The sports betting world may seem like an unlikely laboratory to test theories of investment-strategy psychology, but a team of finance scholars has done just that. In fact, the researchers say, the market for investments and the market for sports wagering are similar in several important ways, including the role of information, expert opinion, sentiment and middlemen in each. The researchers used data from the sports betting market to test a key investment-psychology theory, and discovered that in the real world, sports bettors did not behave the way the theory predicted they would.   "/> Ballpark estimates: Strategies similar for sports bettors, investors

Ballpark estimates: Strategies similar for sports bettors, investors

April 12, 2006

The wide world of sports and Wall Street may not seem to have much in common, but it turns out that sport offers a laboratory environment where scholars can test theories of investment-strategy psychology.  

 

Finance Professor Greg Durham of Montana State University, a graduate of the doctoral program at the W. P. Carey School, joined W. P. Carey professors Michael Hertzel and Spencer Martin in writing "The Market Impact of Trends and Sequences in Performance: New Evidence," published in the October 2005 issue of the Journal of Finance. Durham recently presented the paper at the spring conference of the Chicago Quantitative Alliance, a professional organization of quantitative investment practitioners.

 

The unconventional lab

 

Durham explained that he and his co-investigators tested their ideas about investment psychology by looking at the way people bet on sports. Sports wagering markets, Durham said, share similar characteristics with securities markets.

 

Both the sports and the investment market include users who may have informational advantages, Durham said. For example, an investor who follows Intel carefully by reading analysts' columns and the company's press releases has an informational advantage over investors who are not as diligent about research. The counterpart in sports is the football fan who may have knowledge of a player's injury status, or whose memory stores detailed information going back many seasons concerning the teams, the coaches and the won/lost records.

 

And, both markets are permeated by a plethora of experts, Durham said, such as financial pundit Abby Cohen of Goldman Sachs, who is quoted by media outlets such as Smart Money, or mathematical guru Jeff Sagarin, whose ranking system is used by bettors to predict winners.

 

Sentiment also plays a key role in the behavior of investors and bettors. Investors often put money into brands and companies that are iconic household names: Disney or Intel. Bettors may pick their alma mater, or stand loyally by the hometown Red Sox.

 

And each has a middle party -- the market maker for stocks and the bookmaker for sports bets -- who facilitate transactions, Durham said. For investments, the market maker adjusts price up or down to balance supply and demand for trading; in sports betting the bookmaker uses the point spread to balance the amount of money wagered on each team in a contest.

 

But the two differ in one important way. Unlike investing, where a company's true value changes continuously and is thus difficult to measure, sports wagering offers a known outcome, and therefore known value, in a short period of time. The bettors who participate in this market never have to wonder whether they valued the asset properly, because the outcome of the game provides the answer. Once a game is complete bettors can compare the point spread versus the actual outcome and unambiguously determine whether they are holding an asset (the wager) that is valuable or worthless.

 

Betting/investing on the trend

 

Durham, Hertzel and Martin's research was designed to test the regime-shifting model published in 1998 by Nicholas Barberis, Andrei Schleifer and Robert Vishny (BSV) in the Journal of Financial Economics. This model suggests that despite the fact that companies' earnings performances are random, investors believe that performance is being governed by a "continuation regime" or a "reversal regime," and will make predictions based on past trends or performance.

 

Durham explained that this behavior -- making predictions based on presumed performance regimes despite the fact that performance is random -- is consistent with biases that cognitive psychologists call "representativeness" and "conservatism."

 

Individuals impaired by representativeness tend to incorrectly extrapolate details about a small sample onto what they presume to be its parent population. Conservatism causes an investor to underestimate the material value of new information while over-relying on his previously existing information.

 

The coin-flip example demonstrates representativeness.

 

"People impaired by representativeness who see a coin flip come up heads four times in a row will often incorrectly assume that the coin is more heavily weighted towards heads, when in fact the chances that it will be heads on the fifth flip are still 50-50," Durham explained. "Similarly, if I've seen a company give me four positive earnings surprises in a row, the chances are 50-50 that the next one won't be a positive earnings surprise. However, this person assumes that a few instances are representative of the entire population, and assumes that the next surprise will also be pleasant."

 

Conservatism, on the other hand, underemphasizes new information and over-emphasizes the old. "Let's assume I have received three pieces of good news about a company and the fourth came out as bad news," Durham said. "I should assume the company is 75 percent good and 25 bad, but if I'm under the influence of conservatism, I will undervalue that last piece of information. I don't fully believe in its validity."

 

The BSV hypothesis was validated in a study using Cornell University MBA students by researchers Robert Bloomfield and Jeffrey Hales. The students were told that stocks perform randomly over time. Then they were shown graphs of eight specific historical performance sequences and were asked to predict future performance. Despite the clear explanation that performance is random, the students relied heavily on past performance in making their guesses.

 

Testing theory on the football field

 

Durham, Hertzel and Martin virtually went to the stadium to test the validity of these two studies, including the two fundamental tenets of the BSV model: that performance follows a random walk, and that in formulating beliefs about future outcomes, investors rely upon past performance in a manner suggested by BSV.

 

They used eight seasons (1991-98) of Division I college football games for their data set. For each of the 4,584 games, they recorded the opening point spread, the change over the week in the point spread caused by bettors' and bookmakers' reactions to information, the closing point spread right at kickoff, and the actual outcome -- the score. They also charted winning and losing streaks relative to the point spreads -- the performance history.

 

Durham and his colleagues found that performance in relation to point spreads was random. The number of teams on a given streak length falls by roughly 50 percent with each successive increment in streak length. And, only a handful of teams have histories with low numbers of (necessarily longer) streak -- no more than would normally be expected under randomness.

 

They also found that bettors do not appear to formulate expectations in the specific way hypothesized by BSV. By using teams with the same eight outcome performance histories as used by Bloomfield and Hales, Durham and his W. P. Carey collaborators were able to determine that because the point-spread fluctuations following such patterns were statistically insignificant, "football market participants appear completely insensitive to [recent] performance." The authors reinforced this conclusion by expanding their sample to include teams having every conceivable eight-game historical performance sequence (of which there are 256 possible combinations), and found the same statistically insignificant reactions.

 

In other words, bettors didn't act according to the BSV regime-shift model, or like the Cornell MBA students -- who reacted to historical performance in a manner almost perfectly consistent with what BSV predicted.

 

The authors also wanted to evaluate what bettors thought of performance trends. The BSV model predicts, in the context of sports wagering markets, that investors are increasingly more likely to expect a continuation in performance when the number of games in a winning (or losing) streak against the spread increases. By studying how the point spread changed based on a team's most recent winning or losing streaks, the authors found that the spread moved in the direction of teams on short winning streaks (3 games or less) but counter to those on longer winning streaks.

 

So what does this mean in terms of investor behavior?

 

"Contrary to the prediction of the Barberis, Shleifer, and Vishny model," write the authors, "this finding suggests that investors eventually expect a performance reversal as streak length grows."

 

The paper's conclusions may have implications for future behavioral models that are designed to explain investor behavior. This research also helps to legitimize sports wagering markets as a useful laboratory in which to test financial theory.

 

Bottom line:

  •  Sports wagering markets are very much like stock markets, in terms of participants and mechanics.
  • Performance in the college football wagering market appears to be random, suggesting that wagering in the point-spread betting market is a 50-50 proposition.
  • Theoretical models do not always accurately predict human behavior.
  • College football bettors seem to expect short-run winning and losing streaks (in the point-spread wagering market) to continue and they expect long-run streaks to reverse.