The post-graduate plans of business school graduates is one of our favorite contrary indicators. Each year in November, when Harvard Business School releases statisics about where its graduates have landed jobs, we like to take a look to see what they tell us about the market.

Ray Soifer, a retired executive from Brown Brothers Harriman and a Harvard Business School graduate, designed the famous metric for reading the HBS stats. When 10% or less of a graduating class take Wall Street jobs, it's a long-term buy signal. When 30% or more take market sensitive Wall Street jobs, it's a big flashing sell signal. The idea is that when too many MBAs are flooding into Wall Street firms, the market is probably overheated. Over the years, the HBS metric has proven itself to be a splendid long-term indicator of the American equities market.

This year, for the third year in a row, the HBS stats are sending a loud and clear sell signal. This year, some 41% of Harvard Business School's graduate found work on Wall Street. That's actually a year over year increase, up from 40% last year, 37% the year before that, 30% in 2005 and 26% for the class of 2004.

Meanwhile, following the trend of recent years, the average number of months of work experience of HBS grads slipped again, all the way down to 48 months from 50 last year and 52 the year prior.

Interestingly, we can report that the percentage of HBS grads headed to private equity or leveraged buyout firms climbed this year, going up to 17%. The number had been declining a bit for the past couple of years.  This probably reflects the decline of traditional investment banking in favor of alternatives. Only 9 percent of the HBS grads went to work for investment banks.

DealBook points out that the HBS grads were making decisions about their employment months before the turmoil of September and October.

The index is based on decisions made many months ago, something that Mr. Soifer readily acknowledges.

“Most of these career decisions were made before the market’s recent steep decline — some as early as the fall of 2007,” he wrote. “Next year’s data should be interesting!”