For forecasting oncoming recession, from a purely statistical standpoint, we have two interesting data series to follow, the Piger Recession Probabilities and the Term Spread Probability of Recession
In the latest release of the Piger Recession Probability, the April data (... there is a reporting lag) indicates that the probability of recession has increased to 1.69%.
In 2008, Marcelle Chauvet of the University of California and Jeremy Piger of the University of Oregon published a paper titled “A Comparison of the Real-Time Performance of Business Cycle Dating Methods” which outlined two novel statistical methods (most notably the markov-switching method) for distilling recessionary turning points out of the very same macro data series that the NBER uses to make it’s cycle assessments.
As for the Term-Spread Probability of Recession, the latest data indicates that the probability for recession appears to be on the rise with the January 2016 probability (the probability that there will be a recession by that date) of 4.4%.
Spearheaded by economist Professor Arturo Estrella of the Rensselaer Polytechnic Institute, this method derives a probability of recession from the spread between long and short yields (10-year and 3-month) and is by all accounts the standard for recession probability forecasting.