The Fed's Phillips curve puzzle

The steady recovery in the US economy since the middle of 2009 has pushed the country’s unemployment rate close to its lowest level in nearly half a century.

Ordinarily, that would have been expected to lead to higher prices via rising wages. The fact inflation has failed to accelerate as rapidly as might have been anticipated has been puzzling and led some to question the validity of a central tenet of modern economic theory. In doing so, it potentially poses a major dilemma for policymakers at the Federal Reserve (Fed).

In 1958, New Zealand economist William Phillips published a seminal piece of research in which he identified a historically stable inverse relationship in the UK economy between rates of unemployment and wage inflation that had held for a century. Two years later, US economists Paul Samuelson and Robert Solow went one step further, establishing a link between unemployment and inflation in the United States. 

It wasn’t long before governments and policymakers around the world were putting the theory into practice by adopting activist ‘Keynesian’ policies. This saw them attempting to stimulate activity when the economy was in danger of falling into recession and doing the reverse when it looked like overheating.

The efficacy of these policies became the subject of heated debate in the late 1970s as the high inflation experienced throughout the West led some to suggest the relationship had begun to break down. Nevertheless, the Phillips curve has continued to form the bedrock of Keynesian models that underpin the conduct of monetary policy throughout the developed world. As such, understanding the relationship between employment and inflation and how it may have changed over time is crucial to predicting the future path of monetary policy. 

Measurement matters

While the debate about the merits of the Phillips curve has never really gone away, it has heated up during the past two years due to the apparent failure of tumbling levels of US unemployment to rekindle meaningful growth in wages.

To accurately assess what is going on, it is essential to choose the most appropriate measures of unemployment, wages and inflation. That is far from straightforward. For example, when considering unemployment, the bulk of evidence, not least the fact the headline rate is close to a 50-year low, suggests the US labour market is now extremely tight. At the same time, there is an argument a broader measure of labour market slack – which accounts for some proportion of those inactive, but likely to search for work – provides a better gauge than just the unemployment rate.

The Federal Reserve Bank of Richmond produces a ‘weighted non-employment’ index that measures people out of the labour force as well as those who are officially unemployed. It currently stands at 7.9 per cent, compared to a historical low of 7.6 per cent.¹ Moreover, according to the Bureau of Labor Statistics, the participation rate among prime-age males (aged 25-34) has failed to recover since the global financial crisis (GFC), and at 88.7 per cent remains over three percentage points below the level recorded in 2007.

In truth, the headline rate itself matters less than where it is relative to the non-accelerating inflation rate of unemployment (NAIRU) – also known as the ‘natural’ or ‘equilibrium’ unemployment rate. Even in a healthy economy there will always be some unemployment, since a number of workers will be in-between jobs; there will be a mis-match between workers’ skills and the needs of employers; and because of the impact of factors such as the minimum wage and trade unions. According to the theory, so long as the headline rate is above the NAIRU – in other words there is a positive ‘unemployment gap’ – wages are unlikely to pick up.

The Fed reckons NAIRU, having declined steadily during the 1980s and 1990s due to the changing structure of the US labour market, has fallen further in recent years. It currently estimates the rate to be around 4.6 per cent, whereas as recently as 2015 it was estimated to be 5.1 per cent. With the unemployment rate at 4.1 per cent, that implies an unemployment gap of -0.5 per cent. So while the labour market may be tight, it has not reached the extremes of previous cycles.

Wages under pressure

As for wages, the commonly cited measure is average hourly earnings. However, this is not necessarily the best estimate of the marginal wage (which should be the driver of inflationary pressures), as it is impacted by changes in the composition of the labour market. For example, recent work by the Federal Reserve Bank of San Francisco showed that in the current economic cycle two factors have depressed average wage growth. First, a disproportionate number of lower-skilled, lower-wage workers have returned to the workforce, while other low-skilled workers have moved from part-time to full-time employment. The second factor has been the comparatively large number of higher-skilled, higher-wage baby boomers who have retired.²

Figure 1 shows a representation of the wage Phillips curve for the US, plotting the Federal Reserve Bank of Atlanta’s measure of median wages – which attempts to allow for these compositional effects – against the unemployment gap. It examines the relationship over four distinct terms. We choose to consider these periods separately as each one has unique characteristics due to changes in the structure of the labour market, a long-running decline in inflation expectations, and changing productivity trends. One can see the inverse relationship first identified by Phillips remains intact in each of these periods, including the most recent.³

Indeed, the slope of the curve is somewhat steeper in the most recent period than in the years that preceded the GFC – often referred to as the ‘Great Moderation’.

The slope of the curve is important in determining the extent to which cyclical pressures affect wages and inflation, and ultimately how monetary policy makers adjust interest rates to smooth the cycle. However, one can also see that the level of wage growth consistent with no unemployment gap has shifted down steadily over time.

Why might that be the case? The main explanation seems to be the trend rate of productivity growth has slowed, particularly since the GFC. In equilibrium, one would expect real wage growth to be in line with trend labour productivity growth. While cyclical deviations around that trend would be expected, it should provide a guide to the ‘normal’ rate of wage inflation. 

In the 1990s and early 2000s labour productivity growth was generally between two and 2.5 per cent, whereas in the past five years it has been below one per cent. Since basic economics suggests workers’ real hourly compensation should grow in line with GDP per hour worked over the long run, this decline in trend productivity implies a lower level of nominal wage growth for a given unemployment rate.

Figure 2, which shows the difference between real wage growth and labour productivity growth, illustrates wage growth is actually not weak at all when allowing for the low level of productivity in recent years.⁴ The shaded areas highlight periods when the labour market was tight, when you would expect wage pressures to be greater. The early 1990s and 2000s saw real wage growth rise as the labour market tightened, although it came later in the cycle in the mid-2000s. In the recent period, real wages also rose above productivity growth as the labour market tightened. There has been some moderation in growth in 2017, although the excess over productivity growth remains significant. 

The Phillips curve: not dead yet

So if the linkage between unemployment and wages still holds, what about that with inflation? As with the wage Phillips curve, it is important to consider which measure of inflation to use. In a completely closed economy, with no external trade, the Phillips curve relationship should hold for the broad consumer basket. However, as all modern economies engage in foreign trade, the relationship needs to be adjusted for the role of the exchange rate and the terms of trade (the price of exports relative to imports). Alternatively, one can look at domestically-generated inflation – the part not impacted by trade. One proxy, which we use in this analysis, is service sector inflation, since there is only a limited import component. A number of things stand out from the analysis represented in figure 3, which shows the relationship between the unemployment gap and service inflation.

First, in all but one of the periods considered, there is an inverse relationship between labour market slack and inflation. That suggests the relationship holds, albeit it has been less robust than the one between unemployment and wages. This is most likely due to a decline in inflation expectations. The 1990s and early 2000s were a period of falling inflation expectations as the US continued to pursue disinflationary policies begun in the 1980s to align inflation expectations with the objective of price stability. The structural decline in inflation expectations dominated cyclical developments, which only reasserted themselves once inflation expectations became anchored in the late 1990s.

Secondly, the slope of the price Phillips curve seems to be flatter in the most recent period. In other words, larger movements in unemployment have been necessary to produce the same response in inflation. Indeed, inflation seems to be only around half as sensitive to changes in the unemployment rate as it was in earlier periods.

There are a variety of possible explanations for this. One is that even when looking at a proxy for domestically-generated inflation, there may still be a restraining effect from the globalisation of product and labour markets. Another explanation may be that technological change and rising competition is making it more difficult for businesses to raise prices. Some have dubbed this the ‘Amazon effect’. Neither of these explanations is particularly compelling. The first should also be evident in the wage Phillips curve, while the latter should be reflected in a structural decline in margins, particularly in the retail sector. But there is little evidence of the latter, with margins higher in recent years than before the GFC.

A more likely explanation seems to be that some large elements of the inflation basket have fallen recently due to factors unrelated to the state of the economy. Most importantly, the introduction of the Affordable Care Act in March 2010 led to a steady drop in healthcare inflation. Then, in early 2017, there was a sharp drop in tariffs on data packages for mobile phones that will not be repeated.

In conclusion, it is understandable some should argue the relationship between inflation and unemployment has broken down since using the standard wage and labour slack variables suggests as much.

However, central to the debate is which measure of prices and labour market slack to use. On closer inspection, much of the apparent weakening in the relationship can be explained by mitigating factors. As such, there are strong grounds for believing the relationship first identified by William Phillips remains intact, and is merely lurking beneath the surface. If that is correct, and with the US labour market on balance almost certainly tight, the Fed would be well advised to continue raising interest rates expeditiously to ward off the threat of inflation further down the line.

References

¹ Federal Reserve Bank of San Francisco, March 2016

² What’s up with wage growth?, Federal Reserve Bank of San Francisco, March 2016

³ The analysis excludes the period immediately after the GFC, although the results are not greatly affected by its exclusion.

⁴ The measure of real wages used here is the Federal Reserve Bank of Atlanta’s median wage adjusted for the implicit price deflator for business output.