Economists have recently been abandoning once-core concepts in macroeconomics such as the money multiplier and IS-LM model. This week's figures remind me of something I'd also throw on the bonfire - the Phillips curve.

My chart shows the point. It shows that the relationship between the official measure of unemployment and CPI inflation four quarters later has been perverse. Higher unemployment has led to higher inflation, not lower as the Phillips curve says.

If we control for inflation expectations, as surveyed by the Bank of England, this perverse relationship weakens - but there remains a statistically significant positive relationship between lagged unemployment and inflation.

One reason for this might be that the official measure of unemployment is inadequate. It excludes the 2.3 million of "economically inactive" people who'd like a job and the 1.3 million of part-timers who want a full-time job.

If we take this wide measure and control for inflation expectations, we do get a negative relationship between unemployment and CPI inflation. But it is not statistically significant (p value = 25.7% for quarterly data since 1999Q4). Nor is it economically so; a one million drop in the wide measure of joblessness is associated with only a 0.16 percentage point rise in CPI inflation.

I'd suggest three reasons for this weak relationship. It's because the tendency for a tight labour market to raise wage and price inflation is offset by three other mechanisms:

One is globalization. In a small open economy, inflation depends upon inflation overseas and the exchange rate. Euro area deflation and the strong pound in recent months have offset falling unemployment. As one of the better macro textbooks says:

In an open economy, there is a range of unemployment rates consistent with the absence of inflationary pressure. (Carlin and Soskice, p343)

Secondly, fluctuations in aggregate demand aren't the only story. Supply shocks such as commodity price moves or changes in productivity growth can generate positive correlations between unemployment and inflation.

Thirdly, inflation isn't necessarily cyclical. In a classic paper in 1986, Julio Rotemberg and Garth Saloner pointed out that price wars were more likely (pdf) in booms than slumps. The fact that supermarkets are cutting prices now is consistent with this.

Now, all this said, I'll concede that the conventional Phillips curve might exist outside the UK - though even in the US it is rather flat. For practical purposes, however, unemployment tells us very little about future inflation.