Impuestos específicos al alcohol y mortalidad en carretera: evidencia empírica de la reforma de 2011 en Connecticut

This contribution summarizes the communication awarded with the Award for the best Oral Communication at the XLII Conference, held in Gerona, from July 5 to 7, 2023.

Driving under the influence carries a significant risk, accounting for around 30% of road deaths (WHO, 2022). Alcohol taxes, mainly specific ones, are one of the most cost-effective policies according to some institutions (WHO, 2018). Increasing taxes on alcohol would cause a rise in prices (Nelson and McNall, 2016; Ardalan and Kessing, 2019), which would reduce alcohol consumption (Sharma et al., 2017) and would cause a decrease in driving under the influence of alcohol.

According to the review of Elder et al. (2010), increasing taxes is an effective measure to reduce road mortality (among other health indicators). However, a notable proportion of articles suffered from identification problems. A more recent review, which only focused on quasi-experimental studies, presented less favorable results, with the majority being non-significant (Nelson and McNall, 2016).

There is a position of some decisive politicians (Rehm et al., 2019), supported by the literature, of the greater risk associated with the consumption of beverages with higher alcohol content (Andreuccetti, 2014; January, 2021). This has meant that most of the reforms included in the Nelson and McNall review are not uniform: taxes on drinks with high alcohol content are increased by a relatively larger proportion. Consequently, consumers can avoid the tax increase partially or completely by shifting consumption to certain types of alcoholic beverages. An example is the Illinois reform in 2009, which had no effect on highway mortality (McClelland and Iselin, 2019). Beer consumption increased significantly because the corresponding tax was increased by a trivial amount (Gehrsitz et al., 2021).

In this study we estimate the effect on highway mortality of a 20% increase in specific taxes on alcoholic beverages (based on ethanol content) that occurred in 2011 in Connecticut. The reform is uniform, ruling out the aforementioned possibility of creating incentives to avoid the increase. We implement two robustness tests related to the dependent variable and tax avoidance. We also address possible heterogeneous effects depending on age, gender, and time of accident (Chaloupka, et al. 2002; McCarthy, 2003; Ahlner, 2014; Valen, 2019).

Data and methodology

We use the US Fatality Analysis Reporting System (FARS) database from 2000 to 2019, which contains all fatal traffic accidents including number of deaths, time and place of accident, driver’s blood alcohol level, personal characteristics, among others. We also use administrative data by state to incorporate control variables on number of drivers, age distribution, fuel tax, seat belt use, road maintenance, unemployment, and annual income. We used the proportion of alcohol-related accidents to the total as the dependent variable. This allows us to control some unobservable variables on transportation patterns and driver decisions, which will present similar trends in both types of accidents.

We analyze the effect of the alcohol tax increase using a synthetic differences-in-differences (SDID) model. This model constructs a synthetic state from the states that did not have alcohol tax reforms during the period studied. The procedure is similar to the synthetic control (Abadie, 2010), but incorporates some different features (Arkhangelsky, 2021).

Results

In Table 1 we observe that the result is not significant. We have estimated other models with different specifications and analyzing subgroups, although the tables are not reported due to space limitation. We estimate the model using the rate of deaths per 1,000,000 drivers because of the possibility that the proportion of alcohol-related deaths is introducing biases that may bias the estimate of the causal effect. The effect is also not significant. One factor that may be diluting the effect is the avoidance of the tax increase, taking the form of purchases in neighboring states (Hanson and Sullivan, 2015). We estimate the model including only accidents that take place in non-bordering counties. The effect of the reform is also not significant. The subgroup analysis also does not yield significant results according to gender, age group or time of accident.

Table 1. Effects of the reform on the proportion of road deaths related to alcohol

Note: Synthetic differences-in-differences model with the covariates ‘state’ and ‘fixed effects’. In parentheses, standard errors. Degree of significance: *** 1%, **5%, *10%.

The increase in the tax corresponding to this reform was transferred to the final prices (Conlón and Rao, 2020) and, according to the literature, it should have caused a drop in consumption. However, we did not see a reduction in alcohol-related road deaths. Due to the characteristics of the reform, we can exclude the possibility of tax avoidance, present in the non-uniform increases in previous articles. Consequently, this article provides greater evidence regarding the possible cause of the null effect: the low price elasticity of the demand for alcoholic beverages of the population groups most prone to driving under the influence of alcohol (heavy alcohol consumers, young people, night driving ). , etc.), at least during the period prior to driving.