Research Discussion Paper – RDP 27 Earnings and Award Wages in Australia


The determinants of average earnings are currently an important issue in most industrial countries, if only because of the links between earnings and prices. This paper examines the determinants of earnings in Australia, building on the overseas literature, and previous Australian studies. The approach is econometric and is part of a project to build and refine a quarterly econometric model of the Australian economy. Our basic dependent variable is non-farm wages, salaries and supplements divided by non-farm employment; we chose this variable rather than the Commonwealth Statistician's average weekly earnings series to avoid the need for an extra equation to link it with the national income total in the model. However, for simplicity we call our series ‘average earnings’ in the discussion.

In Australia, there is considerable evidence that the Arbitration Commission plays an important role in the determination of earnings. The Commission makes decisions about key award wages which flow on to various state awards and other agreements. As the factors which influence awards are quite different to those which influence earnings, one has the choice of including award wages or the determinants of award wages in the earnings equation. If the former approach is adopted it is desirable to explain movements in award wages as this is clearly not an exogenous variable, despite the fact that the Commission is not solely concerned with ‘economic’ issues. This exercise yields some interesting insights which are extremely difficult to capture in a single equation for earnings.

It is now conventional to construct equations to explain the rate of change of earnings and usually the annual percentage change of earnings. Although this may be partly a result of historical ‘accident’, it may most nearly reflect the workings of the market; in particular if the change in wages is related to excess demand in the labour market, among other things. Despite the tradition in favour of a rate of change formulation, equations in terms of the level of earnings may perform better. Hence we also constructed equations to explain the level of earnings and compared their performance with the rate of change equations. The levels equations were not satisfactory as they contained severe first order serial correlation, and when this was adjusted by a mechanical procedure the labour market variable became very insignificant, a theoretically implausible result.