Via Greg Mankiw
According to a National Economic Research Board paper:
Monthly US data on payroll employment, civilian employment, industrial production and the unemployment rate are used to define a recession-dating algorithm that nearly perfectly reproduces the NBER official peak and trough dates. The only substantial point of disagreement is with respect to the NBER November 1973 peak. The algorithm prefers September 1974. In addition, this algorithm indicates that the data through June 2008 do not yet exceed the recession threshold, and will do so only if things get much worse.