This was created in partnership with Environment for Development .
This study uses household survey data from four Kenyan towns to examine the effect of households’ characteristics and risk perceptions on their decision to treat/filter water as well as their choice of main drinking water source. Because the two decisions may be jointly made by the household, a seemingly unrelated bivariate probit model is estimated. It turns out that treating non-piped water and using piped water as a main drinking water source are substitutes. The evidence supports the finding that perceived risks significantly correlate with a household’s decision to treat/filter unimproved non-pipe water before drinking it. The study also finds that higher connection fees reduce the likelihood of households connecting to the piped network. Because the current connection fee acts as a cost hurdle that deters households from getting a connection, the study recommends a system where households pay the connection fee in instalments, through a prepaid water scheme or through a subsidy scheme.