Does player mobility lead to higher earnings? Evidence from the NHL.

Vincent, Claude ; Eastman, Byron

I. Introduction

The extensive coverage of player movement at trading and free agency deadlines has generated a perception among fans and journalists that players on the move are consistently moving to higher salaries. A corollary of this perception is that players who have been more mobile during their careers should have higher earnings than their more stable counterparts, ceteris paribus. No one has looked at the veracity of this hypothesis in the context of a major professional sport. This paper does so through an examination of the relationship between mobility and earnings in the National Hockey League. Specifically, we inquire into the cumulative impact of changing teams on the earnings of players.

A cursory look at NHL earnings data shows that players who have changed teams earn 30 percent more, on average, than players who have not, a stylized fact that supports the aforementioned commonly held perception. But does moving systematically lead to higher earnings? To answer this question, we estimate a multivariate earnings function in a human capital framework to ascertain if a player's current salary is a reflection of the impact of all past mobility decisions. We also examine the manner in which free agency shapes the relationship. This is the first investigation of the effect of career mobility on earnings for NHL players.

There is a paucity of literature examining the relationship between free agency and earnings in the NHL with only lip-service being paid to the effects of mobility, per se. In one of the earliest studies, McPherson (1976) examines turnover rates in the NHL from the 1950-1951 to the 1965-1966 seasons. The league formed a closed system during this period and since there was no external competition with respect to personnel recruitment, turnover of players was involuntary. The author finds that involuntary turnover rates varied by position. Goaltenders and defensemen had lower turnover rates than forwards. The study doesn't examine the relationship between turnover and earnings, per se, but the author suggests that the positions with the highest turnover rates are the most highly remunerated. This is the first hint of the hypothesis that earnings and mobility in the NHL may be directly related. Jones, Nadeau, and Walsh (1997, 1999), in two separate investigations, examine NHL earnings for the 1989-1990 season and include a dichotomous variable to capture the free agent effect. It takes a value of one if the player has signed a contract as a free agent (excepting his first contract). Since there was little inter-club competition for players and only marginal quality players changed teams as free agents during the study period, the authors expect a negative relationship between earnings and free agency. Instead, they find no statistically significant relationship. Idson and Kahane (2000) examine earnings in the NHL for the 1990-1991 and 1991-1992 seasons. In their study, they test the premise that a player who will become a free agent at the end of a season may perform better than he might otherwise in order to increase his status and hence future earnings with potential employers. They include in their earnings equation a dichotomous variable which takes a value of one if the player was a free agent in the year previous to the one associated with the current salary. They find a positive relationship between earnings and free agency. In another study, Curme and Daugherty (2004) use data from the 1998-1999 and 1999-2000 seasons to estimate separate earnings equations for older and younger players. They argue that older players are likely to have been eligible for unrestricted free agency sometime in their career while younger players will have never been eligible. Their empirical results suggest the existence of a more competitive market wherein free agency permits greater player mobility, and which has earnings linked more closely to actual performance as experience increases.

This study takes the literature a step further: it examines the effect of career mobility on current earnings. Our analysis incorporates both free agent and non-free agent environments. The rest of the paper is organized as follows. Section II presents the model followed by a description of the data. Section III presents the estimation results. Concluding remarks are presented in the final section.

II. Empirical Specification

In labor markets mobility can be either voluntary or involuntary. Employees can change jobs for three main reasons: layoffs, discharges and quits. For an NHL player, however, moves are completely involuntary until later in his career. Generally, a player's career begins when he is selected by a team in the annual entry draft. The term of his first professional contract is fixed on the basis of his age, and compensation is capped at levels prescribed in the collective bargaining agreement. When his first contract expires the player becomes a restricted free agent. This means that he can then sign with another team but the requirement that compensation be paid to his original team significantly inhibits his ability to initiate team changes. During this part of his career any move to another team is a management decision--from the player's perspective any move is involuntary. The player can be traded to another team for compensation. The team can also leave the player unprotected to be claimed in a waiver or expansion draft. Management's final option is to not offer the player a contract when his current contract expires, thereby permitting him to sign with any team. Players released in this way are typically performing below expectations. Later in his career a player will become a free agent (1). This is the first time in his career that the player can negotiate and sign a contract with any team on his own terms--from the player's perspective the move would be seen as voluntary. His current team cannot interfere with his decision other than through offering him another contract. It may, in fact, compete with other teams to retain his services. We incorporate both pre-free agency and post-free agency moves into our analysis of the relationship between mobility and earnings in the final part of the paper.

1. The Model

We specify an empirical log-linear earnings equation that includes measures of offensive and defensive performance, experience, reputation, amateur background and mobility as explanatory variables. Letting [LNSAL.sub.i] denote the natural logarithm of player i's earnings, [x.sub.i] represent a vector of typical human capital based explanatory variables, and [M.sub.i] refer to a vector of mobility variables, we estimate the following linear relationship:

[LNSAL.sub.i] = [[beta].sub.0] + [[beta].sub.1][X.sub.i] + [[gamma].sub.1][M.sub.i] + [u.sub.i] (1)

where [[beta].sub.0] is the constant term, [[beta].sub.1] and [[gamma].sub.1] are vectors of unknown parameters, and [u.sub.i] is the disturbance term. Estimates of the parameters are obtained using ordinary least squares.

Our model includes the following explanatory variables commonly used in the literature. Experience, which embodies intangible skills, is expected to yield positive returns and is captured by the number of games played (GAMES) over a player's career in the NHL. Any nonlinear effects of experience on earnings are captured by its square (SQGAMES). We expect a positive relationship between earnings and offensive performance which is primarily measured by career points per game (PTSGAME). We also expect a positive relationship between earnings and a player's aggressiveness and his defensive performance. Aggressiveness is captured by career penalty minutes per game (PIMGAME) and defensive skills are captured by career plus-minus record per game (PLUSGAME). We also include the height (HEIGHT) and weight (WEIGHT) of players in the earnings equations because physically larger players may be able to use their size to gain an advantage during games by obstructing the play of opposing players and by creating scoring opportunities for themselves and teammates. We expect a positive relationship between earnings and the size of the player. Since fans are attracted to a player's star power, we include in our model the dichotomous variable STAR which indicates whether a player has ever been selected to either a First or Second All-Star team during his career. We expect a positive relationship between earnings and this measure of reputation. A dichotomous variable DRAFT indicates whether or not the player was selected in the first two rounds of the NHL Entry Draft. A player who is selected in the early rounds of the draft may be able to negotiate a higher salary in his initial contract, especially if his initial skills are a good indicator of future performance that may persist throughout his career.

We also control for a player's amateur background by including dichotomous variables that indicate if he played most of his junior hockey in the Ontario Hockey League (OHL), the Quebec Major Hockey League (QMJHL), the Western Hockey League (WHL), the university or college system in the United States or Canada (USCOLLEGE), or in Europe (EUROPE). Jones, Nadeau and Walsh (1999) argue that university or college trained players and European players may experience higher opportunity costs and could therefore demand higher initial salaries. Junior hockey variables can also be used as proxies for a player's style of play. For example, in models that predict a player's offensive performance in the NHL, Dawson and Magee (2000) suggest that European players are more offensive minded than North American players. Similarly, Voyer and Wright (1998) find that offensive performance in junior hockey (goals or points per game scored) is a significant predictor of NHL performance. However, penalty minutes, weight and reputation are also significant predictors, but with smaller effects, for Canadian and American born players. In addition to style of play, amateur background may also represent unmeasurable forces related to networking possibilities. For example, if Ontario based agents dominate the NHL, this could provide OHL players with a comparative advantage in salary negotiations.

The primary mobility variable (MOVER) is a dichotomous variable that indicates whether a player played for more than one team in his career. Human capital theory provides the basis for the relationship between mobility and earnings. NHL players acquire general training with every game they play. They also acquire specific training unique to each team. When a player changes teams he brings with him his general training and leaves behind any specific training he may have acquired. His current earnings will reflect a combination of his accrued general and specific training. Because a player who has played for more than one team may have more nontransferable human capital, he should have lower productivity and therefore earn less than his counterparts who have played with just one team, ceteris paribus. However, this loss in human capital resulting from the nontransferable specific training may be more than compensated for by a better match and/or teammate effects. Idson and Kahane (2000) discuss such teammate effects on individual player productivity and find that players in the NHL can be considered as complementary inputs and they suggest that this complementarity has positive effects on player salaries. In addition, a move to a new team may result in a better match when the player's style of play corresponds with the coaching philosophy and/or the team's style of play. Such moves would also include considerations related to team staffing requirements wherein a player leaves a team overstaffed in his position for a team that is understaffed in the position.

2. The Data

Our data set contains career information on 625 players: 407 forwards and 218 defensemen. Players are included if they played at least 10 games in the NHL and salary data for the 2003-2004 season are available (2). The reported earnings of players were obtained from the NHLPA website (www. nhlpa.com) and reflect earnings for the 2003-2004 season. Choosing this season avoids the distorting effects of the salary cap, introduced after the lockout of 2004-2005, on the relationship between earnings and productivity. Data on the performance of players, including total points, penalty minutes and plus-minus records refer to their regular season career statistics prior to the 2003-2004 season. The players' physical characteristics, performance and mobility data were obtained from the official guide and record book published by the National Hockey League. Most players in our sample played their entire junior (amateur) hockey in only one of the three major junior hockey leagues in Canada, in Europe, or in the US college system. A relatively small number of players played their amateur hockey in more than one system. These players were assigned to the junior system they last played in before joining the professional ranks.

To incorporate mobility into our analysis, we categorize players as either movers or stayers. Movers are players who changed teams at least once in their careers prior to the 2003-2004 season; stayers are players who played for the same team throughout their careers. The distribution of players by mobility is shown in Table 1. Approximately two thirds of the forwards and defensemen in our sample are movers while the remainder have never changed teams. In terms of the number of moves over a player's career, about one quarter of the players have changed teams only once (24% of forwards and 28% of defensemen). At the other extreme, only 15% of forwards and 12% of defensemen have moved at least four times during their career.

The group means of earnings and other variables are shown in Table 2. Without controlling for the influence of any productivity measures, the common perception holds true: movers earn more, on average, than stayers. For forwards, movers earn approximately $500,000 more than stayers while the difference is $260,000 for defensemen. This may suggest that moving, per se, has a positive effect on earnings. But does the act of changing teams, on its own, increase earnings? The suggestion is misleading because it doesn't take account of the fact that movers have higher productivity, are more aggressive and have more experience. For example, forwards who have moved have scored more points per game. Forwards and defensemen who moved have incurred more penalty minutes per game. Additionally, movers have played more games than stayers. These stylized facts suggest that movers as a group earn more than stayers not simply because they are more mobile but rather because they are more productive.

Tables 3 and 4 present some interesting details related to tenure. The tables show the distribution of the average number of games played with each team. Row (3) of Table 3 illustrates that, on average, a forward plays approximately 200 games with his first team no matter how many times he has moved during his career. (3) Two hundred games represents a little more than two seasons. Similar data for defensemen are presented in Table 4 and display more variation in the mean number of games played with the first team. The descriptive statistics in Tables 2, 3 and 4 provide a good backdrop for the more revealing regression analysis, to which we now turn.

III. Estimation Results

The OLS estimates of the earnings functions are shown in Table 5. Overall, the results related to the variables commonly used in earnings functions for the NHL are consistent with those found in the literature. Experience (GAMES, SQGAMES), points per game (PTSGAME) and reputation (STAR, DRAFT) have the expected sign and are statistically significant for both forwards and defensemen. Aggressiveness and size (PIMGAME, WEIGHT) are significant for forwards only. Further, defensemen from the Quebec Major Junior Hockey League (QMJHL) earn substantially less than defensemen from the Ontario Hockey League (OHL) who represent the omitted group in the analysis. This earnings differential is most likely related to the network effects mentioned earlier. (4)

1. Initial Results

Because our primary purpose is to examine the cumulative effects of player mobility on current earnings, the remainder of this paper focuses on the mobility variables. The first measure of mobility is a dichotomous variable (MOVER) which takes a value of one if the player changed teams at least once in his career and zero otherwise. Contrary to the previous comparison of the mean earnings of movers and stayers, the regression analysis shows that movers earn Jess, not more, than stayers after controlling for productivity and other determinants of earnings. Columns (2) and (5) in Table 5 show that forwards who moved earn 8% less than those who did not; defensemen who moved earn 20% less in the same context. (5)

As interesting as these results are, we glean more information from an analysis of the number of times a player has changed teams instead of examining simply whether or not a player has moved during his career. Columns (3) and (6) in Table 5 present regression results incorporating the number of moves. The omitted group in the regressions is STAYER, players who never changed teams during their careers. The mobility variables therein are then defined as follows: NMOB1 = 1 if a player has changed teams only one time; NMOB2 = 1 if a player has changed teams twice; NMOB3 = 1 if a player has changed teams three times; and NMOB4 = 1 if a player has changed teams four or more times. The results for forwards show that changing teams once or twice has no significant impact on earnings when compared to those who have not moved during their career. However, when a player moves a third time (NMOB3) his earnings are approximately 19% lower than stayers. Forwards who have moved four or more times (NMOB4) can expect to earn 48% less than forwards who have remained with the same team throughout their careers.

For defensemen, changing teams only once has no effect on earnings but moving more than once does. Defensemen who have moved twice (NMOB2) earn approximately 20% less than stayers; defensemen who have moved three times (NMOB3) earn 31% less while four or more moves (NMOB4 results in 59% lower earnings.

2. Player Quality Effects

The estimations related to the mobility variables above may be biased because they do not take account of the possibility that mobility may be related to unmeasurable player talent. (6) Highly talented players may stay with the same team or not move often because such unmeasurable talent is highly valued by their team.

In order to capture player quality effects we selected the top and bottom quartiles of forwards and defensemen by points per game, assuming that points per game and latent talent are related, and reestimated the earnings equations for each quartile by position. The results are presented in Table 6 for forwards and Table 7 for defensemen.

For the top quartile of forwards, the effect of mobility on earnings is complex. The coefficient for MOVER in column (1) of Table 6 is not significant for these highly talented players. However, if mobility is broken down into the number of moves rather than simply whether the player moved or not, moving may have positive or negative effects on the earnings of movers when compared to non-movers depending on the number of moves. Column (2) of Table 6 shows that forwards who move twice (NMOB2) earn 26% more than non-movers, whereas, moving four times (NMOB4) results in 41% lower earnings. For the bottom quartile of players, moving has a negative effect on earnings. The results in column (3) show that players in this quartile who move (MOVER) earn 18% less than non-movers. The results in column (4) show that the impact of mobility on earnings occurs only after two moves. The earnings of forwards who move a third time (NMOB3) are approximately 33% lower than nonmovers. Forwards who have moved four or more times (NMOB4) can expect to earn 52% less than non-movers.

These results suggest that mobility may be linked to latent talent in the case of forwards. When player quality is ignored in Table 5, moving has a negative effect on earnings with a significant impact only after two moves. The results in Table 6 suggest that in the case of less talented forwards, the effect of moving on earnings is qualitatively similar but the magnitude of the effects is larger. When looking at highly talented forwards, moving per se has no effect on earnings while moving has a positive effect after two moves and a negative effect only after four moves.

As for the defensemen, there is also evidence that there is a link between mobility and latent talent. For the top quartile of defensemen, the effect of mobility on earnings is smaller when compared to all defensemen. At the same time, the effect of mobility on earnings is larger for defensemen in the bottom quartile. The MOVER variable for the highly talented defensemen changes from having a negative effect of 22% on earnings to having no significant effect as shown in column (1) of Table 7. The qualitative effect of the number of moves is the same except that it increases in magnitude. Column (2) of Table 7 shows that movers in the top quartile earn 54% less than non-movers after three moves and 55% less after four moves. In the bottom quartile, column (3) of Table 7 shows that the MOVER coefficient morn than doubles to 41%. The estimates in column (4) of Table 7 show that the number of moves for these defensemen is larger and significant for all moves.

3. Free Agency Effects

The analysis so far has ignored the effect of free agency. Free agency as a concept was adopted by the NHL in 1975 with age as the most important determinant of eligibility. Prior to free agency, all moves, from the player's perspective, are involuntary; after it, they become voluntary. We investigate the potential effect of free agency on the relationship between moving and earnings by estimating separate regressions based on player eligibility for free agency. The age requirement under which players become free agents has been modified over the years and we have taken these modifications into account. The results are presented in Table 8 for forwards and Table 9 for defensemen.

The first three columns of Table 8 report the regression results for the subset of forwards who are not eligible for free agency. Column (2) shows, that looked at dichotomously, movers earn approximately 12% less than stayers. When the number of moves is considered, in column (3), moving only once (NMOB1) has no effect on the earnings of movers compared to stayers, but moving more than one time results in lower relative earnings: two moves (NMOB2) results in earnings that are 15% lower. The difference in earnings increases to 27% when there are three moves (NMOB3) mid to 39% when there are four or more moves (NMOB4). The last three columns relate to the subset of forwards who are eligible for free agency. Ignoring the number of moves and considering only whether they moved or not, column (5) shows that there is no difference in the earnings of movers and stayers. When the impact of the number of moves on relative earnings is considered, only forwards who have moved four or more times (NMOB4) earn less than stayers. For forwards, free agent status seems to equalize the movers and stayers as long as they move less than four times in their careers.

The qualitative results for defensemen not eligible for free agency are similar to the results for the same subset of forwards. However, the negative impact of the number of moves on earnings is larger for defensemen. The larger impact of moving on earnings may be greater for defensemen than forwards because of the specialized nature of this position that embodies relatively more nontransferable human capital. The estimates are shown in Table 9. (7) Column (2) shows that movers earn approximately 22% less than stayers. Column (3) shows that defensemen who moved more than once have lower earnings than those who have never moved: defensemen who moved two or three times (NMOB2, NMOB3) earn approximately 22% less than stayers. The disparity increases to 66% when there are four or more moves (NMOB4).

Simply looking at free agency status and player team changes begs the question of movement timing: do moves made before free agency have a different effect on earnings from those made after? To answer this question we group the forwards who are free agents according to their mobility status prior to, and after, becoming unrestricted free agents. Moreover, since it was shown in column (3) of Table 8 that the negative effect of the number of moves on the relative earnings of players not eligible for free agency occurs only after two or more moves, we redefine a stayer before free agency (STAYER1) as a player with less than two moves and a mover before free agency (MOVER1) as a player with two or more moves. This allows us to classify the players into four types and identify them by the following dichotomous variables:

1. STAYER1_STAYER = 1 if the player moved less than twice before and never moved after becoming eligible for free agency.

2. STAYER1_MOVER = 1 if the player moved less than twice before and moved at least once after become eligible for free agency.

3. MOVER1_STAYER = 1 if the player moved more than once before and never moved after becoming eligible for free agency.

4. MOVER1_MOVER = 1 if the player moved more than once before and moved at least once after becoming eligible for free agency.

The regression results are presented in column (1) of Table 10 in which the omitted group is composed of forwards who never moved: the STAYER1_STAYER category.

Generally, players who were mobile before and after free agency (MOVER1_MOVER) earn less than other players. They earn approximately 34% less than players who never moved. Further, the earnings of players who moved either before or after free agency, but not both, are not significantly different from the earnings of players who never moved.

Moving after free agency has a significant impact on current earnings only if the player moved before free agency. We gain further insight by examining the effect of the number of post free agency moves vis-a-vis simply whether or not the player moved. For the pre-free agency moves, we keep the definitions of mobility status used in column (1): STAYER1 is a player with less than two moves and MOVER1 is a player with two or more moves. For the post-free agency moves, we define the dichotomous variables MOB0 = 1 if the player never moved after free agency, MOB1 = 1 if the player moved once, and MOB2 = 1 if he moved two or more times. Column (2) of Table 8 presents the regression results. The omitted group is composed of players with less than two moves before free agency and who never moved after (STAYER1_MOB0). Moving only once after free agency has no significant effect on earnings regardless of whether or not the player moved before free agency. It is only when the player moved two or more times after free agency that there is a negative effect on earnings. The magnitude of this effect on earnings depends on whether or not these players moved before free agency. If they did not move before free agency (STAYER1_MOB2), players earn 36% less than the omitted group whereas if they moved before (MOVER1_MOB2), they earn 64% less.

To summarize, the timing of moves is important. The results in column (1) of Table I0 show that if a player moved both before and after free agency there was a negative effect on earnings. However, the results shown in column (2) reveal a more complicated relationship. Whether or not the player is a stayer before free agency, there are negative repercussions on earnings only when a player has moved more than once after free agency.

IV. Conclusion

The literature has shown mobility to be a very important determinant of salaries in the labor market outside of professional sports. The professional sports labor market is different in one major respect: movement is involuntary for the majority of a career. The basic human capital model argues that a pecuniary gain is the driving force for job change. It is primarily employee initiated. This is not so in the world of sports until complete free agent status is achieved.

This paper has presented a model outlining the effect of labor movement on earnings in a professional sport. Using career data for players in the National Hockey League prior to the 2004-2005 lockout to estimate an earnings function we find that, after controlling for human capital variables, the cumulative effect of moving on earnings is negative. Players who have changed teams during their careers earn less than those who remained with the same team throughout their career. The difference in earnings is larger for defensemen than for forwards.

We gain further insight into the negative impact that changing teams has on earnings when the number of moves is considered. For forwards, changing teams has a negative effect on earnings only if the number of moves is greater than two, while for defensemen, moving more than once has a negative effect on earnings.

Separating our players according to free agency eligibility modifies the results somewhat. For forwards who are not eligible for free agency, the negative effect of the number of moves on earnings is felt after one move instead of two. For defensemen who are not eligible for free agency, the effect of the number of moves does not change when compared with the full sample--the negative effect still occurs after the player moves twice.

As for the forwards who are eligible for free agency, players who change teams more frequently earn less than their more stable counterparts. The timing of the moves is also important. Although movers still earn less than stayers, players who have moved only before or only after free agency, earn more than players who moved both before and after.

The effect of mobility, per se, on earnings may be overestimated in our model because we are not able to incorporate unmeasurable variables. Firstly, mobility may be related to latent talent. Our results show that the negative impact of moving on earnings is particularly strong for less talented players. For highly talented players, however, movers earn less than non-movers only when they change teams often. Otherwise, there is no difference in the earnings of movers and non movers, except for highly talented forwards who earn more when they move twice. Secondly, there is something about players that move often. They do not seem to find a "fit" with a team as quickly as their more stable teammates. Perhaps they have a disruptive influence on a team, or perhaps they are difficult to coach. Mobility may be partially capturing the effect of these difficult to measure and hence omitted variables. To the extent that it does, it may be a proxy for "problem" players that have the skills to keep them in the league but may lack characteristics that endear them to their teammates and coaches.

The finding that changing teams has a negative effect on earnings only after two or more moves may be explained by the institutional environment--players sign multiyear contracts. Earnings are predetermined during a multiyear contract. Therefore, the earnings of a player changing teams will not be responsive to moves until a new contract is negotiated. A player that signs a new contract with his current team is more likely to have negotiated higher earnings than a player who is signing with a new team. An extension of this research incorporating contract length would provide insight into the true relationship between player mobility and earnings. (8)

Acknowledgements

We thank Ashley Ringuette and Tim Eastman for the collection of data. We acknowledge the financial support from the Laurentian University Research Fund.

References

Curme, Michael A. and Greg M. Daugherty. 2004. "Competition and Pay for National Hockey League Players Born in Quebec." Journal of Sports Economics 5(2): 186-205.

Dawson, Don and Lonnie Magee. 2000. The National Hockey League Entry Draft, 1969-1995: An Application of a Weighted Pool-adjacent-violators Algorithm. McMaster University, Department of Economics Working Papers.

Idson, Todd L. and Leo H. Kahane. 2000. "Team Effects on Compensation: An Application to Salary Determination in the National Hockey League." Economic Inquiry 38(2): 345-57.

Jones, J. Colin H., Serge Nadeau and William Walsh. 1997. "The Wages of Sin: Employment and Salary Effects of Violence in the National Hockey League." Atlantic Economic Journal 25(2): 191-206.

Jones, J. Colin H., Serge Nadeau and William D. Walsh. 1999. "Ethnicity, Productivity and Salary: Player Compensation and Discrimination in the National Hockey League." Applied Economics 31(5): 593-608.

Lavoie, Marc. 2000. "The Location of Pay Discrimination in the National Hockey League." Journal of Sports Economics 1(1): 401-411.

Lavoie, Marc. 2003. "The Entry Draft in the National Hockey League: Discrimination, Style of Play, and Team Location." American Journal of Economics and Sociology 62: 383-405.

Lavoie, Marc and Gilles Grenier. 1992. "Discrimination and Salary Discrimination in the National Hockey League: 1977 and 1989 compared." In Advances in the Economics of Sports, vol. 1, edited by Gerald Scully, 151-75. Greenwich, CT: JAI Press.

McPherson, Barry D. 1976. "Involuntary Turnover: A Characteristic Process of Sport Organizations." International Review for the Sociology of Sport 11(4): 5-16.

National Hockey League Official Guide and Record Book 2003. 2004. Toronto, Canada: Dan Diamond and Associates.

Vincent, Claude and Byron Eastman. 2009. "Determinants of Pay in the NHL: A Quantile Regression Approach." Journal of Sports Economics 10(3): 256-277.

Voyer, Daniel and Edward F. Wright. 1998. "Predictors of Performance in the National Hockey League." Journal of Sport Behavior 21(4): 456-473.

Notes

(1.) In this paper free agents refer to those players whose longevity in their career has enabled them to earn unrestricted free agency status. This occurs at the age of 30 to 32 depending on the collective agreement in effect. It does not refer to marginal players who have played only a few games before having reached a minimum age. The former are referred to as Group III free agents and the latter as Group VI free agents in the 1995 collective agreement between the NHL and the NHLPA.

(2.) We excluded two players from our sample. The first player was deleted because he is part of the ownership group. His team and his reported salary may not be entirely related to his performance. The second player was deleted because he signed a contract with a new team for the 2003-2004 season as a free agent for a salary substantially less than the one that he could have obtained from his previous team or from the market.

(3.) The differences across the columns are statistically insignificant.

(4.) Research by Lavoie and Grenier (1992), Lavoie (2000), and Lavoie (2003) found evidence of draft discrimination in the NHL. To test for potential draft discrimination we reestimated the models with the draft variable removed; there was no change in the sign, magnitude, nor significance of the QMJHL coefficients. Further, in a previous investigation Vincent and Eastman (2009) find that the earnings differential between defensemen in the OHL and QMJHL was significant at the upper quantiles of the conditional earnings distribution only and was also unrelated to the draft variable.

(5.) The regression coefficients on the dichotomous variables do not represent actual percentage changes when the dependent variable is expressed in natural logs. All references to percentage changes related to dichotomous variables in this paper are calculated according to the transformation formula 100[exp([??])- 1].

(6.) We thank an anonymous referee for this suggestion.

(7.) Regression results for defensemen who are eligible for free agency are not presented because there are insufficient degrees of freedom.

(8.) Contract information reflecting player history was not publicly available for the period of study.

by Claude Vincent, Assistant Professor, School of Sports Administration, Laurentian University, Sudbury, Ontario. P3E2C6. tel: 705-675-1151, x1043, fax: 705-675-4836. cvincent@laurentian.ca.

and

Byron Eastman, Professor, Department of Economics, Laurentian University, Sudbury, Ontario. P3E2C6. tel: 705-675-1151, x4267, fax: 705-675-4886. bdeastman@laurentian.ca.

TABLE 1.
Mobility Status of Players

                     All Players     Forwards    Defensemen

                      N      %      N      %      N      %

Total                625   100.0   407   100.0   218   100.0
  Movers             407    65.1   264    64.9   143    65.6
  Stayers            218    34.9   143    35.1    75    34.4
Number of moves
  0                  218    34.9   143    35.1    75    34.4
  1                  157    25.1    96    23.6    61    28.0
  2                   99    15.8    66    16.2    33    15.1
  3                   65    10.4    42    10.3    23    10.6
  [greater than or    86    13.8    60    14.7    26    11.9
   equal to] 4

TABLE 2.
Discriptive Statistics by Position and Mobility Status

                       All players            Forwards

                  Stayers     Movers    Stayers     Movers

Total               218        407        143        264
Earnings ('000)   1,538      2,013      1,519      2,105
GAMES               236.62     494.61     225.25     509.39
PTSGAME               0.37       0.40       0.43       0.48
PIMGAME               0.78       0.96       0.70       0.91
PLUSGAME              0.00      -0.02       0.00      -0.02
HEIGHT               73.18      73.24      72.82      72.76
WEIGHT              203.06     205.81     199.18     202.51
DRAFT                 0.51       0.46       0.50       0.46
STAR                  0.05       0.07       0.06       0.07
AMATEUR LEAGUE
EUROPE               83         99         54         62
OHL                  34         95         23         67
QMJHL                18         40         14         27
US COLLEGE           36        100         27         66
WHL                  47         73         25         42

                      Defensemen

                  Stayers     Movers

Total                75        143
Earnings ('000)   1,575      1,841
GAMES               258.29     467.32
PTSGAME               0.25       0.26
PIMGAME               0.94       1.06
PLUSGAME              0.01      -0.02
HEIGHT               73.88      74.13
WEIGHT              210.44     211.88
DRAFT                 0.52       0.45
STAR                  0.04       0.05
AMATEUR LEAGUE
EUROPE               29         37
OHL                  11         28
QMJHL                 4         13
US COLLEGE            9         34
WHL                  22         31

TABLE 3.
Tenure Average Number of Games by Number of
Moves Forwards

                               Moves

                     0     1     2     3     4+

Number of players    143    96    66    42    60
Total Games          225   335   493   673   692
Games per team
First                225   200   190   205   197
Second               --    135   198   195   129
Third                --    --    105   156   120
Fourth               --    --    --    117   113
Fifth                --    --    --    --     89
Sixth and higher     --    --    --    --     44

TABLE 4.
Tenure Average Number of Games by Number of
Moves Defensemen

                             Moves

                     0    1    2     3   4+

Number of players    75   61    33   23   26
Total Games         258  361   421  613  648
Games per team
First               258  222   154  194  163
Second              --   139   130  203  108
Third               --   --    137  134  120
Fourth              --   --    --    81  120
Fifth               --   --    --   --    77
Sixth and higher    --   --    --   --    60

TABLE 5.
Earnings Equations by Position by Mobility Status

                                    Forwards

                      (1)              (2)             (3)

GAMES              0.00209 ***      0.00223 ***     0.00274 ***
                  (0.000)          (0.000)         (0.000)

SQGAMES           -1.43E-06 ***    -1.49 E-06      -1.55 E-06
                  (0.000)          (0.000)         (0.000)

PTSGAME            2.015 ***        1.982 ***       1.795 ***
                  (0.000)          (0.000)         (0.000)

PIMGAME            0.0963 ***       0.0957 ***      0.0767 **
                  (0.007)          (0.007)         (0.013)

PLUSGAME           0.282            0.247           0.0308
                  (0.125)          (0.182)         (0.845)

HEIGHT             0.0241           0.0215          0.0224
                  (0.151)          (0.201)         (0.144)

WEIGHT            -0.00396 *       -0.00366        -0.00414 **
                  (0.080)          (0.106)         (0.048)

DRAFT              0.256 ***        0.247 ***       0.238 ***
                  (0.000)          (0.000)         (0.000)

STAR               0.331 ***        0.329 ***       0.203 *
                  (0.005)          (0.006)         (0.055)

QMJHL             -0.0661          -0.0748         -0.0251
                  (0.489)          (0.432)         (0.759)

WHL               -0.0318          -0.0481          0.00512
                  (0.625)          (0.465)         (0.931)

USCOLLEGE          0.00198         -0.00238         0.0445
                  (0.978)          (0.974)         (0.504)

EUROPE             0.0437           0.0311          0.0852
                  (0.490)          (0.627)         (0.145)

MOVER                              -0.0816 *
                                   (0.065)

NMOB1                                               0.0125
                                                   (0.797)

NMOB2                                              -0.0534
                                                   (0.369)

NMOB3                                              -0.213 ***
                                                   (0.003)

NMOB4                                              -0.658 ***
                                                   (0.000)

_cons             11.46 ***        11.62 ***       11.62 ***
                  (0.000)          (0.000)         (0.000)

N                407              407             407

adj. [R.sup.2]     0.750            0.751           0.798

AIC              487.9            487.6           406.6

BIC              544.0            547.8           478.8

                                  Defensemen

                      (4)             (5)             (6)

GAMES              0.00284 ***     0.00311 ***     0.00375 ***
                  (0.000)         (0.000)         (0.000)

SQGAMES           -1.74 E-06      -1.83 E-06      -2.08 E-06
                  (0.000)         (0.000)         (0.000)

PTSGAME            1.732 ***       1.672 ***       1.554 ***
                  (0.000)         (0.000)         (0.000)

PIMGAME           -0.0508         -0.0418         -0.00506
                  (0.355)         (0.441)         (0.921)

PLUSGAME           0.533 *         0.374           0.115
                  (0.067)         (0.184)         (0.634)

HEIGHT             0.0350          0.0399          0.00487
                  (0.204)         (0.152)         (0.843)

WEIGHT             0.00410         0.00391         0.00514
                  (0.298)         (0.323)         (0.139)

DRAFT              0.185 ***       0.155 **        0.160 ***
                  (0.009)         (0.030)         (0.008)

STAR               0.568 ***       0.549 **        0.380 *
                  (0.007)         (0.010)         (0.067)

QMJHL             -0.312 **       -0.300 **       -0.400 ***
                  (0.032)         (0.030)         (0.001)

WHL               -0.0648         -0.0859         -0.181
                  (0.545)         (0.405)         (0.053)

USCOLLEGE         -0.00519         0.0115         -0.135
                  (0.965)         (0.921)         (0.221)

EUROPE            -0.0199         -0.0435         -0.157
                  (0.862)         (0.698)         (0.102)

MOVER                             -0.222 ***
                                  (0.000)

NMOB1                                             -0.0746
                                                  (0.232)

NMOB2                                             -0.227 ***
                                                  (0.006)

NMOB3                                             -0.378 ***
                                                  (0.003)

NMOB4                                             -0.903 ***
                                                  (0.000)

_cons              9.444 ***       9.217 ***      11.51 ***
                  (0.000)         (0.000)         (0.000)

N                218             218             218

adj. [R.sup.2]     0.654           0.667           0.737

AIC              302.1           294.8           246.1

BIC              349.5           345.6           307.1

p-values in parentheses based on White heteroskedastic standard
errors.

* p<0.10, ** p<0.05, *** p<0.01.

TABLE 6.
Earnings Equations for Forwards by Quality by Mobility Status

                          Top Quartile

                      (1)             (2)

GAMES              0.00312 ***     0.00354 ***
                  (0.000)         (0.000)

SQGAMES           -1.88 E-06      -1.90 E-06
                  (0.000)         (0.000)

PTSGAME            1.587 ***       0.937 *
                  (0.007)         (0.072)

PIMGAME            0.102          -0.0710
                  (0.398)         (0.556)

PLUSGAME          -0.0567         -0.124
                  (0.905)         (0.754)

HEIGHT             0.111 **        0.105 **
                  (0.037)         (0.028)

WEIGHT            -0.0121 **      -0.0113 **
                  (0.026)         (0.017)

DRAFT              0.181 *         0.197 *
                  (0.094)         (0.079)

STAR               0.309 *         0.366 **
                  (0.055)         (0.028)

QMJHL             -0.166          -0.00387
                  (0.512)         (0.985)

WHL               -0.0745          0.0139
                  (0.675)         (0.925)

USCOLLEGE          0.137           0.241
                  (0.440)         (0.118)

EUROPE             0.210           0.338 **
                  (0.157)         (0.015)

MOVER             -0.00224
                  (0.986)

NMOB1                              0.0621
                                  (0.604)

NMOB2                              0.299 **
                                  (0.045)

NMOB3                             -0.119
                                  (0.464)

NMOB4                             -0.533 **
                                  (0.013)

_cons              6.749 **        7.348 ***
                  (0.035)         (0.009)

N                102             102

adj. [R.sup.2]     0.512           0.621

AIC              156.3           133.1

BIC              195.7           180.4

                        Bottom Quartile

                     (3)           (4)

GAMES              0.000750      0.00109 **
                  (0.182)       (0.031)

SQGAMES            2.920E-07     6.020E-07
                  (0.688)       (0.290)

PTSGAME            1.601 **      1.571 **
                  (0.029)       (0.018)

PIMGAME            0.0363        0.0522
                  (0.392)       (0.180)

PLUSGAME           0.419 **      0.250
                  (0.033)       (0.129)

HEIGHT             0.00773       0.0209
                  (0.711)       (0.219)

WEIGHT            -0.000976     -0.00105
                  (0.747)       (0.700)

DRAFT              0.259 ***     0.258 ***
                  (0.001)       (0.000)

STAR

QMJHL              0.0258       -0.0106
                  (0.833)       (0.915)

WHL               -0.00677      -0.00422
                  (0.939)       (0.960)

USCOLLEGE         -0.156        -0.0329
                  (0.102)       (0.684)

EUROPE            -0.0201        0.0869
                  (0.848)       (0.347)

MOVER             -0.200 ***
                  (0.002)

NMOB1                           -0.106
                                (0.136)

NMOB2                           -0.131
                                (0.160)

NMOB3                           -0.408 ***
                                (0.000)

NMOB4                           -0.730 ***
                                (0.000)

_cons             12.55 ***     11.48 ***
                  (0.000)       (0.000)

N                102           102

adj. [R.sup.2]     0.418         0.530

AIC               54.88         35.63

BIC               91.63         80.25

p-values in parentheses based on White heteroskedastic standard
errors.

* p<0.10, ** p<0.05, *** p<0.01.

TABLE 7.
Earnings Equations for Defensemen by Quality by Mobility Status

                          Top Quartile

                      (1)             (2)

GAMES              0.00322 ***     0.00433 ***
                  (0.000)         (0.000)

SQGAMES           -1.76 E-06      -2.25 E-06
                  (0.003)         (0.000)

PTSGAME            1.622 **        1.639 **
                  (0.040)         (0.047)

PIMGAME            0.216           0.124
                  (0.336)         (0.494')

PLUSGAME           0.382           0.391
                  (0.552)         (0.500)

HEIGHT             0.0499          0.00480
                  (0.418)         (0.916)

WEIGHT             0.00473         0.0113
                  (0.548)         (0.139)

DRAFT             -0.135          -0.116
                  (0.450)         (0.472)

STAR               0.353           0.137
                  (0.182)         (0.606)

QMJHL             -0.0801         -0.358
                  (0.763)         (0.263)

WHL               -0.0821         -0.191
                  (0.715)         (0.445)

USCOLLEGE         -0.239          -0.217
                  (0.314)         (0.319)

EUROPE            -0.236          -0.318
                  (0.193)         (0.107)

MOVER             -0.194
                  (0.178)

NMOB1                             -0.0630
                                  (0.645)

NMOB2                             -0.163
                                  (0.319)

NMOB3                             -0.780 **
                                  (0.026)

NMOB4                             -0.809 ***
                                  (0.006)

-cons              8.339 **       10.08 ***
                  (0.027)         (0.001)

N                 55              55

adj. [R.sup.2]     0.537           0.650

AIC               92.03           78.36

BIC              122.1           114.5

                       Bottom Quartile

                     (3)            (4)

GAMES             0.00239 ***    0.00249 ***
                 (0.002)        (0.001)

SQGAMES          -8.900E-07     -0.615
                 (0.358)        (0.609)

PTSGAME          -1.168         -1.301
                 (0.357)        (0.303)

PIMGAME          -0.110 *       -0.106 *
                 (0.096)        (0.097)

PLUSGAME          0.191          0.110
                 (0.656)        (0.785)

HEIGHT            0.0411         0.0324
                 (0.445)        (0.448)

WEIGHT            0.0000318      0.000193
                 (0.995)        (0.965)

DRAFT             0.344 ***      0.366 ***
                 (0.003)        (0.001)

STAR

QMJHL            -0.0704        -0.161
                 (0.716)        (0.292)

WHL              -0.0367        -0.0803
                 (0.760)        (0.471)

USCOLLEGE         0.0902        -0.0468
                 (0.567)        (0.761)

EUROPE           -0.0238        -0.182
                 (0.866)        (0.153)

MOVER            -0.530 ***
                 (0.000)

NMOB1                           -0.362 ***
                                (0.001)

NMOB2                           -0.564 ***
                                (0.000)

NMOB3                           -0.864 ***
                                (0.000)

NMOB4                           -0.747 *
                                (0.060)

-cons            10.41 ***      11.05 ***
                 (0.004)        (0.000)

N                55             55

adj. [R.sup.2]    0.547          0.577

AIC              45.08          43.11

BIC              73.19          77.24

p-values in parentheses based on White heteroskedastic standard
errors.

* p<0.10, ** p<0.05, *** p<0.01.

TABLE 8.
Earnings Equation of Forwards by Free Agency Eligibility

                                  Not Eligible

                      (1)              (2)             (3)

GAMES              0.00244 ***      0.00265 ***     0.00285 ***
                  (0.000)          (0.000)         (0.000)

SQGAMES           -1.20E-06 ***    -1.27 E-06      -1.25 E-06
                  (0.000)          (0.000)         (0.000)

PTSGAME            1.591 ***        1.526 ***       1.471 ***
                  (0.000)          (0.000)         (0.000)

PIMGAME            0.0688 **        0.0681 **       0.0709 **
                  (0.027)          (0.024)         (0.012)

PLUSGAME           0.219            0.165           0.0663
                  (0.217)          (0.346)         (0.668)

HEIGHT             0.00716          0.00244         0.00873
                  (0.643)          (0.875)         (0.555)

WEIGHT            -0.00563 ***     -0.00514 **     -0.00537 ***
                  (0.008)          (0.015)         (0.009)

DRAFT              0.207 ***        0.192 ***       0.202 ***
                  (0.000)          (0.000)         (0.000)

STAR               0.347 **         0.334 **        0.249
                  (0.035)          (0.040)         (0.109)

QMJHL             -0.0394          -0.0578         -0.0585
                  (0.658)          (0.502)         (0.452)

WHL                0.00346         -0.0278         -0.0330
                  (0.955)          (0.647)         (0.567)

USCOLLEGE          0.00896         -0.00506         0.0153
                  (0.901)          (0.944)         (0.825)

EUROPE             0.0409           0.0135          0.0347
                  (0.504)          (0.828)         (0.546)

MOVER                              -0.125 ***
                                   (0.007)

NMOB1                                              -0.0326
                                                   (0.527)

NMOB2                                              -0.165 ***
                                                   (0.008)

NMOB3                                              -0.311 ***
                                                   (0.000)

NMOB4                                              -0.501 ***
                                                   (0.000)

-cons             13.14 ***        13.45 ***       13.01 ***
                  (0.000)          (0.000)         (0.000)

N                312              312             312

adj. [R.sup.2]     0.762            0.767           0.787

AIC              271.9            267.1           241.2

BIC              324.3            323.2           308.6

                                   Eligible

                     (4)             (5)              (6)

GAMES              0.00256 **     0.00260 **       0.00290 ***
                  (0.023)        (0.024)          (0.001)

SQGAMES           -1.93 E-06     -1.95E-06 ***    -1.91 E-06
                  (0.003)        (0.004)          (0.000)

PTSGAME            2.986 ***      2.972 ***        2.601 ***
                  (0.000)        (0.000)          (0.000)

PIMGAME            0.164          0.165            0.0435
                  (0.177)        (0.178)          (0.683)

PLUSGAME           0.253          0.228           -0.379
                  (0.712)        (0.744)          (0.524)

HEIGHT             0.104 **       0.104 **         0.0705 *
                  (0.036)        (0.037)          (0.100)

WEIGHT            -0.00436       -0.00433         -0.000708
                  (0.525)        (0.529)          (0.902)

DRAFT              0.302 **       0.298 **         0.273 **
                  (0.018)        (0.020)          (0.024)

STAR               0.0411         0.0344          -0.0581
                  (0.832)        (0.861)          (0.790)

QMJHL              0.0567         0.0663           0.122
                  (0.813)        (0.785)          (0.546)

WHL               -0.0554        -0.0551           0.153
                  (0.759)        (0.763)          (0.376)

USCOLLEGE          0.0524         0.0652           0.0931
                  (0.766)        (0.722)          (0.559)

EUROPE             0.0106         0.0248           0.0696
                  (0.955)        (0.900)          (0.659)

MOVER                            -0.0858
                                 (0.427)

NMOB1                                              0.0897
                                                  (0.541)

NMOB2                                              0.191
                                                  (0.258)

NMOB3                                             -0.0237
                                                  (0.871)

NMOB4                                             -0.589 ***
                                                  (0.000)

-cons              4.953 *        5.019 *          6.885 ***
                  (0.070)        (0.067)          (0.004)

N                 95             95               95

adj. [R.sup.2]     0.712          0.708            0.783

AIC              165.2          167.1            141.4

BIC              201.0          205.4            187.4

p-values in parentheses based on White heteroskedastic standard
errors.

* p<0.10, ** p<0.05, *** p<0.01.

TABLE 9.
Earnings Equation of Defensemen Not Eligible For Free Agency

                      (1)             (2)             (3)

GAMES              0.00286 ***     0.00322 ***     0.00348 ***
                  (0.000)         (0.000)         (0.000)

SQGAMES           -1.51 E-06      -1.66 E-06      -1.79E-06 ***
                  (0.002)         (0.000)         (0.000)

PTSGAME            1.790 ***       1.712 ***       1.760 ***
                  (0.000)         (0.000)         (0.000)

PIMGAME           -0.0156         -0.00444        -0.0140
                  (0.785)         (0.938)         (0.786)

PLUSGAME           0.298           0.115          -0.126
                  (0.319)         (0.684)         (0.576)

HEIGHT             0.0170          0.0268          0.00144
                  (0.452)         (0.250)         (0.943)

WEIGHT             0.00590         0.00517         0.00596 *
                  (0.122)         (0.178)         (0.062)

DRAFT              0.175 **        0.141 **        0.160 *
                  (0.017)         (0.048)         (0.005)

STAR               0.357           0.327           0.438 *
                  (0.227)         (0.270)         (0.090)

QMJHL             -0.233          -0.191          -0.314 **
                  (0.136)         (0.196)         (0.018)

WHL                0.00941        -0.0145         -0.110
                  (0.936)         (0.898)         (0.259)

USCOLLEGE          0.119           0.166          -0.0371
                  (0.349)         (0.186)         (0.751)

EUROPE             0.0689          0.0380         -0.0841
                  (0.543)         (0.730)         (0.406)

MOVER                             -0.243 ***
                                  (0.000)
NMOB1                                             -0.0874
                                                  (0.147)

NM0B2                                             -0.253 ***
                                                  (0.003)

NM0B3                                             -0.250 *
                                                  (0.079)

NM0B4                                             -1.091 ***
                                                  (0.000)

-cons             10.26 ***        9.767 ***      11.52 ***
                  (0.000)         (0.000)         (0.000)

N                170             170             170

adj. [R.sup.2]     0.690           0.711           0.791

AIC              194.2           183.5           131.1

BIC              238.1           230.5           187.5

p-values in parentheses based on White heteroskedastic standard
errors.

* p<0.10, ** p<0.05, *** p<0.01.

TABLE 10.
Earnings Equations of Forwards Eligible for
Free Agency

                      (1)              (2)

GAMES              0.00353 ***      0.00327 ***
                  (0.001)          (0.001)

SQGAMES           -2.31E-06 ***    -1.93E-06 ***
                  (0.000)          (0.000)

PTSGAME            2.695 ***        2.543 ***
                  (0.000)          (0.000)

PIMGAME            0.0598           0.129
                  (0.610)          (0.142)

PLUSGAME          -0.291            0.298
                  (0.669)          (0.641)

HEIGHT             0.0914 *         0.0491
                  (0.054)          (0.165)

WEIGHT            -0.00144         -0.000254
                  (0.844)          (0.967)

DRAFT              0.242 *          0.0869
                  (0.080)          (0.487)

STAR               0.0464          -0.0488
                  (0.830)          (0.777)

QMJHL              0.0576          -0.0260
                  (0.790)          (0.868)

WHL               -0.0734          -0.0575
                  (0.660)          (0.653)

USCOLLEGE          0.00285          0.121
                  (0.987)          (0.348)

EUROPE            -0.0281          -0.0152
                  (0.866)          (0.911)

MOVER1_STAYER      0.0480
                  (0.703)

STAYER1_MOVER     -0.0208
                  (0.920)

MOVER1_MOVER      -0.421 **
                  (0.013)

STAYER1_MOB1                        0.0319
                                   (0.908)

STAYER1_MOB2                       -0.445 **
                                   (0.041)

MOVER1_MOB0                        -0.0152
                                   (0.881)

MOVER1_MOB1                        -0.171
                                   (0.257)

MOVER1_MOB2                        -1.018 ***
                                   (0.000)

_cons              5.278 **         8.221 ***
                  (0.038)          (0.000)

N                 95               95

adj. [R.sup.2]     0.743            0.814

AIC              156.7            127.3

BIC              200.2            175.8

p-values in parentheses based on White heteroskedastic standard
errors.

* p<0.10, ** p<0.05, *** p<0.01.