Wednesday, 21 October 2009

Some Thoughts on the Phillips Curve

Of all the economists, journalists and assorted financial industry participants who comment on the web – and there’s certainly no shortage of them – the one whose views align most closely with my own is James Hamilton of UC San Diego and Econbrowser. I find that I rarely disagree with him, whether it’s on macroeconomics, oil, securitization, financial markets, or anything else. So I was interested to see him make the case that ‘high levels of unemployment are a factor that will put downward pressure over the next two years’.

His argument is straightforward: he regresses historically realized inflation against unemployment, and also against lagged inflation (the latter is to account for expectations of inflation). He finds a statistically significant negative coefficient for the period from 1948 to 2007, validating the classical Phillips Curve relationship. Since unemployment is currently very high, inflation is (ceteris paribus) likely to be contained over the next few years.

My quibble with this analysis is that I just don’t think historical data, aggregated like this, is a very useful guide to the future. For instance, the period from 1948 to 2007 had several very distinct macroeconomic regimes. Let’s go decade by decade:

1940s: World War II has massive effects on consumption, production, resource allocation and labor markets, effects which do not end with the end of the war. The Fed buys long bonds to finance government spending, but reduces the money supply to pre-empt post-war inflation of the kind seen in 1919-20.

1950s: A boom in capital goods and in consumer credit (think: demobilization and reconstruction) leads to a sustained period of strong employment and benign inflation; the misery index will never be this low again. But a balance of payments deficit hints at trouble to come.

1960s: Operation Twist. The Great Society. Tax cuts combined with spending increases bring about the first persistent, large budget deficits. Congress eliminates the gold reserve requirement for the Fed. Inflation rises. The draft reduces unemployment.

1970s: The US leaves the gold standard in 1971. Nixon imposes wage and price controls from 1971 to 1974. OPEC embargos oil. Arthur Burns argues that it’s okay to countenance ‘temporarily’ higher inflation if that can alleviate economic shocks. He is succeeded by William Miller who thinks Burns kept policy too tight (!).

1980s: Paul Volcker breaks the back of inflation. The Reagan recessions are followed by the Plaza Accord and the devaluation of the dollar.

1990s: China. See my previous post for details.

I think it’s fairly clear that at different times over the last 60 years, very different factors have driven inflation and employment in the US. Certain drivers remain important to this day, but others have changed utterly. (For instance: every decade has seen a different role for the Federal Reserve, from financing government debt in the 1940s, to twisting the yield curve in the 1960s, to fostering stagflation in the 1970s, to satisfying the bond market vigilantes in the 1980s). If causation has changed so much, how can we expect correlations from the past to apply today?

In fairness, I don’t think Prof. Hamilton himself fully believes the case he presents. He merely makes the observation that a forecast for low inflation going forward is not ‘crazy’, while pointing out that other factors (the dollar, commodities) may pull the other way. I can’t disagree with that ... so I guess I’m back to agreeing with him on everything.

Postscript: I’m embarrassed to admit that I only found out recently that the James Hamilton who co-writes Econbrowser is the same James Hamilton who wrote Time Series Analysis, a text which I used a fair bit in my professional career.

No comments:

Post a Comment