Friday, October 23, 2009

Separating the real, the financial, and the sentimental 

A friend asked me a great question today, which I'll paraphrase: If economists think the recession is over but consumer sentiment continues to be pessimistic, who's right? What is most likely to happen next?

As usual, let's look at the graphs, because calling consumer sentiment comes in different flavors. First we have the Michigan survey:
At least on that measure, we see the measure is coming up off its lows earlier this year. But rallying? Hard to say. (BTW, I don't know why the gray band is cut off at August -- that would be the St. Louis Fed's judgment not mine or the NBER's if they do.)

The Conference Board series is also not good or not bad:
This measure views the 50 line as neutral, so since May the Conference Board series has had consumer confidence as "not bad", rather than good.

Let me try to tie those together looking at the latest Leading Economic Indicators (LEI) report, which posted its sixth straight gain in September. I heard Chip Hanlon mention this on Hugh's show on Monday and it got me very curious, so I pulled this data up this morning. LEI is a composite of ten indicators. Hanlon said that (as I heard it) 60% of LEI are things that the Federal Reserve controls. I thought I knew that series and it didn't sound right, but the composition of LEI is a weighted average, and I assumed Hanlon was reading the weights. I'm a geek, but not so geeky that I memorize them, so I looked them up. What I see are standardization factors.

Leading Economic Index, Factor
  1. Average weekly hours, manufacturing 0.2549
  2. Average weekly initial claims for unemployment insurance 0.0307
  3. Manufacturers' new orders, consumer goods and materials 0.0774
  4. Index of supplier deliveries � vendor performance 0.0677
  5. Manufacturers' new orders, nondefense capital goods 0.0180
  6. Building permits, new private housing units 0.0270
  7. Stock prices, 500 common stocks 0.0390
  8. Money supply, M2 0.3580
  9. Interest rate spread, 10-year Treasury bonds less federal funds 0.0991
  10. Index of consumer expectations 0.0282
You can see the most weighted value in there is movement of (real or inflation adjusted) M2. But that weighting is meant to adjust the volatility of each measure. Because M2 doesn't have that great a volatility, it gets a higher weight. You can see, it's not a highly volatile graph. M2 growth has slowed but, because CPI turned negative real M2 is rising.(see alternative measures from ShadowStats for some caution on this figure.)

And with yesterday's report we can add up the impact of the ten. Here is the September reading for contributions of each component to LEI, which grew 1%:

Average work week, production workers
Average weekly initial claims unemployment insurance
Manufacturers' new orders, consumer goods
Index of supplier deliveries (vendor performance)
Manufacturers new orders, nondefense capital goods
Building permits
Stock prices, S&P 500
Real M2 money supply
Interest rate spread (10-year T-bond minus fed funds)
Index of consumer expectations
TOTAL +1.0

I divide that in my mind as six items from the real economy, three financial measures, and a consumer sentiment item. Those first six items contributed only 0.22% of the 1.00% gain came from the real economy, 0.24% came from sentiment using the more favorable Michigan survey, and the other 0.54% came from the three financial indicators. But one of those three are the interest rate spread, which is kept large in part because of the Fed's hitting of the zero interest bound on Fed funds. (Note Allan Meltzer's discussion of profit from the carry trade this morning.) I don't know if that's what Hanlon is referring to, but it makes the argument for some artificiality of the models economists use.

I haven't done a national model forecast for a few years now (I had a model at one time for a classroom exercise but it's old and dusty) but if I did it would make me very nervous right now. The data we are currently plugging into those models are of a different character than before; forecasters using those models without much consideration of the data are extrapolating far away from their models' experiences. Assuming forecasters are not that foolish, we can only assume that they are including a great deal of judgment and autonomous decisions to their extrapolations. Some of that relies on relationships between different economic phenomena, but again that has to look different in this current world. (See the still-useful classic Mincer and Zarnowitz [1969] for thoughts on extrapolation and judgment.)

What worries me here is that the usual relationships would be hard to rely on. Much of that LEI gain is due to the unusual behavior of the interest rate spread -- how reliable is that signal? More data undoubtedly helps, and next week's GDP number on Thursday will tell us whether the financial signal we're getting now is useful.