Are simple tests of son preference useful? An evaluation using data from Vietnam

Son preference is widespread although not universal. Where it occurs it may lead to higher fertility rates. Ideally son preference should be measured in the context of a hazards or parity progression model of fertility, or a logistic model of contraceptive use. Such models require large amounts of survey data, particularly to measure the covariates. Can son preference be discerned reliably using tests which rely on more limited information? The answer is yes, based on applying eight simple tests to data from the Vietnam Living Standards Survey of 1992–93 and comparing the outcomes with the benchmark results from fuller models. Some, but not all, of the simpler tests accurately measure son preference, including estimating a simple hazards or progression parity model, the unisex sibship test, and the sibling differentials test. Received: 28 September 1995 / Accepted: 2 January 1997