So called “top-down” earnings estimates for the S&P 500 are made by brokers and economists, who look at the S&P in aggregate and estimate the index as if it were a single “firm”. Whereas the “bottom-up” estimates are made by security analysts who estimate the earnings of the individual firms in the S&P, then add them up to estimate the total earnings of the index.  Richard Ripp compares top-down to bottom-up estimates and concludes that top-down estimates have proven to be more accurate over time.

Macro Man has a great post in which he compares two versions of top-down estimates for 2010 earnings for the S&P and one bottom-up estimate.

Looking at 2010 earnings estimates yield an incredibly broad range of forecasts. If you believe the crack-smoking bottom-up guys who strip out everything that could be construed as a “loss”, you get a resounding $74 pr share. Not bad!

Taking the same approach (stripping out the quarterly “one offs”), but from a top-down framework, yields a substantially less rosy result: earnings of just $46 per share. And actually counting all the turds for what they are on a top-down basis yields 2010 EPS of just $37 per share.

Here’s his chart (click it to enlarge) that shows various multiples applied to each of these three earnings estimates (the $74 bottom-up, the $46 top-down, and the “as the analysts think earnings will be reported” $37 top-down):

Note that even the most generous multiples applied to the “as reported” estimates still indicates the market is overvalued at present levels.  If analysts are underestimating earnings by say 15% and we adjust their “middle of the road” $46 estimate upward by that 15%, then we get a “corrected” estimate of about $53.  Even then, only if we then apply a multiple of 20X to them do we get a value comparable to the present market level. You be the judge – what’s a “fair” multiple for S&P earnings in 2010?  Is it 20X?  BTW, Jeremy Granthum’s current estimate of fair value for the S&P is 880.