In the United States, tobacco use is estimated to cause approximately a half-million preventable deaths (Office of the Surgeon General, 2012) and $170 billion in tobacco-related healthcare expenditures (Xu et al., 2015) each year.1 Various heart and respiratory diseases and several forms of cancer are among some of the well-known consequences of combustible tobacco use. According to the 2012 Surgeon General's Report, roughly 96 percent of smokers began smoking before age 21 (U.S. Department of Health and Human Services, 2012). There is also evidence that the age of initiation has long-term effects on the intensity of addiction and capacity to quit, as many smokers that began smoking while young become addicted between the ages of 18 and 21 (Taioli and Wynder, 1991; Kwan et al., 2015).

Many public policies have been implemented to reduce smoking, such as imposing higher taxes, issuing statutory warnings on labels and packaging, and raising the minimum legal sales age (MLSA) for tobacco products. In 1992, the MLSA for tobacco products was set to 18 at the federal level. By 2005, only four states (Alabama, Alaska, New Jersey, and Utah) had raised their MLSA to 19. More recently, a significant push has been made to raise the MLSA to 21. As of mid-2019 (the end of our study period), six states and 482 counties or cities had passed Tobacco 21 (T21) laws (American Lung Association, 2019). T21 was signed into law nationally at the end of 2019, though as of November 2022 the FDA has not yet promulgated final rules and so it is unclear the extent to which the law is being enforced.

Our paper contributes by expanding the literature on the effect of T21 in several key dimensions. First, we unpack mechanisms for how T21 laws operate. In particular, T21 laws should increase ID checking when purchasing tobacco for individuals newly covered by the ID requirement, which we will test directly. More speculatively, T21 laws could also raise perceptions of the dangers of tobacco product use, thus reducing use and sales. Additionally, T21 laws increase MLSAs for both cigarettes and e-cigarettes to 21 rather than a combustible-21 only strategy that allows e-cigarettes to continue to be sold at the previous MLSA. Therefore T21 could indirectly increase e-cigarette risk perceptions by more than cigarette risk perceptions if youth previously believed e-cigarettes to be safer products and are now more likely to believe them to be equally as harmful. Higher e-cigarette risk perceptions could offset public health gains that e-cigarettes could otherwise achieve (Viscusi, 2016; Viscusi, 2020).

A second major contribution of our paper is to use cigarette and e-cigarette sales data from the Nielsen Retail Scanner database, in addition to survey data on use. Self-reported tobacco product use information is vulnerable to systematic bias from T21 laws if newly covered individuals become less likely to admit to using a newly restricted product. Sales data thus provide an objective data source to explore key relationships observed from self-reported use information. In addition, we complement recent work on T21 by Hansen et al. (2023) by using a different survey data source – Monitoring the Future (MTF) – to study the effects of T-21 laws on cigarette and e-cigarette use among those newly bound by T-21 laws (18-year-old 12th graders) as well as younger 12th graders, 8th graders, and 10th graders.

In sum, we make unique contributions in terms of studying how T21 laws work.

Our main finding is that T21 reduces cigarette use for all grades, with especially large effects for 12th graders. We also find causal evidence that T21 reduces e-cigarette use for 12th graders. T21 increases perceptions of the risk of both tobacco products, but more for e-cigarettes. We find complementary evidence using sales data that T21 reduces tobacco product sales.

The rest of the paper is organized as follows. In the next section, we provide a brief history of T21 legislation. We also critically review the current literature on T21 and tobacco MLSA laws in general. Section 3 describes the data and empirical specifications. Section 4 reports the main results and those from sensitivity analyses. The last section concludes with a discussion of the policy implications of our findings.