Suppose you had just an hour to teach the basics of macroeconomics, what relationship would you be sure to include? My answer would be the Phillips curve. With the Phillips curve you can go a long way to understanding what monetary policy is all about.
My faith in the Phillips curve comes from simple but highly plausible ideas. In a boom, demand is strong relative to the economy’s capacity to produce, so prices and wages tend to rise faster than in an economic downturn. However workers do not normally suffer from money illusion: in a boom they want higher real wages to go with increasing labour supply. Equally firms are interested in profit margins, so if costs rise, so will prices. As firms do not change prices every day, they will think about future as well as current costs. That means that inflation depends on expected inflation as well as some indicator of excess demand, like unemployment.
Microfoundations confirm this logic, but add a crucial point that is not immediately obvious. Inflation today will depend on expectations about inflation in the future, not expectations about current inflation. That is the major contribution of New Keynesian theory to macroeconomics.
This combination of simple and formal theory would be of little interest if it was inconsistent with the data. A few do periodically claim just this: that it is very hard to find a Phillips curve in the data. (For example here is Stephen Williamson talking about Europe - but see also this from László Andor claiming just the opposite - and this from Chris Dillow on the UK.) If this was true, it would mean that monetary policymakers the world over were using the wrong framework in taking their decisions.
So is it true? The problem is that we do not have good data series going back very far on inflation expectations. Results from estimating econometric equations can therefore vary a lot depending how this crucial variable is treated. What I want to do here is just look at the raw data on inflation and unemployment for the US, and see whether it is really true that it is hard to find a Phillips curve.
The first chart plots consumer price inflation (y axis) against unemployment (x axis), where a line joins one year to the next. We start down the bottom right in 1961, when inflation was about 1% and unemployment 6.7%. Over the next few years we get the kind of pattern Phillips originally observed: unemployment falls and inflation rises.
The problem is that with inflation rising to 5.5% in 1969, it made sense for agents to raise their expectations about inflation. (In fact they almost surely started doing this before 1969, which may give the line from 1961 to 1969 its curvature. For given expectations, the line might be quite flat, a point I will come back to later.) So when unemployment started rising again, inflation didn’t go back to 1%, because expected inflation had risen. The pattern we get are called Phillips curve loops: falling unemployment over time is clearly associated with rising inflation, but this short run pattern is overlaid on a trend rise in inflation because inflation expectations are rising. Of course the other thing going on here is that we had two oil price hikes in 1974 and 1979. The chart finishes in 1980.
Most economists agree that things changed in 1980, as Volker used monetary policy aggressively to get inflation down. The next chart plots inflation and unemployment from 1980 to 2000.
Inflation came down from 13.5% in 1980 to 3.2% in 1983 partly because unemployment was high, but also because inflation expectations fell rapidly. (We do have survey evidence showing this happening.) The remaining period is dominated by a large fall in unemployment. So why didn’t this fall in unemployment push inflation back up? In terms of the chart, why isn’t the 2000 point much higher? Again expectations are confusing things. One survey has inflation expectations at around 5% in 1983, falling towards 3% at the end of 1999. So inflation was being held back for that reason. A Phillips curve, and its loops, is still there, but pretty flat.
The final chart goes from 2000 to 2013. Note that the inflation axis has changed - it now peaks at 4.5% rather than 16%. The interesting point, which Paul Krugman and others have noted, is that this looks much more like Phillips’s original observation: a simple negative relationship between inflation and unemployment. This could happen if expectations had become much more anchored as a result of credible inflation targeting, and survey data on expectations do suggest this has happened to some extent. There are also important changes in commodity prices happening here too.
While the change in inflation scale allows us to see this more clearly, it hides an important point. Once again the Phillips curve is pretty flat. We go from 4% to 10% unemployment, but inflation changes by at most 4%. However from the previous discussion we can see that this is not necessarily a new phenomenon, once we allow for changing inflation expectations.
Is it this data which makes me believe in the Phillips curve? To be honest, no. Instead it is the basic theory that I discussed at the beginning of this post. It may also be because I’m old enough to remember the 1970s when there were still economists around who denied that lower unemployment would lead to higher inflation, or who thought that the influence of expectations on inflation was weak, or who thought any relationship could be negated by direct controls on wages and prices, with disastrous results. But given how ‘noisy’ macro data normally is, I find the data I have shown here pretty consistent with my beliefs.