| "Theory helps us to bear our ignorance of facts." |
| -- George Santayana |
EMEMBER THE "GOLDILOCKS MARKET"? All through 1998 and into 1999, the U.S. economy wasn't too hot or too cold, it was just right. But last summer, the Federal Reserve decided things had heated up a little too much and started raising interest rates in an effort to head off inflation. Ever since then, Fed fretting has been the only game on Wall Street.
As soon as the FOMC meets, attention immediately turns to what will happen at the next meeting. Between meetings, investors hang on every economic statistic. Folks who would normally be parsing though P/E and debt-to-equity ratios, are instead clamoring for the latest purchasing managers' index or equally arcane statistics. After all, these are supposedly the critical factors used by the lead bank in their deliberations.
All of this got us wondering if any of these statistics really make a difference. Presumably rapidly growing GDP and wages, as well as high levels 
of retail spending are indicative of an overheating economy. Without Fed intervention, these would lead to inflation that would be reflected by a rising CPI.
On the other hand, when unemployment falls the job market tightens and wage inflation becomes a concern. That's why bond traders always rejoice when unemployment rises -- you may be out of work but their bonds hold their value. Similarly, an increase in interest rates -- the Fed's tool of choice -- raises the cost of borrowing, slows the economy, and keeps inflation in check. Supposedly these effects would also be reflected in the CPI.
The Ties that Don't Bind
To see how strong these relations are, we turned to the economic statistics provided by Baseline and PDE Economics. Looking back over ten years of data relating various economic statistics and the CPI, we didn't find the strong correlations you'd expect -- especially if you listen the the media fretters. In fact, we found very little relation.In the event you have a life and aren't a statistician, correlation measures relationship between the movements of two variables. If they move in tandem, they are perfectly correlated and have a correlation coefficient of +1. If they always move in opposite directions, they are negatively correlated with a correlation coefficient approaching -1. If their movements bear no relation to one another, they aren't correlated at all and have a correlation coefficient near 0. The strongest relationships fall between 0.7 to 1.0 and -0.7 to -1.0. The closer the coefficient to 0, the more spurious the relationship.
The accompanying table shows the correlation coefficients for CPI and:
- Consumer Spending
- Retail Sales
- Real (Inflation Adjusted) GDP
- Unemployment
- Fed Funds
- S&P 500
- Industrial Production
As you can see from the middle column, none of the relations are really "strong". The greatest correlations are consumer spending and industrial production. Both are negative, implying that they move in the opposite direction as the CPI.
Industrial production makes some sense. If firms are producing more with the same inputs, they're more efficient. More products come to market without additional production costs, so there's little need to increase prices (and inflation) to maintain margins. In other words, the more productive companies are, the less inflationary pressure.
But that's where logic ends. As you can clearly see from the chart above, productivity has been on the rise since the beginning of 1999 -- precisely the time when the Fed started boosting interest rates. If productivity was increasing, there shouldn't be any need to do this. Go figure.
And while you're at it, explain why consumer spending and the CPI are negatively correlated. The Fed certainly doesn't think they are. In fact, this is supposed to be one of their main concerns. They reason -- and it sounds logical -- that high levels of consumer spending indicate strong demand. This allows firms to raise prices for their goods, causing a jump in inflation.
But our ten years of data don't bear this out. On the contrary, they show that consumer spending rises as inflation falls. Could it be that consumers are only comfortable spending when they believe prices are stable? Whatever the explanation, these are the facts.
One other interesting finding was the relationship between the unemployment rate and CPI. Unemployment is supposed to be a gauge of the "tightness" in the labor market, so you'd think it would be negatively correlated with inflation. In other words, when the unemployment rate rises, more people are out of work and searching for a job. It's a buyers market, labor comes cheap. The opposite would be true when unemployment is low and employers are "paying up" for qualified workers.
This inverse relationship between unemployment and inflation was a "fact" in the 60s and 70s. If you paid any attention to economics in those decades, you undoubtedly saw this relationship diagrammed in the infamous "Phillips Curve". But as we've argued before (July 1999 True Facts), this seemingly logical relationship simply doesn't hold. The Fed apparently still believes it does, but the slightly positive(+0.38) correlation between the unemployment rate and CPI says otherwise.
Prediction and Confirmation
Of course you could argue we've looked at the wrong thing. After all, a change in the unemployment or Fed Funds rate doesn't immediately affect the economy. There's a certain lead -- or in some instances lag -- between a change in economic indicators and CPI. A drop in unemployment or an increase in Fed Funds may take several months to be reflected in the CPI.With this in mind, we looked correlations with a three month lead, and three month lag. We didn't go further in either direction simply because the greater the length of time between the indicator and the CPI, the less reliable the relation.
|
Correlations with CPI | ||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||
The results are somewhat surprising. For example, consumer spending, retail sales, real GDP, S&P 500, and Industrial production seem to be better lagging indicators than predictors of the course of inflation. Each has a higher correlation with a three month lag than with a three month lead.
But when you think about it, maybe this isn't so surprising. It indicates that each of these economic series is a better gauge of where inflation has been than where it's going. In other words, rising inflation tends to slow consumer spending, retail sales, and industrial production while serving as a drag on real GDP and stock performance. Who would have thought otherwise?
What many people -- perhaps even the Fed -- may be missing is that inflation affects these series not the other way around. If consumer spending and retail sales are falling, it doesn't mean inflation will rise, it means inflation had previously risen.
The Fed Funds rate also presents an interesting relation. First of all, it's positively correlated with the CPI indicating that inflation and the Fed Fund's rate -- the Fed's weapon of choice against inflation -- move in the same direction. In addition, its correlation is stronger with a three month lead than with a three month lag. This implies when the Fed raises rates to combat inflation, inflation will be higher three months down the road.
While it may at first seem odd, this, too, makes sense. Recall there's a 9-12 month delay between the implementation of monetary policy action and its effect on the economy. If inflation is heating up to the extent the Fed finds it necessary to tighten credit, almost a year will pass before today's action will have an impact, so it stands to reason that inflation will be higher in three months. As the time to the effect draws closer, the positive relation between the Fed Funds rate and inflation grows smaller. This is reflected in the decreasing correlations from three month lead to three month lag.
But what's up with the unemployment rate/CPI correlations? First of all, as pointed out earlier, the relationship is positive. In other words, as unemployment goes up, so does inflation.
Even more surprising is the higher correlation as you move through time. According to our table, unemployment is a better indicator of inflation with a three month lag than with an equal lead. If unemployment rises, inflation probably rose three months ago. Conversely, if inflation rises, unemployment will be up in the next three months.
That flies directly in the face of the Fear-of-Full-Employment reasoning used by the Fed. If any Phillips Curve fan can explain this, please e-mail us.
The Sum of the Parts
Finally, the Fed doesn't just look at each piece of economic data in isolation. Instead, they put them together to get an idea of the overall economic condition.In an attempt to do this, we ran a multiple regression. In plain English, this is a statistical process that compares not just one, but several variables at once. In this case, it could be valuable if the economic series have greater predictive power together than separately.
Now we did take the liberty of simplifying the process by limiting the number of economic series to those with the highest individual correlation to CPI: real GDP, industrial production, and consumer spending. We then ran a stepwise regression on the past ten years of data.
R2=0.5611
Again, if you have a life and aren't a statistics fanatic, a stepwise regression seeks to determine how much of the movement in the dependent variable (CPI in this case) is explained by the movement of the independent variables (the three economic series listed above). The process starts with the independent variable with the best fit and then adds the others to determine if the explanatory power is improved. If the addition of an independent variable fails to improve the relation, it is eliminated as insignificant.
Speaking of which, real GDP turned out not to be statistically significant so was removed from the final relation. The resultant equation appears above. It is statistically meaningful at the 1% level of significance and the R2 value indicates the relationship is relatively strong. In plain English, there's a pretty good relationship here.
Monetary Art and Science
So does this vindicate all the fretting over economic statistics? Not really. It's doubtful the Fed uses many regression equations in setting monetary policy. They do, however, seem to put significant weight on the (more than dubious) Phillips Curve relation.
The one thing that does seem pretty obvious is that the relations between inflation and economic statistics aren't as straighforward as you'd think. That is, of course, if they hold at all.
Despite the appearance of objectively acting on economic facts, monetary policy involves more art than science. Investors should stick to market fundamentals and leave the monetary artistry to the Fed.
Search this site! Just enter you key word or words:
Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
|||||||||||
| Homepage | Return to Top |
