Labour-Market Responses to Volunteering: Regional Differences - June 2000

4. Econometric Results

Previous

Contents

Next

As discussed previously, the econometric analysis of this problem has three separate components: a reduced-form probit model of the decision to volunteer, selectivity-corrected earnings equations for volunteers and for non-volunteers, and a structural probit model that takes account of the expected earnings differential. It thus seems sensible to discuss the decision to volunteer first before turning to the importance question of how regional labour markets respond to volunteering.

4.1 The Decision to Volunteer

The probit model includes all of the standard characteristics which are thought to influence the decision to volunteer, plus a few extra variables that are reported in the SGVP. The variables may be broadly categorized into three groups: personal, family, and labour market. The personal characteristics include sex, age, marital status, educational level, whether or not the individual donates money to charities (GIVE), whether or not the individual classifies him- or herself as "religious" (REL), and whether or not the survey interview was conducted in English or French (ENG). We also include as explanatory variables in the probit model variables that denote if the individual is a recent immigrant of less than four years (NEWLAND), a medium-term immigrant of four to eight years (MEDLAND), or a long-term immigrant of over eight years (OLDLAND). A final personal characteristic is the individual's tenure in his or her current residence: NEWRES is the reference group and denotes individuals who have been in their current residence for less than one year, MEDRES denotes those who have lived for one to five years in their current residence, and OLDRES denotes all others. These variables are included to capture the impact of community attachment, an important indicator of social capital, on volunteering.

Family characteristics include the number of individuals in the household (HHSIZE), and the number of children under six years of age (OWNK05), six to twelve years of age (OWNK0612), thirteen to seventeen years of age (OWNK1317), and over eighteen years of age (OWNK18PL). Labour market characteristics are important because we are estimating a reduced-form probit model that takes account of the fact that the expected earnings differential may influence the decision to volunteer. To this end, we include the occupation of the individual as represented by eighteen different classifications (services are the reference group).

Table 4 reports the reduced-form probit estimates for each of the five Canadian regions. The number of qualitative inter-regional differences that emerge among the various explanatory variables it is actually quite remarkable. For instance, being male has a positive impact on the decision to volunteer in Atlantic Canada and Quebec, but is statistically insignificant elsewhere. Being married has a negative influence in Ontario (at the 10 per cent level of significance) and Quebec, a positive influence (at 10 per cent) in the Prairies, and has no influence in the Atlantic provinces or in British Columbia. The impact of educational level is somewhat less pronounced in Quebec relative to the other regions, while the presence of older children seems to have a mixed impact across of all the regions.

Some interesting regional differences arise in the relationship between being an immigrant and deciding to volunteer. The pattern established elsewhere for Canada as a whole was that being an immigrant has a negative impact on volunteering, an impact that diminishes with time (Devlin, 2000). In the regional analysis, this pattern emerges exactly for British Columbia and is weakly consistent with the results in Ontario. Elsewhere, being an immigrant is largely an insignificant determinant of the decision to volunteer, except in the Prairies where being a medium-term immigrant appears to have a negative impact on this decision. In many ways, these regional differences are not surprising - most immigrants currently land in British Columbia or Ontario; the result in the Prairies may arise because the individual landed elsewhere in Canada but moved later on to the Prairies where he or she needed time to develop the knowledge required to be a formal volunteer.

Another Canada-wide pattern reported in Devlin (2000) relates to the impact of tenure in the same dwelling on volunteering. In every specification of the sub-sample, it was always the case that residence of less than five years did not affect the decision to volunteer, whereas residence of five years or more had a positive and significant impact on this decision. The effect of tenure on the decision to volunteer, however, does appear to differ across regions: the results in the Atlantic region, the Prairies, and Quebec broadly support the established pattern; however, tenure has no impact at all in British Columbia and, rather surprisingly, in Ontario being a medium-term resident has a weakly negative impact on volunteering (at the 10 per cent level of significance) relative to being a new resident. Furthermore, residing five years or more in the same dwelling has no impact on the decision to volunteer in Canada's most populous province. One can only speculate as to what is going on in Ontario. It is clear from Table 1 that, in Ontario, more individuals live in a CITY (an area with a population of 100,000 or more) than elsewhere. People are often more mobile within a city than they are, say, in a rural area: apartment dwellers can easily move from a one-bedroom apartment to a larger one in the same area. Technically, therefore, they could be classified as a "new" resident because they have only recently moved into their current dwelling, but they could, in fact, be established residents in the same community. As a consequence, the distinction between each classification of resident, and the presumed relationship between tenure in one dwelling and tenure in the same community, may be blurred.

Table 4 Reduced-form probits: Decision to volunteer

Variables	BC		Prairies		Ontario		Quebec		Atlantic
	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio
MALE	-0.064	-0.54	0.023	0.36	0.065	1.09	0.315	3.78	0.208	2.59
MARRIED	0.046	0.39	0.119	1.69	-0.111	-1.65	-0.295	-3.30	-0.094	-1.10
HOURS	-0.007	-1.92	-0.010	-5.42	-0.011	-5.44	-0.007	-2.11	-0.012	-4.26
HIGH	0.639	1.29	0.588	3.08	0.542	3.06	0.123	0.69	0.578	2.92
DIPLOMA	0.866	1.75	0.745	3.83	0.726	4.08	0.272	1.51	0.735	3.66
POSTSEC	0.927	1.85	1.011	4.98	0.931	4.96	0.407	2.02	0.875	4.06
UNIV	1.225	2.41	1.116	5.46	1.027	5.47	0.656	3.22	1.268	5.75
OWNK05	-0.106	-1.13	-0.171	-3.38	-0.054	-1.16	0.057	0.83	-0.126	-1.95
OWNK0612	0.315	4.05	0.327	7.15	0.228	5.47	0.204	3.34	0.257	4.79
OWNK1317	0.348	3.47	0.174	3.16	0.330	5.60	0.159	2.24	0.123	1.81
OWNK18PL	0.063	0.57	0.053	0.86	0.055	1.03	0.304	3.89	0.214	2.98
RURAL	0.192	1.56	0.172	2.59	0.164	2.25	0.232	2.47	0.160	2.10
TOWN	0.017	0.13	0.170	1.54	0.071	0.85	0.123	1.06	-0.033	-0.31
MANAGER	-0.087	-0.46	0.118	1.07	-0.046	-0.45	0.041	0.27	-0.027	-0.22
SCIENCE	-0.151	-0.61	0.234	1.54	-0.118	-0.87	-0.084	-0.41	-0.440	-2.23
SOCSC	0.478	1.33	0.463	2.13	0.496	2.44	0.342	1.24	0.451	1.56
TEACH	0.361	1.20	0.528	3.35	0.344	2.32	-0.005	-0.03	0.532	2.86
RELIGION			0.154	0.51	0.796	1.55			1.390	2.18
HEALTH	-0.448	-1.91	0.108	0.80	0.010	0.69	-0.014	-0.08	0.120	0.77
ARTS	-0.542	-1.85	-0.036	-0.17	0.674	3.53	0.236	1.03	0.089	0.36
CLERK	-0.098	-0.55	-0.013	-0.12	-0.187	-1.91	0.199	1.46	0.083	0.70
SALES	-0.186	-1.00	0.096	0.84	0.072	0.66	-0.004	-0.03	0.031	0.24
PRIMARY	-0.140	-0.50	0.239	1.80	0.318	1.73	0.022	0.10	-0.039	-0.21
PROCESS	0.058	0.20	-0.516	-2.02	-0.410	-1.94	0.010	0.04	-0.263	-1.43
MACHINE	-0.195	-0.40	0.076	0.29	-0.541	-2.76	-0.139	-0.43	-0.599	-1.95
FABRIC	-0.514	-1.89	-0.057	-0.40	-0.406	-3.60	-0.364	-2.15	-0.128	-0.79
CONSTRUC	-0.231	-1.00	0.111	0.78	-0.120	-0.81	-0.092	-0.45	-0.190	-1.10
TRANSP	-0.228	-0.86	0.119	0.69	-0.520	-3.07	-0.559	-2.57	-0.096	-0.50
MATERIAL	-0.243	-0.59	-0.312	-1.50	-0.690	-3.39	-0.346	-1.17	0.110	0.39
OTHER			0.830	2.50	-0.645	-2.04	0.094	0.27	-0.020	-0.06
NEWLAND	-1.534	-3.28	-0.225	-0.98	-0.237	-1.34	1.059	1.49	6.288	0.03
MEDLAND	-0.740	-2.39	-1.418	-3.87	-0.306	-1.79	-0.503	-0.73	0.064	0.11
OLDLAND	-0.395	-2.93	-0.116	-1.05	-0.260	-3.60	0.056	0.29	-0.005	-0.02
MEDRES	-0.042	-0.21	0.141	1.35	-0.193	-1.68	0.012	0.06	0.273	1.71
LONGRES	0.087	0.46	0.251	2.55	0.149	1.36	0.325	1.84	0.396	2.69
ENG			-0.370	-0.80	0.229	1.21	0.172	1.36	0.367	3.09
REL	0.090	0.94	0.244	4.25	0.187	3.47	0.168	2.32	0.389	5.31
AGE	-0.004	-0.87	-0.007	-2.62	0.004	1.65	-0.008	-2.04	-0.004	-0.99
GIVE	0.798	5.37	0.938	10.69	0.704	7.51	0.404	3.85	0.738	5.63
CONSTANT	-1.477	-2.47	-1.352	-2.59	-1.868	-6.42	-1.296	-4.23	-2.262	-7.01
# observations	919		2587		3050		1635		1754
Obs. at one:	328		1153		1054		405		721
Log likelihood (0):	-598.69		-1777.9		-1966.2		-915.83		-1187.9
Log likelihood:	-508.72		-1500.9		-1679.3		-838.36		-1033.7

Table 5
Structural probit: Decision to volunteer

Variables	BC			Prairies			Ontario			Quebec			Atlantic
	est.coef.	t-ratio	marg. eff.	est.coef.	t-ratio	marg. eff.	est.coef.	t-ratio	marg. eff.	est.coef.	t-ratio	marg. eff.	est.coef.	t-ratio	marg. eff.
EARNDIF	0.124	7.70	0.04	0.153	18.21	0.06	0.162	22.19	0.05	0.168	16.70	0.03	0.191	16.52	0.07
MALE	-0.078	-0.77	-0.03	-0.005	-0.08	-0.00	-0.049	-0.83	-0.02	0.181	2.13	0.04	0.058	0.77	0.02
MARRIED	0.056	0.48	0.02	0.105	1.47	0.04	-0.060	-0.86	-0.02	-0.039	-0.40	-0.01	-0.062	-0.69	-0.02
HOURS	-0.008	-2.04	-0.00	-0.009	-4.29	-0.00	-0.011	-5.06	-0.00	-0.007	-1.94	-0.00	-0.014	-4.78	-0.01
HIGH	0.744	1.58	0.26	0.653	3.40	0.25	0.400	2.11	0.13	0.246	1.31	0.05	0.597	3.02	0.23
DIPLOMA	0.970	2.07	0.34	0.806	4.14	0.31	0.615	3.25	0.21	0.403	2.14	0.08	0.769	3.91	0.29
POSTSEC	1.130	2.35	0.39	1.065	5.19	0.42	0.715	3.57	0.24	0.672	3.14	0.14	1.089	5.03	0.42
UNIV	1.428	2.99	0.50	1.330	6.58	0.52	1.055	5.41	0.35	0.796	3.96	0.16	1.404	6.69	0.54
OWNK05	-0.121	-1.31	-0.04	-0.140	-2.69	-0.05	-0.019	-0.38	-0.01	-0.017	-0.22	-0.00	-0.162	-2.43	-0.06
OWNK0612	0.314	4.07	0.11	0.287	6.09	0.11	0.231	5.26	0.08	0.233	3.50	0.05	0.249	4.49	0.09
OWNK1317	0.306	2.90	0.11	0.106	1.76	0.04	0.179	2.84	0.06	0.179	2.07	0.04	0.150	2.07	0.06
OWNK18PL	0.075	0.61	0.03	0.011	0.15	0.00	0.050	0.77	0.02	0.288	3.12	0.06	0.173	2.01	0.07
RURAL	0.083	0.67	0.03	0.246	3.84	0.10	0.173	2.39	0.06	0.026	0.27	0.01	0.061	0.76	0.02
TOWN	0.015	0.12	0.01	0.173	1.51	0.07	0.011	0.13	0.00	-0.087	-0.73	-0.02	-0.116	-1.02	-0.04
NEWLAND	-1.161	-2.55	-0.41	-0.018	-0.07	-0.01	-0.249	-1.09	-0.08	0.453	0.52	0.09	6.445	0.02	2.46
MEDLAND	-0.405	-1.22	-0.14	-1.066	-2.63	-0.42	-0.174	-0.83	-0.06	-0.417	-0.59	-0.09	-0.051	-0.09	-0.02
OLDLAND	-0.210	-1.50	-0.07	0.022	0.17	0.01	-0.193	-2.31	-0.06	0.087	0.31	0.02	0.439	1.84	0.17
MEDRES	0.073	0.37	0.03	0.186	1.66	0.07	-0.106	-0.85	-0.04	0.112	0.52	0.02	0.259	1.53	0.10
LONGRES	0.257	1.37	0.09	0.295	2.82	0.11	0.233	1.97	0.08	0.246	1.22	0.05	0.395	2.55	0.15
ENG				-0.076	-0.17	-0.03	0.008	0.41	0.00	0.231	1.48	0.05	0.165	1.34	0.06
REL	0.139	1.44	0.05	0.231	3.83	0.09	0.218	3.72	0.07	0.197	2.40	0.04	0.391	5.04	0.15
AGE	-0.006	-1.32	-0.00	-0.006	-2.13	-0.00	0.003	1.25	0.00	-0.005	-1.15	-0.00	-0.004	-1.13	-0.00
GIVE	0.717	4.75	0.25	0.871	9.36	0.34	0.697	6.63	0.23	0.445	3.74	0.09	0.627	4.47	0.24
CONSTANT	-1.830	-3.19	-0.64	-1.826	-3.59	-0.71	-1.802	-5.71	-0.60	-1.598	-4.73	-0.33	-1.822	-5.47	-0.70
# observations	919			2587			3050			1635			1754
Obs. at one.	328			1153			1054			405			721
Loglikelihood0:	-598.69			-1777.9			-1966.2			-915.83			-1187.9
Log likelihood:	-484.62			-1311.8			-1388.6			-647.35			-879.57

It is interesting to note that the impact of age on the decision to volunteer differs quite markedly across the regions. Age does not appear to matter in the Atlantic provinces or in British Columbia; by contrast, it has a negative impact in the Prairies and Quebec, and a weakly positive one in Ontario. Finally, in all regions, if an individual donates money to charity then he or she is more likely to become a volunteer, suggesting that donating money and time are complementary activities.

In order to ascertain how any expected earnings differential may affect the decision to volunteer, estimates of this differential were computed from the earnings equation and then included in a structural probit analysis of the decision to volunteer. By and large, the results from the structural model accord with the results already presented from the reduced-form model. Table 5 reports the results from the structural probit model. For each region, the estimated coefficients are reported, their t-ratios and the marginal effect of the given variable on the probability of volunteering.

The variable of particular interest, however, is the impact of the differential itself (EARNDIF). In all cases, the estimated coefficient for EARNDIF is positive and statistically significant indicating that the estimated differential attributable to volunteering does indeed matter in the decision to volunteer - however, its impact varies quite a bit across regions. Interpreting the marginal effects for the earnings differential is a bit complicated because the differential is in logarithms. For instance, in Atlantic Canada, if the difference in earnings between volunteers and non-volunteers were to increase by 10 per cent (about $1,800), this would increase the probability of an individual deciding to volunteer by 0.07^*0.10 or 0.7 per cent. In Quebec, the response to a 10 per cent increase in the earnings differential in real terms (as opposed to logarithms) would elicit a much smaller response — a 0.3 per cent increase in the probability of volunteering.

4.2 The Labour Market Response to Volunteers

We are now in a position to assess how the labour market treats volunteers in relation to non-volunteers. To this end, separate earnings equations are determined for each group by region; these equations are estimated using a weighted least squares procedure corrected for sample selection. Any selection bias associated with the choice of whether or not to volunteer is taken into account by the inclusion of the inverse Mill's ration computed from the reduced-form probits previously estimated.

In order to present the results in a manner conducive to inter-regional comparisons, we report all of the earnings equations for volunteers by region in Table 6, and all of the non-volunteers earnings equations in Table 7. The only drawback with this presentation, is that one needs to consult both tables in order to compare across volunteers and non-volunteers; however, it does facilitate inter-regional comparisons, the principal focus of this paper.

Turning first to Table 6, we find several similarities and differences across the regions with respect to the determinants of volunteer earnings. As expected, being male has a positive influence on earnings in all regions. However, being married has no influence in the Atlantic provinces, the Prairies and Quebec, and a positive impact on earnings in British Columbia and Ontario. The number of hours worked has the expected positive sign across all regions. The impact of level of education is rather interesting: earnings increase with educational level in Ontario, having a university degree has a positive impact on earnings in Quebec, but elsewhere the level of education does not seem to be an important determinant of income. For Canada as a whole, however, earnings increase with educational level - a result which is apparently driven by the province of Ontario. That education does not affect earnings, ceteris paribus, is indeed a strange result: part of the explanation may simply be that occupational classification and other human-capital attributes dominate the model; another part may be tied to a variety of factors that explain persistent regional disparities in Canada. We return to this issue in the next section.

Table 6
Weighted OLS regressions: Dependent variable = log income
Volunteers' earning equations

Variables	BC		Prairies		Ontario		Quebec		Atlantic
	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio
MALE	0.405	5.15	0.438	8.66	0.219	5.18	0.352	5.52	0.441	8.24
MARRIED	0.182	2.12	-0.095	-1.54	0.103	2.03	0.084	1.17	0.047	0.80
HOURS	0.025	8.81	0.022	12.38	0.023	14.41	0.023	10.46	0.024	12.02
HIGH	0.233	0.55	-0.311	-1.44	0.202	1.12	-0.072	-0.49	-0.155	-0.81
DIPLOMA	0.349	0.81	-0.158	-0.71	0.331	1.77	0.182	1.18	0.073	0.37
POSTSEC	0.280	0.64	-0.200	-0.86	0.334	1.72	-0.043	-0.25	0.129	0.63
UNIV	0.375	0.84	0.037	0.16	0.495	2.50	0.381	2.09	0.340	1.55
HHSIZE	-0.166	-3.09	-0.120	-4.42	-0.106	-5.36	-0.046	-1.47	-0.102	-4.47
OWNK05	0.210	2.65	0.224	4.62	0.175	4.49	0.094	1.67	0.169	3.43
OWNK0612	-0.012	-0.16	0.066	1.46	-0.023	-0.63	0.042	0.74	0.095	2.13
OWNK1317	0.118	1.29	0.090	1.86	-0.019	-0.42	0.091	1.47	0.111	2.19
OWNK18PL	0.080	0.82	0.157	2.81	0.032	0.73	0.009	0.13	0.174	3.34
EXP	0.033	3.05	0.054	8.09	0.044	7.36	0.061	7.54	0.042	5.42
EXPSQU	-0.000	-2.09	-0.001	-5.07	-0.001	-4.65	-0.001	-5.32	-0.001	-3.35
RURAL	-0.055	-0.65	-0.259	-5.00	-0.318	-6.21	-0.209	-2.98	-0.157	-3.23
TOWN	-0.185	-2.08	-0.229	-2.71	-0.100	-1.70	-0.130	-1.63	-0.023	-0.32
MANAGER	0.237	1.82	0.712	7.91	0.579	8.14	0.361	3.45	0.537	6.33
SCIENCE	0.094	0.54	0.455	3.81	0.734	7.55	0.400	2.88	0.399	2.81
SOCSC	-0.173	-0.88	0.290	1.96	0.410	3.43	0.062	0.38	0.462	3.01
TEACH	0.295	1.76	0.169	1.51	0.454	4.98	0.439	3.64	0.526	5.05
RELIGION			0.263	1.18	-1.283	-4.70			-0.227	-0.98
HEALTH	0.369	2.20	0.509	4.80	0.502	5.42	0.489	3.76	0.478	4.82
ARTS	0.209	0.83	0.057	0.34	0.318	2.63	0.057	0.39	0.062	0.41
CLERK	0.180	1.45	0.446	5.08	0.275	3.83	0.218	2.25	0.250	3.19
SALES	-0.118	-0.89	0.158	1.65	0.184	2.43	0.109	0.96	-0.119	-1.38
PRIMARY	0.125	0.65	0.270	2.63	0.046	0.38	-0.189	-1.23	0.189	1.50
PROCESS	0.040	0.20	0.834	3.39	0.708	3.93	0.449	2.87	0.141	1.00
MACHINE	0.028	0.07	0.245	1.09	0.759	4.26	0.075	0.32	0.307	1.18
FABRIC	0.079	0.36	0.044	0.35	0.618	6.72	-0.004	-0.03	0.397	3.40
CONSTRUC	-0.099	-0.58	0.378	3.21	0.509	4.42	0.178	1.27	0.296	2.30
TRANSP	0.486	2.35	0.304	2.15	0.436	2.97	0.244	1.31	0.031	0.23
MATERIAL	-0.522	-1.51	0.297	1.52	0.800	4.04	-0.278	-1.15	-0.166	-0.88
OTHER			0.360	1.75	0.523	1.92	0.119	0.51	0.338	1.57
INVMILLS	-0.460	-2.79	-0.509	-4.62	-0.317	-3.11	0.021	0.13	0.003	0.03
CONSTANT	9.053	17.76	8.933	30.55	8.537	34.04	8.130	25.60	8.204	30.71
# observations	614		1876		1976		846		1193
Adj.R-square	0.3345		0.3856		0.4099		0.4735		0.5183

Table 7
Weighted OLS regressions: Dependent variable = log income
Non-volunteers' earning equations

Variables	BC		Prairies		Ontario		Quebec		Atlantic
	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio	est.coef.	t-ratio
MALE	0.425	3.84	0.381	6.16	0.323	5.83	0.321	5.42	0.351	5.58
MARRIED	0.117	1.14	0.052	0.82	0.066	1.07	0.231	3.80	0.105	1.65
HOURS	0.019	5.01	0.017	9.40	0.024	10.48	0.018	7.35	0.013	5.17
HIGH	0.366	1.24	-0.060	-0.44	-0.008	-0.06	0.075	0.73	0.064	0.54
DIPLOMA	0.365	1.21	-0.045	-0.30	0.133	0.93	0.216	1.92	0.194	1.46
POSTSEC	0.393	1.29	-0.127	-0.76	0.035	0.23	0.211	1.64	0.201	1.30
UNIV	0.399	1.22	-0.021	-0.12	0.202	1.21	0.505	3.50	0.434	2.38
HHSIZE	-0.134	-3.73	-0.115	-4.29	-0.090	-4.07	-0.042	-1.38	-0.108	-3.30
OWNK05	0.056	0.62	0.185	3.80	0.256	5.58	0.038	0.76	0.147	2.70
OWNK0612	0.144	1.35	-0.005	-0.08	0.040	0.75	0.062	1.13	0.031	0.51
OWNK1317	-0.026	-0.24	0.010	0.15	-0.025	-0.32	0.044	0.76	0.066	1.02
OWNK18PL	0.140	1.28	0.203	3.00	0.044	0.82	-0.039	-0.56	0.080	1.13
EXP	0.042	3.17	0.034	5.24	0.041	5.57	0.046	6.06	0.037	4.72
EXPSQU	-0.001	-2.06	-0.000	-3.93	-0.001	-3.97	-0.001	-4.39	-0.001	-3.32
RURAL	-0.341	-2.90	-0.098	-1.51	-0.297	-3.99	-0.189	-2.87	-0.159	-2.64
TOWN	-0.068	-0.55	-0.061	-0.56	-0.160	-1.98	-0.228	-2.92	-0.017	-0.21
MANAGER	0.662	3.73	0.512	4.91	0.546	5.58	0.460	4.76	0.553	5.89
SCIENCE	0.733	3.32	0.666	4.47	0.473	3.58	0.346	2.60	0.192	1.25
SOCSC	0.684	1.40	0.276	1.03	-0.231	-0.87	0.107	0.45	0.258	0.88
TEACH	0.969	2.47	0.268	1.50	0.455	2.48	0.487	3.97	0.287	1.64
RELIGION			0.186	0.64	0.094	0.12			0.697	0.75
HEALTH	0.977	4.50	0.558	4.46	0.623	4.35	0.625	5.34	0.447	3.58
ARTS	0.654	2.73	0.004	0.02	-0.087	-0.39	0.321	1.84	0.227	1.08
CLERK	0.302	1.85	0.172	1.76	0.279	3.04	0.383	4.13	0.333	3.63
SALES	0.049	0.30	0.165	1.64	0.047	0.45	0.127	1.26	-0.003	-0.03
PRIMARY	0.208	0.88	0.174	1.34	0.113	0.58	0.029	0.21	0.199	1.57
PROCESS	0.349	1.59	0.461	2.75	0.422	2.60	0.162	1.03	0.128	0.98
MACHINE	1.825	4.85	0.837	3.49	0.395	2.66	0.028	0.14	0.246	1.28
FABRIC	0.573	2.79	0.248	2.19	0.406	4.08	0.079	0.77	0.183	1.62
CONSTRUC	0.233	1.14	0.088	0.71	0.177	1.33	0.196	1.48	0.292	2.37
TRANSP	0.098	0.45	0.096	0.65	0.265	1.93	0.299	2.48	0.330	2.38
MATERIAL	0.218	0.70	-0.023	-0.15	0.250	1.69	0.314	1.97	0.106	0.48
OTHER			0.351	0.93	0.403	1.71	0.007	0.03	-0.001	-0.01
INVMILLS	-0.208	-0.93	-0.505	-4.36	-0.587	-3.84	-0.362	-1.71	-0.131	-0.89
CONSTANT	8.075	20.79	8.457	45.01	8.079	42.60	8.110	46.81	8.574	47.21
# Observations	305		711		1074		789		561
Adj.R-square	0.4353		0.4164		0.3910		0.3621		0.3560

Household size has a negative impact on earnings in all regions except Quebec - a result that may be partly explained by the generous child-care subsidies available in that province. The other determinants of earnings behave largely as expected. The only other difference worthy of note concerns the selectivity variable INVMILLS. In all of the analysis undertaken in Devlin (2000), selectivity was a problem. Here, however, we see that for two regions - the Atlantic and Quebec -no selectivity bias is present.

Looking at Table 7 we also find some notable inter-regional differences across the determinants of earnings for non-volunteers, as well as differences in comparison to the volunteer groups. For instance, being married has a weakly positive influence in Atlantic Canada, no influence in British Columbia or Ontario, and a strong positive influence in Quebec - a pattern that differs rather significantly from that displayed for the volunteer groups. The level of education is completely irrelevant in British Columbia, the Prairies and Ontario, while having a university degree exerts a positive impact on earnings in Atlantic Canada and Quebec. Once again, the size of the family has a negative impact on earnings everywhere except Quebec. Finally, some inter-regional differences exist regarding selectivity bias: no bias is found in the Atlantic region or in British Columbia for the non-volunteers, and the estimated coefficient on INVMILLS is significant at the 10 per cent level in Quebec.

It seems clear, therefore, that important differences exist across Canada's five regions -differences that are not revealed when using Canada-wide data. In order to compute the estimated earnings differential attributable to volunteering for each region, we employ the well-known Blinder (1973)-Oaxaca (1973) decomposition procedure which allows one to determine whether earnings increase because an individual has a higher "stock" of human capital relative to average -the stock effect -, or whether earnings increase because an individual earns a greater return to his or her average stock of human capital - the "return" effect. This decomposition procedure has been used extensively in studies of earnings gaps due to, for instance, gender (e.g., Miller, 1987), and entails determining the following:

which can be rewritten as:

where a bar denotes the sample mean, and a hat denotes the OLS estimate of the coefficient. The first term on the right-hand size represents the "stock effect" and second term the "return effect."

Table 8 presents these two effects for each of the five regions. Note that various effects are summed together for the sake of brevity - hence "education" is comprised of the impact associated with each of the four levels of education included in the earnings equations. A positive sign means that the volunteer has the higher stock (or return) in comparison to the non-volunteer, whereas a negative sign means the converse. Thus, for instance, the negative sign for MALE in the stock columns for every region but Quebec means that there are fewer male volunteers in all provinces but Quebec relative to male non-volunteers.

Many of the differences already discussed with respect to the earnings equations are further revealed by this decomposition procedure. Male volunteers earn a lower return to being male in British Columbia and Ontario, and a higher return in the other three regions. The impact of being married also differs across regions. Notice that, in all regions volunteers work fewer hours relative to non-volunteers, and in all but one region - Ontario - volunteers gain a higher return for any given hour worked relative to non-volunteers. Ontario also stands out as the only region where the return to education is higher for volunteers than non-volunteers; the "stock" of education is higher in all regions for volunteers.

The main reason for undertaking this decomposition procedure is that it allows one to calculate the overall difference in expected earnings between volunteers and non-volunteers, taking into account the differences in characteristics across the two groups. The last row entitled "average effect" provides this difference in the log of earnings; these numbers are approximately equal to percentages for small changes.⁴ For Canada as a whole, the estimated differential is 4.25 per cent when the country is separated into its five main regions, we find considerable inter-regional variation in the differentials. They range from 12.52 per cent in British Columbia to 1.17 per cent in the Atlantic provinces. Quebec has the second highest labour-market return to volunteering -6.51 per cent; followed by Ontario (4.91 per cent) then the Prairie provinces (3.13 per cent). Why do regional differences exist in labour-market responses to volunteering? The following section offers some suggestions and concluding remarks.

Table 8
Decomposing the volunteer/non-volunteer earnings differential by region

Variables	BC		Prairies		Ontario		Quebec		Atlantic
	Stock effect (%)	Stock effect (%)	Stock effect (%)	Stock effect (%)	Stock effect (%)	Stock effect (%)	Stock effect (%)	Stock effect (%)	Stock effect (%)	Stock effect (%)
MALE	-4.03	-1.09	-3.90	3.22	-2.33	-5.89	0.43	1.61	-2.18	4.45
MARRIED	0.91	3.90	-1.05	-8.12	0.44	2.27	0.21	-8.79	0.08	-3.89
HOURS	-4.19	21.05	-4.60	18.52	-6.08	-2.28	-1.55	17.25	-6.23	43.20
EDUCATION	2.38	-7.43	3.42	-14.73	4.50	21.97	4.23	-9.99	7.41	-14.00
FAMILY	-2.27	-6.68	-1.38	1.90	-3.90	-8.42	-0.35	0.97	-0.96	5.68
EXPERIENCE	-0.45	-13.00	3.50	27.15	1.60	7.80	-0.04	16.66	-0.38	8.29
POPULATION	-0.79	5.05	-1.80	-7.49	-1.52	0.50	-0.65	1.06	0.30	0.07
OCCUPATION	1.00	-28.62	2.01	8.43	-3.36	12.47	14.23	-7.30	4.41	-0.85
INVMILLS	-63.99	13.08	-69.82	0.30	-44.72	-14.03	3.26	-15.23	0.43	-7.59
CONSTANT		97.71		47.57		45.86		2.02		-37.07
TOTAL	-71.44	83.97	-73.60	76.74	-55.36	60.28	8.25	-1.74	2.88	-1.71
Average effect	12.52		3.13		4.91		6.51		1.17

• ⁴This estimate of the earnings differential is approximated by (lnW_v - ln W_n)^*100. Technically speaking, because earnings are in logs, the actual differential should be calculated as (exp (lnW_v - ln W_n ) -1)^*100. When the differential is small, the approximation method is accurate.

Previous

Contents

Next