The CE and the PCE: a comparison: an analysis of a decline in the ratios of aggregate spending for various categories of expenditures from the BLS Consumer Expenditure Survey and the BEA's Personal Consumption Expenditures over an 11-year period employs a new methodology that takes into account the degree of comparability of those categories

Monthly Labor Review, Sept, 2006 by Thesia I. Garner, George Janini, William Passero, Laura Paszkiewicz, Mark Vendemia

Since the start of the ongoing Consumer Expenditure Survey (CE) in 1980, expenditure estimates from CE data have been compared regularly with corresponding expenditure estimates from other data sources to evaluate both the soundness of the CE estimates at any point in time and the consistency of the estimates over time. In 1987, Raymond Gieseman, the first within the Bureau of Labor Statistics (the Bureau, BLS) to use continuing survey data to conduct this work, stated the aim of the comparisons: "What was expected from these comparisons was a sense of degree and direction of possible survey errors, rather than an exact measure of bias, because the specific estimates from other sources are not necessarily the 'true' values." (1) In conjunction with other evaluation tools, data comparisons are employed to assess the cumulative effects of nonsampling errors on the quality of data obtained from the CE and to develop methodological studies to improve that quality. (2)

In addition to these internal uses, data comparisons have appeared regularly in CE publications. The major biennial releases of the CE program include tables comparing its data estimates with those from other sources. Articles on these comparative measures also have been published in the Monthly Labor Review. (3)

The primary source of independent data for comparison over the years has been the Personal Consumption Expenditures (PCE) of the National Income and Product Accounts, produced by the Bureau of Economic Analysis (BEA); these data are the focus of this study. The PCE affords comprehensive coverage of item categories similar to those of the CE and, in fact, is used as a tool in the process of producing tables for CE publications.

Like all data sources, the CE and the PCE have their strengths and weaknesses. The strength of the CE is that an extensive accounting of expenditures made by consumer units (4) is collected through personal interviews and paper-and-pencil diaries. Separate samples of consumer units participate in the Diary survey and the quarterly Interview survey. A weakness is that the data are collected from samples and thus are subject to sampling errors. Nonsampling errors also may be introduced, in processing the data for final use. The strength of the PCE is that it provides estimates of aggregate expenditures for an extensive list of commodities purchased for consumption by and on behalf of households. However, PCE data are subject to (1) measurement errors in the censuses and sampling and nonsampling errors in surveys that provide source data to the BEA and (2) classification errors by the BEA in its estimation and allocation of production or output to the personal sector and other sectors in constructing the national accounts. Each year, previously released PCE aggregate expenditure estimates are subject to revision, which can result in meaningful differences over time. This alone supports the proposition that there is no "true" value for consumer expenditure estimates, as suggested by Gieseman. (5)

Work by a team of researchers within the Bureau (6) suggested that earlier methods comparing CE data with PCE data needed to be reevaluated. As a part of the reevaluation, the team kept in mind that the CE and alternative data sources were designed to serve different purposes; thus, comparisons of estimates may be affected by differences in scope, definition, and estimation procedure. The team attempted to reconcile these differences as much as possible to construct compatible estimates. The purpose of the current article is to highlight recent work of this team. A quantitative comparison of CE and PCE expenditure estimates is presented, followed by a discussion of differences between the estimates and possible reasons for them.

Outline and summary of findings

The next section highlights previous research comparing the CE and the PCE. Following that, the foundations of the CE and the PCE are presented, including the purposes of the two surveys, the populations they cover, definitions of expenditures, and data collection methods. Then the historical comparison methodology developed and used by the CE is described. (7) Finally, ratios of CE-to-PCE aggregate expenditures from 1984 through 2002 are calculated and shown for categories of expenditures. PCE expenditure estimates are based on 1997 benchmark data, updated to their current levels by periodic revisions that have occurred through 2005.

Exhibit 1 summarizes the trends in the CE-to-PCE ratios over the 1984-2002 period at a disaggregated level. For most categories of spending, the ratios have been decreasing. Appendix table C-1 shows that for two categories of expenditures--clothing for children less than 2 years of age and purchases of vehicles--CE aggregate expenditures are greater than PCE aggregates for earlier periods, but drop to or below PCE estimates in later years. Overall, however, the historical comparison methodology suggests that CE and PCE aggregate estimates are becoming more disparate with time.