health benefits of increases in alcohol and cigarette taxes · the anti-drinking campaign dates to...

British Journal of Addiction (1989) 84, 1193-1204

Health Benefits of Increases in Alcohol andCigarette Taxes

MICHAEL GROSSMAN

City University of New York Graduate School and National Bureau of Economic Research,Mercer Street, New York, NY 10003, U.S.A.

SummaryExcise taxes on alcohol and cigarettes imposed by the Federal government of the United States have been verystable since 1951. This paper summarizes research that shows that increased taxation, which results in higherprices, would discourage alcohol abuse and cigarette smoking. One striking finding is that a policy to raise theFederal excise tax on beer in line with the rate of inflation over the last three decades would cut motor vehiclefatalities of 18 to 20 year olds, many of which are alcohol-related, by about 15%, saving more than 1,000 livesper year. A second is that over 800,000 premature deaths in the cohort of Americans 12 years and older in 1984would be averted if the Federal excise tax on cigarettes were restored to its real value in 1951.

Introduction and BackgroundFor more than a decade, the Federal government ofthe United States and various state and localgovernments have been involved in campaigns todiscourage cigarette smoking and to reduce deathsfrom motor vehicle accidents by curtailing alcoholabuse. The anti-smoking campaign dates to theissuance of the First Surgeon General's Report onSmoking and Health in 1964. This campaign hasconsisted primarily of policies designed to increasepublic knowledge of the harmful effects of cigarettesmoking and to restrict advertising by cigarettemanufacturers. The major elements of the campaignhave been the Fairness Doctrine of the FederalCommunications Commission, which resulted in theairing of anti-smoking messages on radio andtelevision from July 1,1967 to January 1,1971, andthe Public Health Cigarette Act of 1970, whichbanned pro-smoking cigarette advertising on radioand television after January 1, 1971.

Other Federal government policies designed todiscourage smoking include the requirement, begin-ning in July 1966, of a health warning in all cigaretteadvertising and on every package and the strength-

ening of this warning at the time of the impositionof the advertising ban in 1971. In addition, theFederal Trade Commission began monitoring thetar and nicotine content of various brands ofcigarettes in 1967. Subsequently, the cigaretteindustry voluntarily agreed to include the FTCmeasurement in all advertising. Finally, Federalagencies have required the separation of smokersand non-smokers on vehicles in interstate passengertransportation, and many state and local govern-ments have required the provision of no-smokingareas in public places and in the workplace.

The anti-drinking campaign dates to the mid-1970's. One major element of this campaign hasbeen the upward trend in state minimum legal agesfor the purchase and consumption of alcoholicbeverages. This trend began with the increase in thelegal drinking age in Minnesota from 18 to 19 yearsof age in 1976, and an additional 27 states hadincreased their legal drinking age by the time thatCongress passed the Federal Uniform Drinking AgeAct of 1984. This legislation allows the Federalgovernment, through its control of Federal highway

1193

1194 Michael Grossman

funds, to intercede in a legislative area traditionallyreserved for the states. Five per cent of a state'sFederal highway construction fund allocation forthe fiscal year 1987 was withheld if that state'sminimum legal drinking age was below 21 onOctober 1, 1986, and 10% will be withheld fromeach future fiscal year allocation in which itsdrinking age is below 21. In July 1988 Wyomingbecame the 28th state to pass a law complying withthe act, and currently all 50 states and the District ofColumbia have a drinking age of 21.'

A second major element of the anti-drinkingcampaign is refiected by more severe penalties forthe conviction of drunken driving, the allocation ofadditional resources to apprehend drunk drivers,and an easing in the standards for apprehension andconviction. Between 1977 and 1984, the number ofstates with laws that permitted a law enforcementofficer to administer a preliminary breath test to adriver reasonably suspected of driving under theinfluence of alcohol before arrest rose from 13 to 22(Bureau of Justice Statistics, 1979; National High-way Traffic Safety Administration (NHTSA),1985). Ross (1984) reports that, in 1982, 378 billsrelating to drunken driving were introduced in 37states, and 38 of these became law in 25 states.According to NHTSA (1985), 100 new state lawspertaining to drunken driving were enacted betweenDecember 1983 and December 1984.

While the above policies are vehicles to discour-age cigarette smoking and alcohol abuse, increasedtaxation, which results in higher prices, is anotherpolicy that might significantly reduce these behav-iours. These are not simply hypothetical policyoptions. For example, although the Federal excisetax rate on cigarettes remained at 8 cents per packfrom November 1,1951 through to the end of 1982,there were several attempts to increase it during thelate 196O's and the 197O's because of the concernover the health effects of cigarette smoking. More-over, there is evidence that some states increasedtheir cigarette excise taxes as a result of the anti-smoking publicity that followed the issuance of theSurgeon General's Report in 1964. In 1965 therewere 23 state and local tax increases compared withno more than a dozen in any of the preceding 14years (Kellner, 1973). These state and local taxes

' The increases in the legal drinking age documented aboverepresent a dramatic reversal of the downward trend between 1970and 1975. In that period, 29 states lowered their drinking age toconform with a Federal shift in the voting age from 21 to 18 in1970.

continued to increase over time in many states,although in most cases the rates were increased toraise revenue rather than to discourage smoking perse. In recent years the rate of increase has abated inpart because of tax revenue losses due to smugglingof cigarettes from high-tax to low-tax states. Forinstance, states increased their cigarette excise taxrates 34 times between 1970 and 1975 but only 14times between 1975 and 1980 (Lewit, 1982).

Of course, the Federal excise tax rate on ciga-rettes was raised to 16 cents per pack effectiveJanuary 1,1983, as part of the Tax Equity and FiscalResponsibility Act of 1982. As in the case of therecent state excise tax increases, the Federal excisetax was raised to expand tax receipts rather than todiscourage smoking. This increase resulted in a taxthat was 100% smaller than the 32 cent rate requiredto restore the cigarette tax to its real value in 1951.Nevertheless, the real price of cigarettes (thenominal price divided by the Consumer PriceIndex) rose by 36% between 1981 and 1986 (Harris,1987) after falling by 13% between 1960 and 1980(Monthly Labor Review, various years; TobaccoInstitute, 1987).^

Increased taxation of alcoholic beverages has beenvirtually ignored by the Federal and state govern-ments in the anti-drinking campaign. Instead, theFederal excise tax rates on liquor (distilled spirits),beer, and wine remained constant in nominal termsbetween November 1, 1951 and the end of fiscal1985. During this period, the Federal governmenttaxed spirits at the rate of $10.50 per proof gallon(one gallon of 100-proof liquor, which is theequivalent of 50% alcohol by volume), beer atthe rate of $0.29 per gallon (approximately 4.5%alcohol by volume), and wine at the rate of $0.17per gallon (between 11.6% and 21% alcohol byvolume).'

Partly as a result of the stability of the Federalexcise taxes and the modest increases in state andlocal excise taxes, the real price of alcoholicbeverages has declined substantially over time.Between 1960 and 1980, the real price of spirits fellby 48%, the real price of beer fell by 27%, and the

' The legislation which raised the Federal excise tax from 8cents to 16 cents contained a clause which provided for theresumption of the old 8 cent rate at the end of fiscal 1985. Afterhalf-dozen temporary extensions, Congress made the 16 cent ratepermanent in 1986.

' The Federal excise tax rates on beer and wine are specified ona wine gallon of beer and a wine gallon of wine. A wine gallon is ameasure of liquid volume, regardless of alcohol content. Thestandard U.S. wine gallon contains 231 cubic inches at 60°F.

Health Benefits of Increases in Alcohol and Cigarette Taxes 1195

real price of wine fell by 20% (Cook, 1981). While28 states raised their drinking age between 1976 and1984, real alcoholic beverage prices continued tofall: 27% for liquor, 12% for beer, and 19% for wine(Bureau of Labor Statistics, various years). Thus, asargued by Cook & Tauchen (1982), if alcohol abuseis sensitive to price, a government policy ofdeclining real excise tax levels actually may beexacerbating this problem.

Indeed, a number of proposals have been made tocorrect the erosion in the real value of the Federalexcise tax rates on all forms of alcoholic beveragessince 1951 and to prevent future erosion by indexingtax rates to the rate of inflation or by converting toan ad valorem alcoholic beverage excise tax system(Moore & Gerstein, 1981; Luks, 1983; Cook, 1984;Harris, 1984; Becker, 1985; Jacobson & Albion1985). Of course, the Federal excise tax rate ondistilled spirits was raised from $10.50 per proofgallon to $12.50 effective October 1,1985, as part ofthe Deficit Reduction Act of 1984. But this increaseproduced a tax that was over 200% smaller than$42.00 rate required to restore the spirits tax to itsreal value in 1951. Moreover, the tax rates on beerand wine were not changed.

To the extent that smoking participation, thequantity of cigarettes consumed by smokers, andalcohol consumption, particularly excessive con-sumption, are inversely related to price, real in-creases in Federal cigarette and alcohol excise taxrates should reduce premature deaths due to smok-ing and fatal motor vehicle accidents regardless ofthe primary purpose of the tax hikes. Therefore, inthe next section of this paper, I summarize researchby my colleagues and me on the responsiveness ofyouth alcohol use and motor vehicle death rates tovariations in the price of alcohol (Grossman, Coate& Arluck, 1987; Saffer & Grossman, 1987a, 1987b;Coate & Grossman, 1988). To highlight the magni-tude of the price effects, the responsiveness of thesetwo outcomes to price is compared to their responsi-veness to increases in the legal drinking age. In thethird section of the paper, I review research by mycolleagues and me on the impacts of changes incigarette prices on age-specific smoking participa-tion rates and on the quantity of cigarettes con-sumed by smokers (Lewit, Coate & Grossman,1981; Lewit & Coate, 1982). I devote more atten-tion to the alcohol research than to the cigaretteresearch because the former is much more recentand because the implications of the cigarette studiesfor the 1983 increase in the Federal excise tax rateon cigarettes have been considered by Harris (1982,

1987), Lewit (1985), Warner (1986) and by EugeneLewit in this journal. In the final section of thepaper, I qualify some of the conclusions reached inthe previous sections and consider some implica-tions of the rational model of addictive behaviourdeveloped by Becker & Murphy (1988) for theevaluation of alcohol and cigarette excise taxincreases.

It is particularly important to focus on teenagersand young adults in the context of the anti-drinkingcampaign because motor vehicle accident mortalityis the leading cause of death of persons under theage of 35, and alcohol is involved in over half thesefatal accidents. In 1984 persons under the age of 25accounted for 20% of all licensed drivers, but 35%of all drivers involved in fatal accidents (NHTSA1986). These figures are even more dramatic thanthey appear because members of the young drivergroup do not drive nearly as much as older drivers(Voas & Moulden, 1980). It is also important tofocus on youths because alcohol abuse in adoles-cence appears to be associated with alcohol abuse inadult life (Blane & Hewitt, 1977; Rachal et al.,1980). Thus, policies to prevent the onset of thisbehaviour by adolescents might be the most effec-tive means to reduce it in all segments of thepopulation. Similar comments apply to cigarettesmoking since it is an habitual behaviour that beginsearly in life. Moreover, age at onset of smoking isnegatively correlated with the amount smoked andthe incidence of negative health effects (Hammond,1966; Ippolito, Murphy & Sant 1979).

Alcoholic Beverage Prices, Legal DrinkingAges, and Youth Alcohol AbuseIn a project funded by the National Institute onAlcohol Abuse and Alcoholism, Douglas Coate,Henry Saffer, Gregory Arluck, and I present thefirst set of estimates of the responsiveness of youthalcohol use and motor vehicle death rates tovariations in the price of alcohol (Grossman, Coate& Arluck 1987; Saffer & Grossman, 1987a, 1987b;Coate & Grossman, 1988). In addition, we examinethe sensitivity of these two outcome measures toincreases in the legal drinking age. Research on theresponsiveness of youth motor vehicle deaths andalcohol use to the price of alcoholic beverages isparticularly timely in light of the proposals to raiseFederal excise tax rates on alcoholic beveragesdiscussed in the previous section. Moreover, al-though beer is the drink of choice among youths whodrink alcoholic beverages, the alcohol in liquor is


taxed three times as heavily as the alcohol in beer.This has led to suggestions to equalize the tax rateson the alcohol in all forms of alcoholic beverages byraising the tax on beer (Harris, 1984; Jacobson &Albion, 1985)."

Research on the sensitivity of the outcomes atissue to legal drinking ages is valuable given thevolatility in state minimum drinking ages in the197O's and 198O's and the adverse reaction toFederal uniform drinking age legislation. Originally,the penalties imposed on states with a drinking agebelow 21 by the Federal Uniform Drinking Age Actof 1984 were scheduled to expire at the end of fiscal1988. In response to this provision, Texas andNebraska adopted laws that called for a revocationof the 21 drinking age as soon as the legislationexpired. To counteract these laws, the Federallegislation was made permanent in 1986. SouthDakota challenged the constitutionality of the 1984Federal Uniform Drinking Age Act in a suit beforethe Supreme Court, which was supported by 8additional states. In June 1987, the Court ruledagainst South Dakota.

One of the basic aims of our research is tocompare the impact of a uniform minimum age of21 for the purchase of alcohol in all states on youthalcohol use and motor vehicle mortality with that ofone or more of the policies to raise the Federalexcise tax rates on alcoholic beverages describedabove. Of course, an effectively enforced prohibi-tion of alcohol consumption in this age group clearlyshould have a larger impact on their consumptionand fatalities than an increase in excise tax rates.This issue is not clear cut at the empirical level onlybecause of the problem of evasion. Underage youthscan obtain alcohol from their older siblings orfriends. In addition, they can purchase fake identifi-cation cards or buy alcohol in stores that do notbother to demand proof of age. This type of evasionsimply is not possible with an excise tax hike, so thatthe responsiveness of youths to the price of alcoholdetermines the change in consumption and thereforethe motor vehicle death rate.

Even if adult consumers of alcoholic beveragesare relatively unresponsive to price, this need not be

"* Under the Federal excise tax on liquor of $10.50 per gallon ofliquor (50% alcohol by volume) in effect prior to October 1,1985,one gallon of alcohol in liquor was taxed at the rate of $21. Sincethe Federal excise tax on beer is $0.29 per gallon and since onegallon of beer contains 4.5% alcohol by volume, the tax rate on onegallon of alcohol in beer is $6.44. The alcohol in liquor is taxedfifteen times as heavily as the alcohol in wine, and the proposalsmentioned above also contain provisions to correct this distortion.

the case for youths. Given the habitual nature ofalcohol abuse, adult users, who almost always willhave been users for longer periods of time thanyouths, may be less sensitive to price than youths. Inaddition, the fraction of disposable income that ayouthful drinker spends on alcohol probably ex-ceeds the corresponding fraction of an adult drinker.It is well known that the uncompensated (moneyincome-constant) price elasticity of a good rises inabsolute value as the fraction of income spent onthat good rises. Finally, bandwagon or peer effectsare much more important in the case of youthdrinking than in the case of adult drinking. That is,youths are more likely to drink if their peers alsodrink (Blane & Hewitt, 1977; Rachal et al., 1980).As shown by Leibenstein (1950) and by Lewit,Coate & Grossman (1981), the presence ofbandwagon or peer effects increases the priceelasticity of demand.^

The research on youth alcohol use employs twodata sets: the first National Health and NutritionExamination Survey (NHANES I), conducted bythe National Center for Health Statistics (NCHS)between May 1971 and June 1974, and the secondNational Health and Nutrition Examinations Sur-vey (NHANES II), conducted by NCHS betweenFebruary 1976 and February 1980. The research onyouth motor vehicle accident mortality is based on atime series of state cross sections for the period from1975 through 1981. All aspects of the researchcapitalize on substantial differences in legal drink-ing ages among states and on substantial differencesin alcoholic beverage prices among states primarilydue to differences in state excise tax rates on thesebeverages. We concentrate on beer prices and beerexcise tax rates in the research because beer is themost popular alcoholic beverage among youths.State beer excise taxes on a case of 24- 12 oz cansduring the period at issue ranged from a low of 4.5cents in Wyoming to a high of $2.28 in Georgia.

The project contains estimates of demand func-

' In a penetrating economic analysis of rational addiction overthe life cycle, Becker & Murphy (1988) show that the effect ofhabit formation or peer pressure on price responsiveness dependson whether the price variation is permanent or temporary, whetherthe magnitude of the effect if measured by the slope or theelasticity, and whether the outcome pertains to the probability ofconsuming the addictive good or to consumption given participa-tion. In certain cases, adults can be more responsive to price thanyouths in their model, while in other cases the reverse holds. Becker& Murphy's model also questions the conventional wisdomexpressed above that adult price elasticities of demand foraddictive goods should be relatively small. Their work is discussedin more detail in the fourth section.


tions for youth alcohol use in NHANES I (Gross-man, Coate & Arluck, 1987) and NHANES II(Coate & Grossman, 1988) and logit motor vehicleaccident mortality regessions (Saffer & Grossman,1987a, 1987b). In the demand functions we focus onalcohol use by youths aged 16-21 and add alcoholicbeverage prices and legal drinking ages to theNHANES surveys based on a given youth's place ofresidence. In the motor vehicle accident mortalityresearch, logit regressions are estimated for threeage groups: youths aged 15-17, 18-20, and 21-24.The real beer price (the nominal price divided bythe annual Consumer Price Index) is employed as aregressor in the beer demand functions.' Becauseprice data were not available in all years in theperiod from 1975-81 (Saffer & Grossman, 1987a),the cost of beer is given by the sum of the Federaland state excise tax rates on beer divided by theannual CPI for the U.S. as a whole in the mortalityregressions. State-specific beer prices and excise taxrates are highly correlated in years in which bothwere available.'

Youths who reside in a state with a high legaldrinking age may be able to purchase alcohol in aborder state with a lower legal drinking age. To dealwith this phenomenon in the alcohol use studies, wecreate a dichotomous variable that equals one foryouths who live within 20 miles of a state with alower legal drinking age. With the own-state legaldrinking age held constant, the coefficient of theborder age variable in the demand functions shouldbe positive. To deal with the border phenomenon(out-of-state purchases) in the motor vehicle mor-tality research, we create a variable that equals thedifference between the own-state drinking age andthe border-state age (if the difference is positive)multiplied by the fraction of the population of thestate in question who live in counties near theborder state. With the resident-state drinking ageheld constant, an increase in the border variablerefiects a reduction in the border-state drinking age

' In the NHANES I study, beer prices were taken from a specialone-time survey by the Bureau of Labor Statistics (1973). In theNHANES II study, they were taken from unpublished data on theprice of a single leading brand of medium-priced, nationally soldbeer obtained by Stanley Ornstein (see Ornstein and Hanssens1985). The specific brand is confidential. Prices are reported in twounidentified major markets in each state in January and July of1976, 1977, and 1978 and in January of 1979.

' State excise tax rates on wine and liquor are poor proxies forthe prices of wine and liquor in control (monopoly) states becausesuch states derive most of their revenue from the sale of wine andliquor from the price mark-ups rather than from the excise taxes.This comment does not apply to state excise tax rates on beerbecause beer is sold privately in monopoly states.

or an increase in the fraction of the population wholive in counties near the border state, both of whichshould cause the motor vehicle fatality rate to fall.*

In all aspects of the research, efforts are made tocontrol for (hold constant) determinants of alcoholuse and fatal motor vehicle accidents other thanalcoholic beverage prices or taxes and legal drinkingages. All demand functions include as regressors theyouth's age, a dichotomous variable that identifiesblacks, a dichotomous variable that identifies fe-males, and real family income (money familyincome divided by the CPI). The motor vehicleregressions include real per capita income, vehiclemiles traveled per licensed driver, the fraction of thepopulation aged 15-24 who are licensed drivers, anda dichotomous variable that identifies states thatrequire inspection of motor vehicles every year.

In some specifications of the alcohol use andmotor vehicle mortality regressions, we try to takeaccount of the potential role of 'drinking sentiment'in the endogenous determination of alcoholic bever-age prices or taxes, legal drinking ages, and alcoholconsumption. This is accomplished by including thefractions of the population who are Mormons,Southern Baptists, Catholics, and Protestants (ex-cluding Southern Baptists and Mormons) and thefraction of the population who reside in 'wet'counties (counties that permit the sale of alcoholicbeverages) as regressors in some specifications.'Drinking sentiment refers to cultural and tastevariables that may either encourage or discouragealcohol consumption. For example, anti-drinkingsentiment should be relatively widespread in statesin which those religious groups that oppose the useof alcohol, such as Mormons and Southern Baptists,are prevalent. Anti-drinking sentiment also shouldbe an important force in states in which a higher-than-average fraction of the population reside in

' The border age variable described above is modified in cases inwhich there is more than one border state, each of which has alower drinking age than the own state. We do not attempt tocontrol for incentives to purchase alcohol in border areas withlower prices, which tends to understate the price effect in absolutevalue in the demand function. This is a more important problem inaggregate state cross-sectional alcohol demand studies whichemploy a sales variable as opposed to actual consumption. Here theprice effect tends to be overestimated if one does not control forincentives to purchase alcohol in border areas with lower prices.Lewit, Coate & Grossman (1981); Lewit & Coate (1982); andBecker, Grossman & Murphy (1987) discuss this issue in detail inthe context of the estimation of cigarette demand functions at themicro and aggregate levels. For discussions in the context of thedemand for alcohol, see Wales (1968); Cook & Tauchen (1982);Ornstein & Hanssens (1985); Grossman, Coate & Arluck (1987);and Nelson (1988).


'dry' counties (counties that prohibit the sale ofalcoholic beverages). These states may enact highalcoholic beverage excise tax rates as part of thepolitical process. In this situation, the tax coeffici-ents that emerge from regressions that omit drinkingsentiment overstate in absolute value the trueparameters. On the other hand, states in which pro-drinking sentiment is widespread (anti-drinkingsentiment is weak) and alcohol consumption is largemay enact high excise tax rates because the taxationof alcohol is an attractive source of revenue. In thiscase, the tax effects are understated if drinkingsentiment is excluded from the regressions. Similarcomments can be made with respect to drinking ageeffects that do not control for drinking sentimentand with respect to the estimated price and drinkingage effects in the demand functions.

We find that the use of alcohol by youths isinversely related to the prices of alcoholic beveragesand to the legal drinking age in both NHANES Iand NHANES II. The beverage-specific price andlegal drinking age are particularly important deter-minants of beer consumption. This is a key resultbecause beer is the drink of choice among youthswho consume alcoholic beverages. The negative andstatistically significant price and legal drinking ageeffects are by no means limited to reductions in thefraction of youths who consume beer infrequently(less than once a week). Instead, the fractions ofyouths who consume beer fairly frequently (1-3times a week) and frequently (4-7 times a week)fall more in absolute or percentage terms than thefraction of infrequent drinkers when price or thelegal drinking age rises. Along the same lines, thefractions of fairly heavy (3-5 cans on a typicaldrinking day) and heavy (6 or more cans on atypical drinking day) youthful beer drinkers declinemore in absolute or percentage terms than thefraction of light (1-2 cans on a typical drinking day)

' The religion variables are state-specific in the motor vehiclefatality regressions and county-specific in the alcohol consumptiondemand functions, where they pertain to the youth's county ofresidence. The fraction of the population who reside in wetcounties is state-specific and is not used in the demand functions.Saffer & Grossman (1987b) develop an alternative methodology tocontrol for drinking sentiment in the mortality equation. Weconstruct and estimate a simultaneous equations model in which anunobserved variable measuring the pressure to pass a 21-year-oldminimum drinking age law, for example, depends on the mortalityrate in the absence of the law. Although our results are somewhatsensitive to alternative specifications, they suggest that the tax andlegal drinking age effects obtained by more conventional methodsare conservative lower bound estimates. Moreover, the relativeranking of the two estimates is not affected by biases associatedwith endogeneity.

drinkers in response to price or drinking ageincreases. These are striking findings because fre-quent, fairly frequent, heavy, and fairly heavydrinkers are likely to be responsible for a largepercentage of youth motor vehicle accidents anddeaths.

We also find negative and statistically significantreal beer tax effects in the motor vehicle accidentmortality regressions for youths aged 15-17,18-20,and 21-24. Negative and significant drinking ageeffects are obtained for youths aged 18-20. More-over, positive and significant coefficients of theborder variable are obtained for this cohort.

With regard to the magnitudes of the beerconsumption effects at issue in NHANES II (themost recent sample), a Federal policy that simulta-neously taxes the alcohol in beer and liquor at thesame rates and offsets the erosion in the real beer taxsince 1951 would have reduced the number ofyouths who drink beer frequently (approximately11% of all youths) by 32% during the period ofNHANES II and would have reduced the number offairly frequent beer drinkers (approximately 28% ofall youths) by 24%. The enactment of a minimumuniform drinking age of 21 in all states would havereduced the number of frequent drinkers by 28%and the number of fairly frequent drinkers by 11%.

With regard to motor vehicle accident mortality,the drinking age policy would have reduced thenumber of 18 to 20-year-olds killed in motor vehiclecrashes by 8% in the period 1975-81. A policy thatfixed the Federal beer tax in real terms since 1951would have reduced the number of lives lost in fatalcrashes by 15%, while a policy that taxed the alcoholin beer at the same rate as the alcohol in liquorwould have lowered the number of lives lost by 21%.A combination of the two tax policies would havecaused a 54% decline in the number of youths killed.

The preceding figures suggest that, if reductionsin youth alcohol consumption and motor vehicleaccident deaths are desired, both a uniform drinkingage of 21 and an increase in the Federal excise taxrate on beer are effective policies to accomplish thisgoal. They also suggest that the tax policy may bemore potent than the drinking age policy. Indeed,according to our computations, the lives of 1,022youths aged 18 to 20 would have been saved in atypical year in the 1975-81 period if the Federalexcise tax on beer had been indexed to the rate ofinflation since 1951. On the other hand, the lives of555 youths aged 18 to 20 per year would have beensaved if the drinking age had been 21 in all states ofthe U.S.


Micro Cigarette Demand StudiesIn two related studies Eugene Lewit, Douglas Coateand I have obtained age-specific cigarette priceelasticities for smoking participation and for thequantity smoked conditional on smoking (Lewit,Coate and Grossman 1981; Lewit & Coate 1982).The first study is limited to youths aged 12-17. Thesecond study considers persons aged 20-74 andcontains demand functions for three separate agegroups (20-25 years, 26-35 years, and over 35years). Both studies use information on the city andstate of residence of each respondent in a givensurvey to add a measure of the price of cigarettes tothat survey. The resulting series incorporates varia-tions in state and municipal excise and retail salestax rates on cigarettes. It is important to emphasizethat interstate differentials in cigarette prices aresubstantial; retail cigarette prices are approximately50% higher in high-tax states than in low-tax states.

The Lewit-Coate-Grossman (1981) study isbased on Cycle III of the U.S. Health ExaminationSurvey. This is a national sample of 6,768 youthsthat was conducted between March 1966 and March1970 by the National Center for Health Statistics(NCHS). The Lewit-Coate (1982) study is basedon the 1976 U.S. National Health Interview Surveyconducted by NCHS. The sample is comprised of28,033 individuals between the ages of 20 and74.

Based on the estimated demand functions, thesmoking participation price elasticity equals —1.20for teenagers, —0.74 for 20-25 year olds, —0.44 for26-35 year olds, and —0.15 for persons above theage of 35. The corresponding quantity smokedconditional on smoking price elasticities are —0.25for teenagers, —0.20 for the youngest group of adultsmokers, —0.04 for the middle age group, and—0.15 for the oldest age group of adult smokers.

The principal message of the above findings isthat an increase in the Federal excise tax rate oncigarettes is a potent policy to curtail smoking. Thisis because teenagers are more responsive to changesin the price of cigarettes than adults and because theprice elasticity of smoking participation is muchlarger than the price elasticity of the quantitysmoked by smokers. These factors mean thattendencies for smokers to compensate for reduc-tions in the number of cigarettes consumed byswitching to higher tar and nicotine brands, inhalingmore deeply, or reducing idle burn can be ignored inevaluating the impact of excise tax changes. Moreimportantly, the large teenage smoking participationelasticity implies that excise tax increases are very

effective tools to prevent the onset of an habitualbehaviour.

Since the smoking participation rate of all agegroups (the aggregate rate) is dominated by theadult rate, the short-run effect of an increase in theFederal excise tax rate would be modest. The long-run impacts of a tax hike should, however, beconsiderably more substantial. Given the evidencethat individuals are unlikely to initiate smokingafter age 21 (Centers for Disease Control andNational Cancer Institute 1976), it is quite possiblethat the cohort of young persons who do not begin tosmoke as a result of the tax increase would neverbecome regular smokers. If the tax increase ismaintained in real terms, it would continue todiscourage smoking participation by successivegenerations of youths. Thus, it would graduallyimpact the smoking levels of older age groups as thesmoking-discouraged cohorts move through the agespectrum. As a consequence, over a period ofseveral decades, aggregate smoking and its associ-ated detrimental health effects would decline sub-stantially.

Harris (1982, 1987) and Warner (1986) haveused the smoking participation price elasticitiesreported by Lewit, Coate & Grossman to forecastthe impacts of the doubling of the Federal excise taxrate from 8 cents to 16 cents in 1983. They predict adecline in the number of adult smokers that rangesbetween 1.4 million and 2.0 million and a decline inthe number of teenage smokers that ranges between400,000 and 700,000.'" Warner has also evaluated apolicy to restore the excise tax to its real value in1951 by raising the nominal rate from 16 cents to 32cents. This policy would discourage approximately800,000 teenagers from starting to smoke and wouldinduce roughly 2.7 million adult smokers to quit.

Both Warner (1986) and Harris (1987) providecrude estimates of the health benefits of the abovereductions in smoking. Warner assumes that onelifelong smoker out of every four dies prematurelyof smoking-related illness. He calculates that the 8cent tax increase in 1982 will ultimately avert480,000 premature deaths in the cohort ofAmericans 12 years of age and older in 1984.Moreover, this number would rise to 860,000 if thetax rate were set at 32 cents. Harris assumes that 9%of all smokers will not survive to age 65 as a resultof their smoking. This suggests that an additional

'" Due to data limitations and the temporary nature of theoriginal 1982 legislation (see note 2), these predictions have not yetbeen compared with actual trends in the number of smokers.


100,000 persons will live to this age as a result of theprice-induced decline in smoking participation dueto the 1982 tax hike.

Lewit (1985) and Harris (1987) have examinedthe actual decline in per capita cigarette consump-tion following the 1982 tax increase. The real priceof cigarettes rose by 26% between November 1,1981 and November 1, 1984. Based on a priceelasticity of —0.47 for all age groups, per capitaconsumption should have declined by roughly 12%.In fact, the actual decline during this period rangedbetween 11-12%.

Qualifications and Future ResearchThe empirical evidence summarized in the secondand third sections suggests that substantial healthbenefits may accrue to increases in Federal excisetaxes on alcohol and cigarettes. The conclusionsreached in those sections must, however, be inter-preted with caution. For example, consider thecomparison between the legal drinking age policyand an increase in the Federal excise tax on beer.My colleagues and I have not provided enoughevidence to justify the approximately eightfold(thirteenfold based on the 1984 CPI) increase in theFederal excise tax on beer that is implicit in the mostcomprehensive tax policy that we consider (the onethat simultaneously taxes the alcohol in beer andliquor at the same rates and offsets the erosion in thereal beer tax since 1951). Excise tax hikes imposewelfare costs on all segments of the population,while a drinking age policy is targeted at the groupin the population that accounts for a disproportion-ate share of motor vehicle accidents and deaths. Onthe other hand, the enforcement and administrativecosts associated with a uniform minimum drinkingage of 21 may exceed those associated with the taxpolicy. Moreover, our results indicate that an excisetax increase lowers death rates of youths betweenthe ages of 15 and 17 and between the ages of 21 and24. These benefits do not accompany a rise in thedrinking age. In addition, the tax policy may reducefatal crashes involving persons beyond the age of 24.Of course, a substantial tax hike may stimulate thedemand for illegally produced beer, suggesting thatwe have over-estimated the effect of an eightfoldincrease in the Federal excise tax on beer unless thelegal and illegal prices are the same.

Another consideration is that Becker (1968) hasshown that the optimal way for a society to deteroffences is via a system of severe and fairly certain

punishments. In the case of drunk driving, thesemight take the form of loss of driving privileges forlong periods of time, mandatory community service,enrollment in alcohol rehabilitation programs, andprison sentences for repeat offenders. Of course,youthful drunk drivers may respond to an increasein the penalty for this offence only if the probabili-ties of apprehension and conviction are non-trivial.If substantial resources must be allocated to raisingthese probabilities, the excise tax policy may bepreferable to or complementary with a system ofsevere penalties. Moreover, severe and certainpunishments for drunk driving do not address theproblems caused by the link between youth alcoholabuse and adult alcohol abuse.

A third consideration is that there have been noprevious published studies of the effects of beertaxes on youth motor vehicle fatalities. Cook(1981), however, finds that states that raised theirexcise tax rates on liquor between 1960 and 1974experienced below-average increases or above-aver-age reductions in motor vehicle deaths of persons ofall ages, relative to states that did not increase theirtax rates. He also estimates an elasticity of the motorvehicle death rate with respect to the price of liquorof —0.7. On the other hand, Saffer & Grossman(1987a) report an elasticity of the motor vehicledeath rate with respect to the price of beer of —0.7for 15-17-year-olds and - 1 . 3 for 18-20-year-oldsand 21-24-year-olds."

The preceding figures suggest that Saffer &Grossman's estimates are reasonable. But Cook'sresearch is based on the death rate of persons of allages and employs relatively old data. Therefore, itwould be extremely useful to replicate Saffer &Grossman's results with more recent data. This isone of the aims of a new project by Henry Saffer,Frank Chaloupka, and me (Grossman, Saffer &Chaloupka in progress). This research employsmotor vehicle death rates by age, state, and year forthe period from 1975 to 1988. The use of age-specific motor vehicle accident mortality rates willenable us to detect differential responses to alco-holic beverage excise taxes or prices by persons ofdifferent ages. This information will enable us toevaluate more completely alternative policies to

' ' The elasticities presented by Saffer & Grossman assume thatthe beer industry is competitive and has an infinitely elastic supplycurve, so that a tax increase is fully passed on to consumers. Theyalso assume that the sum of the Federal and state excise tax on acase of beer accounted for 13% of the retail price of beer on averagein the period 1975-81. Cook's computation is based on a liquorstate excise tax share of approximately 7%.


reduce fatal motor vehicle crashes by moderatingalcohol abuse.

Very preliminary results from the Grossman-Saffer-Chaloupka project are contained in a paperby Saffer & Chaloupka (1989). Using a time seriesof state cross sections for the period from 1980 to1985, Saffer and Chaloupka focus on four outcomes:the motor vehicle accident mortality rate of personsof all ages; the night-time (between 2400 and 0400)fatality rate of persons of all ages, many of which arealcohol-related; and the same twomortality rates ofyouths between the ages of 15 and 24. The real beertax has a negative and statistically significant effecton each of these four outcomes.

Levy & Reinhart (1988) also have replicatedSaffer & Grossman's beer tax results in a time seriesof state cross sections for the years 1981 through1984. Their dependent variable is the motor vehicleaccident mortality rate of youths aged 16 to 25. Inaddition. Levy & Reinhart report negative andsignificant spirits and wine, but not beer, 'price'coefficients in a multivariate analysis of the deathrate of persons of all ages. They attribute theinsignificant beer effect to multicoUinearity amongthe three prices. Moreover, beer is not as popularamong adults as it is among youths.

Although Levy & Reinhart's results are valuable,more information is required concerning theirbeverage-specific price measures. These measuresare given by beverage-specific total revenue pergallon from combined state and local collections.Two potential problems with these measures arethat they include net profits in monopoly states andlicense fees in all states. Net profit per gallondepends in part on the cost of operation per gallonwhich in turn may be related to the number ofgallons sold. Since the motor vehicle accidentmortality rate also depends on alcohol consumption,the wine and liquor price effects may refiect in partcausality from the death rate to revenue per gallon.Moreover, the sign of the relationship between priceand net profit per gallon is ambiguous.'^ License

'^ Consider a monopolist who produces subject to constantaverage cost. If he faces a demand function with a constant priceelasticity, price and profit per unit output (the difference betweenprice and average cost) will rise in response to an exogenousincrease in average cost. On the other hand, if the monopolist'sdemand function has a constant slope, price will rise but profit perunit output will fall as average cost rises. If the demand functionhas neither a constant slope nor a constant elasticity, the effect ofan exogenous increase in average cost on profit per unit output isambiguous.

fees are one-time, rather than annual, payments, andit is misleading to add them to annual revenuesources such as excise taxes and net profits. A finaldifficulty is that these variables are subject tomeasurement error if total revenue and/or quantityis measured with error. This comment does notemploy to the variable employed by Saffer &Grossman which pertains to the statutory stateexcise tax rate on a case of beer.

A final reason for treating the conclusions in theprevious sections with caution is that the studiessummarized in this paper do not incorporate insightsprovided by economic models of addictive behav-iour. The conventional wisdom, which I expressedin the first two sections, is that the addictive natureof alcohol abuse and cigarette smoking may mutethe role of price for persons other than adults. Avariety of empirical evidence contradicts this view.The best example is a study by Cook & Tauchen(1982). They examine variations in death rates fromcirrhosis of the liver (a standard measure ofexcessive alcohol use) among states of the U.S. aswell as variations in per capita consumption ofdistilled spirits." They find that the state excise taxon distilled spirits has a negative and statisticallysignificant effect on the cirrhosis death rate. More-over, a $1 increase in the state excise tax lowers thedeath rate by almost the same percentage as itlowers per capita consumption (5.4% versus 7.2%).Cook & Tauchen conclude that " . . . liquor con-sumption, including that of heavy drinkers, is quiteresponsive to price (1982, p. 387).'"''

Given the above evidence, future research on theestimation of alcohol and cigarette demand func-tions both in aggregate and in micro data would begreatly enriched if it incorporates insights providedby Becker & Murphy's (1988) model of addictivebehaviour. Becker & Murphy assume that rationalconsumers maximize a utility function that at anymoment in time depends on two goods: c and y.These goods differ because current utility alsodepends on a measure of past consumption of c butnoty. Thus, c is addictive because an increase in thestock of c due to an increase in past consumptionaffects current utility. In particular, in the case ofharmful addiction (the relevant case for cigarettesand alcohol), an increase in the stock lowers current

" Cook and Tauchen's study is limited to license states." Other evidence on the responsiveness of alcohol use and

cigarette smoking by adults or by persons of all ages to price issummarized by Lewit, Coate & Grossman (1981); Lewit & Coate(1982); Grossman, Coate & Arluck (1987); and Grossman (1988).


utility but raises the marginal utility of currentconsumption of c. Consumers are rational in thesense that they take account of these effects inallocating their wealth between c and y. Theymaximize a well-behaved utility function (an indif-ference curve between c and y is convex to theorigin) that is separable over time in c, y, and thestock of c but not in c and y alone.

Although the utility function is separable overtime, Becker & Murphy show that the quantitydemanded of the addictive good should be inverselyrelated not only to the current price of the good butalso to its past and future prices." Past prices arerelevant because they affect past consumption of theaddictive good, and an increase in past consumptiondue to a reduction in past prices raises currentconsumption. The future price is relevant because areduction in it raises future consumption whichlowers the 'shadow price' of current consumption.Put differently, the demand function of a rationalconsumer exhibits the property of symmetry: in-creases in past or future consumption for whateverreason cause current consumption to rise. Becker &Murphy also show that the long-run response to apermanent price change should exceed the short-runresponse in the case of an addictive good. (Theseresponses are defined below). Since this propertydoes not hold for a non-addictive good, the priceelasticity of demand may be larger for the formerthan for the latter.

When the utility function is quadratic and the rateof depreciation on the addictive stock is 100%, theBecker-Murphy model generates a structural de-mand function for consumption at time t (c,) of theform

c,=a,c,_,+)SaiC,+i+a2p,+a3«,+a4M,+i. (1)

Here P is the rate of time preference (timediscount), p, is the price of c,, and u, and M,+ , areunobserved variables that affect utility in periods tand t+1. Equation (1) is the basis of the empiricalimplementation of the model. Note that ordinaryleast squares estimation of the equation would leadto biased estimates of the parameters of interest.The unobserved variables that affect utility in eachperiod are likely to be serially correlated. Even ifthese variables are uncorrelated, c,_i depends on u,.

" The above predictions pertain to a demand function thatholds the marginal utility of wealth constant. This is the standarddemand function employed in life cycle utility maximizationmodels.

and c,+ i depends on u,+i through the optimizingbehaviour. These relationships imply that an ordi-nary least squares estimation of the equation mightincorrectly imply that past and future consumptionaffect current consumption, even when the truevalue of a, is zero. Put differently, past and futureconsumption are endogenous variables and must betreated as such in estimating the model.

Fortunately, the specification in equation (1)suggests a way to solve the endogeneity problem.The equation implies that current consumption isindependent of past and future prices when past andfuture consumption are held constant; any effect ofpast or future prices on current consumption mustcome through their effects on past or futureconsumption. Provided that the unobservables areuncorrelated with prices in these periods, past andfuture prices are logical instruments for past andfuture consumption, since past prices directly affectpast consumption, and future prices directly affectfuture consumption. Therefore, the empirical strat-egy amounts to estimating equation (1) by two-stage least squares, with past and future pricesserving as instrumental variables for past and futureconsumption.

The statistical significance of the coefficient offuture consumption provides a direct test of arational model of addiction against an alternativemodel in which consumers are myopic. In the lattermodel they fail to consider the impact of currentconsumption on future utility and future consump-tion. The parameters of the equation also allow oneto compute the long- and short-run responses ofconsumption to a permanent decline in price. Thelong-run response is the effect on consumption of achange in price in all periods. The short-runresponse is the effect on consumption of a change inprice in the current period and all future periods.(That is, past prices are held constant). Theseeffects are obtained by solving the second-orderdifference equation in (1).

Becker, Murphy and I (Becker, Grossman &Murphy 1987) have applied the above model to thedemand for cigarettes using a time series of statecross sections for the period from 1955 through1985. The estimated coefficient of future consump-tion is positive and statistically significant. Thus, wereject the myopic model of addiction in favour ofthe rational model of addiction. The long-run priceelasticity of demand for cigarettes of —0.77 exceedsthe short-run elasticity of —0.44 by almost 100%.Moreover, our long-run elasticity is at the high endof existing estimates, suggesting that these estimates


are biased because they do not take account of theaddictive nature of cigarette consumption."

Currently, Becker, Murphy and I are using thesame model and data base to fit demand functionsfor distilled spirits and for excessive alcohol con-sumption measured by cirrhosis mortality. Similarmodels can be fit with micro data that containalcohol consumption at several points in the lifecycle. If our results for cigarettes also hold foralcohol, the price elasticity of demand for alcoholicbeverages may be larger than the estimates con-tained in existing research, and the health benefits ofincrease in alcohol and cigarette excise taxes may begreater than those summarized in this paper.

" Chaloupka (1988) obtains similar results in an empiricalapplication of the Becker-Murphy model to a micro data set.

ReferencesBECKER, G . S. (1968) Crime and punishment: an economic

approach, Journal of Political Economy, 76, pp.169-217.

BECKER, G . S. (1985) Don't raise the drinking age, raisetaxes. Business Week, November 25, p. 21.

BECKER, G . S., GROSSMAN, M . & MtmPHy, K M. (1987)An Empirical Analysis of Cigarette Addiction, WorkingPaper (Department of Economics, University of Chi-cago).

BECKER, G . S. & MtmPHY, K. M. (1988) A theory ofrational addiction. Journal of Political Economy, 96, pp.675-700.

BLANE, H . T . & HEwrrr, L. E. (1977) Alcohol and Youth:An Analysis of the Literature. Prepared for the NationalInstitute on Alcohol Abuse and Alcoholism.(Springfield, Virginia, U.S. National Technical Infor-mation Service).

BUREAU OF JUSTICE STATISTICS, U.S. DEPARTMENT OFJUSTICE (1979) Comparative Data Report (Washington,D.C., U.S. Government Printing Office).

BtjREAU OF LABOR STATISTICS, U.S. DEPARTMENT OF LA-BOR (1973) Average Retail Prices of Selected Commodi-ties and Services, Fall 1971 (Washington, D.C., U.S.Government Printing Office).

BLIREAU OF LABOR STATISTICS, U.S. DEPARTMENT OFLABOR. Monthly Labor Review, various years.

CENTERS FOR DISEASE CONTROL & NATIONAL CANCERiNSTrrtJTE (1976) Adult Use of Tobacco-1975 (Wash-ington, D.C., U.S. Department of Health, Education,and Welfare).

CHALOUPKA, F. (1988) An Economic Analysis of AddictiveBehavior: the case of cigarette smoking, Ph.D. disserta-tion (City University of New York Graduate School).

COATE, D . & GROSSMAN, M . (1988) Effects of alcoholicbeverage prices and legal drinking ages on youth alcoholuse. Journal of Law and Economics, 31, pp. 145-171.

COOK, P. J. (1981) The effect of liquor taxes on drinkingcirrhosis, and auto fatalities, in: M. H. MOORE & D. R.GERSTEIN (Eds) Alcohol and Public Policy: beyond theshadow of prohibition (Washington, D.C., NationalAcademy Press).

COOK, P. J. (1984) Increasing the federal alcohol excisetax, in: D. R. GERSTEIN (Ed) Alcohol Problems:government, business, and community action (Washing-ton, D.C., National Academy Press).

COOK, P. J. & TAUCHEN, G . (1982) The effect of liquortaxes on heavy drinking. Bell Journal of Economics, 13,pp. 379-390.

GROSSMAN, M . , COATE, D . & ARLUCK, G . M . (1987) Pricesensitivity of alcoholic beverages in the United States,in: H. D. HOLDER (Ed) Control Issues in Alcohol AbusePrevention: strategies for communities (Greenwich, Con-necticut, JAI Press Inc.).

GROSSMAN, M. (1988) Health Economics of Prevention ofAlcohol-Related Problems. Presented at a workshop onHealth Economics of Prevention and Treatment ofAlcohol-Related Problems sponsored by the NationalInstitute on Alcohol Abuse and Alcoholism and theNational Center for Health Services Research andHealth Care Technology in Washington, DC, Septem-ber 29-30.

GROSSMAN, M . , SAFFER, H . & CHALOUPKA, F . Alcohol,Regulation, and Motor Vehicle Mortality. NationalInstitute on Alcohol Abuse and Alcoholism Grant No. 1ROl AA07593 [in progress].

HAMMOND, E. C. (1966) Smoking in Relation to the DeathRate of 7 Million Men and Women, National CancerInstitute Monograph No. 19. (Washington, D.C., U.S.Government Printing Office).

HARRIS, J. E. (1982) Increasing the federal excise tax oncigarettes. Journal of Health Economics, 1, pp.117-120.

HARRIS, J. E. (1984) More data on tax policy, in: D. R.GERSTEIN (-Ed) Toward the Prevention of AlcoholProblems: Government, Business, and Community Ac-tion (Washington, D.C., National Academy Press).

HARRIS, J. E. (1987) The 1983 increase in the federalcigarette excise tax. in: L. H. SUMMERS (Ed) Tax Policyand the Economy, Volume 1 (Cambridge, Massachus-etts, MIT press).

IppOLrro, R., MtmPHY, R. D. & SANT, D . (1979)Consumer Responses to Cigarette Health Information.Staff Report, Bureau of Economics, Federal TradeCommission (Washington, D.C., U.S. GovernmentPrinting Office).

JACOBSON, M . & ALBION, M . (1985) Raising alcohol taxesis the way to cut drinking and the debt, WashingtonPort, August 11.

KELLNER, I. L. (1973) The American Cigarette Industry: are-examination, Ph.D. dissertation (New School forSocial Research).

LEIBENSTEIN, H . (1950) Bandwagon, snob, and vebleneffects in the theory of consumer behavior. QuarterlyJournal of Economics, 64, pp. 183-207.

LEVY, D . & REINHART, G . R. (1988) Alcohol Prices andFatal Traffic Accidents: an alcohol policy analysis brief.Prepared for the Office of Policy Analysis, NationalInstitute on Alcohol Abuse and Alcoholism (SilverSpring, Maryland, Macro Systems, Inc.).

LEwrr, E. M., COATE, D . & GROSSMAN, M . (1981) Theeffects of government regulation on teenage smoking.Journal of Law and Economics, 24, pp. 545-569.

LEwrr, E. M. & COATE, D . (1982) The potential for usingexcise taxes to reduce smoking. Journal of HealthEconomics, 1, pp. 121-145.


LEwrr, E. M. (1982) The Interstate Market for SmuggledCigarettes. Presented at a Session on Issues in Govern-ment Finance sponsored by the American EconomicAssociation at the annual meetings of the Allied SocialScience Associations in New York, New York onDecember 28-30.

LEWIT, E . M . (1985) Regulatory and Legislative Initia-tives, in: Proceedings of the Pennsylvania ConsensusConference on Tobacco and Health Priorities (Indiana,Pennsylvania, The Pennsylvania Interagency Councilon Tobacco and Health).

LiiKS, A. (1983) Will America Sober Up? (Boston, BeaconPress).

MOORE, M . H . & GERSTEIN, D . R. (Eds) (1981) Alcoholand Public Policy: beyond the shadow of prohibition(Washington, D.C., National Academy Press).

NATIONAL HIGHWAY TRAFFIC SAFETY ADMINISTRATION,U.S. DEPARTMENT OF TRANSPORTATION (1985) A

Digest of State Alcohol-Related Highway Safety Laws.Third Edition. DOT HS 806 745.

NATIONAL HIGHWAY TRAFFIC SAFETY ADMINISTRATION,U.S. DEPARTMENT OF TRANSPORATION (1986) FatalAccident Reporting System, 1984. DOT HS 806 919.

NELSON, J. R. (1988) Effecu of Regulation on AlcoholicBeverage Consumption: Regression Diagnostics and Infl-uential Data, Presented at the Fourteenth AnnualConvention of the Eastern Economic Association,Boston, Masachusetts.

ORNSTEIN, S. I. & HANSSENS, D . M . (1985) Alcoholcontrol laws and the consumption of distilled spirits andheer. Journal of Consumer Research, 12, pp. 200-213.

RACHAL, J. V., Gtress, L. I., HUBBARD, R. L., MAISIO, S. A.,CAVANAUGH, E. R., WADDELL, R. & BENRUD, C . H .

(1980) The Extent and Nature of Adolescent AlcoholAbuse: The 1974 and 1978 National Samlple Surveys,Prepared for the National Institute on Alcohol Abuseand Alcoholism (Springfield, Virginia, U.S. NationalTechnical Information Service).

Ross, H. L. (1984) Deterring the Drinking Driver, LegalPolicy and Social Control. Revised and Updated Edition(Lexington, Massachusetts, D.C. Heath and Company).

SAFFER, H . & CHALOUPKA, F . (1989) Breath testing andhighway fatality rates. Applied Economics, 21, pp.901-912.

SAFFER, H . & GROSSMAN, M . (1987a) Beer taxes, the legaldrinking age, and youth motor vehicle fatalities. Journalof Legal Studies, 16, pp. 351-374:

SAFFER, H . & GROSSMAN, M . (1987b) Drinking age lawsand highway mortality rates: cause and effect. EconomicInquiry, 25, pp. 403-418.

TOBACCO iNSTmrrE (1987) The Tax Burden on Tobacco(Washington D.C, The Tobacco Institute).

VoAS, R. B. & MouLDEN, J. (1980) Historical trends inalcohol use and driving by young americans. in: H.WECHSLER (Ed) Minimum Drinking Age Laws: anevaluation (Lexington, Massachusetts, D.C. Heath andCompany).

WALES, T . J. (1968) Distilled spirits and interstateconsumption, American Economic Review, 58, pp.853-863.

WARNER, K. E. (1986) Smoking and health implications ofa change in the federal cigarette excise tax. Journal ofthe American Medical Association, 255, pp. 1028-32.

health benefits of increases in alcohol and cigarette taxes · the anti-drinking campaign dates to...

Documents