Non-sampling errors in surveys of residential energy consumption

Abstract

Surveys are an important source of information about energy consumption in the residential sector. Most published estimates of survey error only take account of the uncertainty that arises from the random variation that is present when only a sample of the entire population is surveyed. We examine other sources of survey error, called non-sampling errors, and assess the degree to which these sources of error may affect the usefulness of survey data. Two types of error are considered: non-response error which arises from a failure to obtain data from a chosen unit in the sample and response error which arises when erroneous data are obtained from a chosen unit. Response error is found to be more serious than non-response error in the survey results examined here. Response error is found to be most likely to occur when respondents are asked to provide quantitative information such as estimates of the average monthly utility bill or the floor area of the residence. It is suggested that response error can seriously affect the reliability of the results from analytical methods that are frequently applied to survey data. It is argued that this fact justifies increased efforts to reduce the frequency of occurrence of response error in future surveys.