RDP 2006-01: Modelling Manufactured Exports: Evidence from Australian States 1. Introduction

Exports form an important part of the Australian economy, accounting for around 20 per cent of total GDP and around 30 per cent of average annual GDP growth over the past decade. Despite this, there has been a surprising paucity of research that attempts to model the determinants of Australian exports. This is particularly evident when the focus is narrowed to particular broad categories of exports; most recent studies have modelled total exports in a single framework, despite the accepted wisdom that agricultural and resource exports are supply determined, while manufactured exports are largely demand determined.

This paucity of research may in part be due to difficulties finding robust results in export models. Australian estimates, like those of other countries, tend to be characterised by income and price elasticities which are quite sensitive to changes in model specification, even when similar data, methods and sample periods are used. One method that might allow for more robust estimation is to separately model exports from each of the Australian states and then combine these results into a single implied national estimate. This method has the potential to result in more robust estimates of national elasticities by taking advantage of cross-state variation in the regressand and regressors. This strategy has been used successfully in other contexts, such as to reduce the problem of collinearity in estimates of consumption functions (Case, Quigley and Shiller 2005 and Dvornak and Kohler 2003), and to mitigate the effects of technological innovation through time on estimates of the income elasticity of money demand (Fischer 2006).

Such cross-state variation is inherently present in state manufactured exports, but it is not immediately clear that there is variation in the determinants of manufactured exports. However, if allowance is made for differences in the trade orientation of each state's exports, it is possible to produce price and foreign income series that are more closely matched to the conditions facing the average exporter in each state, and which vary across states. This is the approach used in this paper. It is found that there is indeed quite marked cross-state variation in the share of exports going to each trading partner, and that this has a noticeable impact on the profile of state-specific real exchange rates and trading partner GDP. It is also found that there is some variation in the coefficients of each state's model, which further distinguishes this approach from conventional, national, estimates.

The remainder of this paper is structured as follows. Section 2 reviews the previous research on modelling exports. Section 3 discusses the construction of the state-specific data used in the estimation and examines the cross-state variation. Section 4 presents the econometric framework used to model exports, while Section 5 provides results. Section 6 then extends the analysis to include a possible role for domestic influences to affect export outcomes. Section 7 concludes.