RDP 2007-09: Private Business Investment in Australia 1. Introduction

Despite the importance of private business investment as a determinant of the long-run growth potential of the economy and a contributor to short-run fluctuations in the economic cycle, very little empirical work has been undertaken on Australian investment aggregates over the past decade. The rare exceptions include Bond and Hernandez (2003) and Swift (2006).

This paucity of empirical work in Australia probably reflects the well-known fact that modelling investment is difficult. One reason for this is that investment is a highly heterogeneous activity. The need to model equipment and structures investment (which includes new non-residential building and engineering) separately has long been recognised, but some recent papers have also highlighted some potential sources of misspecification if computing equipment investment is not treated separately as well. This is because of the dramatic fall in the price of computing equipment over the past two decades and its relatively high rate of depreciation – in line with strong technological advances in computing. To this end, computers are removed from investment in the analysis that follows.

A second challenge arises when modelling the volatile short-run changes in investment. In this paper, standard techniques are used to exclude observations that are identified as influential in short-run regressions. This can be justified to the extent that investment is especially lumpy and/or suffers from measurement errors at a quarterly frequency. Between 5 and 10 per cent of observations are excluded from the parsimonious regressions, which is found to improve the robustness of the estimates.

Owing to the relative lack of empirical success of more recent innovations to models of investment (such as Euler equation models – see Oliner, Rudebusch and Sichel 1995), the estimations are based on the traditional neoclassical model of investment (Jorgenson 1963). In its strictest form, the traditional neoclassical model implies that the capital-to-output ratio and the cost of capital should be inversely related; however, in this paper no empirical support is found for such a relationship. Instead, we adopt an approximation of the neoclassical model suggested by Bean (1981), which implies an inverse relationship between the investment-to-output ratio and the cost of capital in equilibrium. Evidence is found to support such a relationship for the main components of investment (equipment, building and engineering).

Other relevant factors are found to influence investment (over the short run), although these vary across components. For equipment investment, swings in business confidence (a highly cyclical variable) and movements in the real exchange rate (RER) are both found to be influential. For engineering investment, movements in the terms of trade are particularly relevant. The building investment equation has the least-rich dynamics, with the error-correction term the dominant influence.

As a precursor to the modelling work, Section 2 provides a discussion of investment trends. Over a long history, the most outstanding feature has been the permanent increase in the investment share of GDP which occurred mid last century. Over the past three decades or so, better quality data allow differences across industries and types of investment to be explored. After a discussion of the methodology and empirical literature in Section 3, an empirical examination of the main components of investment is presented in Section 4. The conclusions of the paper are provided in Section 5.