Consumer Confidence viz S&P, from Alston Mabry
Here's a shocker:
Take the Confidence Board Consumer Confidence raw number from May, 2010, through November 2011, and calculate monthly % changes; then regress against that series the previous month's % change (close-close) in the S&P…and you get the previous month's change in S&P being an excellent predictor of the CC change!
multiple R: +0.69
R squared: +0.48
…and just to be humorously over-precise:
CC%chng(month[1]) = -0.00498 + 1.8*S&P%chng(month[0])
Jeff Rollert writes:
I cannot recall when that strong a relationship didn't hold, so much so I stopped doing the calculations.
One of my current stat problems is finding data sets that do not have some strong feedback dynamics, where a strong data point is reflected in market prices and becomes non-predictive. The expansion of SAP, ORCL and other large data set manipulation/analytic engines seems to be one of the culprits.