Conferences

Household expenditure data is vital for many areas of analysis, from the analysis of long-run poverty and inequality to the construction of national price indices. Specialist expenditure surveys tend to be cross-sectional and record detailed expenditures using a short-term diary to minimise respondent burden, but which lack any panel element. Expenditure questions in more general surveys generally ask respondents to recall expenditures on different broad categories, meaning much of the detail is lost and making the figures prone to recall biases or rounding.

An alternative source of expenditure data comes from market research companies. Barcode scanners are installed in the homes of participants which allows the detailed items purchased to be quickly and accurately recorded. This scanner data has three key advantages: it is a panel expenditure data source which allows household responses to socioeconomic events to be tracked; it provides extremely detailed and disaggregate expenditure data; and it combines price and quantity data in one source. However, not much is known about the implications of this mode of data collection in terms of response, attrition and representativeness. There may be concerns that some household types will find it particularly hard to use the technology (such as the very elderly) or that households with little time may quickly tire of scanning all their purchases and attrit. Using data from market research company TNS on the food and grocery purchases of tens of thousands of households over 6 years, our research explores these issues by comparing the expenditure and demographic details of the data to existing data sources more traditionally used in social science research, including the EFS, the Census and the British Household Panel Survey. We aim to document and explore issues around response and representativeness and to assess how far we can attribute differences across the surveys to the mode of data collection.