® Academy of Management Journal 1999, Vol. 42, No. 1, 25-40.

THE PERFORMANCE EFFECTS OF PAY DISPERSION ON INDIVIDUALS AND ORGANIZATIONS
MATT BLOOM University of Notre Dame Pay distribution research is relatively scarce in the compensation literature, yet pay distributions are viewed as critically important by organizational decision makers. This study is a direct test of the relationship between one form of pay distribution—pay dispersion—and performance conducted in a field setting where individual and organizational performance could be reliably observed and measured. Findings suggest more compressed pay dispersions are positively related to multiple measures of individual and organizational performance. Among those who design and administer compensation systems, pay distributions have been an important issue for a long time, yet they have been studied relatively infrequently in the compensation research literature (Gomez-Mejia & Balkin, 1992; Meyer, 1975; O'Reilly, Main, & Crystal, 1988). The term pay distributions refers to the "array of compensation levels paid for differences in work responsibilities, human capital, or individual performance within a single organization" (Milkovich & Newman, 1996: 45). Although there seems to he agreement that pay distributions influence individual and organizational performance, there is little agreement over how or why they matter (Gerhart & Milkovich, 1992). In fact, there has been disagreement in the theoretical literature about the proper distribution of pay for some time: Should pay distributions he compressed and egalitarian, or should they he hierarchical and consecutively increasing like prizes in a golf tournament (Lazear & Rosen, 1981; Mahoney, 1979]? This discussion about what constitutes an equitable distribution of pay is of crucial importance to organizations since, as Thurow put it, "their profits depend upon it" (1975: 106). Indeed, pay distribution decisions have been asserted to have meaningful effects on employee staffing (Livernash, 1957; Rynes, 1992), attitudes (Heneman, 1985), and performance (Becker & Huselid, 1992; Ehrenberg & Bognanno, 1990; Levine, 1993; Pfeffer & Langton, 1993). Recent interest in high-performance work systems (Kochan & Osterman, 1994; Pfeffer, 1994) and concerns ahout the allocation of pay in companies (Bok, 1993; Frank & Cook, 1997) have added emotion and ideology to the debate, yet empirical research is just beginning to inform the controversy (Becker & Huselid, 1992; Ehrenherg & Bognanno, 1990; Pfeffer & Langton, 1993). Although it has contributed substantially to the compensation literature, most existing research
25

has heen limited to situations in which performance occurs over a very short period of time and is primarily dependent upon a single individual (Becker & Huselid, 1992; Ehrenherg & Bognanno, 1990) or a very small group of individuals (Pfeffer & Langton, 1993). Furthermore, published research is virtually silent ahout the relationships hetween pay distributions and organizational performance. Since compensation policy makers ultimately design pay distrihutions to influence organizational success, this is an important area for empirical research. This study was a direct test of the pay distribution-performance relationship in a field setting where individual and organizational performance were ohservahle and could be reliahly measured over an extended period of time. CONCEPTUAL FOUNDATIONS OF PAY DISTRIBUTION MODELS In this study, pay levels (the absolute rates of pay assigned to individuals or jobs in an organization) are considered in terms of how equally they are distributed (their dispersion) in the organization. The central issue is understanding how differences in pay distrihutions influence individual and organizational performance. Two types of pay distributions are common. When a pay distribution is hierarchical, a greater proportion of pay is concentrated in relatively few levels, jobs, or individuals that are near the top of the distrihution. Pay is more widely dispersed and less equal across pay levels. A hierarchical distribution may also have many pay levels and many tiers. A compressed pay distrihution is one in which pay is less dispersed and is spread more equally across johs or individuals, and it may have fewer pay levels than a hierarchical distrihution. The focus of this study was on the spread of pay, or differences in actual pay, not on the number of pay levels.

26

Academy of Management Journal

February

Decisions ahout the equality of intraorganizational pay distributions—whether they will he hierarchical or compressed—are important to organizations and employees. For example, in major league hasehall, Alhert Belle recently signed an $11 million contract with the Chicago White Sox, an award that management presumably hoped will increase the team's performance. Framed in terms of pay dispersion, the important questions are whether Belle's contrihutions to his team will he worth $3.5 million more than those of the next highest paid player on the team (Frank Thomas, whose annual compensation was $7.5 million) and what effects making the distribution of pay on the team more dispersed (unequal) will have on the remaining players and the team's overall success. Alhert Belle's new contract was a large increase in the maximum paid to a White Sox player and created a more hierarchical (dispersed) pay distrihution on the team. In some organizations (e.g.. Fortune 500 companies) the pay ratios of highest-tolowest paid employees can exceed 200:1 (Reingold, 1997); in others (for instance, some law firms and "Big Six" accounting firms) the ratio is closer to 12:1 (White & Associates, 1995). For organizational decision makers, a key question is whether creating a more hierarchical or a more compressed pay distrihution hy paying relatively more or less for top talent has implications for employee and organizational performance. A consistent and long-standing theme in the compensation literature has heen that to induce performance, pay levels must he not only high enough in an absolute sense, hut also must he different enough to reflect the differences in contrihution, human capital, and effort that exist in an organization (Livernash, 1975). Hamilton and Macy distinguished hetween "uniform" and "divergent" pay distrihutions, asserting that intraorganizational differences in pay must reward "excess ahility, knowledge, skill, training, diligence, or whatnot, possessed hy its recipient over the common laborer" (1923: 115). Similarly, Riegel asserted that "wage differentials^ are needed to induce individuals to prepare for and fulfill exacting types of work" (1937: 118). The idea that relative worth is the proper basis for determining pay levels has carried into contemporary views of pay distrihutions. Since an organization has limited compensation resources, the distrihution of pay w^ithin it is inherently a zero-sum matter: Each employee's pay

necessarily limits every other employees' pay. Pay distrihutions are important motivational mechanisms hecause they influence the effort of all employees in an organization (Lazear, 1995; Mahoney, 1979; O'Reilly et al., 1988; Pfeffer, 1994). However, contemporary models offer divergent explanations of how pay levels should he distrihuted and how different pay distrihutions will affect individual and organizational performance. In the hierarchical model, the use of unequal rewards to induce greater individual effort and performance is stressed, hut in the compressed model, the use of pay equalization to motivate cooperation and higher levels of group performance is stressed. At the core of this issue are notions ahout comparative pay and the degree of compensation inequality created hy an organization's pay distrihution. For many years, compensation policy makers have viewed pay compression as a potential negative hecause it may underreward people who have higher-level johs, greater human capital, or higher performance than others (Milkovich & Newman, 1996). Frank (1985) asserted that, as a "positional good," pay affects a person's organizational status and prestige. Thus, more compressed distrihutions may also decrease the positional value of pay. More recently, some writers have proposed that compressed pay distrihutions can he heneficial for group performance hecause they may inculcate feelings of fairness and common purpose, foster cooperative, team-oriented behavior, and support common goal orientations (Cowherd & Levine, 1992; Lazear, 1995; Pfeffer, 1994). These divergent views do not fully inform organizational decision makers ahout how to halance the trade-offs hetween attracting and motivating top talent using a hierarchical pay distrihution and inducing cooperative effort through using a compressed distrihution. These issues are discussed in the sections that follow. Pay Distributions and Individual Performance Under the hierarchical model, pay distrihutions are viewed as networks of incentives (cf. Milkovich & Newman, 1996) that attract talent and motivate individual performance (Milgrom & Roherts, 1992). Hierarchical pay distrihutions are expected to induce higher employee performance hecause they create a meritocracy in which rewards for effort or human capital increase monotonically (Heneman, 1992; Zenger, 1992). A wide spread hetween pay levels (i.e., greater pay dispersion) increases the returns for higher performance, thus creating a positive pay-performance link and inducing higher future performance (Milgrom & Roherts, 1992). Greater pay dispersion is also asserted to motivate

^ The term "wage differential" also refers to the dispersion of pay levels. A pay distribution is more hierarchical when differentials are wider.

1999

Bloom

27

employees whose performance is lower who wish to receive higger pieces of the compensation pie (Milgrom & Roherts, 1992). Because compressed distrihutions create the prohlem of "topping out," where strong performers reach the maximum allowahle pay level, these employees are expected to seek out and remain at organizations with more hierarchical pay distrihutions (Ehrenherg & Smith, 1994). Some research supports this model. Ehrenberg and Bognanno (1990) found that hierarchical prize distrihutions were associated with better player performance in a sample drawn from European professional golf tour data. In a study of auto racing, Becker and Huselid (1992) found that hierarchical distrihutions were positively related to race outcomes. Larger spreads hetween finishing place prizes were associated with better individual racer performance. In summary, according to the hierarchical model, greater dispersion in pay levels within an organization will lead to higher individual performance: Hypothesis la. Greater pay dispersion will be positively related to individual performance. The compressed model focuses on how pay distrihutions affect collahorative work and cooperation. From this perspective, hierarchical pay distrihutions create "disincentives" for cooperation, instill feelings of inequity, promote dissatisfaction, and diminish performance (Kohn, 1993; Pfeffer, 1994). Some research suggests that more dispersed pay distrihutions are associated with greater dissatisfaction, poorerquality work, and increased propensity to leave an organization (Cappelli & Sherer, 1990; Cowherd & Levine, 1992; Pfeffer & Langton, 1993). Since hierarchical distributions may overcompensate those near the top of a pay distrihution, they may give the incumhents of these top positions an unfair advantage in garnering more of an organization's limited compensation resources. Sometimes called the Matthew effect (cf. Merton, 1973)—more is given to those that have, and more is taken away from those that have not—this condition is likely to he considered unfair and "demotivating" hy all employees except those at the top (Kochan & Osterman, 1994; Pfeffer, 1994). Such a situation is especially problematic hecause, hy definition, there will always he more have-nots than haves in a hierarchical pay distrihution. At the extreme, these effects may lead to what Frank and Gook (1997) called winner-take-all contests, which, they argued, are both costly and unproductive. Gompressed pay distrihutions, conversely, are said to inculcate feelings of fairness, foster notions of a common fate, and reduce interpersonal competition (Kochan & Osterman, 1994; Milgrom & Roherts, 1988). Thus, according to the compressed model.

greater dispersion in pay levels within an organization will lead to lower individual performance: Hypothesis lb. Greater pay dispersion will be negatively related to individual performance. Moderating Effects of Pay Distributions These contrasting predictions hased on the hierarchical and compressed models do not provide clear guidance ahout what distrihution is optimal, save implying that more hierarchical or more compressed is better per se. It may he that the answer lies somewhere hetween the two views: the effects of pay dispersion may he moderated hy an employee's position in a pay dispersion. More dispersed pay may he positively related to the performance of those whose pay is nearer the top of a dispersion and negatively related to the performance of those whose pay is nearer the hottom. Frank (1985) asserted that, since pay is a positional good, large differences in pay confer higher status and prestige as well as greater economic henefits. Thus, people near the top should respond favorably to hierarchical pay distrihutions. Conversely, research indicates that when reward distrihutions are unequal, the disadvantaged (those low in the distrihution) will feel deprived and react negatively (Martin, 1981). Given that organizations have limited compensation resources, what is distrihuted at the top is not availahle for those near the hottom, so the lower one is in the distrihution, the lower the prestige, status, and economic henefits conferred hy pay. This formulation suggests that people lower in a pay distrihution will respond unfavorahly to more hierarchical pay dispersion (Gowherd & Levine, 1992; Pfeffer & Langton, 1993). Hypothesis 2. Greater pay dispersion will be positively related to the performance of people higher in a pay distribution and negatively related to the performance of people lower in the pay distribution. Pay Distributions and Organizational Performance Neoclassical economics offers a possihle framework for understanding the relationship hetween pay dispersion and organizational performance (Lazear, 1995; Milgrom & Roherts, 1992). The effort of an organization's employees is viewed as integral to producing the organization's product or service; without the right kind of effort from employees, the organization will have no output to sell (Pindyck & Ruhinfeld, 1995). When properly structured, wages influence employee effort, which in

28

Academy of Management Journal

February

turn influences the organization's output (Gerhart & Milkovich, 1992; Lawler, 1990; Lawler & Jenkins, 1992). According to both the hierarchical and egalitarian views, the distrihution of wages—over and above wage levels—influences hoth the amount and type of effort employees exert. According to the hierarchical view, individuals throughout an organization are motivated to higher performance hy greater pay dispersion; they will work harder to ohtain the prize of high pay (Milgrom & Roherts, 1992). Organizations should henefit from the comhined effort of high performers and, hence, organizational performance (productivity and financial outcomes) should improve commensurahly. However, in organizations with strong work interdependencies, where output depends on collahorative effort, hierarchical pay distrihutions may be detrimental to organizational performance hecause they discourage cooperation (Deutsch, 1985). One reason is that individuals will concentrate only on their own performance—to the exclusion of organizational goals—since their own performance is what matters for moving up in the pay distrihution. People may also engage in deleterious activities like sahotaging the work of other employees in an effort to garner more of the organization's compensation resources (Lazear, 1995). Gompressed pay distrihutions should not only prevent these negative effects, hut also should lead to higher organizational performance by fostering teamwork and creating loyalty and an orientation toward a common goal. Pfeffer argued that "teamwork is fostered hy common fate, and common fate is enhanced to the extent that people in an organization fare comparahly in terms of rewards received" (1994: 50). Where work has strong interdependencies and organizational objectives are dependent upon cooperative effort, less dispersion in pay distrihutions may lead to higher organizational performance, since it not only induces individuals to work harder, hut also causes them to work harder together toward an organization's goals. Hypothesis 3. Greater pay dispersion will be negatively related to organizational performance. METHODS Study Context and Sample Major league hasehall offers a relevant situation in which to study the pay distrihution-performance relationship for several reasons. Like the managements of most organizations, the owners and managers of hasehall teams face the need to optimize individual and group performance. Ge-

teris parihus, better individual player performance is expected to contrihute positively to team performance. Organizational performance in hasehall is also contingent upon work interdependencies (e.g., double plays, defending against hunts, scoring runs). Because major league hasehall teams are characterized hy the separation of ownership and control, motivating workers to exert effort on the owner's hehalf is an important managerial prohlem, and the implementation of an appropriate compensation system is a strategic decision. I used pay and performance information on 1,644 players on 29 teams for the years 1985 through 1993. Two of the teams (the Montreal Expos and the Toronto Blue Jays) were Ganadian. Gompensation data for individual players were not puhlicly availahle hefore 1985. I did not use data from after 1993 hecause of the protracted player strike of 1994. Player Performance Player performance data came from Total Baseball (Thorn & Palmer, 1994) and The Baseball Encyclopedia (Bucek, 1996). For players, I used performance metrics taken directly from the literature on hasehall statistics (Thorn & Palmer, 1994). I used three measures for nonpitchers—adjusted hatting runs, fielding runs, and total player rating—and three for pitchers: adjusted earned run average, pitching runs, and total pitcher rating. All of these measures are commonly used hy professional hasehall teams to judge player performance (Hart, 1995; Thorn & Palmer, 1994). Nonpitchers. A player's offensive performance is a combination of multiple factors, and a standard measure of offensive performance is adjusted hatting runs (Thorn & Palmer, 1994), a measure of how many points (runs) a player contrihuted to his team over the course of a year (season). Since points are positively related to team wins, a higher score indicates better player performance. Adjusted hatting runs is a composite of several different indicators intended to capture the variety of ways a player can contrihute points through his offensive performance and is defined as follows: Adjusted batting runs^ = [0.47 X + [0.78 X doublesijt] + [1.09 X + [1.40 X home runs,y,] + [0.33 X [bases on ballsjji + number of times hit as batsmanji,)] — [0.25 X [at batSijt - hitSjj,)] - [0.5 x outs on base,;,]. j

1999

Bloom

29

where i represents player; /, team; and t, year. The coefficients result from Palmer's analysis of all major league games over a 77-year period (Thorn & Palmer, 1994], Batting runs are normalized and adjusted for characteristics of home baseball parks (a home park is a stadium where a team plays half of its games) that are more (or less) favorable to a hitter's performance. Adjusted batting runs measures a player's impact on team performance in terms of runs (Thorn & Palmer, 1994). Thus, a player neither benefits from nor is penalized for the quality of his team. For example, the individual record of a player who bats for a particularly good team does not benefit because he plays for that team. Adjusted batting runs also provides a standard for comparisons of player performance across time periods and creates a common performance metric for the multiple years in the current data set (Thorn & Palmer, 1994). I measured defensive performance as fielding runs, which converts a player's defensive performance into run equivalents. This statistic takes efforts a player makes to prevent an opposing team from scoring runs and treats them as if they contributed points to his own team. Like adjusted batting runs, it is a composite of several different indicators of defensive performance. It is measured as follows (Thorn & Palmer, 1994): Fielding runSji, = {0.20 X + (2 X assistSjjt) — errorSijt + double j

points. The adjusted earned run average reflects how many of an opposing team's points result from hits made while a particular pitcher was performing. Thus, a higher value indicates poorer pitcher performance since it indicates opposing teams scored more points while the pitcher was pitching. The adjusted earned run average is adjusted for home baseball park because the very factors that make a particular baseball park njore (or less) favorable for batters have the opposite effect on pitchers. Pitching runs, the second measure of pitcher performance, is computed as follows: Pitching runs = innings pitched/9

X [individual earned run average — league-average earned run average). This statistic is analogous to the fielding runs statistic in that it treats efforts a pitcher makes to prevent opposing teams from scoring runs as points contributed to the pitcher's team. Higher scores indicate better performance. The final measure, total pitcher index, reflects a pitcher's overall performance relative to that of all other major league baseball pitchers in a specific year. It is a composite measure that measures overall pitching and fielding performance. Team Performance Two primary groups of performance goals are pursued by owners of professional sports teams: on-field performance and financial performance. Sheehan (1996) noted that it is possible for owners to pursue higher financial performance at the expense of on-field performance, and vice versa. To account for this, I used measures of both goals in this study. I measured on-field performance as the winning percentage [wins/games played), fan attendance [total home attendance/[stadium capacity X home gameSf)], and a team's finishing position [number of games behind the division leader the team was at season's end), which measures the team's performance relative to its most important competitors—division rivals. The higher this value, the poorer the performance. All team revenue and valuation data were obtained from Financial World. These data are not available for years prior to 1990. I used gate receipts, media income, total income, and franchise value as measures of financial performance. Gate receipts included all gate and club receipts but excluded sales taxes. Media income included national, local, and cable television and radio reve-

— fielding runsieague}^{[put-outsjt - strikeouts^] X [innings playedijt/innings playedjt]},

for player J on team ; in year t, where fielding runs^gQgug is the league average for fielding runs at the player's position. The coefficients again come from Palmer's 77-year analysis (Thorn & Palmer, 1994). Adjustments were made for a player's defensive position (e.g., catcher, first base). Like batting runs, fielding runs allows comparisons across time periods. Higher scores indicate better performance. Finally, I used total player rating, a composite of several measures that assess a player's overall offensive and defensive performance (Thorn & Palmer, 1994). This rating assesses a player's performance relative to the performance of all other nonpitchers in major league baseball in a specific year. Higher scores indicate better performance. Pitchers. For pitchers, I measured performance first with the adjusted earned run average, which is earned runs times nine divided by innings pitched. A pitcher's job is to prevent players from the opposing team from getting the hits they need to score

30

Academy of Management Journal

February

nues. Total income was the sum of gate, media, and stadium receipts. The stadium receipts component of total income included suites and luxury seat fees, concessions, parking fees, venue advertising, and licensing and merchandising income. A final measure of financial performance was the assessed value of a team franchise as computed by researchers at Financial World on the basis of total revenues, the team's playing venue, associated leases, and other information. This number represented the estimated value of a team at a season's end.

Independent Variables Player compensation. For this study, annual player pay information was taken from the newspaper USA Today. Each year, USA Today publishes the salaries and performance incentives of the members of all major league baseball teams' rosters. Most player compensation in major league baseball is in the form of contracted salaries, although minimal bonuses for reaching performance milestones do exist. Therefore, I measured player pay as the logarithm of the salary player i received in year t, exclusive of any performance incentives. Pay dispersion. I used four measures to capture pay dispersion on a team. The first, the gini coefficient, is commonly used in economics research to measure the degree of inequality in income distributions (Donaldson & Weymark, 1980). The team gini coefficient is defined as: 1 Team gini coefficient = 1 H
&

2 2^ n^y JJ

XI

(71 + 272 + ••• + " 7 ^ . where y^ . . . y^ is individual player salaries on team j arranged in decreasing order of size, y is the mean salary on team j , and n is the number of players on team /. A separate gini coefficient was computed for each team and each year. The maximum gini coefficient is unity, indicating the greatest degree of dispersion (absolute inequality), and the minimum is zero, indicating the lowest degree of dispersion (absolute equality). Organizations with higher gini coefficients have more hierarchical pay dispersions. I also used the coefficient of variation of pay (Pfeffer & Langton, 1993). A larger coefficient of variation indicates greater pay dispersion. The gini coefficient and coefficient of variation measure variation in an entire pay distribution. An alternative is to use a ratio measure of the pay differences between individuals (for instance, the difference between a player's pay and that of the highest paid player on the team). I used the

simple difference between the maximum pay on a team and a player's pay and also the ratio of a player's pay to the maximum pay on the team as ratio measures of dispersion. Larger values on the former and smaller values on the latter indicated greater dispersion. I included only measures of upward comparisons since research suggests that when people judge their pay against others', the comparisons will be with those who are higher in the organizational hierarchy, not those who are lower (Cowherd & Levine, 1992; Martin, 1981). Pay distribution theories do not provide guidance as to whether different measurements of pay dispersion might produce different effects. Because all four variables measured pay dispersion, I expected the results from all four measures to be consistent with the research hypotheses. Position in pay dispersion. To test the positional hypothesis, I created an interaction term by multiplying the rank of a player's pay on his team [rank of highest-paid player = 1) and the team gini coefficient. A lower value indicated a higher position in the dispersion. This approach tests the moderating effects of a player's position in a pay dispersion on the relationship between team gini and player performance. I used the rank of a player's pay rather than actual pay because I was studying the interaction between relative pay—not actual pay—and pay dispersion. Control variables. Harder (1991) suggested that previous year's performance might be a good indicator of recent performance. I controlled for past performance using a player's total player rating from the previous year (Thorn & Palmer, 1994). I used the player's age and number of seasons in major league baseball (in years) to control for experience and used the number of games played or innings pitched in year t to control for performance opportunities. I used the current winning percentage of the player's team to control for the effects of better (worse) teams on a player's performance. I also used a dummy variable for the years 1985 and 1986. U.S. courts determined that team owners colluded to control player salaries during these years. I controlled for absolute pay levels by including the logarithm of a player's annual salary (Gerhart & Milkovich, 1992; Levine, 1993). To control for external labor market conditions (Ehrenberg & Smith, 1994), I included a control for the total labor costs of each team. This variable also controlled for the possibility that team owners, behaving in line with efficiency wage theory, might decide to pay higher levels to many or all employees in an effort to obtain better talent, induce higher performance, or both (Milgrom & Roberts, 1992). Other team owners might choose to pay below-

1999

Bloom

31

market rates to control labor costs and be willing to forgo some performance or human capital in the process (Milkovich & Newman, 1996). Rather than include the total team payroll, which was very highly correlated with individual player pay, I computed the logarithm of the difference between the total team payroll of team ; and the average total team pay for year t. This measure reflects how much more (less) a particular team's total payroll was than the average team payroll. In the team equations, team talent was controlled for by including the sum of all total player ratings and total pitcher indexes for each team. Each player's total player rating or total pitcher index was weighted by the percentage of total team games in which the player participated. Market size was controlled for by entering the metropolitan area population of a team's home town as provided by the U.S. Bureau of the Census. Market size may be related to a team's income and attendance, although past research indicates the association may be weak (Sheehan, 1996). Canadian population data were collected from maps and tourist information provided by the governments of the cities in which the teams were located. Analysis The data were pooled cross sections and time series. I regressed pay at time t on performance measured at the end of the concurrent year. I assumed that a player's current pay and his relative position in his team's current pay dispersion were more salient than either leading or lagged pay levels and thus influenced his current level of performance. The composition of players on a team and their individual pay levels may change considerably from year to year. This variability makes it difficult for a player to assess his future relative pay. It also makes it less likely that previous relative and absolute pay levels will influence current performance. Measures of past player performance, past pay, and past team performance were taken at year t - 1. I used hierarchical linear modeling in the individual performance analyses to control for team effects, time effects, and cross-sectional heteroscedasticity (Byrk & Raudenbush, 1992). Idiosyncratic factors for each team were accounted for by including a random team effect because the composition of players and managers on teams changes from year to year. I assessed model fit with logarithmic likelihood difference scores and the reduction in Schwarz's Bayesian criterion between models with and without team gini coefficients (Littell, Milliken, Stroup, & Wolfinger, 1996). For the team-level analyses, I used an autoregressive

procedure to control for serial dependence and cross-sectional heteroscedasticity in the data.
RESULTS

Descriptive statistics are presented in Table 1. The average team gini coefficient is .62; however, the maximum is .81 and the minimum is .37, indicating significant range in pay distribution policies. Team ginis and coefficients of variation were correlated at .72. The correlation between the ratio measures [player pay - team maximum and player pay/team maximum) was .67. Pay Dispersion and Individual Performance Statistical results of tests of the relationships between pay distributions and player performance are presented in Table 2. The cell entries in Table 2 are the unstandardized regression coefficients with f-statistics in parentheses below the coefficients. The results provide a great deal of evidence about the effects of pay distributions on player performance: Pay distributions have significant, negative effects on player performance that go over and above the effects of base pay, past performance, age, and experience. Higher gini coefficients are related to lower adjusted batting runs (jB = —12.76, t = -5.19, p < .001), total player rating (j3 = -0.61, t = -2.50, p < .01), pitching runs (/3 = -11.32, t = -3.79, p < .001), and total pitcher index (/3 = -1.10, t = -3.20, p < .01) and to higher (worse) adjusted earned run average (j3 = 36.59, t = 2.91, p s .001). With the exception of defensive performance, the performance of players on teams that had greater dispersion in team pay was lower. These results were also obtained for the team coefficient of variation for all models except the fielding runs model. The results support the compressed model hypothesis about individual performance (Hypothesis lb) and fail to support the hierarchical model hypothesis (Hypothesis la). Results for the ratio measures are also consistent with Hypothesis lb (Table 3). A larger difference between a player's pay and the maximum pay on a team is negatively related to adjusted batting runs (/3 = -0.63, t = -11.59, p < .01), total player rating (i3 = -0.02, t = -4.36, p < .01), pitching runs O = -0.32, t = -4.04, p < .01), total pitcher rating (j8 = -0.04, t = -4.23, p < .001), and adjusted earned run average (/3 = 1.00, t = -3.21, p < .10). A smaller ratio between a player's pay and the maximum jjay on a team is related to poorer performance. One performance measure, fielding runs, is positively related to both ratio measures. The observed effects for pay dispersion account for other

32

Academy of Management Journal

February

TABLE: 1

Descriptive Statistics
Variables 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. Team gini coefficient" Team coefficient of variation" Maximum pay on team - player's pay'''^ Player's pay/maximum pay on team" Total player rating"* Adjusted batting runs'^ Fielding runs'* Total pitcher index" Adjusted earned run average" Pitching runs" Franchise value''^ Gate receipts''* Media income''^ Total income''* Team's winning percentage" Attendance ratio" Finishing position Pay*"'" Adjusted total team pay Age" Seasons in major league baseball" Games played'' Innings pitched" Market size"'' League Mean 0.60 1.01 13.14 0.27 -0.05 -0.04 -0.59 0.15 105.84 0.95 113.75 19.81 27.12 58.05 0.48 0.25 14.58 626,826 658,397 29.75 0.06 96.49 66.15 5.91 0.48 s.d. 0.07 0.21 3.83 0.27 1.39 12.55 8.54 1.32 46.07 11.49 24.33 7.31 9.77 16.75 0.13 0.09 1.07 786,875 744,058 4.71 0.24 43.87 74.79 4.91 0.50 2 3 4 5 6 7 8

.71* .12* -.16* -.05* -.04* -.03 -.04* -.04* -.04* -.28* -.24* -.21* -.26* -.03"^ .00 .11* .05* -.13* -.09* .08* -.03'' -.03 -.10* .14*

.16* -.23* -.04 -.06* .00 -.03* -.05* -.05* -.15* -.05* -.17* -.13* -.04* .11* .08* .05* -.07* -.12* .07* -.05* -.05* -.14* .04*

-.67* -.22* -.28* -.04* -.13* -.12* -.12 .00 .05* .02 .06* .03* .07* .01 -.46* .08* -.13* .06* -.20* -.11* .03 .01

.26* .34* .03* .14* .07* .15* .02 .03* .04* .04* .05* -.01 -.03* .79* .02 .35* -.19 .34* .19* .03 -.02

.73* .54*

-.02 .58* .89* .02 .08* .00 .06* .31* .13* -.17* .13* .02 -.03^

.03 .04* .05* .06* .11* .06* -.11* .23* .03+ -.03"^ -.05* .25* .02 .00

.00 .03 .00 .02 .02 .03 -.12* .30* .02 -.02 -.07* .30* .00 .00

.04* .03 .06* .07* .07* .03* .00 .04* .00 -.01 .00 .02 .03'^ -.03*

.14* .02 -.02

" Measured at the organizational level; n = 236. ^ Logarithm of variable was used in analyses. • Measured at the individual level, n = 8,808. ^ '' Measured at the individual level, n = 5,869. ° Measured at the individual level, n = 2,939. ' Expressed in millions of dollars. * Measured at the organizational level, n = 106. * " p < .10 ^ * p < .05

unique characteristics of each team and indicate that salary distributions, per se, influence individual player performance. However, the results do not indicate that pay levels should he completely equal since higher absolute pay is also related to better player performance. Moderating Effects of Pay Dispersion The results support Hypothesis 2 (Table 4). A player's position in the pay dispersion appears to moderate the relationship between pay dispersion and individual performance. With the exception of fielding runs, the sign of the interaction terms is opposite that of the main (independent) effects, which indicates that pay dispersion moderates individual performance: Greater dispersion is negatively related to the performance of those lower in a dispersion and positively related to the performance of those higher in the dispersion. A graph of the interaction effect for total player rating is pre-

sented in Figure 1. I created the high and low groups for both pay dispersion and position in the dispersion using one standard deviation above and below the mean as cutoffs. Pay Dispersion and Team Performance The low correlations between measures of financial and on-field performance support Sheehan's (1996) assertions that team owners may pursue different, potentially divergent, performance goals. However, among these data the effects of pay distributions on team performance are consistent across these two dimensions of organizational success (Table 5). Adding a team's gini coefficient to the models increases the variance explained (i.e., increased the R^] of all models, suggesting that pay distributions have important effects on firm performance. Pay distributions explain 18 percent of the variation in gate receipts, 11 percent of the variation in total income and attendance ratio, 7 percent

1999

Bloom

33

TABLE 1 (Continued)
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

.61* .04+ .10*
.03

.02

.07*
.02 .07

.08* .23* .10* -.13* .08*
.00 .03

.52* .67* .77"
.02

.35* .13* -.21* .14* .03+
.02

.46* -.14
.04

.07* .05* -.03

.17*
.03 .00

.32* .03+ -.01 .03+ -.02 .59* -.11*

.23* .74* .37* .86* -.34* .14* .11*
.03

.77

-.05* -.02
.04

.15+ .60* -.16+ .13* .18* .34* .30* .07* • .08* -.04 -.07*
.07 .01 .01 .00 .00 .04

-.08

-.52* -.90*
.04

.03*
.02

-.06* .13* .25*
.04

-.01
.01

-.02*
.02 .02

-.01 -.13* -.02 -.02*
.00 .00

.18+ -.14*

.73* -.05*

.61* -.10*

.22* .11+ -.01

.14* -.09*

.07* .08*

.18* .28* -.16* .27* .16* .10* -.04

-.08*
.00

-.03+ -.02 .38* -.02*

-.21* -.04 -.03+ .05* -.03*

-.15* -.10* -.01
.00

.00

.00

-.04

-.01

.04*

of the variation in team value, 5 percent of the variation in winning percentage, 3 percent of the variation in media income, and 1 percent of the variation in finishing position. The results support Hypothesis 3 and indicate that a more hierarchical pay dispersion leads to poorer team performance. The gini coefficient is significant and negative across all team performance measures (Table 5). Teams with greater dispersion in salaries have lower on-field performance [li^innj^g percentage = "0.26, t = - 2 . 3 7 , p < •05; ^attendance ratio = "0.41, t = - 2 . 2 6 , p < .05; position = 18.46, t = 1.98, p < .05). Increasfig p d ing the dispersion by a .01 gini coefficient reduces

-2.15, p < .05) are negatively related to team gini. In addition, team value is significantly and negatively related to pay dispersion (/3 = -169.95, t = -3.04, p < .05).
DISCUSSION

the winning percentage 26 percent and decreases the attendance ratio 41 percent. Financial performance is lower for teams with more hierarchical pay distributions. Gate receipts (/3 = -26.72, t = -2.11, p < .05), media income (/3 = -14.04, t = -1.80, p < .07), and total income 0 = -51.70, t =

Perhaps the most important implication of this study is that dispersion in pay distributions appears to have strong implications for individual performance and organizational success, at least in the context of major league baseball. Across almost all measures, player performance was negatively related to more hierarchical pay dispersion, as were all measures of organizational performance. Dispersion in pay distributions also appeared to be related to the financial value of a franchise, maximization of which is the ultimate objective of most owners. The fact that pay distributions have broad and significant effects on organizational prosperity in

34

Academy of Management Journal

February

TABLE 2 Effect of Pay Structures on Player Performance: Variation Measures"
Adjusted Batting Runs Variable Model 1 (-5.19) -3.48 ** (-4.48) (3.70) 0.18 (-2.84) -0.25 * (-2.84) -2.61 * (11.94) (-0.21) (3.96) (7.67) (-4.99) (29.80) 0.05 ** -0.05 9.62 1.11 ** -0.72 ** 3.49 ** (0.35) -0.42 (-3.07) 0.05 (-3.07) 0.17 (12.09) (-0.10) (4.24) (7.18) (-7.18) (31.86) -0.01^ ' -0.23 1.04 (-1.00) (1.59) (0.27) Model 2 Fielding Runs Model 1 2.00 (1.10) 0.80 -0.52 0.04 0.15 (1.34) (-1.30) (1.58) (0.24) (-3.00) (-0.44) (0.60) 0.02 -0.01** -0.26* 0.004** -0.06 1.07** (0.59) (-3.56) (-2.85) (9.55) (-1.32) (4.93) (1.51) (-2.58) (44.59) Model 2 Total Player Rating Model 1 -0.61** (-2.50) -0.15* 0.02 -0.01** -0.26* 0.004** -0.07 1.13** (-1.97) (0.48) (-3.63) (-3.01) (9.34) (-1.42) (5.00) Model 2

-12.76 *• Team gini coefficient Team coefficient of variation 0.21 Indicator for league*" -0.25 * Age Seasons in major league -2.45 • riacor^O 11

Games played/innings pitched*^ Indicator for collusion years'* Team's winning pUl Lcllld^U

0.05 ** -0.10 9.28 • • 1.23 ** -0.55 ** 3.37 **

(-2.44) - 0 . 0 1 * (-0.67) (0.60) -0.15 1.01 -0.38**' 0.16* 2.06*''

PayAdjusted team pay" Lagged total player rating -2 log likelihood Difference in chi-square Reduction in Schwarz's Bayseian criterion

-0.38** (-3.23) 0.12 (1.51) 2.00^** (24.01)
36 ,123.20 1 ,387.79**

(-3.32) 0.02 (2.15) -0.03* (25.39) 0.52**

0.02 (1.17) -0.04* (-3.49) 0.53** (47.94) 16,291.77 3.87** 1.93

39, 272.86 1, 469.58**

40 ,722.58

27, 509.07

734.80

19.86** 9.90

694.00

1.92 1.00

15,806.95 488.69** 244.38

" Cell entries are unstandardized regression coefficients; numbers in parentheses are t-statistics. ^ Coding: 1 = National League. "^ Variable is games played for batters and innings pitched for pitchers. '' Coding: 1 = 1985 or 1986. ° Logarithm. • Reduction in values indicates a better-fitting model. ^ ^ p < .10 * p < .05 ** p < .01

these data points to the importance of studying them further in samples of other organizations. The results may reflect the fact that wealthier teams paid a few star players more (thus leading to greater dispersion). This may be the case since pay and performance were m^easured in the same year. Implications for Research These data also underscore the importance of understanding contextual factors that mitigate the effects of different pay distributions (Pfeffer & Langton, 1993). In the major league baseball data reported here, a players' position in the pay distribution moderated the pay dispersion-individual performance relationship. In organizations that require collaborative effort, these moderating effects may work against organizational success. In such organizations, purchasing the talents of a highly skilled worker may lead to lower organizational performance because of the way such a purchase changes relative pay inside the organization. To the extent that attempts to purchase talent become bidding contests, the price for talent may be driven

even higher so that owners who win the contest may create extreme dispersion inside their organizations (Frank & Cook, 1995). One potential way to pursue this idea is to study the impact of pay dispersion on justice perceptions. According to McFarlin and Sweeney (1992), perceptions of distributive justice are more strongly associated with attitudes toward specific returns (e.g., pay level satisfaction), and procedural justice perceptions are more strongly related to overall evaluations of organizations (e.g., organizational commitment). It might be informative to investigate whether justice perceptions vary under different pay distributions. For example, people may agree that better performers should get paid more, but at some point the additional rewards conferred by a hierarchical distribution become too much and diminish the pay-performance link. At this point, underrewarded employees may view greater dispersion as distributively or procedurally unjust and produce negative reactions, such as withholding effort, stealing, sabotaging coworkers, and reducing cooperation (Brockner & Wiesenfeld, 1996; Folger, 1993). Pay inequality may also disrupt cooperation

1999

Bloom

35

TABLE 2 (continued)
Adjusted Earned Run Average Model 1 36.59* -1.19 -0.09 -10.03 -0.01 3.35 43.65 ** 2.88 1.79 ** 4.13 ** (2.91) 14.18* (-0.63) (-0.48) (-0.67) (-1.09) (0.47) (10.05) (3.71) (3.39) (6.34) -0.05 -0.05 -1.88 -0.01 1.67 4.00** 3.36** 2.68** 4.00** (3.40) (-0.26) (-0.26) (-0.13) (-1.18) (0.59) (10.25) (4.49) (4.65) (6.19) -0.18 -0.20 0.54 0.01 -0.96 17.67** (-0.24) (-4.25) (0.14) (3.58) (-1.36) (16.01) Model 2 Model 1 -11.32 * (-3.79) -3.53* -0.35 -0.21 -0.16 0.01* 1.67 17.60*'' 0.39^ -0.20 1.60*' (-3.15) (-0.49) (-4.28) (-0.04) (3.65) (-0.59) (16.05) (2.01) (-1.54) (9.68) -0.05 -0.02 0.02 0.001* -0.08 1.77** (-0.67) (-4.48) (0.04) (2.38) (-0.95) (13.77) Pitching Runs Model 2 Total Pitcher Rating Model 1 -1.10* (-3.20) 2.26* -0.07 -0.02 -0.06 0.001* -0.08 1.77** (-2.04) (-0.97) (-4.53) (-0.15) (2.43) (-0.93) (13.80) Model 2

0.48 (2.40) -0.10 (-0.76) 1.57 ** (9.42)

0.07* (3.01) -0.02 (-1.24) (9.60) 0.19** 7,705 .30 49 .57** 24 .79

0.06* (2.66) -0.03* (-2.06) 0.19** (9.86)

•

24

1,184.62 153.97** 77.09

24,:335.62 3.27^ 1.70

17, 805.43

115.22** 57.61

17,914.46 6.19* 3.09

7,753.09 1.78 0.09

TABLE 3 Effect of Pay Structures on Player Performance: Ratio Measures"
Variable Adjusted Batting Runs Fielding Runs 0.15* (3.71) Total Player Rating -0.02** (-4.36) 0.51** (5.84) Adjusted Earned Run Average 1.00* (3.21) Pitching Runs -0.32** (-4.04) 3.56* (3.07) Total Pitcher Index -0.04** (-4.23) 0.39* (2.90)

Maximum pay on —0.63** (-11.59) team — player's pay'' Player's pay/maximum 11.93** (14.04) pay on team

-2.77** (-4.33)

-21.09** (-4.74)

" Models include league, age, seasons, games played/innings pitched, team winning percentage, pay and adjusted team pay (both logarithms), and the lagged total player rating or total pitcher index. Cell entries are unstandardized regression coefficients; t-statistics are in parentheses. * Logarithm, • ^p < .10 * p < .05 **p < .01

and social cohesion, which can impair individual and group performance (Baron & Pfeffer, 1994). These negative reactions may be exacerbated if pay does signal an individual's organizational value (Folger, 1993; Frank, 1985). Thus, too much dispersion may create feelings of social and psychological, as well as economic, injustice (Deutsch, 1985; Folger, 1993). This idea and others—for instance, that more compressed pay distributions inculcate feelings of common fate and team spirit—have not been empirically investigated.

Another contextual factor is whether differences between team-oriented and individual-oriented work environments influence the effects of pay dispersion (Lazear, 1995; Pfeffer & Langton, 1993). Virtually all of the research in support of hierarchical distributions has sampled work in which only individual performance matters (Becker & Huselid, 1992; Ehrenberg & Bognanno, 1990); work interdependencies were either not considered in these studies or were limited to a very small group of people (Pfeffer & Langton, 1993). This study is

36

Academy of Management Journal

February

TABLE 4 Effect of Position in Pay Structure and Pay Structures on Player Performance"
Variable Adjusted Batting Runs Fielding Runs 1.54 (0.51) -0.09 (-1.04) 0.10 (0.78) Total Player Rating -1.24** (-4.40) -0.02+ (0.01) 0.04** (2.55) Adjusted Earned Run Average 73.20** (4.17) 1.32* (2.02) -2.46** (-2.63) Pitching Runs -14.84** (-3.27) -0.26 (-1.56)+ 0.42+ (1.74) Total Pitcher Index -1.41** (-2.69) -0.03 (-1.43) 0.04+ (1.72)

Team gini coefficient -17.28** (-5.53) Rank'' -0.09 (-0.76) Rank X gini*" 0.32+ (1.90)

° All models include league, age, seasons, games played/innings pitched, team winning percentage, pay and adjusted team pay (both logarithms), and the lagged total player rating or total pitcher index. Cell entries are unstandardized regression coefficients; numbers in parentheses are t-statistics. '' These variables were entered "stepwise" into hierarchical regression analyses. + p < .10 * p < .05 **p < .01

FIGURE 1 Interaction of Player's Rank in Pay Dispersion and Team Gini Coefficient on Total Player Rating

High rank

Performance Low rank

Low

Pay Dispersion

High

among the first to test the effects of pay dispersion in a context where work interdependencies are important. Hierarchical pay distributions may be more appropriate when individual characteristics are closely tied to organizational outcomes (as in law, accounting, and consulting firms, research and development units, surgical teams) or when the contributions of individuals are more easily separated from organizational performance (as is the sales performance of stock brokers or research performance in academia; Pfeffer & Langton, 1993). Here, the emphasis on the individual rewards engendered by hierarchical dispersion coupled with the relatively low demands for cooperation may increase both individual and organizational performance. Thus, it may be beneficial for a law firm to

pay a relatively high salary to attract a top attorney or for a university to offer an endowed chair to a particularly productive scholar. In other types of organizations—fire fighting and rescue squads, theatrical casts, manufacturing teams, and hotel customer service staffs, for example—the situation is quite different because the poor performance of a particular worker can be compensated for by the better performance of the other workers, and the outstanding performance of one person is unlikely to influence organizational outcomes over the long term if the performance of others is lacking. For example, in 1991 Calvin Ripken, Jr., was the American League's Most Valuable Player, but his team, the Baltimore Orioles, finished 24 games behind the division leader. In organizations in which work

1999

Bloom

37

TABLE 5 Effect of Pay Structures on Team Performance"
Variable Winning Percentage Attendance Ratio Finishing Position (1.98) (-2.26) (0.93) (-2.38) Gate Receipts -26.72* (-2.11) 1.27 1.95 35.51* (0.65) (0.41) (3.21) Media Income -14.04+ 0.60 3.44* -7.94 (-1.80) (0.34) (2.19) (-0.93) Total Income -51.70'• 0.17 2.54 26.14 Franchise Value (-3.04)

Team gini -0.26* (-2.37) -0.41* (-2.26) 18.46' coeriicient Games played 0.03 (1.62) 0.006 (0.20) -2.62' * Average total team 0.0005 (0.001) -0.05 (-0.68) 2.46 payroll in major league baseball Last season's winning 0.19* (1.94) 0.58*'• (3.64) -23.17' • percentage Average team pay -0.007 (-0.19) 0.02 (0.40) -2.19 Market size -0.008 (-1.05) 0.006 (0.45) 0.00 Team talent 0.004 (5.00) 0.002 (1.42) -0.63' •* Constant -4.01 (-1.42) -1.61 (-0.35) 427.36'• R^ Cbange in R^ from adding team gini coefficient
.33 .05 .27 .11

(-2.15) -169.95* (0.04) (0.28) (1.24)

-13.79 (-1.58) -37.99"^ (-1.80) 68.82 (1.44) (1.80) (6.21) (0.55) (1.31)

(-0.95) 5.26 (1.25) (0.30) 1.85* (2.04) (-6.07) 0.16 (1.80) (2.31) -278.04 (-0.87)
.37 .18

0.11 (0.03) 0.001 ** (11.08) 0.04 (0.57) -112.54 (-0.39)
.59 .03

11.82 (1.48) -33.49+ 11.91'•* (6.90) 24.94** 0.29 (1.70) 0.22 -318.23 (-0.52) 1,853.15
.57 .11 .41 .07

.29 .01

" Cell entries are unstandardized regression coefficients; t-statistics are in parentheses.
•^p < .10

* p < .05 **p < .01

is more highly interdependent and involves a larger group of people, the competition among workers for the higher pay fostered hy hierarchical distrihutions may he deleterious to organizational performance (Hill, Hitt, & Hoskisson, 1993; Lazear, 1995). Pfeffer and Langton (1993) found that academic researchers tended to cooperate less at institutions with more hierarchical pay distrihutions. When outcomes result from many different types of performance, it hecomes harder to "partial out" the contrihutions of any one individual; this limitation complicates the evaluative process and creates room for disagreement, divergent perceptions, and measurement error (Dye, 1984). In these cases, more hierarchical distributions may have negative effects, such as overrewarding some people, a possihility that should he weighed carefully against the potential incentive properties of such distrihutions. Future research should investigate work interdependencies and test how^ they influence the effects of pay distrihutions on individual and organizational performance. The literature on teams (Cohen & Bailey, 1997) and team pay (Gerhart & Milkovich, 1992; Weiss, 1987) may offer informative theoretical models. Future researchers might extend this study hy investigating the homogeneity of work and its relationship to different types of dispersion (Baron & Pfeffer, 1994). Lateral dispersion (Pfeffer, 1994) refers to the spread, or variation, in pay among similar johs or johs at the same organizational level. Vertical dispersion (Milkovich & Newman, 1996) refers to that among different johs at different organizational levels. Virtually all of the research on pay structures addresses vertical dispersion, with

professional auto racers and golfers, university faculties, and now, major league hasehall players all having heen studied. The effects of more hierarchical lateral dispersions might have different effects on justice perceptions, attitudes, and performance since fewer objective or external factors would differentiate johs or employees. Individuals may tolerate vertical dispersion hetter than lateral dispersion, hut empirical research is required to provide answers. The work of Baron and Pfeffer (1994) suggests that effects might he more pronounced with lateral dispersion. Organizational practices might move organizations from having vertical dispersions to having lateral dispersions. For example, as some organizations move to more team-based work, they are also creating broader jobs and fewer organizational levels, factors that might tend to create problems with lateral dispersion. Pay dispersion might operate differently when the salaries of referent others are not known. Research indicates that in such cases people tend to overestimate the pay of others (Lawler, 1971). Overestimates tend to increase as people make judgments ahout the pay of others who are successively more separated from them in a pay distribution. When the salaries of referent others are not known, people may tend to judge a pay distribution as being more hierarchical than it in fact is. Implications for Practice This study also has practical implications. Since pay distrihutions can he created to reflect performance differentials, awarding proportionately greater pay to employees who are performing hetter

38

Academy of Management Journal

February

will likely, over time, create a more dispersed pay distribution (Heneman, 1992). Merit-based pay systems are one of several compensation policies that might have this effect. This study suggests that changing a pay distribution may create motivational problems that work against both an organization's goals and, perhaps, the compensation plan itself. This effect would be most pronounced in a tournament-like pay distribution where differentials in payouts become consecutively larger as an employee moves up the performance-pay hierarchy. The negative effect would be exacerbated if rewards for similar performance levels were not consistent over time, effort, or place, as would occur when, for example, a team's star employee in one year received less than star performers received in previous years. In major league baseball, pay changes are most often determined by past performance (Bretz & Thomas, 1992; Lord & Hohenfeld, 1979), but the amount of reward associated with any performance increment varies with factors such as the team a player is negotiating with and the other players in the market at the same time (Sheehan, 1996). Thus, players with similar performance histories may be rewarded unequally, creating a more hierarchical distribution that carries with it potential distributional injustices. This situation is becoming common elsewhere, as well; for instance, in many accounting, law, and consulting firms the salaries paid to new employees are increasing at a faster rate than those of current employees. Here again, positional effects might be important. For example, the distinction between management and nonmanagement workers may be one important point of demarcation. These data also raise questions about the efficacy of raiding high-priced talent, a practice common in the computer industry. Sheehan (1996) noted that attempts by owners of major league baseball teams to buy success by purchasing the potential performance of high-priced talent have met with mixed results at best. These data suggest an explanation behind this phenomenon—attracting stars creates a more hierarchical pay distribution, which may reduce individual and organizational performance. Organizations may need to pay greater attention to how engaging the imperatives of the external market influences their internal distribution of wages. That internal wage distributions affected performance in major league baseball, where the external market value of wages is well known, suggests that internal equity may matter as much or more in situations where external labor market rates are not as widely known. Securing the talents of a star employee may mean the pay of other employees will also have to be raised.

Limitations One strength of this study is that it analyzed the pay practices of all the organizations competing in an industry. Another is that performance was measured at both the individual and organizational level. However, one limitation is that organizations from only one industry were studied. Future research should not only sample organizations across industries, but also incorporate a sufficient sample from within a single industry to control for competitive and external labor market factors. Future researchers should also strive to measure performance at multiple levels to accurately tap the full range of effects. These results may have differential applicability outside of both major league baseball and the United States. Contextual factors like differences in national culture or public policy may be important for understanding the performance effects of pay distributions. The regulation of pay distributions by national statutes in some countries implies a social norm against hierarchy. Another concern is that there may be few similarities between the pay systems used in baseball and those used in other contexts. The open-endedness of baseball salaries, the restricted ability of some players to move freely from team to team, the relatively short careers of most players, and other factors may make baseball a unique context. To the degree that this is the case, these data have limited generalizability.

CONCLUSION

The results of this study indicate that greater dispersion in pay within an organization is associated with lower individual and group performance, at least where work interdependencies are important. The results also endorse Gerhart and Milkovich's (1992) claim that pay distributions have important performance consequences. This study also highlights the need for more research on how context moderates the effects of pay distributions. This study is one step in that direction,
REFERENCES Baron, J. N., & Pfeffer, J. 1994, The social psychology of organizations and inequity. Social Psychology Quarterly, 57: 190-209. Becker, B. E., & Huselid, M. A. 1992. The incentive effects of tournament compensation systems. Admin-

istrative Science Quarterly, 37: 336-350, Bok, D. C. 1993, The cost of talent: How executives and professionals are paid and how it affects America.
New York: Free Press.

1999

Bloom

39

Bretz, R. D., Jr., & Thomas, S. L. 1992. Perceived equity, motivation, and final-offer arbitration in Major League Baseball. Journal of Applied Psychology, 77: 280-287. Brockner, J., & Wiesenfeld, B. M. 1996. An interactive framework for explaining reactions to decisions: Interactive effects of outcomes and procedures. Psychological Bulletin, 120: 189-208. Bryk, A. S., & Raudenbush, S. W. 1992. Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage. Bucek, J. (Ed.). 1996. The baseball encyclopedia. New York: Macmillan. Cappelli, P., & Sherer, P. D. 1990. Assessing worker attitudes under a two-tier wage plan. Industrial and Labor Relations Review, 43: 225-244. Cohen, S. G., & Bailey, D. E. 1997. What makes teams work: Group effectiveness research from the shop floor to executive suite. Journal of Management, 23: 239-291. Cowherd, D. M., & Levine, D. L 1992. Product quality and pay equity between lower-level employees and top management: An investigation of distributive justice theory. Administrative Science Quarterly, 37: 302320. Deutsch, M. 1985. Distributive justice: A social-psychological perspective. New Haven, GT: Yale University Press. Donaldson, D., & Weymark, J. A. 1980. A single-parameter generalization of the gini indices of inequality. Journal of Economic Theory, 22: 67-86. Dye, R. A. 1984. The trouble with tournaments. Economic Inquiry, 22: 147-149. Ehrenberg, R. G., & Bognanno, M. L. 1990. The incentive effects of tournaments revisited: Evidence from the European PGA tour. Industrial and Labor Relations Review, 43: 74S-88S. Ebrenberg, R. G., & Smith, R. S. 1994. Modern labor economics: Theory and public policy (5th ed.). New York: HarperCollins. Folger, R. 1993. Reactions to mistreatment at work. In J. K. Murnighan (Ed.), Social psychology in organizations: Advances in theory and research: 161183. Englewood Cliffs, NJ: Prentice-Hall. Frank, R. H. 1985. Choosing the right pond: Human behavior and the quest for status. New York: Oxford University Press. Frank, R. H., & Gook, P. J. 1995. The winner-take-all society. New York: Penguin Books. Gerhart, B., & Milkovich, G. T. 1992. Employee compensation: Research and theory. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychology (2nd ed.), vol. 3: 481569. Palo Alto, GA: Consulting Psychologists Press. Gomez-Mejia, L. R., & Balkin, D. 1992. Compensation,

organizational strategy, and firm Cincinnati: Southwestern.

performance.

Hamilton, W., & Macy, S. 1923. The control of wages (2nd ed.). New York: McGraw-Hill. Harder, J. W. 1991. Equity theory versus expectancy theory: The case of major league baseball free agents. Journal of Applied Psychology, 76: 458-464. Hart, S. 1995. Scouting reports: The original reviews of some of baseball's greatest stars. New York: Macmillan. Heneman, H. G., III. 1985. Pay satisfaction. In K. M. Rowland & G. R. Ferris (Eds.), Research in personnel and human resource management, vol. 3: 115139. Greenwich, GT: JAI Press. Heneman, R. L. 1992. Merit pay: Linking pay increases to performance ratings. Reading, MA: AddisonWesley. Hill, G. W., Hitt, M. A., & Hoskisson, R. E. 1992. Cooperative versus competitive structures in related and unrelated diversified firms. Organization Science, 3: 501-521. Kochan, T. A., & Osterman, P. 1994. The mutual gains enterprise: Forging a winning partnership among labor, management, and government. Boston: Harvard Business School Press. Kohn, A. 1993. Punished by rewards: The trouble with gold stars, incentive plans, A's, praise and other bribes. Boston: Houghton Mifflin. Lawler, E. E. 1971. Pay and organizational effectiveness: A psychological view. New York: McGrawHill. Lawler, E. E. 1990. Strategic pay: Aligning organizational strategies and pay systems. San Francisco: Jossey-Bass. Lawler, E. E., Ill, & Jenkins, G. D., Jr. 1992. Strategic reward systems. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of industrial and organizational psychology (2nd ed.), vol. 3: 1009-1055. Palo Alto, GA: Gonsulting Psychologists Press. Lazear, E. P. 1995. Personnel economics. Gambridge, MA: MIT Press. Lazear, E., & Rosen, S. 1981. Rank-order tournaments as optimum labor contracts. Journal of Political Economy, 89: 841-864. Levine, D. I. 1993. What do wages buy? Administrative Science Quarterly, 38: 462-483. Littell, R. G., Milliken, G. A., Stroup, W. W., & Wolfinger, R. D. 1996. SAS system for mixed models. Gary, NG: SAS Institute.' Livernash, E. R. 1957. The internal wage structure. In G. W. Taylor & F. G. Pierson (Eds.), New concepts in wage determination: 143-172. New York: McGraw-Hill. Lord, R. G., & Hohenfeld, J. A. 1979. Longitudinal field assessment of equity effects on the performance of

40

Academy of Management Journal major league hasehall players. Journal of Applied Psychology, 64: 19-26.

February

Pindyck, R. S., & Ruhinfeld, D. L. 1995. Microeconomics (3rd ed.). Englewood Cliffs, NJ: Prentice-Hall. Reingold, J, 1997, Executive pay: Special report. Business Week, April 21: 58-66. Riegel, J. W. 1937. Wage determination. Ann Arhor, MI: Bureau of Industrial Relations. Rynes, S. L. 1992. Recruitment, joh choice; and post-hire consequences: A call for new research directions. In M. D. Dunnette & L. M. Hough (Eds,), Handbook of industrial and organizational psychology (2nd ed.), vol. 2: 399-444. Palo Alto, GA: Gonsulting Psychologists Press. Sheehan, R, G. 1996. Keeping score: The economics of big-time sports. South Bend, IN: Diamond Gommunications. Thorn, J,, & Palmer, P, 1994. Total baseball CD-ROM. Portland, OR: Creative Multimedia. Thurow, L. 1975. Generating inequality: Mechanisms of distribution in the U.S. economy. New York: Basic Books. Weiss, A. 1987. Incentives and worker hehavior: Some evidence. In H. R, Nalbantian (Ed.), Incentives, cooperation, and risk sharing: 137-150. Totowa, NJ: Rowman & Littlefield. White, D. J., & Associates, 1995. The 1995 annual attorney salary survey. Wilmette, IL: White & Associates. Zenger, T. 1992. Why do employers only reward extreme performance? Examining the relationships among performance, pay, and turnover. Administrative Science Quarterly, 37: 198-219,

McFarlin, D. B., & Sweeney, P. D. 1992. Distrihutive and procedural justice as predictors of satisfaction with personal and organizational outcomes. Academy of Management Journal, 35: 626-637. Mahoney, T. A. 1979. Compensation and reward perspectives. Homewood, IL: Irwin. Martin, J. 1981, Relative deprivation: A theory of distrihutive injustice for an era of shrinking resources. In L. L. Cummings & B. M. Staw (Eds.), Research in organizational behavior, vol. 3: 53-107, Greenwich, CT: JAI Press. Merton, R. K. 1973. The sociology of science: Theoretical and empirical investigations. Chicago: University of Chicago Press. Meyer, H. H. 1975. The pay-for-performance dilemma. Organization Science, 33: 39-50. Milgrom, P., & Roherts, J. 1988. An economic approach to influence activities in organizations, American Journal of Sociology, 94: S154-S179. Milgrom, P., & Roherts, ), 1992. Economics, organization, and management, Englewood Cliffs, NJ: Prentice-Hall. Milkovich, G, T., & Newman, J. M. 1996. Compensation (5th ed.), Homewood, IL: Irwin. O'Reilly, C, Main, B. S,, & Crystal, G, 1988. GEO compensation as tournaments and social comparisons: A tale of two theories. Administrative Science Quarterly, 33: 257-274, Pfeffer, J. 1994. Competitive advantage through people: Unleashing the power of the work force. Boston: Harvard Business School Press. Pfeffer, J., & Langton, N. 1993. The effect of wage dispersion on satisfaction, productivity, and working collaboratively: Evidence from college and university faculty. Administrative Science Quarterly, 38: 382407.

Matt Bloom is an assistant professor of management at the University of Notre Dame. He received his Ph.D. from the School of Industrial and Lahor Relations at Gornell University. His current research interests include the causes and consequences of compensation systems. .