RDP 2000-08: Nominal Wage Rigidity in Australia 5. Further Evidence of Downward Nominal Wage Rigidity

5.1 The Relationship with Inflation

To identify whether there are conspicuous differences in the distribution of wage changes in different inflation environments, in Figure 2 we break up our sample into three periods that roughly capture episodes of high, sharply falling and low, stable inflation. At first glance, the features of the distribution of wage changes are different in a high-inflation period. The distribution is clearly more dispersed during the high-inflation episode than when inflation is either falling or low. This is especially so when we make an adjustment for the fact that the spike at zero in the high-inflation episode is affected by changes to tax legislation that encouraged firms to hold base salary fixed when offering employees fringe benefits.^[26]

Summary statistics of the distribution are particularly useful because they can show us how the skewness of annual wage changes has evolved over time. In Figure 3, we present three standard measures of skewness. All point to a general increase in skewness and have the common feature that the rise in skewness is most pronounced during the early 1990s. There has been a general fall in inflation over the period, with a sharp decline in the early 1990s. Consequently, we find a clear negative correlation between the various measures of skewness of wage changes and inflation, consistent with downward nominal wage rigidity (Table 2). This is the case when we consider headline inflation and core inflation (measured by the median price change). The skewness of annual wage changes is also negatively correlated with the inflation expectations for that year. In fact, this correlation is stronger than that between skewness and headline inflation.

5.2 Skewness Near the Median

While an inverse relationship between skewness and inflation is compelling evidence that downward nominal wage rigidity exists, an indication of the extent of this rigidity is important for assessing its macroeconomic consequences. The usual thought experiment is to ask ‘what would the wage-change distribution look like in the absence of rigidity?’. The difference between this counterfactual distribution and the actual distribution captures the extent of wage rigidity.

What should the counterfactual distribution look like? Most attempts have posited that it is symmetric, with Card and Hyslop (1997) providing the most prominent example. But while we cannot ever know what the counterfactual distribution looks like, it may be too restrictive to assume that it is symmetric. It is possible that even if wages were perfectly flexible, shocks to wages may not be symmetrical, so that the underlying distribution of wage changes is skewed. If, as a result, the underlying distribution of wage changes is positively skewed, imposing a mirror-image assumption will exaggerate estimates of downward nominal wage rigidity.

McLaughlin (1999) argues that if a shortage of wage cuts were the main source of skewness, then wage changes close to the median should be symmetric, since these positive observations are not affected by factors that prevent nominal wages from falling. If, instead, skewness is present near the median, it implies that the distribution of wage changes is skewed for reasons other than downward nominal rigidity.

We trim 20 per cent of observations from both tails of the full sample distribution of wage changes. That is, we trim the left-hand tail that encompasses the zero and near-zero observations that may be affected by downward nominal rigidity, and we trim the right-hand tail that encompasses extremely high observations that can inflate measures of skewness. This leaves a central core of observations around the median. We find that there is some skewness near the median, indicated by a small positive skewness coefficient, suggesting a possible role for factors other than downward nominal rigidity in the distribution of wage changes.^[27]

Footnotes

With the initial introduction of the Fringe Benefits Tax in 1988, employers incurred a tax liability for the provision of fringe benefits and many chose not to increase base pay when benefits that attracted the new tax were awarded (MCED, Director of Databases, personal communication, 12 May 2000). Consequently, a more useful indicator of the share of wages that recorded no change can be found from a broader measure of earnings that includes the cost to employers of Fringe Benefits Tax (see Figure 2). [26]

The skewness coefficient of the core observations for the full sample period is 0.37. Even where 30 per cent of observations are trimmed from both tails, a small degree of skewness is evident. [27]