Traffic fatalities and economic growth

Accid Anal Prev. 2005 Jan;37(1):169-78. doi: 10.1016/j.aap.2004.04.006.

Abstract

This paper examines the relationship between traffic fatality risk and per capita income and uses it to forecast traffic fatalities by geographic region. Equations for the road death rate (fatalities/population) and its components--the rate of motorization (vehicles/population) and fatalities per vehicle (F/V)--are estimated using panel data from 1963 to 1999 for 88 countries. The natural logarithm of F/P, V/P, and F/V are expressed as spline (piecewise linear) functions of the logarithm of real per capita GDP (measured in 1985 international prices). Region-specific time trends during the period 1963-1999 are modeled in linear and log-linear form. These models are used to project traffic fatalities and the stock of motor vehicles to 2020. The per capita income at which traffic fatality risk (fatalities/population) begins to decline is 8600 US dollars (1985 international dollars) when separate time trends are used for each geographic region. This turning point is driven by the rate of decline in fatalities/vehicles as income rises since vehicles/population, while increasing with income at a decreasing rate, never declines with economic growth. Projections of future traffic fatalities suggest that the global road death toll will grow by approximately 66% over the next twenty years. This number, however, reflects divergent rates of change in different parts of the world: a decline in fatalities in high-income countries of approximately 28% versus an increase in fatalities of almost 92% in China and 147% in India. The road death rate is projected to rise to approximately 2 per 10,000 persons in developing countries by 2020, while it will fall to less than 1 per 10,000 in high-income countries.

MeSH terms

  • Accidents, Traffic / economics*
  • Accidents, Traffic / mortality*
  • Automobiles / statistics & numerical data
  • Developing Countries*
  • Humans
  • Income*
  • Linear Models
  • Models, Statistical