متا تجزیه و تحلیل از اثر آموزش بر سرمایه اجتماعی

To assess the empirical estimates of the effect of education on social trust and social participation – the basic dimensions of individual social capital – a meta-analysis is applied, synthesizing 154 evaluations on social trust, and 286 evaluations on social participation. The publication bias problem is given special emphasis in the meta-analysis. Our statistical synthesis confirms that education is a strong and robust correlate of individual social capital. The meta-analysis provides support for the existence of a relative effect of education on social participation, and of a reciprocity mechanism between the dimensions of social capital. The analysis also suggests that the erosion of social participation during the past decades has coincided with a decrease of the marginal return to education on social capital. Finally, we find differences in the return to education between genders, between US and other nations, and variations for different education attainments.

The interest in social capital has led to a profusion of studies on its economic and social effects, as well as its sources of origin and its accumulation mechanisms. Social capital as Putnam, 1993, Putnam, 1995 and Putnam, 2000 connotes, is an aggregate concept that encompasses the association networks, norms and trust that facilitate collective interactions for mutual economic and social benefits. Social capital at the individual level is generally seen as an aggregate of two dimensions—trust in general people and personal involvement in social activities. Individual social trust reflects the bond that people share across economic and ethnic groups (Rothstein & Uslaner, 2004). High levels of social trust lead people to expect that others are cooperative and not opportunistic in social and economic exchanges, which reduces transaction cost and helps solve the free-rider problem in providing public goods. Social trust is usually measured by the response to the following question: “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?” This operationalisation of social trust has been widely used in surveys around the world, including the General Social Survey (GSS) and the World Values Survey (WVS). Individual social participation is a general indicator to denote level of personal involvement in social activities. Social participation covers all types of active affiliation with groups outside the family and voluntary activities unrelated to political purposes, such as voting and lobbying. A high level of social participation is supposed to raise civic norms among people and fortify the foundation of a democratic society. Two sub-categories are distinguished to capture the complexity and diversity of social participation: membership in non-political groups (clubs and other organizations) and participation in voluntary activities. Actually, they are both measured as either the probability of joining non-political groups or participating in voluntary activities, or the degree of social involvement, that is, number of memberships or frequency of participation. Education, according to Putnam, 1995 and Putnam, 2000, Brehm and Rahn (1997), Alesina and La Ferrara (2000a), is one of the most important determinants of social capital. It reflects an orientation towards the future by strengthening human capital and social capital for economic and social development. Schooling spreads knowledge – the basic component of human capital, and cultivates social norms – the core of social capital. Schooling is the first non-familial context in an individual's life where moral and cognitive capacities are trained (Offe & Fuchs, 2002). Through civil education from schooling, students learn the basic norms and responsibilities in society, as well as the functioning of democracy. Glaeser, Laibson, Scheinkman, and Scoutter (1999) assert that the most robust correlate of social capital variables is years of schooling. Using the World Values Survey, they find a positive relationship between schooling and membership of organizations in almost every country. Denny (2003) claims that acquiring a 4-year university degree is associated with a 10 percent higher probability of an individual engaging in voluntary works. Putnam, 1995 and Putnam, 2000, Uslaner, 1997 and Uslaner, 1998, Alesina and La Ferrara, 2000a and Alesina and La Ferrara, 2000b also show that more educated people are more likely to have higher trust in other people and they tend to join more social organizations and participate in group activities more frequently. However, the estimated educational returns vary across studies due to heterogeneities in survey sources, and methodological or contextual variations. In this paper we provide a research synthesis of the estimated educational returns to individual social capital from relevant studies. We aim to obtain an inclusive review of educational returns to individual social capital, and evaluate the possible variation sources in the estimated effect of education on social capital. Two criteria were used for the inclusion of the available literature in the meta-analysis: (a) studies should focus on the determinants of at least one dimension of social capital at the individual level with formal education as a covariate in the model; (b) studies should have reported statistical data (t-statistics, p-value or standard error) that allow for estimation by the fixed effects and random effects models. A dataset is created for our analysis that includes estimates from 65 studies. Twenty-eight studies provide estimates of the return to education on social trust and 37 studies provide estimates on social participation. Table 1 presents information on the authors, year of publication and survey period, classified by social trust and social participation.The number of estimates varies markedly from 1 to 88 because some studies offers the estimates of educational return for each nation in the surveys (for instance, Denny, 2003 evaluates the effect of schooling on social participation for 20 countries respectively using International Adult Literacy Survey, and Glaeser et al., 1999 evaluates the effect of schooling on social trust for 20 countries respectively using World Value Survey). The evaluation methodology may differ in the same study as authors may compare the estimates from simple linear model and that from a model that accounts for endogeneity of education on social capital (such as Dee, 2003 and Denny, 2003; Huang, Maassen van den Brink, & Groot, 2008a; Milligan, Moretti, & Oreopoulos, 2003). Contextual variations in the same study are also a key factor for the large number of effect sizes as authors may compare estimates from different specifications of the model. They may investigate the difference in educational returns in the model with and without average education return (such as Glaeser et al., 1999; Helliwell & Putnam, 1999; Marschall & Stolle, 2004), or they may compare the education return for men and for women, for the elderly and for youth, for college education and for high school education. The majority of studies in our dataset do not consider the possibility that the choice of educational attainment and social capital are simultaneously influenced by unobserved heterogeneity specific to the individuals. The ignorance of the endogeneity problem can cause biased estimates of the educational return. For example, it is plausible that people with good relations with parents and friends in their childhood may obtain better education and have higher social capital in adulthood. However, these interactions in childhood usually turn out to be unobservable to researchers. Some studies in our dataset have taken account of the endogeneity problem. Using policy reform dummies as instruments, i.e. the increase in the minimum schooling age and abolition of tuition fees for secondary school, Denny (2003) applies a two-step procedure in the evaluation for Britain, Italy, Northern Ireland and the Republic of Ireland. The results in his study are mixed, although he observes a positive relation between education and altruistic (charity) activities in most West European counties. Dee (2003) employs 2SLS and bivariate probit, by relying on changes in teen exposure to child labor laws, to estimate the educational impact on the probability of joining social groups and volunteering in social services, and the impact on the number of affiliated groups. He confirms the substantial causal effect of schooling on most measurements of social participation, except for the frequency of voluntary work. Changes in compulsory schooling law are also applied in the studies of education and social trust. Milligan et al. (2003), for example, apply this strategy in their study of educational return to trust and other civic outcomes. They do not observe any substantial difference between the estimates from OLS and 2SLS regressions. Huang et al. (2008a) separate a non-systematic health component from the schooling absence length due to illness as a powerful instrument of education choice which is not correlated with adult health. Similarly, this study finds no sign of endogeneity in the college choice of observations. Does the estimation method to account for the endogeneity problem produce considerably different estimates of the educational return to social capital? Our meta-analysis will shed some light on this question. It is also noteworthy that one's social capital can be affected not only by one's own education, but also by that of others around him. The impact of education on social capital can be distinguished into a relative effect and an absolute effect, according to Nie, Junn, and Stehlik-Bany (1996). The relative effect indicates that education is a proxy for relative status, a sorting mechanism for people with higher ability in acquiring social capital. The absolute effect refers to the accumulation of civic values and knowledge. One's own education level is not directly linked to the level of individual social capital, given that education merely serves to sort people of different capabilities in social capital (Helliwell & Putnam, 1999; Nie et al., 1996). Therefore, if more people have a college degree, Nie et al. (1996) argue, perhaps the sociological significance of the credential for high school has been devalued. This issue is covered in our meta-analysis. We will assess both effects of education by evaluating the effects of individual schooling years and average schooling years in the region. We further test whether the impact of 1 more year of schooling on social trust and social participation varies across different levels of education, or whether it rises with education attainments. Coleman (1990) and Putnam, 1993 and Putnam, 1995 suggest a reciprocal effect between trust in general people and participation in social activities: “Social trust, norms of reciprocity, networks of a civic engagement and successful cooperation are mutually reinforcing” (Putnam, 1993, p. 180). Brehm and Rahn (1997) posit an asymmetric causal chain in which trust is the direct outcome of civic engagement. But this asymmetric association is disputed by Uslaner (1997), who argues that trust shapes civic participation. These hypotheses will be tested by including controls for reciprocity between trusting and participating in the meta-analysis. Many studies, such as Ibáñez, Lindert, and Woolcock (2002), Brehm and Rahn (1997), Claibourn and Martin (2000), Li, Savage, and Pickles (2003), Liu and Besser (2003), Newton (2001b), Rahn, Yoon, Garet, Lipson, and Loflin (2003), Uslaner, 1997, Uslaner, 1998, Uslaner, 2003 and Uslaner, 2004b and some others in our dataset provide information of the role of schooling in this reciprocity mechanism. We will address several other questions on the relation between education and individual social capital. For instance, are gender differences a critical factor in explaining the variation in the effects of education? Does education play a role in the so-called erosion of social capital during the past decades? Americans are believed to have more social capital than people in other nations (Putnam, 1995 and Putnam, 2000). By means of a meta-analysis of the estimates taken from the surveys across nations, we are able to examine whether education systems in the US and other nations are related to the social capital inequality. In Section 2 we present a brief introduction of effect size and its role in meta-analysis. Summary statistics of the pooled effect sizes are presented in this section, along with the test statistics for the hypothesis of the fixed effects and random effects model. In Section 3 we extend the appropriate model for further analysis on the influence of study characteristics on the study estimates. Sensitivity tests – the Egger's test and Hedge's correction procedure – are applied to check and correct for publication bias. Section four offers a discussion for the results from the meta-analysis and their relevance to the mentioned topics in Section 1.

In this meta-analysis, we find that one standard deviation of years of schooling accounts for the change in individual social capital by 12–16 percent of the standard deviation in each dimension. The findings confirm that education is a strong and robust correlate of individual social capital. The hypothesis that the fixed effects model is an appropriate synthesis method is rejected by the Q-statistics. Study heterogeneities have a substantial influence on the variation of the effect sizes. Gender differences play a role in the mechanism by which education affects social capital, as women exhibit a statistically negative influence on the effect sizes of both dimensions of individual social capital. The endogeneity problem in schooling achievement and social capital outcome is a critical source of variation of study estimates of social participation, but it does not have any impact in study estimates of social trust. In social participation studies, almost every estimate which accounts for endogeneity turns out to be smaller than the corresponding one which does not account for endogeneity in the same study. The endogeneity problem poses a question of the positive role of schooling on social participation, as the coefficient of endogeneity control offsets the benchmark estimate (the constant term) in our meta-analysis (see Table 4 and Table 7), and the mean effect size of study estimates which accounts for endogeneity is merely 0.01. The possibility that education is not an exogenous determinant of social capital should be taken into account in the study of the sources of social capital, especially when it comes to social participation outcomes. Comparing the effect sizes obtained from surveys conducted before the 1990s with those from surveys in a later period, we observe a decline in the return to education on social participation. The decline in education effect provides a clue for the erosion of civic engagement in the United States (Putnam, 1995 and Putnam, 2000), despite a dramatic increase in educational attainment during the last half century. Our finding is at odds with an upward trend in the wage effect of schooling, as found in a meta-analysis of wage return to education by Ashenfelter, Harmon, and Oosterbeek (1999). One possible clue for this discrepancy is that there is a trade off between the return to education on wage and that on social capital. Increased economic competitions and increased demand for individuals with a high-level human capital in modern society may be seen as a culprit. The emphasis on the role of schooling as a source of human capital could be detrimental to the contribution of schooling on cultivating social capital for collective welfare, as the role of civic education, which do not directly increase competence or income potential in the future, may be dwarfed in school programs or may appear less attractive to students who are anxious for an education that provides opportunities for a good job. We do not see a declining trend in the return of education on social trust. Social trust has its root in individual morality that people follow in social and daily life (Fukuyama, 1995, pp. 153; Uslaner, 1999). People with faith in racial and gender equality, for instance, are more tolerant towards minorities and to others who are not like themselves and these people have a higher level of trust in others. For schools, one of its essential functions is setting moral standards and equipping students with a basic sense of morality. It is possible that this function is not weakened in modern education. However, it is just a hypothesis and more studies are needed to find some clues on this topic. Our analysis provides proof for the view that schooling has a higher effect on social capital in the United States than in the rest of the world. American schools are believed to be more active than schools in other countries in encouraging students into running student offices, participating in civic engagement and joining various associations. The melting pot theory can also help explain why Americans tend to receive a higher educational return on social capital. Encouraging tolerance of ethnic diversity and creating core values of a common American heritage are the main subjects of the social education programs in American public schools. By exposing students to knowledge about ethnic diversity and the contributions of various groups to American civilization development, educators may change negative ethnic group stereotypes, reduce intolerance, and enhance cooperation for the common good. The strongly significant influence of controlling for average education in social participation confirms the existence of a relative effect. For social trust, no evidence is found in support of a relative effect. The descriptive statistics in Table 3 show that the mean effect size for social participation is 0.045 in studies that do not control for average education, and 0.116 in those controlling for average education (both are statistically significant at the 0.0001 level). This provides evidence for both an absolute effect and a relative effect. The relative effect does not dominate the absolute effect of education, so the total effect on social participation is still positive and substantial. It may be difficult to interpret why the effect size is positively associated with the inclusion of average education level control. A simple linear model is chosen to elucidate this insight, assuming education to be the only determinant of social participation (SP): SP=a×edu1+b×(edu1−avedu)SP=a×edu1+b×(edu1−avedu) Turn MathJax on where a represents the marginal absolute effect of years of individual schooling on social participation, b represents the marginal signaling effect or relative effect (years of schooling compared to the average years of schooling in the region). If education has a signaling effect as well as an absolute effect on individual social capital, we expect both a and b to be positive. In addition, years of schooling are positively associated with the average level of education in the region where he or she lives. Mathematically, individual education level is included into the calculation of average education level. Furthermore, higher educated people are more inclined to live in regions with a higher average education level, since people have a preference for a homogenous region with similar social-economic status. More details can be found in Alesina and La Ferrara (2000b), who show that social-economic heterogeneity reduces trust and feelings towards other people. In the restricted model, the covariate of average education level of the region is dropped: SP=c×edu1SP=c×edu1 Turn MathJax on Then the negative effect of average education level will be absorbed by individual schooling years. Thus the estimate of the effect of individual education, coefficient c in the restricted model, will be smaller than that in the full model, which equals a + b. This explains why we observe a positive impact of the control variable for average level of education on the magnitude of the effect of individual schooling years. The size of the schooling effect varies with the levels of education. Effect sizes are significantly higher for people with a college degree or above. The popular one-factor OLS model, where it is assumed that education can always be aggregated into a single measure, say years of schooling, hence may not be a sufficient model to capture the effects of education on social capital. Why the effect of education demonstrates a substantial leap for people with college degree? It is possible that college education is a more efficient and critical stage for individuals to learn to trust other people and cultivate an active civic behavior. Alternatively, a college degree may signal the existence of unobserved ability – individual personality or other inherent psychological attributes – that positively affect both educational achievement and the level of individual social capital. There is evidence to suggest that controlling for a reciprocity mechanism between the two dimensions of individual social capital influences the estimates of educational return. This confirms a “virtuous circle” (Putnam, 1995) in the accumulation of social capital. If social trust is included as an explanatory variable in the social participation equation, the estimate of the marginal effect of schooling years will be lower. The reverse is also true. The intuition behind this is straightforward: since education has a significantly positive effect on both dimensions, and there is a mutual, positive effect between these dimensions, the direct effect of education on a dimension (after controlling for reciprocal effects) will be lower than the total effect (without controlling for reciprocal effects). The significant impact of controlling for reciprocal effect also provides support for the central role of schooling in the generation of social capital. We do not find any substantial difference in the effect sizes across life stages. It is plausible that there is a constant educational return on individual social capital over time. Controlling for media influence has hardly any impact on the estimates of educational returns. It is surprising to observe a negative influence of the inclusion of environment concern (urban or rural differences, population density, average income in the region, etc.) in study estimates of social trust but a positive influence in study estimates of social participation. Do urban schools have better access to resources, facilities and financing and provide better quality social education than rural schools? Do life experiences in urban areas, which are more heterogeneous and complicated, spill over into people's social values and reduces the influence of education on social trust? Unfortunately, there is no explicit answer due to the lack of information in the literacy on these issues. The outcomes from the meta-analysis pose several interesting topics. It will be useful to further explore whether there is a trade off between the return to education on income, and that on social capital. One may also go further and investigate whether a college education is a key treatment for people to obtain more social capital. More studies also need to be done to examine the effects of education for men and for women, for urban region and for rural region, and over the life cycle. We are looking forward to seeing new studies on the association between education and social capital.