The Treasury

Global Navigation

Personal tools

Treasury
Publication

Does Consumer Confidence Forecast Consumption Expenditure in New Zealand? - WP 03/22

4  Forecasting Ability of Consumer Confidence

A two-step process is used to determine the forecasting ability of consumer confidence. The method was developed by Carroll et al (1994) and subsequently used by other researchers, allowing for cross-country comparisons. The first step involves specifying a simple forecasting equation to examine the predictive ability of consumer confidence on consumption expenditure. This is done by examining the from regressions of the growth of various measures of consumption expenditure on lagged values of consumer confidence indexes. This specification takes the following form:

(2)     

where denotes the four different measures of consumption, is a constant, denotes the various consumer confidence indexes, and is the error term. The four measures of consumption are total private consumption, durables consumption, non-durables consumption, and services consumption. Equation (2) essentially amounts to a test of Hall’s (1978) random-walk hypothesis, which states that since the permanent income hypothesis implies that consumption will follow a random walk, consumption growth cannot be forecasted from lagged information.[6] If the ’s in equation (2) are significantly different from zero, Hall’s hypothesis can be rejected. The predictive ability of the Westpac survey component questions is also tested by using the derived Current Conditions Index and Future Conditions Index; and the net responses for each of the five individual questions. The One News survey was also measured for the quarter based on the monthly reading at the end of the quarter instead of using the quarter average, as an alternative. The results, which are not reported in this paper, indicate that this alternative measure do not qualitatively change the estimations.

The second step involves investigating whether consumer confidence has any predictive ability once controls for information contained in other variables are introduced. This is done by modifying equation (2) above by introducing control variables, producing the following form:

(3)    

where is a vector of control variables. Equation (3) allows the predictive ability of consumer confidence to forecast future consumption expenditure to be quantified by examining the incremental *. values. The choice of control variables used is based on existing literature. Carroll et al (1994) used a minimal specification that included four lags of the dependent variable and four lags of the log first difference in real labour income, although they note that the choice of variables is inherently somewhat arbitrary. Bram and Ludvigson (1998) expanded the choice of variables to include four lags of the first difference in the three-month Treasury bill rate and four lags of the log first difference in the real stock price index as measured by the S&P 500 index. The rationale for including the two financial indicators is that they are available on an almost continuous basis and may contain much of the same information captured by consumer confidence.

For this paper, the control variables included in are four lags of the dependent variable, four lags of the log first difference in real labour income, four lags of the first difference in the real 90-day bank bill rate, and four lags of the first difference in the real stock price index as measured by the NZSE40 index. Since the composition of New Zealand household’s wealth is strongly biased towards housing, four lags of the first difference in real house prices were substituted for the stock price index as an alternative. The results, which are not reported, suggest that the stock market index is a better financial indicator for the purposes of this paper.

Notes

  • [6]Hall’s argument is that if previous values of consumption incorporated all information about the wellbeing of consumers at that time, then lagged values of other variables should have no additional explanatory value once lagged consumption in included. Hence, the best forecast of next period’s consumption is this period’s.
Page top