Did the 2005 collective bargaining agreement really improve team efficiency in the NHL?

Buschemann, Arne ; Deutscher, Christian

Introduction

At the end of the 2003-2004 season, the National Hockey League faced serious financial challenges. Player salaries consumed more than 65% of the generated revenues. Hence, more than 20 of the 30 teams were claiming monetary losses. Small market teams in particular suffered from the steady increase in player salaries; they were unable to compete with big market teams for top players and their generous player contracts. In addition, attendance figures decreased to a 4-year low. Consequently, top free agents were signed by teams with large revenue streams, many of which were located in big markets. There was no new agreement between the team owners and the NHL Players' Association (NHLPA) in sight when the 1995 CBA expired. Team owners demanded cost certainty for their teams, whereas the NHLPA initially refused to install salary restrictions in terms of a salary cap. Subsequently, a novelty in hockey sport history occurred when the team owners announced a lockout and eventually cancelled the entire 2004-2005 season. A new agreement was reached in July 2005 and contained several novelties, including a hard payroll cap as well as a revenue sharing plan. The ultimate goal of these measures was to restore financial competitiveness which, as it is proposed, should a priori help financially weak teams to be more competitive. In detail, the CBA includes an upper limit salary cap as well as a lower limit salary floor. Also, the revenue sharing plan is intended to allow low revenue producing teams be more financially competitive. In order to do this, the top ten teams contribute money to a pool where a minimum of 4.5% of league revenues are to be distributed among the bottom 15 teams.

In order to analyze the financial situation of NHL teams before and after the CBA, and to measure the impact of the CBA, this study applied a stochastic frontier analysis (SFA). The objective is to provide empirical evidence on whether or not the new CBA did indeed strengthen financial competitiveness. These impacts of an institutional change can be best observed by analyzing the team efficiency. This methodology has been widely used in the field of sports economics. The most popular choice of output indicators used in reviewed literature has been the sporting performance (measured as wins), winning percentage, or achieved points in a given season. For European soccer, Dawson, Dobson, and Gerrard (2000) applied SFA to estimate the efficiency of managers in English professional soccer. In addition, Frick and Simmons (2008) used SFA to measure the effect of variations in managerial compensation on organizational team success in the German Bundesliga. Barros and colleagues (Barros, Del Corral, & Garcia-del-Barrio, 2008; Barros, Garcia-del-Barrio, & Leach, 2009) analyzed technical efficiency of football clubs in the Spanish Primera Division as well as in the English Premier League with a random frontier model. Concerning U.S. team sport franchises, Zak, Huang, and Siegfried (1979) were the first to analyze efficiencies of 5 NBA teams with a Cobb-Douglas deterministic frontier model. Hofler and Payne (1997) extended this approach and examined a cross-sectional analysis of all 27 NBA teams for the 1992-1993 season in order to observe if teams play up to their potential in terms of actual wins. In a subsequent study, Hofler and Payne (2006) used panel data for the stochastic production frontier model.

In a different strand of literature, Kahane (2005) applied SFA for the NHL and identifies technical inefficiency in production. His results indicate that franchises owned by corporations tend to be more efficient than franchises owned by individuals, and teams with a greater relative presence of French-Canadian players tended to be less efficient. In a similar direction, Fort, Lee, and Berri (2008) applied SFA to address the issue on discrimination in retention of NBA coaches and detected no difference in technical efficiency by race of the coach.

Because team owners postulated over ways to reach cost certainty through the 2005 NHL CBA, it seems obvious that team owners are not solely interested in success on the ice and the glory of victory. From the team owners' perspective, it is imperative that the franchise achieves a positive return on their investment. The present study explores the relationship between the 2005 NHL CBA and the financial success of franchise teams relative to their potential. By using team values, as well as franchises' revenues, as outputs to measure technical efficiencies, the study focuses on economic efficiency. Previous studies, as shown above, predominantly used sporting performance as the output variable for measuring team success. But a sport, even though advocated differently on a regular basis, is not just about winning games. The franchise system of the NHL--which, as stated above, struggled heavily right before the lockout--has to make sure that teams operate efficiently, in terms of financial performance, to ensure the future of the league. Because of this, we deviate from the existing literature by introducing financially important outcome variables.

One method has been ignored in the literature so far: using team values as well as revenues as outputs for measuring technical efficiencies. Thus, the current research is innovative in this context. Efficiency can also be used as a direct benchmark between franchises operating in the same institutional environment. Our article closes this gap through the analysis of the impact of the new CBA on efficiency; specifically, team value maximization and value generation of low performing teams immediately increased efficiencies after the lockout.

Data Description

The data we used includes information on the four seasons prior to the lockout, from 2000-2001 to 2003-2004, and the four seasons immediately following the lockout, from 2005-2006 to 2008-2009. At the beginning of the 2000-2001 season, the NHL expanded from 28 to 30 teams as the Minnesota Wild and the Columbus Blue Jackets joined the league. Because we took the previous season into account, the resulting (unbalanced) panel dataset contained 238 observations on all variables included in the estimates.

Frontier models require identifying inputs and outputs. In order to determine how efficiently the franchises operated, it was essential for us to use a financial ratio as the output. Forbes magazine reports data annually on the sport franchises' team values, as well as the revenues for all major leagues. It breaks down franchise valuation into four categories: sport, market, stadium, and brand management. Team value has been previously applied as a dependent variable to analyze determinants of franchise values (Alexander & Kern, 2004; Humphreys & Mondello, 2008). Therefore, as franchise values are not equally distributed, this study applied the natural logarithm of team values as output. Furthermore, to ensure the robustness of our results, we used the natural logarithm of revenues for each franchise as a second output. This data is also published by Forbes magazine on a yearly basis.

The input variables represented the various factors that were most likely to determine a team's franchise value. Therefore, we included the natural logarithm of the population of each team's metropolitan area in order to account for market-size effects on franchise values. In metropolitan areas with more than one NHL franchise (e.g., Los Angeles and New York), each franchise was credited with the entire population in the metropolitan area--this is because the market cannot be unambiguously separated between each franchise. Data were obtained from the U.S. Bureau of Economic Analysis' Regional Economic Accounts and Statistics, Canada. Since franchises share larger pools of potential fans, we expected a positive relationship between teams located in larger markets and franchise values as well as revenues. It should be noted that, unlike in European soccer, only very few fans join their favorite teams for road games. This is due in part to a greater number of games and a lengthier distance between competing teams.

The team's stadium is another important input factor for multiple reasons. A franchise with a new stadium can expect higher revenues, and hence higher team values, due to state-of-the-art luxury boxes, for example. (1) Hence, we included stadium age, as well as stadium age in quadratic form, in our analysis and expected a negative impact of arena age and an increase in marginal returns on both dependent variables. Data on arena age were collected from Munsey and Suppes' website (http://www.ballparks.com). In addition, the natural logarithm of attendees per game was included. We assumed that, since each attendee generates revenue for the franchise, the higher the number of attendees, the greater the team value. To measure this revenue stream, we used the team marketing annual reports from the Fan Cost Index (FCI), which are constructed for each franchise and year. (2) The FCI tracks the cost of attending a sporting event for a family of four. (3) The more a franchise is able to charge for their tickets and other amenities, the more revenues they generate. Thus, we presumed that the coefficient for the FCI would also be positively related to the team value. To analyze how franchise history affects team value, we included the duration of a team in the league and the squared duration of a team in the league. We expected that teams with a longer franchise history also reported a higher team value. (4)

We also controlled for the athletic achievements of a team. Since NHL standings are based on points and not wins, the rank is not expressed in winning percentages; this is because teams gain a point for an overtime loss. We estimated athletic achievement by dividing the team's total from the previous season by the average points of all teams in the previous season. Following the approach by Miller (2007), points achieved in the previous season are considered to be an important component in determining ticket prices, season ticket sales, media revenues, and advertising prices. We expected a positive coefficient, suggesting that a better athletic achievement in the previous season leads to higher revenues and, therefore, a higher franchise value. One of the most important input factors in professional sports is team expenses. We measure these by including the natural logarithm of the team payroll in our analysis. Data were drawn from USA Today (http://content.usatoday.com/sports/hockey/nhl/salaries/default.aspx). We assumed that a team with high payroll expenses would offer a superior team quality and, therefore, would provide a better utility to fans. Due to this assumption, we anticipated that higher team expenses would positively influence the team value. All monetary magnitudes in this analysis (e.g., team value, FCI, payroll) were deflated by the CPI, which was taken from the U.S. Bureau of Labor Statistics and denoted at prices for the year 2000. Descriptive statistics for all variables introduced above are shown in Table 1.

Empirical Analyses of Efficiencies

We applied a stochastic production frontier model to explore whether the new CBA did indeed improve technical efficiencies within the league. In the present study, the output of the teams in the NHL was measured by the team values as well as revenues after the respective season. To compute technical efficiencies, we applied the model introduced by Battese and Coelli (1995), which allows for time-varying efficiencies. It assumes a log-linear production function for a set of i firms over t time periods and can be presented as follows: (5)

[y.sub.it] = [x.sub.it] [beta] ([v.sub.it] - [u.sub.it]) i = 1, ..., N and t = 1, ... [tau]. (1)

Where [y.sub.it] is the natural logarithm of the franchise value, is a vector of team-specific input quantities, and [beta] is a vector of unknown coefficients over which the likelihood will be maximized. Furthermore, [v.sub.it] represents a random error term that is assumed to be independent and identically distributed (i.i.d.) N(0,[[sigma].sub.v.sup.2)] is i.i.d. and a non-negative random error term that accounts for technical inefficiency in production, and it is further assumed to follow a normal distribution truncated at zero of the N([m.sub.it],[[sigma].sub.u.sup.2)]) distribution. [m.sub.it] is given as

[m.sub.it] = [z.sub.it.sup.[sigma]] (2)

[z.sub.it] is a vector of variables that may influence the efficiency as team value creation, and is a vector to be estimated. Using data from the NHL from 2000-2001 to 2008-2009, we accomplished a total of 238 observations for 30 teams. Table 2 presents the maximum likelihood estimate for our frontier model for franchise values, while Table 3 presents results for revenue generation. All results are robust and not vulnerable to either multicollinearity or heteroskedasticity.

The coherence of the metropolitan area population and both dependent variables was as expected: Market size indeed showed a positive impact. It was not a surprise that both variables--the age of the arena and the squared age--were significant. Although both significance levels were different for the models, both dependent variables decreased as the years played in the facility increased--that is, the arena is not considered to be state-of-the-art after a few seasons, which again reduces fan interest. The positive impact of the squared term can be attributed to relatively old stadiums accommodating nostalgic memories of team history; Madison Square Garden in New York City, for example, is a historic arena that, while not belonging to the most modern arenas around the league, still arouses spectators' interest. As duration in the NHL depicts the tradition of a team, only the squared term significantly impacted both dependent variables. This can be explained by team tradition, which cannot be established within a short period of time. The negative effect of duration on the nonquadratic term can be explained by the honeymoon effect, which diminished after the inauguration. (6) Although these indicators of arena and team history had the expected impact, sporting performance in the previous season apparently has not--that is, it did not impact team value or revenues generated in a significant way. In our model, both indicators for match day revenues affected our dependent variables in a positive way: Average attendance and the FCI exhibit had statistically significant impacts. Finally, the team payroll-depicting team quality and serving as an indicator for the asset the squad displays-has the expected positive impact on both team values and revenue.

After providing insights on indicators influencing team values in the NHL, and establishing a basis for calculating efficiencies for each team and each season, we pursued the initial inquiry to determine whether the lockout in 2004-2005 improved these efficiencies. Figure 1 and Figure 2 provide information on the average efficiencies for a particular season on the 10 most efficient teams, the 10 least efficient teams, and the 8-10 teams in between. One can easily observe that efficiencies increased immediately after the lockout season, providing clear support for the thesis that the new CBA indeed increased efficiencies. In particular, the low performing teams took advantage of the new CBA to close the gap to the high performing teams. This is true for both models, as average efficiency for the 10 least efficient teams improved from 0.73 to 0.84 for Model 1, and from 0.72 to 0.86 for Model 2. Once the new CBA was established, average efficiencies leveled off approximately 7% higher than before the lockout for Model 1 and approximately 9% higher than before the lockout for Model

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

2. The expectation stated in the introduction, which claims a strengthened competitiveness due to the new CBA, was supported by our estimations.

Conclusions

This article has investigated the impact of the new CBA on efficiencies concerning maximizing team values as well as revenue generation. After the lockout season in 2004, we observed an abrupt increase in technical efficiencies after the new collective bargaining agreement was installed--particularly concerning low performing teams benefitting from salary restrictions and revenue sharing.

As our study is the first to use team values and revenues in connection with measuring technical efficiencies of teams, several follow-up questions arise. For example, it would be of great interest to explore if team efficiencies benefit or suffer from having another major league team in the city; in other words, analyzing whether a tougher competition within a local market would serve as a catalyzer and lead to an increase in managerial performance. Going into more detail, it would be interesting to see if a certain combination of major league teams could serve as substitutes or complements. As our analysis provides the somewhat surprising result that sporting performance neither influences team values nor team revenues, future research could compare other major leagues to examine its potential influence in other sports.

References

Alexander, D., & Kern, W. (2004). The economic determinants of professional sports franchise values. Journal of Sports Economics, 5, 51-66.

Barros, C., Del Corral, J., & Garcia-del-Barrio, P. (2008). Identification of segments of soccer clubs in the Spanish league first division with a latent class model. Journal of Sports Economics, 9, 451-469.

Barros, C., Garcia-del-Barrio, P., & Leach, S. (2009). Analysing the technical efficiency of the Spanish football league first division with a random frontier model. Applied Economics, 41, 3239-3247.

Battese, G., & Coelli, T. (1995). A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empirical Economics, 20, 325-332.

Dawson, P., Dobson, S., & Gerrard, B. (2000). Estimating coaching efficiency in professional team sports: Evidence from English Association Football. Scottish Journal of Political Economy, 4, 399-421.

Fort, R., Lee, Y., & Berri, D. (2008). Race and coaching efficiency in the NBA. International Journal of Sport Finance, 3, 84-96.

Frick, B., & Simmons, R. (2008). The impact of managerial quality on organizational performance: Evidence from German soccer. Managerial and Decision Economics, 29, 593-600.

Hofler, R., & Payne, J. (1997). Measuring efficiency in the National Basketball Association. Economics Letters, 55, 293-299.

Hofler, R., & Payne, J. (2006). Efficiency in the National Basketball Association: A stochastic frontier analysis with panel data. Managerial and Decision Economics, 27, 279-285.

Humphreys, B., & Mondello, M. (2008). Determinants of franchise values in North American professional sports leagues: Evidence from a hedonic price model. International Journal of Sport Finance, 3, 98-105.

Kahane, L. (2005). Production efficiency and discriminatory hiring practices in the National Hockey League: A stochastic frontier approach. Review of Industrial Organization, 27, 47-71.

Leadley, J., & Zygmont, X. (2005). When is the honeymoon over? Attendance trends in the National Basketball Association, 1970-2003. Journal of Sports Economics, 6, 203-221.

Leadley, J., & Zygmont, X. (2006). When is the honeymoon over? National Hockey League Attendance, 1970-2003. Canadian Public Policy, 32, 213-232.

Miller, P. (2007). Private financing and sports franchise values: The case of major league baseball. Journal of Sports Economics, 8, 449-67.

Scully, G. (1989). The business of Major League Baseball. Chicago: University of Chicago Press.

Zak, T., Huang, C., & Siegfried, J. (1979). Production efficiency: The case of professional basketball. American Economist, 25, 19-33.

Endnotes

(1) See Alexander and Kern (2004) and Miller (2007) for research of the impact of playing in new stadiums on franchise values.

(2) Information is available at Rodney Fort's website at http://www.rodneyfort.com.

(3) The FCI comprises the prices of four average-price tickets, two small draft beers, four small soft drinks, four regular-size hot dogs, parking for one car, two game programs, and two least-expensive, adult-size adjustable caps.

(4) We did not include the natural logarithm of the duration and age arena input variables since the value of 0 is not defined.

(5) The applied software for the frontier analysis is Stata11 SE.

(6) The honeymoon effect is an increase in attendance after the opening of a new facility, which fades after some time. For literature in major league sports, see for example Leadley and Zygmont (2005, 2006) or Scully (1989).

Arne Buschemann and Christian Deutscher (1)

(1) University of Paderborn, Germany

Christian Deutscher is a postdoctoral research and teaching assistant in the Department of Management at the University of Paderborn, Germany. He studied economics at the University of Bonn, Germany. His research focuses on sports and personnel economics.

Arne Buschemann is a doctoral student in the Department of Management at the University of Paderborn, Germany. He studied international business studies at the University of Paderborn, Germany. His research focuses on sports and personnel economics.

Table 1. Descriptive Statistics of Indicators Influencing Team Values

Variable Operationalization Mean Minimum Maximum

Log value Natural log of the 18.88 18.18 19.78
 team value in dollars
Log population Natural log of 15.13 13.64 16.76
 metropolitan area
 population
Age arena Tenure of the team 12.38 0 47
 in the arena
Age arena (2) Squared tenure of 270.0 0 2,209
 the team in the arena
Duration Duration of the team 34.45 0 99
 in the league
Duration (2) Squared duration of 1,964 0 9,801
 the team in the league
Relative points Achieved points in 1 0.45 1.37
 previous season/average
 points
Log attendance Natural log of 9.72 9.18 9.97
 attendance
Log FCI Natural log of the fan 5.46 4.98 5.98
 cost index
Log pay Natural log of the 17.40 16.28 18.11
 team payroll

Table 2. Stochastic Frontier Estimate for the Dependent
Variable Log(Value)

Variable Coefficients f-Value

Log population .1076 7.36 ***
Age arena -.0130 -3.25 ***
Age arena (2) .0003 3.24 ***
Duration -.0018 -1.00 (+)
Duration (2) .0001 3.62 ***
Relative points .1100 1.46 (+)
Log attendance .7048 7.00 ***
Log FCI .1333 1.94 *
Log pay .2955 5.84 ***

Number of observations 238
Log likelihood 86.5
Chi square 681.83
Probability 0.000

Note. *, **, and *** denote statistical significance at the
0.01, 0.05, and 0.1 level; (+) denotes insignificance.

Table 3. Stochastic Frontier Estimate for the Dependent
Variable Log (Revenue)

Variable Coefficients f-Value

Log population 0.0467 3.63 ***
Age arena -0.0067 -1.86 *
Age arena (2) 0.0002 2.06 **
Duration -0.0032 -1.97 **
Duration (2) 0.0001 4.03 ***
Relative points -0.0432 -0.62 (+)
Log attendance 0.7718 8.43 ***
Log FCI -0.0793 -1.29 (+)
Log pay 0.3609 7.65 ***
Number of observations 238
Log likelihood 107.4
Chi square 505.44
Probability 0.000

Note. *, **, and *** denote statistical significance at the
0.01, 0.05, and 0.1 level; (+) denotes insignificance.