The Superstar Effect: The Case of Tiger Woods

A colleague (Thanks MS!) points me to a neat paper by Jennifer Brown of Northwestern University and NBER published recently in the Journal of Political Economy titled, Quitters Never Win: The (Adverse) Effects of Competing with Superstars. The paper looks at the effects that Tiger Woods had during his prime on the golf scores of field in PGA tournaments. The paper hypothesized that the presence of such a dominant player would lead to diminished performance by other top players, based on a theoretical presumption of reduced effort by those players.

The paper utilizes an impressive set of data from PGA Tour events from 1999 to 2010 and arrives at the following top-line conclusions:

The main results of the paper are as follows:

1.	The presence of a superstar in a tournament is associated with reduced performance from other competitors. In general, the adverse superstar effect is larger for higher-skilled golfers relative to lower-skilled players
2.	Reduced performance is not attributable to the adoption of risky strategies. Players do not appear to be “going for the green” more in the presence of a superstar. Moreover, the variance of players’ hole-by-hole scores in PGA tournaments is not statistically significantly higher when Woods is in the field relative to when he does not participate.
3.	Superstars must be “super” to create adverse effects: The adverse superstar effect is large in periods in which Woods is particularly successful and disappears during periods in which he is performing relatively poorly on the course.

Here are some numbers for the period 1999-2006:

In the first round, the performance of top-ranked players appears affected by the superstar. For major and regular events, top golfers’ first-round scores are 0.6 strokes higher when Woods is in the field relative to when he is not. In an examination of only regular events, the superstar effect is 0.54 strokes for the first round. The magnitude of the effect is substantial, particularly when one considers that an average of two (and as many as eight) players share first place after the first round of tournament play. Moreover, when we account for ties, the top two first-round scores in a tournament differ by an average of only 0.8 strokes. Note that unranked players’ scores are not significantly different when Woods participates. This nonresult aligns with the intuition that players who are low in the distribution of relative skill or who expect to finish in the nearly flat portion of the tournament prize distribution may not be adversely affected by a top competitor. For example, the difference between fortieth-place and forty-first-place prizes in an average regular tournament is less than $1,000; thus, a one-position shift in the distribution has little marginal impact on players’ performances. . .

In an examination of major and regular events, the tournament scores of ranked players are significantly higher when Woods is present: estimates suggest that the effect is between 1.3 and 0.7 strokes, depending on player rank. In only regular events, the superstar effect for top 20 players is positive but not statistically significant. In general, however, the size of the superstar effect is substantial for good PGA Tour golfers: on average, fewer than two strokes separate first and second place in PGA tournaments.

The paper also explores the possible influence of an impressive range of potential confounding factors, such as weather and television coverage and finds the results to be robust.

Why might this effect occur?  The author presents data on Davis Love III who can reportedly earn $100,000 per day making a corporate appearance. The downside of the adverse superstar effect are small by comparison:

To understand the impact of the effect on Woods’s competitors, I ask: What if a single average player were able to overcome his own adverse performance by exerting effort and post scores that were one stroke better in tournaments with the superstar? I simulate the total winnings of each of the ranked players, assuming that his scores were one stroke lower in all events with Woods between 1999 and 2006. On average, a golfer would have earned approximately $28,000 more. Given that an average top 200 player played in 12 events with Woods and earned $3.4 million from the PGA Tour during the 8-year period, the return to effort seems small. For example, in the simulation, a one-stroke improvement by David Love III would have increased his average per-tournament earnings by approximately $10,000—considerably less than his reported daily rate for corporate appearances.

What about Woods?

How much would Tiger Woods’s earnings have been reduced if all of his competitors played as well as they did when he was not in the field? In my main results, I identify a superstar effect of approximately one stroke for players ranked in the top 200 in the world. I simulate the distribution of prizes if all ranked players’ tournament scores had been one stroke better when they competed against Woods; that is, I removed the estimated superstar effect from players’ scores. In 34 of the 136 tournaments that I studied, the simulated improvement in competitors’ performance had no effect since Woods was sufficiently alone in the score distribution to avoid a rank-order shift. In 20 events, the simulation shifted at least one golfer into a tie with Woods in the final tournament standing; in the remaining events, Woods shifted from a tie at a higher prize position to a tie at a lower prize position.

My calculations suggest that Woods’s PGA Tour earnings would have fallen from $54.4 million to $48.4 million between 1999 and 2006 had his competitors’ performance not suffered the superstar effect. By my estimates, Woods pocketed nearly $6 million in additional earnings because of the reduced effort of other golfers—prize money that would otherwise have been distributed to other players in the field. Viewed in this light, the superstar effect is economically substantial.

It is good to be a superstar, and apparently not so bad just being super either.