Gasoline Prices and Traffic Crashes in Alabama

Traffic Injury Prevention, 13:476–484, 2012 C 2012 Taylor & Francis Group, LLC Copyright ISSN: 1538-9588 print / 1538-957X online DOI: 10.1080/15389588.2012.670815

Downloaded by [Mississippi State University Libraries], [Guangqing Chi] at 07:08 01 October 2012

Gasoline Prices and Traffic Crashes in Alabama, 1999–2009 GUANGQING CHI,1 TIMOTHY E. MCCLURE,2 and DAVID B. BROWN3 1

Department of Sociology and Social Science Research Center, Mississippi State University, Mississippi State, Mississippi Department of Sociology, Mississippi State University, Mississippi State, Mississippi 3 Department of Computer Science, The University of Alabama, Tuscaloosa, Alabama 2

Objective: The price of gasoline has been found to be negatively associated with traffic crashes in a limited number of studies. However, most of the studies have focused either on fatal crashes only or on all crashes but measured over a very short time period. In this study, we examine gasoline price effects on all traffic crashes by demographic groups in the state of Alabama from 1999 to 2009. Methods: Using negative binomial regression techniques to examine monthly data from 1999 to 2009 in the state of Alabama, we estimate the effects of changes in gasoline price on changes in automobile crashes. We also examine how these effects differ by age group (16–20, 21–25, 26–30, 31–64, and 65+), gender (male and female), and race/ethnicity (non-Hispanic white, non-Hispanic black, and Hispanic). Results: The results show that gasoline prices have both short-term and long-term effects on reducing total traffic crashes and crashes of each age, gender, and race/ethnicity group (except Hispanic due to data limitations). The short-term and long-term effects are not statistically different for each individual demographic group. Gasoline prices have a stronger effect in reducing crashes involving drivers aged 16 to 20 than crashes involving drivers aged 31 to 64 and 65+ in the short term; the effects, however, are not statistically different across other demographic groups. Conclusions: Although gasoline price increases are not favored, our findings show that gasoline price increases (or decreases) are associated with reductions (or increases) in the incidence of traffic crashes. If gasoline prices had remained at the 1999 level of $1.41 from 1999 to 2009, applying the estimated elasticities would result in a predicted increase in total crashes of 169,492 (or 11.3%) from the actual number of crashes. If decision makers wish to reduce traffic crashes, increasing gasoline taxes is a possible option—however, doing so would increase travel costs and lead to equity concerns. These findings may help to shape transportation safety planning and policy making. Keywords Gasoline prices; Gasoline taxes; Traffic crashes; Traffic safety; Alabama

INTRODUCTION A large body of literature has found that economic factors have important effects in determining the frequency of traffic crashes (Joksch 1984; Sivak and Schoettle 2010; Traynor 2009; Wagenaar 1984). The literature in general suggests that economic conditions have a negative association with traffic crashes. In a bad economy, people drive less by reducing trip frequency and distance, making more multipurpose trips rather than single-purpose trips, and reducing vacation travel in order to save on gasoline expenditures (Chi et al. 2010). All of these intermediate factors decrease people’s exposure to traffic crashes.

Received 27 October 2011; accepted 25 February 2012. Address correspondence to Guangqing Chi, Department of Sociology and Social Science Research Center, Mississippi State University, P.O. Box C, Mississippi State, MS 39762. E-mail: [email protected]

Gasoline prices have also been found to relate to traffic safety in a limited body of literature (Chi et al. 2010). Most of the studies that have found that higher gasoline prices lead to drops in the occurrence of traffic crashes, have been conducted from 3 perspectives. First, a majority of the studies focused on traffic fatalities and found that higher gasoline prices are associated with fewer traffic fatalities (Grabowski and Morrisey 2004, 2006; Leigh and Geraghty 2008; Leigh and Wilkinson 1991; Morrisey and Grabowski 2011; Sivak 2009; Wilson et al. 2009). Although fatalities are the most severe outcome of traffic crashes, they occur in the smallest proportion of all crashes and thus are not representative of the effect of gasoline prices on the overall level of traffic crashes (Kenkel 1993; McGwin and Brown 1999). Second, some of these studies have examined variations in the effects of gasoline prices on traffic crashes by age (Grabowski and Morrisey 2004; Leigh and Wilkinson 1991). Younger adult drivers, who tend to have less driving experience, engage in more risky driving behaviors, drive more at night,

476

477


GASOLINE PRICES AND TRAFFIC CRASHES

and are more likely to be involved in crashes than older drivers (Arnett 2002; Williams 2003; Williams et al. 1998). Although the effects on crashes have been studied by age, the possible variation in gasoline price effects on crashes between male and female drivers as well as across racial groups has rarely been examined (for exceptions, see Chi et al. 2010, 2011). Third, Chi et al. (2011) provided a comprehensive analysis in which they investigated gasoline price effects on all traffic crash types (including fatal, injury, and property damage–only crashes) and by age, gender, and race in Mississippi. Their findings indicated that higher gasoline prices lead to fewer traffic crashes and that the effects vary by age, gender, and race as well as by crash type. Specifically, they found that gasoline prices have greater effects on less severe crashes. However, their data were limited to the time period from April 2004 to December 2008. A longer time period covering economic growth and downturn would provide more robust results. In this study, we focus on the state of Alabama to study gasoline price effects on traffic crashes from 1999 to 2009 by age, gender, and race/ethnicity. Alabama is a neighbor state of Mississippi with similar culture, education, and economic conditions; thus, the results could be compared to the studies by Chi et al. (2010, 2011). However, different from their studies, we use data from a much longer time period—from 1999 to 2009—covering a wide range of gasoline prices, which provides more robust results. This study contributes to the literature by comprehensively examining and comparing gasoline price effects on total traffic crashes across demographic groups throughout a relatively long time period. METHODS Data The data are Alabama monthly traffic crash figures from 1999 to 2009 obtained from the Alabama CARE system at the Center for Advanced Public Safety of the University of Alabama. The crash data are reported by all law enforcement agencies in Alabama

Table I

and include all recorded crashes in the state. Because there is no minimum reporting threshold, only crashes for which law enforcement officers were not called to investigate are omitted. For each crash, law enforcement officers provide data regarding the type of crash, age, race/ethnicity, and gender of each driver and passenger, as well as about 100 other data elements. These data were aggregated to create monthly counts for the total number of crashes. To examine variations by age, gender, and race/ethnicity, we also calculated crash counts and rates for 10 separate demographic groups: crashes involving drivers aged 16 to 20, drivers aged 21 to 25, drivers aged 26 to 30, drivers aged 31 to 64, drivers aged 65+, male drivers, female drivers, non-Hispanic white drivers, non-Hispanic black drivers, and Hispanic drivers. The summary statistics of the variables are presented in Table I. The traffic crash rate (measured as crashes per 100,000 persons) is the highest for crashes involving teenaged drivers aged 16 to 20 and decreases as age increases. The crash rate for crashes involving male drivers is higher than that for crashes involving female drivers; it is also slightly higher for crashes involving non-Hispanic black drivers than for crashes involving non-Hispanic white drivers. The variations in crash rates among these demographic groups generally correspond to the national average. Our primary explanatory variable is the average per gallon price of regular-grade unleaded gasoline, adjusted for inflation (in January 2010 dollars). Data for this variable were obtained from the U.S. Department of Energy’s Energy Information Administration (EIA) for the period from 1999 to 2009. Data were available only on a regional basis, so we approximated the price of gasoline in Alabama by using average prices in the Gulf Coast region. Because existing studies show both short-term and long-term effects of gasoline prices on traffic crashes, we measured gasoline prices at both the month when the crashes occurred and a 1-year lag (Chi et al. 2010; Grabowski and Morrisey 2004). For instance, corresponding to the traffic crash rate in December 2009, the current gasoline price is measured as of

Descriptive statistics of crashes at the monthly level Number of crashes

Driver characteristics

Monthly mean

SD

Minimum

Crashes per 100,000 persons Maximum

Monthly mean

SD

Minimum

Maximum

Age 16–20 21–25 26–30 31–64 65+

2233 1573 1127 4707 974

242 142 81 343 87

1607 1214 902 3913 773

2740 1896 1317 5440 1198

690 494 375 235 162

82 44 36 19 18

484 371 279 190 121

841 579 449 270 208

Gender Male Female

6244 4547

509 346

4914 3718

7168 5279

370 244

37 20

280 193

443 286

Race/ethnicity Non-Hispanic white Non-Hispanic black Hispanic

7727 2847 113

616 235 50

6201 2204 25

8771 3322 262

298 324 155

27 33 46

233 234 48

340 387 282

11, 357

828

9099

13, 021

320

27

247

369

Total


478

CHI ET AL.

December 2009; the gasoline price at a 1-year lag is measured as of December 2008. The former measure represents a short-term effect, and the latter represents a long-term effect. Unemployment rate is used as a control variable because an unemployed person is likely to drive less and thus has less exposure to traffic crashes (Chi et al. 2010; Graham and Glaister 2003; Leigh and Wilkinson 1991; Quddus 2008). Crash rates are likely to be lower when unemployment rates are high (Partyka 1984). Monthly unemployment data were obtained from the U.S. Bureau of Labor Statistics (2009). Climate is also used as a control variable because climate varies seasonally and is found to affect traffic crashes (e.g., Quddus 2008). Data for the mean rainfall and temperature in Alabama were obtained from the Southeast Regional Climate Center (2011). Other variables such as driving behaviors, road conditions, and vehicle characteristics have also been found to affect the likelihood of traffic crash involvement (Fu 2008; Leigh and Geraghty 2008). However, our data show that the relative contribution of these factors (including roadway defects, road composition, roadway slope and curvature, lighting conditions, and vehicle type) did not change significantly at the annual level over the 11-year period of this study. Therefore, the data for these variables were not used. The analysis was based on the crashrelated information available from the Alabama CARE system. The results are not included in this article but are available upon request. Analysis We first visualized the relationships between gasoline price changes and total crashes. We then investigated gasoline price effects on each crash type in regression analyses. In total, there were 11 crash measures: total crashes as well as crashes by age group (16–20, 21–25, 26–30, 31–64, and 65+), gender (male and female), and race/ethnicity (non-Hispanic white, nonHispanic black, and Hispanic). Each crash measure was modeled as a function of average gasoline price during the month when the crashes occurred, gasoline price at a 1-year lag, state unemployment rate, rainfall, and temperature. Temporal variations are controlled for by a time trend variable. Poisson distribution is often used to describe crash counts because the data relating to traffic crashes are random, discrete, and nonnegative events (Kulmala 1995; Long 1997; Miaou 1994). One of its assumptions is that the mean of crashes equals the variance of crashes. If this assumption is violated, negative binomial regression models should be used (Abdel-Aty and Radwan 2000; Lord 2000). Our analysis suggests that all of the 11 crash measures exhibited overdispersion. Thus, negative binomial regression models were used for all crash measures. In addition, populations of each demographic group were used as exposure variables in corresponding regression models. The coefficients could then be interpreted as the effects of gasoline prices on crash rates (the number of crashes per capita) rather than on crash counts. In many previous studies (e.g., Chi et al. 2010; Grabowski and Morrisey 2004), vehicle miles traveled (VMT), or annual average daily traffic (AADT) counts

was used as the exposure variable. However, this study examines the variations in gasoline prices on traffic crashes by age, gender, and race/ethnicity; the VMT or AADT data for each demographic group at the monthly or annual levels are not available. Population can be appropriately used as an exposure variable—population is often used as the denominator in calculating a rate of interest (e.g., mortality due to a certain disease) in public health studies (e.g., Horner et al. 2009). In order to facilitate interpreting gasoline price effects on reducing crashes, elasticities of traffic crashes per capita with respect to gasoline prices were calculated on the basis of coefficient estimates from negative binomial regression models. In addition, elasticity differences by age, gender, and race/ethnicity were statistically tested by using the method addressed in Clogg et al. (1995). RESULTS Gasoline Prices and Total Crashes per Capita We first show the association between gasoline prices and traffic crashes. Figure 1 illustrates the relationship between annual average gasoline prices (adjusted for inflation in January 2010 dollars) and annual total traffic crashes per capita (measured as total traffic crashes per 100,000 persons) in Alabama from 1999 to 2009. It appears that a negative relationship exists between gasoline prices and crashes per capita. For example, in 2000, average gasoline prices were at a 5-year peak and traffic crashes per capita were at their lowest point in 7 years. From 2000 to 2002, gasoline prices decreased but crashes per capita increased. Immediately following that, gasoline prices increased until 2008 but crashes per capita decreased in the same time period, except from 2003 to 2004. The most dramatic decline in crashes occurred from 2007 to 2008 when gasoline prices had the largest increase. The declining trend in crashes had almost reached a plateau when gasoline prices reached their sharpest decline from 2008 to 2009. The results from the negative binomial regression models support the observation that increased gasoline prices lead to reduction in traffic crashes (see Appendix). To facilitate the interpretation, the elasticities of crashes per capita with respect to

Figure 1 Annual average gasoline prices (in January 2010 dollars) and annual total crashes per 100,000 persons, 1999 to 2009, Alabama.

479



Table II

Elasticities of traffic crashes per capita with respect to gasoline prices as predicted from negative binomial regression models, 1999 to 2009, Alabamaa Average gasoline price during the month the crashes occurred (95% confidence interval)

Gasoline price at a 1-year lag (95% confidence interval)

Total

−0.20537∗ (−0.28046, −0.13031)

−0.18045∗ (−0.24646, −0.11442)

Age 16–20 21–25 26–30 31–64 65+

−0.32529∗ −0.21168∗ −0.23543∗ −0.15261∗ −0.16603∗

−0.20974∗ −0.14859∗ −0.21509∗ −0.17728∗ −0.17398∗

Gender Male Female

−0.20199∗ (−0.27300, −0.13097) −0.21441∗ (−0.29880, −0.12999)

−0.17570∗ (−0.23826, −0.11312) −0.18931∗ (−0.26369, −0.11491)

Race/ethnicity Non-Hispanic white Non-Hispanic black Hispanicb

−0.21500∗ (−0.29170, −0.13828) −0.20512∗ (−0.28374, −0.12650) −0.24297 (−0.51184, 0.02588)

−0.17860∗ (−0.24629, −0.11091) −0.20031∗ (−0.26962, −0.13100) −0.15215 (−0.37801, 0.07370)

(−0.42057, −0.23000) (−0.29532, −0.12801) (−0.31765, −0.15321) (−0.22425, −0.08098) (−0.26325, −0.06884)

(−0.29377, −0.12571) (−0.22235, −0.07483) (−0.28753, −0.14265) (−0.24046, −0.11410) (−0.25984, −0.08814)

elasticities were calculated by using the studied period’s average of $2.13 for gasoline prices. Gasoline prices were adjusted in January 2010 dollars. current time nor 1-year lag gasoline prices have significant effects on reducing crashes involving Hispanics at P ≤ .05. ∗ P ≤ .01. aThe

bNeither

gasoline prices were calculated (Table II). The elasticities were measured at the studied period’s mean price of $2.13 for gasoline prices. It was found that gasoline price has both short-term and long-term effects on the total crash rates. A 10 percent increase in the inflation-adjusted gasoline price is associated with a 2.1 percent decrease in the monthly total crash rates immediately (short term) and a 1.8 percent decrease in the monthly total crash rates over a year (long term). The difference between the short-term and long-term effects, however, was not statistically significant (Table III). These results were based on total crashes per capita; variations by age, gender, and race/ethnicity may exist. Therefore, in the following subsections we further examine the effects of gasoline prices on crashes by these demographic groups. Table III Significance tests and 95 percent confidence intervals of elasticity differences between the short-term and long-term effects as predicted from negative binomial regression models, 1999 to 2009, Alabama Elasticity difference (95% confidence interval)a Total

−0.02492 (−0.12490, 0.07506)

Age 16–20 21–25 26–30 31–64 65+

−0.11555 (−0.24262, 0.01152) −0.06309 (−0.17463, 0.04845) −0.02034 (−0.12993, 0.08925) 0.02467 (−0.07084, 0.12017) 0.00795 (−0.12172, 0.13761)

Gender Male Female

−0.02629 (−0.12094, 0.06837) −0.02509 (−0.13761, 0.08743)

Race/ethnicity Non-Hispanic white Non-Hispanic black Hispanic

−0.03640 (−0.13871, 0.06591) −0.00482 (−0.10961, 0.09998) −0.09083 (−0.44196, 0.26031)

aNone of the elasticity differences between the short-term and long-term effects

are significant at P ≤ .05.

Variations by Age We separated traffic crashes into 5 age groups and examined the effects of gasoline prices on crashes involving each group separately. The 5 age groups are drivers aged 16 to 20, drivers aged 21 to 25, drivers aged 26 to 30, drivers aged 31 to 64, and drivers aged 65+. As shown in Table II, gasoline prices have both shortterm and long-term effects on reducing the traffic crash rate of each age group. In the short term, a 10 percent increase in gasoline prices is associated with a 3.3 percent decrease in monthly crashes involving drivers aged 16 to 20, a 2.1 percent decrease in monthly crashes involving drivers aged 21 to 25, a 2.4 percent decrease in monthly crashes involving drivers aged 26 to 30, a 1.5 percent decrease in monthly crashes involving drivers aged 31 to 64, and a 1.7 percent decrease in monthly crashes involving drivers aged 65+. Over the long term, a 10 percent increase in gasoline prices is associated with the following respective decreases in monthly crashes: 2.1 percent for crashes involving drivers aged 16 to 20, 1.5 percent for drivers aged 21 to 25, 2.2 percent for drivers aged 26 to 30, 1.8 percent for drivers aged 31 to 64, and 1.7 percent for drivers aged 65+. The difference between the short-term effect and long-term effect was not significant for each age group (Table III). Within the short-term and long-term effects separately, the elasticity differences across the 5 age groups generally were not statistically significant (Table IV). Only 2 pairs of age groups had significantly different elasticities of crashes per capita in the short term. Gasoline prices have immediately stronger effects on reducing crashes involving drivers aged 16 to 20 than crashes involving drivers aged 31 to 64; a 10 percent increase in gasoline prices is associated with 1.7 percent greater reduction (95% confidence interval [CI]: 0.5%–2.9%) in the former than the latter. Gasoline prices also have immediately stronger effects on reducing crashes involving drivers aged 16 to 20 than crashes involving drivers aged 65+; a 10 percent increase in gasoline

480

CHI ET AL.


Table IV Significance tests and 95 percent confidence intervals of elasticity differences with respect to gasoline prices as predicted from negative binomial regression models, 1999 to 2009, Alabama

Age 16–20 vs. 21–25 16–20 vs. 26–30 16–20 vs. 31–64 16–20 vs. 65+ 21–25 vs. 26–30 21–25 vs. 31–64 21–25 vs. 65+ 26–30 vs. 31–64 26–30 vs. 65+ 31–34 vs. 65+ Gender Male vs. female Race/ethnicity Non-Hispanic white vs. Non-Hispanic black Non-Hispanic white vs. Hispanic Non-Hispanic black vs. Hispanic ∗P

Elasticity difference with respect to average gasoline price during the month the crashes occurred (95% confidence interval)

Elasticity difference with respect to gasoline price at a 1-year lag (95% confidence interval)

−0.11362 (−0.24044, 0.01321) −0.08987 (−0.21575, 0.03602) −0.17268 (−0.29191, −0.05345)∗∗ −0.15926 (−0.29538, −0.02314)∗ 0.02375 (−0.09357, 0.14107) −0.05907 (−0.16921, 0.05108) −0.04565 (−0.17388, 0.08259) −0.08281 (−0.19188, 0.02626) −0.06940 (−0.19671, 0.05792) 0.01342 (−0.10732, 0.13416)

−0.06116 (−0.17298, 0.05067) 0.00535 (−0.10560, 0.11629) −0.03246 (−0.13759, 0.07267) −0.03577 (−0.15589, 0.08436) 0.06650 (−0.03689, 0.16988) 0.02869 (−0.06843, 0.12581) 0.02539 (−0.08779, 0.13857) −0.03781 (−0.13391, 0.05830) −0.04111 (−0.15342, 0.07120) −0.00330 (−0.10987, 0.10327)

0.01242 (−0.09789, 0.12273)

0.01361 (−0.08361, 0.11083)

−0.00989 (−0.11974, 0.09997) 0.02797 (−0.25163, 0.30756) 0.03785 (−0.24226, 0.31796)

0.02171 (−0.07516, 0.11857) −0.02646 (−0.26223, 0.20932) −0.04816 (−0.28441, 0.18809)

≤ .05. ∗∗ P ≤ .01.

prices is associated with 1.6 percent greater reduction (95% CI: 0.2%–3.0%) in the former than the latter. Variations by Gender We examined the effects of gasoline prices on crashes by males and females separately. The results in Table II indicate that gasoline prices have both an immediate effect and a long-term effect on the traffic crash rates of both male and female drivers. In the short term, a 10 percent increase in gasoline prices is associated with a decrease in monthly total crashes per capita of male drivers by 2.0 percent and by 2.1 percent for female drivers. Over the long term, a 10 percent increase in gasoline prices is associated with a decrease in monthly total crashes per capita of male drivers by 1.8 percent and by 1.9 percent for female drivers. The elasticity differences between the short-term and long-term were not statistically significant for both male and female drivers (Table III). Within the short-term and long-term effects separately, the elasticity difference between male drivers and female drivers were not significant either (Table IV). Variations by Race/Ethnicity We examined the effects of gasoline prices on all crashes by race/ethnicity. Table II shows that gasoline prices have both a short-term effect and a long-term effect on reducing crashes involving non-Hispanic whites and blacks. In the short term, a 10 percent increase in gasoline prices is associated with a 2.2 percent decrease in monthly crashes per capita involving nonHispanic white drivers and a 2.1 percent decrease in monthly total crashes per capita involving non-Hispanic black drivers. In the long term, a 10 percent increase in gasoline prices is associated with a decrease in monthly crashes per capita involving non-Hispanic white drivers by 1.8 percent and by 2.0 percent for non-Hispanic black drivers. Gasoline prices do not

have significant effects on crashes involving Hispanics (see Appendix). The results are limited, because Hispanics make up a very small proportion of the Alabama population. Only 4 percent of the population was of Hispanic origin in 2010, which weakens the robustness of the results. The elasticity differences between the short term and long term were not statistically significant for any racial/ethnic group (Table III). Within the short-term and long-term effects separately, the elasticity differences across the 3 racial/ethnic groups were not significant either (Table IV). DISCUSSION A large body of literature has found that economic conditions have important effects on the occurrence of traffic crashes. However, gasoline prices have not been given much attention in existing studies of the economics of traffic safety. A limited number of studies have found that higher gasoline prices lead to drops in the incidence of traffic crashes. Nevertheless, most of the literature has been focused either on fatal crashes only or on all crashes but observed over a short time period. Though fatalities are the gravest outcome of traffic crashes, they occur in only a small proportion of total traffic crashes and therefore are not representative of gasoline price effects on the overall level of traffic crashes. A few studies examined gasoline price effects on all traffic crashes, but they used only a short time period. This study examined gasoline price effects on all traffic crashes in Alabama from 1999 to 2009, a period covering both economic growth and decline, and the possible demographic variations of the effects. Overall gasoline prices seem to have effects on reducing total crashes and crashes involving specific age, gender, and racial/ethnic groups (except Hispanics due to data limitation) in both the short term and the long term. Gasoline prices also have a stronger effect in reducing crashes



involving drivers aged 16 to 20 than crashes involving drivers aged 31 to 64 and 65+ in the short term. The effects, however, were not statistically different across other demographic groups; the effects were not statistically different between the short term and long term for any individual demographic group either. The findings were consistent with those from some previous studies but were inconsistent with those from others. Teenaged drivers are more vulnerable to gasoline price increases because they have lower incomes relative to older drivers; this finding is consistent with those from some previous studies (e.g., Chi et al. 2010; Morrisey and Grabowski 2011). Older drivers have accumulated more assets to weather unexpected economic difficulties such as gasoline price fluctuations and they are more forward-looking on average but also have limited opportunities to reduce their driving due to work and family responsibilities. In terms of gender, gasoline prices have similar effects on crashes involving male and female drivers in both the short term and the long term; this finding differs from the Mississippi study (Chi et al. 2010) that found that the effect more strongly reduced crashes involving female drivers than crashes involving male drivers. In terms of race/ethnicity, gasoline prices do not have significantly different effects on crashes involving nonHispanic white and non-Hispanic black drivers in the short term or the long term: this finding differs from the Mississippi study (Chi et al. 2010) in which the short-term effect was stronger on crashes involving non-Hispanic white drivers than crashes involving non-Hispanic black drivers but the long-term effect was the opposite. The findings suggest that higher gasoline prices lead to fewer traffic crashes. If gasoline prices had remained constant since 1999, how many traffic crashes would not have occurred or how many more crashes would have occurred? The estimated elasticities from the findings can be used to predict the number of crashes at an assumed level of gasoline prices (Figure 2). If gasoline prices had remained at the 1999 level of $1.41 from 1999 to 2009, applying the estimated elasticities would result in a predicted increase in total crashes of 169,492 (95% CI: 107,511–231,478). That is an increase of 11.3 percent (95% CI: 7.2%–15.4%) from the actual number of crashes. The findings suggest that if decision makers wish to reduce traffic crashes, increased gasoline taxes represent a possible option because higher gasoline prices reduce traffic crashes directly. The additional tax revenues can be invested in mass transit systems. This idea echoes recent calls for raising gasoline taxes as one way to reduce the federal deficit. For example, the bipartisan National Surface Transportation Infrastructure Financing Commission of the U.S. Congress (2009) recommended increasing the federal tax on gasoline from 18.4 cents per gallon to 28.4 cents per gallon in order to increase highway funding and reduce budget deficits. If such recommendations are accepted, higher gasoline tax revenues can be used not only for reducing budget deficits but also for investing in mass transit systems, which in turn increases traffic safety.

481

Figure 2 Actual and predicted traffic crashes assuming 1999 gasoline prices, 1999 to 2009, Alabama. The predicted traffic crashes were calculated on the basis of the elasticities for gasoline prices at both current time and 1-year lag from Table II. Error bars are in 95 percent confidence intervals.

However, increasing gasoline taxes would increase travel costs for people who use personal vehicles. People who own personal vehicles and have limited travel budgets would be particularly vulnerable to gasoline price/tax increases. This would not only affect low-income and minority groups but also the middle class, which drives a lot, although high-income drivers and vehicle owners may not feel the pain of higher gasoline prices/taxes. This further leads to equity concerns. Therefore, if increased gasoline taxes are used in order to reduce the incidence of traffic crashes, it would be important to invest in mass transit systems so that people who are vulnerable to gasoline price/tax increases can switch to convenient public transportation. Ample opportunity for further research remains along the thematic lines of this study. First, including additional economic variables could provide a more thorough understanding of the role of gasoline prices as an economic variable in affecting traffic safety. The results of this study are limited by the small number of explanatory variables. Other economic variables such as median household income are also found to affect traffic safety and should be included when data become available (Traynor 2009). Second, this study is focused on the demographic variation in gasoline price effects on total crashes. Analyzing the effects by crash types (such as fatal crashes, injury crashes, property damage–only crashes, and drunk driving crashes), vehicle types (such as newer models versus older models and cars versus motorcycles), and urban–rural differences may provide further insights into the effects of gasoline prices on traffic safety. If gasoline price effects vary greatly by crash types, vehicle types, or along the urban–rural continuum, the policy benefits of increasing gasoline taxes would be very different. For example, if the gasoline price effect on less severe crashes is much higher than that on severe crashes, the policy benefits of increasing

482

CHI ET AL.


gasoline taxes would be discounted. This should be addressed in future research. Third, this study is focused on the state of Alabama only, a relatively rural southern state in the United States. Thus, the results may not be generalizable to other regions and metropolitan areas. Future research could examine the effects in such areas. Doing so would provide a more comprehensive understanding of gasoline price effects on traffic safety. ACKNOWLEDGMENTS The authors thank Angie L. Watkins and Jesse Norris of the University of Alabama for deriving traffic crash data. Appreciation is extended to David C. Viano, the Editor, and 3 anonymous reviewers for their many helpful comments. REFERENCES Abdel-Aty MA, Radwan AE. Modeling traffic accident occurrence and involvement. Accid Anal Prev. 2000;32:633–642. Arnett JJ. Developmental sources of crash risk in young drivers. Inj Prev. 2002;8(suppl 2):17–23. Chi G, Cosby AG, Quddus MA, Gilbert PA, Levinson D. Gasoline prices and traffic safety in Mississippi. J Safety Res. 2010;41:493–500. Chi G, Zhou X, McClure TE, et al. Gasoline prices and their relationship to drunk-driving crashes. Accid Anal Prev. 2011;43:194–203. Clogg C, Petkova E, Haritou A. Statistical methods for comparing regression coefficients between models. AJS. 1995;100:1261–1293. Fu H. Identifying repeat DUI crash factors using state crash records. Accid Anal Prev. 2008;40:2037–2042. Grabowski DC, Morrisey MA. Gasoline prices and motor vehicle fatalities. J Policy Anal Manage. 2004;23:575–593. Grabowski DC, Morrisey MA. Do higher gasoline taxes save lives? Econ Lett. 2006;90:51–55. Graham DJ, Glaister S. Spatial variation in road pedestrian casualties: the role of urban scale, density and land-use mix. Urban Stud. 2003;40:1591–1607. Horner RD, Day GM, Lanier AP, Provost EM, Hamel RD, Trimble BA. Stroke mortality among Alaska native people. Am J Public Health. 2009;99:1–5. Joksch HC. The relation between motor vehicle accident deaths and economic activity. Accid Anal Prev. 1984;16:207–210. Kenkel DS. Drinking, driving, and deterrence: the effectiveness and social costs of alternative policies. J Law Econ. 1993;36:877–913. Kulmala R. Safety at Rural Three and Four-Arm Junctions, Development and Application of Accident Prediction Models. Espoo, Finland: VTT Publications; 1995. Leigh JP, Geraghty EM. High gasoline prices and mortality from motor vehicle crashes and air pollution. J Occup Environ Med. 2008;50:249–254.

Leigh JP, Wilkinson JT. The effect of gasoline taxes on highway fatalities. J Policy Anal Manage. 1991;10:474–481. Long JS. Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks, Calif: Sage Publications; 1997. Lord D. The Prediction of Accidents on Digital Networks: Characteristics and Issues Related to the Application of Accident Prediction Models [dissertation]. Toronto, Ontario: University of Toronto; 2000. McGwin GJ, Brown DB. Characteristics of traffic crashes among young, middle-aged, and older drivers. Accid Anal Prev. 1999;31:181–198. Miaou SP. The relationship between truck accidents and geometric design of road sections: Poisson versus negative binomial regressions. Accid Anal Prev. 1994;26:471–482. Morrisey MA, Grabowski DC. Gas prices, beer taxes and GDL programmes: effects on auto fatalities among young adults in the US. Appl Econ. 2011;43:3645–3654. National Surface Transportation Infrastructure Financing Commission. Paying Our Way: A New Framework for Transportation Finance. Washington, DC: US Congress; 2009. Available at: http://financecommission.dot.gov/Documents/NSTIF Commission Final Report Advance%20Copy Feb09.pdf. Accessed July 25, 2012. Partyka SC. Simple models of fatality trends using employment and population data. Accid Anal Prev. 1984;16:211–222. Quddus MA. Modelling area-wide count outcomes with spatial correlation and heterogeneity: an analysis of London crash data. Accid Anal Prev. 2008;40:1486–1497. Sivak M. Mechanisms involved in the recent large reductions in US road fatalities. Inj Prev. 2009;15:205–206. Sivak M, Schoettle B. Toward understanding the recent large reductions in US road fatalities. Traffic Inj Prev. 2010;11:561–566. Southeast Regional Climate Center. Monthly statewide precipitation and temperature averages 1895–2012. Available at: http://www. sercc.com/climateinfo/monthly seasonal.html. Accessed December 12, 2011. Traynor TL. The relationship between regional economic conditions and the severity of traffic crashes. Traffic Inj Prev. 2009;10:368–374. US Bureau of Labor Statistics. Local area unemployment statistics. 2009. Available at: http://data.bls.gov/PDQ/servlet/SurveyOut putServlet?data tool=latest numbers&series id=LASST28000003. Accessed June 24, 2009. Wagenaar AC. Effects of macroeconomic conditions on the incidence of motor vehicle accidents. Accid Anal Prev. 1984;16:191–205. Williams AF. Teenage drivers: patterns of risk. J Safety Res. 2003;34:5–15. Williams AF, Preusser DF, Ferguson SA. Fatal crashes involving 16-year-old drivers: narrative descriptions. J. Traffic Med. 1998;26:11–17. Wilson FA, Stimpson JP, Hilsenrath PE. Gasoline prices and their relationship to rising motorcycle fatalities, 1990–2007. Am J Public Health. 2009;99:1753–1758.

483


APPENDIX Results of negative binomial regression models for traffic crashes per capita, 1999 to 2009, Alabama Age Total

16–20

Coef.a


Explanatory variables Average gasoline price during the month when the crashes occurred Gasoline price at a 1-year lag State unemployment rate Temperature Rainfall Time trend Exposure (population) Constant Statistics Log likelihood AICc BICd Observations

(95% CI)b

P

Coef. (95% CI)

21–25 P

Coef. (95% CI)

P

−0.00096 (−0.00132, −0.00061)