Infant Mortality Rates in Colombia. From Re-Estimated

In spite of its significance for assessment of improvements in health and of development in general, measurement of infant mortality (IM) in Colombia has been subject considerable variation and inconsistency over the last three decades.  Three measurement periods can be identified.  The first, 1979-1987, is marked by under-reporting/registration of births and deaths. The second, 1988-1997, by no registration of live births.  The third, from 1998 on, is marked by a new system of registration. This produced two series, inconsistent one with the other, with a gap between them.

Objective: The objective of the paper is to reconcile the data from the different periods in order to determine the trend in infant mortality in Colombia over the entire period since 1979.

Methodology: All available data on life births and on infant deaths were aggregate to level of Department (State/Province), providing a series for the first and third periods.  A linear regression model estimating the general relationship IM=αe^βt   where  α is the projected initial level of IM, and β is the average rate of IM decline.  Estimates were applied to each Department and to all combined, assuming fixed effects across Departments and for each of the two series.

Results: The regression estimates suggest that the two series can be aligned and the IM rates for the intermediate period approximated accordingly, so to give a consistent rate of decline in IM for the entire period. This is achieved mainly by eliminating, for each period, Departments with high levels of variations in IM, suggesting bad data and denoted by R²<0.2, and “too low” variations, suggesting “too good” data and denoted R²>0.8.