The value of certain refinements of the infant-mortality rate

Several methods have been advanced for refinement of the conventional infant-mortality rate by relating the deaths to the corresponding annual birth cohorts. The author has tested the accuracy of such methods by assuming various "true" infant mortality rates and applying them to quarterly birth statistics as documented in 39 countries in 1939-49, a decade of considerable fluctuations in the number of births in many countries. Against the background of these actual data, refinement by year of birth is found to give unpredictable and relatively unsatisfactory results. While the error in the conventional rate is, on the average, reduced by about a half, in 25% of the cases the so-called "refinement" actually increases the error. The refinement methods discussed can therefore not be recommended for indiscriminate routine use.