S&P 500 vs Obama's disapproval: a real correlation?

First, some background information. In late August, Jim Cramer of Mad Money ran a segment which showed a relationship between the S&P 500 and Obama’s disapproval rating. It first caught my attention at Hot Air, where moderator Ed Morrissey blogged:

Sharp investors always look for leading indicators to market shifts, and Jim Cramer thinks he’s found a doozy. After watching this segment, it’s hard to disagree, too.

Ed pointed out later in an update, and rightfully so, that this relationship was not necessarily causation but certainly showed some correlation.

Always a bit late to the party, a more mainstream outlet (Gallup) finally looked at this a week later and concluded there was no relationship between the market and Obama’s rating. Moe Lane, blogging the next day, reiterated the lack of relationship but also pointed out some problems with Gallup’s work. For example:

  • Gallup used the Dow instead of the S&P 500.
  • Gallup used Obama’s approval rating instead of disapproval rating (these two are independent).
  • Gallup compared data over the entire range of 2009 instead of starting at the March low (as Cramer suggested).

It should also be pointed out, one important issue (in data analysis) has been largely overlooked: actual calculation of the correlation coefficient. If there really is no relationship, this coefficient should be 0; if there is a relationship, it can be as high as 1. To date, I’ve yet to see anyone present the actual numbers besides my largely unreceived post at Hot Air — it sat in moderation so long, it only published after the headline was off the front page. So I’ll summarize that work here for a different (and hopefully, more aware) audience.

  1. Disapproval rating data can be retrieved from Gallup’s site in csv format.
  2. S&P 500 data is available through Yahoo Finance.
  3. I placed these two data sets in Excel, did necessary formatting (i.e. text to number), and compared the data over the date range from 3/9 to 8/27.
  4. Some days had to be interpolated: no polling (i.e. holidays) or no market (weekends).

Excel will then calculate the correlation coefficient (CORREL function) for the data over any range you desire. I found the coefficient was 0.78 over the entire range (March to August); it also goes through a local maximum in the month of June: more specifically, around June 15, the coefficient hits 0.8. One can speculate why that date may be more important (perhaps, Obama’s mishandling of the Iranian election) but what is fairly certain from the data is that there exists a strong correlation (anything over 0.5) between the S&P 500 and Obama’s disapproval ratings. Is Obama’s failure causing the improvement in the market? No. But what we can conclude is that the same reason his ratings are falling is likely the same reason the market is improving.