- Jan 19, 2010
A Century of Mistakes in Baseball
Although the "wrath of randomness" does rear its head in the study of sports, often the numbers do tell a story. Let's start with a great story that reveals a century of mistakes in Major League Baseball (MLB).
In 1997, the Oakland A's ranked toward the bottom in Major League Baseball, in respect to both team payroll and winning percentage. The next season, Billy Beane became general manager, and part of this story stayed pretty much the same. Specifically, the lack of spending on players didn't change. What did change were the outcomes achieved by the A's. From 1999 to 2002, only the New York Yankees, a team that spent three times more on playing talent than Beane, managed to win more games in the American League. The term "more" is a bit misleading. The Yankees actually won only two more games than the A's across these four seasons.
How was this possible? It's been argued20 that the key was Beane's ability to recognize specific inefficiencies in baseball's labor market. Such inefficiencies allowed Beane to pick up talent that was both cheap and productive.21
At least, that's the story that's been told. For the empirical evidence supporting this tale, we turn to the work of Jahn Hakes and Raymond Sauer. These economists decided to investigate whether the baseball player market was, as they say, "grossly inefficient." Before we get to their answer, however, let's briefly describe an efficient labor market. A basic tenet in economics is that workers are paid in line with their expected productivity, that is, workers who are expected to be the most productive get paid the most. This suggests that baseball players who are expected to perform the best are paid the highest salaries (at least, once they become free agents). In a world where some teams are "rich" and others "poor," the best players typically end up on teams that have the ability to pay the most. In other words, we would expect the Yankees—or the "rich" team—to get the best talent, and a "poor" team like the Oakland A's should end up with the less capable players.
The key to the above reasoning is the phrase "ballplayers who are expected to be the most productive." This tells us that having money isn't enough. Teams have to be able to identify the "most productive" players. If one team can do a better job at identifying the "most productive," then that team might be able to field a very good team that's not very expensive.
To see if the Oakland A's actually followed this blueprint, Hakes and Sauer needed to connect three dots:
- They needed to uncover how various performance characteristics impact wins in Major League Baseball.
- They needed to figure out what individual teams were willing to pay for each performance characteristic.
- They needed to determine whether the salaries that various performance characteristics command is consistent with how those measures impact wins.
To cut to the chase, Hakes and Sauer found that "...hitters' salaries during this period (2000-2003) did not accurately reflect the contribution of various batting skills to winning games." Furthermore, "this inefficiency was sufficiently large enough that knowledge of its existence, and the ability to exploit it, enabled the Oakland Athletics to gain a substantial advantage over their competition."22
How did they reach this conclusion? First, data was collected on team winning percentage, team on-base percentage,23 and team slugging percentage24 for all 30 MLB teams from 1999 to 2003. They then ran a simple regression.
Okay, we get ahead of ourselves. What's a "simple regression?" Regressions25 are essentially the test tubes of economics. When a chemist seeks to understand the world, he or she steps into a laboratory and starts playing around with test tubes. These test tubes allow a chemist to conduct controlled experiments. Hakes and Sauer, though, could not conduct a controlled experiment with Major League Baseball (at least, Major League Baseball probably wouldn't let them do this). What they could do, though, is employ regression analysis. This is simply a standard technique economists employ to uncover the relationship between two variables (like player salary and on-base percentage), while statistically holding other factors constant. When properly executed, regression analysis allows one to see if the relationship between two variables exists; or more precisely, if the relationship between two variables is statistically significant.
Beyond statistical significance, we can also measure the economic significance of a relationship,26 or the size of the impact one variable has on another. Consider how on-base percentage and slugging percentage relate to team wins. Hakes and Sauer found both to be statistically significant. On-base percentage, though, had twice the impact on team wins. Such a result suggests that players should be paid more for on-base percentage. The study of salaries, though, suggested that prior to 2004, it was slugging percentage that got a hitter paid. In fact, in many of the years these authors examined, on-base percentage was not even found to have a statistically significant impact on player salaries.
After 2004, though, the story changed.27 An examination of data from 2004 to 2006 reveals that on-base percentage had a bigger impact on player salaries than slugging percentage. In other words, an inefficiency exploited by Billy Beane was eventually eliminated.28
It's important to note, though, how long this took. The National League came into existence in 1876. All of the data necessary to calculate on-base percentage was actually tracked that very first season in the 19th century. However, it was not until the 21st century—or after more than 100 years—that these numbers were understood by decision-makers in baseball. It appears that decision-makers in baseball made the same mistake in evaluating talent year after year, and this continued for a century. Such a tale suggests that maybe all those fans are on to something. Maybe coaches and general managers are capable of repeating the same mistakes.
Of course, one story from the real world of sports doesn't make a point. What we need is a multitude of stories. And that's what we provide. The stories we tell give insight into how free agents are evaluated, how teams make decisions on draft day, and even how choices are made on game day. We even present evidence that the evaluation of coaches in the National Basketball Association (NBA) is less than ideal.
All of these tales from the world of sports tell one very important story. Decision-making is not often as rational as traditional economics argues. And that story has an impact on our understanding of both sports and economics.