## January 13, 2007

### Attendance

So all the talk about attendance figures made me wonder, how 2006-2007 is different from 2005-2006. Tom Benjamin has been on a tirade, with articles such as: Attendance Again, Records, Attendance and Bure, and Money Talks.

After reading all this I was wondering how attendance is determined, that is to say what is the attendance a function of? There are a few obvious candidates such as winning percentage, opposition winning percentage or skill players (Crosby). My main question is however, is 2005-2006 any different from 2006-2007? In order to accommodate many of the problems I’m making a model with 58 variables for teams (29 home teams, 29 away teams), winning percentage for the home team and away team, day of the week, and year. For this study I looked at three years worth of attendance figures (2003-2004, 2005-2006, 2006-2007) a total of 3117 games. One game in 2003-2004 was rejected as it wasn’t in an official stadium and had astronomical attendance and that would be the Heritage Classic game.

Day of the Week

It appears that the NHL knows Saturday is the best day for attendance so the NHL went through a special effort to get as many games as possible on Saturday to boost attendance. The NHL found a way to make the percentage of games go from 23% Saturdays to 26.4% using Sunday and Monday as rest days. (14% of games occur on Sunday and Monday). The rest of the week is a wash and the NHL loses about 1000 visitors (compared to Saturday games) on the weekdays, with the exception of Friday where attendance is the same as Saturday. Sunday performance is similar to weekdays, but sits at -700, which isn’t as bad.

Winning

Due to the inclusion of individual teams in the regression the winning variable becomes less useful, however including three seasons made winning slightly useful. If you win more games you get better attendance. The winning variable (# of wins per 82 game season) on its own works out to about 120 more guests per game per win; however in the full model it is reduced to 60 more guests per game per win as much of the variability falls onto the individual teams. In other words a team that wins 50 games over the course of a season will attract about 2400 more people per game than a team that wins 30 games over a season. Each additional win for the visiting team is worth an addition 10 extra guests. So a team with 20 wins during the season will draw 300 fewer fans than a team that has 50 wins. In other words fans prefer to see decent teams.

Individual Teams

The worst 5 teams when it comes to attendance (using Washington as a reference) are: Nashville (-939), Chicago (-463), New York Islanders (-1781), New Jersey Devils (-592) and Carolina (-170). The best teams generally have large stadiums such as Montreal (+6260), Detroit (+5056), Philadelphia (+4710), Toronto (+4670), Tampa Bay (+4510). I’m wondering why Tampa does so well, but the others are mostly self explanatory.

Popular Teams

The first group of popular teams is the obvious: the original six all attract statistically significantly more viewers than other teams. Philadelphia and Pittsburg also draw more crowds. Not sure how Philadelphia is drawing crowd. It’s shouldn’t be a shock to see Crosby, I mean Pittsburg, on the list as well. In fact Crosby draws about 950 extra people to the places he visits. It’s interesting to note, but Nashville is the worst team at attracting audiences in foreign arenas at -600 per game. It would be in the best interest of the NHL as a whole to move this team as they’re hurting the NHL by about 25,000 guests (\$1 million) just with Nashville’s road trips (in ticket sales).

Ticket Prices.

Ticket prices could be a useful variable over long periods or with decent market information, in this model increase in ticket prices causes attendance to increase. So it isn’t useful at this present time.

Year

Of course this is the variable that matters most all the above work is to factor out the important variables so we can see how poorly the NHL is doing. And it works out to -300 people per game or \$16 million in ticket sales (not sure how this correlates with booze purchasing at games and merchandising outside of games). That works out to a 1.2% drop in attendance; however there was a 1.6% real increase in ticket prices. So the 98.8% of the people who are still coming are paying 1.6% more, which basically works out to an increase of 0.4% in revenue. However actually doing the math on ticket sales (attendance * average real ticket price) works out to 3% (\$712,152.05 per game vs. \$732,636.94) increase in revenue from ticket sales, which suggests that the loss of attendance is occurring in places where it effects revenue less. It's not all bad news for attendance though, it's still higher than 2003-2004, by about 150 guests per game.