Output variability and economic growth: The case of Australia

Introduction

The determinants of economic growth as well as short run fluctuations (i.e., business cycles) around a non-stochastic trend have been of academic and policymakers' interest for a long time. Generally, business cycle models explain short-run variations in aggregate output and employment and the traditional as well as the new endogenous growth models attempt to explain long-run economic growth. Generally, research in these areas has been carried out independently (of each other). However, Solow (1957) finds that technology shocks influence short-run fluctuations as well as long-run growth rates. Recently, this dichotomy in macroeconomics has been critically reassessed. For example, Black (1987) hypothesizes that economies face a positive risk-return trade-off and, therefore, one should find a positive relationship between volatility and growth. Conversely, Woodford (1990), Bernanke (1983), Pindyck (1991) and economists as far back as Keynes (1936) have argued and found evidence of a negative relationship between volatility and growth. However, Friedman (1968) implicitly argues that business cycles are independent of output around its natural growth rate. Understanding the determinants of economic growth is a central goal of macroeconomics because it has welfare implications.

The objective of this paper is to examine the relationship between output variability and the level of economic growth for Australia.

Theoretical and Empirical Literature

Economic growth and its determinants have been a constant preoccupation of economists for a couple of centuries. For example, Smith (1776), Ricardo (1817), and, much later on, Ramsey (1928), Knight (1944), and Schumpeter (1934) provided the basic building blocks that are now incorporated in the modern growth theories. The main theoretical findings from this literature have centred on the role of diminishing returns and its relationship to the growth of physical capital and the growth rate of the labor force. Other main findings have included the relationship between income and the growth rate of population, the effects of technological progress in terms of increased specialization of labor, and the advantages of having monopoly power in providing an incentive for technological advancement. Solow (1956) and Swan (1956) make a further and more important contribution. A prediction from this work was that per capita income growth must grind to a halt. However, Romer (1986) demonstrates that it is possible to devise mathematical models in which there are spillover benefits of investment that will prevent an economy from grinding to a halt. Over the past 15 years, this technical breakthrough has sparked a great deal of academic interest in unraveling the factors that influence economic growth. Most studies have focused on cross-country comparisons. However, time series analysis on the causes of growth has been quite sparse.

As Caporale and McKiernan (1996, 1998) correctly assert, there is no theoretical consensus regarding the direction of association between output variability and economic growth. In a nutshell, there are three broad schools of thought. The first school of thought can be attributed to Black (1987). Black's hypothesis is that there should be a positive relationship between output growth and volatility-that is, investment will only be undertaken if the expected rates of return are sufficiently high to compensate for the greater risk. Caporale and McKiernan (1996), employing a GARCH-M model and monthly U.K. data for the period 1948-1991, test this hypothesis and find a positive and significant relationship between output variability and U.K. growth. In a related paper, Caporale and McKiernan (1998), employing an ARCH-M model and U.S. annual data for the period 1871-1993, test and reaffirm their broad support for the Fisher business cycle hypothesis. Kormendi and Meguire (1985, p.148) also test Black's hypothesis for 47 countries, and their evidence suggests "that the risk return trade-off facing countries yields approximately 1 percent greater economic growth in exchange for an increase of 2 percent in the standard deviation of the rate of economic growth." Grier and Tullock (1989, p. 264) find, using annual data on 113 countries to construct a pooled cross-section and time series data set and controlling for other explanatory variables (ie., initial per capita real GDP, the growth of government's share of GDP, population growth, inflation, etc.), that the variability of GDP growth (as measured by the standard deviation of growth) "is also positive and significant, indicating there is a modest historical aggregate trade-off between risk and return." However, their empirical results from this study are reported with several caveats imposed on them. Sandmo (1970) in a theoretical paper hypothesizes a positive relationship; however, the basis of the argument is from a different perspective. Sandmo postulates the notion that greater output and income variability leads to an increase in savings which therefore leads to higher growth via increased investment. In a parallel argument, Mirman (1971) has argued that higher volatility will lead to an increase in saving and this will therefore induce a greater rate of investment. Thus, if it can be demonstrated that there is a positive relationship between investment and growth, then this will lead to an increase in growth.

The second broad school of thought argues that there is an inverse relationship between output variability and economic growth. This hypothesis emphasizes the importance of entrepreneurial expectations on investment. It is argued that the level of risk increases when there are fluctuations in economic activity and this may, in turn, reduce the level of investment and output growth. Uncertainty has long been regarded as a significant influence on the decision to invest. Keynes (1936) argued that large fluctuations in investment were inevitable since the future was not observable:

Most, probably, of our decisions to do something positive, the full consequences of which will be drawn out over many days to come, can only be taken as a result of animal spirits-of a spontaneous urge to action rather than inaction, and not as the outcome of a weighted average of quantitative benefits multiplied by quantitative probabilities. (Keynes 1936, p.161)

Woodford (1990) reinvigorates this hypothesis by examining sunspot equilibria. An important negative effect on investment under uncertainty arises when that investment is irreversible and the firm has some discretion as to the timing of the project. Pindyck (1991) and Bernanke (1983) find that this irreversibility in investment results in an inverse relationship between volatility and investment. Ramey and Ramey (1991) find, using data on 92 countries and a sample of OECD countries, that economies with higher volatility have lower economic growth. Also, Zarnowitz and Moore (1986) find, using U.S. data, that real GDP growth rates are highest in periods when the standard deviation of output is relatively lower compared to other periods.

The third broad school of thought is one in which there is no a priori relationship between output variability and economic growth. For example, Friedman (1968) implicitly argues that fluctuations of output about its natural growth path are independent of each other. Fluctuations in output around a non-stochastic trend are caused by price misperceptions due to monetary shocks. In other words, the growth rate of output is determined by real factors such as skills, technology, and other real factors. Speight (1999), using a post-war U.K. monthly industrial production index over the period 1948-1994 and ARMA-GARCH-M models, finds a positive but insignificant effect of output volatility on the growth rate of output. Speight (1999, p. 183) argues that the insignificance of the empirical evidence, "contrary to recent results in support of models stressing their dependence of entrepreneurial investment on the existence of a positive risk-return output trade-off, might therefore be interpreted as more supportive of macroeconomic models which dichotomize the determination of output growth and variability according to real and nominal factors respectively."

The results of the KPSS unit root tests without and with trends are given in Tables 1 and 2. The results are unambiguous. For tests with and without trends, LNIP and LNGDP are found to be nonstationary. We reject the null hypothesis of stationarity because the test statistic exceeds the critical value. On the other hand, GIP and GGDP are found to be stationary. Thus, our analysis is limited to GIP and GGDP.

Once again, the Ljung-Box Q-statistics up to lags of 12 do not indicate any serial correlation. The Jarque-Bera (1987) test of normality is an asymptotic or large sample test. Jarque-Bera showed, under the null hypothesis that the residuals are normally distributed, that the Jarque-Bera statistic follows a chi-square distribution with two degrees of freedom. Our results do not reject the null hypothesis, and therefore we conclude that the evidence suggests that the errors are normally distributed. The mean equation indicates that the growth rate of industrial production is negatively related to the conditional standard deviation. The coefficient is significant at the 5 percent level. This supports the Keynesian position that output variability causes growth to be lower.

Conclusion

This paper examines the relationship between output variability and growth for Australia using ARCH-M models. Our results show that output volatility has a negative and significant impact on economic growth. These results are consistent with the work of Zarnowitz and Moore (1986) and Ramey and Ramey (1991). However, our empirical results are inconsistent with the work of Black (1987), Mirman (1971), and more recently the empirical work of Caporale and McKiernan (1998). Thus, our results tend to support the Keynesian notion that output variability would be expected to have a negative impact on the growth of an economy.