Property rights and resource use: evidence from MLB starting pitchers

ABSTRACT

We conduct a unique test of the efficiency of property rights in major league baseball. The rights to the services of players are resources that can be possessed by the clubs or by the players themselves. This right was effectively reassigned from the club to the individual player when free agency was introduced in 1976. Players, however, only qualify for free agency after 6 years of service, and until that time a temporary property right is possessed by the club. In the absence of efficient bargaining, clubs that possess only a temporary right do not bear the full risk of injury or disability associated with using pitchers. Clubs in this situation can therefore have an incentive to overuse star pitchers. This theoretical prediction is supported by the statistical inferences of our econometric models.

KEYWORDS:

• Property rights
• Coase Theorem
• resource exploitation
• free agency
• major League Baseball

JEL CLASSIFICATION:

• D23
• D52
• J41
• L83

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ See for example Hylan, Lage, and Treglia (1996), Cymrot, Dunlevy, and Even (2001) and Krautmann (2009).

² Moderating pitching work might not actually be effective at reducing injuries or preserving long-term value, but for our purposes we require only that decision makers in MLB do in fact act on the presumed relationship. The Los Angeles Dodgers, after suffering a ‘barrage’ of injuries in 2016, cut back the workload on their starters so that by 2017 the club was pleased that starters were averaging a paltry 89 pitches per game (Kepner 2017).

³ Note that the club can more efficiently bear the risk because each individual player’s production comprises a relatively small fraction of the club’s total income, making the club comparatively less risk averse than the player.

⁴ Agreements to limit work would not be explicitly written into contracts, but would instead be implicit and enforced through the reputation of the club or the team manager. Implicit limits plausibly offer the advantage of flexibility in dealing with positive shocks to the return on additional pitching work, such as a pennant race or a rash of injuries to other members of the pitching staff.

⁵ The source of this market power is the fact that elite baseball talent is highly inelastic in supply and has no close substitutes. Although unfree agents cannot negotiate with other clubs, they can withhold work effort or hold out by refusing to sign contracts. If they have three years of experience, they can elect to take the club to binding arbitration.

⁶ Evidence that the wealth effect of free agency increased demand by pitchers for better treatment is provided by Hylan, Lage, and Treglia (1996) who found that veteran pitchers that had exceeded the six-year service threshold were traded less frequently than were comparable veterans under the reserve system. This finding suggests that free agents effectively purchase protection against the risk of being traded to a less preferred club or city. Such trade protection, like our hypothesized limits on work, essentially restricts the club’s ability to dispose the player’s services. Because trades involve third-party effects on players, restricting them can enhance efficiency and can improve assortative matching of players to clubs. For an example of assortative matching in sports, see Filippin and van Ours (2015).

⁷ Fair (2008), in a model of pitcher performance, uses a roughly comparable selection criterion.

⁸ Perhaps the most direct measure of the intensity of pitcher use is pitches thrown per game. Reliable pitch counts, however, are not readily available before the late 1990s. Pitches can nonetheless be predicted nearly perfectly accurately (r =.99) from a linear projection on four variables: innings, strikeouts, bases on balls, and hits. Predicted pitches from this model correlate very closely (r =.97) with innings per game. Thus innings per game and pitches per game turn out to be essentially equivalent measures of work. We prefer to use innings per game since the data are available, but all our reported inferences also hold using predicted pitches per game.

⁹ Of course, finer distinctions on term are possible in theory but in practice would require contract details, such as club or player options to extend term, that are not readily available.

¹⁰ Note that the definition of Short_i,t also is consistent with an equilibrium model of perfect foresight expectations; that is, a model in which clubs assign an ex ante departure probability of one to players who depart ex post, and zero otherwise. In reality, expectations might roughly approximate perfect foresight since clubs can probably assess fairly accurately whether or not they will retain a player.

¹¹ An issue arises involving mid-season trades that cause sixth-year, imminent free agents to split their season between two clubs. Since a trade implies a departure that the club can anticipate, Short_i,t equals one for the stint with the club that trades the player away. Short_i,t equals one for the stint with the receiving club only if the free agent declines to re-sign. For players traded mid-season during the first year of a free-agent contract, Long_i,t equals one for both team-stints because the club will want to preserve the value in trade of the long-term right.

¹² We also obtained estimates using an alternative performance measure, Fielding Independent Pitching (FIP). Observations on ERA+ and FIP are from www.baseball-reference.com. For details on how these variables are defined, see The Hardball Times at http://www.hardballtimes.com/main/statpages/glossary.

¹³ The small markets also include Cincinnati, Milwaukee, Pittsburgh, St. Louis, Oakland, Tampa Bay, Baltimore, Kansas City, Cleveland, San Diego, and Denver. Small_i,t equals one for Baltimore only from 2005 when Baltimore’s market was divided by relocation of the Montréal franchise to Washington, D.C.

¹⁴ Note that factor groups overlap in the explanatory variables they contain; for instance, the interaction Short×Good×Small allowing overuse to be conditioned by both performance and market size involves three factors, so it appears in three different groups.

¹⁵ Substituting Fielding Independent Pitching (FIP), which correlates about 0.7 with ERA+, as the performance measure yields a slightly higher p-value of.06.

¹⁶ A theoretical possibility is that the small-market effect might have existed but largely disappeared in 1997 after MLB introduced a ‘luxury tax’ that put small market clubs on a more equal footing with other clubs. As a check, we re-estimated our full model in (3) using only data from years prior to 1997. The estimates from this sub-sample continue to find no small market effect (p =.87). Furthermore, even though this sub-sample retains only 46 out of 97 observations on pitchers under short-term control, we continue to obtain a significantly positive, and comparably sized, estimated effect of a short property right.

¹⁷ Our estimated age function implies that pitchers peak at age 26.6, which is remarkably close to Fair’s (2008) point estimate of 26.5. Fair’s dependent variable, however, is ERA+, not innings per game.

¹⁸ To check for robustness, we re-estimate the same model using just the observations involving at least 90% starts. The sub-sample of 2,135 (out of 2,641) observations yields an estimated overuse function nearly identical to equation (4)(4) $\Delta {\text{IPG}}_{\text{i},\text{t}}=\{\begin{array}{l}-.0753+.00181ER{A}_{i,t}^{+},ifER{A}_{i,t}^{+}\ge 105\\ 0,ifER{A}_{i,t}^{+}105\end{array}$(4) estimated from the full sample. Inspecting the estimated year-effects suggests a secular decline in innings per game occurring in 1995. Hence to test the stability of the model over time, we re-estimate using only observations from 1995 and later. Again, the resulting overuse function is almost indistinguishable from (4).

¹⁹ Empirical evidence on the threshold model is mixed, but most recently, Carleton (2013) finds support for the model.

²⁰ This second assumption is rather conservative in that it likely understates the effect of overuse on Stress. Starters have good days and bad days, and on bad days, starters get pulled from the game by the manager. The early exit in poor starts causes pitchers to throw fewer pitches on their bad days than on their good days. Since good performance incentivizes overuse, most overuse would occur on relatively high-pitch, good days, and the cubic function in (5) magnifies the effect on Stress.

²¹ An alternative interpretation is that pitchers themselves can exercise control over their work, and that imminent free agents want to pitch more in order to send a signal of quality to potential employers. We consider this interpretation less plausible, but in any event, the scenario nonetheless implies an inefficiency due to asymmetric information.