LIQUOR PRIVATIZATION Testimony of Antony Davies, Ph.D.

Associate Professor of Economics, Duquesne University Mercatus Affiliated Senior Scholar, George Mason University

PENNSYLVANIA SENATE LAW AND JUSTICE COMMITTEE February 14, 2011

I thank Chairman Pippy and the members of this committee for inviting me to present results of my

research on the privatization of alcohol markets. Good afternoon. I am Antony Davies, Associate Professor of Economics at Duquesne University and

Mercatus Affiliated Senior Scholar at George Mason University. I conduct studies on public policy issues for the Commonwealth Foundation and for the Mercatus Center. My primary area of expertise, however, is in the field of econometrics – the statistical analysis of economic data. Specifically, I invented the methodology that has become the standard for analyzing the most complex data sets – what we call multi-dimensional panel data. My research has appeared in the Journal of Econometrics, considered one of the top academic journals in the field of econometrics, and in academic texts published by the Cambridge University Press and the Oxford University Press.

In the summer of 2009, my co-author, John Pulito, and I conducted a study for the Commonwealth

Foundation on the relationship between the privatization of alcohol markets and alcohol-related social outcomes. Our review of the then existing literature revealed numerous studies that came to no clear consensus. For example, a 2003 study of alcohol outlet density and DUI fatalities in California over an eight year period found that increased outlet density was associated either with no change in DUI fatalities or a decrease in DUI fatalities, depending on how one defines “density”.1 A 2003 study of outlet density among students at eight public universities found a positive relationship between outlet density and self-reported drinking problems.2 A 2005 study of alcohol privatization in Alberta over the period 1950 through 2000 found no relationship between privatization and DUI fatalities.3 A 2006 study looking at privatization across all states for a single year found a positive relationship between privatization and DUI fatalities.4 These and other studies detailed in the literature review I have submitted to the committee provide conflicting stories as to the relationship between privatization and social outcomes.5,6

In response to this cacophony of disparate results, Pulito and I attempted to improve on the existing literature in two ways. First, rather than classifying states as either “control” or “license” as previous authors had done, we ranked states according to the degree of control the states exercised over alcohol markets. States that controlled alcohol sales at both the wholesale and retail levels and that controlled beer, wine, and spirits received the highest control rating. States that controlled sales only at the wholesale or the retail level, or that did not control all forms of alcohol received lower control ratings. 1 McCarthy, P., 2003. Alcohol-related crashes and alcohol availability in grass-roots communities. Applied Economics, 35: 1331-1338. 2 Weitzman, E.R., A.Folkman, K.L. Folkman, and H. Wechsler, 2003. The relationship of alcohol outlet density to heavy and frequent drinking and drinking-related problems among college students at eight universities. Health and Place, 9: 1-6. 3 Trolldal, B., 2005. An investigation of the effect of privatization of retail sales of alcohol on consumption and traffic accidents in Alberta, Canada. Addiction, 100: 662-671. 4 Miller, T., C. Snowden, J. Brickmayer, and D. Hendrie, 2006. Retail alcohol monopolies, underage drinking, and youth impaired driving deaths. Accident Analysis and Prevention, 38: 1162-1167. 5 Davies, A., 2010. Review of studies on liquor control and consumption. Commonwealth Foundation. 6 Davies, A. and J. Pulito, 2010. Binge thinking: A look at the social impact of state liquor controls. Mercatus Center Working Paper, no. 10-70.

Our goal was to extract more information from the data by examining the level of state control, not merely the presence or absence of control. Second, whereas previous studies looked at one state over time, or at several states at single point in time, we looked at all states over time. Our goal was to extract more information from the data by examining changes that occur over time and changes that occur across states. In econometrics, we call a data set like this a “panel data” set. Panel data sets are superior to traditional longitudinal or cross-sectional data sets not merely because they contain more data, but because they capture information and relationships that are impossible to capture in traditional data sets.

I mention all this to bring to the fore the fact that the two studies we wrote and which I have submitted to the committee, stand apart from the previous literature because they are the only studies to date that use the most advanced data sets and the most advanced analytic techniques that can be brought to bear. In our first study, we looked at all states and compared the incidence of underage drinking and underage binge drinking across states with different degrees of alcohol control.7

We found no relationship between alcohol control and underage drinking. For example, among the ten states with the highest rates of underage drinking, seven are license states (i.e., the lowest level of control). But, among the ten states with the lowest rates of underage drinking, six states are license states. We found no statistically significant change in the incidence of underage drinking among the four levels of control. Similarly, we found no difference in the incidence of underage binge drinking among the four levels of control. We also looked at all states over sixteen years and compared DUI fatality rates across states with different degrees of alcohol control. We found that states with the most stringent controls have DUI fatality rates that are significantly greater than in states with less stringent controls.

Our first study used a more comprehensive data set than has been used by previous studies, but employed the same sort of simple cross-state comparison employed in previous studies. In our second study, we looked at forty-nine states over twenty-one years and employed sophisticated panel data analytic techniques.8

In this study, we looked only at DUI fatalities, but we controlled for differences (across states and across time) in the minimum drinking age, mandatory seat belt laws, BAC limits, zero tolerance laws, keg registration laws, preliminary breath test laws, open container laws, and dram shop laws. After filtering out the effects of these laws on DUI fatality rates, we found that states that controlled alcohol markets experienced higher alcohol-involved fatality rates among the legal age population and the same or higher alcohol-involved fatality rates among the underage population.

In more than twenty years of research, numerous studies have failed to reach a consensus as to what social benefits, if any, people derive from their states controlling alcohol markets. The results of our two studies indicate that state control of alcohol markets does not contribute to improved social outcomes and, disturbingly in the case of DUI fatalities, appears to contribute to reduced social outcomes.

7 Pulito, J. and A. Davies, 2009. Government-run alcohol stores: The social impact of privatization. Commonwealth Foundation Policy Brief, 21(3): 1-16. 8 Pulito, J and A. Davies, 2010. State control of alcohol sales as a means of reducing traffic fatalities: A panel analysis. Under review.

REVIEW OF STUDIES ON LIQUOR CONTROL AND CONSUMPTION

Antony Davies, PhD, Duquesne University

Policy Variables and Policy Targets

Studies on the efficacy of alcohol controls have focused on the three broad categories of

market intervention:

Demand mitigation. Laws aimed at mitigating demand target by whom and the manner in

which alcoholic beverages can be consumed. Among others, these laws include age

restrictions, public consumption, and DUI/DWI laws. What the laws have in common is

that they put a restrictive burden on the alcohol buyer.

Supply mitigation. Laws aimed at mitigating supply target by whom and the manner in

which alcoholic beverages can be sold. These laws include government ownership of

retail or wholesale outlets, restrictions on outlet density, and restrictions on hours of

operation. These laws and others like them place restrictive burdens on the alcohol seller.

Transaction mitigation. Laws aimed at mitigating transactions target the act of buying

and selling. These laws principally take the form of taxes and, in more rare cases, price

controls.

While some laws may fall into more than one category (for example, depending on how

they are enforced, keg registration laws could be considered demand mitigating or supply

mitigating or both), the categories broadly reflect differences in policy efficacy noted in the

literature. For example, in their meta studies, Carpenter and Dobkin (2010), Campbell et al.

(2002), Grube and Nygaard (2001), and Her et al. (1999) find that studies report markedly

different degrees of policy success depending on whether the policies are targeting alcohol

demand, alcohol supply, or alcohol transactions.

In addition to various policy variables, studies have looked at various policy targets

including:

Sales/consumption mitigation. Studies that seek to focus on consumption typically

take per-capita sales as a proxy. A reasonable counterargument to using this target

is that alcohol consumption is not universally bad nor is a decline in alcohol

consumption universally good.

Underage drinking mitigation. Underage drinking data typically comes from

surveys and so the data are less reliable. A major source of data in the United

States is the National Survey on Drug Use and Health which asks children ages 12

and over to self-report their alcohol use.

Underage binge drinking mitigation. As with underage drinking data, underage

binge drinking data is also subject to self-report bias. Given the nature of the

question (binge drinking), subjects’ responses are likely to be even less reliable

than responses to underage drinking. Particularly where younger teenagers are

concerned, there may be incentives to lie in either direction (for detection

avoidance or self-aggrandizement). If underage drinking and underage binge

drinking data are subject to bias, so long as the bias is consistent, studies that look

at changes in these measures are likely to obtain results that are at least

directionally correct.

Alcohol-related traffic accident mitigation. In the United States, alcohol-related

traffic accidents are accidents in which at least one person involved in the

accident, and who is not a vehicle passenger, has a blood alcohol content (BAC)

above the statutory maximum. It is not necessary for an automobile driver to be

the one with the high BAC. For example, if a car strikes a pedestrian and the

pedestrian’s BAC is above the minimum, the accident will be classified as

alcohol-related.

Alcohol-related traffic fatality mitigation. An alcohol-related traffic fatality is an

alcohol-related traffic accident in which at least one person died.

Alcohol-involved traffic accident/fatality mitigation. An alcohol-involved traffic

accident is an accident in which at least one automobile driver has a BAC above

the statutory maximum.

Crime. Carpenter and Dobkin (2010) offer a review of studies examining the

causal effects of alcohol consumption on crime. Studies that examine crime focus

on violent and property crimes as opposed to DUI and public intoxication.

The purpose of this paper is to summarize research on the effects of alcohol policy

variables on alcohol policy targets, with a primary focus on privatizations. The goal is to

determine whether there are trends in the research findings that can inform future policy makers.

Outlet Density

Controlling alcohol markets by controlling outlet density (the number of retail

establishments per square mile) is closely related to controlling markets by monopolization.

States that monopolize retail alcohol sales control outlet density by deciding where to place the

state’s retail outlets. States with privatized (or partially privatized) retail markets employ density

controls to restrict the density of private establishments. With respect to controlling retail prices,

the types of beverages sold, hours of operation, and other non-location circumstances,

monopolization provides the state with power that outlet density rules do not. However, if a

state’s goal is merely to control outlet density, monopolization is not necessary – state licensing

within a privatized retail market serves the same purpose.

McCarthy (2003) performs a panel analysis of 111 non-metropolitan California cities

over the period January 1981 through December 1989. He looks at changes in the densities

(establishments per square-mile) of general off-site alcohol licenses (licenses to sell all types of

alcohol for consumption elsewhere), general on-site alcohol licenses (licenses to sell all types of

alcohol for consumption on the retailer’s premises), and off-site and on-site beer/wine licenses

(licenses to sell only beer and wine), and compares these to changes in fatal, non-fatal, and total

alcohol-related traffic accidents. McCarthy finds that an increase in the density of general off-site

licenses is associated with decreases in fatal, non-fatal, and total alcohol-related traffic accidents,

and that an increase in general on-site licenses is associated with increases in non-fatal, but not

fatal, accidents. He also finds that, among beer/wine licenses, an increase in off-site license

density is associated with a decline in total accidents, but that an increase in on-site density is

associated with an increase in non-fatal accidents.

Table 1. McCarthy (2003) results (density = outlets per square mile). Alcohol-Related Traffic Accidents Non-Fatal Fatal Total Increase in density of off-site general licenses Decrease Decrease Decrease Increase in density of on-site general licenses Increase No change No change Increase in density of off-site beer/wine licenses No change No change Decrease Increase in density of on-site beer/wine licenses Increase No change No change

Stockwell et al. (2009) examined the effect of alcohol outlet density and the degree of

privatization among retail alcohol stores on alcohol sales in British Columbia. British Columbia

provides an interesting case study as the province has permitted a gradual increase in the number

of private alcohol stores from 1988 to the present. Unlike McCarthy (2003), who defines density

as number of outlets per square mile, Stockwell et al. define density as number of outlets per

population. They look at 89 regions within British Columbia over the period April 2003 through

March 2008 and claim to find that increased density and increased privatization is associated

with increased per-capita alcohol sales. Their results, however, leave unaddressed the question of

causality – is increased privatization causing increased sales of alcohol, or is increased demand

for alcohol resulting in increased profit opportunities and therefore increased number of private

outlets. Finally, the discussion of their statistical results leaves unaddressed potential technical

errors that would, if present, render their estimates strongly biased in favor of their reported

findings. Stockwell et al. do not discuss whether or not they tested for non-stationarity in their

data. It is reasonable to assume that the time series data that they describe using would be non-

stationary. Further, the results shown in their Table 4 (p. 1832) imply test statistics that are near-

impossibly large for a correctly specified model (maximum = 83.5, average = 22.9).

Table 2. Stockwell et al. (2009) results (density = outlets per population 15 and older). Per-Capita Alcohol Consumption Increase in density of beer outlets Increase Increase in density of wine outlets Increase Increase in density of spirits outlets Increase Increase in density of on-site beer/wine licenses Increase

Weitzman et al. (2003) examine data on drinking habits among college students at eight

public universities and compare the self-reported measures to retail outlet densities. Each of the

3,421 students who participated in the study self-reported to which of the classifications their

drinking behaviors belonged: heavy drinking, frequent drinking, drinking-related problems,

frequent drunkenness, non-binge drinking, binge drinking, drinks-to-get-drunk, and abstention.

Weitzman et al. find positive correlations between retail outlet density and heavy drinking,

frequent drinking, and drinking-related problems. As with related studies, the authors stress that

their results are correlational and that neither causality nor the direction of causality is implied.

Table 3. Weitzman et al. (2003) results (density = outlets per square mile). Correlation with Self-Reported Behaviors Heavy Drinking Frequent Drinking Drinking-Related Problems Outlet density Positive Positive Positive

These studies as well as others (Presley et al., 2002; Douglas et al., 1997; Gruenewald et

al., 1996) point to a possible relationship between retail outlet density and alcohol consumption.

As the studies are not experimental, they leave two important issues unaddressed: (1) Is the

positive correlation between outlet density and alcohol consumption causal? While repeated

correlational studies might suggest causality, there remains the possibility that there is no

causality present. For example, it is possible that college students tend to drink more as a

consequence of age, new-found freedom, and propensity to take risks, and it is possible that the

density of alcohol retail outlets is higher near universities simply because the density of people is

higher near universities. (2) Assuming causality is present, what is the direction of the causality?

While there is a natural tendency to blame markets for people’s behaviors, in fact, markets are

merely the aggregation of people’s behaviors. In other words, markets do not cause behavior;

behavior causes markets. If the correlation between outlet density and alcohol consumption were

shown to be causal, it would be tempting to blame increased consumption on the increased

availability of alcohol. However, an equally compelling (some may argue, more compelling)

argument is that the density of retail outlets is caused by the propensity of the nearby populace to

consume alcohol. Finally, if the relationship between density and consumption is causal, it is

possible that the causality is bi-directional. It may be the case that both increased density causes

increased consumption and that increased consumption contributes to increased density.

From a policy perspective, the unanswered causality question is paramount. If the

relationship between retail outlet density and alcohol consumption is not causal, or if it is causal

but the causality runs from consumption to density or is bi-directional, restrictions on outlet

density will have no effect on alcohol consumption. Worse, as is the case with all social policies,

implementing the policy may lead people to falsely believe that the government is judiciously

spending its resources in pursuit of a valuable social goal, and to erroneously equate spending

and regulation directed toward the goal with the achievement of the goal.

Privatization

In an early review of literature on state monopolization of alcohol markets as a policy

tool for reducing alcohol consumption, Holder (1993) looked at the use of state monopolization

of alcohol markets as a means of combating alcohol consumption and, by extension, alcohol-

related problems. Implicit in Holder’s review, and in many subsequent studies, is the assumption

that alcohol consumption causes, rather than is caused by (or unrelated to), social ills. If, in fact,

the causality is reversed or not present, we would expect that reducing alcohol consumption

would have no effect on social ills. Holder concludes that research demonstrates that limitations

on the availability of alcohol can reduce the consumption of alcohol and that this effect is most

pronounced when alternative, unrestricted forms of alcohol do not exist. However, this result

seems to be tautological in that it is not possible to consume what does not exist.

MacDonald (1986) looked at privatization of wine sales in Idaho and Maine (both in

1971) and found that wine sales increased significantly following privatization, but that beer and

spirits sales did not. In 1969, grocery stores in Washington were allowed to sell imported wines.

Following this privatization, MacDonald detected an increase in wine sales – despite two

mitigating factors: (1) grocery stores already sold domestic wines, and (2) grocery stores charged

prices 25% higher than those in state-owned stores. As confirmed by later studies, MacDonald

found that the wine privatization had no effect on beer and spirits sales. Virginia’s privatization

of fortified wine sales in 1974 was not associated with an increase in wine, beer, or spirits sales.

MacDonald suggests that this result may be due to the fact that fortified wine comprised a very

small portion of the overall market for wine.

Table 4. MacDonald (1986) results. Beer Sales Wine Sales Spirits Sales Privatization of retail wine stores (Idaho, Maine) No change Increase No change Privatization of retail wine stores (Washington) No change Increase No change Privatization of retail fortified wine stores (Virginia) No change No change No change

Holder and Wagenaar (1990) found that, following Iowa’s privatization of liquor stores,

sales of spirits rose significantly (9.5%), sales of wine fell significantly (13.7%), and sales of

beer did not change. Wagenaar and Holder (1995) look at the privatization of wine sales in

Alabama (1973 and 1980), Idaho (1971), Maine (1971), Montana (1979), and New Hampshire

(1978) over the period 1968 through 1991. Employing a Box-Jenkins modeling technique to

measure the relationship between privatization and alcohol sales, they find that each of the states

experienced significant increases in wine sales following privatization. However, they found no

significant change in beer and spirits sales following privatization. These results are consistent

with their 1991 study in which they find similar results for privatization in Iowa (1985) and West

Virginia (1981).

Table 5. Holder and Wagenaar (1990) results. Beer Sales Wine Sales Spirits Sales Privatization of retail spirits stores (Iowa) No change Decrease Increase Table 6. Wagenaar and Holder (1991) results. Beer Sales Wine Sales Spirits Sales Total Sales Privatization of retail wine stores (Iowa, West Virginia)

No change Increase No change Increase

Table 7. Wagenaar and Holder (1995) results. Beer Sales Wine Sales Spirits Sales Privatization of retail wine stores (Alabama, Idaho, Maine, Montana, New Hampshire)

No change Increase No change

Conversely, Mulford, Ledolter, and Fitzgerald (1992), who also examined data before

and after Iowa’s privatization, found that the privatization effect on wine sales was temporary

(dropping to insignificance by two years after privatization), and found no evidence of an

increase in spirits sales following privatization. Mulford et al.’s analysis differs from Wagenaar

and Holder’s in several important respects. Mulford et al. have 29 more months of data following

privatization. With the additional data, Mulford et al. are more likely than Wagenaar and Holder

to detect the temporary nature of the privatization effect, if indeed the effect were temporary.

Mulford et al. also express concern that Wagenaar and Holder’s model was misspecified in that

Wagenaar and Holder’s model implicitly assumes that any change in baseline alcohol sales

following privatization would be permanent. Thus, not only are Wagenaar and Holder’s data less

able to detect temporary effects, but their model expressly assumes that privatization effects are

permanent. Also, Wagenaar and Holder incorrectly include sales of wine coolers in their data.

Because the privatization had no statutory effect on wine cooler distribution, sales of wine

coolers should not be included in their measures of wine sales. Not only should wine cooler sales

not have been included in the data, but, by coincidence, there was a surge in popularity of wine

coolers that occurred around the time of Iowa’s privatization. This coincidence, combined with

the erroneous inclusion of wine cooler sales, causes Wagenaar and Holder’s results to be biased

toward showing a positive privatization effect.

Table 8. Mulford, Ledolter, Fitzgerald (1992) results. Beer Sales Wine Sales Spirits Sales

Privatization of retail wine stores (Iowa) No change Temporary increase;

No change after two years No change

Finally, it is worth noting that Wagenaar and Holder do not mention testing for non-

stationarity – an anomaly that frequently plagues time series data. The presence of non-

stationarity in a time series data set results in spurious parameter estimates – results that are

biased toward significance. In their 1991 paper, Wagenaar and Holder report that they made their

data stationary by using first differences, and, in their 1995 paper, do not mention stationarity at

all. In 1991, they do not report the results for stationarity tests nor discuss performing post-

regression tests for stationarity in the residuals. Were it not for their reported regression results,

this would be less of an issue. However, their regression results (both in the 1991 and 1995

papers) exhibit extremely high multiple correlation coefficients and extremely large test statistics

– both of which can indicate the presence of unaddressed non-stationarity. As discussed later,

unaddressed non-stationarity frequently results in erroneous findings of significant relationships.

Rehm and Gmel (2001) raise concerns about this problem, particularly as it relates to published

research investigating alcohol use.

Trolldal (2005a) looked at alcohol sales and vehicle fatalities in Alberta over the period

of 1950 to 2000. From 1993 to 1994, Alberta privatized all of its retail liquor stores resulting in

an almost tripling of the number of retail outlets selling wine or spirits. Controlling for changes

in after-tax income and alcohol prices, Trolldal finds that sales of spirits increased following

privatization, that sales of beer and wine were unchanged, that the change in total sales of

alcohol were insignificant, and that privatization had no significant effect on vehicle fatalities.

Trolldal notes that these results contradict previous studies and suggests that part of the

difference may lie in the fact that Alberta privatized retail, but not wholesale, markets. Because

the state maintained monopolization of wholesale markets, it is possible that wholesale

restrictions limited the extent to which retail markets could grow following privatization.

Trolldal also notes that total sales did not increase despite an increase in the number of retail

outlets. This, he suggests, can imply the existence of a saturation point beyond which further

increases in the number of outlets has no effect on sales. This result suggests that previous

findings that outlet density affects sales may only hold for lower density geographical areas.

Table 9. Trolldal (2005a) results.

Beer Sales Wine Sales Spirits Sales Total Sales Alcohol-Related Vehicle Fatalities

Privatization of retail spirits stores (Alberta)

No change No change Increase No change No change

Trolldal (2005b) looked at the effect of two rounds of privatization of retail alcohol

markets in Quebec. Starting in 1978, grocery stores were permitted to sell wines that were either

produced in Canada or that were imported and bottled by the Quebec Liquor Board. Starting in

1983, grocery stores were allowed to sell wines that were imported and bottled by private

Quebec manufacturers. Using annual data from Quebec over the period 1950 through 2000, he

finds that wine sales increased by 10% following the 1978 privatization, but that the change was

small enough such that there was no discernable effect on total alcohol sales. He also finds that

the 1983 privatization had no effect on sales of beer, wine, or spirits.

Table 10. Trolldal (2005b) results. Beer Sales Wine Sales Spirits Sales Total Sales Privatization of retail wine stores (Quebec, 1978)

No change Increase No change No change

Privatization of retail wine stores (Quebec, 1983)

No change No change No change No change

In addition to looking at outlet density, Stockwell et al. (2009) look at the relationship

between privatization and alcohol sales after controlling for the possible effect of outlet density

on sales. They find that increases in the proportion of privately owned retail stores were

associated with increases in total alcohol sales.

Table 11. Stockwell et al. (2009) results. Per-Capita Alcohol Consumption Increase in proportion of privately owned retail stores Increase

Miller et al. (2006) examine the relationship between alcohol retail monopolies and the

incidences of underage drinking, underage binge drinking, and DUI involved fatalities. Like

Pulito and Davies (2009), Miller et al. perform difference of means tests for privatized versus

non-privatized states. Among other things, this has the effect of side-stepping the possible

stationarity issues that plague time-series studies. Miller et al. find that the average incidences of

underage drinking, underage binge drinking, and DUI fatalities were higher for states with

privatized alcohol markets. In contrast to Pulito and Davies, Miller et al. look only at a single

year for each state (data for most of the states are from 2001, though some are from as far back

as 1997 or from 2002). Miller et al. perform weighted difference of means tests where the

observations are weighted for each state’s population size. This is odd given that they are

analyzing incidences of drinking. Weighting incidences of underage drinking by a state’s

population size counteracts the division by population size necessary for calculating the

incidence in the first place. This results in an analysis that, in effect, compares the number of

underage drinkers across states rather than the incidence of underage drinking across states,

thereby biasing results in the direction of the largest states. Employing the correct procedure of

comparing the unweighted incidences of underage drinking yields results that are directionally

similar to those found by Miller et al., but which are (marginally) statistically insignificant (p =

0.076 for the difference in underage drinking, p = 0.053 for the difference in underage binge

drinking). The insignificance ceases to be marginal if we remove Utah from the data set – a

reasonable adjustment given the number of people with strong religious views against alcohol

consumption. Among non-privatized states, Utah’s incidence of underage drinking is a

remarkable 7 standard deviations below the mean while Utah’s incidence of underage binge

drinking is 5.7 standard deviations below the mean! Removing Utah yields results of p = 0.124

for the difference in underage drinking and p = 0.096 for the difference in underage binge

drinking, indicating that states with privatized alcohol markets exhibit incidences of underage

drinking and underage binge drinking that are not significantly different from those of states with

state-controlled alcohol markets.

Table 12. Miller et al. (2006) results.

Underage Drinking Underage Binge Drinking DUI Fatalities

Privatization of retail wine or spirits stores (50 U.S. states)

Increase Increase Increase

Like Miller et al., Pulito and Davies (2009) perform difference of means tests for

differences in the incidences of underage drinking, underage binge drinking, per-capita alcohol

consumption, and DUI fatalities among privatized versus non-privatized states. Where Miller et

al. look at the states in a single year, Pulito and Davies look at the states over a period of 16

years (for per-capita consumption and DUI fatalities) and classify states according to the degree

of privatization. They categorize the possible degrees of privatization versus control as full

control (sales of beer, wine, and spirits are controlled at the retail and wholesale levels),

moderate control (sales beer, wine, and spirits are controlled at the wholesale level, and sales of

only one type of alcohol are controlled at the retail level), light control (sales of beer, wine, and

spirits are controlled at the wholesale level, and sales of all three types are privatized at the retail

level), and no control (sales of beer, wine, and spirits are privatized at the wholesale and retail

levels). They compare the degree of control to per-capita alcohol consumption, the incidence of

underage drinking, the incidence of underage binge drinking, and alcohol-related traffic

fatalities. They find that alcohol consumption is significantly greater under no control than under

light control, but statistically identical among full, moderate, and light control. They find no

significant difference in the incidence of underage drinking or the incidence of underage binge

drinking among the four control classifications. While they find no difference in the rate of DUI

arrests among the four control classifications, they do find that the number of alcohol-related

traffic fatalities per DUI arrest are the same under full control and no control, and significantly

lower under moderate and light control.

Table 13. Pulito and Davies (2009) results (changes are as compared with no privatization).

Underage Drinking

Underage Binge Drinking Total Sales DUI Fatalities

Partial privatization of retail stores (48 U.S. states)

No change No change No change Decrease

Full privatization of retail stores (48 U.S. states)

No change No change Decrease Decrease

Full privatization of retail and wholesale stores (48 U.S. states)

No change No change Increase Decrease

Conclusion

These and other studies suggest that there is no clear evidence that privatization of

alcohol markets leads to decreased social measures – whether consumption, underage drinking,

or DUI fatalities. Studies that show relationships are counterbalanced by other studies, of the

same data, that show no relationship. Some studies that show relationships may suffer from

unaddressed statistical anomalies that bias results in favor of finding relationships where none

exist. Studies that show relationships also suffer from unaddressed causality, making the results

useless for guiding policy makers. Future studies can correct some of these shortcomings by

employing more rigorous statistical techniques, though the issue of causality may never be

adequately addressed. Nonetheless, even if causality were left unaddressed, a preponderance of

statistically defensible results in one direction or the other would go a long way to informing

policy.

Currently Pennsylvania, one of the few remaining states that have privatized neither wine

nor spirits stores, is considering privatizing these markets. A bill introduced by Rep. Mike Turzai

(House Bill 2350) would allow Pennsylvania to auction off 750 retail and 100 wholesale

licenses. Segal and Underwood (2007) estimate that Pennsylvania could raise almost $2 billion

from auctioning off its wholesale and retail liquor stores plus an additional $350 million annually

from alcohol sales taxes. In the last fiscal year, Pennsylvania incurred its largest ever budget

deficit – the bulk of which was closed by Federal monies. The money raised from privatizing the

state liquor stores would, by itself, have closed the budget deficit.

Similarly, Robert McDonnell, governor of Virginia, is pushing for that state to privatize

its liquor stores. The governor’s staff is considering four approaches to privatization: selling the

state’s outlets to a single private firm, offering spirits licenses to the 3,000 private firms that

currently sell beer and wine, selling existing state liquor stores to private firms, and auctioning

off a number of licenses. To proceed cautiously, and reflecting on studies sighted in this paper,

Virginia might be best served by auctioning off licenses and, perhaps, coupling the licenses with

options to buy existing stores to which the licenses would apply. This would allow the state to

control outlet density, pending further and more rigorous studies on the effect of outlet density

on alcohol sales, plus unload physical plant that it no longer needs.

Appendix: Non-Stationarity and Time Series Estimates

Time series data frequently exhibits what is known as non-stationarity. Measurements on

things such as population, income, production, and trade will naturally grow (on average) over

time. The result is that the time series will exhibit trends over time. Statistically, the trends cause

the time series to have undefined population means and infinite population variances. For

example, because per-capita income rises (on average) each year, the mean and variance of per-

capita income (from some fixed past point in the past to the present) increase with each passing

year. When analysts attempt to measure the relationship between two non-stationary series, the

natural trend dominates the relationship and drowns out any underlying relationship between the

two series. For example, because both the world population increases over time and the S&P 500

has been increasing over time, one can expect to find a strong correlation between world

population and the S&P 500. The strong correlation results from the fact that the two trends

dominate the analysis. To remove non-stationarity, the analyst can look at the first difference or

the growth rates in the two series. If there is no strong correlation between the historical growth

rate in the S&P 500 and the historical population growth rate, one can conclude that one likely

does not cause the other.

There are formal statistical tests for non-stationarity, but typically the problem manifests

itself in regression correlations close to 1.0, and test statistics that are remarkably high. For

example, the probability of a large asteroid hitting the earth on any given day is about one in 40

billion. That corresponds to a test statistic of about 6.5. An analysis that reports test statistics

above 6.5 is claiming probabilities that are possibly unbelievably small. At that size, it becomes

more likely that the data suffers from non-stationarity than that the analyst has truly found an

immensely improbable relationship.

References

Douglas, K.A., J.L. Collins, C. Warren, L. Kann, R. Gold, S. Clayton, J.G. Ross, and L.J. Kolbe, 1997. Results from the 1995 national college health risk behavior survey. Journal of American College Health, 46(2): 55-66.

Fitzgerald, J.L. and H.A. Mulford, 1992. Consequences of increasing alcohol availability: The Iowa experience revisited. British Journal of Addiction, 87: 267-274.

Gruenewald, P.J., A.B. Millar, and P. Roeper, 1996. Access to alcohol: Geography and prevention for local communities. Alcohol Health and Research World, 20(4): 244-251.

Holder, H.D., 1993. The state monopoly as a public policy approach to consumption and alcohol problems: A review of research evidence. Contemporary Drug Problems, 293-322.

Holder, H.D. and A.C. Wagenaar, 1990. Effects of the elimination of a state monopoly on distilled spirits’ retail sales: A time-series analysis of Iowa. British Journal of Addiction, 85: 1615-1625.

Macdonald, S., 1986. The impact of increased availability of wine in grocery stores on consumption: Four case histories. British Journal of Addiction, 81: 381-387.

McCarthy, P., 2003. Alcohol-related crashes and alcohol availability in grass-roots communities. Applied Economics, 35: 1331-1338.

Miller, T., C. Snowden, J. Brickmayer, and D. Hendrie, 2006. Retail alcohol monopolies, underage drinking, and youth impaired driving deaths. Accident Analysis and Prevention, 38: 1162-1167.

Mulford, H.A., J. Ledolter, and J.L. Fitzgerald, 1992. Alcohol availability and consumption: Iowa sales data revisited. Journal of Studies on Alcohol, September: 487-494.

Pulito, J. and A. Davies, 2009. Government-run alcohol stores: The social impact of privatization. Commonwealth Foundation Policy Brief, 21(3): 1-16.

Presley, C.A., P.W. Meilman, and J.S. Leichliter, 2002. College factors that influence drinking. Journal of Studies on Alcohol, 14: 82-90.

Trolldal, B., 2005a. An investigation of the effect of privatization of retail sales of alcohol on consumption and traffic accidents in Alberta, Canada. Addiction, 100: 662-671.

Trolldal, B., 2005b. The privatization of wine sales in Quebec in 1978 and 1983 to 1994. Alcoholism: Clinical and Experimental Research, 29(3): 410-415.

Rehm, J. and G. Gmel, 2001. Aggregate time-series regression in the field of alcohol. Addiction, 96: 945-954.

Segal, G.F. and G.S. Underwood, 2007. Divesting the Pennsylvania liquor control board. Reason Foundation, April: 1-10.

Stockwell, T., J. Zhao, S. Macdonald, B. Pakula, P. Gruenewald, and H. Holder, 2009. Changes in per capita alcohol sales during the partial privatization of British Columbia’s retail alcohol monopoly 2003-2008: A multi-level local area analysis. Addiction, 104: 1827-1836.

Wagenaar, A.C. and H.D. Holder, 1991. A change from public to private sale of wine: Results from natural experiments in Iowa and West Virginia. Journal of Studies on Alcohol, 52(2): 162-173.

Wagenaar, A.C. and H.D. Holder, 1995. Changes in alcohol consumption resulting from the elimination of retail wine monopolies: Results from five U.S. states. Journal of Studies on Alcohol and Drugs, 56(5): 566-572.

Weitzman, E.R., A.Folkman, K.L. Folkman, and H. Wechsler, 2003. The relationship of alcohol outlet density to heavy and frequent drinking and drinking-related problems among college students at eight universities. Health and Place, 9: 1-6.

Meta Studies

Campbell, C.A., R.A. Hahn, R.Elder, R. Brewer, S. Chattopadhyay, J. Fielding, T.S. Naimi, T. Toomey, B. Lawrence, and J.C. Middleton, 2002. The effectiveness of limiting alcohol outlet density as a means of reducing excessive alcohol consumption and alcohol-related harms. American Journal of Preventive Medicine, 37(6): 556-569.

Her, M., N. Giesbrecth, R. Room, and J. Rehm, 1999. Privatizing alcohol sales and alcohol consumption: Evidence and implications. Addiction, 94(8): 1125-1139.

Carpenter, C. and C. Dobkin, 2010. Alcohol regulation and crime. NBER Working Paper Series, w15828.

Grube, J.W. and P. Nygaard, 2001. Adolescent drinking and alcohol policy. Contemporary Drug Problems, 28(Spring): 87-131.

COMMONWEALTH FOUNDATION for PUBLIC POLICY ALTERNATIVES

POLICY BRIEF from the COMMONWEALTH FOUNDATION

Vol. 21, No. 03 October 2009

Government-Run Liquor Stores The Social Impact of Privatization JOHN PULITO & ANTONY DAVIES, PHD

EXECUTIVE SUMMARY A proposal that has circulated in Harrisburg for years is divesting, selling, or

privatizing Pennsylvania’s state-owned liquor stores. This discussion is particu-larly relevant today given the recent drop in Pennsylvania tax collections and long-term taxpayer obligations. Geoffrey Segal and Geoffrey Underwood of the Reason Foundation estimate that Pennsylvania could raise $1.7 billion from the sale of its wholesale and retail liquor stores. While such a sale would represent only a one-time cash inflow, Nathan Benefield of the Commonwealth Foundation estimates that Pennsylvania would continue to take in close to $350 million annually in alco-hol sales tax.

What gives many pause is the social impact of privatization. Myriad compari-

sons of privatized markets to state-controlled markets suggest that there are unques-tionable advantages to privatization. Of concern are the possible disadvantages. Liq-uor control proponents maintain that, because the state can directly limit the times and locations at which alcohol can be purchased, and because state stores are not profit driven like private firms, privatization would result in increased alcohol con-sumption and problems associated with alcohol consumption, such as impaired driving.

A comparison of states with varying degrees of privatization in the retail and

wholesale markets for alcohol over the period 1970 through 2006 suggests that pri-vatization is associated neither with increased alcohol consumption nor increased traffic fatalities involving impaired drivers.

• States that recently privatized their liquor industries experienced a signifi-

cant decline in per-capita alcohol consumption.

• While not conclusive, we find evidence that is consistent with the existence of a cross-border effect wherein liquor control encourages Pennsylvanians to purchase in neighboring privatized states.

• While states that have liquor controls experience somewhat lower consump-tion of alcohol, we find no evidence that the degree of control matters. Among privatized (license) states and states with varying degrees of control, states with controls on wholesale markets only had the lowest consumption rates.

Privatization is associated nei-

ther with in-creased alcohol

consumption nor increased traffic

fatalities involving impaired drivers.

2

COMMONWEALTH FOUNDATION | policy brief

Divestiture of Pennsylvania’s

state liquor stores would represent a financial windfall

to the state, while posing no threat to public safety,

as it would not result in the so-

cial ills many op-ponents of privati-

zation fear.

• States that have liquor controls experience significantly higher DUI-related fatality rates than states without controls.

• Adjusting for DUI enforcement, states with the highest degree of liquor con-trol exhibited the same alcohol-related driving deaths as did license states. States with lesser controls exhibited significantly fewer DUI-adjusted deaths.

• Evidence shows there is no significant reduction in underage drinking among control states versus license states. Pennsylvania (a full control state) ranks 22nd among the 48 states in the sample for incidence of underage drinking.

Examples at the state and federal levels demonstrate that government is not good at running industries. Repeatedly, the private sector shows that it can provide higher quality goods and services at lower costs. However, arguments might be made for state control as a means of achieving some desired social outcome. In Pennsylvania’s case, advocates claim that the social goals of reducing alcohol con-sumption, underage drinking, and alcohol-related traffic deaths justify controlling wholesale and retail alcohol markets.

Evidence from 48 states over time shows no link between market controls

and these social goals. Divestiture of Pennsylvania’s state liquor stores would rep-resent a financial windfall to the state, while posing no threat to public safety, as it would not result in the social ills many opponents of privatization fear.

3

policy brief | COMMONWEALTH FOUNDATION

INTRODUCTION: ADVANTAGES AND DISADVANTAGES OF PRIVATIZATION The adoption of the 21st Amendment granted states the power to regulate, sell,

and distribute alcoholic beverages. To date, nineteen states have opted to impose some form of control on liquor sales, ranging from controls on wholesale markets only to controls on retail and wholesale markets. Pennsylvania is one of only eight states that control both wholesale and retail markets. From time to time, Pennsyl-vania state policymakers have considered privatizing the state store system. This discussion is particularly relevant today given the recent drop in Pennsylvania tax collections and long-term taxpayer obligations.

Geoffrey Segal and Geoffrey Underwood (2007) of the Reason Foundation esti-

mate that Pennsylvania could raise $1.7 billion from the sale of its wholesale and retail liquor stores.1 While such a sale would represent only a one-time cash inflow, Nathan Benefield of the Commonwealth Foundation estimates that Pennsylvania would continue to take in close to $350 million annually in alcohol sales tax.2 What gives many pause is the social impact of privatization. Pennsylvania currently ranks 34th out of 48 states for per-capita alcohol consumption. Proponents of the state liq-uor system believe that this is due, at least in part, to the control exercised by the state liquor system.

In testimony before the Pennsylvania Senate, Segal and Underwood discussed

the following possible advantages and disadvantages to privatization. Among these are:3

Possible Advantages of Privatization

• Increased efficiency (i.e., lower consumer prices) • Improved customer service • Additional state revenue from the sale of liquor licenses

Possible Disadvantages of Privatization

• Increased alcohol consumption • Increased incidence of underage drinking • Increased alcohol-related motor vehicle accidents (DUI)

Myriad comparisons of privatized markets to state-controlled markets suggest

that there is little question as to the advantages of privatization. Of concern are the possible disadvantages. Liquor control proponents maintain that, because the state can directly limit the times and locations at which alcohol can be purchased, and because state stores are not profit driven as are private firms, privatization would result in increased alcohol consumption and increased problems associated with alcohol consumption such as impaired driving.

In this paper, we examine data from states with varying levels of privatization

over the period 1970 through 2006 to explore the relationship between privatiza-tion, alcohol consumption, and fatalities due to impaired drivers. We find evidence that more stringent state control of liquor markets has the reverse effect, and actu-ally is associated with increased consumption and alcohol-related highway deaths.

Pennsylvania is one of only eight states that con-trol both whole-sale and retail markets.

4

COMMONWEALTH FOUNDATION | policy brief

DEREGULATION: IOWA AND WEST VIRGINIA

Iowa and West Virginia deregulated their retail liquor markets in 1987 and 1990, respectively, providing two recent case studies in the effect of deregulation on con-sumption. Underwood and Segal found no evidence of increased alcohol consump-tion following deregulation in either Iowa or West Virginia. They also noted that though each individual outlet sold less alcohol on average compared to individual state-run stores, alcohol purchases were being distributed over a greater number of stores.

In Figures 1 and 2, the vertical line represents the point in time at which Iowa and West Virginia deregulated their retail liquor markets.

Following deregulation, both states experienced a statistically significant de-cline in average per-capita consumption of alcohol (versus the period prior to de-regulation). Total per-capita consumption of alcohol in Iowa and West Virginia fell 5.9% and 4.1%, respectively, post- versus pre-regulation. Both states also experi-enced an apparent shift in consumption away from higher alcohol-content products. In Iowa, consumption of liquor fell 27%, while consumption of wine and beer rose 29% and 2%, respectively. In West Virginia, consumption of liquor and wine fell by 39% and 12%, respectively, while consumption of beer rose 18%.

One possible explanation is a “convenience effect.” When the retail market is

regulated, it is less convenient for consumers to purchase alcohol (due to restricted hours, restricted retail locations, and a reduced focus on serving the customers’ needs). Consumers will respond to the reduced convenience by increasing the amount of alcohol they purchase per trip so as to reduce the number of trips they make over time, and by buying more high-alcohol products so as to reduce the vol-ume of product they must transport.

0.0

0.5

1.0

1.5

2.0

2.5

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

Gallons of Ethano

l

Total Beer Spirits Wine

Figure 1. Iowa: Alcohol Consumption per Capita (age 14 and older)4

5

policy brief | COMMONWEALTH FOUNDATION

Total per-capita consumption of alcohol in Iowa and West Virginia fell post- versus pre-regulation. Both states also experienced an apparent shift in consumption away from higher alcohol-content products.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

Gallons of Ethano

l

Total Beer Spirits Wine

Figure 2. West Virginia: Alcohol Consumption per Capita (age 14 and older)5

‐

0.2 

0.4 

0.6 

0.8 

1.0 

1.2 

1.4 

1.6 

1.8 

2.0 

2.2 

1970‐1986 1988‐2006

Gallons of Ethano

l Con

sumed

 Ann

ually

Beer Wine Liquor

Figure 3. Iowa: Per Capita Consumption Before and After Deregulation6

6

COMMONWEALTH FOUNDATION | policy brief

In summary, the evidence, while

not conclusive, is consistent with

the existence of a cross-border ef-

fect wherein regu-lation encourages Pennsylvanians to

purchase in neighboring de-

regulated states.

THE CROSS-BORDER EFFECT

It is possible that state control may encourage Pennsylvanians to purchase from border states that are deregulated—the cross-border effect—because of a greater number of retail outlets, more convenient operating hours, and/or lower prices in the deregulated states. If true, we should observe per-capita consumptions among deregulated bordering states to be greater than per-capita consumption in Pennsyl-vania. Figure 5 shows the per-capita consumptions over the period 1970 through 2006 for Pennsylvania and its bordering states. Pennsylvania exhibits a per-capita consumption that is very similar to that of Ohio, the only regulated border state.8 However, Pennsylvania’s per-capita consumption is significantly less than those of the deregulated border states, with the exception of West Virginia.9 With the excep-tion of the comparison to West Virginia, the differences in consumption are what one would expect to observe if there were cross-border effects.

Confounding this comparison, however, is the difference in taxes. Pennsyl-

vania’s tax on a gallon of spirits was $6.48 in 2006 versus $6.44 for New York, $4.40 for New Jersey, $3.75 for Delaware, $1.70 for West Virginia, and $1.50 for Mary-land.10,11 Because Pennsylvania’s tax is greater than the taxes in the deregulated bor-der states, it is unclear whether the difference in per-capita consumptions might be due to deregulation or due to differences in tax rates. Adding weight to the argu-ment that the difference is due to deregulation is the case of Ohio, where the tax ($2.25 per gallon) is one-third that in Pennsylvania. Yet Ohio’s per-capita consump-tion is identical to Pennsylvania’s. In summary, the evidence, while not conclusive, is consistent with the existence of a cross-border effect wherein regulation encour-ages Pennsylvanians to purchase in neighboring deregulated states.

‐

0.2 

0.4 

0.6 

0.8 

1.0 

1.2 

1.4 

1.6 

1.8 

1970‐1989 1991‐2006

Gallons of Ethano

l Con

sumed

 Ann

ually

Beer Wine Liquor

Figure 4. West Virginia: Per Capita Consumption Before and After Deregulation7

7

policy brief | COMMONWEALTH FOUNDATION

CLASSIFICATION OF “CONTROL” STATES:

What we have been describing as “regulated” and “deregulated” states, the Na-tional Alcohol Beverage Control Association (NABCA) classifies as “control” and “license” states, respectively. Specifically, NABCA defines a control state as one in which a controlled distribution system substitutes for the private marketplace in the wholesale and/or retail sale of alcohol. It is reasonable to assume that there are measurably different effects when the control is limited at the wholesale versus re-tail level. To examine these differences, we classify states according to the following levels of regulation: Table 1. Regulation Classifications

NABCA Control Retail or Wholesale Controlled

Alcohol sales are controlled at either the wholesale or retail levels. This is NABCA’s definition of “control”.

Full Control Retail and Wholesale Controlled

Sales of all three types of alcohol (beer, wine, and liquor) are controlled at the retail and wholesale levels.

Moderate Control Partial Retail and Wholesale Con-trolled

Sales of only one type of alcohol (beer, wine, or liquor) are controlled at the retail level, and sales of all three types are controlled at the wholesale level.13

Light Control Wholesale Controlled

No sales are controlled at the retail level, and sales of all three types of alcohol are controlled at the wholesale.

License No Control Alcohol sales are not controlled.

NABCA defines a control state as one in which a controlled distri-bution system substitutes for the private mar-ketplace in the wholesale and/or retail sale of alco-hol.

1.5

1.7

1.9

2.1

2.3

2.5

2.7

2.9

3.1

3.319

70

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

Per‐Capita Ethanol Con

sumed

 (gallons)

DE NJ MD PA OH NY WV

Figure 5. Comparison of Per-Capita Alcohol Consumption, Pennsylvania versus Border States12

8

COMMONWEALTH FOUNDATION | policy brief

Table 2 shows states belonging to the various regulation classifications over the period 1991 through 2006. States not listed are License states. With the exception of Montgomery County, Maryland is a License state. However, because our data is not at the county level, we cannot distinguish between sales occurring within and out-side Montgomery County. For this reason, Maryland is excluded from our analysis.

ALCOHOL CONSUMPTION PER CAPITA

If it is true that regulating the markets for alcohol results in reduced consump-tion, we should expect to see greater consumption in states in lower control catego-ries. The figure below shows average per-capita alcohol consumption over the pe-riod 1991 through 2006. Average per-capita consumption in states that NABCA de-fines as “controlled” is 5.5% lower than in License states (2.2 gallons versus 2.3 gal-lons). This difference is statistically significant, suggesting that control is associated with reduced consumption. However, when we break down the states by degree of control (the blue bars in Figure 6), we see that consumption rises as we reduce regu-lation from Full Control (2.1 gallons) to Moderate Control (2.3 gallons), though the increase is not statistically significant. Consumption then falls significantly as we further reduce regulation from Moderate Control to Light Control (1.9 gallons), and then rises again as we reduce to License.

Consumption then falls signifi-cantly as we fur-

ther reduce regu-lation from Mod-erate Control to

Light Control, and then rises again as we reduce to

License.

NABCA Control Full Control Moderate Control Light Control

Retail or Wholesale Retail and Wholesale Partial Retail and Wholesale Wholesale

Alabama Alabama Idaho Iowa

Idaho Maine Iowa Michigan

Iowa Mississippi Michigan West Virginia

Maine Montana New Hampshire

Michigan Pennsylvania North Carolina

Mississippi Utah Ohio

Montana Vermont Oregon

New Hampshire Wyoming Virginia

North Carolina Washington

Ohio West Virginia

Oregon Pennsylvania

Utah

Vermont

Virginia

Washington

West Virginia

Wyoming

Table 2. States According to Regulation Classification

9

policy brief | COMMONWEALTH FOUNDATION

Although we are examining data over 16 years, because there are only three states in the Light Control classification, it is possible that the sharp decrease in consumption for that group is due to some state-specific effects. In summary, after breaking states down by level of control, evidence suggests that, while regulating liquor at the wholesale level may contribute to reduced consumption, there is no clear evidence that regulating liquor at the retail level affects consumption.

UNDERAGE DRINKING

Advocates of state control argue that because private retailers will be less dili-gent about carding, control is a necessary protection against underage drinking. If true, we would expect to see an increase in the incidence of underage drinking as we move from control to license states. The National Survey on Drug Use and Health asks respondents aged 12 and over to report their alcohol use over the previ-ous thirty days. The results of this survey show no significant reduction in underage drinking among control states versus license states. Regardless of the degree of con-trol, the average incidence of underage drinking is between 29% and 31% (the dif-ferences are not statistically significant). For example, Pennsylvania (a Full Control state) ranks 22nd among the 48 states in the sample for incidence of underage drink-ing. Of the top-10 states for underage drinking, seven are license states. But, of the bottom-10 states for underage drinking, six are license states. Whether or not the purpose of retail and wholesale control is the mitigation of underage drinking, the data are clear that control has no effect on underage drinking.

In summary, after breaking states down by level of control, evidence suggests that, while regulating liquor at the wholesale level may contribute to reduced con-sumption, there is no clear evidence that regulating liquor at the retail level affects con-sumption.

‐

0.3 

0.5 

0.8 

1.0 

1.3 

1.5 

1.8 

2.0 

2.3 

2.5 

NABCA Full Control Moderate Control

Light Control License

Gallons of Ethano

l Con

sumed

 Ann

ually

Figure 6. Average Per-Capita Consumption by Regulation Classification, 1991-200614

10

COMMONWEALTH FOUNDATION | policy brief

Similarly, there is no statistically significant difference in the incidence of binge drinking across the regulation classifications.16 States with greater control of alcohol retail and wholesale markets do not experience lower average incidences of binge drinking among 12- to 20-year-olds.

States with greater control of alcohol retail and

wholesale mar-kets do not ex-perience lower

average inci-dences of binge drinking among 12- to 20-year-

olds.

Figure 7. Incidence of Underage Drinking by Regulation Classification, 200315

0%

5%

10%

15%

20%

25%

30%

35%

NABCA Full Control Moderate Control

Light Control License

Alcho

hol U

se Amon

g 12

‐20 Ye

ar Olds

Figure 8. Incidence of Underage Binge Drinking by Regulation Classification, 200317

0%

5%

10%

15%

20%

25%

NABCA Full Control Moderate Control

Light Control License

Binge Drinking Amon

g 12

‐20 Ye

ar Olds

11

policy brief | COMMONWEALTH FOUNDATION

DRIVING FATALITIES

Another argument against privatization is the concern that the number of fatali-ties resulting from driver impairment would increase. Figure 9 shows the number of drivers (per 100,000 population) involved in fatal accidents for which a driver had a blood alcohol content (BAC) of 0.01 or higher. To clarify, the figure shows the num-ber of impaired drivers involved in fatal accidents, not the number of impaired driv-ers killed. The latter measure, while available, is less meaningful as it ignores fatali-ties that were caused by another driver’s impairment. If regulating the markets for alcohol results in reduced fatalities, we should expect to see a lower fatality rate in states in higher control categories.

The average fatality rate in states that NABCA defines as “controlled” is higher

(6.1) than that of License states (5.9). This finding is consistent with Rees (1997) who, in a study of Iowa, Ohio, and West Virginia over the period 1985 to 1995, con-cluded that deregulation would not lead to increased traffic-fatalities.18 Though the difference is not statistically significant, if we break the states down by degree of control, we obtain a strongly significant difference in fatality rates. States with Full Controls have the highest average fatality rate (7.4), states with Moderate Controls have the lowest (5.1), while License states (5.9) and states with Light Controls (5.6) are statistically identical.

A possible confound is the degree of enforcement of drunk driving laws. If states

with more regulations also spend more resources on monitoring and punishing drunk drivers, then we would expect to see reduced fatality rates simply as a result of the more stringent monitoring and enforcement. In summary, not only do control

Not only do con-trol states exhibit higher fatality rates than do li-cense states, Full Control states ex-hibit fatality rates which exceed fa-tality rates in li-cense states by 25%.

Figure 9. Average Number of Drivers (per 100,000 population) Involved in Fatal Crashes with BAC 0.01 or Higher, 1991-200619

‐

1.0 

2.0 

3.0 

4.0 

5.0 

6.0 

7.0 

8.0 

NABCA Full Control Moderate Control

Light Control License

Drivers In

volved

 in Fatal Crashes

12

COMMONWEALTH FOUNDATION | policy brief

Full Control states exhibit the high-

est rate of DUI arrests, and Li-

cense states ex-hibit the lowest,

though the differ-ences are not sta-

tistically signifi-cant

states exhibit higher fatality rates than do license states, Full Control states (states that regulation proponents would expect to have the lowest fatality rates) exhibit fatality rates that not only are the highest among the control categories, but which exceed fatality rates in license states by 25%.

DRIVING UNDER THE INFLUENCE (DUI) ARRESTS

It is possible that the fatality results shown above are influenced by the enforce-ment of DUI laws. For example, states with more DUI checkpoints and/or more DUI arrests will likely experience a lower fatality rate due to removing impaired drivers from the road before they can cause accidents. DUI arrests are shown in the follow-ing figure arranged by degree of regulation. We exclude Delaware from this analysis because its DUI arrest rate is atypical. Over the period 2002 through 2006, the aver-age number of DUI arrests per 1,000 population nationwide (excluding Delaware) was 4.04. Over the same period, Delaware’s DUI arrest rate was 0.3 per 1,000 popu-lation, or 1/16th that of the rest of the states.

Full Control states exhibit the highest rate of DUI arrests (4.6 arrests per 1,000

population), and License states exhibit the lowest (4.2), though the differences are not statistically significant. One way to adjust for differences in DUI enforcement is to examine the ratio of fatalities to DUI arrests. All other things being equal, the greater the problem a state has with impaired driving, the greater this metric will be. These results are shown in Figure 11.

Figure 10. Average DUI Arrests (per 1,000 population), 2002-200620

‐

0.5 

1.0 

1.5 

2.0 

2.5 

3.0 

3.5 

4.0 

4.5 

5.0 

NABCA Full Control Moderate Control

Light Control License

DUI A

rrests

13

policy brief | COMMONWEALTH FOUNDATION

After accounting for DUI enforcement, we see that License states exhibit the

same average fatalities-per-DUI rate as do Full Control states (1.8 alcohol-related fa-talities per DUI arrest). Moderate Control and Light Control states exhibit the rates that are statistically identical (1.1 and 1.3, respectively). These fatality rates are sig-nificantly lower than those of Full Control and License states. These results remain consistent with our previous results that Full Control states exhibit the greatest problem with alcohol-related highway fatalities. While evidence suggests that Light Controls (i.e., controlling the wholesale markets) may reduce DUI-enforcement ad-justed fatalities, increasing controls beyond this level can actually increase the fatal-ity rate to the same level observed in License states.

A counter argument is that there is a causal relationship underlying the data. It

is possible that states with a greater number of fatalities have the incentive to devote more resources to DUI enforcement. This does not explain, however, why the num-ber of fatalities-per-DUI would rise as we move from states with Light Controls to states with Full Controls.

Advocates claim that the social goals of reducing alcohol consump-tion, underage drinking, and al-cohol-related traf-fic deaths justify controlling whole-sale and retail alcohol markets. Evidence from 48 states over time shows no link be-tween market controls and these social goals.

Figure 11. Drivers Involved in Fatal Crashes with BAC 0.01 or Higher per 100 DUI Arrests, 2002-200621

‐

0.2 

0.4 

0.6 

0.8 

1.0 

1.2 

1.4 

1.6 

1.8 

2.0 

NABCA Full Moderate Light License

Drivers in

 Fatal Crashes per 100

 DUI A

rrests

14

COMMONWEALTH FOUNDATION | policy brief

CONCLUSION

Myriad examples at the state and federal levels illustrate that government is not good at running industries. Repeatedly, the private sector shows that it can provide higher quality goods and services at lower cost. However, arguments might be made for state control as a means of achieving some desired social outcome. In Pennsyl-vania’s case, advocates claim that the social goals of reducing alcohol consumption, underage drinking, and alcohol-related traffic deaths justify controlling wholesale and retail alcohol markets.

Evidence from 48 states over time shows no link between market controls and

these social goals. While alcohol consumption in license states is slightly higher than in controlled states, among controlled states, greater levels of control are actu-ally associated with increased consumption rates. Rates of underage drinking and underage binge drinking are virtually identical in license and control states. Simi-larly, there is no difference in alcohol-related traffic deaths in license versus control states. However, among control states, states with the most controls also exhibit the highest rates of alcohol-related traffic deaths—even after adjusting for differences in enforcement of DUI laws. In short, evidence suggests that control of alcohol markets does not imply control of alcohol consumption.

Evidence from 48 states over time

shows no link be-tween market

controls and these social

goals.

15

policy brief | COMMONWEALTH FOUNDATION

ENDNOTES  1. Segal, Geoffrey F. and Geoffrey S. Underwood, 2007. Divesting the Pennsylvania Liquor

Control Board. Reason Foundation. 2. Benefield, Nathan. Public Benefits from Private Liquor. Commonwealth Foundation. 3. Segal, Geoffrey F. and Geoffrey S. Underwood, 2007. Divesting the Pennsylvania Liquor

Control Board. Reason Foundation. 4. Iowa deregulated the retail sale of alcohol on July 1, 1987. Data source: National Insti-

tute on Alcohol Abuse and Alcoholism. 5. West Virginia deregulated the sale of alcohol on February 27, 1990. Data source: Na-

tional Institute on Alcohol Abuse and Alcoholism. 6. Data Source: National Institute on Alcohol Abuse and Alcoholism 7. Data Source: National Institute on Alcohol Abuse and Alcoholism 8. There is no statistically significant difference between Pennsylvania’s and Ohio’s aver-

age per-capita consumptions over the indicated period. Differences between Pennsyl-vania and each of the other border states are statistically significant.

9. West Virginia was regulated prior to 1990 and deregulated, at the retail level, from 1990 on.

10. The implied tax rate for Pennsylvania is estimated by the Distilled Spirits Council of the United States.

11. Montgomery County is the only county in Maryland that controls the sale of alcohol. For our purposes, we count Maryland here as deregulated.

12. Data Source: National Institute on Alcohol Abuse and Alcoholism 13. At the retail level, Idaho regulates all beverages that exceed 16% alcohol and Ohio regu-

lates all beverages that exceed 21% alcohol. We classify both of these states as Partial Retail and Wholesale regulated.

14. Data Source: The National Institute of Alcohol Abuse and Alcoholism 15. Data Source: The NSDUH Report, U.S. Department of Health and Human Services, Of-

fice of Applied Studies, Issue 13, 2006. 16. Binge drinking is the consumption of five or more units of alcohol on a single occasion. 17. The NSDUH Report, U.S. Department of Health and Human Services, Office of Applied

Studies, Issue 13, 2006. 18. Rees, Rebecca, 1997. Privatization of Liquor Stores: No Threat to Public Safety. Com-

monwealth Foundation. 19. Data Source: Fatality and Analysis Reporting System 20. Data Source: Sourcebook of Criminal Justice Statistics. Data is only readily available for

2002 through 2006 and does not include Florida. Data for Delaware are excluded. 21. Data Source: Fatality and Analysis Reporting System and Sourcebook of Criminal Justice

Statistics. Data is only readily available for 2002 through 2006 and does not include Florida.

COMMONWEALTH FOUNDATION | policy brief

COMMONWEALTH FOUNDATION for PUBLIC POLICY ALTERNATIVES | 225 State Street, Suite 302 | Harrisburg, PA 17101 717.671.1901 phone | 717.671.1905 fax | Info@CommonwealthFoundation.org | CommonwealthFoundation.org

ABOUT THE AUTHORS AND THE COMMONWEALTH FOUNDATION Mr. John Pulito is a Commonwealth Foundation Research Fellow at the Department of Economics and Quantitative Science at Duquesne University. Dr. Antony Davies (antony@antolin-davies.com) is Associ-ate Professor of Economics at Duquesne University and adjunct scholar with the Commonwealth Founda-tion. The Commonwealth Foundation is an independent, non-profit research and educational institute that develops and advances public policies based on the nation’s founding principles of limited constitutional government, economic freedom, and personal responsibility for one’s actions. More information is avail-able at www.CommonwealthFoundation.org

GUARANTEE OF QUALITY SCHOLARSHIP The Board of Directors and Staff of the Commonwealth Foundation is dedicated to providing the highest quality and most dependable research on public policy issues in the Keystone State. To this end, the Commonwealth Foundation guarantees that all statements of fact presented in our publications are verifi-able, and information attributed to other sources is accurately represented. Committed to providing Pennsylvanians with reliable information, the Commonwealth Foundation wel-comes critical review of its work. If the accuracy of our research is questioned and brought to the Founda-tion's attention with supporting evidence in writing, the Foundation will respond. If an error exists, the Commonwealth Foundation will issue an errata sheet that will accompany all subsequent distributions of the publication, which constitutes the complete and final remedy under this guarantee. For additional information or questions on this policy email info@commonwealthfoundation.org

1

STATE CO�TROL OF ALCOHOL SALES AS A MEA�S OF REDUCI�G

TRAFFIC FATALITIES: A PA�EL A�ALYSIS

John Pulito

Commonwealth Foundation Fellow

Duquesne University

Antony Davies

Associate Professor of Economics

Duquesne University

JEL classifications: H7, I18, K2, L5, R1

Key words: alcohol, regulation, policy, traffic fatalities, DUI, impaired, driving, underage,

privatization

We thank Lyn Cianflocco of the National Highway Traffic Safety Administration and Bill

Ponicki of the Prevention Research Center at UC Berkeley for help in obtaining data.

2

Today, nineteen states control alcohol sales at the wholesale and/or retail levels.

Previous research has compared alcohol-related fatality rates to the �ational Alcohol Beverage

Control Association’s binary classification of states as controlled versus privatized. What has

not been explored is the relationship between alcohol-related fatalities and the degree of control

states exercise over wholesale and retail alcohol markets. This paper presents a panel study of

alcohol-related fatality rates for 49 states from 1982 to 2002 that, after controlling for related

alcohol control policies, compares four measures of fatality rates to the degree of state control

versus privatization of alcohol markets. Whereas studies using binary classifications found no

relationship between fatality rates and privatization, we find a significant relationship that

varies based on the degree of privatization. We find that states with privatized alcohol markets

experience significantly lower alcohol-related traffic fatality rates as compared to states that

control alcohol markets, and that the magnitude of the difference varies according to the degree

of privatization and the definition of alcohol-related fatality.

3

I. Alcohol-Control Policies and Alcohol-Related Traffic Fatalities

The 21st

Amendment to the U.S. Constitution granted power to the individual states to

regulate, sell, and distribute alcohol. To date, nineteen states have opted to regulate alcohol

markets to varying degrees, some doing so as a means of generating revenue (Room 1987), and

others as a reflection of the state’s cultural values (Holder 1993).

Two states, West Virginia and Iowa, serve as case studies of the effect of privatizing

liquor sales on alcohol-related traffic fatalities since these states moved from controlled retail

and wholesale markets for alcohol to privatized markets. Macdonald (1986) compares Iowa’s

retail alcohol markets pre- and post-privatization and finds increased per-capita consumption of

wine post-privatization, but no change in consumption of beer or spirits. Fitzgerald and Mulford

(1988) look at state survey data and monthly alcohol sales for Iowa between July 1983 and

August 1987 in an attempt to explain the increase in wine consumption. They find that, in the

months following Iowa’s decision to privatize, state stores began to liquidate their wine

inventory at discount prices, creating a temporary surplus of wine and commensurately lower

prices on retail markets. Holder and Wagenaar (1990) also examine Iowa’s privatization and find

no change in beer consumption, a 13.7 percent decrease in wine consumption, and a 9.5 percent

increase in spirits consumption following privatization.

The National Alcohol Beverage Control Association (NABCA) classifies states that

impose any regulation on alcohol sales, whether at the wholesale or retail level, as “control,” and

states that impose no regulations as “license” (i.e., privatized). Rees (1997) looks at traffic

fatalities in Iowa, West Virginia, and Pennsylvania between 1985 and 1995, and finds no

significant difference in alcohol-related traffic fatalities between control and license states.

4

In a study of alcohol-control policies and the driving under the influence (DUI) fatality

rate, Kenkel (1993) hypothesized that the DUI fatality rate would be lower in control states than

in license states. He fits data from the 1985 Health Interview Survey to a consumer behavior

model but finds no evidence of a relationship between DUI fatality rates and whether a state is

control or license. Kenkel’s (1993) model is based on Becker’s (1969) which assumes that an

individual will only commit a DUI offense when the expected utility from the offense outweighs

his utility from any other activity. Becker finds that the number of criminal offenses committed

by any individual is a function of the probability of conviction, the punishment if convicted, and

other personal factors such as a willingness to commit illegal acts. As does Becker, Kenkel finds

that the discounted expected cost of drunk driving is predominantly a function of the probability

of arrest and conviction rather than a function of alcohol availability.

This finding is consistent with Chaloupka and Saffer (1989) who compare DUI fatality

rates before and after laws that allowed police to administer blood alcohol content (BAC) tests.

Chaloupka and Saffer find a statistical decrease in the number of DUI fatalities in states with

preliminary BAC test laws. Results from Chaloupka, Saffer, and Grossman (1993) suggest that,

if every state had a preliminary breath test law, average annual DUI fatalities would decrease by

approximately 3.4 percent, and that 18 to 20 year-olds would account for 20 percent of this

reduction.

In October 2000, the U.S. Congress passed the Department of Transportation’s 2001

Appropriations Act which mandated that states consider driving to be illegal if the driver’s BAC

is at least 0.08 g/dl (NHTSA, 2001b). By 2005, every state had enacted a 0.08 g/dl BAC law.

Zador et al. (2000), Dee (2001), and Lund et al. (2007) found that setting a BAC of 0.08 g/dl

resulted in a reduction in DUI fatality rates. Lund et al.’s (2007) results suggest that if all drivers

5

had a BAC less than 0.08 g/dl in 2005, nearly 9,000 deaths could have been avoided. Similarly,

Dee (2001) finds that 0.08 g/dl BAC laws have reduced traffic fatality rates by approximately

16.5 percent, and have reduced weekend and weekday fatality rates by 8.6 percent and 5.8

percent, respectively. Confounding this apparent effect, however, is the increasing prevalence of

airbags and primary seatbelt laws (McCartt et al., 2008).

Cook and Tauchen (1984) find that states that reduced the minimum drinking age from

21 to 18 experienced an 11.1 percent increase in the traffic fatality rate among 18 to 21 year-

olds. Dee and Evans (2001) hypothesized that this change in the minimum drinking age would

only shift driving fatalities from teens to adult drivers. Instead, they find that the movement

significantly decreased alcohol-related traffic fatalities for teen drivers. This finding is similar to

Dee (1990) who finds that changing the minimum drinking age from 18 to 21 substantially

reduced abusive teen drinking by 8 percent and teen driving fatalities by 9 percent. In a meta

study, Wagenaar and Toomey (2002) reviewed 241 empirical analyses on the minimum legal

drinking age and conclude that the body of research suggests an inverse relationship between the

mandatory legal drinking age and both per-capita alcohol consumption and the rate of traffic

accidents.

Kahane (2000) finds that seatbelt laws are 45 percent and 60 percent effective in reducing

traffic fatalities in passenger cars and light trucks, respectively. Chaloupka et al. (1993) and

Eisenberg (2003) find similar results. Sobel and Nesbit (2007) and Risa (1994) look at the

relationship between seat belt laws and traffic fatality rates and find that the safety gain from

wearing a seat belt is offset by the driver’s willingness to drive faster and more recklessly.

Other factors that have been identified as being related to traffic fatalities include the

price of unleaded gasoline (Sivak 2009; Grabowski and Morrisey 2004), speed limits and vehicle

6

miles traveled (Park et al. 2008), and rainfall, temperature, and terrain (Yakovlev and Inden

2009).

What has not been explored is the relationship between alcohol-related fatalities and the

degree of control states exercise over wholesale and retail alcohol markets. We perform this

analysis in a panel data setting. Whereas studies using NABCA’s binary classifications of control

vs. license found no relationship between control and alcohol-related fatalities, we find a

significant relationship that varies based on the degree of market control.

II. Data

We use a panel of the fifty U.S. States over the period 1982 through 2002. We divide

alcohol control variables into two categories: alcohol sales and use controls and alcohol market

controls. The former category includes laws that regulate the use of alcohol (e.g., DUI laws), the

sales or purchase of alcohol (e.g., keg registration laws, dram shop laws), or are tangentially

related to DUI fatality rates (e.g., seat belt laws). The latter category is the focus of our study and

reflects whether wholesale and retail markets are state run or privatized. Variable definitions and

summary statistics are shown in Table 1.

[Insert Table 1 here]

The Fatality Analysis Reporting System (FARS) defines an alcohol-impaired traffic

fatality as one in which a person, who is either a driver or a vehicle occupant, is killed within 30

days of a motor vehicle accident in which at least one driver had a BAC of at least 0.08 g/dl. An

alcohol-involved traffic fatality is one in which a person, who is either a driver, a vehicle

occupant, or a non-motorist, is killed within 30 days of a motor vehicle accident in which at least

one person (driver, passenger, or non-motorist) had a BAC of at least 0.01 g/dl (ANSI, 1996;

7

NHTSA, 2007). The definition of an alcohol-impaired fatality is narrower than the definition of

an alcohol-involved fatality in that the former requires a minimum BAC of 0.08 g/dl (versus 0.01

g/dl) and requires that the person with the BAC be a driver.

Prior to the National Minimum Legal Drinking Age Act of 1984, drinking ages varied

across states and by beverage type. Due to high multicollinearity among the minimum drinking

ages for the various forms of alcoholic beverages (beer, wine, spirits), we create a variable that

reflects the youngest age one can purchase and consume alcohol of any form. Due to high

multicollinearity between primary and secondary seat belt laws (Yakovlev and Inden 2009), the

mandatory seat belt variable reflects the first year in which any mandatory seat belt law (either

primary or secondary) was adopted. The BAC limit variable reflects the year in which a state

enacted a law making it illegal for an automobile operator to drive with a BAC greater than or

equal to 0.08 g/dl. This zero tolerance law represents the year in which a state enacted a BAC

limit of 0.02 g/dl or less for drivers under the age of 21. The tolerance of so called “zero

tolerance” laws vary by state. For example, for underage drivers, Texas mandates a BAC limit of

0.00 g/dl, while Pennsylvania mandates a BAC limit of 0.02 g/dl (NHTSA, 1997).

Keg registration laws require retailers to record the serial numbers of kegs along with

personal identifications of people who purchase the kegs. Preliminary breath test laws allow

police officers to administer a BAC test if the driver is suspected of being intoxicated. Open

container laws prohibit alcohol from being readily accessible by either the driver or passenger of

a moving vehicle. Dram shop laws establish liability for any establishment that serves alcohol to

an obviously intoxicated person who is subsequently involved in an alcohol-related traffic

fatality.

8

We base our classifications according to degree of alcohol market control on Pulito and

Davies (2009). These classifications are shown in Table 2.

[Insert Table 2 here]

In Maryland, only Montgomery County regulates alcohol sales. Due to the inability to

differentiate between the sale of alcohol in Montgomery County and sales in the rest of the state,

we drop Maryland from the data set. Table 3 shows the states according to their control

classifications. With the exception of Maryland, states not appearing in Table 3 are classified as

�o Control (i.e., privatized) states.

[Insert Table 3 here]

III. Results and Discussion

For each of four dependent variables, we estimate the following model in a fixed-effect

panel framework correcting for heteroskedasticity and first-order autocorrelation:1

3 8

1 1

j j

it j it j it i it

j j

Y C X hα γ β ε= =

= + + + +∑ ∑ (5)

where the hi are state-specific fixed effects and εit is a normal disturbance. We estimate this

model for each of four outcomes, Yit: the alcohol-involved traffic fatality rate, the alcohol-

impaired traffic fatality rate, the underage alcohol-involved traffic fatality rate, and the underage

alcohol-impaired traffic fatality rate. The variables j

itX represent the following alcohol sales and

use control policies: the minimum legal drinking age, mandatory seat belt law, 0.08 g/dl BAC

limit, zero tolerance law, keg registration law, preliminary breath test, open container law, and

1 The Hausman and Breusch-Pagan random effects test supports the choice for a fixed effects model. The residuals

from each of the regressions are stationary.

9

the dram shop law. The variables j

itC are dummies representing the alcohol market control

classification for state i in year t. Results are shown in Table 4.

[Insert Table 4 here]

In almost all cases, alcohol sales and use controls (minimum drinking age, mandatory

seat belt laws, etc.) show the expected relationship with alcohol involved and impaired fatalities:

more stringent policies are associated with reduced fatality rates. Whereas previous studies using

NABCA classifications of “control” versus “license” states have found no relationship between

alcohol market controls and DUI fatality rates, our results not only show a significant

relationship, but one that is the opposite of the policy intent. In moderate control states, fatality

rates are significantly greater than in states with privatized alcohol markets according to all

measures: underage, legal age, alcohol involved, and alcohol impaired. For light control states

versus no control states, fatality rates are significantly greater among those of legal age for

alcohol involved fatalities but not for the narrower definition of alcohol impaired fatalities. For

light control states, there is no significant difference in underage fatality rates regardless of the

fatality definition. These results suggest that light controls are associated with either no effect on

fatality rates or an increase in fatality rates, depending on the fatality definition. Meanwhile,

moderate controls are associated with greater fatality rates according to all fatality definitions.

One possible explanation (cf., Pulito and Davies 2009) is that controlling alcohol markets

increases the fixed cost of alcohol purchases. For example, state controlled retailers will tend to

keep less convenient hours and have fewer and less convenient retail locations. The

inconvenience increases the fixed cost of purchasing alcohol and so encourages consumers to

purchase more alcohol per trip than they would in the absence of the fixed cost. In purchasing

10

more alcohol per trip, consumers would then tend to have more alcohol on hand and so tend to

consume more.

These results contrast with those of heavy control states versus no control states. In heavy

control states and among those of legal age, alcohol involved fatality rates are greater while

alcohol impaired fatality rates are significantly lesser. One possible explanation is that policies in

heavy control states end up shifting risk from the drivers to passengers and non-motorists. For

example, “designated driver” campaigns may have the effect of encouraging people who

accompany a designated driver to drink more by inadvertently framing the DUI risk as being

solely a function of the driver’s BAC. Once in the car, unruly drunk passengers could present a

hazard to the driver’s concentration thereby partially negating the effect of the driver’s sobriety.

A fatality resulting when the driver was sober but the passengers were not would be recorded as

an alcohol involved fatality but not an alcohol impaired fatality. Similar, though weaker, results

are found among underage fatality rates wherein heavy control states exhibit the same fatality

rates as no control states for alcohol involved fatalities, but a significantly less rate for alcohol

impaired fatalities. Perhaps most disturbing is the finding that underage fatality rates (both

alcohol involved and alcohol impaired) are significantly higher among moderate control states

than among privatized states.

IV. Conclusion

This study utilizes a panel of 49 states from 1982 to 2002 in an attempt to measure the

relationship between privatization of alcohol sales and alcohol-related traffic fatalities.

Controlling for other alcohol-control policies, our results demonstrate that states with privatized

alcohol markets generally experience lower alcohol-related traffic fatalities. Whereas previous

11

studies using classifications of states as “controlled” versus “privatized” found no relationship

between fatality rates and privatization, we find a significant relationship that varies based on the

degree of privatization and the definition of alcohol-related fatality. We find that states with

privatized alcohol markets experience significantly lower alcohol-related traffic fatality rates

among consumers of legal age, and weaker though directionally similar results among underage

consumers. Our results suggest that concerns that privatization of state-controlled markets would

lead to an increase in alcohol-related traffic fatalities may be unfounded, and that the state-

control of alcohol markets itself may be a factor in increasing alcohol-related traffic fatalities.

12

V. References

American National Safety Council (ANSI). 1996. “Manual on Classification of Motor Vehicle

Traffic Accidents.” Sixth Edition (ANSI D16.1-1996).

Becker, G. 1969. “Crime and Punishment: An Economic Approach.” Journal of Political

Economics. 76(2): 169-217.

Chaloupka, F. and Saffer, H. 1989. “Breath Testing and Highway Fatality Rates.” Journal of

Applied Economics. 21(7): 901-912.

Chaloupka, F.; Saffer, H.; and Grossman, M. 1993. “Alcohol-Control Policies and Motor-

Vehicle Fatalities.” The Journal of Legal Studies. 22(1): 161-186.

Cook, P. and Tauchen, G. 1984. “The Effect of Minimum Drinking Age Legislation on Youthful

Auto Fatalities.” The Journal of Legal Studies. 13(1): 169-190.

Dee, T. 2001. “Does Setting Limits Save Lives? The Case of 0.08 BAC Laws.” Journal of Policy

Analysis and Management. 20(1): 111-128.

Dee, T. 1990. “State Alcohol Policies, Teen Drinking, and Traffic Fatalities.” Journal of Public

Economics. 72(2): 289-315.

Dee, T. and Evans, W. 2001. “Behavioral Policies and Teen Traffic Safety.” American Economic

Review. 91(2): 91-96.

Eisenberg, D. 2003. “Evaluating the Effectiveness of Policies Related to Drunk Driving.”

Journal of Policy Analysis and Management. 22(2): 249-274.

Fatality Analysis Reporting System (FARS). 2009. Fatalities and Fatality Rates by State, 1982 -

2006. Electronic Data. www-fars.nhtsa.dot.gov.

Fitzgerald, J.L. and Mulford, H.A. 1992. “Consequences of Increasing Alcohol Availability: the

Iowa experience revisited.” British Journal of Addiction. 87(2): 267-274.

Fitzgerald, J.L. and Mulford, H.A. 1988. “Consequences of Increasing Off-Premise Wine Outlets

in Iowa.” British Journal of Addiction. 83(11): 1271-1279.

Freeman, D. 2007. “Drunk Driving Legislation and Traffic Fatalities: New Evidence on BAC 08

Laws.” Contemporary Economic Policy. 25(3): 293-308.

Grabowski, D. and Morrisey, M. 2004. “Gasoline Prices and Motor Vehicle Fatalities.” Journal

of Policy Analysis and Management. 23(3): 575-593.

National Highway Traffic Safety Administration (NHTSA). 2000. “On DWI Laws in Other

Countries.” DOT HS 809 037. Washington, DC.

13

National Highway Traffic Safety Administration (NHTSA). 1997. “National Drunk and Drugged

Driving (3D) Prevention Program Planner.” Washington, DC.

National Institute on Alcohol Abuse and Alcoholism (NIAAA). 2009. “Keg Registration.”

apis.niaaa.nih.gov/index.asp. (accessed November 11, 2009).

Park, K.; Myeong, K.; and Stapleton, H. 2008. “Determinants of Traffic Fatalities in the U.S.”

Journal of Undergraduate Research. Vol. 8.

Peltzman, S. 1975. “The Effects of Automobile Safety Regulation.” Journal of Political

Economy. 83(4): 667-725.

Plümper, T and Troger, V. 2007. “Efficient Estimation of Time-Invariant and Rarely Changing

Variables in Finite Sample Panel Analyses with Unit Fixed Effects.” Political Analysis.

15(2): 124-139.

Ponicki, W. R. 2004. Statewide Availability Data System II: 1933 - 2003. �ational Institute on

Alcohol Abuse and Alcoholism. Research Center Grant P60-AA006282-23. Berkeley,

CA: Pacific Institute for Research and Evaluation, Prevention Research Center.

Pulito, J. and Davies, A. 2009. “Government Run Liquor Stores: The Social Impact of

Privatization.” The Commonwealth Foundation Policy Brief. 21(3): 1-16.

Rees, R. 1997. Privatization of Liquor Stores: No Threat to Public Safety. The Commonwealth

Foundation.

Risa, A. 1994. “Adverse Incentives from Improved Technology: Traffic Safety Regulation in

Norway.” Southern Economic Journal. 60(4): 844-857.

Room, R. 1987. “Alcohol Monopolies in the U.S.: Challenges and Opportunities.” Journal of

Public Health Policy. 8(4): 509-530.

Sivak, M. 2009. “Mechanisms Involved in the Recent Large Reduction in U.S. Road Fatalities.”

Injury Prevention. 15(3): 205-206.

Sobel, R. and Nesbit, T. 2007. “Automobile Safety Regulation and the Incentive to Drive

Recklessly: Evidence from NASCAR.” Southern Economic Journal. 74(1): 71-84.

U.S. Department of Health and Human Services. 2005. “Helping Patients Who Drink too Much.”

pubs.niaaa.nih.gov/publications/Practitioner/CliniciansGuide2005/guide.pdf (accessed

September 16, 2009).

Wagenaar, A., and Toomey, T. 2002. “Effects of Minimum Drinking Age Laws: Review and

Analyses of the Literature from 1960-2000.” Journal of Studies on Alcohol. Supplement

14: 206-225.

14

Yakovlev, P. and Inden, M. 2009. “Life in the Fast Lane: A Panel Data Analysis of U.S. Traffic

Fatalities.” Working Paper.

Young, D. and Likens, T. 2000. “Alcohol Regulation and auto fatalities.” International Review

of Law and Economics. 20(1): 107-126.

Zador, P.; Krawchuk, S; and Voas, R. 2000. "Relative Risk of Fatal Crash Involvement by BAC,

Age, and Gender," DOT HS 809 050. Washington, DC: National Highway Traffic Safety

Administration.

15

Table 1. Variable Descriptions and Summary Statistics

Variable Description Mean (Stdev)

Outcome Variables

Alcohol-involved traffic fatalities per 1,000 people among 21+ year-olds 0.09 (0.04)

Alcohol-involved traffic fatalities per 1,000 people among 15 to 20 year-olds 0.04 (0.02)

Alcohol-impaired traffic fatalities per 1,000 people among 21+ year-olds 0.08 (0.03)

Alcohol-impaired traffic fatalities per 1,000 people among 15 to 20 year-olds 0.03 (0.02)

Alcohol Market Controls

Dummy variable for the classification of a state as Heavy Control

(1=State is classified as Heavy Control, 0=Otherwise) 0.11 (0.31)

Dummy variable for the classification of a state as Moderate Control

(1=State is classified as Moderate Control, 0=Otherwise) 0.18 (0.39)

Dummy variable for the classification of a state as Light Control

(1=State is classified as Light Control, 0=Otherwise) 0.07 (0.26)

Alcohol Sales and Use Controls

Lowest minimum legal drinking age among all alcoholic beverage types 20.69 (0.79)

Dummy variable for Enforcement/Adoption of a mandatory seat belt law

(1=Enforcement/Adoption, 0=Otherwise) 0.66 (0.48)

Dummy variable for Enforcement/Adoption of a 0.08 g/dl BAC limit for

drivers (1=Enforcement/Adoption, 0=Otherwise) 0.19 (0.39)

Dummy variable for Enforcement/Adoption of a 0.02 g/dl BAC limit for

drivers under the age of 21 (1=Enforcement/Adoption, 0=Otherwise) 0.37 (0.48)

Dummy variable for Enforcement/Adoption of a keg registration law

(1=Enforcement/Adoption, 0=Otherwise) 0.17 (0.38)

Dummy variable for Enforcement/Adoption of a preliminary breath test law

(1=Enforcement/Adoption, 0=Otherwise) 0.51 (0.50)

Dummy variable for Enforcement/Adoption of a open container law

(1=Enforcement/Adoption, 0=Otherwise) 0.49 (0.50)

Dummy variable for Enforcement/Adoption of a dram shop law

(1=Enforcement/Adoption, 0=Otherwise) 0.83 (0.38)

Outcome data come from the Fatality Analysis Reporting System (2009). Data on alcohol sales and use controls come from

Ponicki (2009). Alcohol market control classifications come from Pulito and Davies (2009).

16

Table 2. Alcohol Market Control Classifications

Heavy Control

Sales of at least two types of alcohol (beer, wine, and liquor)

are controlled at the retail level, and sales of at least one type of

alcohol is controlled at the wholesale levels.

Moderate Control

Sales of one type of alcohol (beer, wine, or liquor) are

controlled at the retail level, and sales of at least one type of

alcohol are controlled at the wholesale level.*

Light Control No sales are controlled at the retail level, and the sale of at least

one type of alcohol is controlled at the wholesale level.

No Control Alcohol sales are not controlled. This is NABCA’s definition

of “license”.

* At the retail level, Idaho regulates all beverages that exceed 16 percent alcohol, and Ohio regulates all beverages that exceed 21

percent alcohol. Wine and Spirits have an average alcohol content of 12 percent and 40 percent, respectively (U.S. Department of

Health and Human Services, 2005). Based on these content levels, we classify Ohio and Idaho as moderate control states.

17

Table 3. States According to Alcohol Market Control Classification

Heavy Control Moderate Control Light Control

Maine Alabama Iowa*

Pennsylvania Idaho Michigan

Montana Iowa* Mississippi

Utah New Hampshire West Virginia*

North Carolina Wyoming

Ohio

Oregon

Vermont

Virginia

Washington

West Virginia*

* Iowa was Moderate Control prior to 1987 and Light Control after. West Virginia was

Moderate Control prior to 1990 and Light Control after.

18

Table 4. Alcohol Related Traffic Fatalities as a Function of Alcohol Policies

Legal Age Traffic Fatality Rates Underage Traffic Fatality Rates

Factor Alcohol Involved Alcohol Impaired Alcohol Involved Alcohol Impaired

Alcohol Market Controls

Heavy Control 0.026 *** -0.057 *** 0.017 -0.028 ***

(0.004) (0.015) (0.011) (0.007)

Moderate Control 0.063 *** 0.049 *** 0.021 ** 0.024 ***

(0.011) (0.015) (0.009) (0.007)

Light Control 0.024 *** -0.004 0.018 * 0.000

(0.006) (0.003) (0.010) (0.002)

Alcohol Sales and Use Controls

Minimum Drinking Age -0.004 *** -0.004 *** -0.005 *** -0.004 ***

(0.001) (0.001) (0.001) (0.001)

Mandatory Seat Belt -0.012 *** -0.008 *** -0.009 *** -0.007 ***

(0.002) (0.002) (0.002) (0.002)

BAC Limit -0.007 *** -0.006 ** -0.004 ** -0.003 **

(0.002) (0.002) (0.001) (0.001)

Zero Tolerance -0.011 *** -0.007 *** -0.009 *** -0.006 ***

(0.003) (0.002) (0.002) (0.002)

Keg Registration -0.009 *** -0.006 *** -0.009 *** -0.001

(0.002) (0.002) (0.002) (0.002)

Preliminary Breath Test -0.005 ** -0.010 *** -0.001 -0.004 ***

(0.002) (0.002) (0.002) (0.001)

Open Container -0.011 *** -0.007 *** -0.006 *** -0.001

(0.002) (0.002) (0.002) (0.001)

Dram Shop -0.010 *** -0.005 ** -0.008 *** -0.002

(0.003) (0.002) (0.002) (0.001)

Constant 0.165 *** 0.191 *** 0.131 *** 0.134 ***

(0.022) (0.022) (0.017) (0.017)

R2

0.865 0.700 0.760 0.542

Dependent variables: Alcohol-involved/impaired traffic fatalities and underage alcohol-involved/impaired traffic fatalities per

1,000 people. Significant levels: *** at 1%, ** at 5%, and * at 10%. Standard errors are reported in parentheses. Each regression

is estimated using 1,029 observations (49 states over 21 years).