Wages Growth and the NAIRU in Australia

The Wage Price Index grew by 0.5 per cent in the September quarter (2.2 per cent through the year). A lot of discussion was around how weak the numbers were, especially considering the latest Labour Force outcome for October (participation rate down, unemployment rate up). The latest figures continued a period of relatively weak WPI growth in the last few years, especially compared to the period 2000-2010 where the WPI averaged around 0.9 per cent per quarter. The latest figures also cast doubt on forecasts in the Federal Budget which have wages growing by 2.75 by mid-2020.

Let's investigate the wage growth numbers in the context of a simple model.

The model we use is a simplified expectations-augmented wage Phillips Curve. This states that the growth in wages is a function of expected inflation, labour productivity and the gap between the unemployment rate and the NAIRU.

\(w=\pi^e+\alpha.prod+\theta(urgap)\)

This specification presents some difficult measurement issues: how do we measure the 3 RHS variables.

1) Labour productivity is the simplest one because we can find the variable directly in the national accounts. However, there are non-trivial issues for the ABS to measure it and it is quite a noisy variable, so often we use a smoother or a trend version. In the RBA's recent MARTIN model, they use state-space models to contruct trend labour productivity, which is currently growing around 1 per cent (annualised). You can find the code and data to recreate the analysis below, which gives a roughly similar view (although trend growth is slightly lower at around 0.8 per cent).

Around 2012, trend labour productivity was growing at a quarterly rate of around 0.3-0.35 per cent per quarter. However, in the most recent period, the rate has slowed to around 0.25 per cent. Therefore one potential reason for the slowing in the wage price index in recent periods has been the slowing growth of labor productivity (not however, that the WPI is somewhat quality adjusted so the pass-through from productivity to wages is probably less than one).

2) Inflation expectations: we don't have a direct measure of inflation expectations so instead we generally use survey measures or implicit measures from financial markets. These can be either directly asking groups or inferring expectations from financial markets.

As seen in the above chart from the RBA's November 2019 SMP, survey measures indicate that inflation expectations are for inflation to remain roughly at the mid-point of the target band (2.5pct annual). Looking at the second chart in the panel, we see that measures inferred from financial markets indicate that expectations have recently fallen to be below the target band (see this bulletin article for issues around market measures of inflation: LINK). The inflation expectation from markets is currently 1.3 per cent, well below the target band.

3) NAIRU: The NAIRU is another unobserved variable which needs to be estimated. The standard method is to use a state-space model (LINK). In this framework, a Wage or Price Phillips Curve is used, with the Kalman Filter used to extract an estimate of the NAIRU. Effectively we invert the above relationship to calculate the NAIRU to best fit the wage or price data.

For this post I want to focus on 2) and 3) taking as given the trend labour productivity numbers. I have recreated the RBA NAIRU model in the Eviews code at this link (Eviews NAIRU Codes). In the baseline model, the estimated final value of the NAIRU is 4.6 per cent vs the RBA estimate of 4.5 per cent. This would imply an unemployment gap of around 0.13 per cent (or 0.6 percentage points).

In the Eviews code I also experiment with using a different measure of inflation expectations (the RBA model uses a 'trend' measure which has expectations stable at around 2.5 per cent for most of the inflation targeting period).

As an alternative, we can instead estimate the same system but now use bond market expectations of inflation. For the purposes of comparing the impact of changing measures of inflation expectations, I fix the coefficients of the model at the values in the RBA version. Using these parameters and bond-market expectations the NAIRU measure is now 5.2 per cent, more or less equal to the actual unemployment rate.

What are the implications of these two estimates of the NAIRU for interpreting wage growth? If we use the RBA measure of inflation expectations, one reason for slow wage growth is the negative unemployment gap (moving along the Phillips Curve). In this case, the slowing wage growth will help lower producer real wages and encourage employment (until the gap is closed). If however, we use the bond-market measure, the interpretation changes. Now, the wage numbers are not showing labour market slack but rather falling inflation expectations (a shifting inwards of the Phillips Curve).

Which of these interpretations is correct is of course impossible to know. We don't directly observe any of the RHS variables of our simple Phillips Curve and the most recent data is always subject to potential data revisions. However, given the run of recent below-target inflation figures (which the RBA forecasts to continue until end 2021), the second interpretation becomes more likely and predicts continued wage growth outcomes like we saw in the September 2019 data release.