Analysis of Non-Response Bias in the Nigeria Demographic and Health Survey Using Statistical Adjustment Methods

Non-response in household surveys such as the Nigeria Demographic and Health Survey (NDHS) can introduce systematic bias into population estimates if non-respondents differ systematically from respondents, yet formal statistical analysis of non-response bias in the NDHS has received limited methodological attention. This study analyses the extent and statistical impact of non-response bias in the 2018 NDHS using household response rate data, non-response tracking records, and external administrative population benchmarks. Unit non-response rates by stratum were computed. Logistic regression identified predictors of household non-response using available frame characteristics. Post-stratification weighting and calibration weighting were applied and compared for bias reduction effectiveness by benchmarking to census marginal distributions. The NDHS achieved an overall household response rate of 96.6 percent, with substantially lower rates in North East urban strata (88.4 percent). Non-response was significantly predicted by urban location (OR 2.1), security-affected LGA designation (OR 3.8), and high-income area classification (OR 1.9). Unadjusted estimates for selected fertility and child health indicators deviated by up to 3.8 percentage points from calibration-weighted estimates in North East strata. Calibration weighting outperformed post-stratification in bias reduction across all assessed indicators. The study contributes methodological evidence on NDHS non-response bias and recommends adoption of calibration weighting as the standard for NDHS estimates in low-response security-affected strata. Keywords: non-response bias, NDHS, calibration weighting, survey statistics, Nigeria