Do expenditures on tobacco control decrease smoking prevalence?

Abstract

Effectiveness of tobacco control programmes in reducing smoking prevalence during 2001 to 2005 is examined. Tobacco control spending is found to exert no significant effects on smoking prevalence across the 50 states. Cigarette prices are found to lower prevalence of daily smokers, but exert no effect on nondaily smoking prevalence. Several reasons are suggested for why these results might conflict with previous research. These include that most previous studies examined two states (California and Massachusetts) with long-standing tobacco control programmes and that most studies examined periods in which many of the states in their samples did not actively fund their programmes. Another reason may be that, unlike most previous studies, this study controls for the possibility that tobacco control spending is endogenous when, for example, states exhibiting relatively low smoking prevalence are also states with relatively high distaste for smoking and accordingly fund tobacco control programmes more generously. A negative relation between tobacco control spending and smoking prevalence does not necessarily indicate that higher spending causes lower prevalence when spending is endogenously determined.

Notes

¹ Orzechowski and Walker (2005).

² ‘Smoking Prevalence Among U.S. Adults,’ October 2006; taken from http://www.cdc.gov/tobacco/research_data/adults_prev/prevali.htm.

³ For a recent study of how cigarette tax increases in Japan are predicted to effectively lower smoking, see Wan (2006).

⁴ Farrelly et al. (1999) estimates that smoking bans on all US workplaces would reduce smoking prevalence by 10%. Lye and Hirschberg (2004) provide support for the view that governments may enhance worker productivity, and therefore wages, by regulating smoking and drinking. Boyes and Marlow (1996) find evidence of many smoke-free restaurants prior to passage of a smoking ban in one of the earliest cities in California to adopt such a ban and suggest that smoking bans may exert little or no effect on tobacco use if bans follow changes in smoking behaviour rather than cause such changes.

⁵ CDC (2001) and CDC (2002).

⁶ Ibrahim and Glantz (2003). Per capita calculation computed by the author using average California population over this period.

⁷ See American Nonsmokers’ Rights Foundation (2006) for the list of local jurisdictions and their enactment dates.

⁸ See Koh et al. (2005).

⁹ See American Nonsmokers’ Rights Foundation (2006) for the list of local jurisdictions and their enactment dates.

¹⁰ Luccasasen et al. (2005) argue that smuggling mitigates much of the health benefits of higher cigarette taxes. Tauras et al. (2007) find that smokeless tobacco products and cigarettes are economic complements in consumption and therefore policies that lower cigarette consumption will lower smokeless tobacco consumption as well.

¹¹ Tauras and Chaloupka (2004) find that, after controlling for many other factors, higher spending on tobacco control changes youth's attitudes and beliefs towards smoking. Pechmann and Reibling (2006) find that spending on antismoking advertisements lowers youths’ smoking intentions as long as they focus on tobacco-related disease and suffering. However, counterindustry advertisements were not found to lower intentions. These studies are not included in this discussion because they focused on intentions rather than outcomes.

¹² Stevenson and Shughart (2006) provide additional evidence that might suggest that tobacco control spending might not vary systematically across states. They find that monies awarded to states from the Master Settlement Agreement were not entirely based on the stated purposes of recovering health care costs attributable to smoking and funding of smoking-prevention programs.

¹³ Of course, an economically efficient allocation does not necessarily mean that states with high prevalence should spend more on control if, for instance, programmes are not equally effective or when states have heterogeneous policy objectives. Differences in demographics across states might also lead to spending differences across states under a ‘rational needs’ approach. For example, Yen (2005) finds that smoking falls with education, but older smokers consume more cigarettes than younger smokers, in the USA. Goel and Nelson (2005) find significant differences in tobacco consumption across age and gender as well in the USA.

¹⁴ See, for example, Charmistha and Grabowski (2003).

¹⁵ Centers for Disease Control and Prevention (CDC). ‘Behavioral Risk Factor Surveillance System Survey Data’, Atlanta, Georgia: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, (2006).

¹⁶ Source: Regional Economic Information System, Bureau of Economic Analysis, U.S. Department of Commerce http://www.bea.gov/bea/regional/spi/spi_download.csv?table_id=544846.

¹⁷ Source: http://www.census.gov/population/socdemo/education/cps2005/tab13.xls

¹⁸ Source: http://www.census.gov/popest/states/asrh/SC-EST2005-02.html

¹⁹2004 National Surveys of Religion and Politics, 2001 American Religious Identification Survey, 2000 Glenmary Religious Congregations and Membership in the U.S., United States Census, and 2002 Gallup Polls; see http://www.beliefnet.com/politics/religiousaffiliation.html

²⁰ See www.tobaccofreekids.org, ‘History of State Spending for Tobacco Prevention’, 7 December 2005.

²¹ Farrelly et al (2003) ran separate regressions on these four states along with their examinations of all states together. However, regressions were not run on the 46 states without long histories of active funding.

²² http://lungaction.org/reports/key304.html

²³ Percent of voting-age population voting for Presidential candidates was considered as an alternative instrument. It was hypothesized that higher percents would indicate states with more citizens comfortable with attempting to lower smoking through tobacco control programmes, but, at the same time, there is no reason to expect that this percent would directly cause smoking prevalence. Percents for elections in 2000 were used for years 2001 and 2002, while percents in the 2004 election were used in each of remaining years and were obtained from the U.S. Statistical Abstract (2007). While the coefficient on the percent voting in Presidential elections was positive, it was significant at the 5% level and therefore believed to be inferior to tobacco settlement funds as an instrument which was significant at the 1% level.

²⁴ See GAO (2004).

²⁵ See, for example, Gross et al. (2002), Johnson (2004) and Sloan (2005).

²⁶ First-stage estimates of coefficients (t-statistics) are:–25.36 (3.22) + 4.31 (3.02) PRICE–0.0002 (3.34) INCOME + 0.06 (1.23) COLLEGE + 0.17 (1.64) AGE–0.32 (1.92) UNEMPLOYMENT + 0.02 (0.96) MORMON + 0.08 (4.41) TS–0.42 (0.81) DUM2002 + 0.54 (1.00) DUM2003 + 0.67 (1.26) DUM2004 + 0.66 (1.25) DUM2005.

²⁷ See Stock and Watson (2003).

²⁸ As discussed previously, percent of the voting-age population voting in Presidential elections was also considered as an alternative instrument. Estimations using this instrument yielded nearly identical coefficients of the same significance on this interaction variable between tobacco control spending and the four states with long-standing programmes.

²⁹ For example, Escario and Molina (2004) find that, based on a sample of 12 EU countries over 1983 to 1993, a 10% surtax on tobacco would reduce lung cancer mortality by 1.2% in the first year.

³⁰ Goel (2007) examines state-level US data of quitting behaviour by smokers and shows that, while workplace smoking restrictions do not affect behaviour, quite decisions are significantly influenced by cigarette prices and home smoking restrictions.