Missing Data Can Be a Geographic Phenomenon

Missing data is a problem across all types of analysis, with survey nonresponses just one example of a universal problem. This is particularly the case with survey questions on sensitive topics, such as ethnicity or income, as these are often not answered. This can introduce bias into the data and present a challenge to analysts, particularly as they can lead to incorrect conclusions and policy recommendations. Accordingly, methods have been developed to fill these gaps, but their effectiveness depends on how random these omissions are. This article contributes to the topic by first reviewing the theoretical background of missing data and missing spatial data. It then examines demographic and income answers to the UK Financial Lives Survey and investigates the geographic component to the missing data using geographically weighted regression (GWR). We found that there is a geographic component to nonresponse in the financial survey, and this geographic component can be leveraged to improve missing-data estimation. Although location information is not available in every data set, the use of GWR for imputing missing data could provide a useful additional analytical tool in other fields of research when it is.