Does the level of infant mortality affect the rate of decline? Time series data from 21 countries

Some scholars have suggested that an inverse care law holds for infant mortality--that infant mortality reductions are more rapid in regions with lower infant mortality. This theory has not been subjected to proper quantitative analysis. This paper examines time series data on infant mortality from 21 countries to test whether percentage reductions in infant mortality are larger when infant mortality is lower. We apply the Dickey-Fuller generalized least squares (DFGLS) unit root test to infant mortality rate (IMR) time series data from 21 mostly European nations for 1870-1988 to test the statistical significance of beta in a regression analysis of Deltaln IMR(t)=alpha+beta ln IMR(t-1)+epsilon(t). Evidence that beta is significant and negative would support the claim that infant mortality declines more rapidly when substantial mortality reductions have already been achieved. With the preferred specification, the DFGLS unit root test rejected the inverse benefit hypothesis in all countries except the Netherlands at the 5% and 10% cutoffs and Australia at the 10% cutoff. The rejection of the inverse benefit hypothesis was robust to alternative specifications of the lag structure of IMR and to transformations of IMR other than logarithmic. Based on late 19th and 20th century data from countries in Europe and Latin America, there is scarce evidence that percentage reductions in infant mortality are generally smaller in higher mortality countries. These findings suggest that large percentage reductions in infant mortality are possible for countries at any stage in economic development and are reflective of durable advances in human knowledge, social institutions and physical capital.