The provision of clean water is mentioned as an important factor in many studies dealing with the decline of mortality in Europe during the late nineteenth and early twentieth centuries. In developing countries too, improved water supply is assumed to have a strong impact on mortality. When studying the effect of water supply on public health, researchers are confronted with many methodological problems. Most of these also apply to historical studies of the subject. We review the evidence from this historical research, taking into account the methodological problems observed in contemporary impact evaluation studies, and we use more refined data from the Dutch city of Tilburg, enabling us to overcome many of these shortcomings. Finally, we discuss some factors which may explain why we failed to discover an effect of the availability of piped water on the level of childhood mortality.
PIP: The UN designation of the 1980s as the International Drinking Water Supply and Sanitation Decade has stimulated research into the effect of water supply on public health. This paper opens with a review of studies that have attempted to quantify the impact of improved water supply (improved quality and/or increased quantity) on infant and child mortality in 19th- and early 20th-century France, the Netherlands, Germany, Sweden, and the US. The next section reviews studies on the health effects of water supply on contemporary populations in the developing world. The paper continues by discussing some of the methodological problems contained in such research including the difficulty of predicting the sign of the correlation between waterworks and morality rates, the difficulty in choosing which dependent variable to use, data limitations, problems finding appropriate control groups, unequal conditions prevailing before adoption of a specific public health improvement, failure to analyze mortality by age, and failure to determine seasonal differences. Next, a study of the effect of water supply on the population in the Dutch city of Tilburg in the early 20th century is presented. This study answers some of these methodological problems because it uses available individual-level data on health indicators and water supply. Analysis of the data, including use of a hazard rate model with the independent variables of water supply, socioeconomic status, father's income level, age of mother at birth, birth order of child, interval to next child, season of birth, number of rooms in the house, and district of birth, was performed for three age groups from birth to five elapsed years. This study revealed no discernable effect of the availability of piped water on the level of childhood mortality. The paper ends by proposing various explanations for these findings.