Chu: Winning, scoring hardly matter to attendance levels

Chicago, Toronto, and Minnesota have the highest (median) annual attendance rates from 2006 to 2014. While the Leafs and Wild have die-hard, loyal hockey fans, the Chicago Blackhawks is the only team that has found post-season success during those years. But their 2006-07 attendance rate is one of the worst in the league since 2006 and that coincided with a team record of under .500.

This brings us to these questions: Does winning and team location matter? And if they do have effects, are the magnitudes relatively large enough to make us care enough about them.

About the data

Annual average attendance figures are pulled from ESPN.com, while team performance metrics are collected from NHL.com. Each record or observation is a team season. For example, Washington’s 2009-10 season is one observation.

The merged dataset spans from 2006 to 2014, consisting of 240 observations. Note that the Atlanta Thrashers moved to Winnipeg in 2011-12 into a smaller arena of 15,004 seating capacity.

Attendance figures in this analysis are an average annual (or hockey seasonal) rate statistic, a percentage of a team’s arena seating capacity. There are instances of rates of over 100% due to some arenas containing standing room.

Attendance Analysis 

Let’s take a look at how attendance data is distributed:

Chu1

 

 

A non-symmetrical distribution with negative skew is seen. The median annual (average) NHL team attendance since 2006 is 98.95% of seating capacity, approximately equivalent to 18,300 fans in a 18,500-seating capacity medium-sized NHL arena. Some outliers lie on the left, where single season attendance for some teams fell in the range of 55% to 70%:

 

Team

Season

Avg Attendance

Attendance %

W %

W% Prev Season

St Louis

2006-07

12520

59.6

41.5

25.6

Chicago

2006-07

12727

62.1

37.8

31.7

NY Islanders

2010-11

11059

67.9

36.6

41.5

Phoenix

2009-10

11989

68.5

61.0

43.9

 

The Coyotes franchise had a troubling 2009-10 season off the ice when it was placed into bankruptcy. During the off-season, the uncertainty of the team’s future home city was probably more than enough to scare off Phoenix season ticket holders from renewing. Average attendance for the season fell by 19.4% from the previous season. Since the Phoenix situation was an unusual case, this observation is removed from the statistical analysis later on.

Despite Chicago having a very low attendance rate in 2006-07, they still scored first place in median attendance during the 2006-2014 period:

Team

Median Attendance (% of arena capacity)

Chicago

108.50

Toronto

103.25

Minnesota

102.75

Ottawa

101.20

Pittsburgh

101.10

Vancouver

101.10

Philadelphia

100.70

Calgary

100.0

Edmonton

100.0

Montreal

100.0

NY Rangers

100.0

San Jose

100.0

Boston

99.50

Detroit

99.50

Buffalo

99.35

Washington

97.25

Los Angeles

96.70

Tampa Bay

95.40

Anaheim

93.95

Dallas

92.50

Nashville

91.75

St. Louis

89.50

New Jersey

88.20

Carolina

88.05

Winnipeg/Atlanta

86.50

Colorado

85.95

NY Islanders

81.80

Columbus

81.35

Florida

81.35

Phoenix

80.85

Since most of the statistical distributions of the team attendance data are skewed, I use the interquartile range, better known as the IQR, as an alternative measure to the standard deviation. The IQR is the range of the middle 50% of the sorted data. (The units are still in percentage points.) Here is the annual team attendance IQRs sorted from least stable to most stable attendance:

 

Team

Median

IQR

Winnpeg/Atlanta

86.50

22.48

Dallas

92.50

14.98

Phoenix

80.85

12.60

Chicago

108.50

11.35

St. Louis

89.50

11.03

Tampa Bay

95.40

10.53

Nashville

91.75

9.48

Anaheim

93.95

8.83

Los Angeles

96.70

8.30

Washington

97.25

7.43

Colorado

85.95

6.40

Ottawa

101.20

5.60

Boston

99.50

5.35

NY Islanders

81.80

5.05

Carolina

88.05

4.65

Columbus

81.35

4.50

Florida

81.35

3.95

New Jersey

88.20

2.43

Detroit

99.50

2.25

Minnesota

102.75

1.53

Vancouver

101.10

1.30

Philadelphia

100.70

1.15

Pittsburgh

101.10

1.00

Buffalo

99.35

0.93

Toronto

103.25

0.63

San Jose

100.0

0.53

NY Rangers

100.0

0.13

Calgary

100.0

0.00

Edmonton

100.0

0.00

Montreal

100.0

0.00

Chicago’s recent success in the stands only began when Jonathan Toews and Patrick Kane entered their first NHL season in 2007-08. In that season, Chicago’s win rate increased by 29%, while their attendance rate rose in a similar fashion by 32% (or 19.9 percentage points) from 62.1% a year earlier. In the following season, team performance and attendance improved again. The team had improved their win-loss record by 48% and saw a boost in their attendance rate to 111.2%, a 79% increase in just two years.

Team Performance, Location, and Attendance

For the Hawks, regular season wins and attendance in the same season seem to increase together. How about for the rest of the league? And does the team’s country and other team performance metrics matter?

To find out, I run an ordinary least squares linear regression of various team performance metrics in both the observed and previous seasons, and team location on the attendance rate. A subset of the original dataset is used for analysis. The 2012-13 and 2013-14 seasons are removed since both contain data of the shortened 2012-13 season. (Chicago and Pittsburgh had winning percentages of 75%, which is likely unsustainable over a full season.)

Phoenix’s 2010-11 and 2011-12 seasons, when the franchise continued to experience financial challenges, turn out to be outliers for the statistical model. Above .500 records in both the observed and previous seasons weren’t enough to draw some fans back. The franchise was too financially unstable for fans to commit to seasons tickets.

Since the Phoenix bankruptcy years are too extreme for the model and are not the kind of seasons we want to use to study the statistical relationship between winning and attendance, these seasons are removed from the analysis. As you may recall from earlier, Phoenix’s 2009-10 season, the first season of NHL ownership, is removed.

The following variables in the linear regression are statistically significant:

  • The previous season’s winning percentage. (Prev W%)
  • The observed season’s goals for per game average. (GF/GP)
  • Country of team’s home
  • Whether the team made the playoffs in the observed season
  • Country AND whether the team appeared in the playoffs in the observed season.

The r-squared of the model is 52%. It means approximately 52% of the variation in the attendance rate can be attributed to the variation in these variables. The linear regression model isn’t a spectacular one, but it’s not a terrible one either.

Previous W%

Out of these variables, the previous season’s record of the team may be the only one with some causality simply based on the fact that it’s data from the previous season. The previous season’s performance is a factor in a fan’s decision on whether to purchase season tickets.

One additional win, on average, suggests an increase in the attendance rate by 0.49 percentage points when all other factors are held constant. In a 18,500-seat arena, it’s about 90 more fans, which, perhaps surprisingly, is not a relatively large number.

GF/GP in observed season

Fans love watching goals and this is just further proof. Scoring an additional goal per game over a full season is an unrealistic improvement in one year. So, let’s, instead, take the median goals per game and subtract it from the 75th percentile value. The difference gives us a more realistic increase of roughly 0.2 goals per game, a 16-goal difference over a whole season, with an attendance rate increase of 0.96 percentage points, on average, when all other factors are held constant. Like the previous season’s winning percentage, the estimated goal scoring effect on attendance is very small.

Country and playoff appearance

There appears to be an interaction between the team’s country and whether the team made the playoffs. Here’s how the marginal effects of these explanatory variables are broken down when all other factors are controlled (figures are in comparison to a non-playoff American team):

  • An American playoff team sees an attendance increase of 6 percentage pointson average or roughly 1100 fans in a 18,500-seat arena.
  • A Canadian non-playoff team expects to attract, on average, a 14.4 percentage point increase in fans at their home games.
  • A Canadian playoff team draws an increase in attendance of 10.4 percentage points.

How can a Canadian non-playoff team attract more fans than a Canadian playoff team? Well, it turns out some teams, like the Leafs and Oilers, rarely or never made the playoffs between 2006-2012, but continued to have had loyal fans. So, the playoff variable among Canadian teams didn’t work exactly as expected. There needs to be similar variation of Canadian teams between the playoff and non-playoff teams in order to better evaluate any kind of playoff appearance or regular season performance effect.

No matter which country the team is located in, making the playoffs matters. The playoff appearance effect is one measure of regular season success and happens to be more relevant than the winning percentage in the observed season, which is not statistically significant as you’ll see below.

Not statistically important variables

What about the other variables that didn’t make the cut? In the initial linear regression, most variables had coefficients that are not statistically significant. Here are a handful of those:

  • Win percentage in the observed season
  • Goals scored for per game average in the previous season
  • Playoff appearance in the previous season
  • Any variable that measured playoff performance in the previous season (conference final appearance, Cup final appearance, Cup win)

There isn’t enough statistical evidence to suggest that the previous season’s playoff appearances or performance has an effect on the following season’s attendance rate. In other words, when a team makes the playoffs, advances a few rounds, or wins the Stanley Cup, a boost in next season’s attendance isn’t certain.

Conclusion

To recap, regular season wins in the previous season and goals-scored-for over the course of the season have only a seemingly, very small effect on the team attendance rate on average. Of course, there are exceptions like Chicago, where the attendance rate grew by nearly 80% in two years as the team improved, and Toronto, where sell outs are common regardless of performance. No matter which country your favourite team is in, the arena will be closer to (or beyond) capacity if your team has a great shot at making the playoffs.

There are other factors not accounted for in the model including the size of the hockey market, fan loyalty, and star players, but we’ll leave that work to the team owners.

Leave a Reply

Your email address will not be published. Required fields are marked *

*