YoY change of employment is my favorite recession indicator. Floyd Norris on NYT explains his recession indicator with perfect record (like all labor markets indictors, this is a lagging indicator):

One unfortunate milestone reached today was that the 12-month change in private sector jobs, before seasonal adjustment, is down 125,000 jobs. (After seasonal adjustment, there is a small increase, for some reason.)

Over the last nine recessions, dating back to 1953, that indicator has a perfect record. Every time it has turned negative, the economy is already in recession. And once it turns negative, it stays that way for quite a while. (Caution: These numbers will be revised, and this history is based on the revised figures, not the originally reported ones. But I think the revisions, when they come, are much more likely to indicate the figure went negative early rather than later.)

In December 1953, the figure turned negative six months after the recession was later determined to have begun. It remained negative for 14 months.

In October 1957, it went negative three months after the recession began, and stayed that way for 15 months.

In December 1960, the negative jobs figure came 9 months after the recession started, and it stayed negative for 10 months.

In July 1970, it turned negative 8 months after the recession began, and stayed negative for 13 months.

In November 1974, the first negative number came a year into the recession. It stayed negative for 14 months.

In June 1980, the recession was 6 months old when the negative number arrived. It stayed negative for just 6 months.

In January 1982, the negative number came 7 months after the recession started, and it stayed negative for 17 months.

In December 1990, the first negative number came 6 months into the recession, and the figure stayed negative for 17 months.

In June 2001, the recession was 4 months old. The job change number stayed negative for 30 months, the longest streak ever.

During that stretch, the figure never went negative without a recession being under way. That is not a complete coincidence, of course, since the employment figures are one factor the National Bureau of Economic Research weighs in setting recession dates.

In every case, the recession was over before the job change figure turned up. Perhaps it is interesting that the lag between the official recession end and the first 12-month increase in private sector jobs was significantly longer than in earlier recessions. The lag was one year after the recession ended in March 1991, and two years after the most recent recession ended in November 2001.

If we assume that this year falls within the previous parameters, then the recession started sometime between May 2007 and March 2008, with October 2007 the most likely. Based on other data, October is probably the earliest date that the recession could have begun.

Again assuming the previous data is predictive, the last negative job number will arrive between November 2008 and November 2010, with August 2009 the most likely month.