5. RESULTS FROM THE LOGISTIC REGRESSION
Factors associated with volunteering may be inter-related. For example, age structure and family composition may affect the volunteer rates of people living in rural or urban areas. A multivariate analysis helps to understand these inter-relationships by looking at effects while holding all other factors constant.
For this multivariate analysis GSS data was chosen. This is because the characteristics of volunteers were similar between the two datasets, and because the GSS provides more accurate estimates for voluntary work. The descriptive analysis in tables 4.1 and 4.2 show that characteristics such as sex, education, employment and proficiency in spoken English are all associated with volunteering. However, these characteristics may also be associated with each other. In order to investigate the individual effects of each factor on volunteering, independent of the other factors, logistic regression has been used. This determines the effect of individual characteristics (such as sex, age, education, employment etc.) on volunteering while holding the effects of all the other variables constant.
Logistic regression is a type of multivariate analysis. By assigning a value of 1 if the person volunteered and 0 otherwise, a binary dependent variable is created that is suitable for logistic regression modelling. The results of this type of modelling are easier to interpret than other non-linear techniques (such as probit modelling). The listing of logistic coefficients and model diagnostic statistics used in this analysis are in Appendix C and detailed technical information on the logistic model is in Appendix D.
The odds ratios obtained from the logistic regression analysis are presented in table 5.1. As the name suggests, odds ratios are the ratio of two odds. The odds ratio for a female is obtained by dividing the odds of being female in the volunteering group by the odds of being male in the volunteering group. Any variable that has an odds ratio of greater than 1.0 indicates an increased propensity to volunteer, while a variable with an odds ratio of less than 1.0 has a reduced propensity to volunteer. If the confidence interval for an odds ratio includes the value 1.0, then we consider the estimated odds ratio not to be significantly different from 1.0 (i.e. the resulting propensity to volunteer based on that characteristic is not significant).
In the GSS model, the odds ratio for a female is 1.28. Since the confidence interval surrounding this estimate (1.18–1.38) does not include the value 1.0, it can be asserted that, at a 95% significance level, being female increases the propensity to volunteer, all other variables being held constant. Another way to put this is that the odds ratio result shows the effect of being female increases the odds of volunteering by 28% over males, while controlling for the effects of all other variables (age, family composition, marital status, proficiency in spoken English, education level, employment status and area of residence).
5.1 ODDS RATIO DESCRIBING THE PROPENSITY TO VOLUNTEER, 2006 GENERAL SOCIAL SURVEY
The size of the odds ratio indicates which variables have the strongest association with volunteering. However, since the odds ratios are not linear, it is not possible to make relative comparisons between characteristics. For example, an odds ratio of 2.42 for having an Advanced Diploma/Diploma qualification or above does not mean these people are 1.6 times more likely to volunteer than people whose highest qualification was Year 12 or a Certificate III or IV (odds ratio 1.52).
The variables that contribute most strongly to the propensity to volunteer are educational attainment and family composition, based on the model. Having progressively higher levels of educational attainment contributed markedly to the propensity to volunteer compared with those whose highest level of educational attainment was Year 11 or below. The results also show that respondents from a family where there was a dependent child under the age of 15 years had twice the odds of volunteering than a person from a family with no dependent children (the base category).
People who do not speak English well, or do not speak English at all, had a lower propensity to volunteer than people who speak English, even after controlling for variables such as education and employment status.
Being employed part-time increased the propensity to volunteer over being employed full-time. In addition, as described above, women had a higher propensity to volunteer than did men.
Finally, the results indicate that people who lived outside the capital cities had an increased propensity to volunteer compared with their capital city counterparts. Similar results have been observed in Canada, where both formal and direct personal volunteering were found to be highest in intermediate urban, small town and rural communities (Reed and Selbee, 2001).