The Case of the Immaculate Disflation
Trust the Markov Process
One of the biggest stories of 2023 was the fairly rapid disflation that occurred over the year. Underlying inflation fell by more than 2 percentage points in both the US and Australia - even more surprisingly this occurred without a substantial rise in the unemployment rate!
This immaculate disinflation surprised most economists. But should it have?
Given economists were surprised by how 2023 unfurled, one might expect the model that underpins most of their thinking to be wrong or at least misleading.
I have previously written about RBA’s NAIRU model and how supply-chain effects can be added to better take into account the effects of Covid-19. This is a very mainstream, conventional model of the economy in which inflation is driven by a combination of inflation expectations, slack in the labour market and supply side shocks.
So how did this normie, mainstream model perform over the past year?
There is no simple way to determine whether a model is right or not. But a sensible health check every economist should do is look at the equation error terms and see whether they seem to be relatively “normal”. Residuals effectively represent everything about economic data that the model cannot explain. If the residuals are large or persistent, or seem to explain a large share of economic data, this is probably a sign that your model isn't performing too well.
As you can see from the plots below there is one clear period in which the model performs terribly. In March 2020, the unit labour cost rose by 5% in a single quarter. This was largely driven by changes in the composition of the labour market caused by the Covid-19 lockdowns.
I think this is a good example of how a large, outlier residual indicates a failure of the model. This is fairly unsurprising - lockdowns have never been well modelled by conventional pre-pandemic economic models for the obvious reason that they had never occurred before!
But when we look at the past four quarters the residuals look much more normal. Indeed if anything they are smaller than the historical norm. The residuals for the inflation equation are particularly small. This indicates that the current, moderately high inflation rate is fairly well explained by the combination of:
an unemployment rate that is slightly below the NAIRU (which helped keep it high),
well anchored inflation expectations (which helped stabilise it),
and the dissipating supply shock shocks combined with lagged effects of higher inflation earlier in the year (which mostly explain the recent fall).
Indeed, if an economist had followed this model, they would have forecast the current rate of inflation remarkably well (assuming they could forecast the unemployment rate fairly well)
Finally, when we look at the residuals for the NAIRU equation (the models best guess at how the NAIRU is fluctuating over time) we again see a very small deviation from the historical norm.
In short trusting a conventional NAIRU model of inflation would have mostly steered you in the right direction over 2023 - at least for Australian data. A good model of the economy should allow for both supply and demand shocks. When you create a model that accounts for both supply and demand shocks, the case of the immaculate deflation becomes fairly straightforward to explain.
Which is probably why the RBA did a fairly good job of forecasting it!