Endnotes

View the most recent version.

1. This study uses the definition of culture and culture activities established in the Canadian Framework for Culture Statistics .

2. Literature searches were carried out during autumn 2006 and winter 2007 using various combinations of search words in a number of different research databases. We noted that past Canadian research on culture participation is essentially composed of survey studies which describe the socioeconomic and demographic characteristics of culture participants (see, for instance, The Heritage and Arts final report, 2000; reports from Canada Council for the Arts, 1999; 2001; 2002; and Ogrodnick, 2000). Although studies of this kind may show how culture participation differs by personal and family factors, they may not be able to determine the key factors. The two exceptions, which we mention, are Bourdeau (2002), and Fisher and Preece (2003) who conducted multivariate analyses for selected culture activities to determine which socioeconomic attributes mostly contribute in the explanation of culture attendance.

3. In addition, since Fisher and Preece (2003) did not rely on population-weighted estimates to study the social correlates of participation in musical activities in Canada, they cannot verify if their results are robust to population weights.

4. We use Bourdieu (1984) to measure a person's social capital by his or her family background, i.e., parents schooling.

5. Detailed information concerning this survey is available upon request at Statistics Canada's Social and Aboriginal Statistics Division.

6. All details concerning the recoding of each response variable and other variables derived from the 2005 GSS to model culture participation are listed in Tables 1 and 2 in the Appendix.

7. But, in order to rigorously gauge the impact potential near dependencies may have on the estimation of the standard error of the estimate of the regression parameters, we also performed the variance inflation factors (VIF) test in tables not shown. The general rule for the VIF test is that the VIF should not exceed 10 (Belsley, Kuh and Welsch, 1980). As the largest VIF obtained was 4.7, we concluded once more that none of our econometric models has been deteriorated by collinearity.

8. The proportional odds assumption assumes that all explanatory variables other than the constant terms have an equal effect on the odds of crossing the threshold between never participating and participating only occasionally and the threshold between participating only occasionally and participating regularly or frequently.

9. The SOR model generates (κ-1) standard multinomial intercept parameters for the y-response, (κ-2) independent Øs and a single β parameter for each explanatory variable.

10. This scaling parameter denotes the distance between the outcome levels in terms of the linear prediction Ø_qβX_i.

11. According to Statistics Canada' 1998/99 Performing Arts Survey, four provinces have the largest number of non-profit performing arts organizations in Canada: Quebec (226); Ontario (210), British Columbia (70) and Alberta (42).

12. As model coefficients are invariant to centering of first order terms, marginal effects are calculated by centering all continuous independent variables at their means. The reader should thus interpret marginal effects as the change in predicted probability associated with percent changes in the continuous independent variables. Also, we calculated the marginal effect of a dummy variable as the discrete change in the dependent variable as the dummy variable changes from 0 to 1. Under this specification, therefore, marginal probabilities should be interpreted as indicating how much the baseline probability to participate in a given cultural activity varies with each unit change in the corresponding explanatory variable, holding all other variables constant.

13. We note that the values of Ø's are all distinct with 99% level of statistical significance. Moreover, the ability of our model to predict culture participation is greater than would be expected by chance since χ- squared is statistically significant for each equation.