Monday, September 11, 2006

What's the best inflation measure? Depends on who's asking.

I'm back from a brief hiatus. Our office has moved, and it hasn't left much time for blogging. Now everything is in place, and its back to work.

There is considerable debate on the blogosphere about what's the best inflation measure. There is CPI, Core CPI, Median CPI, Trimmed Mean CPI, as well as each of these in PCE form.

There are many who argue that core inflation understates inflation. If Joe Consumer has to pay more for gas, then his cost of living has clearly gone up. I think that's right. So if you are looking for a measure of cost of living increases, headline CPI is your number. That is a perfectly legitimate way of looking at inflation.

But if you are concerned with monetary policy, then better to take a monetarist's view on inflation. So if we assume...

P = M / Q

Where P is the price level, M is the money supply, and Q is the quantity of goods.

... then inflation (or change in the price level) is a function of the change in M and Q.

Economists try to measure M and Q directly, but for various reasons most economists agree that we can't get accurate enough results to calculate P from these figures.

So we have to try to measure P by observing market prices. But by looking at my equation, we see that if a change in P is caused by a change in M, all prices would be impacted similarly. Unfortunately, when we look at real market prices, we don't see them all moving in the same way. That's because each good is subject to its own supply and demand factors in addition to a universal money supply factor. If we are concerned with whether we have the right supply of money, then what we really want to do is isolate the money factor from the good-specific factors.

Here is where this logic train is heading. Obviously if you are trying to isolate the money supply factor, then taking a straight average of the prices of all goods doesn't make much sense. The average would be influenced by outlier prices, and we know the outliers must be the ones most impacted by good-specific factors. Take gasoline prices. Gas prices have risen in recent periods due to increased cost of inputs (oil) and supply constraints (e.g., refinery capacity shortages). This doesn't have much to do with a rising money supply.

Core CPI/PCE attempts to isolate this factor by ignoring food and energy, which tend to be quite volatile. That's a start, but it could well be that any number of prices are unusually influenced by good-specific factors.

So this leads us to the trimmed mean measure. This is where the most volatile PCE components are eliminated each month, regardless of what they might be, and an average is taken from the remaining. I think that's as close as we're going to get to isolating the money effect without going through some serious data massaging. Currently, the FRB of Dallas is keeping the measure. They've recently published a Power Point presentation on the subject, available here.

As you can see on page 21 of the presentation, the trimmed mean has been stubbornly high, suggesting to me that the Fed is more likely to at best hold rates where they are. Maybe the bond market is coming around to this way of thinking. The 10-year has fallen 3/4 of a point in the last 6 trading days.

Technorati Tags:

, ,

No comments: