Innovations in measuring household finances

NCRM news
Annette Jäckle, University of Essex

Developing feasible and efficient methods for collecting accurate information about household finances poses many challenges. The “Understanding household finance through better measurement” project, funded by the ESRC Transformative Research scheme and NCRM, has been working on several new developments during 2017.

This research, which is being carried out using the Understanding Society Innovation Panel, includes:

(1) Experimenting with ways of collecting data about the whole household budget constraint in a single  interview, that is, about income, spending and changes in assets and debts;
(2) Experimenting with ways of improving the reporting of income in a survey questionnaire, by showing respondents a summary of all income they have reported in the survey, and asking them to review and edit the summary within the interview;
(3) Reviewing different ways in which data about household finances can be collected that do not involve asking survey questionnaires, and reviewing the likely implications for the representativeness and measurement properties of data collected in this way (for example, linking to data collected by financial aggregators, credit card data, credit rating data, loyalty card data, or scanning till receipts or barcodes);
(4) Analysing the willingness of the general population to participate in a range of survey data collection tasks using their own smartphones or tablets (for example using the camera of their device, installing apps to enter data, or to track information about how they use their device, tracking GPS positioning, tracking movement using the in-built accelerometer, etc.);
(5) Trialling an app to measure spending, where respondents were asked to take pictures of their shopping receipts, or to enter the value and description of purchases directly in the app, for one month.

The focus of (1) has been on developing a feasible method of collecting information about the household budget constraint within a survey interview. Few existing household surveys collect such information: almost all surveys focus on only one or two elements of the household budget constraint. Having all of income, spending, and changes in wealth for the same household should lead to both improvements in data quality and new research opportunities. We trialled a questionnaire module in wave 9 of the Innovation Panel where we asked respondents to report on all money coming into and going out of the household in the last month. To make this feasible, we used the very detailed information about income already collected in the survey; for other aspects such as spending we asked only for aggregate information. For changes in savings and credit accounts we experimentally compared two different sets of questions. The results suggest that it is possible to collect information about the entire budget within an interview. In the face-to-face interviews, showing respondents the balance of all their inflows and outflows in the last month, which by definition should match, led to respondents correcting their reports and resulted in improvements in the data. Respondents who completed the survey online however did not seem to engage with this reconciliation task as often.

The focus of (5) has been to assess the Total Survey Error implications of using an app to collect spending data. The potential benefits of collecting scanned till receipts to measure spending is that these provide objective data on purchases of goods and services, and do not rely on respondents to recall this information. However, not all respondents have access to a compatible mobile device, or are able and willing to do such a task for a survey. Data collected with such an app might therefore not be representative of the general population. Combining data from the spending app with data collected previously in the Innovation Panel interviews shows that a relatively small proportion of the sample were able and willing to complete the task: 13% used the app at least once. Surprisingly, however, those who decided to participate tended to complete the task for the entire month: after 28 days 82% of participants were still using the app. Younger and more educated sample members were more likely to participate, as were women, and sample members who already do similar tasks for their own purposes, for example using apps to check their bank balances. Crucially, however, there were no differences between participants and non-participants in measures of income and spending collected in the Innovation Panel. Over time respondents shifted somewhat from scanning receipts to directly reporting spending in the app, suggesting that they preferred reporting their spending in that way. Work on examining the quality of spending data captured with the app is ongoing.

Further information on the project, including links to collaborators, presentations and working papers, is available at