joining us now is david baylon and jeff rosenberg.e saw an adp number that was a little bit better. we see these things and people complain about quarterly results being every three months because there's regression to the -- that it's too short of a period. we do this monthly and we do seem to see a reversion to the mean from time to time as we do with steve's work on the -- which you just looked up. >> i did. >> some months you're off by 2,000. >> i've got five and seven this year. my average error this year is 41,000. >> are you proud of that? >> i'm very proud of that. are you kidding? >> i'd like to point out economists excited about being less wrong than other economists. >> that's as good as you're going to get. look. what's you're model say, caruso-cabrera? >> i don't have a model. >> okay. well, put yourself out there. i've got an 84,000 error in february. and a 7,000 error in january. >> you could drive a truck through that spread. >> so the reason for bringing this up -- >> and a 5,000 error in april. you brought it up, joe