Julian LaFarge

Labor Force and Employment

06/07/2012

Employment, Social Capital, and Health:

An Empirical Analysis

The United States labor force has been undergoing massive changes in over the past 30 or so years. More people are working as the population continues to rise, but even more so people with more diverse backgrounds and demographics are entering the work force or clamoring for jobs, as how the economy is currently structured seems to predict that individuals who would like to enter the labor force won’t be able to. Sociologists White and Cunneen contend that “structural unemployment and underemployment, privatization of state services and withdrawal of income support” (White and Cunneen) inherently exist in neo-liberal ideologies, making this disparity between the employed and the unemployed seemingly unavoidable. With this increase in size and diversity of the labor force and the accompanying restrictions to entry, the effects of employment merit analysis. One could possibly make the argument that of employment's causal relationships, perhaps none are more important than employment's affect on one's health. Indeed, “occupational conditions have consequences for physical health outcomes” (Wickrama Lorenz 363). Further, “A central theoretical proposition of sociology states that social structural positions have dramatic effects on life chances” (Link Lennon 1351). However, one is left wondering how is this possible, that is, how exactly is it that whether or not one is employed is able to influence one's health? One possible mechanism through which employment influences health is the relatively new sociological concept of social capital. Robert Putnam writes, “Mounting evidence suggests that people whose lives are rich in social capital cope better with traumas and fight illness more effectively” (289). His 2000 book Bowling Alone is the most cited work amongst social capital literature, according to citation analysis by Moore, Haines, Hawe, and Shiell speaking to social capital's relatively late emergence as a concept meriting analysis on the sociological scene (731). Furthermore, social capital's influence on health has been explored in various other research projects, such as Sara Ferlander's "The Importance of Different Forms of Social Capital for Health” where she concludes “high levels of trust in society can facilitate faster and wider diffusion of information which may in turn promote healthier behaviours and control uhealthy ones” (123). While employment's effect on health and social capital's effect on health have both been documented, an effort to link the three in a dependent → mediator → independent causal relationship. Noting this lack in the literature, this study uses data from the 2010 General Social Survey to examine the relationship between social capital and health and the mechanisms through which employment and social capital influence health, hypothesizing that employment affects health through the creation and promotion of social capital in the workplace.

Quantitative Employment and Health Analysis In order to assess the effects of employment status on health, data was taken from the 2010 General Social Survey (GSS), in order to provide the most contemporary, and thusly accurate, picture of the relationship between employment and health. The independent variable, our predictor, is employment status. The 2010 GSS contains a non-dichotomous “Labor Force” variable which classifies status as, “Part-Time”, “Full-Time”, “Temporarily Not Working”, “Unemployed”, “Retired”, “Student”, and “Homemaker”. The variable was then recoded into a dichotomous, discreet variables, with the part-timers, full-timers, and those who were temporarily not working classified as “employed” and everyone else falling into the “unemployed” category. The United Stated Bureau of Labor includes those who are temporarily not working, but still employed, as members of the labor force, thusly justifying the inclusion of those respondents who classified themselves as temporarily not workinguslaborforce. The dependent variable in this analysis is health, whose measure is taken from the General Social Survey question asking respondents to self-report their health, with results coded as, “Poor”, “Fair”, “Good”, or “Excellent”. Additionally, this analysis is concerned with developing a possible mechanism through which employment status affects health, hypothesizing employment’s influence on health to be a function of its creation of social capital; an employee has coworkers whom they interact with, network with, and form relationships with. Over the last decade social capital literature has increased significantly, due in part to Robert Putnam, whose work entitled Democracy in America is the most cited work amongst social capital literature, according to citation analysis by Moore, Haines, Hawe, and Shiell (731). Social capital’s meteoric rise to sociological prominence has brought with it many questions regarding its conceptualization and measurement, as the term itself was first used in 1917 by Lyda Hanifan only cursorily and afterwards only sporadically and secondarily until its reemergence following Putnam (Farr 7-8). Therefore to analyze it's effects objectively, this analysis must offer a framework through which it can quantify social capital. In Bowling Alone, Putnam defines social capital as “features of social organization such as networks, norms, and social trust that facilitate coordination and cooperation for mutual benefit” (67). Putnam identifies the main tenets of social capitals as networks, norms, and trust. Implicit in this is that networks, norms, and trust are vital in the creation of maintenance of the collective asset that is social capital. When combining these three aspects in a sort of logical narrative on social capital, it would appear that social capital is “complexly conceptualized as the network of associations, activities, or relations that bind people together as a community via certain norms and psychological capacities, notably trust, which are essential for civil society and productive of future collective action or goods, in the manner of other forms of capital.” (Adler 9). Putnam’s theory also falls in line with what Adler and Kwon refer to as the “core intuition” guiding all social capital research, that “the goodwill that others have toward us is a valuable resource” (18). Thusly, the conceptualization offered by this analysis is one where social capital produces tangible, measurable benefits (whether it be an increase in productivity or as a predictor of good health), i.e. that social ties matter, granting considerable weight to associational membership, levels of trust, and norms of reciprocity as both key indicators and producers of social capital. Following this study’s theoretical framework, measurements of social capital reflect its central tenets of social trust, associational membership, and norms of reciprocity. Level of societal trust was measured by the responses to a General Social Survey item that asked, “Can people be trusted?”. This variable is self-evident in the way that it measures and individual’s level of trust, with responses being coded as either “They cannot be trusted”, “They can be trusted”, or “It depends”. To calculate the respondent’s attitude towards norms of reciprocity, responses from the General Social Survey question, “Are people fair or do the try to take advantage?”, with responses coded as “They take advantage”, “They are fair”, or “It depends”. Whether or not a respondent believes his fellow man to be fair or opportunistic is a measure of reciprocity in that equality and fairness, assessed by the question, are reciprocal in nature; if a respondent feels like his fellow man is out to take advantage of him, not only is it likely that trust will decline, but also this perceived lack of fairness is equal to an absence of high levels of norms, as high levels of reciprocal norms equate to a sense of equality. Thus, expectations exist that all laws, actions, reactions, etc. will be met equally, i.e. there exists a high level of fairness. As Kawachi et al state in their study on income inequality, social capital, and mortality rates, “Belief in the goodwill and benign intent of others facilitates collective action and mutual cooperation and therefore adds to the stock of a community’s social capital. Collective action, in turn, further reinforces community norms of reciprocity” (1492) Additionally, level of associational membership was measured in a more holistic manner. In creating an aggregate variable of civic engagement i.e. associational membership, participation in religious services, union membership, membership in environmental groups, and voter turnout rates from the 2004 and 2008 presidential elections were combined to create a scale civic participation variable, with values ranging from 0 to 5, a result of recoding the 5 individual civic participation variables into dichotomous, discrete variables. While union and environmental group membership, along with participation in religious services, are logical forms of associational membership, voter turnout rates appears to be the “odd variable out”, in that it is not an indicator of associational membership. Of the declining rates in voter turnout, Putnam writes, “Declining electoral participation is merely the most visible symptom of a broader disengagement from community life” (35). Participation or lack thereof, in the electoral process is reflective of social capital. The three components of social capital were then combined into an aggregate, scale variable dubbed “Social Capital”. In addition to the variables already mentioned, years of education and age were found to be highly correlated with both labor force status and health among the traditional control variables. Any statistical models used in this paper will control for these variables, also gathered from the 2010 GSS In hypothesizing that the social capital is the mechanism through which employment affects health, employment status acts as an independent variable affecting both the dependent variable of health and the mediating variable of social capital, through which labor force participation is able to influence health outcomes. Thusly we must not only test for association, but also for potential mediation effects. In their 1986 article “The Moderator-Mediator Variable Distinction”, Barron and Kenny establish three fundamental criteria necessary for mediation to be occurring; (1) the independent variable must predict the dependent, (2) the independent must predict the mediator, and (3) the mediator must predict the dependent (1176). Thusly, this statistical analysis begins by establishing these three premises before all else. Firstly then, the relationship between employment and health was tested using multiple linear regression and controlling for years of education and age. Results are listed in table 1. (Below)

|Coefficientsa |
|Model |Unstandardized Coefficients |Standardized |t |Sig. |
| | |Coefficients | | |
| |B |Std. Error |Beta | | |
|1 |(Constant) |1.316 |.135 | |9.747 |.000 |
| |LABOR FORCE FIXED |.164 |.059 |.089 |2.780 |.006 |
| |AGE OF RESPONDENT |-.004 |.002 |-.065 |-2.083 |.037 |
| |HIGHEST YEAR OF SCHOOL COMPLETED |.053 |.008 |.210 |6.597 |.000 |
|a. Dependent Variable: HEALTH 3 FIXED |
|TABLE 1: OLS REGRESSION OF EMPLOYMENT AND HEALTH |

After controlling for the potential confounders, we find the relationship between labor force status and health to be statistically significant, with a p-value of .006. The Betas also make sense as well; employment improves health while increases in age decreases one's health. Secondly, we examine the relationship between employment and our mediating variable of social capital, hoping again to find a statistically significant relationship to fulfill the requirements for mediation. The results of the regression model are:
|Coefficientsa |
|Model |Unstandardized Coefficients |Standardized |t |Sig. |
| | |Coefficients | | |
| |B |Std. Error |Beta | | |
|1 |(Constant) |-1.585 |.248 | |-6.391 |.000 |
| |LABOR FORCE FIXED |.220 |.102 |.069 |2.152 |.032 |
| |AGE OF RESPONDENT |.024 |.003 |.256 |8.146 |.000 |
| |HIGHEST YEAR OF SCHOOL COMPLETED |.205 |.014 |.458 |14.372 |.000 |
|a. Dependent Variable: Ultimate Social Capital |
|TABLE 2: OLG REGRESSION FOR EMPLOYMENT ON SOCIAL CAPITAL |

Here we can see that even in our model employment still correlates with social capital, with a p-value of .032, and a beta of .069, implying a positive relationship between the two. While our confounders, correlate higher with social capital and produce higher Betas than employment does, it is encouraging that they do not cause the relationship between employment and social capital to become statistically insignificant. The third requisite for mediation is that our mediator predicts our dependent; that social capital correlates with and predicts health. Another regression yielded;

|Coefficientsa |
|Model |Unstandardized Coefficients |Standardized |t |Sig. |
| | |Coefficients | | |
| |B |Std. Error |Beta | | |
|1 |(Constant) |1.326 |.218 | |6.090 |.000 |
| |Ultimate Social Capital |.138 |.031 |.223 |4.444 |.000 |
| |HIGHEST YEAR OF SCHOOL COMPLETED |.057 |.014 |.204 |4.217 |.000 |
| |AGE OF RESPONDENT |-.011 |.002 |-.214 |-4.631 |.000 |
|a. Dependent Variable: HEALTH 3 FIXED |
|TABLE 3: OLS REGRESSION FOR THE EFFECTS OF SOCIAL CAPITAL ON HEALTH |

With a p-value of .000, we see that our mediator of social capital correlates significantly with our independent variable health. Thusly we have established the three conditions required for mediation as given by Barron and Kenny; employment status predicts health, employment predicts social capital, and social capital predicts health. After verifying that these three conditions for mediation are met, Barron and Kenny state that the final measure of mediation is, “when Paths a and b [independentàmediator and mediatoràdependent respectively] are controlled, a previously significant relation between the independent and dependent variables is no longer significant” (1176). Therefore, in order to continue and complete our mediation analysis, we must re-run our first model for employment’s effect on health, however this time, controlling for the effects of the hypothesized mediator, social capital. Once we include our hypothesized mediator into the model, the statistical significance of social capital in predicting the variance in health should either decrease partially, suggesting partial mediation, or completely, becoming statistically insignificant, suggesting complete mediation through our mediator, social capital. Here is the full model, including our mediator, independent, and control variables:

|Coefficientsa |
|Model |Unstandardized Coefficients |Standardized |t |Sig. |
| | |Coefficients | | |
| |B |Std. Error |Beta | | |
|1 |(Constant) |1.314 |.241 | |5.462 |.000 |
| |Ultimate Social Capital |.099 |.034 |.172 |2.873 |.004 |
| |HIGHEST YEAR OF SCHOOL COMPLETED |.034 |.015 |.130 |2.224 |.027 |
| |AGE OF RESPONDENT |-.007 |.003 |-.116 |-2.182 |.030 |
| |LABOR FORCE FIXED |.359 |.099 |.188 |3.637 |.000 |
|a. Dependent Variable: HEALTH 3 FIXED |
|TABLE 4: OLS REGRESSION, EMPLOYMENT & HEALTH W/ SOCIAL CAPITAL |

Unfortunately, our initial hypothesis that social capital served as a mediating variable between employment and health does not hold up; employment's P value decreases and Beta actually increases with the inclusion of social capital, and social capital's P value increases while it's Beta decreases. Despite the fallacy of our initial hypothesis, this in and of itself is very interesting to note, for this model would indicate that our initial hypothesis was only wrong in the path of the proposed causal relationship (employment → social capital → health). Social Capital's relationship with health is less statistically significant with labor included, satisfying Barry and Kenny's final test for mediation (the significance of the dependent → independent relationship being reduced with the inclusion of our supposed mediator, now employment). Thusly, we now must consider the new possibility causal relationship social capital → employment → health. Of the other requirements for mediation, only one new regression is needed; table 3 satisfies the independent → dependent, table 1 the mediator → dependent, and table 4 provides the final proof of mediation (social capital's influence on health is reduced by the inclusion of employment, the purposed mediator). The only requirement for mediation missing is the independent → mediator test, i.e whether or not the social capital → health causal relationship is statistically significant. Table 5 below contains the regression:

|Coefficientsa |
|Model |Unstandardized Coefficients |Standardized |t |Sig. |
| | |Coefficients | | |
| |B |Std. Error |Beta | | |
|1 |(Constant) |.581 |.088 | |6.571 |.000 |
| |Ultimate Social Capital |.028 |.013 |.089 |2.152 |.032 |
| |HIGHEST YEAR OF SCHOOL COMPLETED |.025 |.006 |.177 |4.387 |.000 |
| |AGE OF RESPONDENT |-.005 |.001 |-.166 |-4.528 |.000 |
|a. Dependent Variable: LABOR FORCE FIXED |
|TABLE 5: OLS REGRESSION OF SOCIAL CAPITAL'S EFFECT ON LABOR |

The table above illustrates the statistically significant relationship between social capital and labor, satisfying the only missing requirement of the Barry and Kenny's criteria for mediation between social capital, employment, and health. Table 4 represents the final test; social capital's previously highly significant relationship with health (Beta of .223, P value of .000) is lessened with the inclusion of respondent's employment status (Beta of .172, P value of .004) suggesting partial mediation. To take this model for social capital’s effects on health further, a sobel test was conducted to test the significance of the indirect relationship between social capital and health through healthy behaviors. Data required by the sobel test is the regression coefficient (A) and standard error (SEA) for the relationship between social capital and the regression coefficient (B) and standard error (SEB) for the relationship between healthy behaviors and for the relationship between healthy behaviors and health (while controlling for social capital). Inputting these values (A =.115 SEA = .038, B = -.097, SEB = .022) into a sobel test calculator available online at , yields a Sobel test statistic of 1.70254841 with a p-value of .044, meaning the indirect effect of social capital on individual health through health behaviors is significant (the .044 P value also supports our assertion of partial mediation).
Discussion and Conclusion From a theoretical standpoint, there is no lack in the literature a wealth of material on the effects of social capital on employment. Rochelle Parks-Yancy writes that, “compared to men and whites, women and blacks often do not know as many people who can provide them with useful information, influence, or opportunities and who can help them with career advancement” (Parks-Yancy 516). This is one way that social capital effects employment status; the higher one's social capital is, the more access to better information regarding job openings and availabilities and better chances through recommendations of obtaining that job. Furthermore, “ the fact that employers often pay monetary bonuses to their employees for successful referrals suggests that employers view workers' social connections as resources in which they can invest, and which might yield economic returns in the form of better hiring outcomes” (Castilla 1289). In a study on the effects of social capital on employment, Yu-Chieh Lo writes, “For a typical person, if everything being equal except that his social capital is enhanced from the lower to the upper quartile, then his chance of being employed will be raised from .80 to .92” (10). Thusly, the discovery of this paper that social capital affects employment is one that has been elaborated on and analyzed in previous scholarly work. However,employment's mediating effect in the social capital → health relationship is interesting indeed. In discussing the relationship between social capital and health, Markku Hyppa writes: “Health-related behaviors are very seldom direct determinants of health; rather, they interact with social environment, psychological features, and biological factors [. . .]. In the light of the lifestyle theory, statistically adequate population health surveys identifying direct and indirect relationships between health behaviors and social capital are urgently needed (114).
Answering Hyppa’s request, this research project has strived to empirically analyze the multi- faceted relationship between social capital and healthy behaviors, i.e. individual health. While there has been studies done on the effects of social capital and health, such as Berkman and Syme’s “Social Networks, Host Resistance, and Mortality” and House JS, Landis KR, Umberson D. “Social Relationships and Health” , employment has been excluded from all of these studies as a significant part of the equation in the social capital à health causal relationship. The results of the empirical portion of this paper call for a merging of the literatures on social capital and employment and social capital and health. In no way, shape, or form however, does this analysis definitely prove one way or another that social capital definitively causes good or bad health vis-a-vi it’s influence on employment. Taking up rhetoric like that not only undermines the entire quantitative method, but also places constraints on such an intricate web of causality, especially considering the mediating effect is only partial and not whole. Thusly, one should not view “causal” analyses such as these as explanatory; rather, one should walk away from a analysis like this not with the idea that social capital causes good health through employment, but rather that social capital, employment, and health are connected, important to one another without being wholly responsible one way or the other.

Works Cited

Baron, Reuben M., and David A. Kenny. "The Moderator-mediator Variable Distinction in Social

Psychological Research: Conceptual, Strategic, and Statistical Considerations." Journal

of Personality and Social Psychology 51.6 (1986): 1173-182. Print.

Farr, James. "Social Capital: A Conceptual History." Political Theory 32.1 (2004): 6-33. Print.

Ferlander, S. "The Importance of Different Forms of Social Capital for Health." Acta

Sociologica 50.2 (2007): 115-28. Print.

Fernandez, Roberto M., Emilio J. Castilla, and Paul Moore. "Social Capital at Work: Networks

and Employment at a Phone Center." American Journal of Sociology 105.5 (2000): 1288-

356. Print.

Hyyppä, Markku T. Healthy Ties Social Capital, Population Health and Survival. Dordrecht:

Springer, 2010. Print.

Link, Bruce G., Mary Clare Lennon, and Bruce P. Dohrenwend. "Socioeconomic Status and Depression: The Role of Occupations Involving Direction, Control, and Planning." American Journal of Sociology 98.6 (1993): 1351. Print.
Lo, Yu-Chieh. Human Capital, Social Capital, Ethnic Capital and Employment Outcomes. Thesis. Princeton University, 2006. Print.
Kawachi, I., B. P. Kennedy, K. Lochner, and D. Prothrow-Stith. "Social Capital, Income Inequality, and Mortality." American Journal of Public Health 87.9 (1997): 1491-498. Print.
Moore, S., V. Haines, P. Hawe, and A. Shiell. "Lost in Translation: A Genealogy of the "Social Capital" Concept in Public Health." Journal of Epidemiology & Community Health 60.8 (2006): 729-34. Print.
Parks-Yancy, R. "The Effects of Social Group Membership and Social Capital Resources on Careers." Journal of Black Studies 36.4 (2006): 515-45. Print.

Putnam, Robert D. Bowling Alone: The Collapse and Revival of American Community. New

York: Simon & Schuster, 2000. Print.

White, R. and Cunneen, C., “Social Class, Youth Crime and Justice.”

Wickrama, K.A.S, Frederick Lorenz, Rand D. Conger, Lisa Matthews, and Glen Elder. "Linking Occupational Conditions to Physical Health through Marital, Social, and Intrapersonal Processes." Journal of Health and Social Behavior 38.4 (1997): 363-75. Print.