A score statistic for testing the Hardy-Weinberg law for data sampled from several populations with different allele frequencies is derived assuming a common coefficient of disequilibrium (odds ratio-type) across strata. The assumption can be examined by a chi-square goodness-of-fit test. It is unlikely such a model would be rejected in typical studies involving human populations. Under this model, a Wald-type test has substantially less power than the score test. An approximate formula for the sample size required for a specific power of the stratified score test for detecting departure from equilibrium is given.