This article was downloaded by: [University of Calgary] On: 19 March 2015, At: 13:35 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Anxiety, Stress, & Coping: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gasc20
Financial strain, social capital, and perceived health during economic recession: a longitudinal survey in rural Canada a
a
Christine Frank , Christopher G. Davis & Frank J. Elgar
b
a
Department of Psychology, Carleton University, Ottawa, ON, Canada b
Click for updates
Institute for Health and Social Policy and Douglas Institute, McGill University, Montreal, QC, Canada Accepted author version posted online: 20 Nov 2013.Published online: 10 Dec 2013.
To cite this article: Christine Frank, Christopher G. Davis & Frank J. Elgar (2014) Financial strain, social capital, and perceived health during economic recession: a longitudinal survey in rural Canada, Anxiety, Stress, & Coping: An International Journal, 27:4, 422-438, DOI: 10.1080/10615806.2013.864389 To link to this article: http://dx.doi.org/10.1080/10615806.2013.864389
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Downloaded by [University of Calgary] at 13:35 19 March 2015
Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions
Anxiety, Stress & Coping, 2014 Vol. 27, No. 4, 422–438, http://dx.doi.org/10.1080/10615806.2013.864389
Financial strain, social capital, and perceived health during economic recession: a longitudinal survey in rural Canada Christine Franka*, Christopher G. Davisa and Frank J. Elgarb a
Department of Psychology, Carleton University, Ottawa, ON, Canada; bInstitute for Health and Social Policy and Douglas Institute, McGill University, Montreal, QC, Canada
Downloaded by [University of Calgary] at 13:35 19 March 2015
(Received 27 May 2013; accepted 5 November 2013) Although the health consequences of financial strain are well documented, less is understood about the health-protective role of social capital. Social capital refers to a sense of community embeddedness, which is in part reflected by group membership, civic participation, and perceptions of trust, cohesion, and engagement. We investigated whether perceptions of social capital moderate the relation between financial strain and health, both mental and physical. This longitudinal study surveyed adults in two communities in rural Ontario where significant job losses recently occurred. Data were collected on financial strain, social capital, perceived stress, symptoms of anxiety and depression, and physical health on three occasions over 18 months (N’s = 355, 317, and 300). As expected, financial strain positively related to perceived stress, poor physical health and symptoms of anxiety and depression, whereas social capital related to less stress, better physical health, and fewer symptoms of anxiety and depression. Effects of financial strain on perceived stress and depressive symptoms were moderated by social capital such that financial strain related more closely to perceived stress and depressive symptoms when social capital was lower. The findings underscore the health-protective role of community associations among adults during difficult economic times. Keywords: social capital; financial strain; health; stress; community psychology
Current economic trends of job losses and rising unemployment rates have had an impact on health in many industrialized countries (Bernard, 2009; Pilat, Cimper, Olsen, & Webb, 2006). In Spain, physicians have treated more mental health and alcohol problems since the recession hit, particularly in patients who were unemployed or had difficulty paying their mortgages (Gili, Roca, Basu, McKee, & Stuckler, 2013). Self-rated mental health has deteriorated in Britain (Katikireddi, Niedzwiedz, & Popham, 2012). In Italy, suicides and suicide attempts due to economic stressors have increased (De Vogli, Marmot, & Stuckler, 2013). Layoffs and unemployment during times of economic decline are sources of considerable stress for individuals and families, and the resulting financial or economic strain has negative consequences for mental and physical health (Dekker & Schaufeli, 1995; Fox & Chancey, 1998; Kinnunen & Pulkkinen, 1998; Sverke, Hellgren, & Näswall, 2002). However, in some cases, the relation between financial strain and health (mental and physical) is not particularly strong, which suggests that some individuals, families, or communities possess qualities or resources that protect them from the *Corresponding author. Email: [email protected] © 2013 Taylor & Francis
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
423
negative effects of such strain. This study explored the protective role of social capital in two rural communities in Ontario, Canada that were dealing with a large number of job losses and heightened economic uncertainty from 2008 to 2010. A large body of research has documented the health consequences of involuntary job loss. Studies have found that involuntary job loss followed by sustained periods of unemployment is associated with symptoms of anxiety and depression, alcohol and drug use, and physical health problems (Iversen & Sabrow, 1988; Keefe et al., 2002; Pearlin, Menaghan, Lieberman, & Mullan, 1981; Price, Choi, & Vinokur, 2002; Vinokur, Price, & Caplan, 1996). For example, a study by Studnicka et al. (1991) of adults who were laid off following a factory closure found that those who were still unemployed 12 months later were eight times more likely to report poor psychological health compared to those who had regained employment. In the USA during the 1980s, a series of studies led by Rand Conger documented social and psychological consequences of an economic downturn in farming communities in the Midwest as falling grain prices and high interest rates forced many farmers into bankruptcy and foreclosure (Conger et al., 1990, 1992, 1993; Conger, Reuter, & Elder, 1999). Their longitudinal study of these families found that economic pressure led to increased depression and demoralization in parents, which in turn contributed to marital conflict, parenting difficulties, and adjustment problems in their adolescent children (Conger et al., 1992, 1993). These studies highlight the pervasive and devastating effects of job loss and financial strain on health, well-being, and social relationships. It could be argued that the negative health effects of job loss should be somewhat mitigated in countries with comprehensive welfare programs, particularly in countries where unemployment does not impose barriers to accessing health care or social services. A study by the Organization for Economic Cooperation and Development (OECD) found that the health impact of economic cycles related directly to international differences in public service expenditures (Gerdtham & Ruhm, 2006). Specifically, the study found smaller health impacts during periods of economic growth and recession in countries that shared stronger labor protections and larger social investments. Nevertheless, within these countries with extensive welfare systems, the evidence suggests that job loss and financial uncertainty continue to negatively affect health and well-being. For instance, a longitudinal study in Sweden by Voss, Nylén, Floderus, Diderichsen, and Terry (2004) found that unemployment significantly increased the risk of disease at 10- and 24-year follow-ups. In a national longitudinal study of adults in Germany, Lucas, Clark, Georgellas, and Diener (2004) found that job loss was associated with a multi-year decline in life satisfaction that persisted even after reemployment. These results speak to the long-lasting effects job loss can have on individual health and well-being, even in countries where job loss does not limit access to health care and social services. It is important to note that most of the research assessing the psychological and physical health effects of job loss and economic downturn has emphasized the role of perceived economic or financial strain as the mediating construct between job loss and well-being (e.g., Broman, Hamilton, & Hoffman, 1990; Conger et al., 1990; Pearlin et al., 1981). Financial strain, defined as a feeling of unease about one’s perceived ability to “make ends meet,” is conceptualized as a psychological appraisal that one’s demands (e.g., bills, financial obligations) exceed available resources (Lazarus & Folkman, 1984). The research cited above suggests that the health effects of lost income or unemployment are at least partly mediated by the perception of financial strain.
Downloaded by [University of Calgary] at 13:35 19 March 2015
424
C. Frank et al.
Although job loss, economic hardship, and financial strain increase the risk of physical and mental health problems, some individuals are resilient in the face of these challenges. Perceived social capital is one factor that appears to buffer individuals from the negative health effects of financial strain. Social capital refers to the value accrued from belonging to groups, whether they are families, friendship networks, faith, work, or recreational groups (e.g., Coleman, 1988, 1990; Putnam, 1995). Not only do these groups contribute to a sense of identity, safety, trust, and belonging, they also help create norms of reciprocity and bonds of support that can be called upon in times of need. Social capital stems from the social connections and networks people establish in everyday interactions, which in turn, both forms and reinforces the individual’s perception of trust, safety, cohesion, engagement, and reciprocity within his or her community (Ahnquist, Wamala, & Lindstrom, 2012; Giordano & Lindström, 2010; Lochner, Kawachi, & Kennedy, 1999; Onyx & Bullen, 2000). This perception is reflected in terms of membership in community organizations (e.g., bowling leagues, political parties, religious groups), volunteerism, and interpersonal and intergroup trust (e.g., Putnam, 1993, 1995). Social capital is conceptually related to social support, but is distinct insofar as it refers to aspects of social contexts and not individuals per se. Whereas perceived social support refers to the sense that help is available if needed (usually from close others), social capital refers to the extent of integration in broad social networks. Social capital may also be conceptualized and operationalized at the community or state level (e.g., Kawachi, Kennedy, & Glass, 1999; Putnam, 1993, 1995). From this more sociological perspective, social capital may be reflected in terms of the scope and depth of community organizations and degree of civic engagement and intergroup conflict (e.g., Kawachi & Berkman, 2000). No doubt, the psychological and sociological perspectives on social capital are closely related. Estimates of community-level social capital often are based, at least in part, on aggregated perceptions from groups of individuals (e.g., Kawachi, 1999; Kawachi et al., 1999). However, because social capital is sometimes assessed only at an ecological level (e.g., proportion of the population that votes; Putnam, 1995) and other times only at an individual level (e.g., perceptions of trust; Kennelly, O’Shea, & Garvey, 2003), prior research has not always separated the compositional effects of individual perceptions of social capital from the contextual effects of social capital (Elgar et al., 2011). Therefore, we adopt a multilevel approach to investigating individuals’ perspectives on social capital in their community, while taking into account community-level differences. Like financial capital, perceived social capital can be regarded as a resource or currency that can be leveraged in times of need or stress. Individuals who perceive higher levels of social capital likely have wider social networks from which aid and support are available, and these networks might be more willing to provide assistance because it is good for the community and not just for the individual in need. Those who have higher levels of perceived social capital may also feel less embarrassed to receive support to the extent that they see their role in the community to be more than merely a tax-payer and service-user. Population studies have found that high levels of perceived social capital confer advantages in terms of individual health and well-being (Elgar et al., 2011; Subramanian, Kawachi, and Kennedy, 2001). Kawachi et al. (1999) found that even when controlling for other risk factors (e.g., obesity, smoking, lack of access to health care) individuals who lived in US states with higher levels of social capital reported better
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
425
health. Kawachi and Berkman (2000) argued that the availability of social capital benefits health because it promotes health-related behaviors, facilitates access to services and amenities, and provides psychosocial processes that support health (e.g., perceptions of safety, social support). These influences on health may be especially beneficial in economically deprived areas that lack material resources (Cramm, Møller, & Nieboer, 2012; De Silva, Huttly, Harpham, & Kenward, 2007; Gore, 1978; Sapag et al., 2008). Whereas research has established links between perceived social capital (measured at the individual level) and both mental and physical health (e.g., Giordano, Björk, & Lindström, 2012; Giordano & Lindström, 2010; Verhaeghe & Tampubolon, 2012), we are not aware of any studies that conceptualized social capital as a moderator of the financial strain – health/well-being link. Some research has found that perceived social support moderates the effect of financial strain on health and well-being. For instance, Schwarzer, Hahn, and Jerusalem (1993) found in a study of former East Germans who migrated to West Germany following the collapse of the Berlin Wall that social support buffered the effect of prolonged unemployment on health. Likewise, in their study of farming communities in the US Midwest during the economic recession of the 1980s, Conger et al. (1999) found that for both men and women, spousal support buffered the effects of perceived economic strain on emotional distress. Perceived social capital is a broader concept than perceived social support insofar as it emerges from one’s embeddedness within a community as opposed to one’s embeddedness within a social relationship. People with high levels of social capital are well integrated into their community, they are civically minded and active participants in their community. And so when they need help, their community is there for them. Social capital, as a measure of the extent of integration within social networks, is a protective resource that can be leveraged to maintain health and well-being once coping resources have been depleted. Protective resources are relationships, objects, or personal characteristics that are either valued directly (e.g., close attachments), or indirectly allow one to obtain other valued means (e.g., social support; Diener & Fujita, 1995; Hobfoll, 2002). As such, social capital might be particularly important when financial resources are depleting, and therefore might buffer the negative effects of financial strain on health and well-being. Although social capital may be regarded as a personal resource, it is also likely to be a resource that is shared within families. Couples share many of the same values and worldviews (Antonovsky & Sourani, 1988; Ransom, Fisher, & Terry, 1992; Roest, Dubas, Gerris, & Engels, 2009), and tend to interpret adversity through the same lens (Davis, Harasymchuk, & Wohl, 2012). Family systems theorists have long argued that coping and resilience are family-based processes (Patterson, 2005; Rosenblatt, 2000; Walsh, 1996). Indeed, Coleman’s (1988) original writings on the concept emphasized the important role that family connections play in children’s health and education. Therefore, social capital, with its emphasis on social embeddedness or integration, can also be considered a family-based resource. In this report, we assess the role that family-level social capital plays as individuals cope with financial uncertainty brought on by the economic recession of 2008. Our focus is on two towns in Ontario, Canada that suffered significant job losses due to the recession. Previous research shows that social capital is correlated positively with health and wellness (e.g., Elgar et al., 2011; Giordano et al., 2012; Kawachi et al., 1999). Here we assess the extent to which social capital buffers people from the negative effects of financial strain.
426
C. Frank et al.
Downloaded by [University of Calgary] at 13:35 19 March 2015
Method Participants Adults were recruited by telephone using a random dialing procedure from two predominantly English-speaking communities in rural Ontario. The communities had similar demographic and economic characteristics, and populations were comparable (between 7800 and 8800) at the time of data collection (Statistics Canada, 2007). Our plan was to compare one town facing significant job losses to the other. However, in the early stages of data collection, the economic recession also hit the second town. Therefore, the decision was made to combine the samples. A sample of 355 adults from 240 families was recruited initially and 89.3% (n = 317) of the original sample was followed up at nine months and 84.5% (n = 300) of the original sample was followed up again at 18 months. Only individuals who provided complete data for at least two waves of testing were included in the analyses. The gender and age composition of two community samples did not differ significantly over the course of the study. The sample was 62.5% female and had an average age of 49.5 years at Time 1 (SD = 14.3 years). In Town A, over the course of the study, 58.6% of participants were employed, 18.2% were unemployed, and 23.2% were retired. In Town B, 66.9% of participants were employed, 13.4% were unemployed, and 19.7% were retired. Families were reimbursed with a $50 gift card redeemable at a local store for each assessment. A university research ethics board approved the study procedures.
Measures Financial strain was assessed with two items adapted from Pearlin et al.’s (1981) seminal work on job disruption, coping, and stress. Respondents were asked the extent to which they have had difficulty paying bills in the past 12 months (with response options ranging from 1 = “a great deal of difficulty” to 5 = “no difficulty at all”) and the extent to which they have enough money left at the end of the month (with response options ranging from 1 = “more than enough money left over” to 5 = “not enough to make ends meet”). After reverse scoring the first item, the two items were averaged so that higher scores indicated higher financial strain. The correlation between the two items was strong (r’s = .66–.72; Spearman–Brown = .80–.83), and test-retest correlations were quite high (rtime1, time2 = .68; rtime2, time3 = .65). Financial strain was moderately and negatively correlated with perceived SES using a modified version of Cantril’s (1965) ladder where participants were asked to rank themselves based on the amount of money they have, their level of education, and how respected their job is relative to other people in Canada (rtime1 = −.47, rtime2 = −.42, rtime3 = −.42, p’s < .001). Financial strain was also moderately and negatively correlated with family income (assessed as self-reported household income; rtime1 = −.44, rtime2 = −.31, rtime3 = −.31, p’s < .001) providing evidence for convergent validity. Social capital was measured with Onyx and Bullen’s (2000) Social Capital Scale which assessed eight aspects of the construct: participation in the local community, social agency, feelings of trust and safety, neighborhood connections, work connections, family and friends connection, tolerance of diversity, and value of life. We excluded “work connection” items from the scale because many of the individuals in the study were not employed and thus unable to answer questions about work connections. Rather than parsing social capital into its components or manifestations, research on social capital has tended to focus on the concept in its whole (e.g., Sarracino, 2010). That is, although
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
427
participation in the local community, feelings of safety and trust, tolerance for diversity, and family and friendship connections each may be seen as conceptually distinct, social capital is the common thread. Past research confirms that although the subscales capture distinct aspects, they cohere in a common second-order factor (Bullen & Onyx, 2007; O’Brien, Burdsal, & Molgaard, 2004). Therefore, we use total score for social capital in our analyses. Responses to the 30 items were on a 4-point Likert scale (from 1 = “no, not much” to 4 = “yes, very much”). Internal consistency of the scale was high (α = .82–.84), as were its test-retest correlations (rtime1, time2 = .79; rtime2, time3 = .75). The extent to which families shared a common sense of social capital was assessed by means of an intercepts-only hierarchical linear model (HLM). The data are structured hierarchically such that assessment cycle (times 1, 2, and 3) is nested within individuals, and individuals are nested within families. The intercepts-only model apportions the total variance into within-person variance (i.e., within-person fluctuations), between-person variance (reflecting individual differences), and between-family variance (i.e., differences between families). Confirming our expectation that social capital is a shared family resource, the intercepts-only HLM indicated that 47.9% of the total variance in social capital was located at the third (family) level, with the remainder divided approximately equally between the first two levels. Therefore, assessments of social capital were relatively stable across time and across family members, but differed considerably from one family to the next. Perceived stress was assessed with the Perceived Stress Scale (Cohen, Kamarck, & Mermelstein, 1983), which is a widely used general measure of appraised stress during the previous month. It contained 14 items – 7 positive (e.g., “In the last month, how often have you felt that things were going your way?”) and 7 negative (e.g., “In the last month, how often have you felt that you were unable to control the important things in your life?”), each rated using a scale ranging from 1 (“never”) to 4 (“very often”). The seven positively valenced items were reversed scored and a mean score on the 14 items was calculated with higher scores indicating more stress. Consistent with past research (e.g., Cohen et al., 1983), the instrument was internally consistent at all three times (α = .84–.91) and test-retest correlations were strong (rtime1, time2 = .69; rtime2, time3 = .60). General anxiety was assessed with the Zung Anxiety Scale (Zung, 1971), which consisted of 20 items measuring symptoms experienced over the past week. Participants rated each item on a 4-point scale ranging from 1 (“none or a little of the time”) to 4 (“most or all of the time”) to indicate how much the statement describes how they felt during the past week. A mean score of the 20 items was calculated with higher scores indicating more anxiety. The scale was internally consistent (α = .83–.88) and anxiety scores were consistent over time (rtime1, time2= .71; rtime2, time3= .68). Symptoms of depression were assessed with the Center for Epidemiologic Studies – Depression Scale (CES-D; Radloff, 1977), which consisted of 20 items that describe depressive feelings and behaviors experienced over the past week. Participants rated each item on a 4-point scale ranging from 0 (“rarely or none of the time”) to 3 (“all the time”). Scores were averaged so that higher scores indicated more symptoms. The scale was also internally consistent (α = .93–.94) and yielded high test-retest correlations (rtime1, time2= .75; rtime2, time3= .66). Physical health was measured using the Balanced Physical Health Scale (BPHS; Godin & Mantler, 2007). The BPHS is a 20-item instrument assessing general health in terms of mobility issues (e.g., “During the past three weeks, have you had any difficulty
428
C. Frank et al.
walking?”), digestive issues (e.g., “During the past three weeks, have you been bothered by indigestion/heartburn?”), general fitness (e.g., “If I have to exert myself [e.g., run for the bus or take the stairs] I find I have difficulty catching my breath”), resistance (e.g., “I seldom get sick”), and energy (e.g., “I have the energy to do things I enjoy after I’m done work for the day”). Some items were reverse-scored and an average was calculated with higher scores indicating worse physical health. Cronbach’s alpha for the 20-item scale ranged from .87 to .91. In the current study, the BPHS highly correlated with the general health question, “In general, would you say your health is excellent, very good, good, fair, or poor?” (Ware, Kosinski, & Keller, 1996) rtime1 = −.69, rtime2 = −.70, rtime3 = −.63.
Downloaded by [University of Calgary] at 13:35 19 March 2015
Procedure Randomly selected families from the two communities were invited to participate in the community participation and health outcomes study. Our invitation to participate was extended to all adults living in the household. Self-completion surveys were delivered by a research assistant to those who agreed to participate. Prior to the first assessment, home visits were arranged for the purposes of obtaining informed consent. Completed surveys were returned by mail.1 Three assessments were carried out between Fall 2008 and 2010 with 9-month intervals. The data were inspected for normality and outliers. At Time 1, two outliers were identified: one on anxiety and the other on depression. At Time 2, four outliers were identified: one on anxiety and three on depression. At Time 3, one participant was identified as an outlier on stress. In all cases, participant’s scores were adjusted to within three standard deviations of the sample mean.
Analysis overview As noted above with respect to social capital, the data are analyzed using a three-level HLM (Raudenbush & Bryk, 2002) where the first level was the assessment cycle (Times 1, 2, and 3) and reflects within-person variability, the second level was individual differences, and the third level refers to family-level variance. We first assessed how much variance in the data was attributable to within-person differences (Level 1), the interpersonal differences (Level 2), and between families differences (Level 3) by means of an intercepts-only HLM. To the extent that variance is negligible at one level for a given variable, it suggests that at that level the variable is constant or invariant. Four three-level models were fitted to the data to determine the extent to which financial strain at timet for personi (of familyj) related to each of the four dependent variables (depression, anxiety, perceived stress, and physical health) at timet, taking into account the time of assessment. Therefore, the Level 1 (within person) model can be represented as: ytij ¼ B0ij þ B1ij Financial Straintij þ B1ij Timetij þ etij where ytij represents the dependent variable at timet for personi (of familyj), B0ij represents the intercept for personi (in familyj), B1ij represents the slope for (grand mean centered) financial strain on the dependent variable for personi (in familyj), and etij is the
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
429
random error for personi (in familyj). At Level 2, we modeled the effect of gender and age (grand mean centered) on the Level 1 intercept and the Level 1 slope for financial strain. This assesses whether the person-level intercepts and effects of financial strain vary as a function of gender or age of the participant. At Level 3 (between families), we assessed the extent to which the Level 1 intercepts and slopes for financial strain vary as a function of the community within which one resides (Town A vs. Town B) and one’s family’s perceived level of social capital (averaged over time and over members of the family). To the extent that community or social capital is associated with one’s Level 1 intercept, it suggests that community or social capital has a main effect on the dependent variable. The extent to which community or social capital relates to one’s Level 1 slope for financial strain indicates the degree to which community or social capital moderates the effect of financial strain on the outcome. These moderated effects, where significant, were followed up with simple slope analyses. To give a sense of the magnitude of effects, we provide in Table 2 the proportional reduction in residual variances (PRV) in the dependent variables from the intercept-only model (with no predictors) to the final model. The PRV may be interpreted as a pseudo-coefficient of determination (R2; see Peugh, 2010; Raudenbush & Bryk, 2002). However, since variance has been apportioned to each level, a separate PRV is calculated for each of the three levels. These estimated proportions indicate the proportion of variance at the particular level explained by predictors at that and lower levels.
Results Attrition We found no significant differences between individuals who dropped out of the study at Time 2 or Time 3 and those who remained in age, gender, employment status, anxiety, perceived stress, depressive symptoms, physical health, financial strain, or social capital (t’s < 1.8, p’s > .10).
Basic models An intercepts-only multilevel model of financial strain showed substantial betweenfamily variance (44.5%) but almost all the remaining variance (50.0%) was within-person variance, suggesting fluctuations over time within individuals. The small percentage of between-person variance (5.5%) suggests that families are internally consistent in their perceptions of financial strain. Financial strain did not differ between the two communities (B = .15, SE = .10, t = 1.49, p = .14), and negatively related to social capital (B = −.51, SE = .15, t = –3.46, p = .001). Financial strain was higher among females than males (B = .22, SE = .06, t = 3.41, p = .001), and among younger participants (B = −.02, SE < .01, t = −5.69, p < .001), and increased over the course of the study (B = .35, SE = .03, t = 12.65, p < .001).
Predictive models Based on these findings, we constructed a series of multilevel models where each outcome (depressive symptoms, anxiety symptoms, perceived stress, and physical health) for person i at time t was explained by financial strain at time t, time, and a random
430
C. Frank et al.
Table 1. Descriptive statistics on variables. Baseline (n = 315)
Downloaded by [University of Calgary] at 13:35 19 March 2015
Financial strain Social capital Perceived stress Anxiety symptoms Depressive symptoms Physical health Subjective SES
9 months (n = 312)
18 months (n = 285)
M
SD
α
M
SD
A
M
SD
α
2.37 2.67 1.72 1.73 0.74 3.62 5.55
0.81 0.40 0.65 0.42 0.58 0.66 1.91
.79 .83 .90 .83 .93 .87 –
2.99 2.76 1.64 1.64 0.66 3.66 5.60
1.07 0.40 0.45 0.45 0.55 0.67 1.62
.81 .82 .91 .88 .93 .89 –
2.99 2.71 1.89 1.63 0.67 3.66 5.49
1.05 0.41 0.48 0.43 0.54 0.63 1.57
.83 .84 .84 .87 .93 .87 –
intercept. At level 2 (between persons), we tested whether the level 1 random intercept and the effect of financial strain varied by gender and age. At level 3 (between families), we tested whether the level 2 intercept and effect for financial strain varied by average family social capital and community. Age and gender were excluded from these models because these variables did not relate to social capital. Descriptive statistics on all the variables are presented in Table 1. As shown in the regression analyses presented in Table 2, differences in mental and physical health between the two communities represented in the sample ranged from nonsignificant to modest once differences in other variables were held constant. Older respondents tended to report somewhat less stress (p < .10) but more physical health problems. Age did not significantly predict symptoms of depression or anxiety. Women reported more stress, depressive symptoms, and anxiety symptoms than men, and marginally more physical health problems. As well, assessment cycle related negatively to depressive, anxiety, and physical symptoms (Table 1). Most relevant to our study hypothesis, financial strain related positively to all four outcomes indicating that those with greater financial strain perceived more stress, had more symptoms of depression, anxiety, and ill-health. However, the main effects of financial strain on stress and depressive symptoms were moderated by social capital. In both cases, simple slopes analyses indicated that the effects of financial strain on stress and depressive symptoms were stronger when social capital was lower (see Figure 1). When social capital was low (−1 SD), the effect of financial strain on stress was B = .17 (SE = .03, t = 6.19, p < .001) whereas when social capital was high (+1 SD), the effect of financial strain on stress was smaller by about one-half (B = .09, SE = .03, t = 2.89, p < .001). Similarly, when social capital was low (−1 SD), the effect of financial strain on depressive symptoms was B = .11 (SE = .02, t = 4.90, p < .001), whereas when social capital was high (+1 SD), the effect of financial strain on depressive symptoms was negligible (B = .03, SE = .02, t = 1.09, p = .28). Although social capital did not significantly moderate the effect of financial strain on health and anxiety, in both cases a significant main effect of social capital was observed such that those reporting higher social capital also reported lower anxiety (B = −.35, SE = .06, t = −5.90, p < .001) and better health (B = −.63, SE = .10, t = −6.63, p < .001). The proportional reduction in variance estimates provided in Table 2 suggest that very little variance is accounted for at Level 1, and almost all variance is accounted for at Level 3. These values should not be taken to mean that financial strain (a Level
Stress
Depressive symptoms
Anxiety symptoms
Physical illness
B (SE) t
B (SE) t
B (SE) t
B (SE) t
0.68 (.02) .10 (.05)2.02* −.55 (.07) −7.54*** .001 (.002) .51 .11 (.05) 2.29* −.06 (.01) −4.32*** .07 (.02) 4.31*** .002 (.03) .06 −.12 (.05) −2.54* .001 (.001) .50 −.001 (.03) −.03
1.66(.02) .07 (.04) 1.82† −.35 (.06) −5.90*** .001 (.001) .71 .10 (.04) 2.54** −.06 (.01) −6.01*** .05 (.01) 3.84*** .01 (.03) .44 −.05 (.04) −1.51 −.000 (.001) −.39 −.02 (.03) −.72
2.36 (.03) .07 (.06) 1.13 −.63 (.10) −6.63*** .01 (.002) 3.33*** .11 (.06) 1.84† −.03 (.01) −2.45* .06 (.02) 3.43*** −.01 (.03) −.33 −.02 (.05) −.49 .002 (.001) 1.30 −.01 (.03) −.17
Constantt 1.72 (.03) Communityj .08 (.05) 1.61 Social capitalj −.43 (.08) −5.49*** Agei .004 (.002) −1.92† Gender (female)i .12 (.05) 2.28* Assessment cyclet −.01 (.02) −.68 Financial straint .13 (.02) 6.16*** Straint × communityj .01 (.04) .19 Straint × social capitalj −.12 (.06) −2.21* Straint × agei .002 (.001) 1.20 Straint × gender (female)i −.02 (.04) −.42 Variance (intercept only) Level 1 (within subject) .14 Level 2 (between subject) .19 Level 3 (between family) .06 Total variance .38 Pseudo R-squared (proportional reduction in variance)a Level 1 (within subject) .02 Level 2 (between subject) .28 Level 3 (between families) .95
.09 .17 .06 .32
.06 .13
Anxiety, Stress, & Coping: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gasc20
Financial strain, social capital, and perceived health during economic recession: a longitudinal survey in rural Canada a
a
Christine Frank , Christopher G. Davis & Frank J. Elgar
b
a
Department of Psychology, Carleton University, Ottawa, ON, Canada b
Click for updates
Institute for Health and Social Policy and Douglas Institute, McGill University, Montreal, QC, Canada Accepted author version posted online: 20 Nov 2013.Published online: 10 Dec 2013.
To cite this article: Christine Frank, Christopher G. Davis & Frank J. Elgar (2014) Financial strain, social capital, and perceived health during economic recession: a longitudinal survey in rural Canada, Anxiety, Stress, & Coping: An International Journal, 27:4, 422-438, DOI: 10.1080/10615806.2013.864389 To link to this article: http://dx.doi.org/10.1080/10615806.2013.864389
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Downloaded by [University of Calgary] at 13:35 19 March 2015
Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions
Anxiety, Stress & Coping, 2014 Vol. 27, No. 4, 422–438, http://dx.doi.org/10.1080/10615806.2013.864389
Financial strain, social capital, and perceived health during economic recession: a longitudinal survey in rural Canada Christine Franka*, Christopher G. Davisa and Frank J. Elgarb a
Department of Psychology, Carleton University, Ottawa, ON, Canada; bInstitute for Health and Social Policy and Douglas Institute, McGill University, Montreal, QC, Canada
Downloaded by [University of Calgary] at 13:35 19 March 2015
(Received 27 May 2013; accepted 5 November 2013) Although the health consequences of financial strain are well documented, less is understood about the health-protective role of social capital. Social capital refers to a sense of community embeddedness, which is in part reflected by group membership, civic participation, and perceptions of trust, cohesion, and engagement. We investigated whether perceptions of social capital moderate the relation between financial strain and health, both mental and physical. This longitudinal study surveyed adults in two communities in rural Ontario where significant job losses recently occurred. Data were collected on financial strain, social capital, perceived stress, symptoms of anxiety and depression, and physical health on three occasions over 18 months (N’s = 355, 317, and 300). As expected, financial strain positively related to perceived stress, poor physical health and symptoms of anxiety and depression, whereas social capital related to less stress, better physical health, and fewer symptoms of anxiety and depression. Effects of financial strain on perceived stress and depressive symptoms were moderated by social capital such that financial strain related more closely to perceived stress and depressive symptoms when social capital was lower. The findings underscore the health-protective role of community associations among adults during difficult economic times. Keywords: social capital; financial strain; health; stress; community psychology
Current economic trends of job losses and rising unemployment rates have had an impact on health in many industrialized countries (Bernard, 2009; Pilat, Cimper, Olsen, & Webb, 2006). In Spain, physicians have treated more mental health and alcohol problems since the recession hit, particularly in patients who were unemployed or had difficulty paying their mortgages (Gili, Roca, Basu, McKee, & Stuckler, 2013). Self-rated mental health has deteriorated in Britain (Katikireddi, Niedzwiedz, & Popham, 2012). In Italy, suicides and suicide attempts due to economic stressors have increased (De Vogli, Marmot, & Stuckler, 2013). Layoffs and unemployment during times of economic decline are sources of considerable stress for individuals and families, and the resulting financial or economic strain has negative consequences for mental and physical health (Dekker & Schaufeli, 1995; Fox & Chancey, 1998; Kinnunen & Pulkkinen, 1998; Sverke, Hellgren, & Näswall, 2002). However, in some cases, the relation between financial strain and health (mental and physical) is not particularly strong, which suggests that some individuals, families, or communities possess qualities or resources that protect them from the *Corresponding author. Email: [email protected] © 2013 Taylor & Francis
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
423
negative effects of such strain. This study explored the protective role of social capital in two rural communities in Ontario, Canada that were dealing with a large number of job losses and heightened economic uncertainty from 2008 to 2010. A large body of research has documented the health consequences of involuntary job loss. Studies have found that involuntary job loss followed by sustained periods of unemployment is associated with symptoms of anxiety and depression, alcohol and drug use, and physical health problems (Iversen & Sabrow, 1988; Keefe et al., 2002; Pearlin, Menaghan, Lieberman, & Mullan, 1981; Price, Choi, & Vinokur, 2002; Vinokur, Price, & Caplan, 1996). For example, a study by Studnicka et al. (1991) of adults who were laid off following a factory closure found that those who were still unemployed 12 months later were eight times more likely to report poor psychological health compared to those who had regained employment. In the USA during the 1980s, a series of studies led by Rand Conger documented social and psychological consequences of an economic downturn in farming communities in the Midwest as falling grain prices and high interest rates forced many farmers into bankruptcy and foreclosure (Conger et al., 1990, 1992, 1993; Conger, Reuter, & Elder, 1999). Their longitudinal study of these families found that economic pressure led to increased depression and demoralization in parents, which in turn contributed to marital conflict, parenting difficulties, and adjustment problems in their adolescent children (Conger et al., 1992, 1993). These studies highlight the pervasive and devastating effects of job loss and financial strain on health, well-being, and social relationships. It could be argued that the negative health effects of job loss should be somewhat mitigated in countries with comprehensive welfare programs, particularly in countries where unemployment does not impose barriers to accessing health care or social services. A study by the Organization for Economic Cooperation and Development (OECD) found that the health impact of economic cycles related directly to international differences in public service expenditures (Gerdtham & Ruhm, 2006). Specifically, the study found smaller health impacts during periods of economic growth and recession in countries that shared stronger labor protections and larger social investments. Nevertheless, within these countries with extensive welfare systems, the evidence suggests that job loss and financial uncertainty continue to negatively affect health and well-being. For instance, a longitudinal study in Sweden by Voss, Nylén, Floderus, Diderichsen, and Terry (2004) found that unemployment significantly increased the risk of disease at 10- and 24-year follow-ups. In a national longitudinal study of adults in Germany, Lucas, Clark, Georgellas, and Diener (2004) found that job loss was associated with a multi-year decline in life satisfaction that persisted even after reemployment. These results speak to the long-lasting effects job loss can have on individual health and well-being, even in countries where job loss does not limit access to health care and social services. It is important to note that most of the research assessing the psychological and physical health effects of job loss and economic downturn has emphasized the role of perceived economic or financial strain as the mediating construct between job loss and well-being (e.g., Broman, Hamilton, & Hoffman, 1990; Conger et al., 1990; Pearlin et al., 1981). Financial strain, defined as a feeling of unease about one’s perceived ability to “make ends meet,” is conceptualized as a psychological appraisal that one’s demands (e.g., bills, financial obligations) exceed available resources (Lazarus & Folkman, 1984). The research cited above suggests that the health effects of lost income or unemployment are at least partly mediated by the perception of financial strain.
Downloaded by [University of Calgary] at 13:35 19 March 2015
424
C. Frank et al.
Although job loss, economic hardship, and financial strain increase the risk of physical and mental health problems, some individuals are resilient in the face of these challenges. Perceived social capital is one factor that appears to buffer individuals from the negative health effects of financial strain. Social capital refers to the value accrued from belonging to groups, whether they are families, friendship networks, faith, work, or recreational groups (e.g., Coleman, 1988, 1990; Putnam, 1995). Not only do these groups contribute to a sense of identity, safety, trust, and belonging, they also help create norms of reciprocity and bonds of support that can be called upon in times of need. Social capital stems from the social connections and networks people establish in everyday interactions, which in turn, both forms and reinforces the individual’s perception of trust, safety, cohesion, engagement, and reciprocity within his or her community (Ahnquist, Wamala, & Lindstrom, 2012; Giordano & Lindström, 2010; Lochner, Kawachi, & Kennedy, 1999; Onyx & Bullen, 2000). This perception is reflected in terms of membership in community organizations (e.g., bowling leagues, political parties, religious groups), volunteerism, and interpersonal and intergroup trust (e.g., Putnam, 1993, 1995). Social capital is conceptually related to social support, but is distinct insofar as it refers to aspects of social contexts and not individuals per se. Whereas perceived social support refers to the sense that help is available if needed (usually from close others), social capital refers to the extent of integration in broad social networks. Social capital may also be conceptualized and operationalized at the community or state level (e.g., Kawachi, Kennedy, & Glass, 1999; Putnam, 1993, 1995). From this more sociological perspective, social capital may be reflected in terms of the scope and depth of community organizations and degree of civic engagement and intergroup conflict (e.g., Kawachi & Berkman, 2000). No doubt, the psychological and sociological perspectives on social capital are closely related. Estimates of community-level social capital often are based, at least in part, on aggregated perceptions from groups of individuals (e.g., Kawachi, 1999; Kawachi et al., 1999). However, because social capital is sometimes assessed only at an ecological level (e.g., proportion of the population that votes; Putnam, 1995) and other times only at an individual level (e.g., perceptions of trust; Kennelly, O’Shea, & Garvey, 2003), prior research has not always separated the compositional effects of individual perceptions of social capital from the contextual effects of social capital (Elgar et al., 2011). Therefore, we adopt a multilevel approach to investigating individuals’ perspectives on social capital in their community, while taking into account community-level differences. Like financial capital, perceived social capital can be regarded as a resource or currency that can be leveraged in times of need or stress. Individuals who perceive higher levels of social capital likely have wider social networks from which aid and support are available, and these networks might be more willing to provide assistance because it is good for the community and not just for the individual in need. Those who have higher levels of perceived social capital may also feel less embarrassed to receive support to the extent that they see their role in the community to be more than merely a tax-payer and service-user. Population studies have found that high levels of perceived social capital confer advantages in terms of individual health and well-being (Elgar et al., 2011; Subramanian, Kawachi, and Kennedy, 2001). Kawachi et al. (1999) found that even when controlling for other risk factors (e.g., obesity, smoking, lack of access to health care) individuals who lived in US states with higher levels of social capital reported better
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
425
health. Kawachi and Berkman (2000) argued that the availability of social capital benefits health because it promotes health-related behaviors, facilitates access to services and amenities, and provides psychosocial processes that support health (e.g., perceptions of safety, social support). These influences on health may be especially beneficial in economically deprived areas that lack material resources (Cramm, Møller, & Nieboer, 2012; De Silva, Huttly, Harpham, & Kenward, 2007; Gore, 1978; Sapag et al., 2008). Whereas research has established links between perceived social capital (measured at the individual level) and both mental and physical health (e.g., Giordano, Björk, & Lindström, 2012; Giordano & Lindström, 2010; Verhaeghe & Tampubolon, 2012), we are not aware of any studies that conceptualized social capital as a moderator of the financial strain – health/well-being link. Some research has found that perceived social support moderates the effect of financial strain on health and well-being. For instance, Schwarzer, Hahn, and Jerusalem (1993) found in a study of former East Germans who migrated to West Germany following the collapse of the Berlin Wall that social support buffered the effect of prolonged unemployment on health. Likewise, in their study of farming communities in the US Midwest during the economic recession of the 1980s, Conger et al. (1999) found that for both men and women, spousal support buffered the effects of perceived economic strain on emotional distress. Perceived social capital is a broader concept than perceived social support insofar as it emerges from one’s embeddedness within a community as opposed to one’s embeddedness within a social relationship. People with high levels of social capital are well integrated into their community, they are civically minded and active participants in their community. And so when they need help, their community is there for them. Social capital, as a measure of the extent of integration within social networks, is a protective resource that can be leveraged to maintain health and well-being once coping resources have been depleted. Protective resources are relationships, objects, or personal characteristics that are either valued directly (e.g., close attachments), or indirectly allow one to obtain other valued means (e.g., social support; Diener & Fujita, 1995; Hobfoll, 2002). As such, social capital might be particularly important when financial resources are depleting, and therefore might buffer the negative effects of financial strain on health and well-being. Although social capital may be regarded as a personal resource, it is also likely to be a resource that is shared within families. Couples share many of the same values and worldviews (Antonovsky & Sourani, 1988; Ransom, Fisher, & Terry, 1992; Roest, Dubas, Gerris, & Engels, 2009), and tend to interpret adversity through the same lens (Davis, Harasymchuk, & Wohl, 2012). Family systems theorists have long argued that coping and resilience are family-based processes (Patterson, 2005; Rosenblatt, 2000; Walsh, 1996). Indeed, Coleman’s (1988) original writings on the concept emphasized the important role that family connections play in children’s health and education. Therefore, social capital, with its emphasis on social embeddedness or integration, can also be considered a family-based resource. In this report, we assess the role that family-level social capital plays as individuals cope with financial uncertainty brought on by the economic recession of 2008. Our focus is on two towns in Ontario, Canada that suffered significant job losses due to the recession. Previous research shows that social capital is correlated positively with health and wellness (e.g., Elgar et al., 2011; Giordano et al., 2012; Kawachi et al., 1999). Here we assess the extent to which social capital buffers people from the negative effects of financial strain.
426
C. Frank et al.
Downloaded by [University of Calgary] at 13:35 19 March 2015
Method Participants Adults were recruited by telephone using a random dialing procedure from two predominantly English-speaking communities in rural Ontario. The communities had similar demographic and economic characteristics, and populations were comparable (between 7800 and 8800) at the time of data collection (Statistics Canada, 2007). Our plan was to compare one town facing significant job losses to the other. However, in the early stages of data collection, the economic recession also hit the second town. Therefore, the decision was made to combine the samples. A sample of 355 adults from 240 families was recruited initially and 89.3% (n = 317) of the original sample was followed up at nine months and 84.5% (n = 300) of the original sample was followed up again at 18 months. Only individuals who provided complete data for at least two waves of testing were included in the analyses. The gender and age composition of two community samples did not differ significantly over the course of the study. The sample was 62.5% female and had an average age of 49.5 years at Time 1 (SD = 14.3 years). In Town A, over the course of the study, 58.6% of participants were employed, 18.2% were unemployed, and 23.2% were retired. In Town B, 66.9% of participants were employed, 13.4% were unemployed, and 19.7% were retired. Families were reimbursed with a $50 gift card redeemable at a local store for each assessment. A university research ethics board approved the study procedures.
Measures Financial strain was assessed with two items adapted from Pearlin et al.’s (1981) seminal work on job disruption, coping, and stress. Respondents were asked the extent to which they have had difficulty paying bills in the past 12 months (with response options ranging from 1 = “a great deal of difficulty” to 5 = “no difficulty at all”) and the extent to which they have enough money left at the end of the month (with response options ranging from 1 = “more than enough money left over” to 5 = “not enough to make ends meet”). After reverse scoring the first item, the two items were averaged so that higher scores indicated higher financial strain. The correlation between the two items was strong (r’s = .66–.72; Spearman–Brown = .80–.83), and test-retest correlations were quite high (rtime1, time2 = .68; rtime2, time3 = .65). Financial strain was moderately and negatively correlated with perceived SES using a modified version of Cantril’s (1965) ladder where participants were asked to rank themselves based on the amount of money they have, their level of education, and how respected their job is relative to other people in Canada (rtime1 = −.47, rtime2 = −.42, rtime3 = −.42, p’s < .001). Financial strain was also moderately and negatively correlated with family income (assessed as self-reported household income; rtime1 = −.44, rtime2 = −.31, rtime3 = −.31, p’s < .001) providing evidence for convergent validity. Social capital was measured with Onyx and Bullen’s (2000) Social Capital Scale which assessed eight aspects of the construct: participation in the local community, social agency, feelings of trust and safety, neighborhood connections, work connections, family and friends connection, tolerance of diversity, and value of life. We excluded “work connection” items from the scale because many of the individuals in the study were not employed and thus unable to answer questions about work connections. Rather than parsing social capital into its components or manifestations, research on social capital has tended to focus on the concept in its whole (e.g., Sarracino, 2010). That is, although
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
427
participation in the local community, feelings of safety and trust, tolerance for diversity, and family and friendship connections each may be seen as conceptually distinct, social capital is the common thread. Past research confirms that although the subscales capture distinct aspects, they cohere in a common second-order factor (Bullen & Onyx, 2007; O’Brien, Burdsal, & Molgaard, 2004). Therefore, we use total score for social capital in our analyses. Responses to the 30 items were on a 4-point Likert scale (from 1 = “no, not much” to 4 = “yes, very much”). Internal consistency of the scale was high (α = .82–.84), as were its test-retest correlations (rtime1, time2 = .79; rtime2, time3 = .75). The extent to which families shared a common sense of social capital was assessed by means of an intercepts-only hierarchical linear model (HLM). The data are structured hierarchically such that assessment cycle (times 1, 2, and 3) is nested within individuals, and individuals are nested within families. The intercepts-only model apportions the total variance into within-person variance (i.e., within-person fluctuations), between-person variance (reflecting individual differences), and between-family variance (i.e., differences between families). Confirming our expectation that social capital is a shared family resource, the intercepts-only HLM indicated that 47.9% of the total variance in social capital was located at the third (family) level, with the remainder divided approximately equally between the first two levels. Therefore, assessments of social capital were relatively stable across time and across family members, but differed considerably from one family to the next. Perceived stress was assessed with the Perceived Stress Scale (Cohen, Kamarck, & Mermelstein, 1983), which is a widely used general measure of appraised stress during the previous month. It contained 14 items – 7 positive (e.g., “In the last month, how often have you felt that things were going your way?”) and 7 negative (e.g., “In the last month, how often have you felt that you were unable to control the important things in your life?”), each rated using a scale ranging from 1 (“never”) to 4 (“very often”). The seven positively valenced items were reversed scored and a mean score on the 14 items was calculated with higher scores indicating more stress. Consistent with past research (e.g., Cohen et al., 1983), the instrument was internally consistent at all three times (α = .84–.91) and test-retest correlations were strong (rtime1, time2 = .69; rtime2, time3 = .60). General anxiety was assessed with the Zung Anxiety Scale (Zung, 1971), which consisted of 20 items measuring symptoms experienced over the past week. Participants rated each item on a 4-point scale ranging from 1 (“none or a little of the time”) to 4 (“most or all of the time”) to indicate how much the statement describes how they felt during the past week. A mean score of the 20 items was calculated with higher scores indicating more anxiety. The scale was internally consistent (α = .83–.88) and anxiety scores were consistent over time (rtime1, time2= .71; rtime2, time3= .68). Symptoms of depression were assessed with the Center for Epidemiologic Studies – Depression Scale (CES-D; Radloff, 1977), which consisted of 20 items that describe depressive feelings and behaviors experienced over the past week. Participants rated each item on a 4-point scale ranging from 0 (“rarely or none of the time”) to 3 (“all the time”). Scores were averaged so that higher scores indicated more symptoms. The scale was also internally consistent (α = .93–.94) and yielded high test-retest correlations (rtime1, time2= .75; rtime2, time3= .66). Physical health was measured using the Balanced Physical Health Scale (BPHS; Godin & Mantler, 2007). The BPHS is a 20-item instrument assessing general health in terms of mobility issues (e.g., “During the past three weeks, have you had any difficulty
428
C. Frank et al.
walking?”), digestive issues (e.g., “During the past three weeks, have you been bothered by indigestion/heartburn?”), general fitness (e.g., “If I have to exert myself [e.g., run for the bus or take the stairs] I find I have difficulty catching my breath”), resistance (e.g., “I seldom get sick”), and energy (e.g., “I have the energy to do things I enjoy after I’m done work for the day”). Some items were reverse-scored and an average was calculated with higher scores indicating worse physical health. Cronbach’s alpha for the 20-item scale ranged from .87 to .91. In the current study, the BPHS highly correlated with the general health question, “In general, would you say your health is excellent, very good, good, fair, or poor?” (Ware, Kosinski, & Keller, 1996) rtime1 = −.69, rtime2 = −.70, rtime3 = −.63.
Downloaded by [University of Calgary] at 13:35 19 March 2015
Procedure Randomly selected families from the two communities were invited to participate in the community participation and health outcomes study. Our invitation to participate was extended to all adults living in the household. Self-completion surveys were delivered by a research assistant to those who agreed to participate. Prior to the first assessment, home visits were arranged for the purposes of obtaining informed consent. Completed surveys were returned by mail.1 Three assessments were carried out between Fall 2008 and 2010 with 9-month intervals. The data were inspected for normality and outliers. At Time 1, two outliers were identified: one on anxiety and the other on depression. At Time 2, four outliers were identified: one on anxiety and three on depression. At Time 3, one participant was identified as an outlier on stress. In all cases, participant’s scores were adjusted to within three standard deviations of the sample mean.
Analysis overview As noted above with respect to social capital, the data are analyzed using a three-level HLM (Raudenbush & Bryk, 2002) where the first level was the assessment cycle (Times 1, 2, and 3) and reflects within-person variability, the second level was individual differences, and the third level refers to family-level variance. We first assessed how much variance in the data was attributable to within-person differences (Level 1), the interpersonal differences (Level 2), and between families differences (Level 3) by means of an intercepts-only HLM. To the extent that variance is negligible at one level for a given variable, it suggests that at that level the variable is constant or invariant. Four three-level models were fitted to the data to determine the extent to which financial strain at timet for personi (of familyj) related to each of the four dependent variables (depression, anxiety, perceived stress, and physical health) at timet, taking into account the time of assessment. Therefore, the Level 1 (within person) model can be represented as: ytij ¼ B0ij þ B1ij Financial Straintij þ B1ij Timetij þ etij where ytij represents the dependent variable at timet for personi (of familyj), B0ij represents the intercept for personi (in familyj), B1ij represents the slope for (grand mean centered) financial strain on the dependent variable for personi (in familyj), and etij is the
Downloaded by [University of Calgary] at 13:35 19 March 2015
Anxiety, Stress & Coping
429
random error for personi (in familyj). At Level 2, we modeled the effect of gender and age (grand mean centered) on the Level 1 intercept and the Level 1 slope for financial strain. This assesses whether the person-level intercepts and effects of financial strain vary as a function of gender or age of the participant. At Level 3 (between families), we assessed the extent to which the Level 1 intercepts and slopes for financial strain vary as a function of the community within which one resides (Town A vs. Town B) and one’s family’s perceived level of social capital (averaged over time and over members of the family). To the extent that community or social capital is associated with one’s Level 1 intercept, it suggests that community or social capital has a main effect on the dependent variable. The extent to which community or social capital relates to one’s Level 1 slope for financial strain indicates the degree to which community or social capital moderates the effect of financial strain on the outcome. These moderated effects, where significant, were followed up with simple slope analyses. To give a sense of the magnitude of effects, we provide in Table 2 the proportional reduction in residual variances (PRV) in the dependent variables from the intercept-only model (with no predictors) to the final model. The PRV may be interpreted as a pseudo-coefficient of determination (R2; see Peugh, 2010; Raudenbush & Bryk, 2002). However, since variance has been apportioned to each level, a separate PRV is calculated for each of the three levels. These estimated proportions indicate the proportion of variance at the particular level explained by predictors at that and lower levels.
Results Attrition We found no significant differences between individuals who dropped out of the study at Time 2 or Time 3 and those who remained in age, gender, employment status, anxiety, perceived stress, depressive symptoms, physical health, financial strain, or social capital (t’s < 1.8, p’s > .10).
Basic models An intercepts-only multilevel model of financial strain showed substantial betweenfamily variance (44.5%) but almost all the remaining variance (50.0%) was within-person variance, suggesting fluctuations over time within individuals. The small percentage of between-person variance (5.5%) suggests that families are internally consistent in their perceptions of financial strain. Financial strain did not differ between the two communities (B = .15, SE = .10, t = 1.49, p = .14), and negatively related to social capital (B = −.51, SE = .15, t = –3.46, p = .001). Financial strain was higher among females than males (B = .22, SE = .06, t = 3.41, p = .001), and among younger participants (B = −.02, SE < .01, t = −5.69, p < .001), and increased over the course of the study (B = .35, SE = .03, t = 12.65, p < .001).
Predictive models Based on these findings, we constructed a series of multilevel models where each outcome (depressive symptoms, anxiety symptoms, perceived stress, and physical health) for person i at time t was explained by financial strain at time t, time, and a random
430
C. Frank et al.
Table 1. Descriptive statistics on variables. Baseline (n = 315)
Downloaded by [University of Calgary] at 13:35 19 March 2015
Financial strain Social capital Perceived stress Anxiety symptoms Depressive symptoms Physical health Subjective SES
9 months (n = 312)
18 months (n = 285)
M
SD
α
M
SD
A
M
SD
α
2.37 2.67 1.72 1.73 0.74 3.62 5.55
0.81 0.40 0.65 0.42 0.58 0.66 1.91
.79 .83 .90 .83 .93 .87 –
2.99 2.76 1.64 1.64 0.66 3.66 5.60
1.07 0.40 0.45 0.45 0.55 0.67 1.62
.81 .82 .91 .88 .93 .89 –
2.99 2.71 1.89 1.63 0.67 3.66 5.49
1.05 0.41 0.48 0.43 0.54 0.63 1.57
.83 .84 .84 .87 .93 .87 –
intercept. At level 2 (between persons), we tested whether the level 1 random intercept and the effect of financial strain varied by gender and age. At level 3 (between families), we tested whether the level 2 intercept and effect for financial strain varied by average family social capital and community. Age and gender were excluded from these models because these variables did not relate to social capital. Descriptive statistics on all the variables are presented in Table 1. As shown in the regression analyses presented in Table 2, differences in mental and physical health between the two communities represented in the sample ranged from nonsignificant to modest once differences in other variables were held constant. Older respondents tended to report somewhat less stress (p < .10) but more physical health problems. Age did not significantly predict symptoms of depression or anxiety. Women reported more stress, depressive symptoms, and anxiety symptoms than men, and marginally more physical health problems. As well, assessment cycle related negatively to depressive, anxiety, and physical symptoms (Table 1). Most relevant to our study hypothesis, financial strain related positively to all four outcomes indicating that those with greater financial strain perceived more stress, had more symptoms of depression, anxiety, and ill-health. However, the main effects of financial strain on stress and depressive symptoms were moderated by social capital. In both cases, simple slopes analyses indicated that the effects of financial strain on stress and depressive symptoms were stronger when social capital was lower (see Figure 1). When social capital was low (−1 SD), the effect of financial strain on stress was B = .17 (SE = .03, t = 6.19, p < .001) whereas when social capital was high (+1 SD), the effect of financial strain on stress was smaller by about one-half (B = .09, SE = .03, t = 2.89, p < .001). Similarly, when social capital was low (−1 SD), the effect of financial strain on depressive symptoms was B = .11 (SE = .02, t = 4.90, p < .001), whereas when social capital was high (+1 SD), the effect of financial strain on depressive symptoms was negligible (B = .03, SE = .02, t = 1.09, p = .28). Although social capital did not significantly moderate the effect of financial strain on health and anxiety, in both cases a significant main effect of social capital was observed such that those reporting higher social capital also reported lower anxiety (B = −.35, SE = .06, t = −5.90, p < .001) and better health (B = −.63, SE = .10, t = −6.63, p < .001). The proportional reduction in variance estimates provided in Table 2 suggest that very little variance is accounted for at Level 1, and almost all variance is accounted for at Level 3. These values should not be taken to mean that financial strain (a Level
Stress
Depressive symptoms
Anxiety symptoms
Physical illness
B (SE) t
B (SE) t
B (SE) t
B (SE) t
0.68 (.02) .10 (.05)2.02* −.55 (.07) −7.54*** .001 (.002) .51 .11 (.05) 2.29* −.06 (.01) −4.32*** .07 (.02) 4.31*** .002 (.03) .06 −.12 (.05) −2.54* .001 (.001) .50 −.001 (.03) −.03
1.66(.02) .07 (.04) 1.82† −.35 (.06) −5.90*** .001 (.001) .71 .10 (.04) 2.54** −.06 (.01) −6.01*** .05 (.01) 3.84*** .01 (.03) .44 −.05 (.04) −1.51 −.000 (.001) −.39 −.02 (.03) −.72
2.36 (.03) .07 (.06) 1.13 −.63 (.10) −6.63*** .01 (.002) 3.33*** .11 (.06) 1.84† −.03 (.01) −2.45* .06 (.02) 3.43*** −.01 (.03) −.33 −.02 (.05) −.49 .002 (.001) 1.30 −.01 (.03) −.17
Constantt 1.72 (.03) Communityj .08 (.05) 1.61 Social capitalj −.43 (.08) −5.49*** Agei .004 (.002) −1.92† Gender (female)i .12 (.05) 2.28* Assessment cyclet −.01 (.02) −.68 Financial straint .13 (.02) 6.16*** Straint × communityj .01 (.04) .19 Straint × social capitalj −.12 (.06) −2.21* Straint × agei .002 (.001) 1.20 Straint × gender (female)i −.02 (.04) −.42 Variance (intercept only) Level 1 (within subject) .14 Level 2 (between subject) .19 Level 3 (between family) .06 Total variance .38 Pseudo R-squared (proportional reduction in variance)a Level 1 (within subject) .02 Level 2 (between subject) .28 Level 3 (between families) .95
.09 .17 .06 .32
.06 .13
