NEW YORK (CNN/Money) - Well, we can all breathe easy. With weekly jobless claims below 400,000 again, the employment picture is surely improving.

Thursday's report that the number of workers filing for unemployment benefits came in at 386,000 was the fourth week in five where claims have been below the magic 400,000 level. That's good news as far as investors are concerned, because it suggests that there will be no further softening in the labor market -- something that could hamper economic recovery.

The only problem is that nobody seems to be sure why, exactly, 400,000 should be considered such an all-important level.

"400,000 is perceived as representing the dividing point between contraction and stability in the labor market, but really it's an arbitrary line in the sand," said Morgan Stanley economist Bill Sullivan. "It's not unlike big, round numbers in the Dow Jones -- it doesn't have any significance the fact that it captures your attention."

At first glance, Sullivan's statement appears to fly in the face of history. Look through the numbers over the last 30 years and you'll see that weekly claims over 400,000 have signaled a weakening job market while claims under 400,000 have signaled strength.

What's key? Adjust for the growth in the labor force, and the 400,000 level on claims would be much higher now. Year  "Key" level  1974  400,000  1984  490,000  1994  570,000  2003  630,000

But dig a little deeper and you find that the fact that this "indicator" has worked doesn't really make sense. Its accuracy appears to be due more to chance than anything else.

Back in 1974, for instance, claims jumped from less than 300,000 at the beginning of the year up to over 500,000 by year's end -- a spike that was coincident with a move in the unemployment rate from 5.1 percent to 7.2 percent. The indicator worked like a charm, right? But know also that the U.S. labor pool in 1973 was around 92 million workers. Now it's about 145 million.

A bit of simple math (145 million/92 million, times 400,000) tells us that for the claims indicator to really be consistent, the key level should really be about 630,000 now. That's way above the peak level of 459,000 claims hit back in April. Yet, somehow, the labor market has been less than stellar. So what gives?

The answer, according to Credit Suisse First Boston bond strategist Mike Cloherty, probably lies in the way that the labor force has changed. Back in 1974, manufacturing accounted for 24 percent of the nonfarm workers on U.S. labor rolls. Today, it accounts for just 11 percent.

Why does that matter? Turns out there's a big skew toward manufacturing workers in the weekly claims data -- not because they have a higher tendency to be unemployed, but because they have a higher tendency to know what to do when they're unemployed. An hourly employee who misses one shift due to, say, a blackout shutting down his work can file for unemployment. Most people don't know to do this -- or even if they do, bother to fill out the forms -- but the manufacturing workforce, which is heavily unionized, does.

"Heavily unionized industries have a much higher filing rate because the union does all the paperwork for you and lets you know how the system works," said Cloherty.

The manufacturing bias in the claims numbers can make them extremely choppy, even with the Labor Department doing its best smooth them with seasonal adjustments. July and August are particularly dodgy, because there are a lot of plant shutdowns. "This is a number you're not supposed to pay attention to in the summer," said Cloherty.

Power outages that shroud the entire Northeast in darkness can also be a problem -- last week's blackout didn't factor in Thursday's claims report, but it will heavily distort the numbers in the weeks to come.

Maybe this will get people to stop focusing on the key 400,000 level and set them to work on figuring out where the U.S. labor market is headed. At least for a little while.

"I don't know where 400,000 came from," said Northern Trust chief U.S. economist Paul Kasriel. "I guess sometime in the last 20 years somebody decided that 400,000 was important. In the economics profession -- and I use that term advisedly -- these things take on lives of their own."