ABSTRACT
<p><b>OBJECTIVE</b>To assess the data quality and estimate the provincial infant mortality rate (1q0) from China's sixth census.</p><p><b>METHODS</b>A log-quadratic model is applied to under-fifteen data. We analyze and compare the average relative errors (AREs) for 1q0 between the estimated and reported values using the leave-one-out cross-validation method.</p><p><b>RESULTS</b>For the sixth census, the AREs are more than 100% for almost all provinces. The estimated average 1q0 level for 31 provinces is 12.3‰ for males and 10.7‰ for females.</p><p><b>CONCLUSION</b>The data for the provincial 1q0 from China's sixth census have a serious data quality problem. The actual levels of 1q0 for each province are significantly higher than the reported values.</p>