The interactive effect of alcohol and altitude on traffic fatalities.

I. Introduction

The fundamental maxims of driving are "speed kills" and "drinking and driving don't mix." A series of recent articles have analyzed these speed adages and based on the work of Lave[8; 9] and others, we now know that both speed and its variance kill.(1) In this paper we expand the analysis to investigate the potential for an interactive effect between alcohol and altitude on automobile fatality rates. We also examine a number of other policy and driving related variables. That drinking and driving don't mix is supported by Loeb [12; 13; 14], Levy and Asch [11], Sommers [18], and to some extent, Cook and Tauchen[4]. That the deleterious effects of alcohol on reaction time is more pronounced due to the diminished oxygen intake to the brain at higher elevations is physiologically supported in studies by Mazess, et al.[15] and Newman [16].(2) Our main finding is that the interaction between altitude and alcohol consumption may have a pronounced effect on traffic fatalities and largely explains the disproportionate fatality rates found in western regions of the United States.(3) In other words, drinking and driving don't mix, especially in the Rocky Mountains.

II. Data and Model

The model used in this paper is based on elementary cross sectional state data for the contiguous 48 states.(4) The dependent fatality variable is an interstate motor vehicle fatality rate computed by the ratio of the number of traffic fatalities to the number of vehicle miles driven. Independent variables include:(5)

AGE: Minimum legal drinking age.

BEER: Per capita consumption of malt beverages.

SPEED: Weighted average speed of free-moving vehicles driven on rural and urban interstate

highways.(6)

VAR: Weighted average speed variability of free-moving vehicies driven on rural and

urban interstate highways.

D: Dununy variable for inspection program; 1 for states with inspection program in

effect, 0 otherwise.

YOUTH: Percentage of the population between the ages of 18 and 24.

ALT: An average of altitudes of metropolitan areas within a state.(7) ALTBEER: The interaction of ALT and BEER.

Regression results for several model specifications are provided in Tables I (linear) and II (logarithmic).(8) Models are of the form:

where [X.sub.ji] represent socioeconomic and driving related independent variables and [beta.sub.j] (j = 0, 1, . . .) represent parameters to be estimated.

[TABULAR DATA OMITTED]

Bayesian extreme bounds analysis (EBA) results, with attention to the interaction term, are provided in Table III.(9) Bayesian EBA is designed to expose problems of inferential fragility due to multicollinearity when model specifications are altered. The extreme bounds represent the maximum and minimum coefficients that could be obtained via maximum likelihood estimation when all possible linear combinations of variables that might not be included in a specification (doubtful variables) are added or dropped in a model specification. Variables are classified as either free or doubtful. Free variables have a diffuse prior probability distribution and doubtful variables have a prior location set to zero. The prior variance/covariance matrix is set to the identity matrix and results are computed for a scalar multiple of this matrix as the scalar is swept from zero to infinity.(10)

[TABULAR DATA OMITTED]

In terms of conventional statistical analysis, the coefficient on the interaction term is statistically significant over the range of specifications. These results in driving related fatality rate of a Western dummy variable to account for possible differences in driving related fatality rates between Western and non-Western states.(11) The Bayesian results further demonstrate the robustness of the results since the upper and lower bounds for the alcohol/altitude interaction, speed, variance, and inspection variables do not cover zero and lie within the theoretically anticipated region.

III. Conclusions

The empirical results, both classical and Bayesian, support a positive interaction effect between alcohol and altitude. More importantly, the results are non-fragile across a large set of reasonable model specifications. We also find positive, significant, and non-fragile effects from speed and the variance of speed which is consistent with Fowles and Loeb[6], Levy and Asch [11], and Synder [19]. In addition we find a negative and non-fragile effect associated with the presence of state safety inspection programs which are also statistically significant in the logarithmic models. We conclude that govemment agencies in high altitude regions of the United States should seriously consider intensified drunk driving law enforcement to reduce fatality rates. In addition, differential taxing policies might be employed to reduce alcohol sales in high risk areas provided access to their purchase in low risk areas is small due to, for example, travel distances.

(1.) In addition to speed and variance effects on traffic fatalities, Fowles and Loeb[6] studied socioeconomic and policy related variables. Levy and Asch [11] examined a similar model. Snyder[19] and Lave [8] focused attention on the effects of speed and the variance of speed, as did Rodriguez[17]. These papers all pointed to the importance of speed and/or speed variance as a relevant determinant of highway deaths. (2.) The interaction between altitude and alcohol appears to be more pronounced for seasoned subjects than for newcomers to higher altitudes. See, for example, Mazess, et al.[15]. Freedman, et al.[7], Collins, Mertens, and Higgins [3], and Collins and Mertens[2] question the interactive effects for subjects newly introduced to high altitudes. (3.) Fowles and Loeb[6] discovered a robust regional dummy effect for western states when analyzing speed and speed variance effects on fatalities. Also see Sommers[18], who attributes the peculiar aspects of Rocky Mountain fatalities to such factors as noncompliance of speed limits, unusual roadway characteristics, and recreational dtiving. (4.) Analysis is based on data for the year 1979. The authors greatly appreciate an anynomous reviewer's assistance in providing data on traffic fatalities for Louisiana, Maryland, and Rhode Island. (5.) Data on fatality rates and highway mileage are from Highway Statistics, 1979. Data on the inspection variable are from Digest of Motor Laws, 1980, and data on personal income per capita, population density, population by age, and non-metropolitan area populations are from The Statistical Abstract, 1980, 1981. The data on states' areas and metropolitan altitudes are from The World Almanac, 1981. (6.) See Fowles and Loeb[6] for a discussion of weighted speed. In addition, statistical results were obtained using the average speed on rural interstate highways with similar results to those reported here (available from the authors). (7.) The basic results are not sensitive to various altitude measures such as maximum altitude, minimum altitude, or other statewide altitude charactcristics. (8.) Other model specifications that included median income, hospital access, and population density were also considered. Results in terms of speed, variance, inspection, and alcohol/altitude interaction are not different from those reported in this paper. Results are available from the authors. (9.) EBA results are presented only for the linear model. Perfect collinearity occurs in the logarithmic specification, being interactive by design, when interactive terms are introduced. (10.) For a theoretical discussion of extreme bounds analysis, see Leamer[10]. For an application, see Fowles and Loeb[6]. Computations were performed using Micro EBA, see Fowles[5]. (11.) See Fowles and Loeb[6] and Sommers[18] for an additional investigation into this phenomenon. The statistical results are available from the authors.

References

[1.] Asch, Peter and David T. Levy, "Does the Minimum Legal Drinking Age Affect Traffic Fatalities?" Journal of Policy Analysis and Management, Winter 1987, 180-92. [2.] Collins, William E. and Henry W. Mertens, "Age, Alcohol, and Simulated Altitude: Effects on Performance and Breathalyzer Score." Aviation, Space, and Environmental Medicine, November 1988, 1026-33. [3.] _____, Henry W. Mertens, and E. Arnold Higgins, "Some Effects of Alcohol and Simulated Altitude on Complex Performance Scores and Breathalyzer Readings." Aviation, Space, and Environmental Medicine, April 1987, 328-32. [4.] Cook, Philip J. and George Tauchen, "The Effect of Minimum Drinking Age Legislation on Youthful Auto Fatalities, 1970-1977." Journal of Legal Studies, January 1984, 13, 169-90. [5.] Fowles, Richard, "Micro EBA." The American Statistician, November 1988, 274. [6.] _____ and Peter D. Loeb, "Speeding, Coordination, and the 55-MPH Limit: Comment." American Economic Review, September 1989, 916-21. [7.] Freedman, Robert, Merilyne C. Waldo, Lawerence E. Adler, Neil Baker, Dennis Levin, and Richard Deitrich, "Electrophysiological Effects of Low Dose Alcohol on Human Subjects at High Altitude." Alcohol and Drug Research, 6, 1985-86, 289-97. [8.] Lave, Charles A. "Speeding, Coordination, and the 55-MPH Limit: Reply." American Economic Review, September 1989, 926-31. [9.] _____, "Speeding, Coordination, and the 55-MPH Limit." American Economic Review, December 1985, 1159-64. [10.] Leamer, Edward E., "Sets of Posterior Means with Bounded Variance Priors," Econometrica, May 1982, 725-36. [11.] Levy, David T. and Peter Asch, "Speeding, Coordination, and the 55-MPH Limit: Comment." American Economic Review, September 1989, 913-15. [12.] Loeb, Peter D., "The Determinants of Motor Vehicle Accidents - A Specification Error Analysis." Logistics and Transportation Review, March 1988, 24, 33-48. [13.] _____, "The Determinants of Motor Vehicle Fatalities with Special Consideration to Policy Variables." Journal of Transport Economics and Policy, September 1987, 279-87. [14.] _____, "The Efficacy and Cost-Effectiveness of Motor Vehicle Inspection Using Cross-Sectional Data - An Econometric Analysis." Southern Economics Journal, October 1985, 500-09. [15.] Mazess, R. B., Emilio Picon-Reategui, R. Brooke Thomas, and Michael A. Little, "Effects of Alcohol and Altitude on Man During Rest and Work." Aerospace Medicine, April 1968, 403-06. [16.] Newman, Henry W., "The Effect of Altitude on Alcohol Tolerance." Quarterly Journal of Studies on Alcohol, December 1949, 398-403. [17.] Rodriguez, Ricardo J., "Speed, Speed Dispersion, and the Highway Fatality Rate." Southern Economic Journal, October 1990, 349-56. [18.] Sommers, Paul M., "Drinking Age and the 55 MPH Speed Limit." Atlantic Economic Journal, March 1985, 43-48. [19.] Synder, Donald, "Speeding, Coordination, and the 55-MPH Limit: Comment." American Economic Review, September 1989, 922-25.