volunteering and well-being cristina rosemberg new directions in welfare ii 8 july, paris

14
Volunteering and well-being Cristina Rosemberg New Directions in Welfare II 8 July, Paris

Post on 19-Dec-2015

213 views

Category:

Documents


0 download

TRANSCRIPT

  • Slide 1
  • Volunteering and well-being Cristina Rosemberg New Directions in Welfare II 8 July, Paris
  • Slide 2
  • Motivation Explore potential positive effects of participating on civic engagements and of taking a more active role in society. Literature have established a positive correlation between volunteering and well-being (Li&Ferraro, 2005, Helliwell&Putman, 2004): Formal volunteering have beneficial effects on subjective well-being, particularly on depression among older people. Civic engagements have a robust positive correlation with happiness and life satisfaction However, the positive correlation found in the literature could be spurious given three main problems: 1.Reverser causality: does volunteering increases subjective well-being, or is it that people with higher levels of well-being is more willing to engage in this type of activities? 2.Self-selection: are there underlying characteristics that make individuals to selected themselves into the volunteering that are also correlated with their well- being? 3.Omitted variables: are there factors which can not observed- that determines a both, a higher propensity to volunteers and to report higher levels of well being? (e.g. personality traits).
  • Slide 3
  • Methodology (I) Instrumental variables Need to find an instrument (Z) that affects Mental Health indirectly just through its effects on volunteering. More precisely, the instrument has to full-fill two requirements: 1.Corr(X,Z)!=0 2.Corr(Z,ni)=0
  • Slide 4
  • Methodology (II) Data British Household Panel Survey (BHPS) 18 waves, random sample of aprox. 10,000 individuals (5,500 British households), 15 years and older. Includes measures of well-being, volunteering, social characteristics How to measure well-being? Preference satisfaction, hedonic accounts, evaluation accounts Combined measures: mental health GHQ12 Brief self-report measure, with excellent properties as a screening instrument for psychiatric disorders in nonclinical settings (Goldberg & Williams, 1988). Use extensively in medical, psychological and sociological research. GHQ-12 comprises six positive and six negative items concerning the past few weeks. Presence or intensity of the state is ranked by the respondent using a 4-point scale. It cover issues of social functioning (feeling capable of making decisions), anxiety and depression (being able to sleep well ) and confidence (thinking of oneself as worthless). Likert GHQ score: obtained by assigning the value of 3 to the most negative answer and the value of 0 the most positive ones. Score: from 0 (most posittive outcome) to 36. How to measure volunteering?: Memberships (W1-W5, W7, W9, W11, W13, W15, W17): Q.: Are you currently a (n active) member of any of the kinds of organisations [...]? It is not clear what are the resources (money, time) that individuals contribute to these organisations: what does active mean? Variable seems to be capturing a broad measure of social capital better than volunteering.
  • Slide 5
  • Methodology (III) How to measure volunteering? (cont): Unpaid voluntary work (W6,W8,W10,W12,W14,W16,W18): Q: We are interested in the things people do in their leisure time, I'm going to read out a list of some leisure activities. Please look at the card and tell me how frequently you do each one... Do unpaid voluntary work. Main concern: unpaid voluntary work questions could be capturing participation in informal volunteering or the existence of family strategies such as caring for a family member that lives inside or outside the household. According to the literature, this kind of volunteering might be detrimental to carers mental health (Li&Ferraro, 2005). However, caring for a family member does not seem to driven the responses to this question: Volunteering among individuals that do care for a household member is similar to volunteering among individuals that do not report providing that kind of support (20.6% and 20.7% respectively). And the difference is not statistically significant.
  • Slide 6
  • GHQ12: 36 point Likert scale Wave 6 Volunteering Average 7 waves Average score: 11.20
  • Slide 7
  • Methodology (IV) Instrument: Percentage of people in the region that engages in volunteering, per year. Positively correlated with volunteering...but not reason to believe that it is correlated with any underlying factors determining individual mental health. Other controls:. Second stage (Mental health): sex, age, age^2, physical health, marital status, financial strain, log annual income. First stage: instrument and covariates of 2 nd stage.
  • Slide 8
  • Slide 9
  • Results (I)
  • Slide 10
  • Results (II) Fixed-effects (within) IV and GLS regressions Model1 Model 2 Model 3 IV OLS IV OLS IV OLS b/se Volunteering -0.513 -0.281*** -1.324 -0.379*** -1.246 -0.383*** (0.377) (0.050) (2.251) (0.100) (2.254) (0.100) Age 0.017 0.016 0.030 0.026 0.027 0.023 (0.015) (0.026) (0.023) (0.026) (0.023) Age2 -0.000 - - - - (0.000) Base: livingcomfortably Finan.sit=doing alright 0.434*** 0.435*** 0.489*** 0.495*** 0.484*** 0.489*** (0.047) (0.094) (0.092) (0.094) (0.092) Finan.sit=jus aboutgetting by 1.410*** 1.408*** 1.576*** 1.574*** 1.575*** 1.573*** (0.057) (0.112) (0.111) (0.112) (0.111) Finan.sit=finding it quiet difficult 3.004*** 3.000*** 3.231*** 3.216*** 3.226*** 3.212*** (0.091) (0.177) (0.173) (0.178) (0.173) Finan.sit=finding it very difficult 4.879*** 4.873*** 5.433*** 5.429*** 5.432*** 5.428*** (0.136) (0.258) (0.257) (0.259) (0.258) Number ofhealth problems 0.577*** 0.650*** 0.654*** 0.648*** 0.652*** (0.019) (0.038) (0.036) (0.037) (0.036) Base:never married Marital status=married 0.300*** 0.302*** 0.271 0.269 0.300 0.298 (0.109) (0.108) (0.198) (0.197) (0.198) Marital status=separated 0.957*** 1.627*** 1.601*** 1.639*** 1.617*** (0.181) (0.345) (0.339) (0.345) (0.339) Marital status=divorced -0.117 -0.115 -0.123 -0.110 -0.098 -0.085 (0.156) (0.156) (0.283) (0.280) (0.283) (0.281) Marital status=widowed 1.332*** 1.329*** 1.487*** 1.495*** 1.524*** 1.531*** (0.189) (0.335) (0.333) (0.335) (0.334) Ln(Income) 0.044*** 0.046*** 0.024 0.035 0.020 0.030 (0.013) (0.037) (0.026) (0.037) (0.026) Trust -0.384*** -0.394*** -0.388*** -0.397*** (0.091) (0.088) (0.091) (0.088) Frequency talkstoneighbours(weekly or more=1) -0.234** -0.238** (0.094) (0.093) Frequencymeets people (weekly or more=1) 0.056 0.042 (0.110) (0.104) Constant 8.439*** 8.392*** 8.412*** 8.358*** 8.661*** 8.631*** (0.337) (0.328) (0.534) (0.516) (0.544) (0.537)
  • Slide 11
  • Results (III) Validity of the instruments: Weakness: first-stage regression shows a strong (positive) correlation between the instrument and volunteering. Over identification: We cannot reject the null that the instruments are valid. Hausman test of endogeneity: There are no systematic differences between IV and OLS estimates. If endogeneity is ruled out, then OLS provides consistent and efficient estimators, while IV provides consistent but inefficient estimators. Fixed effects seem to be removing problems of omitted variables and reversed causality.
  • Slide 12
  • Results (III) What about self-selection? A treatment effect model The idea behind the model is to regress two equations simultaneously: The first is the probability of volunteering controlling by personality traits (Big 5: extraversion, openness, neuroticism, agreeableness and conciousteness). The second is the outcome regression (mental health) as a function of the treatment variable (volunteering). To simultaneously estimate the two regressions we have to assume that the error terms are jointly normally distributed. Estimate treatment effect model using Wave 16. Wald-test tests the null that the correlation between the error terms of the two equations is biased towards zero. With a chi2(1)= 119.26, p-value=0.000, we can conclude that there is selection bias in our model. However, once the model have been corrected, volunteering is still positive and significantly correlated with mental health. E(Mental Health volunteering=1)= 11.25 E(Mental Health volunteering=0)= 11.53
  • Slide 13
  • Results (IV) What are the mechanisms through which volunteering generates a positive effect on mental health? Hypothesis: Volunteering as a buffer mechanism to deal with potentially negative personal episodes/situations: Retirement Financial strain Termination of marriage
  • Slide 14
  • Conclusions Fixed effect models seem to be successfully dealing with issues of reverse causality and omitted variables. Self-selection problem is not tackle with OLS estimations, however: Treatment effects model provide similar OLS estimators once estimation have been corrected by selection bias. Volunteering has a positive effect on mental health. Volunteering seems to be playing a role on alleviating potential negative effects of personal episodes/situations: It increases well-being among retirees: Hypothesis: Helps volunteers to find a sense of purpose after their working life. Decreases the negative effects of being on financial strain: Hypothesis: Helps volunteers to see things in perspective/Helps volunteers to achieve personal satisfaction that is not related to monetary rewards. Deludes the negative effect of being separated, divorced or widowed (as opposed to being married). Hypothesis: Helps volunteers to see things in perspective Further research: Test this results with other measures of well-being such as life satisfaction. More in-depth analysis needed to understand how those mechanism work in the field work