n?u=RePEc:tow:wpaper:2016-17&r=tre

repec.nep.tre

n?u=RePEc:tow:wpaper:2016-17&r=tre

Towson University

Department of Economics

Working Paper Series

Working Paper No. 2016-17

The Effects of False Identification

Laws with a Scanner

Provision on Underage Alcohol-

Related Traffic Fatalities

by Erik Nesson and Vinish Shrestha

October 2016

© 2016 by Author. All rights reserved. Short sections of text, not to exceed two

paragraphs, may be quoted without explicit permission provided that full credit, including

© notice, is given to the source.


The Effects of False Identification Laws with a Scanner

Provision on Underage Alcohol-Related Traffic Fatalities

Erik Nesson ∗

Department of Economics

Ball State University

Vinish Shrestha †

Department of Economics

Towson University

Abstract

We examine the effects of false identification laws with a scanner provision (FSP laws)

on alcohol-related fatal accidents involving underage drivers using data on traffic fatalities

from the National Highway Traffic Administration from 1998 to 2014 and information

on alcohol control policies from the Alcohol Policy Information System. We find

that the implementation of FSP laws reduced alcohol-related traffic fatalities among

16-18 year olds without a statistically significant change in non alcohol-related fatalities

among 16-18 year olds or in alcohol-related fatalities among 21-24 year olds. Our

results are stable across a number of different specifications. A back-of-the-envelope

calculation suggests that if all remaining states passed FSP laws, the reduction in underage

alcohol-related traffic fatalities among 16-18 year olds would generate nearly

$250 million in annual economic benefits.

Keywords: Underage alcohol consumption; Drunk driving; DWI; False ID laws; Scanner

provision

JEL Codes: I12, I18

∗ Department of Economics, Miller College of Business, Ball State University, Muncie, IN (E-Mail: etnesson@bsu.edu)

† Corresponding author. Department of Economics, Towson University, Baltimore, MD (E-Mail:

vshrestha@towson.edu)

1


1 Introduction

Motor vehicle accidents are the leading cause of death among 16 to 20 year olds (National

Highway Traffic Safety Administration, 2008), and one out of three crashes among 16 to 20

year olds are attributed to alcohol. Although all 50 states and the District of Columbia

follow the federal minimum legal drinking age (MLDA) and zero tolerance policies (blood

alcohol content (BAC) limit of 0.02 or lower), a significant portion of youths still consume

alcohol and participate in drinking and driving (Substance Abuse and Mental Health Services

Administration, 2014). Many studies have examined a wide variety of demand-side policies

aimed at reducing alcohol use and alcohol-related outcomes, including increases in the MLDA

(Carpenter and Dobkin, 2009; McCartt et al., 2010; Yörük and Yörük, 2011, 2013; Voas et al.,

2003), zero-tolerance policies (Carpenter, 2004; Zwerling and Jones, 1999), and alcohol taxes

(Ruhm, 1996; Chaloupka et al., 2002; Markowitz and Grossman, 2000; Markowitz, 2000,

2005; Markowitz et al., 2012; Cook and Durrance, 2013).

However, there has been little recent variation in most of these demand-side policies.

For example, the MLDA of 21 was implemented in 1988, since 1991 the nominal federal

excise alcohol tax has remained constant, only a handful of states have increased alcohol

taxes in recent decades, and by the late 1990s all 50 states and the District of Columbia had

implemented zero tolerance policies. More recent alcohol-control policy variation has been in

supply-side policies, especially policies targeted at the use of false identification (false ID) to

purchase alcohol. Relatively few studies examine these growing number of false ID policies

and their effects on youth alcohol consumption. Two notable exceptions are Yörük (2014),

who studies laws providing incentives for retailers to adopt scanners to detect false IDs (FSP

laws) and Bellou and Bhatt (2013), who study the implementation of vertical ID laws. Both

of these studies find that these policies reduce underage alcohol consumption. However, no

studies have examined these policies’ effects on alcohol-related outcomes, notably alcoholrelated

traffic accidents.

In this paper, we aim to fill this gap in the literature. We use within state variation in

2


FSP laws in a differences-in-differences model to identify whether FSP laws reduce alcoholrelated

driving fatalities among underage drinkers. We use data from the National Highway

Traffic Administration (NHTSA) from 1998 to 2014 to measure the number of alcoholrelated

traffic fatalities among 16-18 year olds, 16-20 year olds, and 21-24 year olds and

information on alcohol control policies from the Alcohol Policy Information System (APIS).

We also examine other state level alcohol laws including distinctive licenses, other provisions

to support alcohol retailers in not selling alcohol to minors, BAC 0.08 limits, Sunday sales

bans, and beer taxes.

In our main results, we find that the implementation of FSP laws reduced alcohol-related

traffic fatalities among 16-18 year olds by between 10 and 16 percent, with less evidence

that FSP laws affected alcohol-related traffic fatalities among 18-20 year olds. Our results

are stable across a number of different specifications, and we also conduct two falsification

tests, examining alcohol-related traffic fatalities among 21-24 year olds and non alcoholrelated

traffic fatalities among 16-18 year olds. We do not find statistically significant FSP

law coefficients for either of these populations, and in many specifications we can bound

the 95 percent confidence intervals relatively close to zero. We implement an event study

specification to examine the dynamic effects of the policies.

In our event study, we find

some evidence of a decrease in alcohol-related traffic fatalities in the quarter preceding the

implementation of FSP laws, which may be explained by retailer adopting ID scanners prior

to implementation.

Our study makes contributions to different areas of the economics literature examining

alcohol use and traffic fatalities. Most directly, our study builds on previous work which

suggests that policies which aid retailers in identifying false ID reduce underage alcohol consumption

(Yörük, 2014; Bellou and Bhatt, 2013). We extend this research to examine one

of, if not the largest, negative consequence of underage alcohol consumption, alcohol-related

traffic fatalities. While we focus on FSP laws, we also examine other retailer support provisions,

notably distinctive licenses and affirmative defense laws, which aid retailers in legal

3


issues related to selling alcohol to a minor. Although we find negative and statistically significant

relationships between these two policies and underage alcohol-related traffic fatalities,

there is limited variation in these laws within states during our period of analysis and thus

we interpret these results with caution.

More broadly, we contribute to a number of studies that have examined the relationship

between alcohol control policies, alcohol consumption, and negative consequences from alcohol

consumption, including MLDA laws (Cook and Tauchen, 1984; Carpenter and Dobkin,

2009; Dee, 1999; McCartt et al., 2010; Yörük and Yörük, 2011, 2013; Voas et al., 2003; Zhang

and Caine, 2011), zero-tolerance policies (Carpenter, 2004; Zwerling and Jones, 1999), alcohol

taxes (Chaloupka et al., 1993; Ruhm, 1996; Chaloupka et al., 2002; Cook and Moore,

1993; Cook, 1987; Cook and Durrance, 2013; Pacula, 1998; Markowitz and Grossman, 2000;

Markowitz, 2000, 2005; Markowitz et al., 2012; Mast et al., 1999; Son and Topyan, 2011),

Sunday alcohol sales (Lovenheim and Steefel, 2011; Stehr, 2010), and BAC laws (Dee, 2001;

Eisenberg, 2003; Liang and Huang, 2008; Grant, 2010). Finally, our study also contributes to

research examining the determinants of traffic fatalities and public policies which may intentionally

or unintentionally affect traffic fatalities. Previous research in this literature have

examined the relationship between traffic fatalities and smoking bans (Adams and Cotti,

2008), the minimum wage (Adams et al., 2012), casinos (Cotti and Walker, 2010), and cell

phones (Abouk and Adams, 2013; Cheng, 2015).

The rest of this paper is outlined as follows. Section 2 covers our data sources, Section 3

outlines our identification strategy, Section 4 describes our results, and Section 5 concludes.

2 Data

2.1 Alcohol-Related Traffic Fatalities

This study uses data from the Fatality Analysis Reporting Systems (FARS) of the

NHTSA from 1998 to 2014.

The data reported in FARS is a nationwide census of fatal

4


motor vehicle crashes.

We follow the NHTSA procedure, used to generate their official

statistics, to calculate the number of alcohol-related traffic fatalities for each state and quarter

(NHTSA, 2012).

The NHTSA classifies accidents in three categories with respect to

alcohol involvement: (1) fatal accidents not involving alcohol; (2) accidents where a driver

had a BAC level less than 0.08 but greater than zero, and; (3) accidents where a driver had

a BAC level of at least 0.08. Although federal mandates require the measurement of BAC

levels in all traffic accidents resulting in a fatality, BAC levels are unreported for roughly

half of fatal accidents. The missing BAC levels in the FARS dataset are imputed by using a

“general location model,” which models the probability of having a positive BAC level (see

Rubin et al. (1998) for more details). The imputed value of the BAC level depends on factors

like “age, gender, safety belt or helmet use, license expiration, prior traffic convictions, day

of the week, time of the day, the role of the vehicle in the accident, whether the car remains

on the road, the type of vehicle driven, and whether police at the accident believed drinking

was involved” (Adams et al., 2012) 1 . To minimize the impact of any measurement error

resulting from the imputation, we focus on accidents where a driver had a BAC level of at

least 0.08, as a lower percentage of these accidents have imputed BAC levels. Our results are

robust to using accidents with BAC level greater than 0, which we discuss more in Section

4.2.

For each state, year, and quarter cell we count the number of accidents involving a driver

with a BAC level that exceeds 0.08 in three age categories: (1) ages 16-18; (2) ages 16-

20; and (3) ages 21-24. The first age category, ages 16-18, will measure the effects among

underage drinkers most responsive to FSP laws (Yörük, 2014). The second age category,

ages 16-20, will measure the response among all underage drivers, and the third age category

serves both as a falsification test (to see if the rate decreases in response to a policy only

targeting underage drinkers) and a test of whether a decrease in underage alcohol-related

traffic fatalities results in an increase in legal drinking age fatalities a few years later, as

1 The measure of imputed BAC level has been used by adams2012; and abouk2013

5


more at-risk drivers reach legal drinking age.

2.2 Alcohol Control Policies

We collect state-level information regarding FSP laws and alcohol-control policies from

the Alcohol Policy Information System (APIS). We record the quarter in which FSP laws

went into effect, and we additionally collect the quarters in which four additional provisions

supporting alcohol retailers went into effect: 1) Retailers’ power to seize identification without

fear of prosecution (even if the identification is valid); 2) Affirmative defense (reducing

retailers’ liability if they provide alcohol to minors); 3) Providing retailers with the right to

sue a minor; and 4) Providing retailers with the right to detain a minor. Finally, we also

collect the quarter in which states began mandating distinctive underage licenses.

Table 1 shows the dates that these laws when into effect. Currently, 11 states have a

FSP law, and all of the states enacted such provisions between 1998 and 2014, the years

corresponding to this study. This allows for sufficient within state variation in FSP laws

to identify the effect of such law on alcohol-related traffic fatalities. Other retailer laws are

prevalent in only a handful of states. Only five states have a law that provides retailers

with the right to sue minors and only three states give retailers the right to detain minors.

Although many states allow a retailers’ provision of affirmative defense, there is little

within-state variation during the time frame of our study. Similarly, although 26 states have

implemented a distinctive license provision, there exists little within-state variation in these

laws. Due to a lack of within-state variation, we interpret the effects of these other alcohol

control policies with caution.

2.3 State Level Variables

We account for other state-specific alcohol control policies which may be correlated with

the introduction of FSP laws. First, we include the combined real federal and state beer tax

from the U.S. Brewers Association (2014) and APIS. We also include indicators for whether

6


states had a ban on Sunday alcohol sales (including a local option) and a BAC limit of 0.08

for drunk driving from APIS. 2

Following the findings of Ruhm (1996), we control for annual state unemployment rates

and the natural log of real per-capita income from the Bureau of Labor Statistics and the

Bureau of Economic Analysis, respectively. We also include the quarterly state-level real

minimum wage, originally compiled by Allegretto et al. (2016), following the results from

Adams et al. (2012). To account for costs associated with driving we include the real statelevel

annual gasoline taxes (per gallon) from the Office of Highway Policy Information. We

account for implementation of state-level primary seatbelt laws using data from the Centers

for Disease Control and Prevention (CDC) and we control for state policies on texting while

driving using data from Abouk and Adams (2013). We convert all nominal dollar values to

real 2014 dollars using the consumer price index.

Because our dependent variable is a count, we control for the natural log of the age-specific

population collected from the National Cancer Institute’s SEER population database (National

Cancer Institute, 2016) and organized by the National Bureau of Economic Research.

3 Identification Strategy

We follow a basic difference-in-differences framework with multiple treatment groups

and time periods where identification is generated by within-state variation in FSP laws

over time. The regression model is written as:

D stq = exp(α 1 + α 2 F SP stq + A stq α 3 + X stq α 4

+ α 5 P op stq + σ s + γ t + η q + θ s t) + e stq

(1)

where D stq represents the count of alcohol-related traffic fatalities for state s in year t and

quarter q, F SP stq is an indicator variable taking the value “1” if a state has a FSP law in

2 Local option laws are enacted by local governments and are less restrictive than the state law.

7


a given year and quarter and “0” otherwise, A st is a vector of other alcohol control policies

discussed in Sections 2.2 and 2.3, X stq is a vector of other state characteristics discussed in

Section 2.3, P op stq is the state population of 16-18 year olds (or respective age group), σ s , γ t ,

and η q refer to state, year, and quarter fixed effects, respectively, θ s t represents state-specific

linear time trends, and e stq is an error term. We cluster our standard errors at the state

level.

Figure 1 presents a histogram of alcohol-related driving fatalities pertaining to 16-18 year

olds. The figure demonstrates that the density of alcohol-related driving fatalities is skewed

towards the right and consists of a preponderance of zeros. Given the discrete nature of

our dependent variable with zero values, we estimate Equation (1) by using a Fixed Effects

Poisson (FEP) estimator.

The FEP estimator is a quasi-maximum likelihood estimator

that belongs to the linear exponential family. An advantage of using a Poisson regression

model is that the FEP estimates are consistent even if the count does not follow a Poission

distribution. This model relies under a much weaker assumption that requires conditional

mean, exp(X ′ β), to be correctly specified Cameron and Trivedi (2009). A drawback of the

Possion regression model is equidispersion, the assumption that the conditional mean and

conditional variance are the same. To account for this aspect, we test the robustness of our

results to alternative estimation strategies, described in more detail in Section 4.2.

Our identification rests on the assumption that absent a FSP law, the trend in alcoholrelated

traffic accidents in states with FSP laws would be similar to states without FSP

laws. A potential theat to the indentification strategy is that the decision to implement FSP

law can be endogenous. Specifically, it is plausible that states experiencing a high level of

underage drinking and alcohol-related fatalities may decide to adopt FSP law compared to

states with lower level of underage drinking and alcohol-related traffic accidents. To account

for other potential time varying factors that might be correlated with both the FSP laws and

alcohol-related traffic accidents, we control for other alcohol control policies and state-level

variables.

We additionally include state-specific linear time trends in alternative specifi-

8


cations, which should alleviate at least some concerns of not accounting for pre-treatment

trends in alcohol related traffic fatalities. As another test of our identifying assumption, we

estimate a dynamic event-study model to investigate the effect of the FSP law in the time

periods prior to and after the enactment of a FSP law. Instead of the FSP law variable

taking a value “1” after the adaption of the FSP law as shown in Equation (1), the model

specification includes binary indicators for the time periods prior to and after the FSP law.

More specifically, we group the time periods before and after the law into half-year bins,

beginning three years before the law implementation and running until three or more years

after the law implementation, using time periods more than three years before implementation

(or states which never implemented the law) as our omitted category. We also parse out

the half-year immediately preceding implementation to more carefully analyze anticipatory

effects.

The implementation of FSP laws generate three different variations: 1) Over time, 2)

Across states; and 3) Age (under age versus older individuals). We therefore estimate the

effect of FSP laws on traffic fatalities by using a difference-in-differences-in-differences (DDD)

model.

Unlike a DID model, a DDD model uses two control groups: underage alcoholrealted

fatal accidents in states without the law and alcohol-related fatal accidents pertaining

to older individuals (21-24 year olds) in states with and without FSP laws. In the DDD

specification, the only threat to identification would be factors correlated with FSP laws that

are differentially affecting alcohol-related fatal accidents of underage and those pertaining

to drivers of legal drinking age. The DDD specification is defined as follows:

D stqa = exp(β 1 + β 2 F SP stq + β 3 Age a + β 4 F SP stq × Age a + A stq β 5 + X stq β 6

+ β 7 P op astq + σ s + γ t + η q + σ s Age a + γ t Age a + η q Age a + θ s Age a t) + e astq .

(2)

Here, D stqa represents the count of alcohol-related traffic fatalities for age group a in state

s, year t and quarter q, Age a is an indicator variable for an age group that is underage, and

F SP stq Age a is an interaction term between the indicator for whether a state has passed a

9


FSP law and the indicator for an underage group. The model also includes state by age group

fixed effects (to control for pattern of alcohol-related accidents which are different between

underage and drivers of legal drinking age across states irrespective of time), year by age

group fixed effects (accounts for differences in trend in alcohol-related fatalities between

underage and older individuals overtime irrespective of state), quarter by age group fixed

effects, and state-specific linear time trend by age group effects (accounts for linear time

varying unobserved factors which may be correlated with both FSP laws and alcohol-related

fatalties that may affect underage and older drivers differently). The coefficient of interest

is β 4 , showing the differential effect of a FSP law on alcohol-related traffic fatalities among

underage drivers compared to drivers 21-24 years old and under-age drivers in states without

the FSP law.

4 Results

4.1 Main Results

Table 2 provides descriptive statistics of the variables used in this study. The average

number of alcohol-related fatalities (per quarter) for 16-18 year olds is 2.25. The averages

pertaining to 16-20 and 21-24 year olds are 5.57 and 8.21, respectively. Approximately 23

percent of states had a FSP law over the time span of this study, and the average beer tax

per gallon is $0.31, which includes both state and federal taxes.

Table 3 presents the results estimating Equation (1) for 16-18, 16-20, and 21-24 year olds.

We present the findings from three different specifications for each age group. Specification

(1) includes non alcohol-related control variables (real gasoline tax, real minimum wage,

unemployment, log of per capita income, and log of respective age specific population), and

state, year and quarter fixed effects. Specification (2) adds variables pertaining to alcoholcontrol

policies (listed in 2), and specification (3) adds state-specific linear time trends. The

coefficients from Poission regressions can roughly be interpreted as semi-elasticities, and we

10


additionally report the semi-elasticity for the FSP laws in brackets as given by the formula

exp(β) − 1. Our results suggest that FSP laws reduced alcohol-related fatal traffic accidents

by between 10 and 16 percent for 16-18 year olds. These coefficients from specifications (1)

and (2) are statistically significant at a 1 percent level and the coefficient from specification

(3) is significant at a 5 percent level. We see less evidence that FSP laws lead to statistically

significant reductions in alcohol-related traffic accidents for all underage people, although

the point estimates are not small. The findings suggest that the reduction in alcohol-related

fatalities associated with FSP laws is mainly concentrated among 16 to 18 year olds. Such

a finding is consistent with Yörük (2014)’s results, which indicate that FSP laws decreased

alcohol consumption, including binge drinking, among younger teens (16-17 year olds).

In the last set of results, FSP laws generally lead to very small and statistically insignificant

changes in alcohol-related traffic accidents among 21-24 year olds. This provides

suggestive evidence that the reduction in alcohol-related traffic fatalities among 16-18 year

olds is due to FSP laws and not due to other unobservable state characteristics which affect

overall alcohol-related traffic fatalities.

This also suggests that the reduction in alcoholrelated

traffic fatalities among 16-18 year olds does not translate into an increase in fatalities

as more at-risk minors reach legal drinking age.

We find statistically significant relationships between other retailer-focused alcohol control

policies and alcohol-related traffic fatalities. In particular, distinctive licenses, ID seizure

laws, and affirmative defense laws are all consistently related to reductions in alcohol-related

traffic fatalities among 16-18 year olds. However, as mentioned before, these estimates should

be interpreted with caution due to limited within-state variation in such laws. We report the

effects of the other control variables in Appendix Table A1. We find evidence that seat belt

laws reduce alcohol-related traffic fatalities and that gas taxes have a small positive effect

on alcohol-related traffic fatalities among 16-18 year olds, but the expected negative effect

for 21-24 year olds. The unemployment rate is negatively related to alcohol-related traffic

fatalities among 16-20 year olds and 21-24 year olds.

11


Although the effects of FSP laws are concentrated among underage drivers, as expected,

it is possible that the results shown in Table 3 are driven by a reduction in fatal driving

accidents in general among states implementing the FSP laws. If this is the case, even in

the absence of a FSP law, states that implemented a FSP law would have experienced a

reduction in alcohol-related fatal accidents. To test this possibility, we present falsification

tests where we replace the dependent variables in Table 3 with non alcohol-related traffic

fatalities. Table 4 shows results from these regressions, and has the same structure as Table

3. In all but one specification, we see very small coefficients which are not statistically

significant. In many cases, we can bound the effects to be relatively small compared to the

effects in Table 3. For example, in specification (3) for 16-18 year olds, the lower bound of

the 95 percent confidence interval is roughly -0.085.

4.2 Robustness Checks

We next employ a set of robustness checks to further examine the stability of our results

of the effects of FSP laws on alcohol-related traffic accidents among 16-18 year olds. These

results are displayed in Table 5. In the first column, we examine the effect of controlling for

non alcohol-related traffic fatalities by including the count of age-specific non alcohol-related

fatalities as an independent variable, following Adams et al. (2012). Adams et al. (2012)

argue that inclusion of non alcohol-related fatalities help explain within state variation in

alcohol-related fatal accidents which can be attributed to factors such as changes in lighting,

speed limits, weather, and other unobserved state-specific changes which may affect driving

accidents in general. 3

In the second column, we use as our dependent variable all alcoholrelated

traffic fatalities among 16-18 year olds, not just those where the driver had a BAC

level greater than 0.08.

In the third and fourth columns, we estimate the OLS models

using the log rate of alcohol-related traffic accidents the dependent variable. Column (3)

3 We also ran specifications where we include the number of alcohol-related fatalities among 21-24 year

olds as an independent variable. Here again, the coefficient for the effect of FSP laws on alcohol-related

fatalities among 16-18 year olds is stable. These results are available upon request.

12


includes the state specific population of 16-18 year olds as weights and Column (4) does not.

Columns (5) and (6) follow specifications in Adams et al. (2012) where the dependent variable

is the inverse normal function of the rate of alcohol-related traffic accidents. Column (5)

includes population weights and Column (6) excludes them. In all our columns, the negative

statistically significant relationship between FSP laws and alcohol-related traffic fatalities

among 16-18 year olds is stable.

4.3 Dynamic Response

In this section, we present results from dynamic response models which estimate the

immediate versus the long run-impact of FSP laws. We show these results in figures which

display the coefficients and 95 percent confidence intervals for the effects of FSP laws on

traffic fatalities in the quarters leading up to the law’s implementation and the quarters

after implementation. The omitted time period is more than 12 quarters before the law (or

states that never passed a FSP law), and the other independent variables are those included

in specification (3) from Table 3. The results from these models are plotted in Figures 2, 3,

and 4. Figure 2 uses alcohol-related traffic fatalities among 16-18 year olds as the dependent

variable, Figure 3 uses non alcohol-related traffic fatalities among 16-18 year olds as the

dependent variable, and Figure 4 uses alcohol-related traffic fatalities among 21-24 year olds

as the dependent variable.

Examining Figure 3, there is not an obvious trend in non alcohol-related traffic fatalities

among 16-18 year olds before, during or after a FSP law’s implementation. In Figures 2

and 4, there is a decrease in alcohol-related traffic fatalities in the quarter preceding the

implementation of FSP laws. This decrease may result from retailers adopting ID scanner

technology before the FSP law or increased interest in drunk driving resulting from press

regarding the FSP laws. There appears to be a similar pattern for binge drinking in Yörük

(2014), where measures of binge drinking trend downwards in the years before a FSP law’s

implementation. However, alcohol-related traffic fatalities among 21-24 year olds rebound

13


up to their pre-FSP law levels shortly after the FSP law’s implementation while alcoholrelated

traffic fatalities among 16-18 year olds stay at a level roughly 40 percent below the

level three years before implementation. So while these results hint that other factors may

drive decreases in alcohol-related fatalities immediately before a FSP law’s implementation,

it appears that FSP laws lead to an immediate and long-lasting decrease in alcohol-related

traffic fatalities among 16-18 year olds. Additionally, the dynamic response models indicate

that our results in Table 3 may understate the effects of FSP laws among 16-18 year olds. 4

4.4 DDD Model

Finally, we present results from the DDD models estimating the effect of FSP laws on

alcohol-related traffic fatalities among underage drivers in states with FSP laws compared

to drivers of ages 21-24 and underage drivers in states without FSP laws. As our dynamic

event study models find some evidence of a temporary decrease in alcohol-related traffic

fatalities among 21-24 year olds, these specifications will determine whether alcohol-related

traffic fatalities fell among underage drivers in states with FSP laws relative to drivers of

legal drinking age and underage drivers in states without FSP laws. Similar to models in

Table 3, we examine two groups of underage drivers: ages 16-18 and ages 16-20. Here again,

we find that the effects are concentrated among 16-18 year old drivers, with FSP laws leading

to about a 10 percent decrease in alcohol-related traffic fatalities, statistically significant at

the five percent level in our first two specifications. When we include age-group by state

linear time trends, our point estimate of the effect of scanner laws on alcohol-related traffic

fatalities among 16-18 year olds does not change, but the standard error increases by a factor

of two, and the coefficient loses its statistical significance. 5

The DDD specification alleviates concerns that other unobserved factors which are corre-

4 The results from model specifications without the state-specific linear time trends are presented in the

Appendix section of the paper in Figure A1.

5 This may be due to an increase in multicollinearity among regressors after adding state-specific linear

time trends. The difference in semi-elasticities from the interaction effect, reported in brackets, is given by

exp(β 4 + β 2 ) − exp(β 2 ).

14


lated with the implementation of FSP laws are driving the results. For example, consider a

scenario of states implementing FSP laws because of campaigns against drunk driving. For

the results from the DDD model to be biased upwards, such campaign would have to affect

underage drivers in states with FSP laws more, in comparison to drivers of legal drinking age

(21-24 year olds) in those states. This is unlikely to happen given that 21-24 year olds suffer

from the highest proportion of alcohol-related accidents. 6

If anything, such a campaign is

likely to be targetted towards 21-24 year olds, which will create a downward bias in results

from the DDD models.

5 Conclusion

In this paper, we examine the effects of FSP laws, which promote technology enabling

alcohol retailers to more easily identify false ID and not sell alcohol to minors, on alcoholrelated

traffic fatalities. Using within-state variation in the quarter in which states adopted

FSP laws, we find that FSP laws led to about a 10 to 16 percent reduction in alcohol-related

traffic fatalities among 16-18 year olds.

Falsification tests suggest that FSP laws do not

affect non alcohol-related traffic fatalities or alcohol-related traffic fatalities among those of

legal drinking age. Additionally, we subject our results to other robustness tests and find

these effects to be stable.

Finally, results from dynamic event studies suggests that the

decrease in alcohol-related traffic fatalities among 16-18 year olds from FSP laws begins in

the quarters immediately preceding the law’s implementation but is sustained long after

the law’s implementation. We see no such sustained decreases in either non alcohol-related

traffic fatalities among 16-18 year olds or in alcohol-related traffic fatalities among 21-24

year olds.This is a modest but plausible effect, as FSP laws have been shown to reduce the

probability of binge drinking by 4.4 percentage points among teens (Yörük, 2014)

To put our findings in context, Carpenter and Dobkin (2009) calculate that lowering the

6 NHTSA (2013) reports that the proportion of fatal crashes was highest for 21-24 year olds with 33

percent of the total alcohol-related fatal accidents. https://crashstats.nhtsa.dot.gov/Api/Public/

ViewPublication/812102 (accessed October, 28 2016)

15


MLDA will lead to an additional 0.77 alcohol related fatal deaths (for every 100,000 18-20

year olds), and Eisenberg (2003) suggests that a reduction in Blood Alcohol Content (BAC)

limit from 0.10 percent to 0.08 percent and graduated licensing (for drivers under 21) reduces

fatal crash rates by 2.6 and 11.5 percent, respectively. The effect of provision of graduated

licensing on alcohol-related fatal accidents translates to a reduction in average annual fatal

crashes by 0.42 fatal crashes (per 100,000 18-20 year olds). 7 This suggests that our estimate

of reduction in alcohol-related crashes by 0.58 accidents per year are similar to the effects of

the MLDA, graduated licensing, and BAC limits on alcohol-related crashes among underage

drivers.

To present our results another way, the 40 states which have not passed FSP laws had a

combined population of about 8.5 million 16-18 year olds in 2014 and a total of 172 alcoholrelated

traffic fatalities. If these 40 states passed a FSP law, we would expect about 27 fewer

alcohol-related traffic fatalities per year. Adopting the U.S. Department of Transportation’s

economic value of a statistical life of $9.2 million in 2014, this reduction would translate to

$248 million dollars in annual economic benefits, which translates to about $29 for every

16-18 year old. 8

Our work calls for future research into the effects of FSP laws and other supply-side

alcohol-control policies on other alcohol-related outcomes such as crime, education attainment,

child bearing and birth outcomes, and sexually transmitted infections.

7 This calculation uses average state population pertaining to 16 to 20 year olds of 471,274.7. Then,

0.42 = 0.115 ∗ 17.2 ∗ (100, 000/471, 274).

8 Please see this document for the U.S. Department of Transportation’s economic value of a statistical

life: https://www.transportation.gov/sites/dot.gov/files/docs/VSL2015_0.pdf (accessed October,

10 2016)

16


References

Abouk, R. and S. Adams (2013). Texting bans and fatal accidents on roadways: Do they

work? or do drivers just react to announcements of bans? American Economic Journal:

Applied Economics 5 (2), 179–199.

Adams, S., M. L. Blackburn, and C. D. Cotti (2012). Minimum wages and alcohol-related

traffic fatalities among teens. Review of Economics and Statistics 94 (3), 828–840.

Adams, S. and C. Cotti (2008). Drunk driving after the passage of smoking bans in bars.

Journal of Public Economics 92 (5), 1288–1305.

Allegretto, S., A. Dube, M. Reich, and B. Zipperer (2016). Credible research designs for

minimum wage studies: A response to neumark, salas, and wascher. Industrial and Labor

Relations Review Forthcoming.

Bellou, A. and R. Bhatt (2013). Reducing underage alcohol and tobacco use: Evidence

from the introduction of vertical identification cards. Journal of Health Economics 32 (2),

353–366.

Cameron, A. C. and P. K. Trivedi (2009). Microeconometrics using Stata, Volume 5. Stata

press College Station, TX.

Carpenter, C. (2004). How do zero tolerance drunk driving laws work? Journal of Health

Economics 23 (1), 61–83.

Carpenter, C. and C. Dobkin (2009). The effect of alcohol consumption on mortality: regression

discontinuity evidence from the minimum drinking age. American Economic Journal:

Applied Economics 1 (1), 164–182.

Chaloupka, F., H. Saffer, and M. Grossman (1993). Alcohol-control policies and motor

vehicle fatalities. Journal of Legal Studies 22 (1), 161–186.

Chaloupka, F. J., M. Grossman, and H. Saffer (2002). The effects of price on alcohol

consumption and alcohol-related problems. Alcohol Research and Health 26 (1), 22–34.

Cheng, C. (2015). Do cell phone bans change driver behavior? Economic Inquiry 53 (3),

1420–1436.

Cook, P. J. (1987). The impact of distilled-spirits taxes on consumption, auto fatalities, and

cirrhosis mortality. Advances in Substance Abuse: Behavioral and Biological Research,

Supplement 1 , 159–167.

Cook, P. J. and C. P. Durrance (2013). The virtuous tax: lifesaving and crime-prevention

effects of the 1991 federal alcohol-tax increase. Journal of Health Economics 32 (1), 261–

267.

Cook, P. J. and M. J. Moore (1993). Violence reduction through restrictions on alcohol

availability. Alcohol Research and Health 17 (2), 151.

17


Cook, P. J. and G. Tauchen (1984). Effect of minimum drinking age legislation on youthful

auto fatalities, 1970-1977, the. Journal of Legal Studies 13, 169.

Cotti, C. D. and D. M. Walker (2010). The impact of casinos on fatal alcohol-related traffic

accidents in the united states. Journal of Health Economics 29 (6), 788–796.

Dee, T. S. (1999). The complementarity of teen smoking and drinking. Journal of Health

Economics 18 (6), 769–793.

Dee, T. S. (2001). Does setting limits save lives? the case of 0.08 bac laws. Journal of Policy

Analysis and Management 20 (1), 111–128.

Eisenberg, D. (2003). Evaluating the effectiveness of policies related to drunk driving. Journal

of Policy Analysis and Management 22 (2), 249–274.

Grant, D. (2010). Dead on arrival: Zero tolerance laws don’t work. Economic Inquiry 48 (3),

756–770.

Liang, L. and J. Huang (2008). Go out or stay in? the effects of zero tolerance laws on alcohol

use and drinking and driving patterns among college students. Health Economics 17 (11),

1261–1275.

Lovenheim, M. F. and D. P. Steefel (2011). Do blue laws save lives? the effect of sunday

alcohol sales bans on fatal vehicle accidents. Journal of Policy Analysis and Management

30 (4), 798–820.

Markowitz, S. (2000). The price of alcohol, wife abuse, and husband abuse. Southern

Economic Journal, 279–303.

Markowitz, S. (2005). Alcohol, drugs and violent crime. International Review of Law and

Economics 25 (1), 20–44.

Markowitz, S. and M. Grossman (2000). The effects of beer taxes on physical child abuse.

Journal of Health Economics 19 (2), 271–282.

Markowitz, S., E. Nesson, E. Poe-Yamagata, C. Florence, P. Deb, T. Andrews, and S. B. L.

Barnett (2012). Estimating the relationship between alcohol policies and criminal violence

and victimization. German Economic Review 13 (4), 416–435.

Mast, B. D., B. L. Benson, and D. W. Rasmussen (1999). Beer taxation and alcohol-related

traffic fatalities. Southern Economic Journal, 214–249.

McCartt, A. T., L. A. Hellinga, and B. B. Kirley (2010). The effects of minimum legal

drinking age 21 laws on alcohol-related driving in the united states. Journal of Safety

Research 41 (2), 173–181.

National Cancer Institute (2016). Surveillance, Epidemiology, and End Results (SEER) Program

Populations (1969-2014). National Cancer Institute, DCCPS, Surveillance Research

Program, Surveillance Systems Branch.

18


National Highway Traffic Safety Administration (2008). Fatality analysis reporting system.

Retrieved February 23, 2009.

Pacula, R. L. (1998). Does increasing the beer tax reduce marijuana consumption? Journal

of Health Economics 17 (5), 557–585.

Rubin, D., R. Subramanian, and D. Utter (1998). Multiple imputation of missing blood

alcohol concentration (BAC) values in FARS. NHTSA Technical Report DOT HS 808

816.

Ruhm, C. J. (1996). Alcohol policies and highway vehicle fatalities. Journal of Health

Economics 15 (4), 435–454.

Son, C. H. and K. Topyan (2011). The effect of alcoholic beverage excise tax on alcoholattributable

injury mortalities. The European Journal of Health Economics 12 (2), 103–

113.

Stehr, M. F. (2010). The effect of sunday sales of alcohol on highway crash fatalities. BE

Journal of Economic Analysis & Policy 10 (1), 1–22.

Substance Abuse and Mental Health Services Administration (2014). Results from the 2012

national survey on drug use and health: Summary of national findings (no. nsduh series

h-46, hhs publication no.(sma) 13–4795). Technical report, Substance Abuse and Mental

Health Services Administration, Rockville, MD.

U.S. Brewers Association (2014). Brewers’ almanac. U.S. Brewers Association, Washington,

D.C..

Voas, R. B., A. S. Tippetts, and J. C. Fell (2003). Assessing the effectiveness of minimum

legal drinking age and zero tolerance laws in the united states. Accident Analysis &

Prevention 35 (4), 579–587.

Yörük, B. K. (2014). Can technology help to reduce underage drinking? evidence from the

false id laws with scanner provision. Journal of Health Economics 36, 33–46.

Yörük, B. K. and C. E. Yörük (2011). The impact of minimum legal drinking age laws on alcohol

consumption, smoking, and marijuana use: evidence from a regression discontinuity

design using exact date of birth. Journal of Health Economics 30 (4), 740–752.

Yörük, B. K. and C. E. Yörük (2013). The impact of minimum legal drinking age laws

on alcohol consumption, smoking, and marijuana use revisited. Journal of Health Economics

32 (2), 477–479.

Zhang, N. and E. Caine (2011). Alcohol policy, social context, and infant health: The

impact of minimum legal drinking age. International Journal of Environmental Research

and Public Health 8 (9), 3796–3809.

Zwerling, C. and M. P. Jones (1999). Evaluation of the effectiveness of low blood alcohol

concentration laws for younger drivers. American Journal of Preventive Medicine 16 (1),

76–80.

19


Figure 1: Histogram of Alcohol-Related Traffic Fatalities Ages 16-18

Percent

0 10 20 30

0 10 20 30

Alc. Rel. Crashes (BAC>0.08) Ages 16 to 18

Notes: Data from FARS. This graph shows a histogram of quarterly alcohol-related fatalities among 16-18

year old drivers. Alcohol-related fatalities are defined as those fatalities with a BAC level>0.08.

20


Figure 2: The Effect of FSP Laws on Alcohol-Related Traffic Fatalities Ages 16-18

-.8 -.6 -.4 -.2 0 .2 .4

-12 to -11

-10 to -9

-8 to -7

-6 to -5

-4 to -3

-2

-1

0 to 1

2 to 3

Centered Quarters

4 to 5

6 to 7

8 to 9

10 to 11

12+

Notes: This graph displays coefficients and the 95 percent confidence interval for a fixed effects poisson

regression where the dependent variable is the number of alcohol-related fatalities among 16-18 year old

drivers and the independent variables of interest are indicators for the time periods leading up to and after

the passage of a FSP law. In addition to the FSP law indicators, the regression controls for the presence of

distinctive license laws, false ID seizure laws, affirmative defense laws, right to sue minor laws, and right to

detain minor laws, the real beer tax, the presence of BAC 0.08 laws, the presence of Sunday sales bans, the

presence of seatbelt laws, the real gasoline tax rate, the real minimum wage, the unemployment rate, the

natural log of state per-capita income, state, year, and quarter fixed effects, and state-specific linear time

trends. The log of the number of 16-18 year olds in each state is included in the regression with a

coefficient constrained to one. Standard errors are clustered at the state level.

21


Figure 3: The Effect of FSP Laws on Non Alcohol-Related Traffic Fatalities Ages 16-18

-.8 -.6 -.4 -.2 0 .2 .4

-12 to -11

-10 to -9

-8 to -7

-6 to -5

-4 to -3

-2

-1

0 to 1

2 to 3

Centered Quarters

4 to 5

6 to 7

8 to 9

10 to 11

12+

Notes: This graph displays coefficients and the 95 percent confidence interval for a fixed effects poisson

regression where the dependent variable is the number of non alcohol-related fatalities among 16-18 year

old drivers and the independent variables of interest are indicators for the time periods leading up to and

after the passage of a FSP law. In addition to the FSP law indicators, the regression controls for the

presence of distinctive license laws, false ID seizure laws, affirmative defense laws, right to sue minor laws,

and right to detain minor laws, the real beer tax, the presence of BAC 0.08 laws, the presence of Sunday

sales bans, the presence of seatbelt laws, the real gasoline tax rate, the real minimum wage, the

unemployment rate, the natural log of state per-capita income, state, year, and quarter fixed effects, and

state-specific linear time trends. The log of the number of 16-18 year olds in each state is included in the

regression with a coefficient constrained to one. Standard errors are clustered at the state level.

22


Figure 4: The Effect of FSP Laws on Alcohol-Related Traffic Fatalities Ages 21-24

-.8 -.6 -.4 -.2 0 .2 .4

-12 to -11

-10 to -9

-8 to -7

-6 to -5

-4 to -3

-2

-1

0 to 1

2 to 3

Centered Quarters

4 to 5

6 to 7

8 to 9

10 to 11

12+

Notes: This graph displays coefficients and the 95 percent confidence interval for a fixed effects poisson

regression where the dependent variable is the number of alcohol-related fatalities among 21-24 year old

drivers and the independent variables of interest are indicators for the time periods leading up to and after

the passage of a FSP law. In addition to the FSP law indicators, all regression controls for the presence of

distinctive license laws, false ID seizure laws, affirmative defense laws, right to sue minor laws, and right to

detain minor laws, the real beer tax, the presence of BAC 0.08 laws, the presence of Sunday sales bans, the

presence of seatbelt laws, the real gasoline tax rate, the real minimum wage, the unemployment rate, the

natural log of state per-capita income, state, year, and quarter fixed effects, and state-specific linear time

trends. The log of the number of 21-24 year olds in each state is included in the regression with a

coefficient constrained to one. Standard errors are clustered at the state level.

23


Table 1: Alcohol Control Policy Effectiveness Dates

State

FSP Law

Distinctive

Licenses

Seizure of

ID

Affirmative

Defense

Right to

Sue Minor

Right to

Detain

Minor

Alabama

Alaska 1Q 1998 1Q 1998 3Q 2004

Arizona 3Q 2005 1Q 1998 3Q 2007

Arkansas 1Q 1998 3Q 2005

California 1Q 1998 1Q 1999 1Q 1998

Colorado

Connecticut 4Q 2001 1Q 1998 1Q 1998

Delaware

District of Columbia 1Q 1998 1Q 1998

Florida

Georgia

Hawaii 1Q 1998 1Q 1998

Idaho 1Q 1998 1Q 1998

Illinois

Indiana 1Q 2002 1Q 1998

Iowa

Kansas

Kentucky

Louisiana

Maine 1Q 1998 1Q 1998

Maryland

Massachusetts

Michigan

Minnesota 1Q 1998 3Q 2000 1Q 1998

Mississippi 1Q 1998 1Q 1998

Missouri 1Q 1998 1Q 1998

Montana

Nebraska 3Q 2010 1Q 1998 1Q 1998

Nevada

New Hampshire 1Q 2008 1Q 1998 1Q 2003

New Jersey

New Mexico

New York 3Q 1999 1Q 1998 1Q 1998

North Carolina 4Q 2001 1Q 1998 1Q 1998 1Q 1998

North Dakota 1Q 1998 3Q 2011 1Q 1998

Ohio 3Q 2000 1Q 1998 1Q 1998

Oklahoma 1Q 1998

Oregon 1Q 2000 1Q 1998 1Q 1998

Pennsylvania 4Q 2002 1Q 1998

Rhode Island 1Q 1998 1Q 1998 1Q 1998

South Carolina

South Dakota 1Q 1998 3Q 2000

Tennessee

Texas 3Q 2005 1Q 1998 1Q 1998

Utah 3Q 2009 1Q 1998 1Q 1998 1Q 1998 2Q 2009 1Q 1998

Vermont 3Q 2000 3Q 2000

Virginia

Washington

West Virginia 2Q 2003 2Q 2000 1Q 1998

Wisconsin 1Q 1998 1Q 1998 1Q 1998 4Q 2013

Wyoming

Notes: Data from APIS.

24


Table 2: Summary Statistics

Variable Mean Std.Dev

Alc. Rel. Crashes (BAC>0.08) Ages 16 to 18 4.318 4.408

Alc. Rel. Crashes (BAC>0.08) Ages 16 to 20 11.102 10.350

Alc. Rel. Crashes (BAC>0.08) Ages 21 to 24 18.599 16.073

Non-Alc. Crashes Ages 16 to 18 25.490 20.513

Non-Alc. Crashes Ages 16 to 20 47.860 38.686

Non-Alc. Crashes Ages 21 to 24 35.434 30.068

FSP Law 0.234 n.a.

Distinctive Licenses 0.508 n.a.

Seizure of ID 0.197 n.a.

Affirmative Defense 0.538 n.a.

Right to Sue Minor 0.022 n.a.

Right to Detain Minor 0.016 n.a.

Real Beer Tax 0.311 0.276

BAC Level of 0.08 0.846 n.a.

Sunday Alcohol Sales Ban 0.216 n.a.

Seatbelt Law 0.795 n.a.

Texting Ban 0.246 n.a.

Real Gas Tax 25.053 7.004

Real Minimum Wage 7.471 0.769

Unemployment Rate 6.164 2.138

Log Real Per-Capita Income 3.753 0.137

N 3468

Notes: Data from FARS, APIS, U.S. Brewers Association (2014), the Bureau

of Labor Statistics, the Bureau of Economic Analysis, Allegretto et al. (2016),

the Office of Highway Policy Information and the CDC. Summary statistics are

weighted by state population.

25


Table 3: The Effect of FSP Laws on Alcohol-Related Traffic Fatalities

Ages 16-18 Ages 16-20 Ages 21-24

(1) (2) (3) (1) (2) (3) (1) (2) (3)

26

FSP Law -0.141*** -0.123*** -0.160** -0.085* -0.056 -0.079* -0.022 -0.008 -0.061

(0.042) (0.039) (0.069) (0.047) (0.042) (0.047) (0.037) (0.032) (0.039)

Distinctive Licenses -0.157*** -0.316*** -0.040 -0.110 -0.048 -0.299***

(0.057) (0.091) (0.037) (0.072) (0.047) (0.055)

Seizure of ID -0.269*** -0.299** -0.188*** -0.170*** 0.114 0.004

(0.103) (0.142) (0.044) (0.064) (0.137) (0.055)

Affirmative Defense -0.164*** -0.194*** -0.311*** -0.387*** -0.223*** -0.168***

(0.049) (0.044) (0.038) (0.066) (0.040) (0.027)

Right to Sue Minor 0.018 -0.267 -0.092 -0.116 -0.147*** -0.054

(0.135) (0.263) (0.108) (0.174) (0.043) (0.113)

Right to Detain Minor 0.360** 0.524* 0.205*** 0.249 -0.074 -0.149***

(0.159) (0.275) (0.052) (0.173) (0.081) (0.025)

State Fixed Effects X X X X X X X X X

Year and Quarter Fixed Effects X X X X X X X X X

Other Controls X X X X X X X X X

Other Alcohol Controls X X X X X X

State-Specific Time Trends X X X

Semi-Elasticity [-0.131] [-0.116] [-0.148] [-0.081] [-0.054] [-0.076] [-0.022] [-0.008] [-0.059]

Num Obs. 3468 3468 3468 3468 3468 3468 3468 3468 3468

Notes: This table shows coefficients and standard errors clustered at the state level in parentheses for fixed effects poisson regressions where the

dependent variable is the number of alcohol-related traffic fatalities among different age groups as given in the column titles. The independent variable

of interest is an indicator for the presence of a FSP law, and semi-elasticities for the FSP Law coefficients, calculated as exp(β)−1, are given in brackets.

Other controls include the presence of seatbelt laws, the presence of texting while driving bans, the real gasoline tax rate, the real minimum wage, the

unemployment rate, and the natural log of state per-capita income. Other alcohol controls include the real beer tax, the presence of BAC 0.08 laws,

and the presence of Sunday sales bans. The log of the population of the relevant age group in each state is included in the regression with a coefficient

constrained to one. Standard errors are clustered at the state level. Stars denote statistical significance levels: ∗ : 10%, ∗∗ : 5%, and ∗∗∗ : 1%.


Table 4: Falsification Test: The Effect of FSP Laws on Non Alcohol-Related Traffic Fatalities

Ages 16-18 Ages 16-20 Ages 21-24

(1) (2) (3) (1) (2) (3) (1) (2) (3)

27

FSP Law -0.027 -0.025 -0.017 -0.032 -0.024 -0.035 -0.046 -0.029 -0.089***

(0.044) (0.043) (0.033) (0.032) (0.033) (0.031) (0.033) (0.029) (0.030)

Distinctive Licenses -0.031 -0.097* 0.014 -0.078* 0.151** 0.022

(0.039) (0.053) (0.028) (0.042) (0.065) (0.069)

Seizure of ID 0.024 0.017 0.090 0.076*** 0.009 -0.041

(0.039) (0.029) (0.070) (0.024) (0.057) (0.054)

Affirmative Defense -0.076 -0.129** -0.104*** -0.206*** -0.176*** -0.215***

(0.049) (0.056) (0.030) (0.051) (0.020) (0.030)

Right to Sue Minor 0.060 0.075 0.024 0.128 -0.181** -0.155***

(0.062) (0.066) (0.053) (0.078) (0.091) (0.054)

Right to Detain Minor 0.015 -0.037 0.011 -0.054 0.026 0.159*

(0.057) (0.055) (0.055) (0.112) (0.025) (0.085)

State Fixed Effects X X X X X X X X X

Year and Quarter Fixed Effects X X X X X X X X X

Other Controls X X X X X X X X X

Other Alcohol Controls X X X X X X

State-Specific Time Trends X X X

Semi-Elasticity [-0.027] [-0.024] [-0.016] [-0.031] [-0.024] [-0.034] [-0.045] [-0.029] [-0.085]

Num Obs. 3468 3468 3468 3468 3468 3468 3468 3468 3468

Notes: This table shows coefficients and standard errors clustered at the state level in parentheses for fixed effects poisson regressions where the

dependent variable is the number of non alcohol-related fatalities among different age groups as given in the column titles. The independent variable of

interest is an indicator for the presence of a FSP law, and semi-elasticities for the FSP Law coefficients, calculated as exp(β) − 1, are given in brackets.

Other controls include the presence of seatbelt laws, the presence of texting while driving bans, the real gasoline tax rate, the real minimum wage, the

unemployment rate, and the natural log of state per-capita income. Other alcohol controls include the real beer tax, the presence of BAC 0.08 laws,

and the presence of Sunday sales bans. The log of the population of the relevant age group in each state is included in the regression with a coefficient

constrained to one. Standard errors are clustered at the state level. Stars denote statistical significance levels: ∗ : 10%, ∗∗ : 5%, and ∗∗∗ : 1%.


Table 5: Robustness Tests: The Effect of FSP Laws on Alcohol-Related Traffic Fatalities

Controlling for

Non-Alc

Fatalities

Using All

Suspected Alc.

Fatalities as

Dependent

Variable

Using Logged

Rate of Alc.

Fatalities w/

Weights

Using Logged

Rate of Alc.

Fatalities w/o

Weights

Using Inv.

Normal Rate of

Alc. Fatalities

w/ Weights

Using Inv.

Normal Rate of

Alc. Fatalities

w/o Weights

28

FSP Law -0.144* -0.138*** -0.101* -0.114* -0.024** -0.028**

(0.077) (0.047) (0.051) (0.059) (0.012) (0.014)

Distinctive Licenses -0.284*** -0.297*** -0.225** -0.153 -0.052** -0.036

(0.098) (0.056) (0.086) (0.092) (0.020) (0.022)

Seizure of ID -0.330** -0.190* -0.230* -0.079 -0.053** -0.020

(0.155) (0.113) (0.115) (0.092) (0.026) (0.021)

Affirmative Defense -0.203*** -0.137* -0.071* -0.142** -0.018** -0.034**

(0.041) (0.077) (0.039) (0.055) (0.009) (0.013)

Right to Sue Minor -0.273 -0.274 -0.196 -0.132 -0.044 -0.030

(0.257) (0.301) (0.237) (0.156) (0.054) (0.036)

Right to Detain Minor 0.540* 0.378 0.158 0.281* 0.040 0.071*

(0.287) (0.274) (0.120) (0.161) (0.031) (0.040)

State Fixed Effects X X X X X X

Year and Quarter Fixed Effects X X X X X X

Other Controls X X X X X X

Other Alcohol Controls X X X X X X

State-Specific Time Trends X X X X X X

Adjusted R-Squared n/a n/a 0.680 0.722 0.683 0.723

Num Obs. 3468 3468 3468 3468 3468 3468

Notes: This table shows coefficients and standard errors clustered at the state level in parentheses. The dependent variable in the first two columns is the number of

alcohol-related fatalities among 16-18 year olds. In the second two columns it is the logged rate of alcohol-related fatalities among 16-18 year olds, and in the final two

columns it is the inverse normal distribution transformation of the rate of alcohol-related fatalities among 16-18 year olds. The independent variable of interest is an

indicator for the presence of a FSP law. In addition to the FSP law indicator, the regressions control for the presence of distinctive license laws, false ID seizure laws,

affirmative defense laws, right to sue minor laws, and right to detain minor laws, the real beer tax, the presence of BAC 0.08 laws, the presence of Sunday sales bans,

the presence of seatbelt laws, the presence of texting while driving bans, the real gasoline tax rate, the real minimum wage, the unemployment rate, the natural log

of state per-capita income, state, year, and quarter fixed effects, and state-specific linear time trends. In the fixed effects poisson regressions, the log of the number of

16-18 year olds in each state is included in the regression with a coefficient constrained to one. Standard errors are clustered at the state level. Stars denote statistical

significance levels: ∗ : 10%, ∗∗ : 5%, and ∗∗∗ : 1%.


Table 6: DDD Models: The Effect of FSP Laws on Alcohol-Related Traffic Fatalities

Ages 16-18 Ages 16-20

(1) (2) (3) (1) (2) (3)

Scanner x Underage -0.118*** -0.120*** -0.105 -0.053 -0.054 -0.030

(0.045) (0.044) (0.087) (0.044) (0.043) (0.054)

Underage -0.881*** -0.885*** -0.341*** -0.731*** -0.734*** -0.490***

(0.067) (0.068) (0.066) (0.053) (0.053) (0.046)

FSP Law -0.022 -0.008 -0.061 -0.026 -0.006 -0.056

(0.038) (0.033) (0.040) (0.038) (0.034) (0.040)

Distinctive Licenses -0.074* -0.301*** -0.046 -0.223***

(0.042) (0.032) (0.033) (0.058)

Seizure of ID 0.022 -0.071 -0.014 -0.070

(0.080) (0.057) (0.093) (0.043)

Affirmative Defense -0.211*** -0.173*** -0.256*** -0.256***

(0.031) (0.021) (0.033) (0.030)

Right to Sue Minor -0.115** -0.099 -0.129** -0.077

(0.058) (0.139) (0.064) (0.128)

Right to Detain Minor 0.024 0.011 0.039 0.014

(0.023) (0.082) (0.031) (0.078)

State Fixed Effects X X X X X X

Year and Quarter Fixed Effects X X X X X X

Other Controls X X X X X X

Other Alcohol Controls X X X X

State-Specific Time Trends X X

Semielasticity [-0.109] [-0.113] [-0.093] [-0.050] [-0.052] [-0.028]

Num Obs. 6936 6936 6936 6936 6936 6936

Notes: This table shows coefficients and standard errors clustered at the state level in parentheses for fixed effects

poisson regressions where the dependent variable is the number of alcohol-related fatalities among different age groups

as given in the column titles. The independent variable of interest is an indicator for the presence of a FSP law

interacted with an indicator for an underage group, and difference in semi-elasticities for the interaction coefficients,

calculated as exp(β INT + β F SP ) − exp(β F SP ), are given in brackets. Other controls include the presence of seatbelt

laws, the presence of texting while driving bans, the real gasoline tax rate, the real minimum wage, the unemployment

rate, and the natural log of state per-capita income. Other alcohol controls include the real beer tax, the presence of

BAC 0.08 laws, and the presence of Sunday sales bans. The log of the population of the relevant age group in each

state is included in the regression with a coefficient constrained to one. Standard errors are clustered at the state level.

Stars denote statistical significance levels: ∗ : 10%, ∗∗ : 5%, and ∗∗∗ : 1%.

29


Appendix Tables

30


Figure A1: The Effect of FSP Laws on Categories of Traffic Fatalities

Not Including State Time Trends

-.8 -.6 -.4 -.2 0 .2 .4

-12 to -11

-10 to -9

-8 to -7

-6 to -5

-4 to -3

-2

-1

0 to 1

2 to 3

Centered Quarters

4 to 5

6 to 7

8 to 9

10 to 11

12+

(a) Alcohol-Related Traffic Fatalities Ages 16-18

-.8 -.6 -.4 -.2 0 .2 .4

-12 to -11

-10 to -9

-8 to -7

-6 to -5

-4 to -3

-2

-1

0 to 1

2 to 3

Centered Quarters

4 to 5

6 to 7

8 to 9

10 to 11

12+

(b) Non Alcohol-Related Traffic Fatalities Ages

16-18

-.8 -.6 -.4 -.2 0 .2 .4

-12 to -11

-10 to -9

-8 to -7

-6 to -5

-4 to -3

-2

-1

0 to 1

2 to 3

Centered Quarters

4 to 5

6 to 7

8 to 9

10 to 11

12+

(c) Alcohol-Related Traffic Fatalities Ages 21-24

Notes: This graph displays coefficients and the 95 percent confidence interval for a fixed effects poisson

regression where the dependent variable is given by the figure caption and the independent variables of

interest are indicators for the time periods leading up to and after the passage of a FSP law. In addition to

the FSP law indicators, the regression controls for the presence of distinctive license laws, false ID seizure

laws, affirmative defense laws, right to sue minor laws, and right to detain minor laws, the real beer tax,

the presence of BAC 0.08 laws, the presence of Sunday sales bans, the presence of seatbelt laws, the real

gasoline tax rate, the real minimum wage, the unemployment rate, the natural log of state per-capita

31

income, state, year, and quarter fixed effects. The log of the number of individuals in the relevant age

groups in each state is included in the regression with a coefficient constrained to one. Standard errors are

clustered at the state level.


Table A1: The Effect of FSP Laws on Alcohol-Related Traffic Fatalities

Other Alcohol Policies and Control Variables

Ages 16-18 Ages 16-20 Ages 21-24

Real Beer Tax 0.259 -0.024 -0.153

(0.177) (0.108) (0.098)

BAC Level of 0.08 -0.077 -0.007 -0.007

(0.077) (0.039) (0.028)

Sunday Alcohol Sales Ban -0.112 -0.027 -0.026

(0.098) (0.042) (0.040)

Seatbelt Law -0.112** -0.086** -0.039

(0.045) (0.037) (0.033)

Texting Ban -0.100 -0.016 -0.025

(0.063) (0.043) (0.052)

Real Gas Tax 0.010** 0.005* -0.007***

(0.004) (0.003) (0.002)

Real Minimum Wage -0.025 0.017 -0.021

(0.031) (0.025) (0.021)

Unemployment Rate -0.021 -0.046*** -0.043***

(0.019) (0.014) (0.011)

Log Real Per-Capita Income 2.077** 1.753*** 0.923

(0.983) (0.586) (0.699)

State Fixed Effects X X X

Year and Quarter Fixed Effects X X X

State-Specific Time Trends X X X

Semi-Elasticity [-0.148] [-0.076] [-0.059]

Num Obs. 3468 3468 3468

Notes: These regressions show coefficients and standard errors for the other

controls not shown in Table 3 and refer to Specification (3) for each age group.

Stars denote statistical significance levels: ∗ : 10%, ∗∗ : 5%, and ∗∗∗ : 1%

32


Table A2: The Effect of FSP Laws on Alcohol-Related Traffic Fatalities

Event Study Specification

Alcohol-Related

Fatalities Ages 16-18

Non Alcohol-Related

Fatalities Ages 16-18

Alcohol-Related

Fatalities Ages 21-24

(1) (2) (1) (2) (1) (2)

11Q-12Q Before Law -0.003 -0.039 0.001 0.000 -0.057 -0.073

(0.106) (0.103) (0.054) (0.057) (0.037) (0.046)

9Q-10Q Before Law -0.125 -0.188** -0.153*** -0.153*** -0.095 -0.134*

(0.082) (0.085) (0.048) (0.043) (0.064) (0.069)

7Q-8Q Before Law 0.050 -0.025 -0.089* -0.095 -0.067 -0.112**

(0.090) (0.079) (0.048) (0.063) (0.061) (0.051)

5Q-6Q Before Law -0.095 -0.180** -0.135*** -0.162** 0.076 0.033

(0.061) (0.084) (0.048) (0.063) (0.072) (0.072)

3Q-4Q Before Law -0.183 -0.274** -0.081* -0.112** -0.027 -0.067

(0.131) (0.129) (0.046) (0.057) (0.055) (0.059)

2Q Before Law -0.055 -0.175** -0.105 -0.129 -0.298*** -0.348***

(0.086) (0.082) (0.083) (0.090) (0.094) (0.096)

1Q Before Law -0.341** -0.461*** -0.120 -0.148 -0.184** -0.235***

(0.168) (0.163) (0.079) (0.091) (0.075) (0.071)

0Q-1Q After Law -0.248*** -0.383*** -0.102** -0.141** -0.218*** -0.273***

(0.067) (0.065) (0.046) (0.061) (0.039) (0.065)

2Q-3Q After Law -0.164 -0.320** -0.117** -0.165*** -0.066 -0.138*

(0.134) (0.142) (0.055) (0.063) (0.069) (0.077)

4Q-5Q After Law -0.212*** -0.361*** -0.075* -0.125* -0.132*** -0.209***

(0.072) (0.103) (0.041) (0.066) (0.038) (0.066)

6Q-7Q After Law -0.089 -0.256** -0.117*** -0.170** -0.081 -0.165*

(0.095) (0.107) (0.042) (0.069) (0.076) (0.095)

8Q-9Q After Law -0.341*** -0.514*** -0.088 -0.132 -0.048 -0.123*

(0.107) (0.115) (0.110) (0.103) (0.063) (0.072)

10Q-11Q After Law -0.177* -0.366*** -0.057 -0.097 0.020 -0.059

(0.107) (0.137) (0.059) (0.078) (0.060) (0.083)

At Least 12Q After Law -0.178*** -0.450*** -0.093** -0.116* -0.008 -0.109

(0.068) (0.091) (0.045) (0.064) (0.045) (0.093)

State Fixed Effects X X X X X X

Year and Quarter Fixed Effects X X X X X X

Other Controls X X X X X X

Other Alcohol Controls X X X X X X

State-Specific Time Trends X X X

Num Obs. 3468 3468 3468 3468 3468 3468

Notes: This table displays coefficients and standard errors clustered at the state level in parentheses for fixed effects

poisson regressions where the dependent variable given by the column title and the independent variables of interest are

indicators for the time periods leading up to and after the passage of a FSP law. In addition to the FSP law indicators,

the regression controls for the presence of distinctive license laws, false ID seizure laws, affirmative defense laws, right

to sue minor laws, and right to detain minor laws, the real beer tax, the presence of BAC 0.08 laws, the presence of

Sunday sales bans, the presence of seatbelt laws, the presence of texting while driving bans, the real gasoline tax rate,

the real minimum wage, the unemployment rate, the natural log of state per-capita income, state, year, and quarter

fixed effects, and state-specific linear time trends. The log of the population of the relevant age group in each state is

included in the regression with a coefficient constrained to one. Standard errors are clustered at the state level.

33

More magazines by this user
Similar magazines