Industry-level stock returns volatility and aggregate economic activity in Australia

Abstract

Drawing upon rationales from the theories of investment and consumption under uncertainty and the models of sectoral reallocation, we assess the implications of industry-level stock returns volatility for the future state of the Australian economy in terms of real Gross Domestic Product (GDP) growth, inflation and unemployment. By explicitly modelling the cyclical pattern of industry-level volatility and relating it to corresponding cyclical behaviour of macroeconomic variables, we show that industry-level volatility is a leading indicator of the movements in output growth and inflation. We find complementary evidence from a Vector Autoregression (VAR) based multi-step Granger causality test and impulse response analysis. However, the forecast error variance decompositions suggest that although the industry-level volatility accounts for a significant fraction of the forecast error of inflation, this explains only a small fraction of output and unemployment uncertainties. Further analysis indicates that industry-level volatility contains better information about the future state of the economy than does aggregate stock market volatility

Acknowledgements

The author gratefully acknowledges the assistance of Dr Craig Ellis and Dr Girijasankar Mallik while working on this article. This article has also benefited from helpful comments on earlier drafts from Dr Kevin Daly and other participants at the College of Business Research Symposium 2006 at the University of Western Sydney. The usual disclaimer applies.

Notes

¹ See also Hwang and Satchell (2001) for a discussion and related evidence from a more statistical perspective about why cross-sectional measures of volatility (e.g. industry-level volatility) may contain better information than a time series measure of volatility (e.g. aggregate market volatility).

² We also had to apply some additional filters such as removing nontrading days from the raw data to keep our measures of volatility from potential data errors.

³ As 90-day dealer bill rate is not available from DS until late January 1976, we derive daily risk-free rate from monthly yield on 90-day bank-accepted bills available from OECD database over the period 2 January 1973 to 30 January 1976. In doing so we assume, as in Campbell et al. (2001), that the daily risk-free rate is the rate which compounds to the monthly 90-day bank-accepted bills rate over the number of calendar days in a month.

⁴ It is widely held that Paul Keating's, the then Treasurer of Australia, remark on 14 May 1986 in a radio interview had an almost immediate negative effect on the value of the Australian dollar and, so far as the economic statistics over the next year are concerned, had a significant effect on the broader Australian economy.

⁵ Stationarity requirement is not the only deficiency of cross-correlation analysis. A high correlation may be the result of a single shock common to two series that are otherwise independent (McDermott and Scott, 2000). Concordance statistic, on the other hand, does not suffer from sudden jumps in series.

⁶ Though Harding and Pagan (2001) are generally against any kind of detrending of data, they suggest subtracting only a linear deterministic term if any component at all has to be removed from the series.

⁷ We include a trend term in the VAR following an evidence of trend component in INF and DUR while performing unit root test. Later, on the basis of formal likelihood ratio test we could not accept the null of no trend in the estimated VAR.

⁸ The estimated VAR is found to be stable as all the eigenvalues of the stacked companion matrix remain inside the unit circle. Further, VAR residual serial correlation test does not find evidence of significant first- or fourth-order serial correlation.

⁹ Lag order of the extended VAR model is also 4 as suggested by the AIC.

¹⁰ According to the definition of Brainard and Culter (1993), industry-specific returns include both the deterministic and stochastic components of Equation 3 presented in Section III.

¹¹ We do not report the results of such exercises to conserve space in the paper, but they are available on request.