Barrons

## Of Reason and Reasons

### Economists clueless about ways of market

By GENE EPSTEIN

The trouble with economists, says economist Richard Thaler, is that they assume the world is as rational as they are, while he believes it can often be as naive as he is. This penchant for seeing reason where none exists can get our dismal scientists into all manner of trouble, which Thaler has long delighted in exposing (see his 1992 book, The Winner's Curse). But his specialty is poking fun at their unique difficulty in understanding the markets.

It's surely no coincidence that Street economists are allowed to predict gross domestic product and the consumer price index, but when it comes to more important variables like the Dow Jones Industrial Average, they're kept out of harm's way. That's because, as Thaler puts it, the trick of forecasting prices in any given year is to be one step ahead of the market; two or three steps and you've only outsmarted yourself. Back in the early 'Eighties, a lot of economists believed the dollar was overvalued, and they were right. It was overvalued in terms of trade balances. But they were too many steps ahead of the market, which kept pushing the dollar higher and higher anyway.

John Hancock Mutual Life Insurance chief economist Bill Cheney cheerfully admits that he was recently confounded by a little numbers game that Thaler has long been using to illustrate his point. At the Program on Investment Decisions and Behavioral Finance held this past November at Harvard, Cheney and others were asked by Thaler to submit a number between zero and 100. From those figures, Thaler would then compute a simple average and the winner would be the person whose number fell closest to two-thirds of that average.

Cheney's laser-like mind immediately went to work. It first occurred to him that the winning number had to be less than 67 because even if everyone guessed 100, then the average would be 100, and two-thirds of 100 was 67. Then he realized that if everyone else realized that, then the average would be less than 67, and two-thirds of 67 was 44... . Working through the logic of this infinite regress, he submitted the number zero.

Once all the figures were gathered and the average computed, Thaler announced that a couple of the submissions were in the high 50s, which meant that these people were asleep if not brain dead. Too many steps behind the market.

He then noted that a lot of bets clustered around 33. Those people were awake, he remarked, but had assumed everyone else was sleeping. They had obviously assumed that the estimates would be distributed around the midrange of 50 and had taken two-thirds of that. But that didn't make any allowance for second-guessing among the other participants. At least one step behind the market. Then Thaler got to the two or three zeros. "You guys have had too much economics," he declared. "You solved it analytically when that's not what this is about."

Cheney recalls that the winning number was somewhere in the low 20s (Thaler says it usually ranges between 15 and 25). Let that be a lesson in how to be one step ahead.

Ironically, Thaler teaches at the University of Chicago, which spawned some of the most ultra-rational economic theories the world has ever known. That's where Nobel laureate Milton Friedman developed his "permanent income hypothesis," which was the wacky idea that the average person uses a present value model to determine how much he'll spend on consumption. It's also where another Nobel laureate, Gary Becker, cooked up his theory that families were like firms (not my family and probably not his).

But then again, Thaler is at Chicago's Graduate School of Business rather than in the economics department, so his chance of corrupting those impressionable Ph.D. candidates is quite limited.

Friday, the Bureau of Labor Statistics' employment report confounded Wall Street's consensus for the sixth straight month, but this time in a different direction from before. The BLS announced that after five months of strong growth, nonfarm payrolls slipped by 36,000 workers in March (against a consensus estimate of a gain of 250,000), with declines mainly in construction and in retail trade. The March unemployment rate also ticked up, to 4.7% from 4.6%.

But perhaps the most prudent reaction to the news was to pass and wait 'til next month.

To begin with, the idea that the economy shed any jobs at all in March doesn't square with the numbers on initial unemployment insurance claims, which have been bumping along at the exceedingly low level of 300,000 per week. As HSBC Securities chief economist Ian Shepherdson points out, initial-claims figures have been a pretty reliable guide to underlying trends in the labor market.

Moreover, the estimate may be plagued by faulty seasonal adjustments, which can be problematic in the first quarter, when weather can become an important factor. The report said that there was an 88,000 loss in construction jobs, but before seasonal adjustment it showed a 50,000 gain. It said that there was a 48,000 loss in retail trade, but before seasonal adjustment it showed a 40,000 increase. So -- will the true figure please stand up?

The seasonals are trying to suggest that since these gains should have been greater under "normal" conditions, it's only proper to put them in negative territory. But what's normal when you're talking about March? This year it reversed the usual routine by coming in like a lamb and going out like a lion, with cold weather through most of its stay. That had to have slowed growth in construction employment and probably in retail as well, since hiring in eating and drinking places was especially hurt.

But more than anything else, the reported decline follows on the heels of some huge gains in nonfarm payrolls. So even after factoring it into the picture, we have an average rise of 205,000 in the first quarter compared with 267,000 per month through all of 1997.

Manufacturing employment, which ought to be most vulnerable to the effects of the Asian crisis, was essentially unchanged for the second straight month, although the factory workweek continued to decline from its record highs. For the time being at least, manufacturers are cautiously hoarding their workers lest they lose them in this still-tight labor market.

Some people think the official unemployment rate is based on the number of people currently receiving unemployment insurance, which leads to all manner of suspicion that the jobless are being undercounted. Since I heard this once from a Street economist, no less (whose name I withhold out of kindness), the confusion over this point is worth clearing up.

To begin with, there are currently 2,182,000 people receiving unemployment insurance versus 6,529,000, or just about three times as many, that the Bureau of Labor Statistics says were jobless in the month of March. So there doesn't seem to be too much reason to worry that anyone who has exhausted his benefits will somehow fall off the BLS radar screen.

But actually, the Bureau pays no attention at all to the unemployment-insurance number. Through its monthly household survey, which involves knocking on doors and asking questions of a representative sample of more 50,000 households nationwide, it's able to arrive at its own estimate. According to its criteria, in order to qualify as unemployed, respondents have to state that they didn't work over the prior week and that they were available for work; and answer a battery of questions showing an effort to seek work sometime over the prior four weeks.

Now, all statistics, especially those that emanate from the government, should be regarded as guilty until proven innocent. But this particular number seems reasonably worthy of parole so long as we don't ask too much of it. For sure, it doesn't tell us a lot of things, such as what kinds of jobs respondents would be willing to accept and at what wage, whether they'd be willing to relocate and where-or, of course, whether they're telling the truth to begin with. (Although the lying bias, if there is one, is probably fairly stable over time.)

But it does provide an indication of the degree to which that all-important resource called labor is being utilized, not to mention an often useful check on how the economy is really doing. Without it, we wouldn't have known to call 1992 -- when the economy grew while unemployment rose -- the year of the jobless recovery.

Skeptics might want to question the plausibility of the 6.5 million figure precisely because it is so much higher than the 2.2 million who are receiving benefits. But it isn't very difficult to account for the difference.

The March report shows that 900,000 of the unemployed had been without work for 27 weeks or more, and so would have exhausted their benefits, or be close to exhausting them, assuming they were receiving them to begin with. Add another 2.2 million jobless who were re-entrants into the labor force and another 600,000 new entrants; neither of these groups would qualify for benefits because they weren't laid off from jobs.

Then throw in some of the 700,000 who quit voluntarily (the folks who used to cause Mr. Greenspan so much consternation), many of whom wouldn't be eligible to file. Also add some of the 600,000 who were fired for cause, not to mention others (for whom no count is available) who worked at their jobs for too short a time to qualify for benefits.

Roughly speaking, this yields four to five million unemployed souls who aren't on the rolls -- which should be more than enough to lend plausibility to the BLS figure.

Finally, the civilian labor force of 137.5 million consists of the people who have jobs (131 million) and those who are unemployed (6.5 million). Divide the number of unemployed by the civilian labor force and get the March unemployment rate of 4.7%.

E-mail: gene.epstein@news.barrons.com.

Questions:

1. Is it necessary for investors to be rational for markets to be efficient?
2. Under what conditions does irrationality among investors yield opportunities for beating the market?