2. Data and Methodology
The Survey of Giving, Volunteering and Participating (SGVP) recently released by Statistics Canada provides an ideal source of data on various aspects of an individual's philanthropic activities. This survey was undertaken in 1997 and covers the activities of individuals in the November 1996 to October 1997 period. It is the most complete survey of its kind, containing responses from some 18,301 individuals who are representative of the Canadian population at large.1 Because this study is interested in the labour-market responses to volunteering, the sample is restricted to those individuals who are employed on either a part-time or full-time basis at the time of the survey. Further restricting the sample to individuals who responded to questions of interest, resulted in a sub-sample of 9 945 individuals; the labour-market response to volunteers in this sub-sample formed the basis of the analysis in Devlin (2000) and lead to the conclusion that volunteers earn about 4 per cent higher incomes than do non-volunteers. In this paper, we continue to use this sub-sample to estimate any regional differences in this labour-market response.
The basic econometric model entails estimating earnings equations for volunteers and non-volunteers by region, as expressed in (1) and (2):
where Xji
is a vector of individual i's characteristics (including the stock of human capital) and
represents a normally distributed random component, j=v,n. The subscript v indicates that the individual is a volunteer, while n indicates a non-volunteer.
To ensure that sufficient numbers exist, Canada is separated into its five well-known regions: British Columbia, the Prairies (Alberta, Saskatchewan and Manitoba), Ontario, Quebec, and the Atlantic region (New Brunswick, Nova Scotia, Prince Edward Island, and Newfoundland and Labrador).2 Equations (1) and (2) are estimated for each of these five regions in order to generate some estimates of the earnings differentials that exist across the country between volunteers and non-volunteers.
The problem with the model as depicted in (1) and (2), which has been discussed in detail in Devlin (2000), is that volunteers may not be drawn randomly from the population, introducing a selectivity bias into the earnings equations. Thus, we use a Heckman two-stage procedure in which inverse Mill's ratio is calculated from a probit estimation of the decision to volunteer, and introduced into the earnings equations. These earnings equations are used to estimate the earnings differential associated with volunteering, which is then included in a structural probit model to determine whether the expected earnings differential affects the probability of volunteering. In other words, to the extent that the expected earnings differential is a measure of the economic benefits associated with volunteering, we can determine if these benefits actually motivate individuals to engage in this sort of activity. Having established in Devlin (2000) that individuals do respond to the earnings differential, this paper explores any regional variation in this response.
The final point to note is that volunteers are over-represented in the SGVP data set because of the particular sampling technique used. As a result, it is extremely important to use the sample weights accompanying this survey in order to adjust for how "representative" each observation is relative to the Canadian population at large. Thus, all of the estimation procedures undertaken in this analysis take account of sample weights.
- 1Details of this data set are available in Statistics Canada (1998). How this data set
compares to the earlier VAS survey is discussed in Devlin (2000).
- 2Data are not available for the Canadian territories.