Seppo Suominen Essays on cultural economics
Download 2.66 Mb.
|
3.5EstimationConventional regression analysis is used here but the results might be biased due to heterogeneity. Time-series and cross-section studies that do not control heterogeneity might yield biased results. An alternative for conventional regression analysis is panel data methods. The benefits of using panel data are that (1) individual heterogeneity can be controlled, (2) estimated parameters are more efficient and (3) with panel data the dynamics of adjustment can be studied better (Baltagi 2008, 6-7). There are a large number of different approaches for the panel data. With pooled regression the heterogeneity effect contains only a constant term and the OLS provides consistent and efficient estimates of the slope coefficients. If the heterogeneity is unobserved but correlated with observable variables, the OLS is biased due to omitted variables but a fixed effects approach estimates the group-specific (cross-section dimension) constant term άi. If the individual heterogeneity is uncorrelated with the included variables, the random effects approach specifies that there is a group-specific random element (Greene 2008, 183). However, the main method is a pooled regression. The estimation is also done using perpetual inventory expectations but the variable had no statistical significance and is not reported. Results are adjusted for heteroskedasticity and heteroskedasticity is tested with the White test. The first model uses the price of the ticket (LogPrice), the population in home and visitor’s town (LogHomePop and LogVisPop), round (LogHGame), teams’ success or winning ratio (LogHOmePoints and LogVisPoints) and the maximum day temperature in home town (Temp, note: not logarithm) or alternatively the difference from the long-run average in the day (TempDiff) and dummies for Tuesday (TU) and Thursday (TH). All parameter estimates of the temperature difference in the pooled estimation are statistically significant, have the right sign and are plausible. The distance and the income measures are not used as controls. The fixed and random effects models are not plausible since either the home town population variable or the price variable get unrealistic coefficient estimates. The Saturday effect is substantial: the audience is about 10 – 11 percent larger than on Tuesdays or Thursdays. Other weekday dummies (Monday, Wednesday, Friday or Sunday) are not significant (not reported here).
Table 3 : Model 2 estimation results, including distance and income variables
Download 2.66 Mb. Share with your friends:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||