Variations in NHL attendance: the impact of violence, scoring, and regional rivalries - Discrimination and The NHL

American Journal of Economics and Sociology, The, April, 2003 by Rodney J. Paul

Attributes of the visiting team are also included in the model. Points per game (VPPG), goals per game (VGPG), and fights per game (VFPG) were included as the average of that variable going into the current game. These signs are ambiguous a priori as it is not certain what fans wish to see from opposing clubs. If they wish to see the best visiting teams, the sign on visiting points per game should be positive. If they expect the home team to lose, however, this sign could be negative. Similar logic holds for visiting goals per game. Visiting fights per game also has an ambiguous sign as more violence may be expected to increase attendance, although the home fans more than likely prefer to see their team invoke violence on the opponent, not vice versa.

The NHL has moved to regional divisions and conferences to promote rivalries and take advantage of the excitement generated by local teams playing more often. In 1998-1999, the league moved from four divisions to six divisions, with closer geographic ties. In the season studied, 1999-2000, division rivals played each other five times during the year in the Eastern Conference and six times a year in the Western Conference. The rest of the conference rivals played four times a year and non-conference opponents played either once or twice, except for Canadian teams, which can play other non-conference Canadian teams three times in a given year. The coefficient on DIV, the division rivals, is expected to be positive due to a variety of factors that may include local media coverage of more than one team, more familiarity with the opposing players, and the existence of friends and co-workers who may be fans of this opposing team. A dummy is included for non-conference opponents (NONCON). Possible abolishment of int er-conference play has been noted by periodicals such as The Hockey News. This would be to lower transportation costs to teams and further enhance local rivalries. If it is true that fans do not want to see opposing conference teams play, the sign of this coefficient should be negative. On the other hand, if fans are excited to see an opposing squad that might only come to town once a year, the sign should be positive. A variable for Canadian opponents (CANOPP) is included for a breakdown of the sample into just Canadian teams. In respect to the idea that Canadian fans like seeing other Canadian teams, the sign on the coefficient should be positive.

Data from 1999-2000 consists of 1,148 games. Teams that do not have game-to-game variation in attendance due to sellouts are excluded from the set, to bring the total sample size to 984. The Colorado Avalanche, Detroit Red Wings, and Dallas Stars all sold out each game during the 1999-2000 season. Equations (1) and (2) are estimated by the Zellner method of seemingly unrelated regressions due to the possibility of the error terms not being independent. The model is estimated for the league as a whole and then broken down to observe the U.S. and Canadian teams separately. The main focus of this research is the attendance equation (1); the second equation is estimated as in Jones, Stewart, and Sunderman (1996), but the discussion that follows focuses on attendance rather than price.