Seppo Suominen Essays on cultural economics

3.6Robustness tests

In this section a wide range of robustness checks is performed. Instead of the regional unemployment rate, households incomes (regional, NUTS4-level) are used to control for spectators’ incomes in specifications 3 and 4. For the fall season from September to December 2007 the annual household incomes for the year 2007 are while for the spring season from January to March 2008 the annual household incomes for the year 2008 are used. Due to coarse nature of annual incomes, a more detailed effect of incomes is obtained by using a monthly variable measuring consumer confidence (CCI). The spefications 3 to 6 reveal that the incomes have a negative impact on attendance. This result is in line with several studies (Borland and Lye 1992, Baimbridge, Cameron and Dawson 1996, Falter and Perignon 2000). Finally the gender effect is studied by using a consumer confidence index separately for men and women in specifications 7 and 8. Attendance is more sensitive for mens’ consumer confidence than for womens’ confidence. The temperature has interesting effects on attendance since low temperature seems to attract a bigger audience. The different speficiations confirm the impact of temperature:, The impact is however not very large: one degree Celcius lower temperature results in roughly 0,5 percent higher attendance. Since the average attendance is about 5000, this means 25 spectators. Ice hockey is a game played indoors, thus the effect of temperature is different than in football or other outdoors sports.

Home team’s success or winning ratio (Home Points or Home3Last) seems to attract more spectators since these variables get a positive coefficient regardless of the model. The visitor’s success on the contrary seems to lessen attendance (Visitor Points), but the recent success in terms of last three games (Vis3Last) does not seem to be significant as outlined in specification 10. Ice hockey audience seems to favour an assured win and not even games. If the home team falls into a losing circle, the attendance also falls. In these models visitor’s last three games points (Vis3Last) do not seem to significantly explain the attendance of the game, while home team’s last three games points seem to explain.

Most of the hypotheses are verified. The price and distance variable parameters are negative as proposed. The home town population has a larger positive coefficient than the visitor’s town population. The previous performances of the teams seem to explain attendance as suggested in hypothesis 3. Some spectators take into account the previous performance and therefore the ice hockey game is at least partially a search good. The weather conditions have a small but significant impact on attendance.

The pooled regression, fixed effects and random effects estimation results of the different specifications are in appendix.