|Ivars Peterson's MathTrek|
July 3, 2000
Economist Paul M. Sommers of Middelbury College in Vermont has analyzed the data for the years from 1967 to 1998. He reports his findings in the May College Mathematics Journal.
Sommers defined six variables, one for each of the divisions (Eastern, Western, Central) of the National Football Conference (NFC) and the American Football Conference (AFC). If the Super Bowl winner was from a particular division, the relevant variable was assigned the value 1; otherwise, it was 0. The dependent variable (DJIA) denoted the percentage change in the Dow Jones Industrial Average between the closing average on the Friday before Sunday's championship game and the closing average on the last day of that calendar year.
Super Bowl winners and changes in the Dow Jones Industrial Average (DIJA), 1967-2000:
|Year||Winner||Conference||Current Division||Total Points||Percentage|
Change in DJIA
|1969||New York Jets||AFC||Eastern||23||-13.52|
|1987||New York Giants||NFC||Eastern||59||-7.74|
|1991||New York Giants||NFC||Eastern||39||+19.16|
|2000||St. Louis Rams||NFC||Western||39||?|
Here's Sommers' mathematical model:
DJIA = b0 + b1AFCE + b2AFCC + b3AFCW + b4NFCE + b5NFCC + b6TOTALPTS + b7TOTALPTS*ROOTS + e.
TOTALPTS gives the total number of game points, and TOTALPTS*ROOTS represents an "interaction term" where ROOTS equals 1 for a team from the original National Football League, including Cleveland, Indianapolis (formerly Baltimore), and Pittsburgh); otherwise, ROOTS is 0.
Sommers' model suggests that the stock market posts a significantly higher gain if the Super Bowl winner is from the National Football Conference Western Division. Moreover, the higher the combined team point totals are, the lower the percentage change in the Dow Jones Industrial Average.
At the same time, the model suggests that the Super Bowl theory's predictive power has declined precipitously in recent years. The victory by the Denver Broncos in 1998 implied that the percentage change in the Dow Jones Industrial Average would be -20.85 percent. In fact, the Dow closed out the year 1998 up 20.61 percent.
In general, the explanatory power of the model has fallen from nearly 82 percent to less than 53 percent in the last decade. Factoring in the 1999 results, when the Denver Broncos defeated the Atlanta Falcons, makes the fit even worse--only 46 percent.
"Despite the Super Bowl theory's surprisingly strong early record, reading the sports page now is not making it any easier to read economic tea leaves," Sommers concludes.
What accounts for the surprisingly strong predictive power of the Super Bowl theory in the early years? "Statistical fluke," Sommers suggests.
Copyright 2000 by Ivars Peterson
Norris, F. 1997. The Super Bowl predicts the market, and vice versa. New York Times (Jan. 24).
Peterson, I. 2000. Super Bowls and stock markets. Science News 157(June 17):399.
Sommers, P.M. 2000. The Super Bowl theory: Fourth and long. College Mathematics Journal 31(May):189.
You can find Super Bowl results at http://www.allsports.com/superbowl/results.htm. Additional information from the National Football League is available at http://www.nfl.com/.
Information about the Dow Jones Industrial Average can be found at http://indexes.dowjones.com/.
Comments are welcome. Please send messages to Ivars Peterson at email@example.com.