The Information Content of Consumer Surveys
Pain, Nigel, Weale, Martin, National Institute Economic Review
Considerable attention is given in the media to changes in consumer sentiment, as recorded in regular monthly surveys. It is widely felt that changes in sentiment provide useful information about future expenditure patterns. Given the lags that occur in the production of national accounts data, a reliable coincident or leading indicator of consumption might prove especially useful at times of particular uncertainty about the short-term economic outlook.
This supposition can of course be tested statistically. It may be that consumer sentiment just reflects current changes in the underlying structural determinants of expenditure, so that sentiment falls at times when asset prices drop or incomes decline. Equally it is possible that there is no consistent relationship at all between sentiment and expenditure across the economic cycle. In this note we examine the information content of consumer surveys in the UK and the United States.
The UK measure we use comes from the European Commission, and is used also by the OECD. The indicator combines replies to five separate questions by taking the arithmetic average of the individual balance of (weighted) positive to negative responses for each of the questions. The questions themselves relate to the current and expected financial condition of the household, the current and expected general economic situation and whether respondents are planning to make major purchases at present.
Chart 1 shows the long-term trend in this series since 1974, which is the first available data point, along with the quarterly rate of growth of consumers' expenditure at constant prices. It is clear that there is a general tendency for confidence to be weaker at times of subdued growth. At present confidence remains well above the longer-term average level, even though it has slipped a little since the start of this year. (Somewhat surprisingly the long-run average is well below zero, suggesting either some bias in the survey, an inappropriate weighting of the separate questions, or innate pessimism.)
In passing, it is of interest to note an example of where consumer confidence provided a misleading signal. Sentiment in the UK declined sharply in the latter part of 1992, coinciding with the withdrawal of sterling from the Exchange Rate Mechanism, and fell to a point well below that in the previous two years of recession. Yet, as we correctly forecast at the time using a structural model rather than a survey indicator, the enforced devaluation of sterling helped to bring about an economic recovery rather than a deeper recession.
Our US index is the consumer sentiment indicator that has been compiled by the Survey Research Centre of the University of Michigan since 1952. Again five questions are used. These cover similar ground to those in the UK, relating to current and expected trends in personal finances and business conditions and current buying conditions for durable goods. Each question has three possible answers, which are weighted 0, 1 or 2. The aggregate index is calculated as an average of the scores for the individual indicators. Chart 2 shows the behaviour of this indicator over the past 30 years, again alongside the quarterly rate of growth of expenditure. The lowest points of the series, in the mid-70s, early 1980s and early 1990s, coincide with periods in which the US economy was in recession. The marked falls in consumer confidence this year are readily apparent, but because of the high level at the end of last year, confidence has yet to fall to the levels seen in past recessions.
Is consumer sentiment a leading or a cyclical indicator?
One test of the information content of the consumer survey measures is to run simple dynamic regressions of the growth in consumers' expenditure in the current quarter on lagged expenditure growth and current and lagged survey responses. The growth of expenditure has to be used rather than the level as the bounded survey responses cannot be expected to account for the long-term upward trend in the level of expenditure over time. …