Constructing Continuous Household Income Measurement on the British Election Study Internet Panel

This research note describes the procedure for constructing continuous measurements of income on the British Election Study Internet Panel using several imputation steps. As with most social surveys not focusing directly on income measurement, the British Election Study Internet Panel uses a household income question that asks respondents to place their income within discrete bins. The most common analysis practices are either to group incomes into coarser bins such as quintiles which loses information or to impute a continuous income measure based on bin midpoints, which ignores the likely underlying distribution of continuous income, which will not be uniform within each bucket. We adapt a recently developed technique to interpolate a cumulative density function from discrete categories for aggregate counts to apply to the individual level data. Using this approach will allow the incorporation of external information about mean incomes within subgroups to be further integrated into the imputation approach