Highway safety has steadily improved during the past several decades, but traffic fatalities, exceeding more than 30,000 annually, are still one of the leading causes of non-disease deaths in the United States. The United States also has the highest traffic accident fatality rate in developed countries among people of age 24 and younger, despite laws that ban drinking until the age of 21. In addition to those direct costs, traffic accidents account for a large share of highway congestion and delays and increase insurance premiums for all motorists. It is not an exaggeration to suggest that reducing MVCs and their associated costs should be among the nation’s most important microeconomic policy goals. The top line in Figure 1 shows that automobile fatalities have followed a downward trend since the 1970s, and have fallen especially rapidly during recessions, which are shaded in the figure. A natural explanation is that those declines are simply a consequence of the decrease in vehicle miles travelled (VMT) that typically accompanies a recession. With a smaller labor force commuting to work, fewer goods being shipped along overland routes, and less overall economic activity, a decline in traffic facilities is no surprise. But the darker line in the figure shows that the fatality rate (fatalities per VMT) has also decreased more sharply during recessions than during other parts of the business cycle. This implies that factors other than declining VMT contribute to the reduction in fatalities that tends to occur when real economic activity contracts. The purpose of this paper is to document those factors. Researchers have documented that cyclical fluctuations in economic conditions affect most major sources of accidental deaths, including motor vehicle accidents, and they have concluded that fatalities resulting from most of those sources decline approximately in proportion to the severity of cyclical contractions in economic activity found that overall death rates rose when unemployment fell, and argued that this relationship could be explained by labor shortages that resulted in elderly people receiving worse health care in nursing homes as the economy expanded. But, as noted by Peterman (2013), this line of research does not explain why the percentage decline in automobile fatalities during recessions has been so much greater than the percentage decline in VMT. In 2009, for example, VMT declined less than 1%, but fatalities declined fully 9%. There are several leading hypotheses that have been proposed to explain the sharper decline in the automobile fatality rate during recessions, including: *motorists drive more safely because they are under less time pressure to get to various destinations (especially if they are unemployed and have a lower value of travel time than they have when they were working); *households try to save money by engaging in less discretionary or recreational driving, such as Sunday drives into the country on less-safe roads; *motorists become risk averse because of the economic strain during a recession and drive more carefully to avoid a financially devastating accident; *recessions may cause a change in the mix or composition of drivers on the road that results in less risk and greater safety because the most dangerous drivers account for a smaller share of total VMT. To the best of our knowledge, researchers have not tested those hypotheses empirically because the data on individual drivers’ VMT, socioeconomic and vehicle characteristics, and safety records during recessionary and non-recessionary periods that would be required to do so are not publicly available. Publicly available data on VMT are generally aggregated to at least the metropolitan area or state level and suffer from potentially serious measurement problems. For example, nationwide VMT statistics that are released by the federal government are not based on surveys of individual drivers’ VMT; instead, they are estimated from data on gasoline tax revenues to determine the amount of gasoline consumed which is then multiplied by an estimate of the average fuel efficiency of the nation’s vehicle fleet. This paper avoids the problems associated with the publicly available data and instead analyzes motorists’ safety over the business cycle using a novel, disaggregated data set obtained by State Farm Mutual Automobile Insurance Company (hereafter State Farm®) from a private supplier. It relies on a new generation of information technologies, referred to as telematics, which remotely record individual vehicles’ exact VMT from odometer readings and store information about drivers and their safety records. This unique data set enables us to identify heterogeneous adjustments in vehicle use to changes in local economic conditions across a large number of motorists and to assess how differences in their adjustments may affect overall road safety. Capturing motorists’ heterogeneous responses turns out to be important because we find that changes in local unemployment do not affect the average VMT per driver across all drivers, but they do affect the composition of drivers on the road. In particular, we find that the drivers in our sample who are likely to pose the greatest risks to safety, as indicated by several observable characteristics such as the driver’s age and accident history, reduce their VMT in response to increasing unemployment. As a result, safer drivers account for a larger share of VMT during periods when aggregate unemployment is high. This finding reconciles the two key safety-related phenomena observed during recessions: the large (and typically permanent) decline in aggregate automobile fatalities and the modest (and usually transient) decline in aggregate VMT, which we note may be spuriously correlated. 5 Moreover, the quantitative effect of the change in driver composition on automobile safety is economically significant: our estimates suggest that the change in the composition of VMT that results from an increase in the nationwide unemployment rate of 1 percentage point could save nearly 5,000 lives nationwide per year. (Author/publisher)
Samenvatting