NEW YORK (CNN/Money) - Boy, oh, boy is the stock market expensive.

Thanks to its 18.4 percent rally since the end of the first quarter, the price-to-earnings ratio on the S&P 500 has jumped to 20.3 from 17.6. Other than in the gaga years of the late 1990s, it has rarely been so high.

And that's just when you look at the profit figures supplied by First Call, which exclude many of the often-specious "one-time items" that companies take out of earnings. Look at the earnings under generally accepted accounting principals (GAAP) that companies provide in the documents they file with the Securities and Exchange Commission, and the S&P's P/E swells to 33.1. That's more than twice the 67-year average of 15.47. There are only a handful of periods when its been so high -- all of them, alas, coming since 1999.

So it's time to sell, right? Not necessarily.

"The market does have a valuation problem," said First Albany chief investment officer Hugh Johnson. "But I don't put much stock in any valuation model, including my own. It's very rational -- it tells you when the stock market is overvalued or undervalued -- but it doesn't work."

Sacrilege, of course. But history appears to back it up. Consider P/Es -- the absolute bedrock of most valuation analysis. According to research from Yale School of Management professor Matthew Spiegel, if you only invested when the S&P's P/E was below its historical mean you would have only been in the market for 19 of the past 77 years. Nor were those the "right" years to invest. At the end of 1973, for example, the P/E was 13.5, its lowest level in years. Yet the market dropped almost 28 percent.

The problem is that P/Es have risen over time, meaning that the market has tended to look overvalued relative to its history. Even if, starting in 1925, you employed a strategy where you looked just at the past 10 years and invested only when P/Es fell below their mean, you would have a compounded annual return of just 4.2 percent.

"You always see people saying the market's P/E is high relative to history so the market must be expensive," said Spiegel. "It's sort of taken as an article of faith that P/Es are good to look at."

The error people are making, thinks Spiegel, is that they are looking at P/Es with the benefit of perfect hindsight. We think the S&P's P/E of 15 at the end of 1994 was cheap because we're comparing it to what we've seen since then. But investors back then didn't know what would come next. They did know, however, that the S&P's P/E at the end of 1984 was 10.

The other problem with using P/Es based on historical earnings is that you are looking at the past and the market is supposed to price on the future. Some valuation models try to circumvent that by looking at future earnings estimates, which is nice if you have a very good idea of what the future will bring. If you are one of those people we hope that you are enjoying the tropical island which you've bought and retired to very much.

Finally, there's just the problem that the stock market is driven by people, and people aren't particularly beholden to the rules of logic.

"When you say the stock market is overvalued, you imply that the stock market is going down," said Johnson. "But there is a touch of lunacy in every bull market -- stocks can become even more overvalued. A rational investment policy in an irrational world is suicide."