Although economic squabbling is fun to follow, a lot of it is the result of the use of fuzzy language. As a result, there is no way of advancing the conversation; arguments are just people clinging to different definitions. The use of mathematics in economics is supposed to eliminate this squabbling; unfortunately, the mathematical models themselves rarely fit reality. However, we need to translate the debates into operational discussions, to see whether they can be applied to the real world. If we turn to my previous article about NAIRU, we need to ask ourselves -- does the definition of NAIRU we are using pass the Santa Claus test?
One defense of NAIRU is that the concept is OK, but we just have a hard time measuring it. This fails the Santa Claus Test. It is essentially saying:
If the unemployment rate is below some unknown value, inflation will start to accelerate higher.
This statement conveys no more useful information than saying:
If a country is on Santa Claus' naughty list, inflation will start to accelerate higher.
It should be noted that NAIRU was not supposed to be a hand-waving concept, like those darned post-Keynesians throw around.* Instead, NAIRU is a well-defined time series, as shown at the beginning of this article. When I am referring to "NAIRU," it is that family of well-defined time series. And as was discovered in the 1990s, NAIRU offered concrete predictions -- and failed. (I discussed this in Section 3.4 of Interest Rate Cycles: An Introduction.)
I have a low opinion of the value of mainstream research, and so I no longer attempt to follow it closely. My understanding is that modern researchers have given up on NAIRU -- in much the same way that the NAIRU replaced the analytical failure that was the "natural rate of unemployment." These new measures may be qualitatively similar, but are calculated in a different fashion. Since they are generated by different modelling techniques, I feel makes no sense to refer to them as "NAIRU models." Whether or not they are useful depends on analysing each model.
Why 4%?I want to get back to a passage from Simon Wren-Lewis that I previously quoted.
If we really think there is no relationship between unemployment and inflation, why on earth are we not trying to get unemployment below 4%? We know that the government could, by spending more, raise demand and reduce unemployment. And why would we ever raise interest rates above their lower bound?I did want to be critical in my original article, but I think I need to offer a longer explanation why this passage represents questionable analysis.
I’ve been there, done that. While we should not be obsessed by the 1970s, we should not wipe it from our minds either. Then policy makers did in effect ditch the NAIRU, and we got uncomfortably high inflation.
- Where did this magical 4% number come from? (My guess is that he is referring to the United Kingdom.) That is exactly like the magical levels of the unemployment rate that hawks said would cause accelerating inflation in the 1990s. (Those levels were repeatedly revised lower in response to inflation stability.)
- Just because policymakers made a mistake in the 1970s, does not mean that policy could be done correctly. The Old Keynesians either did not understand, or they did not care about the inflationary bias of their policies.
For example, if the government actually cared about reducing underemployment, it could do the following policy changes.
- A programme of direct employment at a relatively low wage; in the limit, a Job Guarantee programme.
- Raising income tax rates.
- Tweaking other spending programmes to eliminate the tendency of price rises to be propagated (indexation).
The objective of the first step is to raise incomes for those at the bottom of the income distribution. This could easily cause a one-time shock to the price level, as businesses that rely on underpaid labour get wiped out. This is why the last step is necessary -- to prevent that one-time shock turning into a sustained inflation.
I am not saying that calibrating the size of tax hikes to keep inflation in check would be an easy task; my feeling is that the initial tax changes would be a guesstimate. However, once the dust settles, the country could have an unemployment rate below the magical 4% level while at the same time having a stable inflation rate.
NAIRU With Dead Zones
The possibility that we can use NAIRU with a "dead zone" was raised in the comment section. Although this is certainly a reasonable idea, I feel that it has problems when put into practice.
The figure above shows the standard conception of NAIRU; it plots the effect of the unemployment rate on "inflation acceleration" (technically, the acceleration of the price level).
- If the unemployment rate is greater than NAIRU (UR* in the figure), there is a "negative acceleration in inflation" (falling inflation rate) -- all else equal. (Throughout this article, I am ignoring the fact that most NAIRU models have other variables that affect inflation, such as inflation expectations. Each method to estimate NAIRU uses different variables; and we often have no idea of how to forecast them in the first place.)
- If the unemployment rate is below UR*, inflation is rising.
The slope of the line (theta) is a second free parameter if we want to fit this model to data. If we allowed for a nonlinear relationship between the plotted variables, we would end up with extra free parameters that define the nonlinearity.
It should be noted that this figure only shows what is happening at a given point in time; the NAIRU level (UR*) is allowed to move around over time (as seen in the chart at the beginning of this article). The fact that this value is mobile means that a NAIRU model can always be fit to data (and hence there is no way of rejecting it). This inability to reject the model explains why it was useless for policymakers in the 1990s: in order to forecast inflation, you needed to forecast UR* as well as the unemployment rate. Even though they got their forecast of the measurable unemployment rate right, the unmeasurable UR* moved on them, and hence their inflation forecast was still wrong.
There are now three (possibly four) free parameters to this model.
- UR_LO: the level beyond which low unemployment generates upward inflationary pressures.
- UR_HI: the level beyond which high unemployment generates downward pressure on inflation.
- The slope of the lines (theta). If we allowed two different slopes, we would have a fourth free parameter.
The problem with this characterisation is that we have too many parameters, which are presumably all time-varying. The figure above shows a few curves that would be impossible to disentangle in practice.
- A baseline curve (solid line).
- A curve with no dead zone, but lower slope (dashed line).
- A curve with a wider dead zone, and a higher slope (dots).
This flexibility will allow us to have any number of potential fits to the same data, with widely separated UR_LO values. This means that we have no idea when the economy will start to overheat.
We need to look at operational differences between points of view in order to avoid being stuck in squabbling about terminology. However, the inability of NAIRU proponents to come up with a useful model makes it difficult to have a more concrete discussion.
* My attitude towards qualitative approaches to economics is somewhat complicated. Earlier in my career, I was not patient with qualitative approaches to economics -- I was an applied mathematician, and a quantitative analyst. After a couple of decades of looking at the wreckage of failed mathematical models, I now see the value of qualitative approaches. That said, qualitative economics has an unfortunate tendency to degrade into squabbling about terminology.
(c) Brian Romanchuk 2017