As discussed above, car accidents, and especially the fatal ones, have a negative impact on country’s economy. Many researchers have been trying to develop models which could explain their causes in order to identify and reduce them. Effects of numerous policies were also examined and, in some cases, predictions of future fatalities had been made.
One of the earliest models estimating road fatalities known as Smeed’s Law was linking deaths with vehicles and population. Smeed (1949) analyzed sample of 20 countries by using data from 1930-1946 period and found that the relationship between number of fatalities and vehicle ownership rate was stable across the sampled countries. The formula was expressed in the
𝐷 continued to rise although the actual number in some countries started to decline from the 1960s.
Kopits & Cropper (2005) found evidence that the income level at which fatality rate per population first declined is approximately 8,600 USD (measured in 1985 international prices) and this income level was attained by countries such as New Zealand in 1968, Belgium, the United Kingdom, and Austria in the early 1970s, and South Korea in 1994. For the United Kingdom, the Smeed’s prediction was moving correctly and had approximately the right magnitude until 1966. By 2000, the prediction was about four times too high (Koren & Borsos, 2010). Other researchers attempted do develop more complex models to cope with changing situation by adding more variables in order to reflect subtle nuances affecting road safety.
Peltzman (1975) developed a model which tested the effect of National Traffic and Motor Vehicle Safety Act of 1966 on death toll on US roads. The Act which made it compulsory for new vehicles to be equipped with several safety features like seat-belts, energy absorbing steering column, and penetration resistant windshield, did not prove to be effective in reducing number of road causalities.
In a time-series analysis (1947-1972) Peltzman divided data into two periods: a pre-regulatory (1947-1965) and a post-regulatory (1966-1972) period. By using log-log multiple regression, he estimated parameters of socioeconomic variables like accident cost, income, time trend, alcoholic intoxication, driving speed and driver age for the pre-regulatory period and then, using the same model, projected fatality rates for the post-regulatory period. There was no significant difference between projected and actual fatality rates during the post-regulatory period and he therefore concluded that the Act had not been effective. He argued that safer cars produced an offsetting behavior on part of drivers as drivers became driving less cautiously while relying on safety equipment installed in their cars.
This conclusion finds support in Wilde’s (1998) paper who argues that there is a constant injury risk target which drivers seek to maintain, therefore, safety regulations are usually being offset.
Evidence of an offsetting behavior was also found by Garbacz (1990) and Keeler (1994).
On the other hand, other researchers have reservations about Peltzman’s findings. Robertson (1977) considers Peltzman’s model insufficiently robust and sensitive to changes in pre-regulatory period. When he re-estimated the equation based on a shorter period (1947-1959 instead of 1947-1965), he obtained different coefficients and projected fatality rates were higher compared with actual ones. Joksch (1976) criticized possible multicollinearity among some variables (income, time trend, and speed) which might have led to biased coefficients. After re-estimating an adjusted model he concluded that the original model was unstable. Studies by Crandall & Graham (1984) and Zlatoper (1984) also indicate that offsetting behavior does not exist. These studies were examining situation in the United States. Several researchers performed studies based on Peltzman’s model in Asian countries.
Garbacz (1989) examined determinants of traffic fatalities in Taiwan from 1964 to 1984. The situation was, nevertheless, very different from what it is nowadays; road causalities were on the rise, there were no regulations on seat-belt or helmet usage, drunk driving did not appear to be a severe problem although alcohol law was in place already. The study was undertaken relatively shortly after the industrialization of Taiwan had begun, Garbacz, therefore, had to consider characteristics of a transitioning economy. He modified Peltzman’s model and used following variables to explain traffic fatalities: real income, accident price, numbers of registered motorcycles and automobiles, real price of gasoline, truck-ton-kilometers and ratio of agricultural employment to total employment. His study suggested that disposable income, number of registered motorcycles and truck-ton-kilometers were positively related to fatalities while the price of accidents, price of gasoline and number of registered automobiles had a negative impact on fatalities. Another conclusion was that an economy based on labor-intensive agriculture reduced fatalities.
McCornac (1993) estimated the effectiveness of seat-belt usage and government safety policies in Japan. In his study he used real income, accident price, highway miles travelled by automobiles, real expenditures on alcohol, percentage of speed limit violations, seat-belt usage and safety policy to estimate their impact on death toll. The data on seat-belt usage were obtained from annual surveys of the National Police Agency and expressed as a percentage of drivers who used seat-belts on a regular basis. Safety policies were a proxy for improvements in traffic safety equipment and were estimated as a number traffic signals, road signs and other road safety elements weighted by length of roadways. Seat-belt usage, safety policies and
accident price had a significant negative effect whereas income and total miles travelled had a significant positive effect.
Another paper based on Peltzman’s model was written by Wong & Wu (1998). Their study tested the efficacy of safety policies (namely seat-belt regulation, usage of breathalyzers, and circuit training and testing system25) along with other socioeconomic determinants (income, accident price, population, number of motorcycles and automobiles, and time trend) on total, occupant, and non-occupant fatalities in Singapore. Dummy variables were used to represent existence or non-existence of a safety policy in a given year. Breathalyzers and seat-belt law were not found to be effective in reducing traffic fatalities, the only safety policy showing a significant negative impact was the implementation of circuit training system. Time trend proved to be negatively related to traffic causalities while number of registered automobiles had a positive effect.
In summary, the various studies of the efficacy of government safety policies in reducing traffic fatalities have not been conclusive. This paper examined effects of safety policies and other socioeconomic determinants on fatalities in Taiwan. While studies depicted in literature presented important aspects in general and antecedents linked significantly to traffic fatalities in particular, they at the same time paid limited attention to the effects of both legal variables (e.g., alcohol law and helmet law) and socioeconomic variables (e.g., real price of fuel, unemployment rate, number of vehicles) as a whole on the number of fatalities. This work is unique in that it studies both groups of variables individually in order to estimate their unbiased effect on road fatalities.
25 This system was introduced in April 1985 with the aim of ensuring that the new pools of drivers and riders were adequately trained and prepared for the roads. The circuits cater for the training and testing of learner drivers and riders. Furthermore, all learner riders were banned from public roads in October 1985.