Section 1 - Introduction

Motor vehicle fatalities in the United States have declined dramatically during the last two decades due to a number of factors including technological improvements in vehicles and roadways, increased use of seatbelts, and decreased alcohol-impaired driving. Nonetheless, motor vehicle fatality rates among teenaged and older drivers, passengers and pedestrians continue to exceed that of the middle-aged, sometimes by a factor of more than five to one (see Figure 1-1).

Figure 1-1: Fatal crash involvement by age: 1995-96

Much of the success in reducing motor vehicle related fatalities has been as a result of declines in deaths among teens. As Figure 1-2 shows, the death rate for teens has declined by more than one-third since its height in 1979. Much of this can be attributed to the enactment of higher minimum drinking ages and higher alcohol taxes (Dee, 2001; Dee and Evans, 2001), although the tax effects have been seriously challenged by recent work (e.g., Dee, 1999). In contrast, motor vehicle fatality rates among older drivers have been increasing since 1980, particularly among those aged 75 and above. The aging of the U.S. population over the coming 25 years makes these rates particularly distressing. The Insurance Institute for Highway Safety reports: “Drivers aged 65 and older … are expected to account for as much as 25 percent of total driver fatalities in 2030, compared to 14 percent currently” (Lyman et al. 2002).

While the states have been active in implementing graduated drivers license and other programs directed at younger drivers, they have done relatively little with respect to older drivers. A handful of states have imposed shorter intervals between license renewals for older residents, often with a requirement for in-person renewal. Some states require vision and road tests for older drivers seeking to renew their licenses. Most states, however, treat all those over 18 or 21 to the same licensure and renewal provisions, although there are major differences across states. The states have, of course, enacted a series of relatively new laws that apply to all drivers. These include mandatory seatbelt use, lowering of the legal definition for driving while intoxicated from a blood alcohol concentration (BAC) of .10 to .08, and provisions for the administrative suspension of a driver’s license for driving under the influence (DUI). Contrasted with these actions has been the increase in legal speed limit to 65, 70 or 75 miles per hour on rural interstate highways.

Figure 1-2: Driver fatality rates, 1975-99

There has been almost no research on the effects of these policy changes on older drivers (Grabowski and Morrisey 2001). Elderly individuals may be affected directly, by getting less safe older drivers off the road, or they may be affected indirectly as BAC laws, for example, may keep younger alcohol-impaired drivers off the road. The purpose of this paper is to examine the effects of the changes in state motor vehicle laws on the number of fatalities among older drivers and older persons more generally. This is accomplished by estimating a series of negative binominal regressions of motor vehicle fatalities for those aged 65 to 74, 75 to 84 and 85 and older using the Fatality Analysis Reporting System (FARS) data over the period 1985 to 2000.

Section 2 -Background

There has been remarkably little research on the effects of state driving laws on the fatality rates of older people. The principal exceptions are in the areas of license renewal policies and vision testing. Rock (1998) examined the effects of shorter licensure periods for older adults and the removal of road tests for those under age 75 in Illinois in 1989. Comparing changes 1987-89 to 1995, he found no effects of these changes on crashes, fatal crashes, crash rates or licensure rates of older drivers.

Shipp (1998) compared occupant motor vehicle fatalities for those over age 60 in states with and without a vision-related re-licensing statute. He found that there would be a 12.2 percent reduction in fatalities over the 1989-1991 period if 8 of the 10 states without such a law were to enact one. Levy, Vernick and Howard (1995) used 1985 to 1989 FARS data to examine the effects of license renewal periodicity, visual acuity, knowledge and road tests on older (aged 70+) fatalities. They concluded that state-mandated tests of visual acuity, adjusted for license renewal period, were associated with lower fatal crash risk for older drivers. Knowledge tests provided further, albeit statistically insignificant reductions. Levy (1995) used the same data to examine licensure rates among older adults. He concluded that vision and road tests were associated with reduced elderly licensure.

There are two concerns with this literature, one methodological and one interpretative. Methodologically, with the exception of Rock (1998), the studies are identified solely by the cross-sectional variation in state laws, which may result in biased estimates if unobserved factors are correlated with this cross-state variation in the adoption of the state laws. With a cross-sectional research design, one cannot rule out the presence of a third unobserved factor that affects both the implementation of the state law and the motor vehicle fatality rate. The Rock study has the related problem that it only examines law changes within a single state, and thus does not control for any broader national trends in motor vehicle safety over the study period. Although the estimation strategy will be discussed in detail below, this study addresses both of these issues by employing a fixed-effects model, which exploits within-state variations in the regressors and outcomes, and as a result, automatically controls for time-invariant factor that differ across states.

Interpretively, the more clinical literature has argued that visual acuity, as measured in state driver’s license laws, is not the key factor. Instead, Ball et al. (1993) and Owsley et al. (1998) demonstrate that “useful field of view” is the key visual factor. This suggests that the effect of the state’s vision test may be to encourage some older residents to seek medical care before or after a driver’s license exam. This may result in decisions to stop or limit their driving experience, or to obtain corrective vision-related care that improves their vision in relevant dimensions. Recent work by Owsley et al. (2002), for example, found that those older adults who had undergone cataract surgery had a crash involvement ratio nearly half that of similar older adults who had not.

There are remarkable differences across the 48 contiguous states in the nature of their motor vehicle laws. Importantly, most states do not distinguish between older and younger drivers in the license renewal policies. As of 2000, 45 states require in-person renewals for older drivers, 40 require vision exams upon renewal and three require a road test. A handful of these states, four, have changed one or more of these laws at least once since 1985. The average renewal cycle in 2000 was 4.4 years for older adults and four states have shortened the period for older drivers since 1985. However, most states have the same renewal provisions for all adult drivers and there has been a trend towards lengthening the renewal period for adult drivers in recent years. Thus, 15 states lengthened the renewal period for all adults (including elderly drivers) between 1985 and 2000. Only 12 states have accelerated renewal procedures for older individuals. In part, this is because legislation limiting older drivers is controversial. Connecticut, for example, required a vision test for older drivers in 1992, but delayed its implementation by legislation in virtually every legislative session after that (Connecticut General Statutes 2001).

Currently all states except New Hampshire have mandatory seatbelt use laws covering front-seat occupants. All of these laws were enacted in the 1985 to 1995 period. Only 18 of the laws are “primary” allowing the police to stop vehicles solely for belt-law violations. Maximum fines for a first offense vary from $5 to $75 dollars. While there have been a number of studies examining the effects of seatbelt use (see particularly Evans and Graham 1991) none has examined the effects on older drivers.

Maximum speed limits of 55 miles per hour (mph) were set by federal law in the early 1970s. In 1987 states were given flexibility to raise limits to 65 mph on rural interstates. In 1995 Congress repealed the federal statute. Today, Hawaii is the only state with a 55 mph limit on rural interstates. Twenty states have rural limits of 65 mph, 14 have set 70 mph and 11 have 75 mph. Farmer, Retting and Lund (1999) reported that the 1995 repeal increased interstate fatalities by 15 percent in the states that raised their limits. Dee and Evans (2001) did not find a statistically significant effect of the initial move to 65 mph limits following the 1987 change in federal legislation. Others have argued that higher interstate speed limits, while increasing deaths on interstates may reduce overall deaths because secondary roads become safer. Lave and Elias (1994) concluded that fatality rates across all roads actually declined from 3.4 to 5.1 percent with the increase to 65 mph. There appears to be no research on the effects of speed limit laws on older drivers.

By 1988 all states except Massachusetts and the District of Columbia have had “per se” laws that make it a crime to drive with a BAC above a prescribed level. Between 1988 and 2000, 14 states lowered the legal threshold from .10 to .08; in 2000 Congress enacted legislation making this lower level the national standard. In addition, 41 states and the District of Columbia have administrative license suspension, in which licenses are taken before conviction if a driver’s BAC exceeds a specified level or if the driver refuses to take the test. There have been a number of studies examining the effects of these laws, and others dealing with dram shop, zero tolerance, and minimum drinking ages, among others (Grabowski and Morrisey, 2001). In general, the evidence suggests that the laws, particularly minimum drinking ages, lower BAC levels and administrative suspension of licenses have reduced motor vehicle fatalities. There are methodological concerns with many of these studies, but again, there has been no attention devoted to the effects on older drivers.

Section 3 - Data

The source of all motor vehicle fatality information within this study is the Fatality Analysis Reporting System. The FARS, collected by the National Highway Traffic Administration, is a census of all motor vehicle crashes involving a fatality. See www.fars.nhtsa.dot.gov. To be included in this census of crashes, a crash had to involve a motor vehicle traveling on a traffic-way customarily open to the public and must result in the death of a person (either the occupant of a vehicle or a non-motorist) within 30 days of the crash. The FARS contains detailed information on the vehicles, drivers, occupants, and non-occupants involved within the crash. We construct age-cohort state and year specific fatality counts from the FARS.

Table 3-1 contains summary information from our database. The variable means and standard deviations are provided. The driver and traffic (drivers plus passengers) fatality rates are highest for the 75-84 cohort. The driver fatality rate is 15.30 per 100,000 individuals aged 75-84 in the population and the overall fatality rate is 27.39. For the 65-74 cohort, the driver fatality rate is 11.29 and the overall fatality rate is 18.78. Finally, for the 85+ cohort, the driver fatality rate is 12.45 and the overall rate is 26.15.

Table 3-1: Descriptive statistics: state panel data 1985-2000 (N=768)

Variable	Mean	SD
Driver fatality rate per 100,000, ages 65-74	11.29	4.74
Driver fatality rate per 100,000, ages 75-84	15.30	6.00
Driver fatality rate per 100,000, ages 85+	12.45	7.44
Traffic fatality rate per 100,000, ages 65-74	18.78	6.71
Traffic fatality rate per 100,000, ages 75-84	27.39	8.49
Traffic fatality rate per 100,000, ages 85+	26.15	11.42
In-person license renewal required	0.88	0.27
Vision test required for license renewal	0.78	0.39
Road test required for license renewal	0.06	0.24
License renewal period (years)	4.19	1.13
Mandatory seat belt law – primary enforcement	0.19	0.39
Mandatory seat belt law – secondary enforcement	0.58	0.49
65 MPH speed limit	0.52	0.50
70+ MPH speed limit	0.17	0.38
Illegal per se at 0.08 BAC	0.16	0.36
Administrative license suspension	0.61	0.48
State unemployment rate	5.58	1.79
Real state personal income per capita	14.40	2.32

Note: Alaska, Hawaii and the District of Columbia are omitted.

Data on motor vehicle laws and their dates of enactment were obtained from several sources. We began with compilations of laws from the Insurance Institute for Highway Safety (2002). These were compared with data collected (and shared) by Thomas Dee, an economist at Swarthmore College who has authored several papers on the economics of motor vehicle safety (e.g., Dee, 1999, Dee 2001). We then conducted a telephone survey of all state departments of motor vehicles to confirm the laws, resolve inconsistencies, and to obtain the dates of changes in the laws. In several instances we used codes of annotated state statutes and specific legislative acts available on the web to determine when laws were enacted.

Annual unemployment rate data were obtained from the U. S. Department of Labor, Bureau of Labor Statistics (2002). Annual per capita income information was obtained from the U. S. Department of Commerce, Bureau of Economic Analysis (2002) and adjusted for inflation using the Consumer Price Index, all items. Population estimates by age group were obtained from the U.S. Census Bureau (1999, 2001).

Section 4 – Empirical Methods

The empirical model exploits the panel nature of the FARS data to examine the effect of state motor vehicle laws on elderly motor vehicle fatalities. During the period January 1985 through December 2000, we had access to 768 observations (i.e., 48 contiguous states times 16 years). The basic specification for the empirical results presented here is:

(1)

where F_st refers to the motor vehicle fatality count in state s of year t, L_st is a vector of state motor vehicle laws, Z_st includes an intercept and a set of exogenous controls, v_s is a state fixed effect, w_t is a year-specific intercept, and ε_st is the randomly distributed error term. β represents the coefficients on the state motor vehicle laws and γ represents the coefficients on the control variables within the model. The state fixed effects control for any fixed state-specific omitted variables correlated with the propensity to change the motor vehicle fatalities. Such variables may include, for example, basic political and religious sentiments and geographic characteristics. The year dummies control for national trends in motor vehicle fatalities that may be correlated with changes in state laws such as federal motor vehicle policies and the progress of motor vehicle and road safety technology. Thus, the basic identification strategy implicit in equation (1) purges the unobserved and potentially confounded cross-sectional heterogeneity by relying on the within-state variation in state motor vehicle laws and by using states that did not change their laws as a control for unrelated time-series variation.

Full fixed effects models of this sort have seldom been used to examine the effects of state laws on vehicular fatalities. See Dee (2001) and Dee and Evans (2001) for exceptions. As a consequence, we present more traditional models without year or state fixed effects and then add these vectors to examine the implications of the specification on the findings and policy implications.

The specification includes ten major state motor vehicle laws that may be important in explaining changes in motor vehicle fatality rates among the elderly population. The first is whether in-person renewal is required for older drivers. Some have argued that in-person renewal allows the examiner to determine whether the applicant has obvious physical or cognitive impairments that can trigger a requirement for a medical examination prior to the renewal of a license. If so, fatalities will be reduced. We also include a measure of the periodicity of license renewal. States with a longer period between renewals reduce the opportunities to observe the physical and mental condition of older license applicants whose abilities may rapidly decline over time. If so, states with longer renewal periods will have more elderly motor vehicle fatalities.

Two binary variables are included for whether a vision or road test is required of older adults at the time of renewal. In some states only elderly applicants are subject to these conditions. In some others, all drivers are. As with the renewal cycle, we make no distinction between whether the laws apply only to older adults. Rather, we focus on the constraints imposed upon older individuals regardless of whether others are also affected. We expect that shorter renewal cycles will reduce fatalities.

Two binary indicators are included for mandatory seat-belt laws. Seat-belt laws with primary enforcement allow the police to directly cite a motorist for not wearing a belt. Under secondary enforcement of a seat-belt law, a motorist can only be cited for a violation if they are pulled over for some other reason. We expect that seatbelt laws will reduce fatalities among older drivers and others and that primary enforcement will have a larger effect than secondary enforcement.

The vector of laws also includes two binary indicators related to drunk driving. The first represents those states that make it illegal per se to drive with a blood alcohol concentration (BAC) of 0.08. The second represents whether the state’s licensing authority is allowed to suspend before any court action related to a charge of drunk driving. There is recent evidence that the enactment of all these policies may be important in decreasing motor vehicle fatalities (Dee, 2001), thus we expect that they will reduce motor vehicle fatalities among older adults. In early specifications, we experimented with the inclusion of dram shop laws that make it illegal for a bartender or waiter to serve alcohol to a drunken patron and mandatory jail time for DUI offenses. They were not statistically significant and did not affect the coefficients of other variables.

Finally, two other binary indicators identify those states that have increased their maximum rural speed limit to 65 miles per hour or to 70 or more miles per hour. The effect of these laws is ambiguous. On the one hand, such laws should increase the consequences of crashes by raising average speed. On the other hand, higher interstate speed limits may induce some time-conscience drivers to use secondary roads less, increasing their safety. Moreover, at higher speed limits more older drivers may give up driving, or forego driving on interstate highways.

The exogenous controls introduced in the vector Z_st include two macroeconomic factors: the annual unemployment rate and the annual per capita income level. Several studies have recognized the important of including these macroeconomic factors within analyses of state motor vehicle fatality rates (e.g., Evans and Graham, 1988; Ruhm, 1996).

We estimate fatality models for three cohorts of older drivers, those aged 65 to 74, 75 to 84 and 85 and older. This is done because driving skills deteriorate as people age and the effects of various state statutes may vary substantially across age groups. In addition, within each age cohort, we estimate the model on all age-specific motor vehicle fatalities to explore the extent to which there are effects of the laws on all older adults in addition to drivers.

In the fatality models, F_st represents the count of annual motor vehicle fatalities in the state-year. In modeling the motor vehicle fatality rate, a semi-log specification is often employed due to skewness in the motor vehicle fatality rate (e.g, Dee, 2001). In age specific fatality models; however, there are some cells with zero fatalities. In order to accommodate these zero cases, fatalities are treated as a count on the left-hand side of the model and the natural log of the age-specific population is included on the right-hand side of the model. Conventional count data models do not generate consistent estimates when cross-sectional fixed effects are introduced because of the “incidental parameter” problem. Thus, we employ the conditional maximum likelihood approach for negative binomial models developed by Hausman, Hall and Griliches (1984). The negative binomial model is less restrictive than a Poisson regression because it accommodates the presence of over dispersion in the counts. Similar to a semi-log model, the estimates generated by the negative binomial model can be interpreted as the proportionate change in the given motor vehicle fatality count.

A final methodological point concerns the likely presence of heteroskedasticity in grouped state-year data, which may bias the estimates of the parameter standard errors. We addressed this potential issue by adjusting the standard errors using the Huber-White robust estimator.

Section 5 – A Primer on Interpreting Regression Results

For those readers less familiar with interpreting the results from regression analyses, this section provides a brief guide to facilitate interpretation of the results presented in Section 6 of this report. This report uses an analytic tool termed regression analysis. Basically, this approach allows the analyst to model the relationship between two variables, controlling for other important variables of interest. For the purposes of our study, we are interested in the relationship between state motor vehicle laws and the fatality rate for older individuals. Thus, we have estimated regression models to examine the effect of these state laws on older driver fatalities for the age cohorts 65-74, 75-84 and 85+. We also estimate these models for all older traffic deaths, regardless of whether the older individual was a driver or passenger.

For each of the models, the regression analysis controls for a series of other variables and allows us to examine the effect of a particular state law on the fatality rate of interest. For example, we can examine the effect of a primary seatbelt law on traffic fatalities holding all those other variables in the model constant. Tables 6-1, 6-2, and 6-3 present the results from our regression models. The results in the table are the estimates of the coefficients β from the equation in the previous section. Because we use a particular regression model (i.e., the negative binomial model), the coefficients presented in the next section can be interpreted as a proportionate change in the given fatality rate. Let’s take the third column of Table 6-1 as an example. The coefficient estimate for the effect of a primary seatbelt law is –0.136. This result implies that the adoption of a primary seatbelt is associated with a 13.6% reduction in the fatality rate for drivers aged 64-74, controlling for all the other variables shown in the left-hand column of the table. Additionally, we control for what are termed year and state fixed effects in certain specifications of the model. Basically, these effects control for any “unobserved” time trend or state-specific factors that may influence the fatality rate. By unobserved effects, we mean factors that are unobserved to the analyst, but either changing nationally over time (thus, the year fixed effects) or constant to the states over time (thus, the state fixed effects). An example of an unobserved time-trend that may affect motor vehicle fatalties would be the introduction of safety airbags in recent years. An example of an unobserved state-specific effect that may affect motor vehicle fatalities would be inclement weather in northern states. Clearly, the specifications in columns 3 and 4 of the tables that control for these effects will be preferred to the specifications in columns 1 and 2 that do not.

A final important issue in interpreting our results involves the precision of the coefficient estimates. The number in parentheses below each coefficient estimate is a standard error of the estimate. For example, the standard error on the estimate of the effect of a primary seatbelt is 0.038 in the third column of Table 6-1. As a basic rule of thumb, the smaller the standard error, the greater the precision of the estimate. The asterisks within the tables help identify whether the result is statistically different from zero. The coefficient estimates without asterisks are not statistically different from zero. The estimates with asterisks are statistically different from zero.

Section 6 - Results

Table 6-1 presents the results for older adults aged 65 to 74. The first three columns relate to fatalities among older drivers only, while column (4) reports our preferred specification for all older adults in the age cohort. With this and Tables 6-2 and 6-3, we present three alternative specifications. Column (1) reports results that do not control for either year or state fixed effects. Column (2) adds a vector of year fixed effects; and column (3) includes both state and year fixed effects.

Comparison of the parameter estimates in columns (1) and (2) illustrates the importance of controlling for national trends. The effect of raising the speed limit to 70 mph or greater is a case in point. Column (1) indicates that such a policy raised driver fatalities in the 65-74 age cohort by over 28.8 percent, relative to those states and years when rural speed limits were set at 55 mph. However, this estimated effect is reduced by half, to 14.4 percent, when one includes year dummies to control for national trends as is done in Column (2). Over time, motor vehicles and roadways have become safer due to technological improvements. The results in Column (2) suggest that the failure to control for this trend will tend to bias the estimates of the effectiveness of state actions.

Column (2) essentially compares states that have changed their policies with those that have not (and with the pre-change period of those states that did adopt a change). Many studies have used this approach and it may be reasonable to do so if there are no unobserved factors that influence the adoption of these state laws. This is a high standard to meet. We noted earlier, for example, the controversy associated with the enactment of vision testing for older adults in Connecticut. As another example, one might argue that a 70 mph speed limit is enacted in one state because drivers are already routinely traveling at that speed. If so, generalizing from the effects in that state to others, where most drivers travel at, say 60 mph, is likely to give misleading results. Thus, the parameter estimate for the enactment of a 70 mph maximum rural speed limit in Column (3), which does control for state effects, indicates a 1.6 percent increase in fatalities among older drivers, but is no longer statistically significant. If drivers were driving faster prior to the legal change in these states, then this result should not be overly surprising. Thus, we focus for the most part on the findings from the full fixed-effects specifications in Columns (3) and (4).

We find that the enactment of mandatory seat belt laws has reduced motor vehicle fatalities among drivers aged 65 to 74. Laws that allow police to stop vehicles solely for a seat belt violation, on average reduce fatalities by 13.6 percent among drivers and 7.6 percent among all older adults in the cohort. Even secondary enforcement is effective. Fatalities are reduced by 6.5 percent for drivers and 4.6 percent for all older adults in the age range.

Table 6-1: Determinants of auto fatalities ages 65-74, 1985-2000 (N=768)

Independent Variable

Driver Fatalities

(1)

Driver Fatalities

(2)

Driver Fatalities

(3)

Overall Fatalities

(4)

In-person renewal

0.007

(0.038)

-0.010

(0.039)

----

Vision test

-0.099***

(0.030)

-0.073**

(0.030)

----

Road test

-0.087**

(0.039)

-0.091**

(0.039)

----

Renewal cycle (years)

0.003

(0.008)

-0.0006

(0.0079)

0.007

(0.016)

0.014

(0.013)

Primary seat belt law

0.037

(0.037)

-0.035

(0.041)

-0.136***

(0.038)

-0.076**

(0.031)

Secondary seat belt law

0.032

(0.031)

-0.027

(0.034)

-0.065**

(0.026)

-0.046**

(0.021)

65 MPH speed limit law

0.080***

(0.025)

-0.015

(0.041)

-0.077***

(0.027)

-0.059***

(0.021)

70+ MPH speed limit law

0.288***

(0.036)

0.144***

(0.055)

0.016

(0.043)

0.035

(0.035)

Illegal per se at 0.08 BAC

-0.006

(0.028)

-0.018

(0.028)

-0.019

(0.027)

0.007

(0.022)

Administrative license suspension

-0.006

(0.024)

-0.007

(0.024)

-0.031

(0.025)

-0.036*

(0.022)

State unemployment rate

-0.043***

(0.007)

-0.037***

(0.008)

-0.020***

(0.007)

-0.026***

(0.006)

Real per capita income ($1,000s)

-0.106***

(0.006)

-0.117***

(0.006)

0.047**

(0.020)

0.025

(0.017)

Log (Population 65-74)

0.921***

(0.014)

0.934***

(0.014)

0.647***

(0.173)

0.715***

(0.152)

Year Fixed Effects

Yes

State Fixed Effects

Yes

Pseudo R-Squared

0.25

0.26

0.34

0.35

Dependent Variable Mean

36.18

63.22

Notes: Huber-White standard errors are presented in parentheses.

* = Statistically significant at 10% level; ** = statistically significant at 5% level;

*** = Statistically significant at1% level.

Implementation of a 65 mph speed limit on rural roads is estimated to reduce older driver fatalities rates by 7.7 percent and motor vehicle fatalities among all older adults aged 65 to 74 by 5.9 percent. This may result from faster drivers moving back from weakly patrolled secondary roads to interstates when the speed limits were raised. Some have speculated that this could reduce total fatalities (Grabowski and Morrisey 2001). The data suggest that there was no statistically significant effect of raising the speed limit to 70 mph or higher in those states that did so.

States that enacted BAC levels of .08 did not reduce age 65 to 74 driver fatalities; the effect is small in magnitude and lacks statistical significance. The use of administrative license suspension similarly did not reduce driver fatalities, but did reduce motor vehicle fatalities among those aged 65 to 74 by 3.6 percent. This suggests that the effects were indirect, keeping other intoxicated drivers off the road.

The full fixed-effects models find no effect of shorter renewal periods on fatalities for this age cohort. The fixed effects models omit the vision test, road test and in-person renewal variables because there were not enough states that implemented changes in these policies over the study period to allow a meaningful analysis. An examination of Column (2), however, suggests that the states with vision and road tests for older drivers may be reducing fatalities substantially, but the inability to control for other relevant state variables limits the confidence one can put in these findings.

Many studies have shown that the state of the local economy influences driving behavior and fatality rates. Our results are no exception. When the state’s economy is better, measured as a lower unemployment rate and/or a higher real per capita income, fatalities increase. This may be due to more drivers on the road, more travel per person, and/or more reckless behavior.

Finally, a larger population of older adults in the age cohort does increase fatalities and does so less than proportionately. A one percent increase in the age 65-74 population in the average state increases older adult fatalities by .715 percent.

Table 6-2 presents the same analyses for those aged 75 to 84. In general, the results are similar, although the effects tend to be somewhat smaller and often are less statistically significant. Primary enforcement of seat belt laws reduces driver fatalities (nearly 11 percent).

Table 6-2: Determinants of auto fatalities ages 75-84, 1985-2000 (N=768)

Independent Variable

Driver Fatalities

(1)

Driver Fatalities

(2)

Driver Fatalities

(3)

Overall Fatalities

(4)

In-person renewal

-0.068*

(0.039)

-0.065*

(0.039)

----

Vision test

-0.047

(0.033)

-0.025

(0.032)

----

Road test

-0.161***

(0.041)

-0.158***

(0.041)

----

Renewal cycle (years)

0.004

(0.008)

0.001

(0.008)

0.008

(0.014)

0.017

(0.011)

Primary seat belt law

0.051

(0.038)

-0.021

(0.040)

-0.109**

(0.048)

-0.044

(0.037)

Secondary seat belt law

0.043

(0.032)

-0.016

(0.033)

-0.046

(0.032)

-0.018

(0.025)

65 MPH speed limit law

0.173***

(0.025)

0.088**

(0.041)

0.033

(0.029)

0.013

(0.023)

70+ MPH speed limit law

0.344***

(0.038)

0.221***

(0.058)

0.076*

(0.045)

0.111***

(0.035)

Illegal per se at 0.08 BAC

-0.005

(0.030)

-0.018

(0.030)

0.004

(0.032)

-.005

(0.023)

Administrative license suspension

0.013

(0.024)

0.005

(0.029)

-0.020

(0.029)

-0.022

(0.022)

State unemployment rate

-0.029***

(0.007)

-0.028***

(0.008)

-0.002

(0.008)

-0.016***

(0.006)

Real per capita income ($1,000s)

-0.084***

(0.005)

-0.096***

(0.007)

0.041*

(0.022)

0.027*

(0.016)

Log (Population 75-84)

0.944***

(0.013)

0.955***

(0.014)

0.363*

(0.158)

0.295**

(0.139)

Year Fixed Effects

Yes

State Fixed Effects

Yes

Pseudo R-Squared

0.27

0.28

0.34

0.35

Dependent Variable Mean

30.04

56.15

Notes: Huber-White standard errors are presented in parentheses.

* = Statistically significant at 10% level; ** = statistically significant at 5% level;

*** = Statistically significant at1% level.

However, the speed limit laws affect this age cohort differently than the younger-older adults. Here the 70+ speed limits increased fatalities by 7.6 percent among drivers and 11.1 percent among 75-84 year old adults. There was no statistically significant effect of a 65 mph limit, although the point values suggest that deaths were increased. Again, the lower BAC level and administrative license suspension had no effects of motor vehicle fatalities.

Table 6-3 presents the same analyses for those aged 85 and older. The principal finding in this age cohort is that the primary enforcement of seat belt laws again reduces fatalities substantially, by 18.3 percent for very-old drivers and 14.4 percent for adults, overall. The rest of the parameter estimates lack statistical significance at the conventional levels. This is probably the result of the relative scarcity of fatalities in this age cohort and therefore, the potentially large random component associated with their fatalities.

Table 6-3: Determinants of auto fatalities ages 85+, 1985-2000 (N=768)

Independent Variable

Driver Fatalities

(1)

Driver Fatalities

(2)

Driver Fatalities

(3)

Overall Fatalities

(4)

In-person renewal

-0.151**

(0.064)

-0.143**

(0.060)

----

Vision test

-0.025

(0.049)

0.057

(0.048)

----

Road test

-0.039

(0.066)

-0.064

(0.064)

----

Renewal cycle (years)

0.042***

(0.011)

0.038***

(0.011)

0.020

(0.025)

0.032

(0.020)

Primary seat belt law

0.104*

(0.059)

-0.003

(0.065)

-0.183**

(0.078)

-0.144**

(0.057)

Secondary seat belt law

0.092*

(0.054)

0.020

(0.057)

-0.039

(0.057)

0.011

(0.040)

65 MPH speed limit law

0.312***

(0.041)

0.112*

(0.063)

0.034

(0.061)

-0.042

(0.040)

70+ MPH speed limit law

0.567***

(0.056)

0.230***

(0.084)

0.075

(0.085)

0.008

(0.061)

Illegal per se at 0.08 BAC

-0.026

(0.044)

-0.053

(0.041)

0.0001

(0.062)

0.0003

(0.042)

Administrative license suspension

0.054

(0.037)

0.029

(0.036)

-0.056

(0.049)

-0.015

(0.034)

State unemployment rate

-0.034***

(0.011)

-0.006

(0.013)

0.020

(0.015)

-0.005

(0.011)

Real per capita income ($1,000s)

-0.078***

(0.010)

-0.106***

(0.010)

-0.015

(0.039)

0.001

(0.028)

Log (Population 85+)

0.963***

(0.022)

0.978***

(0.022)

0.732**

(0.312)

0.694***

(0.203)

Year Fixed Effects

Yes

State Fixed Effects

Yes

Pseudo R-Squared

0.27

0.28

0.31

0.32

Dependent Variable Mean

8.23

18.06

Notes: Huber-White standard errors are presented in parentheses.

* = Statistically significant at 10% level; ** = statistically significant at 5% level;

*** = Statistically significant at1% level.

Section 7 - Discussion

This paper has examined the effects of a number of state motor vehicle policies on fatalities among older adults. Five issues emerge from the analysis.

First, mandatory seat belt laws have been a major factor in reducing fatalities among older adults, and particularly among older drivers. Primary laws, which allow police officers to stop vehicles solely for a seat belt offense, were estimated to reduce older driver fatalities by 13 to 18 percent. The magnitude of the life savings is comparable in the other older age cohorts. Even secondary enforcement seat belt laws work, although their effects are roughly half that of the primary enforcement laws. This suggests that while knowledge of the benefits of seat belt laws may reduce fatalities, the incentive that comes from mandatory requirements appears to be enough to get many older adults to wear seat belts. The states impose fines for first offence seat belt violations of $5 to $75 dollars. It would be instructive to see whether the mandate itself is sufficient, or whether compliance increases with the size of the penalty.

Second, we found no effect of lowering per se rules on BAC laws or administrative license suspension provisions on fatalities among older adults. It may be that these laws are rarely binding on older drivers in as much as older adults tend to drink less. More importantly, perhaps, people’s reduced tolerance for alcohol with advancing age may mean that the legal limit may not be sufficiently low to capture the same level of impairment in older drivers as in younger ones. Clearly, more work is needed on the policy implications of the differential effects of alcohol on the ability of older adults to drive safely.

Third, we found suggestive evidence that vision and road tests for 65 to 74 year olds, road tests and in-person renewals (where road or vision tests were not required) for those aged 75 to 84 and in-person renewals for those aged 85 and older were associated with fewer driver fatalities in their respective age cohorts. However, these results stem from cross-state comparisons and may over or under estimate the true impacts because we were not able to control adequately for differences across the states. As is clear from the analysis of other state motor vehicle policies reported here, the failure to control of state level differences could result in substantial biases. Obviously, more research of these programs is warranted. In the meantime, the available evidence does suggest that these re-licensure policies for the elderly population save lives.

Fourth, we found mixed effects of higher speed limits on motor vehicle fatalities among older adults. This may seem surprising but in our view it merely reflects the complex relationships among interstate speed limits, historic enforcement efforts, the relative use of primary and secondary roads and defensive behavior on the part of older drivers. Higher speeds may indeed result in more deaths on a given road. However, if lower speed limits were seldom enforced and the driving community routinely drove at a faster rate, then raising the speed limit is unlikely to have an impact, it merely reflects the existing reality. In addition, however, if lower interstate highway speeds are enforced, many drivers may travel on less regulated secondary roads. Interstate deaths may be reduced with the lower limit, but total fatalities may be higher with more traffic on secondary roads. Finally, higher posted speed limits may result in fewer motor vehicle fatalities among older adults if they adjust their driving practices to avoid high-speed interstates. Our analysis implicitly examines the net effect of all these factors, and so depending upon their relative importance, any relationship between speed limits and fatalities is possible.

Finally, we found that the number of motor vehicle fatalities among older adults increases with the size of the older adult population. This comes as no real surprise. However, the increases are less than proportional. Between 2001 and 2015 the number of adults in the 65-74, 75-84 and 85+ age groups are projected to increase by 44.9, 6.3, and 44.2 percent, respectively (Census 2000). Our estimates suggest that motor vehicle deaths in these cohorts will increase by 32.1, 1.9 and 30.7 percent, respectively. Even with less than proportional increases, the analysis suggests the gravity of the issue of older adults and motor vehicle related deaths.

To date there have been only a small handful of studies that have explored the effects of policy alternatives to reduce motor vehicle fatalities among older adults. This study has shown that, clearly, there is a need for more work.

Section 8 – Conclusions

This report used state-level data from the 1985 through 2000 Fatality Analysis Reporting System to examine the effects of changes in state laws dealing with seatbelt use, speed limits, driving while intoxicated, and license renewal on fatalities among drivers and others aged 65 and over. Negative binomial regressions were estimated using state and year fixed effects.

The primary findings were as follows:

· Seat belt laws, particularly with primary enforcement, have been effective in reducing older driver fatalities.

· There is no evidence that DUI laws have had an effect on older driver fatalities.

· Maximum speed limit laws have had a mixed (or inconclusive) effect on older driver fatalities.

· The results suggest that strengthening license renewal provisions for older drivers may have been effective in reducing older driver fatalities.

Section 9 - References

Ball, K, C Owsley, ME Sloane, DL Roenker, and JR Bruni (1993), “Visual Attention Problems As a Predictor of Vehicle Crashes in Older Drivers,” Investigative Ophthamology: Visual Science 34(11):3110-3123.

Connecticut General Statutes Revised to 2001 See www.cslib.org/psaindex.htm, last accessed August 27, 2002.

Dee, TS (1999), “State Alcohol Policies, Teen Drinking and Traffic Fatalities,” Journal of Public Economics 72(2):289-315.

Dee, TS (2001), “Does Setting Limits Save Lives? The Case of 0.08 BAC Laws,” Journal of Policy Analysis and Management 20(1):111-118.

Dee, TS and WN Evans (2001), “Teens and Traffic Safety,” in J. Gruber, ed. An Economic Analysis of Risky Behavior Among Youths pp:121-165 (Chicago: University of Chicago press).

Evans, WN and JD Graham (1988), “Traffic safety and the Business Cycle,” Alcohol, Drugs and Driving 4:31-38.

Evans, WN and JD Graham (1991), “Risk Reduction or Risk Compensation? The Case of Mandatory Safety-Belt Use Laws,” Journal of Risk and Uncertainty 4(1):61-73.

Farmer, CM, RA Retting and AK Lund (1999), “Changes in Motor Vehicle Occupant Fatalities after Repeal of the National Maximum Speed Limit,” Accident Analysis & Prevention 31(5):537-543.

Grabowski, DC and MA Morrisey (2001), “The Effect of State Regulations on Motor Vehicle Fatalities for Younger and Older Drivers: A Review and Analysis,” Milbank Quarterly 79(4):517-545.

Hausman, J, BH Hall and Z Griliches (1984), “Econometric models for Count Data with an Application to the Patents-R&D Relationship,” Econometrica 52(4):909-938.

Insurance Institute for Highway Safety (2002), “State Laws,” www.hwysafety.org/safety_facts/safety.htm, last accessed August 27, 2002.

Lave, C and P Elias (1994), “Did the 65 MPH Speed Limit Save Lives?” Accident Analysis & Prevention 26(1):49-62.

Levy, DT (1995), “The Relationship of Age and State License Renewal Policies to Driving Licensure Rates,” Accident Analysis & Prevention 27(4):461-467.

Levy, DT, JS Vernick and KA Howard (1995), “Relationship between Driver’s License Renewal Policies and Fatal Crashes Involving Drivers 70 Years or Older,” Journal of the American Medical Association 274(13):1026-1030.

Lyman, S, SA Ferguson, ER Braver, and AF Williams (2002), “Older Driver Involvement in Police Reported Crashes and Fatal Crashes: Trends and Projections,” Injury Prevention 8(2):116-120.

Owsley, C, K Ball, G McGwin Jr, ME Sloane, DL Roenker, MF White, and ET Overley (1998), “Visual Processing Impairment and Risk of Motor Vehicle Crash Among Older Adults,” Journal of the American Medical Association 279(14):1083-1088.

Owsley, C, G McGwin Jr, M Sloane, J Wells, BT Stalvey, and S Gauthreaux (2002), “Impact of Cataract Surgery on Motor Vehicle Crash Involvement by Older Adults,” Journal of the American Medical Association 288(7):841-849.

Rock, SM (1998), “Impact from Changes in Illinois Drivers License Renewal Requirements for Older Drivers,” Accident Analysis & Prevention 30(1):69-74.

Ruhm, CJ (1996), “Alcohol Policies and Highway Vehicle Fatalities,” Journal of Health Economics 15:435-454.

Shipp, MD (1998), “Potential Human and Economic Cost-Savings Attributable to Vision Testing Policies for Driver License Renewal, 1989-1991,” Optometry and Vision Science 75(2):103-118.

U.S. Bureau of Economic Analysis (2002), “Annual State Personal Income,” www.bea.gov/bea/regional/data.htm, last accessed Augsut 27, 2002.

U.S. Bureau of Labor Statistics (2002), “Historical State Labor Force Data,” www.bls.gov/lau/home.htm, last accessed August 27, 2002.

U.S. Census Bureau (1999), “Population Estimates,” eire.census.gov/popest/archieves/ 1990.php#state, last accessed August 27, 2002.

U. S. Census Bureau (2000), “Projections of Total Resident Population by 5-Year Age Groups, Race, and Hispanic Origin with Special Age Categories: Middle Series, various years, www.census.gov/population/projections/nation/summary/np-t4-d.txt, last accessed August 27, 2002.

U.S. Census Bureau (2001), “Population Estimates,” eire.census.gov/popest/data/states/ tables/ST-EST2001-01.php, last accessed August 27, 2002.

Section 10 - Acknowledgements

This project was funded by a grant from the University Transportation Center for Alabama. As part of our larger research initiative, we also recognize support from the Injury Control Research Center at The University of Alabama at Birmingham. We thank Qing Li and Prerna Gala for able research assistance, Jack Nelson for research advice and Stephen Mennemeyer, Bisakha Sen, Ted Gayer and Thomas Dee for comments on an initial draft of this report.

Effects of State Laws to Reduce Auto Fatalities

UTCA

University Transportation Center for Alabama

List of Figures

Executive Summary

Figure 1-1: Fatal crash involvement by age: 1995-96

Section 2 -Background

Section 3 - Data

Section 4 – Empirical Methods

Section 5 – A Primer on Interpreting Regression Results

Section 6 - Results

Independent Variable

Independent Variable

Independent Variable

Section 9 - References

Section 10 - Acknowledgements