The real problem with the Phillips curve

Conservatives love to bash Phillips curve thinking.  They are right that the model is flawed, but they are criticizing it for the wrong reason.  It is a model that works under extremely limited conditions:

1.  Stable inflation expectations.

2.  Demand shocks are much bigger than supply shocks

3.  The government doesn’t intervene much in the labor market

Thus it does reasonably well in a large diversified capitalist economy with its currency pegged to gold.  Gold keeps expected inflation stable, the large diversified economy means most supply shocks are of minor importance, and non-intervention in the labor market keeps the natural rate of unemployment fairly stable.

In the fiat money world it works best in Hong Kong, which has had a stable expected rate of inflation since about 1984, because they are pegged to the dollar of the inflation-targeting US.  A t the same time their actual rate of inflation is quite volatile (unlike the US), because Hong Kong’s real exchange rate is unstable (due to factors such as the East Asian crisis and the Balassa-Samuelson effect.)  They also have relatively little labor market intervention.

The real problem with the Phillips curve is very different.  It has led many economists to think that inflation “normally” falls when unemployment is high.  But that’s not quite right.  There is nothing in economic theory that says inflation should fall when unemployment is high, it depends entirely on whether the cause of unemployment is a decline in AS or AD.  Remember my maxim “never reason from a price change.”

Only AD shocks create the typical Phillips curve correlation.  But that means only bad economic policy creates the Phillips Curve pattern.  With stable growth of NGDP you get a reverse Phillips curve (upward-sloping.)

Because people tended to think it was “normal” for inflation to fall sharply in 2009 as unemployment soared, they didn’t realize that this pattern showed gross negligence by the Fed.  Instead of conservatives bashing the Phillips curve model, they should have been saying “the fact that we are observing the typical Phillips curve pattern suggests that we need massive monetary stimulus.”  But of course that’s not what they were saying.

PS.  If anyone wants to construct a HK Phillips curve graph using annual inflation and unemployment data (1985-2010) for HK, send it to me and I’ll add it to the post, with your name attached (unless you don’t want it attached.)  Also let me know the correlation coefficient.  It should be fairly high (around 80%?) if you’ve done it right.

Update:  Integral provided just what I asked for:



8 Responses to “The real problem with the Phillips curve”

  1. Gravatar of Integral Integral
    6. July 2011 at 17:56

    Prof. Sumner,

    Ask and you shall receive. The graphic is here:

    (credit “Integral” if you please)

    Source: “Hong Kong Census and Statistical Department” website, Tables 006 “Labor Force, Unemployment and Underemployment” and 052 “Consumer Price Indicies.”

    Inflation is year-on-year percentage change in CPI. Unemployment is yearly average.

  2. Gravatar of Nick Rowe Nick Rowe
    6. July 2011 at 18:10

    Yep. If you can estimate a Phillips Curve, it shows that monetary policy failed. HK monetary policy is set in Washington DC, so naturally it does fail — for HK.

    Published a paper once, with econometrician Gabriel Rodriguez. “Why US money does not cause US output, but does cause Hong Kong output”. The title says it all, if you understand the basic idea of Granger causality.

  3. Gravatar of Scott Sumner Scott Sumner
    6. July 2011 at 18:59

    Integral, Thank you, that is very helpful.

    Nick, Yes, I got this idea from my students, who often do papers on the PC. I noticed it fit HK very well, and then started to ask myself why.

    Looks like you were way ahead of me.

  4. Gravatar of Joe Calhoun Joe Calhoun
    7. July 2011 at 07:46

    I could be wrong about this but I believe the original Phillips research only looked at wage inflation. What it showed was that when unemployment was high, real wage inflation was low and when unemployment was low, real wage inflation was higher(er). Duh.

  5. Gravatar of dirk dirk
    7. July 2011 at 12:32

    This was a helpful post. I had seen some conservative posts attacking the Phillips Curve recently and was wondering what your response might be. Thanks.

  6. Gravatar of The real problem with the Phillips curve « Economics Info The real problem with the Phillips curve « Economics Info
    8. July 2011 at 05:01

    […] Source […]

  7. Gravatar of Scott Sumner Scott Sumner
    8. July 2011 at 17:23

    Joe, It looked at nominal wages not real. And arguably wage inflation is the right variable.

    Thanks Dirk.

  8. Gravatar of Head Scratching Inflation – GaiaMoney Head Scratching Inflation – GaiaMoney
    21. October 2019 at 07:02

    […] love to bash Phillips curve thinking “, writes  Scott Sumner.  “They are right that the model is flawed, but they are criticizing it for the wrong reason. […]

Leave a Reply