NEGOTIATED SETTLEMENT UNDER MLB FINAL-OFFER SALARY ARBITRATION SYSTEM

NEGOTIATED SETTLEMENT UNDER MLB FINAL-OFFER SALARYARBITRATION SYSTEM

J. RICHARD HILL and NICHOLAS A. JOLLY

This paper provides a detailed analysis of negotiated salaries under Major LeagueBaseball’s final-offer arbitration process using data from the 2007–2010 seasons.There is a wage premium of 25% for hitters and 14% for pitchers filing for arbitration.Interestingly, there is an additional premium for exchanging offers for hitters but notfor pitchers. The additional premium in salary for hitters who exchange offers withtheir clubs amounts to 7%. (JEL J31, J52)

I. INTRODUCTION

In negotiated settlements, the costs of reach-ing an agreement can be expensive in termsof both time and money. If both sides couldrealize the objective, unbiased, end result fromthe beginning, then an accord could be reachedquickly. This is rarely the case. In industrialrelations, the use of mediators and arbitrators isdesigned to assist union and management nego-tiators in reaching a contract settlement. Theuse of mandatory interest arbitration to resolvebargaining disputes is found predominantly inthe public sector; this is designed to reducethe negotiation costs to taxpayers. Unable toreach a settlement, unionized police, firefight-ers, teachers, and the respective governing bod-ies in charge of the budgets of these employeegroups are often forced to allow a neutral,third party to decide the major fiscal contentsof their new collective bargaining agreements(CBAs). In rare cases such as in New Jersey,the wages of police and firefighters are deter-mined by final-offer arbitration, a system inwhich the arbitrator is constrained to chooseeither the offer proposed by the union or theoffer proposed by the management. Final-offerarbitration was designed to reduce the use ofexpensive arbitration as, in theory, the require-ment that the arbitrator must accept the final

Hill: Department of Economics, Central Michigan Univer-sity, Mount Pleasant, MI 48859. Phone 989-774-3706,Fax 989-774-2040, E-mail [email protected]

Jolly: Department of Economics, Marquette University, POBox 1881, Milwaukee, WI 53201-1881. Phone 414-288-7576, Fax 414-288-5757, E-mail [email protected]

offer of one side or the other would push bothsides closer together and allow for a negotiatedsettlement.

Major League Baseball (MLB) has used asystem of final-offer salary arbitration (FOSA)since 1973 to determine the salaries of playersof a certain experience level. Baseball’s FOSAprocess was thought to provide a “laboratoryexperiment” that would provide greater insightinto the workings of such a system of interestarbitration because of the level of data availablefor analysis. The historical record of these caseshas been studied extensively by economists toanalyze topics such as the determinants of arbi-trators’ decisions, the determinants of the prob-ability that the dispute will end in arbitration,and the effect of each party’s willingness to bearrisk on the salary outcome. Much of the previ-ous research into baseball’s final-offer systemhas centered on the final offers themselves andthe arbitrators’ decisions. However, under base-ball’s system, a negotiated salary agreement cancome at any point before the arbitration panelrenders a decision.

This paper uses data from the 2007 through2010 seasons and analyzes negotiated salariesunder three out of the four separate stagesof the FOSA system: those who are eligiblefor arbitration and do not file for it; thosewho file for arbitration and do not exchange

ABBREVIATIONS

CBA: Collective Bargaining AgreementFOSA: Final-Offer Salary ArbitrationMLB: Major League Baseball

533

Contemporary Economic Policy (ISSN 1465-7287)Vol. 32, No. 2, April 2014, 533–543

doi:10.1111/coep.12048© 2014 Western Economic Association International

534 CONTEMPORARY ECONOMIC POLICY

salaries; and those who exchange salaries anddo not reach the arbitration stage of the process.Specifically, the analysis provides estimates ofthe average treatment effect for players movingto each subsequent, aforementioned stage inthe bargaining process. No other studies ofthe FOSA process in MLB have examined theeffects of filing for arbitration on the salaries forplayers. Other published studies have used thelist of players who filed for arbitration as theirdataset. This ignores the first step in the baseballarbitration process.1 Burgess and Marburger(1993) concluded that arbitration awards wonby players were higher and those won byowners were lower than negotiated settlementsfor comparable players. Miller (2000b) foundthat negotiated salaries that occurred after theexchange of final offers under arbitration weresignificantly different from those for free agents,while Marburger (2004) found that the averagefree agent salary was a significant determinant ofboth player and management final offers. Otherresearchers (Farmer, Pecorin, and Stango 2004;Miller 2000a) have examined the role of risk inshaping the outcome of negotiated and arbitratedsettlements after the exchange of final offers.Therefore, the majority of the earlier literaturehas focused on the stages of the process aftereach side has made and exchanged final offers.Because the majority of players who are eligiblefor salary arbitration today settle before theexchange of offers, it is important to understandthe earliest stages of the process, which is thefocus of this paper.

The rest of this paper proceeds by examin-ing the FOSA process in MLB in Section II.Theory and the previous literature are discussedin Section III. Section IV provides a discussionof the data, model, and empirical methodology,while Section V details the results of the analy-sis. Section VI offers conclusions.

II. THE FINAL-OFFER SALARY ARBITRATIONPROCESS

The FOSA system began with the 1973 CBA.The FOSA system has four distinct stages:players being eligible for arbitration; filing forarbitration and not exchanging offers; exchang-ing offers and settling prior to arbitration; and

1. Beginning with 2007, an internet website (Cot’sContract website at http://mlbcontracts.blogspot.com) madeavailable a list of players who were eligible for arbitrationbased on service time.

arbitrating salaries.2 Service time determineswhether a player is eligible for FOSA.3 Cur-rently, eligible players are those who havebetween 3 and 6 years of service time.4 Play-ers with 2 years of service who had at least86 days of service during the preceding sea-son and ranked in the top 17% in total serviceamong this group are eligible as well (theseplayers are known as the super-twos).5 Play-ers know if they meet these eligibility crite-ria shortly after the conclusion of the season.Either eligible players or owners can file forarbitration between January 5 and 15 follow-ing a season. Players and clubs exchange offersby January 18. The league then schedules arbi-tration hearings between February 1 and 20. Athree-member arbitration panel renders a deci-sion usually within 24 hours after the conclusionof the hearing.6 The panel must choose either thesalary offer of the player or the club and cannotrender a compromised settlement.

The parties can negotiate a salary any timebefore the arbitration panel renders its decision,and this occurs in the majority of cases. Forexample, between 2007 and 2010, 567 play-ers were eligible for FOSA. Of those players,467 (approximately 81% of the eligible) filed forarbitration, 182 players (about 32% of the eli-gible) filed for arbitration and exchanged offers,and 22 players (about 4% of the eligible) wentthrough the entire arbitration process. In fact,385 players (68%) settled before the exchangeof offers. Using data from 1993 to 1996, Farmer,Pecorin, and Stango (2004) had a sample of527 players who filed for arbitration, with only82 (16% of those who filed) players settlingbefore an exchange of offers, 374 (71% of those

2. Players who are eligible for free agency and offeredsalary arbitration by their team are not considered here.

3. The CBA defines one day of service as each day aplayer is on a team’s active roster. It takes 172 days to getone service year. The days begin with the first regularlyscheduled game in a season and conclude with the lastregularly scheduled game in a season.

4. Before the 1985 CBA, players used to be eligibleafter two years of service time.

5. Super-twos were added to the 1990 CBA.6. The use of a three-member panel rather than a sin-

gle arbitrator began with the 1997 CBA. The arbitrationpanel can only consider the following six criteria whenrendering a decision: (1) the quality of the player’s contri-bution to his team in the past season, including performance,leadership, and public appeal; (2) the length and consis-tency of the player’s career performance; (3) the recordof the player’s past compensation; (4) comparative baseballsalaries; (5) mental or physical player defects; and (6) recentperformance by the club, including league standing andattendance.

HILL & JOLLY: MLB FINAL-OFFER SALARY ARBITRATION SYSTEM 535

who filed) exchanging offers, and 71 (13% ofthose who filed) receiving arbitrated settlements.Because the authors only use a sample contain-ing players who filed for arbitration, it is difficultto get a direct comparison of the numbers. Still,however, there is a dramatic contrast in some ofthese figures: 374 players exchanging offers inthe 1993–1996 period versus 182 players in the2007–2010 period; 71 settlements reached byarbitration from 1993 to 1996 versus 22 settle-ments reached by arbitration in the 2007–2010period.

The contrast in these figures illustrates thebelief that a change in the approach to the useof salary arbitration has taken place despite nochange in the rules and regulations governingthe process. If there is an accepted relationshipbetween salary and past performance, then thismay explain why some agreements on salaryare being reached during earlier stages of thenegotiation process. Another potential explana-tion follows the reasoning of Brown and Link(2010) that labor relations are more harmoniousnow between the owners and the players follow-ing previous decades of discord over collusionby owners and strikes by the players.

III. THEORY AND PREVIOUS LITERATURE

A beginning point for almost all of thetheoretical models on FOSA in baseball is thework of Farber (1980). Using his basic constructfor final-offer arbitration, it is assumed thatan arbitrator will choose the offer made by aclub if:

|yA − yc| < |yA − yp|,(1)

where yA represents the arbitrator’s calculationof an objective salary for the player given thecriteria set forth in the CBA, yc is the offerrendered by the club, and yp is the offer fromthe player or his agent; it is assumed that yc andyp are a function of yA. The player’s offer willbe chosen if the inequality holds in the oppositedirection.

Under the assumption that clubs’ and players’salary offers are risk-neutral and arbitratorsare unbiased and interchangeable, Faurot andMcAllister (1992) find that four of the criterialisted in the CBA are all significant determinantsof the expected value of the arbitrators’ fairsettlement. Fizel (1996) finds racial bias in thedecisions rendered by arbitrators. Marburger andBurgess (2004b) use a probit model to predict

the winning offers in arbitration cases. Theyconclude that the FOSA process not only favorsreasonable offers but also creates an incentiveto settle before arbitration to avoid unfavorablerulings.

Farber’s (1980) model can also be used toanalyze the effect of each side’s level of risk-aversion on the outcome. If F(yA) represents thedistribution of the arbitrator’s fair settlement andis assumed to be known by both sides, then theclub chooses a bid yc that will maximize

[1 − F(yc + yp)

2

]yc + F

[yc + yp

2

]yp,(2)

assuming that utility from the bids are strictlyconcave and increase monotonically. The playerselects his bid, yp in a symmetrical fashion.From these, each side derives a reaction functionto the other’s optimal offer and the simultaneoussolution of each yields Nash equilibrium.

Farber (1980) discusses the so-called contractzone as an area in which parties may reachan agreement because of a convergence oftheir final offers. This contract zone can bedefined as:

yc < yL < yU < yp,(3)

where yL is the lower bound of the zone and yUis the upper bound. Some research has focusedon the contract zone. Marburger (2004a) sug-gests that final offers by both management andplayers are a weighted average of the player’spast season compensation and the average freeagent salary of the current season; this sameresult holds using only cases that ended in arbi-tration. Hadley and Ruggiero (2006) use non-parametric analysis and conclude that arbitratorsand the FOSA process are approximately mim-icking the free agent process.

Faber (1980) contends that if uncertainty overyA, the arbitrator’s fair settlement, disappears,then the contract zone shrinks to yA as bothyc and yP are a function of yA. However, ifuncertainty increases the gap between the finaloffers, then arbitrated settlements may increase.These are said to be of “low quality” byFarber (1980) as they fall outside the contractzone. The conclusion by Burgess and Marburger(1993) that arbitrated salaries won by playerswere higher and those won by managementwere lower in the baseball FOSA system thannegotiated salaries for comparable players offersempirical verification of Farber’s conclusion.


Miller (2000b) incorporates negotiating costsinto his theoretical model to allow for nego-tiated settlements in baseball’s FOSA system.Farber (1980) concludes that costs of arbitra-tion widen the contract zone and increase thelikelihood of a negotiated settlement. However,if costs are asymmetrical, then the party sub-ject to the higher costs will become more risk-averse and the resulting contract zone will moveunfavorably for this party and get larger over-all. According to baseball’s CBA, the hearingcosts of arbitration are borne equally by theclub and player, and each is responsible forhis own expenses and those of his counsel orother representatives. Since a club may reapsome economies of scale if hearings are set formore than one player, it is possible that costsare asymmetrical.

The basic Farber model can be used to framethe current empirical research, but some interest-ing questions arise. To what extent does filingfor arbitration differ from simple eligibility forarbitration in the effect on a negotiated salary?Second, to what extent does exchanging offersdiffer from simply filing in the effect on anegotiated salary? Previous studies of baseball’sFOSA system have not dealt with these issues.Marburger and Scoggins (1996) use a probitmodel to determine that higher quality playersare more likely to file for arbitration and pressfor an arbitrated settlement. This means thatselectivity bias is a likely problem in the esti-mation of salary equations for a self-selectinggroup of players. Farmer, Pecorin, and Stango(2004) use a two-stage process to adjust for anyselectivity bias to isolate the effect of aggres-sive bargaining behavior on negotiated versusarbitrated salaries. Likewise, Miller (2000a) usesprobit models for arbitration-eligible and freeagent players to correct for any selectivity biasin salary regressions.

This paper has reviewed the Farber (1980)model in this section. It discusses the Miller(2000a) model in Section 4B. Both serve asstarting points for the research presented here.Recall that the purpose of this paper is to ana-lyze negotiated salary outcomes at the threeearly stages of MLB’s FOSA process. Farber(1980) focuses on how final offers are createdalong with the consequences for negotiated set-tlements. Miller (2000a) examines bargainingafter final offers have already been exchanged.None of these papers models the bargaining thattakes place before the exchange of final offers,however. Therefore, there are no implications

from these models that provide a hypothesisthat is being tested in this paper. The estimatedequation presented below would require a bar-gaining model that allows negotiations to occurbefore the exchange of final offers. Many bar-gaining models already exist that can be usefullyextended to include bargaining behavior beforethe exchange of final offers, and, specifically, thedecision to file for arbitration. While this typeof extension is beyond the scope of this paper,it would be a fruitful area for future research.7

IV. DATA, MODEL, AND EMPIRICALMETHODOLOGY

A. Data

The dataset for this study includes the salariesof all arbitration-eligible players from the 2007through 2010 seasons. The data also include pre-vious season and career performance statisticsfor these players.8 The data is disaggregated intotwo subgroups. The first subgroup contains hit-ters and the second subgroup contains pitchers.Recall that the purpose of this study is to inves-tigate the effect of advancement through eachof the first three stages of the FOSA process onplayers’ salaries. To this end, each subgroup isdelineated by the four stages; players who areeligible; players who file for arbitration; thosewho exchange salary offers; and those who pro-ceed through the entire arbitration process.

Table 1 provides summary statistics by sub-group and arbitration stage. The top panelof the table contains information for hitters,while the bottom panel is for pitchers. Eachstage of the FOSA process listed in Table 1is mutually exclusive. In other words, the col-umn labeled “Eligible” provides statistics forthose players who are eligible for arbitrationand do not file for arbitration; the column“Filed” contains statistics for those who file forarbitration and do not exchange salary offerswith the teams’ owners. The column labeled“Exchanged” displays statistics for those players

7. The authors thank an anonymous referee for thepoints made in this paragraph.

8. The salary data comes from the USA Today onlinedatabase found at http://content.usatoday.com/sportsdata/baseball/mlb/salaries/team. Most data on service time andarbitration eligibility come from the Cot’s Contract web-site at http://mlbcontracts.blogspot.com/. Some biographi-cal data, arbitration eligibility data, and service time dataon players are from the Baseball Reference database athttp://www.baseball-reference.com/. Performance data andsome biographical data come from the Baseball Almanac athttp://www.baseball-almanac.com/.


TABLE 1Descriptive Statistics by Player Type and Eligibility Status

Hitters

Eligible Filed Exchanged Arbitrated

Previous salary 772,491 700,595 637,329 578,874Current salary 925,868 1,249,929 1,556,606 1,969,986Service time 4.03 3.55 3.44 3.53Team attendance 2,443,523 2,538,231 2,604,545 2,135,266Team winning percentage 0.49 0.51 0.51 0.51At bats 316.37 365.33 490.00 471.73Career at bats 1404.92 1401.84 1540.56 1616.91Slugging average 0.39 0.43 0.45 0.43Career slugging average 0.41 0.42 0.44 0.44On base percentage 0.32 0.33 0.35 0.34Career on base percentage 0.33 0.33 0.34 0.34Sample sizes 51 127 77 11

Pitchers

Eligible Filed Exchanged Arbitrated

Previous salary 651,451 606,065 614,656 912,857Current salary 957,774 1,054,612 1,304,721 1,832,635Service time 3.89 3.80 3.72 3.74Team attendance 2,435,287 2,536,583 2,614,206 2,697,242Team winning percentage 0.50 0.50 0.50 0.47Saves 3.31 4.67 3.67 12.45Innings pitched 75.61 87.25 121.18 95.37Career innings pitched 365.97 380.94 445.37 400.89Earned run average 4.73 4.10 3.80 3.57Career earned run average 4.31 4.12 4.05 3.81Relief pitcher 36 111 48 6Sample sizes 54 153 83 11

Notes: The categories listed in the column are mutually exclusive. In other words, the category Eligible indicates thatthe players are eligible for arbitration but did not file. Those who filed for arbitration did not exchange offers. Those whoexchanged offers did not go through the arbitration process.

Source: Authors’ calculations taken from the data. See text for discussion.

who exchange salary offers and do not pro-ceed through the arbitration process; “Arbi-trated” provides calculations for those playerswho proceed through the entire FOSA process.9

Finally, the first column of the table lists thevariables of interest.

Table 1 shows that for both hitters and pitch-ers, the previous season’s salary is lower andthe current season’s salary is higher as playersproceed through each stage of final-offer arbitra-tion. This provides transient evidence that thereis a premium associated with each FOSA stage.Table 1 also shows that those who receive moreplaying time are more likely to proceed through

9. Although this paper does not concentrate on arbitratedsalaries, the descriptive statistics are provided here for com-pleteness. Only 22 players proceed through the arbitrationstage. Therefore, any econometrics performed on this sub-group would be very imprecise.

each stage of the arbitration process. This istrue for hitters and pitchers as evidenced by theincreases in at bats and career at bats for hittersand innings pitched and career innings pitchedfor pitchers. Hitters and pitchers with less ser-vice time appear to be more likely to proceedthrough each stage of the arbitration process.Finally, better players are more likely to enterinto the later stages of the FOSA process asevidenced by the changes in the performancestatistics displayed in Table 1. These descrip-tive statistics indicate that there is selection,potentially, into the various stages of the FOSAprocess.

B. Model and Methodology: Treatment Effectsof Filing and Exchanging Offers

As stated previously the underlying model forthe previous work in this area is the theoretical


model developed by Farber (1980) in whicha contract zone is developed based on theobjective salary for the player estimated by aneutral, unbiased arbitrator, and the final offersof the player and the club. Miller (2000a)offers an approach that better fits the focusof this research but it must be adapted toallow for negotiated settlements both beforeand after the exchange of final offers. Consideran intertemporal version of the Miller (2000a)cooperative bargaining model in which Nashequilibrium is found from

y∗ns = arg maxyns(Ucs(1 − yns) − dcs)(4)

×(Ups(yns) − dps),

where yns is the salary reached through negotia-tion in stage s of the FOSA process. The utilityof the club in step s, Ucs, and the utility of theplayer in step s, Ups, are a function of the bar-gained salary and the respective disagreementoutcomes in step s, dcs and dps, the expectedutility of proceeding to the next step, s +1, inthe arbitration process. Let ycs and yps be thesalary offers of the club and player respectivelyin stage s; these offers are unobservable in theearly stages of the process but represent the finaloffers in the latter stages of the process. Fromthe disagreement functions10 and Equation (4)above, the first-order condition defines the func-tion for the negotiated salary

y∗ns = y∗

ns[dcs(ycs, yps)dps(ycs, yps)].(5)

Miller (2000a) notes how Equation (5) is anincreasing function in the disagreement pointfor players and a decreasing function in thedisagreement point for clubs. Recall that thereare three stages under scrutiny in this research:negotiated settlements for eligible players whodo not file for arbitration, negotiated settlementsfor players who file for arbitration but settlebefore an exchange of offers, and negotiatedsettlements for players who file and exchangeoffers but settle without arbitration. Therefore,if there is an increase in negotiated salary duringeach subsequent stage in the bargaining process,then it can be interpreted that the arbitrationprocess is either increasing the disagreementpoint for players or decreasing the disagreementpoint for the clubs.

Obviously, the wage offers of the club andthe player are a function of the player’s past

10. These are not shown but can be found in Miller(2000a), 42.

performance and years of experience. Asymmet-ric information may cause a divergence betweenthe player and club offers. These differencesmay shrink as more information is exchangedbetween parties in subsequent stages of the pro-cess. The willingness to assume risk in the nego-tiation process will shape offers as well. It seemslikely that both sides may be willing to assumemore risk in the earlier stages of the processbefore the exchange of final offers. The act ofmaking a final offer leaves either party vulnera-ble to an adverse decision should the opposingside opt for arbitration. The costs of negoti-ation, both real and psychic, can also play arole in the utility maximization of each side.Real costs of negotiation may be minimal dur-ing early stages where offers can be exchangedvia phone or fax. Costs of an arbitrated decisionare borne equally by parties and include air-fare and/or hotel stays for the arbitrator, expertwitness, and/or lawyers/agents. Psychic costsinclude the psychological stress for players notknowing where they and their families may beliving next season. For club general managersand coaches there is stress from dealing withplayers disgruntled by the negotiation process.

The cost of disagreement favors managementin the pre-filing stage of the process as playersmust accept a contract from management if theydo not file for arbitration. This may cause betterplayers to advance to the filing stage to increasetheir bargaining power, that is, their disagree-ment point. Marburger and Scoggins (1996) findthat higher quality players are more likely to filefor arbitration and press for an arbitrated set-tlement. After filing, but before the exchangeof final offers, it is unclear which side has anadvantage. Once final offers are exchanged, bothsides face the prospect of an adverse decision bythe arbitration panel. Perhaps better players feelthat they have an advantage because clubs maywant to keep them happy so their performanceis not negatively affected by the rancor that cansurround negotiations. This implies that thereshould be a relatively higher salary premiumassociated with filing for arbitration instead offor exchanging final offers.

Recall that the purpose of this paper is to ana-lyze how the process of moving to each subse-quent stage of MLB’s FOSA process influencesthe salary of players. As suggested in Table 1,there appears to be a salary premium associ-ated with moving to each stage. Therefore, itis logical to think of the process of filing forarbitration, or the act of officially exchanging


TABLE 2List of Variables for Regressions

Common Variables Hitters Only Pitchers Only

Log of previous salary At bats SavesService time Career at bats divided by service time Innings pitchedTeam attendance (1,000,000) Slugging average Career innings pitched divided by service

timeTeam winning percentage Career slugging average Earned run averageYear dummy variables On base percentage Career earned run averageFirst time eligibility dummy Career on base percentage Relief dummyMultiyear contract dummy First base dummy Relief dummy × innings pitched

Second base dummy Relief dummy × (career inningspitched/service time)

Third base dummy Relief dummy × savesCatcher dummyShort stop dummy

offers with the team as a type of program inwhich players participate, and these various pro-grams should influence salaries in one way oranother. The empirical methodology will esti-mate the average treatment effects on players’salaries of moving to each subsequent stage inthe FOSA process. Table 1 also shows that bet-ter players appear to be more likely to move onto each stage. Therefore, the estimated equationsneed to account for selection into the partic-ipation of each program, that is, filing and/orexchanging final offers. If selection on observ-able characteristics is assumed, then a stan-dard model for estimating the average treatmenteffects is the following:

ln(yit ) = β0 + β1Dit−1 + β2xit−1(6)

+ β3Dit−1(xit−1 − x) + uit .

In Equation (6), ln(yit ) is the natural log ofplayer i’s salary in season t . The xit−1 containsa set of previous season performance character-istics, team-specific variables, and dummy vari-ables for position played, year, a player beingeligible for arbitration for the first time, and anegotiated salary resulting in a multi-year con-tract. The full set of variables contained in xit−1is in Table 2.

Equation (6) is estimated separately for hit-ters and pitchers. For each sample, Equation (6)is estimated twice. The first time, the sampleused includes players who are eligible for arbi-tration and those who filed for it and did notofficially exchange salary offers with their team.The second time uses the sample of playerswho filed for arbitration and exchanged salaryoffers but did not move to the final stage of

the arbitration process. Given that only 22 play-ers (11 hitters and 11 pitchers) actually movedto the arbitration stage, they are excluded fromthe analysis sample. The variable Dit−1 is adummy variable equaling 1 if the player filedfor arbitration (for the first estimation) or offi-cially exchanged a salary offer with his team(during the second estimation) between seasonst − 1 and t . The estimate of β1 is the averagetreatment effect for moving to each subsequentstage in the arbitration process. The interactionsbetween Dit−1 and the de-meaned variables con-tained in Table 2 help to control for selection onobservable characteristics.11

V. RESULTS

The results from Equation (4) are presentedin Tables 3 and 4. Table 3 presents the resultsfor hitters, and Table 4 is for pitchers. The firstcolumn in each table contains the independentvariables used in the analysis. The second andthird columns present the parameter estimatesby different subsamples of the data. Model1 uses the sample of players who did notexchange salary offers. Put another way, Model1 focuses on the movement from being eligibleto filing for arbitration but not exchanging salaryoffers. Model 2 uses the sample of playerswho file for arbitration and do not go all ofthe way through the arbitration process. Inother words, Model 2 examines the movementfrom filing for arbitration to exchanging salaryoffers.

11. See Wooldridge (2002) for a complete discussion ofthis model.


TABLE 3Treatment Effects of Filing and/or Exchanging

on Salary—Hitters

Sample

NoExchangeModel 1

File andExchangeModel 2

Log of previous salary 0.218 0.245(1.70)∗ (4.91)∗∗∗

Service time 0.081 0.174(0.89) (4.15)∗∗∗

Team attendance (1,000,000) −0.069 −0.045(1.08) (1.37)

Winning percentage 1.063 0.575(1.91)∗ (1.70)∗

At bats 0.002 0.002(3.27)∗∗∗ (8.07)∗∗∗

Career at bats/service time 0.001 0.002(2.05)∗∗ (7.24)∗∗∗

Slugging average −2.049 0.751(1.56) (1.00)

Career slugging average 7.092 3.174(4.65)∗∗∗ (2.75)∗∗

On base percentage 3.447 1.261(2.50)∗∗ (0.89)

Career on base percentage −3.162 −1.985(1.50) (0.94)

First base 0.423 −0.075(2.11)∗∗ (0.55)

Second base 0.047 −0.092(0.48) (1.45)

Third base −0.214 0.050(2.18)∗∗ (0.66)

Catcher 0.019 0.137(0.13) (2.51)∗∗

Short stop −0.106 0.017(1.01) (0.25)

First time eligible −0.012 0.048(0.10) (0.76)

Multiyear contract 0.023 −0.111(0.29) (1.18)

Filed 0.223 —(3.54)∗∗∗ —

Exchange — 0.072— (2.06)∗∗

Observations 178 204R2 0.93 0.94

Notes: Robust t statistics in parentheses. Standarderrors clustered at the team level used in the calculations.All regressions include year dummy variables and inter-actions between the treatment dummy variable and thedemeaned independent variables listed in Table 2. See textfor discussion.

∗Significant at 10%; ∗∗significant at 5%; ∗∗∗significantat 1%.

Before analyzing the average treatment eff-ects, the coefficients associated with the vari-ables found in Table 2 are discussed. Results

TABLE 4Treatment Effects of Filing and/or Exchanging

on Salary—Pitchers

Sample

NoExchangeModel 1

File andExchangeModel 2

Log of previous salary 0.323 0.346(3.01)∗∗∗ (5.89)∗∗∗

Service time 0.163 0.111(2.66)∗∗ (4.66)∗∗∗

Team attendance(1,000,000)

−0.005 −0.019(0.07) (0.49)

Winning percentage 0.860 −0.657(1.20) (1.21)

Saves −0.019 −0.019(0.09) (0.10)

Innings pitched 0.007 0.006(3.66)∗∗∗ (4.48)∗∗∗

Career inningspitched/service time

4.45E-04 0.002(0.19) (1.85)∗

Earned run average −0.002 −0.069(0.18) (3.91)∗∗∗

Career earned runaverage

−0.132 −0.130(1.72)∗ (4.26)∗∗∗

Relief dummy −0.082 0.414(0.38) (1.91)∗

Relief × saves 0.043 0.046(0.22) (0.24)

Relief × innings pitched −0.003 −0.003(1.10) (1.43)

Relief × (career inningspitched/service time)

0.003 −0.002(0.98) (1.17)

First time eligible −0.025 0.015(0.30) (0.19)

Multiyear contract 0.152 0.030(0.60) (0.35)

Filed 0.134 —(2.16)∗∗ —

Exchanged — −0.028— (0.82)

Observations 207 236R2 0.88 0.88

Notes: Robust t statistics in parentheses. Standarderrors clustered at the team level used in the calculations.All regressions include year dummy variables and inter-actions between the treatment dummy variable and thedemeaned independent variables listed in Table 2. See textfor discussion.

∗Significant at 10%; ∗∗significant at 5%; ∗∗∗significantat 1%.

in Tables 3 and 4 show some similaritieswith regards to pay between hitters and pitch-ers. Service time and the previous season’ssalary appear to influence positively the cur-rent season’s salary. Additionally, those playerswho receive more playing time receive higher


salaries. This can be seen through the positiveand significant coefficients associated with atbats for hitters and innings pitched for pitch-ers. As expected, better performance also resultsin higher negotiated salaries as evidenced bythe significant coefficients associated with careerslugging average for hitters and career earnedrun average for pitchers. Interestingly, beingeligible for arbitration for the first time andsalary negotiations resulting in multi-year con-tracts have no discernible impact on players’current salaries. One difference between hittersand pitchers in Tables 3 and 4 is the effect ofteam variables on salaries. The team’s previ-ous winning percentage positively affects cur-rent season’s salary for hitters but not pitchers.

As Table 1 suggests, there may be a salarypremium for those players who proceed througheach stage of the FOSA process. Table 3presents estimates of the average treatmenteffect for hitters who move to each subsequentstage. Focusing on Model 1, Table 3 indicatesthat there is a positive and highly significantgain for hitters who file for arbitration but donot exchange salaries with their clubs. Thosewho file and do not exchange offers increasetheir salary by 25%.12 Model 2 indicates thatthose who move from filing to exchanging offersreceive a salary premium of 7%, which is sta-tistically significant at the 5% level. Therefore,the conclusion presented by these results is thatfiling for arbitration substantially increases aplayer’s salary when compared to simply beingeligible for it. Furthermore, exchanging finaloffers does provide an additional increase insalary. It is not nearly as large, however, as theinitial increase from filing for arbitration.

Table 4 presents the treatment effects forpitchers. Focusing on Model 1, the results showthat those pitchers who move from eligibilityto filing and not exchanging offers increasetheir salary by 14%. This treatment effect isstatistically significant; however, it is muchsmaller in magnitude when compared with theresults in Table 3 for hitters. Results in Model2 show that there is a negative and statisticallyinsignificant relationship for moving from filingto exchanging offers. Combined, these resultsimply that there is a premium for filing forarbitration as opposed to being eligible, andthere is no salary premium for exchanging offersonce a pitcher has already filed for arbitration.

12. The exact percentage changes come from the for-mula eβ̂ − 1.

The results in Table 4 are different from thosefound in Table 3 for hitters. Hitters receivesignificant salary gains when moving to eachstage of the arbitration process, and the gain islarger when filing for arbitration. Pitchers, onthe other hand, only receive a gain in salaryfor filing, and this gain is approximately 11percentage points lower than that for hitters.This finding suggests that hitters and pitchersare treated differently during the negotiationprocess.

Overall, the empirical results suggest thatboth hitters and pitchers can improve theirsalaries by filing for arbitration rather than justbeing eligible to do so. It is possible thatfiling for arbitration spurs owners into moregood faith negotiations through the threat ofpossibly moving forward to the final stage of thearbitration process, which is costly.13 This resultis somewhat surprising. Why would playersnot file for arbitration, a relatively costlessprocess, if it meant an increase in salary?Previous studies have not analyzed separatelythe filing phase of the FOSA process. Perhapsthe adjustment for selectivity bias used here isnot accounting for all of the potential selection.Instead, the result could be capturing the effectof better players filing for arbitration whilelesser players do not. However, different setsof explanatory variables and different methodsfor accounting for the potential selection intoeach stage of the arbitration process have beenused. The quantitative and qualitative results aresimilar and available upon request. Therefore,the results presented in Tables 3 and 4 are robustto different empirical specifications.

A potential explanation for the large salarypremium associated with filing and the small/nopremium associated with exchanging offers forhitters/pitchers, respectively, is that there is onlya short amount of time between these stages,typically 3 days. Perhaps the rush to finishnegotiations before the official exchange ofoffers, threat of arbitration, and the cost incurredby such, are responsible for the larger bumpin salary for filing as opposed to exchanging.Once offers are exchanged, valuable informationis then available for all participants in theprocess to see and analyze. There is a longertimeframe in which to continue negotiationsbefore arbitration hearings are held, typically 2weeks to a month.

13. The authors thank an anonymous referee for makingthis suggestion.


The size of the salary increases in the fil-ing stage and smaller and/or lack of significantincreases for the exchange phase suggest that theprocess is starting to push agreement to resolu-tion much earlier. Ashenfelter and Dahl (2005)analyze settlements determined by final-offerarbitration for wages of police and firefightersin New Jersey. The authors conclude that indi-viduals who use experts during the arbitrationprocess fair better than those who do not, andknowledge of this benefit develops over time.The extensive use of agents by players duringthe negotiation process in MLB and the historicinstitutional knowledge created over the yearsmay be finally creating the intended result ofthe process design.

The difference in the size of the salaryincreases for hitters versus pitchers is not sur-prising. Pitchers are more susceptible to careershortening/ending shoulder or arm injuries. Thisrisk creates greater variability in their perfor-mances and lessens the expected value for thefuture from good performances in the past.Miller (2000a) also found differences in risk-assuming behavior by negotiators for positionplayers versus starting pitchers.

VI. CONCLUSIONS

The number of salaries actually determinedby arbitration in MLB has declined in recentyears, suggesting that there has been a shiftin the use of the FOSA system even thoughthere has been no change in the operation ofthe system. Because of this, the focus of thispaper is on the salary determination processat the three initial stages of the FOSA pro-cess. Treatment effects are estimated for filingfor arbitration and exchanging offers. Resultsindicate that, among hitters who become eli-gible for arbitration, those who file for arbi-tration receive a wage premium of 25%; theexchange of offers increases salaries by 7%.For pitchers, there is a salary premium of 14%for filing for arbitration and no premium asso-ciated with exchanging offers. The differencein findings for hitters and pitchers suggests thatthey are treated differently during the negotia-tion process.

This paper methodologically contributes tothe literature in a number of ways. This is thefirst paper on MLB to use information from themost recent seasons, 2007 to 2010, a time periodin which salary negotiation cases that were ulti-mately determined by arbitration were at historic

lows. Additionally, the analysis makes use ofstatistics on service time as opposed to expe-rience. This is important since the CBA statesthat it is service time, as opposed to experience,that dictates eligibility for arbitration. Finally,this is the first paper to focus explicitly on thesalary determination process of players in eachof the first three stages of the arbitration process:being eligible for arbitration; filing for arbitra-tion and not exchanging offers; and exchangingoffers and not continuing through to arbitration.

The results of the analysis offer interestingconclusions that the previous literature has notshown. There is a salary premium for both hit-ters and pitchers for filing for arbitration; how-ever, there is only a premium for exchangingoffers for hitters, and this premium is substan-tially smaller than that for filing for arbitration.The previous literature has not shown this. Thisparticular result is similar to the literature on freeagency, which has shown that the threat of freeagency seems to increase salaries, while goingto the free agent market does not.

Given the rise in the use of final-offer arbitra-tion, particularly in the public sector, the resultsfrom this analysis may shed some light on theconstruct of the final-offer arbitration processes.The analysis does not necessarily give insightinto the evolution of the FOSA system usedin MLB. However, the limited number of casesthat proceed to the arbitration stage in the timeperiod of this study and the finding that playersbenefit significantly, on average, from filing forarbitration suggests that the design of this sys-tem may be finally achieving the overall goal offinal-offer arbitration to encourage cooperativebargaining.

REFERENCES

Ashenfelter, O., and G. B. Dahl. “Strategic Behavior, Self-Serving Biases, and the Role of Expert Agents: AnEmpirical Study of Final-Offer Arbitration.” NationalBureau of Economic Research, Inc., NBER WorkingPapers: 11189, 2005.

Brown, D. T., and C. R. Link. “Final Offer Arbitration inMajor League Baseball: An Empirical Analysis of Bar-gaining Failure.” Unpublished Manuscript, Universityof Delaware, 2010.

Burgess, P. L., and D. R. Marburger. “Do Negotiated andArbitrated Salaries Differ Under Final-Offer Arbi-tration?” Industrial and Labor Relations Review, 46,1993, 548–58.

Farber, H. S. “An Analysis of Final Offer Arbitration.”Journal of Conflict Resolution, 24(4), 1980, 683–705.

Farmer, A., P. Pecorin, and V. Stango. “The Causes ofBargaining Failure: Evidence from Major LeagueBaseball.” Journal of Law and Economics, 47, 2004,543–68.


Faurot, D. J., and S. McAllister. “Salary Arbitration andPre-Arbitration Negotiation in Major League Base-ball.” Industrial and Labor Relations Review, 45, 1992,697–710.

Fizel, J. “Bias in Salary Arbitration: The Case of MajorLeague Baseball.” Applied Economics, 28(2), 1996,255–65.

Hadley, L., and J. Ruggiero. “Final-Offer Arbitration inMajor League Baseball: A Nonparametric Analy-sis.” Annals of Operations Research, 145, 2006,201–09.

Marburger, D. R. “Arbitrator Compromise in Final OfferArbitration: Evidence from Major League Baseball.”Economic Inquiry, 42, 2004, 60–68.

Marburger, D. R., and P. L. Burgess. “Can Prior Offers andArbitration Outcomes Be Used to Predict the Winners

of Subsequent Final-Offer Arbitration Cases?” South-ern Economic Journal, 71(1), 2004, 93–102.

Marburger, D. R., and J. F. Scoggins. “Risk and Final OfferArbitration Usage Rates: Evidence from Major LeagueBaseball.” Journal of Labor Research, 17(4), 1996,735–45.

Miller, P. “An Analysis of Final Offers Chosen in Baseball’sArbitration System: The Effect of Pre-ArbitrationNegotiation on the Choice of Final Offers.” Journalof Sports Economics, 1, 2000a, 39–55.

. “A Theoretical and Empirical Comparison of FreeAgent and Arbitration-Eligible Salaries Negotiated inMajor League Baseball.” Southern Economic Journal,67(1), 2000b, 87–104.

Wooldridge, J. M. Econometric Analysis of Cross Sectionand Panel Data. Cambridge, MA: MIT Press, 2002.

Documents

NEGOTIATED SETTLEMENT UNDER MLB FINAL-OFFER SALARY ARBITRATION SYSTEM