Menthol and non-menthol smoking: the impact of prices and smoke-free air laws

Authors

  • John A. Tauras,

    Corresponding author
    1. University of Illinois at Chicago and National Bureau of Economic Research, Chicago, IL, USA
      John Tauras, Department of Economics (m/c 144), University of Illinois at Chicago, 601 S. Morgan, Chicago, IL 60607-7121, USA. E-mail: tauras@uic.edu
    Search for more papers by this author
  • David Levy,

    1. University of Baltimore and Pacific Institute for Research and Evaluation, Baltimore MD, USA
    Search for more papers by this author
  • Frank J. Chaloupka,

    1. University of Illinois at Chicago and National Bureau of Economic Research, Chicago, IL, USA
    Search for more papers by this author
  • Andrea Villanti,

    1. The Schroeder Institute for Tobacco Research and Policy Studies at Legacy and The Johns Hopkins Bloomberg School of Public Health, NW Washington, DC, USA
    Search for more papers by this author
  • Raymond S. Niaura,

    1. The Schroeder Institute for Tobacco Research and Policy Studies at Legacy and The Johns Hopkins Bloomberg School of Public Health, NW Washington, DC, USA
    Search for more papers by this author
  • Donna Vallone,

    1. The Schroeder Institute for Tobacco Research and Policy Studies at Legacy and The Johns Hopkins Bloomberg School of Public Health, NW Washington, DC, USA
    Search for more papers by this author
  • David B. Abrams

    1. The Schroeder Institute for Tobacco Research and Policy Studies at Legacy and The Johns Hopkins Bloomberg School of Public Health, NW Washington, DC, USA
    Search for more papers by this author

John Tauras, Department of Economics (m/c 144), University of Illinois at Chicago, 601 S. Morgan, Chicago, IL 60607-7121, USA. E-mail: tauras@uic.edu

ABSTRACT

Aims  To examine the relationship between menthol and non-menthol prices and smoke-free air laws and the choice between menthol and non-menthol cigarettes among current smokers.

Design, setting and participants  Data were extracted from the nationally representative (USA) 2003 and 2006/07 Tobacco Use Supplements to the Current Population Survey. A total of 57 383 adult smokers (aged 18+) were examined.

Measurements  A regression model was used to estimate the probability of being a menthol smoker conditional on being a current smoker who had a distinct preference for either non-menthol or menthol cigarettes. Cigarette prices, smoke-free air laws and socio-economic and demographic characteristics were examined as covariates.

Findings  The prices of menthol and non-menthol cigarettes were associated with the choice between menthol and non-menthol cigarettes. However, smokers did not find menthol and non-menthol cigarettes to be close substitutes for one another. Non-menthol cigarettes were found to be less of a substitute for menthol cigarettes than vice versa. Young adults and African Americans were less responsive to prices with respect to switching between menthol and non-menthol cigarettes than were older adults and non-African Americans, respectively.

Conclusions  The US Food and Drug Administration (FDA) is grappling with the issue of whether or not to ban menthol cigarettes. The findings from this study suggest that smokers do not find menthol and non-menthol cigarettes to be close substitutes. The strong preference for mentholated cigarettes may serve as a lever to reduce smoking prevalence when combined with increased access to effective cessation treatments.

INTRODUCTION

On 22 June 2009, President Obama signed legislation into law granting the US Food and Drug Administration (FDA) the authority to regulate tobacco products. The legislation also banned candy and spice-flavored cigarettes, but exempted mentholated cigarettes from the ban. Now the FDA is grappling with the issue of whether to ban menthol cigarettes, which is the most popular cigarette flavoring smoked by millions of Americans daily. In 2006, the market share of menthol cigarettes was 20% [1].

Menthol cigarettes are popular with African Americans and young smokers. According to the 2006/07 Tobacco Use Supplement to the Current Population Survey (TUS CPS), approximately 70% of African American smokers (approximately 3 million African Americans) smoke menthol cigarettes [2]. While the evidence on the relative risks of smoking menthol cigarettes is unclear, African Americans smoke fewer cigarettes on average than white smokers, yet have higher rates of lung cancer and other tobacco-related diseases [3,4]. Moreover, the rate of smoking menthol cigarettes among young smokers has been increasing in recent years. According to the National Survey on Drug Use and Health (NSDUH) [5], the rate of smoking menthol cigarettes among past month smokers increased from 43.5 to 47.7% among adolescents aged 12–17 years and from 34.1 to 40.8% among young adults aged 18–25 years between 2004 and 2008.

Some recent research suggests that menthol flavoring may play a role in smoking initiation and cessation. For example, Hersey et al. [6] found that adolescent smokers who had smoked for less than 1 year had a higher preference (51%) for menthol cigarettes than those who had smoked for more than 1 year (43.6%), and that menthol smoking prevalence was higher among middle school students than high school students. In the NSDUH adolescent and young adult smokers who had initiated smoking in the past year were more likely to smoke menthol cigarettes than those who had smoked for more than a year [5]. Hersey and colleagues [6] speculated that menthol cigarettes may serve as starter cigarettes because they are perceived as less harsh to smoke. Some studies have examined the effects of menthol on smoking cessation. While a few studies have found no statistically significant differences in rates of smoking cessation by menthol status [8,9], a number of recent studies have. In particular, studies by Okuyemi and colleagues (2003, 2004, 2007), Harris et al. (2004), Foulds et al. (2006) and Gandhi et al. (2009) all found that individuals who smoke menthol cigarettes are significantly less successful in quitting than individuals who smoke non-menthol cigarettes [10–15].

Given the sizeable market share of menthol cigarettes, the popularity of menthol cigarettes among certain subgroups in the United States and the potential role menthol plays in smoking initiation and cessation, a ban on menthol cigarettes would probably have both economic and public health implications. This study attempts to inform the debate by looking at the economic relationship between the demands for mentholated and non-mentholated cigarettes among current smokers using the two most recent waves of the TUS CPS. In particular, this study examines how menthol and non-menthol prices and smoke-free air laws are associated with the choice between menthol and non-menthol cigarettes among current smokers and it provides evidence on the degree of substitutability that exists between menthol and non-menthol cigarettes.

METHODS

Data

In our analyses, we used data extracted from the three TUS CPS surveys conducted in 2003 (February, June and November) and the three TUS CPS surveys conducted in 2006/07 (May and August, 2006; January, 2007). The CPS is a monthly survey of approximately 50 000 households that has been conducted for more than 50 years by the Bureau of Labor Statistics of the US Census Bureau. Representing the civilian non-institutional population, the CPS focuses on labor force characteristics of the US, including employment, occupation, economic status, earnings, hours worked and other indicators. During September 1992, questions regarding tobacco use were asked of all adult respondents as part of the first TUS CPS. Additional questions regarding tobacco use were added in later waves of the TUS CPS. In the last two waves of the TUS CPS (2003 and 2006/07), self-responding smokers were asked questions about the type of cigarette smoked (i.e. mentholated or non-mentholated) and asked about the price they paid for the last pack/carton of cigarettes that they purchased. These questions, in conjunction with the large sample size, nationally representative design with sampling in each state and DC, make the TUS CPS an ideal data set to examine the economic determinants of menthol and non-menthol cigarette demand.

The question asked among smokers about the type of cigarette they smoke was used to create the dependent variable that was used in all the model specifications. In particular, the dependent variable took on a value of 1 for respondents who were current smokers and indicated that their usual brand was a mentholated brand and took on a value of zero for respondents who were current smokers and indicated that their usual brand was a non-mentholated brand. [Individuals who did not provide information on their brand preference (i.e. menthol or non-menthol) were excluded from the analyses. Also, a small fraction of smokers did not have a usual brand preference (i.e. they did not have a preference for menthol or non-menthol cigarettes). These observations were also excluded because the focus of the analyses is on choice between menthol and non-menthol consumption.]

Of particular importance to this research was the information collected in the TUS CPS on cigarette prices. All smokers were asked if they usually purchase cigarettes by the pack, carton, or both pack and carton. Based on this response, the smokers were asked about the price they last paid for either a pack or carton. Combining the price information with usual type of cigarette smoked, we constructed two alternative cigarette price measures: average state-specific price for a pack of menthol cigarettes and the average state-specific price for a pack of non-menthol cigarettes. [Using individual-level prices would lead to biased price coefficients (away from zero) due to endogeneity. The endogeneity problem arises because of the availability of different brands and different quality of cigarettes. In essence, the consumer exercises some choice over the price they pay rather than the price being completely exogenous. This implies that the price variable may be correlated with unobserved differences in preferences, leading to biased price coefficients.] The average state-specific cigarette prices were constructed by eliminating the top and bottom 2% of state-specific menthol and non-menthol pack prices to avoid keying errors and then deriving mean estimates employing the supplement self-response adjustment weight (prices reported for carton were divided by 10 to obtain an average pack price). Finally, the state-specific average menthol and non-menthol pack prices were deflated by the consumer price index (1982–84 = 1) to adjust the prices for inflation over time. [The current consumer price index is based on a 1982–84 base of 1. A CPI index value of 1.85 would indicate that prices have increased by 85% compared to the average prices across the base years of 1982–84.] The average price of menthol cigarettes in our sample was slightly higher than the average price of non-menthol cigarettes. Adjusting the prices to 2010 dollars, the average price for a pack of menthol cigarettes across both waves of data was $3.88 and the average price for non-menthol cigarettes was $3.79. Table 1 provides the state-specific average prices and standard deviations for a pack of menthol cigarettes and non-menthol cigarettes across both waves of data.

Table 1.  State-level menthol and non-menthol cigarette price descriptive statistics (in 2010 dollars).
StateMean menthol priceSD menthol priceMean non-menthol priceSD non-menthol price
  1. SD: standard deviation.

AK$5.2040.522$5.2590.504
AL$3.3330.164$3.1840.101
AR$3.3730.192$3.1930.078
AZ$4.4550.463$4.4130.272
CA$4.2870.126$4.2330.092
CO$3.4850.233$3.3960.279
CT$4.8800.076$4.9810.127
DC$4.2620.164$4.5290.331
DE$3.1990.093$3.1790.071
FL$3.2730.120$3.1100.080
GA$3.2930.055$3.0570.082
HI$4.6270.262$4.6750.332
IA$3.1220.111$3.0780.065
ID$3.3850.222$3.3120.110
IL$4.3920.256$4.1370.083
IN$3.3990.092$3.3870.077
KS$3.4380.094$3.5560.084
KY$2.8130.323$2.5970.155
LA$3.4220.098$3.2120.066
MA$5.1160.209$5.0840.155
MD$4.0150.112$3.9280.137
ME$4.5370.367$4.6090.431
MI$4.5500.255$4.4950.294
MN$4.0470.413$3.8630.295
MO$3.0780.273$2.9420.188
MS$3.0320.090$2.9220.105
MT$3.8200.597$3.8680.588
NC$3.0700.212$2.8690.106
ND$3.2550.151$3.2000.067
NE$3.5720.138$3.4710.118
NH$3.5910.177$3.5680.126
NJ$5.5030.373$5.4730.355
NM$3.6530.346$3.5690.211
NV$3.7160.249$3.6340.159
NY$5.4460.223$4.9520.197
OH$3.7550.267$3.6850.216
OK$3.1460.334$3.0980.283
OR$4.2660.235$4.0440.100
PA$4.1640.199$4.1040.172
RI$5.2280.388$5.1990.342
SC$2.8980.087$2.8110.122
SD$3.4220.417$3.5250.361
TN$2.9660.191$2.8590.082
TX$3.7190.421$3.5390.341
UT$3.9750.320$3.8760.133
VA$3.1060.196$3.0490.153
VT$4.3640.369$4.4960.360
WA$4.9710.278$4.6590.189
WI$3.8050.212$3.6830.150
WV$2.8210.188$2.9470.118
WY$3.3710.241$3.3160.243

Research commissioned by the National Cancer Institute examined different measures of US cigarette prices obtained from alternative data sources and collected using different methods. These included the widely used cigarette prices from the Tax Burden on Tobacco (TBOT), the self-reported prices collected as part of the TUS CPS and other price measures [16]. The final report concluded that the self-reported prices from the TUS CPS are valid measures of price. They are correlated highly with state and local cigarette taxes and with the most widely used TBOT cigarette prices.

Several other independent variables were constructed to control for factors thought likely to affect menthol/non-menthol choice. These factors included: indicators for age of the respondent (18–24, 25–34, 35–64 and 65 years and older reference category); gender (male and female: reference category); indicators of race/ethnicity [Hispanic, non-Hispanic (NH) African American, NH Asian, NH Native American/Alaskan native, NH Hawaiian/Pacific Islander, NH multiple race and NH white: reference category]; indicators of education (high school graduate, some college: no degree, earned associates degree, earned bachelor's degree, earned master's degree, earned doctorate degree and less than high school degree: reference category); indicators of marital status (widowed, divorced, separated, single and married: reference category); and employment status indicators (unemployed, not in the labor force, employed: reference category). In addition, an inflation-adjusted (1982–84 = 1) quasi-continuous family income variable was created using the mid-points of the categorical responses. [The highest income category that could be chosen in February and June 2003 was $75 000 or more. The highest income category that could be chosen in November 2003, May 2006, August 2006 and January 2007 was $150 000 or more. For individuals who chose the highest income category, a value of $87 499.5 was chosen for those surveyed in February and June 2003 and a value of $165 000 was chosen for those surveyed in November 2003, May 2006, August 2006 and January 2007. Sensitivity analyses around the value assigned to the top category was conducted with only small changes in coefficient estimates and significance levels occurring. The results from the sensitivity analyses are available upon request.] Because the highest income category is right-censored in the TUS CPS, a dichotomous indicator was created to take into account this right-censoring. The indicator took on a value of 1 if the individual chose the highest income category and took on a value of zero otherwise.

Finally, US Census division indicators (North East, Mid Atlantic, West North Central, South Atlantic, East South Central, West South Central, Mountain and West: reference category) and survey indicators (June 2003, November 2003, May 2006, August 2006; January 2007 and February 2003: reference category) were constructed from the survey data.

Using state identifiers, smoke-free air laws were merged with the survey data. To capture the overall magnitude of each state's smoke-free air legislation and to decrease the collinearity associated with simultaneously including highly correlated state level policies in the equations, a smoke-free air index variable was created. The index variable accounted for smoke-free air laws in three venues: private work-sites, restaurants and bars. For each of the venues a restriction rating of between 0 and 2 was assigned. The restriction rating was based on the strength of protection. Specifically, if smoking was prohibited in a particular venue, the restriction rating for that venue was assigned a value of 2; if smoking was restricted in a particular venue, but was less comprehensive than a total ban, the restriction rating for that venue was assigned a value of 1; finally, if there were no restrictions on smoking in a certain venue, then the restriction rating for that venue was assigned a value of 0. The index was derived by adding up the equally weighted restriction ratings for each of the three venues, such that the minimum and maximum values for the index in any given state/month were 0 and 6, respectively.

Finally, given the popularity of smoking menthol among African Americans and young smokers, interaction variables were constructed to see if African Americans or young adults responded differently to cigarette prices than did non-African Americans or older adults. In particular, the following interaction terms were created: non-menthol price × African American; menthol price × African American; non-menthol price × age 18–24; and menthol price × age 18–24. If the coefficients for the interaction terms were statistically significant, then this would imply that a difference existed in price responsiveness between the groups.

Empirical methods

A linear probability model was used to estimate the probability of being a menthol smoker conditional on being a current smoker who has a distinct preference for either non-menthol or menthol cigarettes (current smokers who have no preference for either menthol or non-menthol cigarettes were excluded from the equations). The unit of analysis is at the individual level. A total of 57 383 adult smokers (aged 18+) were examined. A linear probability model was chosen because the interaction terms are more readily interpreted than in non-linear techniques such as logit or probit models. [Unlike linear models, in non-linear models the interaction effect cannot be evaluated simply by looking at the sign, magnitude or statistical significance of the coefficient on the interaction term. Instead, the effect of the interaction between two variables in non-linear models requires computing cross derivatives. We estimated logit equations and found the results to be comparable to those obtained using a linear probability model.] Three equations were estimated. The first equation included the following covariates: inflation-adjusted price of menthol cigarettes, inflation-adjusted price of non-menthol cigarettes, smoke-free air index, inflation-adjusted family income, income right-censoring indicator, age indicators, gender indicator, marital status indicators, race/ethnicity indicators, education indicators, employment indicators, US Census division indicators and survey month indicators. Model 2 was identical to model 1, but added the two interaction variables between cigarette prices and African American status. Model 3 was identical to model 2, but replaced the African American price interactions with the age 18–24 price interactions. Table 2 contains the estimates from the three equations. In all the equations that were estimated, the linear probability model uses variation in cigarette prices both between states and within states over time to identify the impact of menthol and non-menthol prices on the choice between menthol and non-menthol consumption among smokers. Table 1 provides descriptive statistics on state-specific menthol and non-menthol prices.

Table 2.  Probability of being a menthol smoker, conditional on being a smoker.
Independent variablesModel 1Model 2Model 3
  1. t-Statistics are in parentheses. *P < 0.10; **P < 0.05; ***P < 0.01, based on a two-tailed test. The dependent variable is the probability of being a menthol smoker conditional on being a smoker with a distinct preference for either menthol or non-menthol cigarettes. Model 1 includes the following covariates: inflation-adjusted price of menthol cigarettes, inflation-adjusted price of non-menthol cigarettes, smoke-free air index, inflation-adjusted family income, income right censoring indicator, age indicators, gender indicator, marital status indicators, race/ethnicity indicators, education indicators, employment indicators, US Census division indicators and survey month indicators. Model 2 is identical to model 1, but adds the two interaction variables between cigarette prices and African American status. Model 3 is identical to model 2, but replaces the African American price interactions with the age 18–24 price interactions. NH: non-Hispanic; SFA: smoke-free air.

Menthol price−0.03* (–1.69)−0.05** (−2.24)−0.07*** (−3.09)
Non-menthol price0.07*** (3.30)0.09*** (4.06)0.10*** (4.44)
Menthol price × black 0.09 (1.37) 
Non-menthol price × black −0.16** (−2.37) 
Menthol price × age 18–24 years  0.29*** (4.85)
Non-menthol price × age 18–24 years  −0.27*** (−4.27)
SFA index0.003** (2.46)0.003*** (2.59)0.003** (2.49)
Real family income0.0000003** (2.38)0.0000003** (2.40)0.0000003** (2.34)
Income right censored0.002 (0.19)0.001 (0.15)0.002 (0.18)
Age 18–24 years0.04*** (3.90)0.04*** (3.88)−0.02 (−0.70)
Age 25–34 years−0.01 (−1.07)−0.01 (−1.09)−0.01 (−1.08)
Age 35–64 years0.02*** (2.92)0.02*** (2.90)0.02*** (2.91)
Male−0.09*** (−26.85)−0.09*** (−26.88)−0.09*** (−26.85)
Widow−0.00 (−0.27)−0.00 (−0.26)−0.00 (−0.27)
Divorce0.01* (1.85)0.01* (1.88)0.01* (1.83)
Separate0.02** (2.35)0.02** (2.36)0.02** (2.35)
Single0.03*** (5.93)0.03*** (5.95)0.03*** (5.92)
Hispanic0.10*** (12.77)0.10*** (12.73)0.10*** (12.79)
NH: black0.51*** (71.82)0.62*** (17.35)0.51*** (71.79)
NH: American Indian/Alaskan Native0.002 (0.18)0.002 (0.15)0.003 (0.20)
NH: Asian0.15*** (9.18)0.15*** (9.12)0.15*** (9.18)
NH: Hawaiian/Pacific Islander0.38*** (9.51)0.38*** (9.48)0.38*** (9.57)
NH: multiple race0.09*** (7.41)0.09*** (7.38)0.09*** (7.40)
High school0.01** (2.42)0.01** (2.41)0.01** (2.45)
Some college0.01 (1.45)0.01 (1.47)0.01 (1.47)
Associate degree0.02*** (2.86)0.02*** (2.86)0.02*** (2.88)
Bachelor degree−0.02*** (−3.07)−0.02*** (−3.07)−0.02*** (−3.01)
Masters degree−0.001 (−0.08)−0.001 (−0.09)−0.001 (−0.05)
Doctoral degree0.01 (0.27)0.01 (0.26)0.01 (0.29)
Unemployed0.02*** (2.77)0.02*** (2.79)0.02*** (2.75)
Not in labor force−0.005 (−1.07)−0.005 (−1.02)−0.005 (−1.10)
North East0.04*** (5.62)0.04*** (5.42)0.04*** (5.65)
Mid-Atlantic0.13*** (14.96)0.13*** (15.04)0.13*** (15.08)
East North Central0.11*** (15.29)0.12*** (15.50)0.11*** (15.34)
West North Central0.03*** (4.22)0.04*** (4.49)0.03*** (4.32)
South Atlantic0.10*** (11.81)0.10*** (12.07)0.10*** (11.84)
East South Central0.05*** (4.62)0.05*** (4.79)0.05*** (4.64)
West South Central0.03*** (2.87)0.03*** (3.10)0.03*** (2.91)
Mountain−0.01 (−0.96)−0.01 (−0.69)−0.01 (−0.91)
June 20030.004 (0.63)0.004 (0.65)0.004 (0.68)
November 20030.01 (0.99)0.01 (1.00)0.01 (1.00)
May 2006−0.003 (−0.50)−0.003 (−0.55)−0.003 (−0.50)
August 2006−0.004 (−0.58)−0.004 (−0.64)−0.004 (−0.58)
January 2007−0.01 (−0.86)−0.01 (−0.95)−0.01 (−0.84)
Menthol price elasticity−0.236−0.326−0.455
Non-menthol price elasticity+0.475+0.615+0.669
n57 38357 38357 383

We also estimated several models that contained interaction terms between income (and low-income status, where low-income status was defined as being in the lowest quartile) and both cigarette price measures and gender and both cigarette prices. Moreover, we estimated models that contained interaction terms between the smoke-free air index variable and African American status, between the smoke-free air index and gender and between the smoke-free air index and young adult status. In none of these equations were the interaction terms significantly different from zero, suggesting no differential response by these factors to prices or smoke-free air laws. Results are available upon request.

We also estimated equations that included a variable measuring state-level inflation-adjusted tobacco control spending per capita and the coefficient was not significant, but did not appreciably affect coefficient of other variables. In addition, we estimated equations for the two waves (2002/03 and 2006/07) separately and found similar results to those presented in this paper that combined the two waves. We also included various specifications of dummy variables for the classification of states, which also did not affect results substantially, except when separate indicators were included for all states (less 1 as the benchmark). This latter result reflects the fact that the combination of state and time indicators alone explained nearly 94% of the variation in the price measures, leaving very little independent variation in prices to affect the dependent variable.

Finally, in all specifications, we used a robust method of calculating the variance–covariance matrix of the estimators developed by Huber (1967) [7]. All equations were estimated using STATA version 9.2 software (http://www.statacorp.com).

RESULTS

The total number of respondents who provided information on usual cigarette type across both waves of data was 65 221 people. This included: 16 294 menthol smokers; 46 899 non-menthol smokers; and 2028 no usual type smokers. Applying the replicate weights provided by the National Cancer Institute, and excluding smokers with missing information on usual cigarette type it was estimated that 26.63%, 70.03% and 2.34% of current smokers smoke menthol, smoke non-menthol and have no usual type, respectively, in the United States.

The results from all the estimated equations suggest that cigarette prices were associated with the choice between menthol and non-menthol cigarettes among current smokers. In all equations, higher menthol prices were related inversely to the probability of being a menthol smoker and related positively to the probability of being a non-menthol smoker. Similarly, higher non-menthol prices were associated positively to the probability of being a menthol smoker and related inversely to the probability of being a non-menthol smoker. The significant price associations imply that some smokers substituted between menthol and non-menthol as the relative prices of menthol and non-menthol cigarettes changed. Price elasticities of cigarette choice were calculated in an attempt to measure how closely menthol and non-menthol cigarettes substituted for each other. [Using derivatives, the price elasticity is (dq/dp) × (p/q), where q is the dependent variable and p is price. If the percentage change in the dependent variable is greater than the percentage change in price, then the price elasticity will be greater than 1 in absolute value (known as price elastic). If the percentage change in the dependent variable is less than the percentage change in price, the absolute value of the price elasticity will be smaller than 1.0 (known as price inelastic).] In particular, we estimate a conditional menthol to non-menthol price elasticity which measured the impact of menthol prices on the probability of a smoker choosing menthol cigarettes, holding all other factors constant. In model 1, the conditional menthol to non-menthol price elasticity was estimated to be −0.23. This suggests that a 10% increase in the price of menthol cigarettes was associated with a decreased probability of being a menthol smoker by 2.3%. We also estimated a conditional menthol to non-menthol cross-price elasticity which measured the impact of non-menthol prices on the probability of a smoker choosing menthol cigarettes, holding all other factors constant. In model 1, the conditional menthol to non-menthol cross-price elasticity was estimated to be +0.47. This suggests that a 10% increase in the price of non-menthol cigarettes was associated with an increase in the probability of being a menthol smoker by 4.7%. The inelastic conditional menthol to non-menthol cross-price elasticity suggests that menthol and non-menthol cigarettes were not close substitutes for one another.

Using the elasticity estimates from model 1 above, the population effects associated with raising the price of menthol and non-menthol prices by 10% can be compared. Estimates from the 2006/07 TUS CPS suggest that 10 662 780 individuals are menthol smokers in the United States and 28 000 075 individuals are non-menthol smokers. The elasticity estimates from this study suggest that holding non-menthol prices (and all other factors) constant, a 10% increase in the price of menthol cigarettes would be associated with 245 244 individuals (2.36%) who would switch from menthol to non-menthol cigarettes. A comparable 10% increase in the price of non-menthol cigarettes holding menthol prices (and all other covariates) constant would be associated with 501 150 individuals (4.75%) who would switch from non-menthol to menthol smoking. These results imply that a non-menthol price increase would be associated with more people switching from non-menthol to menthol than an equal percentage price increase for menthol cigarettes—which would be associated with fewer people switching from menthol to non-menthol.

With respect to the differential response by African Americans, the interaction effects in model 2 suggest that the relationship between the prices of non-menthol cigarettes and African Americans is different to that of other race/ethnicities. Specifically, the interaction effect between the non-menthol price and African American was negative, and significant suggesting that African Americans were less likely to switch between menthol and non-menthol cigarettes than other races and ethnicities when non-menthol prices change. We conducted Wald tests for linear hypotheses about the interaction coefficients. We first tested whether the coefficients on both interaction variables were jointly equal to zero. The results strongly rejected the null hypothesis that the coefficients were jointly equal to zero. We also tested whether the coefficients on the interaction variables were equal to one another. The equality of coefficient hypothesis was rejected at the 5.9% level.

Similarly, with respect to the differential response by young adults, the interaction effects in model 3 suggest that the relationship between cigarette prices and young adults was different to the relationship between cigarette prices and older adults. Specifically, the interaction effect between the non-menthol price and young adult status was negative and significant and the interaction effect between the menthol price and young adult status was positive and significant, suggesting that young adults were less likely to switch between menthol and non-menthol than older adults when the relative prices changed. Again, we conducted Wald tests on the coefficients of the interaction variables. The results strongly rejected the null hypothesis that the coefficients were jointly equal to zero and strongly rejected the null hypothesis that the coefficients on the interaction variables were equal to one another.

The smoke-free air index was found to be related positively to the probability of being a menthol smoker in all the models that were estimated. Inflation-adjusted family income was found to be associated positively to the probability of being a menthol smoker. Males were less likely to choose menthol smoking than females. Individuals who were divorced, separated or single were more likely to be menthol smokers than those who were married. Individuals who were aged 18–24 and 35–64 years were more likely to be menthol smokers than those aged 65+. Hispanics, African Americans, Asians, Hawaii and Pacific Islanders and individuals of multiple races were significantly more likely to be menthol smokers than whites. Individuals with high school degrees and associate degrees were more likely to be menthol smokers than those who had less than a high school education. Individuals with bachelor degrees were less likely to be menthol smokers than those who had less than a high school education. Individuals who were unemployed were more likely to smoke menthol cigarettes than those who were employed. Finally, individuals residing in the West division were significantly less likely to be menthol smokers than were individuals who reside in all other divisions of the country, except individuals who resided in the Mountain division.

DISCUSSION

The prices of menthol and non-menthol cigarettes were found to be related to the choice between menthol and non-menthol cigarettes. Economists would expect a cross-price elasticity greater than 1 if two related products are close substitutes. The elasticities reported in this study are substantially less than 1, implying that menthol and non-menthol cigarettes are not close substitutes. Moreover, the decision to smoke menthol is asymmetric. The size of the relationship for non-menthol prices was about twice that of menthol prices on the choice between menthol and non-menthol cigarettes. This result suggests that non-menthol smokers are more likely to switch to menthol cigarettes from an increase in non-menthol prices than are menthol smokers switching to non-menthol cigarettes in response to an increase in menthol prices. Thus, non-menthol cigarettes are less of a substitute for menthol cigarettes than vice versa.

There were no gender or income differences in price responsiveness (results for the gender and income interactions are available upon request). However, those who are young and African American were less responsive to price with respect to switching between menthol and non-menthol cigarettes. This result indicates that menthol cigarettes are even less of a substitute for non-menthol cigarettes to these groups than among non-African Americans and older smokers. This is consistent with tobacco industry documents citing the brand loyalty of young smokers [17,18] and national studies showing the stability of cigarette brand use among adolescent compared to adult smokers [19]. Studies of African American adolescents, young adults and adults also confirm strong menthol cigarette brand preference [5,20–23]. Another important finding is relatively more use of menthol cigarettes in states that have stronger laws restricting smoking. This may be due to menthol smokers being less likely to quit in response to smoke-free air laws or some other unexplained third variable. Interaction terms were also considered between smoke-free air laws and whether African American, young, or male, and did not find that these groups were more likely than others to be affected by smoke-free air laws in their choice between menthol and non-menthol cigarettes.

There are several limitations to this study. Only menthol smokers relative to non-menthol smokers were considered. Some smokers in the sample did not classify themselves as favoring either type, but these smokers constituted 2.3% of smokers, further indicating that the vast majority of smokers have definite preferences toward menthol or non-menthol cigarettes. Another limitation is that the TUS CPS is cross-sectional in nature. A final limitation is that we estimate models of brand choice among individuals aged 18+. [Only self-respondents answered the brand choice and cigarette price questions in the TUS CPS. The tobacco questions in the TUS CPS were asked only of those aged 18+ in January 2007. However, in May and August 2006 and waves in 2003, tobacco questions were asked of those aged 15+. To be consistent across waves of data, we used self-responding individuals aged 18+ in our analyses.] Future research on brand choices among younger adolescent smokers is needed.

There are several implications of the results of this study in the context of the Food and Drug Administration's (FDA) mandate to consider the recommendation of a scientific advisory committee report on the status of Menthol as a characterizing flavor, as provided by the Family Smoking Prevention and Tobacco Control Act. One option is for the advisory committee to recommend that the FDA ban menthol in cigarettes, and this study suggests several public health advantages of recommending such a ban. First, this strong preference for mentholated cigarettes may serve as a lever to reduce smoking prevalence when combined with increased access to effective cessation treatments. While past studies indicate that higher cigarette prices reduce smoking, our study suggests that smokers do not find menthol and non-menthol cigarettes to be close substitutes. This tendency appears to be especially strong among younger smokers and African Americans. Our results also suggest that menthol smokers are less likely to be affected by stronger tobacco control policies such as stronger smoke-free air laws. If menthol smokers are less likely to substitute to non-menthol cigarettes following restrictive regulation, this could result in decreased smoking prevalence among younger smokers, African Americans and adult menthol smokers, in general. Such beneficial changes in overall patterns of use would contribute to the broad public health standard of reducing or preventing the harms of smoking and protecting health.

Secondly, from a consumer perspective, perceptions of how a planned menthol ban is construed and implemented will probably influence patterns of tobacco use behavior in complex ways. In order to maximize the broad public health benefit of an outright ban or a gradual reduction in menthol, the public will need to be prepared with appropriated tailored messages and interventions directed at prevention of uptake among youth as well as to promote cessation among current users. Direct to consumer education campaigns should be planned and tested well in advance of implementation of a ban. Such campaigns will help to capitalize on the ‘teachable moment’ by educating the public about the opportunity to either not start smoking or to quit, and by providing easy and affordable access to proven cessation interventions. There may also be feelings of protest among some users, and thus an unintended backlash will need to be anticipated and averted in the educational campaigns that prepare the public for the ban. A carefully planned educational campaign will maximize the broad public health benefit of a menthol ban and minimize any misunderstandings, unintended consequences or attempts by the tobacco industry to encourage the public to resist any FDA action.

Finally, FDA-mandated post-implementation surveillance will be needed to track the impact on consumers, smoking prevalence, public health benefits and any possible unintended harm. Regardless of the decisions the FDA will make, this study provides useful information to consider when weighing the evidence of the benefits and harms of various options to ensure the greatest public health benefit.

Declarations of interest

Support for this research has been provided by the American Legacy Foundation. Some of Tauras' time has also been supported by the National Cancer Institute and American Legacy Foundation sponsored Tobacco Research Network on Disparities (TReND).

Ancillary