Survey bias and the poor: How survey responses overstate electricity spending

Accurately estimating household electricity expenditures is essential for assessing energy poverty and informing subsidy and affordability policies. This study combines household survey data with administrative utility billing records for households in Rwanda to examine the extent and nature of misreporting in self-reported electricity spending, using a sample size of 650 households. We find systematic over-reporting among poorer households, primarily due to mismatches between survey recall periods and the irregular, prepaid nature of electricity purchases. To correct for this bias, we adjust reported expenditures using empirical distributions of monthly purchase frequency derived from utility data. Additionally, for unmatched households, we develop predictive models based on household characteristics, though these perform less reliably in data-sparse settings. Extending the analysis to Uganda, we apply both the frequency-based correction and predictive models. Across countries, adjusted estimates suggest electricity burdens for most households lie between 0.5% and 2.5% of total household expenditure across all wealth groups. These findings underscore the limitations of relying solely on household surveys to measure electricity spending in prepaid systems. They highlight the value of integrating administrative utility data with statistical correction methods to produce more accurate and policy-relevant assessments of electricity affordability. • Surveys overstate electricity spending by 2–3X among poor households. • Poorer households purchase electricity less frequently. • Survey recall periods often fail to align with actual electricity purchase durations. • Utility billing data enable more accurate estimation of purchase frequency and correction of survey bias. • After frequency adjustment, electricity burdens account for approximately 1.5% ±1 of household budgets.