"The stock market quaked Thursday at new data showing labor costs soaring at the fastest pace in eight years, which fanned fear that the Federal Reserve will raise interest rates again soon to pre-empt inflation."
Merury News 7/30/99
"PRICES RISE BY A BIG, FAT ZERO...ECONOMISTS ARE SURPRISED"
Mercury News, 7/16/99
"The statistics show the American job machine chugs along with
little evidence of inflation, but also that the labor market is so
tight inflation might erupt at any time.... The Fed is particularly
worried that wage inflation may erupt, igniting a general
inflationary trend that could burn the entire economy, Chairman Alan
Greenspan has made clear. "
Mercury News 6/5/99
"NO. 2 AT FED WILL STEP DOWN ... SHE BELIEVED IN DRIVING DOWN
UNEMPLOYMENT, DESPITE RISK OF INFLATION. "
Mercury News 6/4/99
As exemplified in the quotes above taken just from the last few months, the business news has been filled for several years with remarks of concern, from various "economists" and government officials up to Fed Chairman Alan Greenspan, about the persistence of low unemployment in the United States. You would think that the Federal Government doesn't want people to have jobs. If you were to ask these officials, they'd probably piously inform you that that isn't it at all: no, no, there's a "law of economics" that if unemployment gets to low, inflation will somehow suddenly cut loose like a wild beast. They'd talk knowingly about the "Phillips Curve" and look at you with condescension if you appear puzzled. They'd have all the stigmata of a knowing elite looking after your interests with greater wisdom than you can muster. But is it true?
To understand whether a model is correct or not, a scientist looks at the data: more specifically, at whether the model can correctly predict the behavior of some aspect of the real system. The model we want to test is the idea that when unemployment goes down, inflation must go up: the model is called the "Phillips curve" after the author of a paper published in 1958 in the British journal Economica, in which Mr. Phillips analyzed some data for Britain's economy, on the basis of which he proposed this relation. A picture of the idea is shown below as a "plot": that is, a relationship between unemployment and inflation, depicted by showing a line on which the measured data is supposed to lie. Each point on the line is identified with a particular value of unemployment (small values on the left, large ones on the right, as shown by the numbers on the bottom "axis"), and a particular value of inflation (small at the bottom, large at the top, as shown by the numbers on the left side "axis").
The model says that if we choose a point corresponding to the unemployment and inflation for any given year, it will lie on the dark line. Thus point "B" depicts some year in which unemployment is about 7% and inflation about 2%. Unemployment can only go down if inflation goes up: thus, to reduce unemployment to 4% (point "A") we have to increase inflation to 6%.
To test the model, we need to put in the measured data: the real inflation and unemployment levels for the United States, or for other countries, for various years. For each year, we locate the spot on the plot with the corresponding values of unemployment and inflation, and put a mark (a circle, for instance) there. If the model is right, when we plot a point for each year, they would all fall on or close to the line. Naturally, even if the model is right, we wouldn't expect the points to be exactly on the line: errors in the measurements, unusually good or bad weather, any number of things might contribute some noise. However, if the model is really a "law of economics" it should show up very clearly despite a bit of noise: we might expect to find a graph like the example shown below, in which we have made up some data to show what a good model fit looks like:
(Scientists and statisticians have invented all sorts of ways to decide just how close the fit between model and measured data is -- "correlation coefficient", "mean squared error", "F statistics" and "t statistics" and so on -- but don't let that worry you: the human eye and brain are still the best tools for deciding whether models agree with data, which is why most scientific papers are full of graphs and plots like these. If it looks right, it is right.)
Now we're ready to look at the real thing: in the next plot we show the unemployment and inflation data for the United States of America, for the period 1961 to 1997, again showing for comparison the "Phillips curve" model:
Unlike our made-up data, the real data shows no particular tendency to be on the line or even close to it. Moving the line up or down wouldn't help: there's just no agreement between the data and any model that shows inflation going up as unemployment goes down. Any scientist at this point would conclude: the model is wrong. There's no evidence in the behavior of the U.S. economy for the last 37 years for any economic law linking low unemployment with high inflation. In fact, if there's any trend at all it is the other way around: high unemployment is correlated with high inflation.
Now, you can go deeper into the data and find that over certain periods of time a sort of Phillips curve is right (economists like to call this the "short run Phillips curve")-- but you can choose other segments of time when it's equally wrong. Believing that somehow the data that agrees with your theory is valid and the data that doesn't is erroneous is convenient in politics or a courtroom -- but it isn't science. What are you going to believe: sophistry or your own eyes?
So next time Alan Greenspan lectures Congress on the terrible danger impending because too many of us have jobs, you'll know that you're hearing an idealogy unrelated to the real world. Perhaps with time we can convince our friends, our representatives and -- who knows? -- even economists that what matters is not how many of us are working, but whether what we're working on actually adds enough value to the world to pay our salaries. And remember: no matter how many fancy degrees, elegant arguments, or important goverment officials come attached to a theory, the proof is in the data.
The example of the Phillips curve, and the data for the US economy, shown above come from Gregory Mankiw's Principles of Economics, chapter 33, supplemented by World Economic and Social Survey 1998 put out by the United Nations. A more thorough discussion of some of the blatant fallacies of contemporary economics can be found in Paul Ormerod's book, The Death of Economics.