.st] = ([[gamma].sup.c.sub.t] - [[gamma].sup.b.sub.t]) + ([[epsilon].sup.c.sub.st] - [[epsilon].sup.b.sub.st]).

Two assumptions are needed for the identification of the MSA effects in this model. First, as stated, we assume that the MSA-related time periods do not have a direct influence on beer excise taxes (i.e., [[delta].sup.b] = 0). (4) Second, we assume that the time-specific effects for cigarettes and beer are equal (i.e., [[gamma].sup.c.sub.t] = [[gamma].sup.b.sub.t])) or at least that the difference is not correlated with the timing of the MSA.

Data

Our data on state excise taxes and determinants of state excise taxes come from several sources for the period 1990 to 2002. (5) There was substantial variation in cigarette excise tax rates across states and over time (Table 1). (6) Virginia had the minimum excise tax of $0.025/pack in 2002, whereas Massachusetts had the maximum rate of $1.51/pack in 2002. Figure 1 shows the mean excise taxes across states over time relative to their initial levels in 1990. There was a clear positive linear trend over the sample period for real cigarette taxes with a spike in 2002. The mean real excise tax on beer fell during 1990 to 2002 except for a small increase in 1998.

Political economy models of taxation and public spending have extended the median voter results of Downs (1957) to include other sources of influence, such as lobbying by special interest groups and distributions of preferences such as multidimensional preferences or ideology-based cohorts as in models of electoral competition by candidates (Becker 1983; Stigler 1972; see Persson and Tabellini 2002 for a review of this literature). An example equilibrium in these models has (1) blocs of voters with high proportions of policy-sensitive voters (swing voters) and (2) organized groups (lobbies) that are overrepresented in the political process relative to the socially optimal benchmark in which the total marginal benefit of spending across groups equals the social marginal cost of raising the funds. In addition, there is an inverse relationship between the extent of overrepresentation of lobbies and the size of the lobbies.

Smokers, like abstainers in alcohol control policy, represent important swing voters in the issue of cigarette excise taxes because they have a larger stake in the specific policy and therefore more incentive to rely less on party ideology. We include state-level smoking rates in the analysis. On average, we expect smokers to oppose additional taxes leading to a negative correlation between the smoking rate and the excise tax. (7) Of course, a negative correlation between smoking rates and excise taxes could also be due to higher excise taxes leading to lower smoking rates. We consider this possible endogeneity in the specification tests to follow. Additionally, some research has noted that smokers may rationally vote for cigarette tax increases as a way to regulate their own smoking and the negative health consequences associated with it (Crain et al. 1977; Gruber and Koszegi 2001). The argument is that state cigarette excise taxes are a self-control device. If favored by smokers, one would expect that higher percentages of smokers would lead to higher rather than lower excise tax rates.

The major tobacco manufacturers are a small and "overrepresented" group. The presence of cigarette producers in the state as a special interest group is accounted for using the volume of tobacco leaf production. In competitive markets, producers have an incentive to oppose taxes on their products as these raise marginal costs of production. However, the nature of the supply side of the market for cigarettes and its implications for tax overshifting reduces the incentives for producers to avoid taxes.

Other variables control for taxes, laws, and regulations that may affect state policy goals for cigarette excise taxes. The federal excise tax on cigarettes, the presence of a "smoker protection" law, an index for clean air regulations, and MSA settlement payments per capita are included to test whether these substituted for or complemented excise taxes in state policy goals. Federal excise taxes could crowd out state excise taxes if state taxes were near an optimal level before any change in the federal level. However, if political factors, such as lobbying at the state level, had prevented an optimal level of taxes, then an increase in the federal rate could provide an opportunity (for example, weakening the political clout of the industry) for the state to approach the optimum (crowd in). Smoker protection laws require smoking areas in public locales, usually a response to clean air regulations elsewhere. Clean air regulations are aggregated into one categorical variable based on the number and type of public places where smoking was restricted: none (omitted), nominal, basic, moderate, and extensive (U.S. Department of Health and Human Services 1989).

The minimum and maximum excise tax rate in bordering states captures the impact of neighboring states' excise tax rates. Due to issues such as border crossing, smuggling, and lobbying by producers, the political equilibrium in which excise taxes are set is influenced by taxes in neighboring states (Benjamin and Dougan 1997). (8) The mean taxes in bordering states were much larger (maximum: $0.57) and smaller (minimum: $0.20) than mean taxes ($0.41), implying that states have some discretion in setting excise taxes on cigarettes and are not merely tax takers (Table 1).

We also include three sets of variables as indicated by the political science literature on tax determination. (9) That literature has considered the following hypotheses: tax increases are more likely (1) when states are facing fiscal crises; (2) early in a governor's term to minimize the impact on reelection bids; (3) when political control of the state is conducive for passage, such as when all branches are controlled by the same party (Berry and Berry 1992; Winters 1996).

First, the fiscal health of the state is measured by the appropriated ending balance for the fiscal year net of appropriated tobacco revenues. This definition most closely matches the information government officials have when they decide on whether to raise cigarette excise tax rates. It also clearly identifies the role that tobacco taxation must play to balance states' budgets. Large budget deficits could lead to increases in excise taxes, although other avenues for decreasing deficits, such as other revenue sources and cuts in expenditures, might weaken this relationship.

Second, the election cycle is measured using dummies for gubernatorial election years and off years that did not have an election or did not immediately follow an election year; years following an election year are the omitted category. For example, a state might have elections in 1992, 1996, and 2000. The years following the elections would be 1993, 1997, and 2001, respectively. The off years are 1994-95, 1998-99, and 2002-2003. Berry and Berry (1992) found that politicians could minimize the negative political consequences of tax increases by maximizing the time between the tax increase and the next election. (10)

Third, political control is measured by an ideology index that increases in the number of branches of government (governor, house, and senate) controlled by Democrats and by an institutional control dummy indicating years in which all three branches were controlled by the same party, Democrat or Republican. Democrats have tended to prefer a larger role for government services and changes in legislation are easier to implement when the same party controls all branches of the government (Berry and Berry 1992).

Two groups in the population receive special attention in the tobacco policy debate: teens due to the fact that the majority of smokers start at this age and the elderly who experience the health consequences of smoking (Gruber and Zinman 2001). We include the proportion of the states' population ages 10 to 19 and 65 and over to test the sensitivity of excise taxes to these important segments of the population. Finally, wealth is measured by income per capita.

We reproduce the model for cigarettes ([X.sup.c.sub.st]) as closely as possible for excise taxes on beer ([X.sup.b.sub.st]). There are a few minor differences. The presence of swing voters is measured using the share of the state population that abstained from any alcohol consumption (abstainer rates). To make this variable directly comparable with smoking rates, we include drinking rates as 100 minus the abstainer rate. We test for endogeneity of drinking rates in specification tests described below. The regulatory laws for beer include the blood alcohol concentration considered illegal per se and the presence of an open container, anticonsumption, and/or dram shop law (by statute or case law). Dram shop liability laws hold alcohol servers responsible for harm that intoxicated or underage patrons cause to other people (or, in some cases, to themselves).

In the results to follow, the state fixed effects ([[beta].sub.s]) are suppressed. In addition, the reported coefficients correspond to their structural counterparts: For variables that only appear in the cigarette equation the coefficients are [[beta].sup.c], for variables that only appear in the beer equation the coefficients are [[beta].sup.b], and for variables that appear in both equations the coefficients are ([[beta].sup.c] - [[beta].sup.b]). Tests reveal the presence of serial correlation in the error terms; so robust standard errors are reported.

III. RESULTS

Excise Taxes

The fixed effect results indicate that state excise taxes on cigarettes were 9.9 cents higher on average in the year of the settlements than in 1990 to 1992 (Table 2). Since implementation, cigarette excise taxes were 10.5 cents higher than in 1990 to 1992. These results suggest that litigation and excise taxes are complements in tobacco control policy. These results are also consistent with the hypothesis that the litigation and resulting settlements shifted the political equilibrium and reduced the constraints to higher excise taxes on cigarettes.

The smoking rate is negatively related to cigarette excise taxes. The drinking rate is also negatively related to beer taxes, but the relationship is not statistically significant. The correlation could have been due to the importance of smokers as swing voters (who oppose higher taxes) or to people responding to higher prices through smoking cessation (see below for discussion of instrumental variables estimation).

Tobacco leaf production is included to measure the strength of the cigarette industry as a lobbying force in each state. Within each state, decreases in the level of tobacco production are associated with decreases in the real state excise tax. What we observe in tobacco-producing states for most of the sample is that nominal cigarette tax rates have been held constant leading to a decrease in the real excise tax at the same time that tobacco quotas have been cut dramatically.

The federal excise tax on cigarettes, as the settlements, is a complement to state excise taxes. This could mean that federal increases in excise taxes relax political constraints at the state level opposing an increase. For example, state politicians could infer from the federal legislation that the tobacco producers' lobby has lost influence in the promise of votes or campaign contributions. With a reduced lobbying influence, state excise taxes can approach the efficient level. The federal excise tax on beer is also positively associated with increases in the state beer excise tax, but again, this relationship is not significant.

The results for consumption laws are mixed with evidence of both substitutability with excise taxes and complementarity. Extensive clean air laws lead to lower excise taxes on cigarettes (substitutability), but basic clean air laws relative to no clean air laws lead to higher excise taxes (complementarity). The presence of an open container law leads to lower excise taxes on beer (substitutability), but dram shop laws by statute lead to higher beer taxes (complementarity). However, we lack many observations in which laws/taxes changed to identify these effects in the fixed effect specification. In particular, only eight states changed their clean air status between 1990 and 2002, and only one (North Carolina) had a change in clean air status since the implementation of the MSA in 1998. (11)

States consider excise tax rates in their neighboring states in setting their own excise tax rates. The fixed effects estimates imply that as the maximum cigarette excise tax in bordering states increased by $1, the state's own cigarette excise tax increases by $0.22, supporting the view that decisions of neighboring states influence a state's choice of its excise tax rate. Taxes in bordering states are insignificant in the analysis of excise taxes on beer.

Lower ending balances for the states' fiscal years leads to higher excise taxes for both cigarettes and beer. This corresponds with anecdotal evidence that sin taxes were used as stopgaps for severe budget crises in states in the late 1990s and early 2000s.

Of the political variables, Democratic control of the state government leads to relatively higher excise taxes on cigarettes, perhaps because Democrats have traditionally favored higher public expenditures. Timing in gubernatorial election cycles and single-party control of the state do not affect excise taxes.

Larger youth populations (as a share of the total state population) lead to higher cigarette excise taxes relative to beer excise taxes. This relationship could indicate the use of excise taxes as a deterrent to smoking initiation. In fact, most teens in this age group cannot vote and thus cannot form a coalition to oppose such action. Higher elderly populations as a share of the state population did not significantly affect excise taxes. Higher income per capita was weakly associated with higher cigarette taxes relative to beer.

Specification Tests

The decision whether to smoke or drink depends on price, which includes excise taxes. Thus, a negative correlation between excise taxes and smoking rates (drinking rates) could be due to smokers (drinkers) responding to taxes, not taxes responding to smokers (drinkers). We conduct a number of tests to assess the empirical importance of the endogeneity of these variables.

A fixed-effects approach assumes that the included explanatory variables are strictly exogenous; that is, they are orthogonal to the error term, including the state-specific unobserved effect, in any time period. One way to test for strict exogeneity is to add to the original specification variables from future periods. Under the maintained assumption these variables should not be significant. We include t + 1 values of smoking and drinking rates in (3) and find these were not significantly correlated with the difference in excise taxes in period t (F = 1.12, p = 0.33).

We also use an instrumental variables (IV) method to account for endogeneity of current smoking rates, drinking rates, and state MSA payments because they are partly determined by cigarette sales in the state. Instruments excluded from the state excise tax equation (3) include state-level averages/proportions for demographic variables shown to influence the smoking decision: marital status, pregnancy, and exercise (Sloan and Trogdon 2004). (12) The instruments are jointly significant in the first stage. We also do not reject the null hypothesis that the instruments can be excluded from the excise tax equation using a Hansen test (chi-square = 0.06, p = 0.81).

Using IV with fixed effects reduces the magnitude of the post-MSA time indicators, and they are no longer statistically significant (Table 3). Importantly, a Hausman test fails to reject the null hypothesis that the fixed effect and IV estimates are equivalent (chi-square = 0.44, p = 1.00). Another test proposed by Davidson and MacKinnon (1993) also fails to find evidence of endogeneity (F = 0.22, p = 0.88). Therefore, we focus on the fixed effect D-D results in Table 2.

IV. DISCUSSION AND CONCLUSION

State excise taxes increased by approximately $0.10 per pack in 1998, the same year as the MSA, and remained at that level in 1999 to 2002. This evidence is consistent with the view that the litigation altered the political equilibrium by reducing the constraints to raising excise taxes that previously prevented an optimal level of excise tax.

Overall, is litigation used as a device to improve the public health ultimately in the public interest? In terms of the health benefits resulting from the price increases, one would answer "yes," but improved health must be offset at least in part by the welfare loss to smokers through reduced consumption (Sloan and Trogdon 2004). Quantifying the magnitude of this loss is controversial (Gruber and Koszegi 2001; Manning et al. 1991; Viscusi 1999), and debating this issue is beyond the scope of this study. Furthermore, how the receipts from the MSA have been allocated has been a subject of great controversy, the discussion of which is a study in itself (Gross et al. 2002; Sloan et al. 2005a, b). But given the political clout of tobacco, at least historically, it seems improbable that legislatures would set policies, both tax and other, at levels that would be set by a social dictator. Litigation may be advantageous in changing the political equilibrium and in enforcing agreements.

Facing a recession after 2000, states depended on settlement payments to fill budget deficits in part, giving states incentives to maintain the solvency of cigarette companies. The companies have maintained that the implementation of the MSA has had a negative impact on their profits. However, several oligopoly models suggest that profits could actually increase in the short run, and evidence from financial reports indicate that this might have happened (Sloan et al. 2004).

A limitation of this study is that other tobacco-related events occurred concurrently with the MSA. Many of these events, such as the Federal Trade Commission litigation against Joe Camel, revealed new information to the public about the harmful and additive nature of cigarettes and the operations of the cigarette producers. Such information could affect smoking sentiment at the state level around the time of the settlement. To the extent that changes in public sentiment affect smoking rates and restrictions on public smoking, we have controlled for such changes in the analysis. In addition, because the lawsuits were filed by the states, tobacco companies admitted that smoking is harmful for the first time. Thus, new information and accompanying changes in public sentiment due to these revelations should be rightly attributed to the settlements. However, our measure of the effect of the MSA could have resulted from changes in voter preferences due to litigation other than the MSA.

APPENDIX: DATA SOURCES

Price and tax data for cigarettes by state and year are from the Tax Burden on Tobacco (Orzechowski and Walker 2002). Beer taxes are from the Alcohol Epidemiology Program at the University of Minnesota and from the Tax Foundation Web site (Alcohol Epidemiology Program 2003; Tax Foundation 2003). These data are used to calculate the maximum and minimum excise taxes in bordering states for each state.

Data on tobacco leaf production are from the U.S. Department of Agriculture (1991 2002). The 16 states that produce tobacco leaf had missing values for 1990.

State-level smoking rates, abstainer (from alcohol) rates, and instruments for these rates are calculated using data from the Center for Disease Control and Prevention's Behavioral Risk Factor Surveillance System (BRFSS) for 1990-2002. The BRFSS is collected annually from a nationally representative sample of the U.S. adult population. A person is considered to be a smoker if he or she reported to have smoked every day; occasional or irregular smokers were considered nonsmokers. A person is considered to have abstained from alcohol if he or she reported zero drinks in the past month. State-level rates were computed using the sampling weights provided in the survey.

There are missing observations for states that did not participate in the BRFSS in the early years of our sample. The alcohol questions were in a rotation core and not used in every year in every state. We interpolate state-level abstainer rates for these states and years using results from a regression of smoking rate on time and time squared.

Fiscal variables are from various editions of The Fiscal Survey of States from 1989 to 2002 (National Governors Association and National Association of State Budget Officers 1989-2002). Three state/years are missing from this data source.

Election cycle and political control variables are from a number of sources: Statistical Abstract of the United States, Book of the States, and the National Governors Association Web site (U.S. Department of Commerce 2001; Council of State Governments & American Legislators' Association 1990-1991; National Governors Association 2003). The political control variables have 13 missing state/years.

State income per capita is from the Bureau of Economic Analysis (U.S. Department of Commerce 2003).

Variables concerning states' tobacco regulation are from the CDC's State Tobacco Activities Tracking and Evaluation System. The presence of smoker protection laws is from State Legislated Actions on Tobacco Issues (Coalition on Smoking or Health 2001).

States' alcohol regulatory laws are from various editions of the Sourcebook of Criminal Justice Statistics from 1990 to 2002 (U.S. Department of Justice 1990-2002). There are 38 missing values for illegal levels of blood alcohol content.

The amount of money collected from the MSA by state is from the publication, Show Us the Money: An Update on the States' Allocation of the Tobacco Settlement Dollars (Campaign for Tobacco-Free Kids et al. 2002).

The fraction of the state population ages 10-19 and 65 and over is from the Bureau of the Census. Age breakdowns by state were not yet available for youth from the 2000 census. Thus, the fraction of the state population ages 10-19 is predicted using state-specific linear trends for 2000-2002.

ABBREVIATIONS

BRFSS: Behavioral Risk Factor Surveillance System

D-D: Difference-in-Difference

IV: Instrumental Variables

MSA: Master Settlement Agreement

REFERENCES

Alcohol Epidemiology Program. Unpublished data. University of Minnesota: School of Public Health, 2003.

Becker, G. S. "A Theory of Competition among Pressure Groups for Political Influence." Quarterly Journal of Economics, 93(3), 1983, 371-400.

Benjamin, D. K., and W. R. Dougan. "Efficient Excise Taxation: The Evidence from Cigarettes." Journal of Law and Economics, 40, 1997, 113-36.

Berry, F. S., and W. D. Berry. "Tax Innovation in the States: Capitalizing on Political Opportunity." American Journal of Political Science, 36, 1992, 715-42.

Campaign for Tobacco-Free Kids, American Cancer Society, American Heart Association, and American Lung Association. Show Us the Money. An Update on the States' Allocation of the Tobacco Settlement Dollars. January 15, 2002. Available online at tobaccofreekids.org/reports/settlements/2002/ fullreport.pdf; accessed April 29, 2002.

Coalition on Smoking or Health. State Legislated Actions on Tobacco Issues. Washington, DC: Coalition on Smoking or Health, 2001.

Coate, D., and M. Grossman. "Effects of Alcoholic Beverage Prices and Legal Drinking Ages on Youth Alcohol Use." Journal of Law and Economics, 31, 1988, 145-71.

Cook, P. J., and M. J. Moore. "This Tax's for You: The Case for Higher Beer Taxes." National Tax Journal 47, 1994, 559-73.

Council of State Governments and American Legislators' Association. The Book of the States. Lexington, KY: Council of State Governments, 1990-1991.

Crain, M., T. Deaton, R. Holcombe, and R. Tollison. "Rational Choice and the Taxation of Sin." Journal of Public Economics, 8, 1977, 239-45.

Davidson, R., and J. G. MacKinnon. Estimation and Inference in Econometrics. New York: Oxford University Press, 1993.

Dee, T. S. "State Alcohol Policies, Teen Drinking and Traffic Fatalities." Journal of Public Economics, 72, 1999, 289-315.

Downs, A. An Economic Theory, of Democracy. New York: Harper and Row, 1957.

Gross, C. P., B. Softer, P. B. Bach, R. Rajkumar, and H. P. Forman. "State Expenditures for Tobacco-Control Programs and the Tobacco Settlement." New England Journal of Medicine, 347(14), 2002, 1080-86.

Grossman, M., D. Coate, and G. M. Arluck. "Price Sensitivity of Alcohol Beverages in the United States," in Control Issues in Alcohol Abuse Prevention." Strategies for States and Communities, edited by H. D. Holder. Greenwich, CT: JAI, 1987.

Grossman, M., F. J. Chaloupka, H. Saffer, and A. Laixuthai. "Effects of Alcohol Price Policy on Youth: A Summary of Economic Research." Journal of Research on Adolescence, 4(2), 1994, 347-64.

Gruber, J., and B. Koszegi. "Is Addiction Rational? Theory and Evidence." Quarterly Journal of Economics, 116, 2001, 1261-303.

Gruber, J., and J. Zinman. "Youth Smoking in the U.S.: Evidence and Implications," in Risk), Behavior among Youth." An Economic Analysis, edited by Jonathan Gruber. Chicago: University of Chicago Press, 2001, pp. 69-120.

Kahn, M. E., and J. G. Matsusaka. "Demand for Environmental Goods: Evidence from Voting Patterns on California Initiatives." Journal of Law and Economics, 40, 1997, 137-73.

Kelder, G. E., and R. A. Daynard. "The Role of Litigation in the Effective Control of the Sale and Use of Tobacco." Stanford Law and Policy Review, 8, 1997, 63-98.

Kenkel, D. S. "Drinking, Driving and Deterrence: The Social Costs of Alternative Policies." Journal of Law and Economics, 36, 1993, 877-914.

Manning, W. G., E. B. Keeler, and J. P. Newhouse. The Cost of Poor Health Habits. Cambridge, MA: Harvard University Press, 1991.

Manning, W. G., E. B. Keeler, J. P. Newhouse, E. M. Sloss, and J. Wasserman. "The Taxes of Sin. Do Smokers and Drinkers Pay Their Way?" Journal of the American Medical Association, 261(11), 1989, 1604-9.

National Association of Attorneys General. Tobacco Settlement Documents. Available online at naag.org/ issues/tobacco/index.php?sdpid=399; accessed August 12, 2003.

National Governors Association. Web site, www.nga.org; accessed January 28, 2003.

National Governors Association and National Association of State Budget Officers. The Fiscal Survey of States. Washington, DC: National Governors Association and National Association of State Budget Officers, 1989-2002.

Orzechowski and Walker. The Tax Burden on Tobacco: Historical Compilation 2002, vol. 37. Arlington, VA: Orzechowski and Walker, 2002.

Parmet, W. E., and R. A. Daynard. "The New Public Health Litigation." Annual Review of Public Health, 21, 2000, 437-54.

Peltzman, S. "Toward a More General Theory of Regulation." Journal of Law and Economics, 19, 1976, 211-40.

Persson, T., and G. Tabellini. "Political Economics and Public Finance," in Handbook of Public Economics, vol. 3, edited by A. J. Auerbach and M. Feldstein. Amsterdam: Elsevier Science, 2002, pp. 1549-659.

Sloan, F. A., and J. G. Trogdon. "The Impact of the Master Settlement Agreement on Cigarette Consumption." Journal of Policy Analysis and Management, 23, 2004, 843-55.

Sloan, F. A., C. A. Mathews, and J. G. Trogdon. "Impacts of the Master Settlement Agreement on the Tobacco Industry." Tobacco Control, 13, 2004, 356-61.

Sloan, F. A., J. Ostermann, and G. Picone. The Private and Social Cost of Smoking. Manuscript. Durham: Duke University, 2003.

Sloan, F. A., J. S. Allsbrook, L. K. Madre, L. E. Masselink, and C. A. Mathews. "States' Allocations of Funds from the Tobacco Master Settlement Agreement." Health Affairs, 24(1), 2005a, 220-27.

Sloan, F. A., E. S. Carlisle, J. R. Ratliff, and J. G. Trogdon. "Determinants of States' Allocations of the Master Settlement Agreement Payments." Journal of Health Politics, Policy and Law, 30(4), 2005b, 643-86.

Stigler, G. J. "Economic Competition and Political Competition." Public Choice, 13, 1972, 91-106.

Tax Foundation. Various State Tax Rates. Available online at www.taxfoundation.org/variousrates.html; accessed January 7, 2003.

U.S. Department of Agriculture. Crop Production-Annual Summary. Washington, DC: U.S. Department of Agriculture, 1991-2002.

U.S. Department of Commerce. Statistical Abstract of the United States. Washington, DC: U.S. Department of Commerce, Bureau of the Census, 2001.

U.S. Department of Commerce, Bureau of Economic Analysis. Regional Accounts Data, available online at www.bea.doc.gov/bea/regional/data.htm; accessed March 4, 2003.

U.S. Department of Health and Human Services. Reducing the Health Consequences of Smoking: 25 Years of Progress. Atlanta: U.S. Dept. Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, National Center Chronic Disease Prevention and Health Promotion, Office of Smoking and Health, 1989.

U.S. Department of Justice. Sourcebook of Criminal Justice Statistics. Washington DC: U.S. Dept. of Justice, Bureau of Justice Statistics, 1990-2002.

Viscusi, W. K. "The Governmental Composition of the Insurance Costs of Smoking." Journal of Law and Economics, 42, 1999, 575-609.

--. Smoke-Filled Rooms: A Postmortem on the Tobacco Deal. Chicago: University of Chicago Press, 2002.

Winters, R. F. "The Politics of Taxing and Spending," in Politics in the American States: A Comparative Analysis, 6th ed., edited by V. Gray and J. Herbert. Washington, DC: Congressional Quarterly Press, 1996, 319-60.

(1.) The literature is inconclusive regarding the issue of whether existing excise taxes are at the socially optimal level. There is some empirical evidence that state cigarette excise taxes are set at the level on average that accounts for the externalities that cigarette consumption causes, unlike state beer taxes, which are too low given the externalities from excess alcohol consumption (see Manning et al. 1989, 1991). But these results depend on treatment of secondary smoke within households (see Sloan, Ostermann, and Picone 2003) and assumptions of rationality. Assuming that smokers lack self-control, Gruber and Koszegi (2001) argued that for taxes to account for the internal costs of smoking in terms of life/years lost, taxes be set so as to yield a price of $30.45 per pack.

(2.) These results are sensitive to controls for unobserved factors at the state level through state fixed effects; see, for example, Dee (1999). All of our specifications include state fixed effects.

(3.) We thank an anonymous referee for suggesting such a flexible model.

(4.) If these time periods did impact beer excise taxes, then the estimated coefficients represent the difference in their impact on cigarette taxes relative to beer taxes.

(5.) See the appendix for a thorough discussion of data sources and variable construction.

(6.) All dollar (cents) values were converted to real 2002 dollars (cents) for the analysis.

(7.) Voter participation is also important. On average, smokers tend to be poorer, and the poor are less likely to vote.

(8.) Benjamin and Dougan (1997) developed a model of cigarette tax determination in which the relative ease with which cigarettes can be transported across state boundaries constrains their efficient taxation. From this model, they obtain the prediction that excise taxes rise at a decreasing rate as one moves outward from the point of production. Although they emphasized North Carolina as the key cigarette-producing state, there are three major tobacco-producing states: North Carolina, Virginia, and Kentucky.

(9.) Although our analysis is fundamentally economic, there is a history of whether public policy choices can be understood as a conventional economic good or alternatively whether it is desirable to introduce noneconomic preference variables into the study (Kahn and Matsusaka 1997; Peltzman 1976).

(10.) Some governors in these elections are termlimited, which would reduce the likelihood of finding a significant impact from the election cycle variables. However, the intuition for the model would still apply to these cases if the governor cared about the party's outcome in the next election and voters associated tax increases with the party as well as the individual governor.

(11.) The eight states are California, Delaware, Louisiana, Maryland, Missouri. North Carolina, Tennessee, and Virginia.

(12.) There is one variable for the proportion of state residents that exercised regularly and another indicator variable for state/year combinations where this information is missing. Thus, the equation is overidentified. Additional instruments could have easily been included using the lags of these instruments.

JUSTIN G. TROGDON and FRANK A. SLOAN *

* This research was supported in part by a grant from the Robert Wood Johnson Foundation administered by the foundation's Substance Abuse Policy Research Program titled Economic Analysis of Tobacco Litigation.

Trogdon: Department of Economics, Duke University, Box 90097, Durham, NC 27708-0097. Phone 1-919-4913503, Fax 1-919-684-8047, E-mail justin.trogdon@ duke.edu

Sloan: Department of Economics, Duke University, Box 90097, Durham, NC 27708-0097. Phone 1-919-613-9358, Fax 1-919-684-6246, E-mail fsloan@duke.edu

TABLE 1
Summary Statistics: 1990-2002

Variable Mean SE

Cigarette
State excise tax on cigarettes 40.67 26.24
 (cents/pack)
Smoking rate (%) 23.16 2.96
Tobacco leaf production (millions lbs.) 25.66 96.08
Federal excise tax on cigarettes 29.34 4.46
 (cents/pack)
Smoker protection law 0.55 0.50
Clean air index: nominal 0.13 0.33
Clean air index: basic 0.15 0.36
Clean air index: moderate 0.26 0.44
Clean air index: extensive 0.42 0.49
Real tobacco settlement payments 8.51 18.20
 ($/capita)
Maximum tax in bordering state 56.90 27.28
Minimum tax in bordering state 19.94 16.29
Appropriated ending balance net of 0.03 0.04
 tobacco revenues (millions $)

Beer
State excise tax on beer 2.71 2.13
 (cents/12-ounce drink)
Federal excise tax on beer 6.10 0.74
 (cents/12-ounce drink)
Drinking rate 52.07 10.66
Appropriated ending balance net of 0.03 0.04
 alcohol revenues (millions $)
Maximum tax in bordering state 4.00 2.70
Minimum tax in bordering state 1.03 0.88
Blood alcohol concentration: illegal 0.10 0.01
 per se
Open container law by statute 0.62 0.49
Anticonsumption law by statute 0.80 0.40
Dram shop law by statute 0.77 0.42
Dram shop law via case law 0.11 0.32

Explanatory variables in both
 cigarette and beer analysis
Gubernatorial election year 0.29 0.45
Gubernatorial off year 0.47 0.50
Index of Democratic control (0 to 1) 0.48 0.35
Single-party control 0.42 0.49
Real income ($1000/capita) 27.66 4.28
Population ahe 10-19 years (%) 14.66 1.46
Population ahe 65 and over (%) 12.61 2.02
N 572

TABLE 2
Fixed Effect Regression Analysis: Cigarette
Tax (cents/pack) - Beer Tax (cents/drink)

 Robust
 Coefficient SE

1993 to 1997 3.165 2.122
1998 9.911 ** 3.425
1999 to 2002 10.482 ** 3.658
Cigarettes ([[beta].sup.c])
Smoking rate (%) -1.358 ** 0.458
Tobacco leaf production 0.053 ** 0.017
 (millions lbs.)
Federal excise tax on cigarettes 0.421 (+) 0.243
 (cents/pack)
Smoker protection law -2.436 2.873
Clean air index: nominal 5.786 4.187
Clean air index: basic 17.304 * 7.066
Clean air index: moderate -8.461 9.359
Clean air index: extensive -11.028 ** 3.300
Real tobacco settlement payments -0.044 0.033
 ($/capita)
Maximum tax in bordering state 0.220 ** 0.067
Minimum tax in bordering state 0.190 0.171
Appropriated ending balance net -384.572 * 162.381
 of tobacco revenues (millions $)

Beer ([[beta].sup.b])
Drinking rate (%) -0.146 0.110
Federal excise tax on beer (cents/ 0.643 0.868
 12-ounce drink)
Blood alcohol concentration: 91.775 111.490
 illegal per se
Open container law by statute -11.496 * 4.478
Anticonsumption law by statute 0.727 3.531
Dram shop law by statute 59.946 ** 16.067
Dram shop law via case law -0.051 2.761
Maximum tax in bordering state 0.170 0.456
Minimum tax in bordering state 1.902 4.883
Appropriated ending balance net -349.884 * 164.900
 of alcohol revenues (millions $)

Cigarettes and beer ([[beta].sup.c]-
 [[beta].sup.b])
Gubernatorial election year -0.202 1.636
Gubernatorial off year -1.226 1.345
Index of Democratic control (0 to l) 11.198 ** 3.383
Single-party control -1.047 1.542
Population age 10-19 years (%) 7.381 ** 1.804
Population age 65 and over (%) 2.271 2.714
Real income ($1000/capita) -1.493 (+) 0.858
Constant: [[beta].sub.1] = -27.173 47.661
 ([[gamma].sup.c]-[[gamma].sup.b])
N 572
[R.sup.2] 0.81

Notes: The specification includes fixed effects for states.

(+) Significant at the 0.10 level based on a two-tailed test.

* Significant at the 0.05 level based on a two-tailed test.

** Significant at the 0.01 level based on a two-tailed test.

TABLE 3
Fixed-Effect IV Regression Analysis: Cigarette
Tax (cents/pack) - Beer Tax (cents/drink)

 Coefficient SE

1993 to 1997 -0.813 5.680
1998 5.558 6.756
1999 to 2002 6.545 6.576
Cigarettes ([[beta].sup.c])
Smoking rate (%) -3.465 3.900
Tobacco leaf production (millions lbs.) 0.048 0.045
Federal excise tax on cigarettes (cents/pack) 1.017 1.204
Smoker protection law 0.017 4.766
Clean air index: nominal 2.129 9.916
Clean air index: basic -- --
Clean air index: moderate -7.880 10.679
Clean air index: extensive -11.902 7.504
Real tobacco settlement payments -0.300 0.393
 ($/capita)
Maximum tax in bordering state 0.186 * 0.079
Minimum tax in bordering state 0.171 0.124
Appropriated ending balance net -449.441 * 217.710
 of tobacco revenues (millions $)

Beer ([[beta].sup.b])
Drinking rate (%) -0.820 1.380
Federal excise tax on beer (cents/ 1.804 2.108
 12-ounce drink)
Blood alcohol concentration: illegal 108.688 158.942
 per se
Open container law by statute -12.631 ** 3.739
Anticonsumption law by statute 2.353 4.004
Dram shop law by statute -- --
Dram shop law via case law 0.643 6.025
Maximum tax in bordering state 0.720 1.020
Minimum tax in bordering state -0.154 7.052
Appropriated ending balance net -413.614 (+) 216.787
 of alcohol revenues (millions $)

Cigarettes and beer ([[beta].sup.c]-
 [[beta].sup.b])
Gubernatorial election year -0.876 1.985
Gubernatorial off year -1.160 1.602
Index of Democratic control (0 to 1) 10.867 ** 3.081
Single-party control -0.087 1.986
Population age 10-19 years (%) 7.580 ** 2.126
Population age 65 and over (%) 3.565 2.964
Real income ($1000/capita) -0.886 1.277
Constant: [[beta].sub.1] = ([[gamma].sup.c]- -79.564 74.189
 [[gamma].sup.b])
N 572
Hausman test for endogeneity 0.44
Davidson and MacKinnon test for 0.22
 endogeneity

Notes: The specification includes fixed effects for
states. Smoking rates, drinking rates, and state settlement
payments are instrumented with state-level pregnancy
rates, exercise rates, indicator variable for missing exercise
rates, and marriage rates. The instruments are jointly sig-
nificant at the 1% confidence level in the first stages. We
fail to reject the null hypothesis that the instruments are
orthogonal to the error term in the estimation equation.

(+) Significant at the 0.10 level based on a two-tailed test.

* Significant at the 0.05 level based on a two-tailed test.

** Significant at the 0.01 level based on a two-tailed test.