RBA Annual Conference – 1998 Discussion

1. Philip Lowe

I would like to start off with a question. It is a question that John's paper hints at, but one that never quite gets asked directly. The question is motivated by John's observation that while the distribution of wages has widened in Australia, it is still more compressed than in some other countries with lower unemployment rates. My question is the following: to what extent is the persistently high unemployment rate in Australia attributable to too high an aggregate real wage, and to what extent is it due to a lack of relative wage flexibility?

Figure 1 helps make the question a little more concrete. It shows the unemployment rates for workers with post-school qualifications and for those without post-school qualifications. Three observations can be made. First, both unemployment rates are high. Second, the unemployment rate for workers without post-school qualifications (largely, unskilled workers) is much higher than for those with post-school qualifications. Third, this difference between the two rates has tended to increase through time.

Figure 1: Unemployment Rates by Level of Education

The concrete question is to what extent are the high unemployment rates of both categories of labour, but particularly the high unemployment rate of unskilled workers, attributable to too high an aggregate real wage, and to what extent are they attributable to too high a relative wage for unskilled workers (on average, those without post-school qualifications earn around two-thirds of their more highly educated workmates). Putting this another way, if it were possible to reduce the average wage, but keep wage relativities constant, would we fix the unemployment problem for unskilled workers? Or is the solution, an adjustment in relative wages, and if it is, why hasn't it happened?

I don't think that anybody can be confident that they really know the answers to these questions, so what I would like to do is to use John's paper to try to throw some light on the question. In particular, I would like to pick up on four points which the paper makes which I think are relevant:

  • while relative wages tend not to move much, they do at least help in the adjustment process;
  • not a lot is known about the wage elasticities for different types of labour;
  • quantity signals play an important part in resolving disequilibrium in the labour market; and
  • credential creep is a common phenomenon when unemployment is high.

I will say a few words about each of these points.

John makes the point that relative wages tend to be stable through time, but that they do change to help correct disequilibrium. In general, the changes occur only slowly, sometimes very slowly. However, the examples that John gives point to an asymmetry: if there is a shortage of a particular type of labour – say world-class leg-spin bowlers or visionary chief executives – relative wages can adjust very quickly and by large amounts, but if there is excess supply of some type of labour – say unskilled workers – the adjustment tends to be slower and less dramatic. Reading behind the lines, John mainly attributes this slow adjustment in Australia to social and political factors. These factors are reflected in our institutions and in policy approaches that are acceptable to the electorate. I think it is difficult to argue with the proposition that a different endowment of these social and political factors would significantly change the microeconomics of the labour market, and could make a major contribution to reducing unemployment. The experience of the United States suggests a possible path, but it is one that many people feel uncomfortable with.

The second relevant point that John makes is that we know relatively little about the wage elasticities for different types of labour. However, we do know a few stylised facts. Amongst these are the following:

  • it is difficult to substitute unskilled labour for skilled labour;
  • capital and skilled labour are not good substitutes, and in some cases may well be complements; and
  • unskilled labour and capital are reasonably good substitutes.

These ‘facts’ suggest that the primary effect of reducing the relative wage of unskilled workers would be to lead to a substitution away from capital and towards unskilled labour. Unfortunately we have no robust estimates of exactly how strong this effect is, and what the lags are likely to be, and some people might even argue that my stylised facts would better be termed ‘stylised assertions’. This makes the debate very difficult, for nobody can confidently say how much the wage distribution would need to widen to significantly reduce the unemployment rate for unskilled workers. As a result, one sees guesstimates all over the place, and I suspect this makes it harder for the issue to be debated productively.

The third point is that quantitative signals play a key allocative role. A way of restating this point is that supply-side adjustments are important. For example, if the unemployment rate for unskilled workers is high, some of the unskilled workers will turn themselves into skilled workers, and workers entering the workforce for the first time will be more likely to seek post-school qualifications than would otherwise have been the case. In some sense this quantitative signal has been having the right effect for the past two decades, as young people respond to the high unemployment rates of unskilled workers by staying at school longer and enrolling in tertiary education. Another example is the migration of labour from regions or states with relatively poor employment prospects (say Victoria in the early 1990s) to those with brighter prospects (e.g. Queensland).

These supply-side adjustments are important in limiting the divergence of unemployment rates across states and between skilled and unskilled workers. But the adjustments can be painfully slow, as evidenced by the persistently high unemployment rate for unskilled workers, and the persistently high unemployment rate for Tasmanians. Despite the slow adjustments, the fact that relative wages also change only slowly means that quantity adjustments play a vital role in the adjustment process. You can see this clearly in the employment data by state. When unemployment in a state is higher than the national average, there is very little adjustment in the state's relative wage, with most of the adjustment occurring through internal migration (Debelle and Vickery 1998). In general, if these supply-side adjustments were quicker, a lack of relative wage flexibility may not be too costly, and the real issue affecting unemployment would then be the average real wage, and not the distribution of wages.

The fourth point is the existence of ‘credential creep’ when unemployment is high. This effect might mean that a decline in the aggregate real wage is the primary solution to the absolutely and relatively high unemployment rate of unskilled workers. An important question is: could the unemployment rate for unskilled workers be disproportionately high, even though relative wages are consistent with full employment? I think the answer is probably yes. If the aggregate wage is too high, some firms are likely to choose skilled workers to do unskilled jobs – so that when unemployment is high one needs a PhD to get a lecturing job at a university, but when unemployment is low, a Masters degree is sufficient. As a result of this credential creep, unskilled workers end up being disproportionately unemployed. The solution is not to lower their relative wage, but instead to lower the aggregate wage.

So where do John's four points leave us? As usual, there is no clear answer. My reading of John's paper is that the evidence suggests that a fall in the relative cost of unskilled labour would increase the number of unskilled workers employed. But so too would a fall in the aggregate real wage, even if the wage relativities remain unchanged. Further, it is a plausible, although admittedly untested proposition, that aggregate wage restraint is more effective than allowing the wage distribution to widen further.

If this proposition is true, how do we achieve the necessary aggregate wage restraint. One solution is an incomes policy – such as the Accord. However, rather than focus on the merits and costs of this approach, I would like to think about another possibility; that is, does allowing greater flexibility of relative wages generate aggregate real wage restraint. In my comments earlier, I have talked about the two things as being alternatives, but as John reminds us a few times in his paper, everything is connected to everything, so perhaps they are complements, rather than substitutes.

I think it is a reasonable proposition that allowing the bottom end of the real-wage distribution to fall puts pressure back on the rest of the distribution, leading to greater aggregate restraint. Certainly, over recent years, those countries with high and/or growing wage inequality have tended to experience restrained growth in aggregate labour costs. If those at the lower end of the wage distribution are experiencing negative or low real wage growth, this is likely to put some pressure on the wages just above them in the distribution, and in turn increase the pressure on the wages above them and so on. Obviously one cannot push this argument too far: wage restraint for cleaners is unlikely to lead through a cascading of competitive effects to wage restraints for chief executives. Nevertheless, arguably, one of the benefits of easing the social and political constraints on the dispersion of wages, is to deliver aggregate wage outcomes which ensure that on average, more of those who want to have jobs actually have them. Again, this proposition is untested, but I think is worthy of exploration. There may be other alternatives as well.

Finally, I would like to end where I should probably have started. That is to congratulate John on an enjoyable and comprehensive paper. John's analysis gives us a useful framework for thinking about the microeconomics of the Australian labour market. But at the same time it reminds us of that fact that we know relatively little about the strength and speed of the underlying adjustment mechanisms that most of us believe are out there operating in the marketplace. As usual, more work is needed!


Debelle, G. and J. Vickery (1998), ‘Labour Market Adjustment: Evidence on Interstate Labour Mobility’, Reserve Bank of Australia Research Discussion Paper No. 9801.

2. General Discussion

Discussion of this paper focused on two issues:

  • the interaction of education and the supply of skilled labour with the degree of wage dispersion; and
  • the evidence on the effects of changes in the minimum wage on employment.

Participants noted that both the supply and the demand for skilled or educated labour is important in determining the returns paid to skill and education. In Australia, there is not much evidence of an increase in the education premium because the increase in demand has been generally matched by an increase in supply. However, in the US and the UK, the widening in wage dispersion in the 1980s and 1990s was in part due to an increase in the return to skill. This was the result of a deceleration in the supply of educated labour over that period, following the large increase in supply in the 1970s – in part a function of demographics.

Some participants noted that despite the rise in the education premium in the US and the UK, a large part of the increase in wage dispersion has been caused by a rise in within-group inequality. That is, the dispersion of wages has also increased considerably within the group of (say) educated workers. To date, it has been hard to discern what has been driving the widening in within-group inequality. This poses problems in determining the appropriate policy response.

One participant noted that, given the relatively high return to additional years of schooling, it was difficult to ascertain why unemployment should be so much higher in Australia today – when the workforce had on average two years of extra schooling – than it was in the 1960s. The rise in education should have translated into a decrease in real unit labour costs which should have boosted employment. This may indicate that the education level of the workforce may not be a critical cause of unemployment (at least for a workforce with the education level of that in Australia).

The controversial findings of Card and Krueger on the impact of changes in the minimum wage on employment generated much discussion. One participant argued that their much-cited results based on the fast food outlets in New Jersey and Pennsylvania were not that persuasive. However, the other evidence that Card and Krueger presented using time-series evidence was more robust and suggested that rises in the minimum wage had modest negative effects on employment. Another participant noted that this conclusion was supported by the results of a panel study across a number of OECD countries.

The discussion also highlighted that it is important to remember that the minimum wage research does not necessarily estimate aggregate labour demand elasticities. The research is only focusing on a wage change that affects a small section of the labour force. In determining an aggregate labour demand elasticity, it is necessary to consider the impact of minimum wage changes on the average wage in the economy. Therefore, one needs to assess how changes in the minimum wage are likely to flow through to other wage rates in the economy.

Finally, some participants questioned the relevance of the evidence from the US on the effect of changes in the minimum wage for Australia, given the different set of labour market institutions in the two countries, given that the minimum wage is much lower in the wage distribution in the US, and given that Australia does not have a single minimum wage but rather a myriad of minima.