The effects of labour strikes on consumer demand in professional sports: revisited

Previous research has concluded that the 1981 and 1994–1995 Major League Baseball (MLB) strikes have caused short-term losses in attendance but have not resulted in any long-term effects on attendance. While total attendance at MLB games following the 1994–1995 strike has recovered to its pre-strike levels, this has been done only through the construction of new stadiums at an unprecedented pace which cannot continue into the future. After accounting for stadium effects, average MLB baseball attendance has dropped significantly since the 1994–1995 strike.

In their previous work (Schmidt and Berri, [2], [3]), Martin Schmidt and David Berri question whether labour strikes in professional sports in the United States have resulted in permanent shifts in consumer demand. Their analysis of the 1981 and 1994–1995 strikes in Major League Baseball (MLB) as well as the 1994 National Hockey League (NHL) and 1982 and 1987 National Football League (NFL) strikes suggest that these events have had only a temporary effect on attendance and not permanently harmed the leagues. The apparent conclusion is that sports teams are deeply ingrained into consumer preferences and that leagues sell a relatively unique product with no close substitutes so that strikes have little bearing on future demand. This result is surprising in that it contradicts the popularly held notion that millionaire players arguing with billionaire owners about compensation is inherently distasteful to the average sports fan, and therefore strikes lead to consumer dissatisfaction and lower future attendance. For example, while the popular media suggest that the 2004–2005 NHL lockout, which led to the termination of the entire 2004–2005 season as well as the cancellation of the Stanley Cup for the first time since 1919, may result in a significant weakening or even the death of the NHL, Schmidt and Berri's results suggest no such long-run concerns.

Schmidt and Berri's results also have importance outside the field of professional sports. It is easy to envision that strikes, particularly in service industries, may have significant effects on future consumer demand. For instance, the inconvenience suffered by a traveller caught in an airline strike may prompt that traveller to avoid that particular airline in the future. If Schmidt and Berri's results hold outside the sports industry, however, labour and management need to consider only the short-run costs of labour strife but not any future ramifications since strikes cause no long-run changes in consumer demand.

The evidence is relatively clear that in most cases sports strikes have caused only temporary disruptions in demand. For example, the 1981 MLB strike resulted in the loss of 717 games (34% of the season total). The average attendance per team fell from 1 654 390 in 1980 to 1 020 938 in 1981. With the full resumption of games in the 1982 season, however, attendance more than recovered to an average of 1 714 918 per team (see Fig. 1). Schmidt and Berri's analysis of the 1982 and 1987 NFL strikes and the 1994 NHL lockout also confirm the hypothesis that labour disputes have only caused temporary disruptions in fan interest. While not formally analysed by Schmidt and Berri, a cursory look at the 1998 NBA lockout also shows the NBA attendances quickly returned to their pre-lockout levels following the resumption of games.

Graph: Fig. 1. Major League Baseball attendance, 1901–2003

The 1994–1995 MLB strike appears to be an exception. The league has never recovered to the attendance record of 2 509 159 per team set in 1993, the season immediately prior to the strike. The 1994–1995 strike was unusual in several ways. First, the strike was the second one for the league that resulted in a significant loss of games. Second, the strike was exceptionally long, indeed until the 2004–2005 NHL lockout the longest interruption in major American professional sports history, and spanned parts of two seasons. Third, the strike resulted in the loss of the 1994 post-season and the cancellation of the World Series for the first time since 1904. Again, until the 2004–2005 NHL lockout, no other strike in any of the major American sports had resulted in the loss of post-season games.

Even despite the lack of recovery to the 1993 attendance heights, Schmidt and Berri argue that the 1994–1995 strike still had no permanent effects on MLB attendance. They suggest that the 1993 season was an anomaly due, in part, to a remarkably successful expansion in the league. In 1993, the league added teams for the first time in 16 years including the Colorado Rockies who drew a league record 4.5 million fans in their inaugural season. Instead, the post-strike period should be compared with the comparatively normal 1992 season when MLB drew only 2 148 934 fans per team. While average attendances fell by 350 000 in both 1994 and 1995 compared with 1992, league attendances had recovered to their 1992 levels by the 1996 season and continued to grow from there. By this reasoning, fans' threats to stay away from the game following labour unrest are again shown to not be credible. Overall, Schmidt and Berri ([2]) conclude that 'the evidence prevented suggests that although the most protracted periods of labor discord had short-term impacts on attendance, there is no empirical evidence that these exogenous shocks had any long-term effects.'

While it is undeniable that MLB attendances in the period after the 1994–1995 strike did, in fact, rise to nearly their pre-strike levels, it is not nearly so clear that the strike had no permanent effects on attendance after accounting for other factors affecting professional baseball in the past decade, in particular the construction of new MLB infrastructure during the 1990s and early 2000s. The past 15 years have witnessed an unprecedented transformation of baseball's infrastructure, and by the 2004 season, 19 of MLB's 30 teams were playing in facilities constructed or significantly renovated since 1989. This pace of construction cannot possibly continue into the future. Indeed, by the end of 2004, only one new facility was under construction (in St. Louis), the lowest number since the early 90s. See Table 1 for a list of newly built or renovated stadiums.

Table 1. New stadiums and attendance effects since 1989

New stadiums nearly always lead to at least a temporary increase in attendance for the host city. The duration of the increase in attendance varies from team to team. When a new stadium is combined with renewed success by the home team on the field of play (as in the case of the Cleveland Indians following the 1994 construction of Jacob's Field), the increase in attendance may last for many years while the 'novelty effect' of a new facility can be as little as a single year when the host team is unable to improve its win/loss record (as in the case of the Detroit Tigers and Milwaukee Brewers in 2000 and 2001). In all cases, new stadiums result in at least a one season significant attendance boost. In the 14 stadiums constructed for non-expansion teams between 1989 and 2004, the increase in attendance (after accounting for strike years) between the debut season and the previous season averaged 780 000 fans or roughly 34%. This figure even understates the true impact of new stadiums on attendance since franchises, particularly older teams such as the Chicago White Sox, Cleveland Indians and Detroit Tigers, often experience a bump in attendance, or a 'nostalgia effect', in the last season played in an old stadium. Over the same period, the increase in attendance between the debut season for a new stadium and the season two years prior averaged over 1 050 000 fans or over 50%. See Table 1.

It is possible that Major League Baseball has managed to avoid declining attendance due to fan dissatisfaction with labour/management strife only through the provision of expensive new stadiums. If the novelty effect from new stadiums is temporary in nature, MLB has only managed to postpone the inevitable decline in attendance as a result of baseball's frequent strikes as the recent pace of stadium construction cannot continue into the future. Furthermore, the strikes may have resulted in a wasted opportunity for MLB to solidify its hold on the American sports psyche. Instead of increasing the popularity of the sport, the $5 billion in new stadium construction (two-thirds of which was provided by the public sector) simply allowed MLB to keep from falling backwards. In the next sections, the 1981 and 1994–1995 strikes are re-examined more closely paying special attention to the issue of stadium construction or renovation.

Data on team by team and total MLB attendance were obtained from Major League Baseball for the 1901 to 2003 seasons. As in Schmidt and Berri ([3]), to account for expansion, the total MLB attendance data are divided by the number of teams in the leagues, and the first difference of the data is taken for each season in order to correct for non-stationarity of the original time series. With the exception adding three seasons at the end of the data set, the data are roughly comparable to that used by Schmidt and Berri.

Again following the procedure outlined in Schmidt and Berri, the attendance data are analyzed using intervention analysis as in Box and Tiao ([1]). Schmidt and Berri consider the following equation:

Graph

where yt represents the first-differenced average attendance, z(46) is a dummy variable representing the exogenous shock of American servicemen returning from the First World War to the baseball stands, and z(81) and z(82) are dummy variables representing the 1981 strike year and the 1982 return to normalcy year. z(93) is a dummy variable accounting for the unusual expansion year in 1993 immediately prior to the strike, z(94) is a dummy variable modelling the effect of the strike in 1994, and z(96) is a dummy variable for the strike recovery in 1996. The results of OLS regression on the data are found in Table 1 and listed as model 1.

While the magnitude of the of z(96) coefficient is much smaller than that of the z(94) coefficient, indicating that the size of the attendance recovery in 1996 was less than that of the initial attendance drop in 1994 due to the strike, Schmidt and Berri argue that the size of the attendance drop in 1994 was exaggerated due to usually large attendances in 1993. A Wald test performed on the z(93), z(94) and z(96) coefficients finds that a null-hypothesis that the absolute value of the z(94) coefficient is equal to the sum of the z(93) and z(96) coefficients cannot be rejected. In other words, as found by the previous authors, it is reasonable to conclude that the size of the attendance drop in 1994 was matched by a combined attendance gain of an equal magnitude in the years 1993 and 1996.

An alternative model that answers whether the attendance effects of the 1994–1995 strike were permanent in nature is tested by adding to model 1 an additional intervention dummy variable, z(94–03), that takes a value of 1 for the entire 1994–2003 period and 0 elsewhere. The coefficient on the z(94–03) variable shows that average attendance during the 1994–2003 period, after accounting for the actual strike and recovery years, grew at a statistically insignificant rate of 108 fans per year faster following the strike than before the strike. A comparison of models 1 and 2 clearly shows model 1 as the superior explanatory model lending credence to Schmidt and Berri's hypothesis that the 1994–1995 strike likely had only temporary effects on MLB attendance.

Next, the two competing models can be re-examined while considering the effects of stadium construction on average attendance. An additional 'Stadiums' variable is added to Equation 1 representing the number of new stadiums opened in MLB during the corresponding year. An analysis of the data shows that prior to 1960, the construction of new stadiums seems to have had no effect on attendance for the tenant team, perhaps due to the limited number of additional amenities provided by new stadiums prior to this period, and, therefore, only the construction of stadiums since 1960 are included in the stadiums variable. When expansion franchises enter into the league, the stadium is counted as a new stadium only if it was newly constructed as a permanent facility for the new franchise. Finally, in cases where a facility opened during the middle of a season, such as Safeco Field in Seattle, the stadium is counted in the stadium variable as a new stadium only in the year it opened. Alternatively, the facility could be counted as one-half of a new facility in the year it opened and one-half of a new facility in the following year when it is available for an entire season. Both methods were tested, and it made little difference in the outcome.

The data in Table 2 for models 3 and 4 represent OLS regression results for Equation 1 plus the stadiums variable for a models with and without the z(94–03) variable, respectively. In both models, the coefficient on the stadium variable is positive and statistically significant at the 1% level, and the inclusion of the stadium variable improves model fit. The magnitude of the stadium coefficients in models 3 and 4 predicts that each stadium built between 1960 and 2003 has resulted in a permanent increase in average league-wide attendance of 30 000 to 40 000 fans per team.

Table 2. (Sample 1902–2003) intervention analysis: the 1994–95 strike

What is particularly interesting is the fact that the intervention variable in Equation 4 for the 1994–2003 period is now statistically significant at the 10% level. This indicates that over the entire strike, recovery, and post-strike period, attendance growth has been significantly lower, by over −50 000 fans per team per year, than experienced by MLB over the 1902–1993 period. Since 14 new stadiums were constructed over this period, however, each of which contributed to an increase in average attendance of 41 000 fans, the decline in the underlying attendance was masked by the attendance boost provided by new playing facilities. While a Wald test on the combination of the 1993 shock, 1994 strike, and 1996 recovery still fails to reject the hypothesis that the combination of 1993 and 1996 coefficients is equal to the absolute value of the 1994 coefficient, model 4 lends strong support to the conclusion that MLB has suffered not only temporary but permanent effects from the labour troubles in 1994–1995.

The 2002 season presents clear evidence of MLB's problem. That season's average attendance of just under two-and-a-quarter million was the lowest since 1996 and over a quarter of a million fewer fans per team than in the record year of 1993. Not coincidentally, the 2002 season was also the first since 1996 in which no team opened a new stadium. As the stadium pipeline dries up, MLB may be doomed to ever lower average attendances. The threat of a mid-season player's strike also hung over the 2002 season. While last-minute negotiations averted a walk-out, fans apparently stayed away from baseball none the less. As noted by Schmidt and Berri ([2]), previous episodes of labour/management strife which did not result in the loss of regular season games did not seem to have any significant effects on attendance. The 153 000 reduction in average attendance between the 2001 and 2002 seasons is by far the largest during any period of labour troubles which did not lead to a loss of a significant number of regular season games.

Of course, models 1–4 present only two of the intervention and response possibilities, but Box and Tiao ([1]) suggest other alternatives. Schmidt and Berri's proposal, as shown in models 1 and 3, demonstrates one extreme where the strike can be seen as a series of pure 'pulse' events causing a one-period increase in the growth in attendance due to expansion in 1993, a drop to strike levels in 1994 and 1995 and an immediate and complete recovery in attendance growth to its pre-1993 level in the following year as shown in Fig. 2a. (Note that all examples in Fig. 2 show the intervention and response effects after accounting for time trend and stadium effects.) In these models, where attendance is first differenced, the intervention variables take values of 1 in 1993, 1994 and 1996 and a value of 0 in all other periods. On the other extreme as proposed in models 2 and 4, the event can be seen as a pure 'step' function with the strike causing attendance growth to fall to a permanently lower level after the strike as in Fig. 2b. In these alternative models, the additional z(94–03) intervention variable is added. As noted previously, the data approximate Fig. 2a when stadium effects are not accounted for but more closely match Fig. 2b when the construction of new playing facilities is considered. Therefore, the inclusion of a stadium variable changes the interpretation of the attendance patterns following the 1994–1995 strike from one where the attendance effects are temporary to one where lasting effects are observed.

Graph: Fig. 2. Various impulse-response possibilities

Another possible response to the strike would include an immediate negative impact followed by a gradual return to the original level over a number of periods that so that z(94–03) = 1 in 1994, z(94–03) = 1 − n(1/N) in period n over the subsequent N periods, and z(94–03) = 0 elsewhere (corresponding to linear recovery model as in Fig. 2c), or z(94–03) = 0 prior to 1994, z(94–03) = 1 in 1994, and z(94–03) = (1/ α)n for all periods subsequent to 1994 (corresponding to an exponential recovery model as in Fig. 2d). Alternatively, the response could be an immediate negative impact followed by a partial recovery to a level that is either higher or lower than that existing before the strike as in Fig. 2e and f.

Without a detailed a priori expectation of the long-run effects of the strike on attendance, one is simply left to the task of examining multiple models for fit. Table 2 shows regression results for a linear complete recovery model over the 1994–2003 period (model 5), a linear partial recovery model where MLB recovers one-half of the attendance decline by 2003 (model 6), and a linear full recovery model over a shorter time period from 1994–1999 (model 7). All three models show worse fit than model 4 which presumes a permanent decline in attendance growth. No other variations in the linear recovery model showed fit statistics that were superior to model 4.

In addition, models similar to Fig. 2d and f showed poor model fit under nearly all values of α and percentages of recovery. For comparison, the regression results for a model with α = 2 and full recovery is shown in Table 2, model 8.

The primary limitation of intervention model examined in this paper is the simplistic manner in which new stadiums are assumed to affect average attendance. The model presented in this study presumes that the construction of a new stadium will result in a one-time change in average attendance with no subsequent effects. In other words, the new stadium results in an immediate and permanent increase in average league-wide attendance. While the stadiums variable could be modified in a similar fashion to the z(94–03) strikes intervention variable to allow for effects that either appear or dissipate over time, as before any such modifications would be generally ad hoc in nature.

A more advanced technique would be to use the existing time-series data to generate reaction functions that would allow for a more detailed examination of the construction of a new stadium on attendance and would have the secondary benefit of allowing a test of whether the most recently constructed stadiums have experienced a shorter 'novelty effect' than stadiums constructed earlier in the 1990s. Such a model, however, is beyond the scope of this short research note.

The purpose of this study has been to determine whether Major League Baseball's frequent period of labour disputes have resulted in permanent or temporary reductions in average attendance at games. Previous research has concluded that the 1994–1995 Major League Baseball strike caused short-term losses in fan interest but did not result in any long-term effects on attendance. The results in this study suggest a different conclusion. While total attendance at MLB games following the 1994–1995 strike has recovered to its pre-strike levels, this has been done only through the construction of new stadiums at an unprecedented pace, which cannot continue into the future. After accounting for stadium effects, the average growth in MLB team attendance has dropped significantly since the 1994–1995 strike.