Measuring Consumption and Saving: Introduction*

Transcription

1 FISCAL STUDIES, vol. 30, no. 3/4, pp (2009) Measuring Consumption and Saving: Introduction* THOMAS F. CROSSLEY University of Cambridge; Institute for Fiscal Studies In order to think about a great deal of fiscal and tax policy, and, indeed, a great deal of public policy more generally, it is necessary to know how households financial resources are allocated across time (between current consumption and saving), across goods and across the different individuals in the household. It is also necessary to understand how those allocations come about: how decisions are made and what influences those decisions. The policy areas to which data on spending and saving are relevant include: indirect taxes; incentives for saving and retirement preparation; the degree to which households are insured against idiosyncratic shocks, and the efficacy of social insurance programmes; the measurement and abatement of poverty and inequality; and the micro foundations of aggregate demand. 1 Such data are also important for the study of the emerging obesity epidemic, and other issues in nutrition and health. Of course, all of these areas have been, and continue to be, the focus of a great deal of research. The greatest obstacle to research progress in many of these areas is data. On many questions, theory is well ahead of measurement. Further progress rests heavily on bringing more and better data to bear, to assess alternative theories and to generate the new empirical insights that will stimulate new theoretical developments. This Special Issue of Fiscal Studies brings *The author is grateful for the support of the ESRC-funded Centre for Microeconomic Analysis of Public Policy at the Institute for Fiscal Studies (grant number RES ) for funding this work. Thanks also to Cormac O Dea for permission to reproduce Figure 1 from their joint work and to Judith Payne for superb editorial work on this Special Issue. Keywords: consumption, saving, measurement, data, survey methods. JEL classification numbers: D12, C81, C83. 1 For further discussion, and references, see Browning, Crossley and Weber (2003).. Published by Blackwell Publishing Ltd, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA

2 304 Fiscal Studies together a collection of papers that attempt to push back the frontier of the measurement of consumption and saving. Microdata on expenditure and saving suffer from two kinds of limitations. First, we often do not have the data we need. Second, the data that we do have suffer from quality issues that are important and seem to be growing over time. With respect to the first point, the traditional way of collecting data on consumption (more correctly, expenditure) has been household budget surveys. Typically, these surveys were designed to support the construction of price indices, and not for more general economic research. An obvious limitation is that while we are often interested in the well-being of individuals, these surveys collect data on consumption at the household level, making it very difficult to study allocations within households. Further, these surveys are perceived to have very high respondent burden. As a result, little content is collected beyond expenditures and basic demographics, and households are typically asked to participate only once, or for a short period, so the data are cross-sectional, or at best a short panel. The lack of longitudinal information means that we cannot measure household-level dynamics in consumption, yet saving and spending decisions are both idiosyncratic and inherently dynamic. The lack of information on other characteristics and on choices in other domains limits our ability to study the causes and consequences of changes in spending and saving. How household consumption and saving choices are related to household circumstances and to choices in other domains is central to what we need to know. Health is a very good example. It is important to understand how spending patterns at older ages are influenced by changes in health status, and, equally, how allocation decisions by households affect their health outcomes. Detailed expenditure surveys rarely collect much (if any) health information, while surveys with good information on health collect, at best, fairly limited information on consumption and saving. Turning to data quality, it is now widely known that measures of data quality in traditional budget surveys are exhibiting alarming trends. Figure 1, which is typical, provides an illustration. The solid line plots the response rate to the UK Family Expenditure Survey (FES) and its successor, the Expenditure and Food Survey (EFS), over time. It appears to be increasingly difficult to get households to participate in this survey. The dashed line shows the ratio of total expenditure in these surveys, grossed up to the national level (using appropriate survey weights), to the corresponding figure in the National Accounts. Less expenditure is recorded the microdata than in the National Accounts, and the discrepancy is growing over time. This problem is not unique to the UK data; in fact, the corresponding discrepancy between the US Consumer Expenditure Survey (CEX) and

3 Measuring consumption and saving: introduction 305 FIGURE 1 The quality of UK household expenditure data Source: Crossley and O Dea, forthcoming. National Income and Product Accounts (NIPA) personal consumption expenditure (PCE) data is larger and began to grow earlier (Attanasio, Battistin and Ichimura, 2004). The papers in this Special Issue address both the problem of not collecting the data we need and the problem of quality in the data we do collect. For example, Bonke and Browning report on a groundbreaking attempt to collect information on the allocation of consumption within households. Hurd and Rohwedder report on an attempt to collect consumption data in a supplement to a major longitudinal survey (the Health and Retirement Study) that collects detailed information on health and many other domains. The papers by Leicester and Oldfield and by Griffith and O Connell study a source of longitudinal data on household expenditures. Each of these papers addresses a critical data need. With respect to the quality of consumption and saving data, the papers in this issue consider the impact on data quality of a wide range of survey design choices. The papers by Leicester and Oldfield, by Griffith and O Connell and by Essig and Winter consider differences in data collected by different survey modes. Hurd and Rohwedder pay particular attention to the trade-off between purchase infrequency and recall bias that arises when choosing the length of the period to which expenditure questions should refer. Pradhan considers how the quality of consumption data is affected by the level of disaggregation in the expenditure questions. Comerford, Delaney and Harmon consider a number of survey design issues including question wording, response scale, recall period and the level of disaggregation. In addition to important particular findings that each of the papers in this Special Issue reports, several broad and important messages emerge from the

4 306 Fiscal Studies issue as a whole. First, the papers emphasise that when we set out to measure household consumption and saving, we have a rich and growing toolbox on which to draw. Most obvious in this regard is the multiplicity of modes of data collection that might be employed. Leicester and Oldfield compare home scanner data on expenditures and data collected in a traditional budget survey, and Griffith and O Connell consider the utility of home scanner data for research on nutrition in particular. Essig and Winter compare computerassisted personal interviews and drop-off questionnaires. Hurd and Rohwedder describe experience with a mail survey, as well as experiments in a survey conducted over the internet (in the American Life Panel). The papers in this Special Issue also illustrate that a significant body of knowledge has accumulated about measuring consumption and saving in particular, and about the nature of survey response more generally. They build on past research, not only in economics but also in psychology, statistics and other related disciplines. A final broad theme of the papers is the value of randomised experiments in generating new knowledge of the survey response process and in improving our data collection efforts. The papers by Essig and Winter, by Comerford, Delaney and Harmon and by Hurd and Rohwedder all report on deliberate controlled experiments in survey design. In these experiments, treatments differences in mode of data collection or question format are randomly assigned to respondents. In addition, Pradhan exploits an existing random allocation of survey respondents to a detailed consumption module as a controlled experiment. The important insights in these papers will no doubt spur further experimentation. If we wish to measure how households financial resources are allocated across time, across goods and across the different individuals within the household, we must overcome many difficult problems. Nevertheless, it is far too important a project to be abandoned. The papers in this Special Issue demonstrate that it is possible to make progress on these problems and they highlight promising directions for future research. References Attanasio, O., Battistin, E. and Ichimura, H. (2004), What really happened to consumption inequality in the US?, National Bureau of Economic Research (NBER), Working Paper no Bonke, J. and Browning, M. (2009), The allocation of expenditures within the household: a new survey, Fiscal Studies, vol. 30, pp (this issue). Browning, M., Crossley, T. F. and Weber, G. (2003), Asking consumption questions in general purpose surveys, Economic Journal, vol. 113, pp. F Comerford, D., Delaney, L. and Harmon, C. (2009), Experimental tests of survey responses to expenditure questions, Fiscal Studies, vol. 30, pp (this issue). Crossley, T. F. and O Dea, C. (forthcoming), Wealth and Saving of UK Households on the Eve of the Crisis, Report, London: Institute for Fiscal Studies.

5 Measuring consumption and saving: introduction 307 Essig, L. and Winter, J. K. (2009), Item non-response to financial questions in household surveys: an experimental study of interviewer and mode effects, Fiscal Studies, vol. 30, pp (this issue). Griffith, R. and O Connell, M. (2009), The use of scanner data for research into nutrition, Fiscal Studies, vol. 30, pp (this issue). Hurd, M. and Rohwedder, S. (2009), Methodological innovations in collecting spending data: the HRS Consumption and Activities Mail Survey, Fiscal Studies, vol. 30, pp (this issue). Leicester, A. and Oldfield, Z. (2009), Using scanner technology to collect expenditure data, Fiscal Studies, vol. 30, pp (this issue). Pradhan, M. (2009), Welfare analysis with a proxy consumption measure: evidence from a repeated experiment in Indonesia, Fiscal Studies, vol. 30, pp (this issue).