Does social capital determine health? Empirical evidence from MENA countries

Does social capital determine health? Empirical evidence from MENA countries

G Model ARTICLE IN PRESS SOCSCI-1211; No. of Pages 9 The Social Science Journal xxx (2014) xxx–xxx Contents lists available at ScienceDirect The ...

487KB Sizes 0 Downloads 24 Views

G Model

ARTICLE IN PRESS

SOCSCI-1211; No. of Pages 9

The Social Science Journal xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

The Social Science Journal journal homepage: www.elsevier.com/locate/soscij

Does social capital determine health? Empirical evidence from MENA countries Moheddine Younsi ∗ , Mohamed Chakroun University of Economics and Management of Sfax, Unit of Research in Development Economics (URDE), Airport Road, km 4.5, LP 1088, Sfax 3018, Tunisia

a r t i c l e

i n f o

Article history: Received 17 December 2013 Received in revised form 30 August 2014 Accepted 30 August 2014 Available online xxx Keywords: Social capital indicators Health Instrumental variable estimates Middle East and North Africa Countries

a b s t r a c t The paper attempts to evaluate the causal relationship between health and trust used as a proxy of social capital, both at the individual and at the community level for a sample of eighteen countries from the Middle East and North Africa. The data were drawn using the World Values Survey for eighteen countries, supplemented by regional level data. The results suggest a reversed causal effect: individual social capital has a causal beneficial impact on health and vice versa. However, the effect of health on social capital appears to be significantly higher than the social capital effect on health. The findings demonstrate that the people in good health have a higher propensity to take part in social activities and to benefit from it. Conversely, the other part of the population in poor health may see its health worsening faster because of the missing beneficial effect of social capital. © 2014 Western Social Science Association. Published by Elsevier Inc. All rights reserved.

1. Introduction The literature on social fact and health has for a long time been a well-established research topic in public health (Berkman & Syme, 1979; Brown & Harris, 1978; Cobb, 1976; Lynch, 1977). It is only since the 1990s that subsequent studies dealing with social connectedness and social cohesion have systematically been referred as “social capital”. In the sociological literature, several definitions of social capital have been proposed. Bourdieu (1986) defines social capital as a capital of social connections, mutual acquaintance and social recognition. Coleman (1988) refers to social capital as all those features of the social structure which might facilitate actions of individuals within the social structure itself. Putnam, Leonardi, and Nanetti (1993) define social capital as “those features of

∗ Corresponding author. Tel.: +216 74 27 97 10; fax: +216 74 27 91 39. E-mail addresses: [email protected] (M. Younsi), [email protected] (M. Chakroun).

a social organization, such as trusts, norms, and networks that can improve the efficiency of the society by facilitating coordinated actions.” Although all definitions refer directly or indirectly to social connections or social network, as elements of social capital, Putnam’s definition points to the role of social capital as a catalyst of coordination, an essential device to achieve better outcomes, either social or economic. Coordination reduces transaction costs, to overcome difficulties due to incomplete or asymmetric information, and to establish efficient transactions in presence of incomplete contracts (Alesina & La Ferrara, 2002). It may be that the emphasis on trust as an indicator of social capital is well placed because trust favors cooperation, without the need of creating long standing personalized relationships and processes of reputation building. Moreover, trust is a determinant of social connections, as a minimum amount of trust is required to initiate a social interaction (Ghosh & Ray, 1996; Kranton, 1996). Nevertheless, empirical research undoubtedly provided some thriving conceptual and theoretical developments (Kawachi, Subramanian, & Kim, 2008). For the pros, social

http://dx.doi.org/10.1016/j.soscij.2014.08.008 0362-3319/© 2014 Western Social Science Association. Published by Elsevier Inc. All rights reserved.

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model SOCSCI-1211; No. of Pages 9

ARTICLE IN PRESS M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

2

capital is an encompassing umbrella under which unprecedented patterns of thinking have emerged. According to robustness checks based on changes in cut-points in health measures (D’Hombres, Rocco, Suhrcke, & McKee, 2010; Folland, 2007; Islam, Merlo, Kawachi, Lindstrom, & Gerdtham, 2006; Poortinga, 2006a), the social participation on health appeared in some cases important pathways through which health may be improved. The crucial question is whether this relationship reflects a causal impact from social capital to health a reverse causal impact or the influence of other factors simultaneously affecting social capital and health. One of the most salient new strands of research investigates joint individual and contextual effects of social capital on health. Several recent studies have already emphasized the positive influence of multilevel measures of social capital on individual health outcomes (Olsen & Dahl, 2007; Scheffler, Brown, & Rice, 2007). A common finding of these studies suggests that the influence of social capital is underestimated when multi-level influence is not taken into account. Another important contribution of the social capital literature has been to go beyond correlations. The recourse to instrumental variables shed some light on the until-then ill-known causal relationships between social capital and health (D’Hombres et al., 2010; Folland, 2007). This paper contributes to the debate from an empirical perspective, first, by briefly assessing the level and recent trend in selected social capital indicators and, second, by examining whether social capital has impacted upon health in eighteen Middle East and North Africa (MENA) countries for which data has been available. We investigate the impact of social capital on individual selfreported health for a sample of eighteen Middle East and North Africa (MENA) countries, using the sixth round of the World Values Survey (WVS, 2010–2012). This survey offers large possibilities to tackle some of the econometric challenges involved. The paper is structured as follows. Section 2 presents a review of the literature. Section 3 reports empirical evidence on the level and trends in social capital in the MENA countries, using the World Value Survey data. Section 4 presents the in-depth analysis of the causal impact of social capital on health in eighteen MENA countries. Section 5 concludes by summarizing the main results. 2. Literature review The relationship between social capital and health has been documented since 1901, when Emile Durkheim identified a relationship between suicide rates and the level of social integration. Since then research has continued to demonstrate that higher social capital and social cohesion are associated with improved health conditions. Recent research shows that the lower the trust among citizens, the higher the average mortality rate (Berkman & Syme, 1979; Kawachi & Berkman, 2000). Until quite recently, the literature that measures social capital at the area level, has been much smaller than the literature that measures social capital at the individual level (D’Hombres et al., 2010; Folland, 2007; Poortinga, 2006b). Findings somewhat vary with respect to the strength of the

observed associations depending on the geographical context, on the choice of the geographical unit at which the social capital is measured as well as the specific measures of the social capital and the health conditions employed in empirical analyses (Brown, Scheffler, Seo, & Reed, 2006; Mohan, Twigg, Barnard, & Jones, 2005). Furthermore, measurements of the community social capital that are based simply on an aggregation of individual social capital indicators tend to be viewed as the second-best, although they are often used as proxies for more relevant community-based measures. This view leads to a concept of community social capital that is not merely a sum of the social capital of the individual members of that community. It is argued that social capital can impact health through various channels: From a macro level of analysis, social capital may facilitate health care delivery. The better the social network among and between each group of health care providers (the government, the market and the family/community), the more efficiently and effectively health care could be delivered. Community and volunteer organizations play a central role in providing services to patients in both developing and industrialized nations. Social capital may also support prevention efforts. Prevention can only be effective, if it is supported by formal and informal networks through which people receive information and medicine. From a meso and micro level of analysis, social capital can improve health through enforcing or changing social norms. A more cohesive society, with a strong feeling of group identity tends to be attentive to common wellbeing (Kawachi & Berkman, 2000; Lochner, Kawachi, Brennan, & Buka, 2003): this implies that environment-damaging behaviors are avoided and entrepreneurs are more likely to take care of a healthy workplace and work environment in their firms. Moreover, smoking, sanitation, and risky sexual practices are behaviors, which often negatively impact public health: all such behaviors are less likely within a socially cohesive society (Brown et al., 2006). Finally, shared values and norms can also have an impact on the level of community violence and, therefore, on the frequency of injuries and violent deaths. From a very micro/individual perspective, intensive social interactions offer a privileged channel for information transmission and sharing of past experiences on health facilities, doctors, drugs and diseases, thus reducing the cost of health information (Ghosh & Ray, 1996; Kranton, 1996). Moreover, trust by facilitating cooperation, gives access to support, aid and care services provided by informal institutions based on reciprocity, which provide insurance in case of health shocks (Alesina & La Ferrara, 2002; La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1997). In this study, we add some evidence to the existing literature since we test the relationship between social capital and health taking explicitly into account the circular relationship between the two measures and modeling the errors in reporting the true level of individual social capital. Anticipating the results, first we corroborate the hypothesis of a positive causal impact running both from social capital to health and from health to social capital, second, we suggest that community social capital plays a minor role in determining people’s health and that it is the individual

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model

ARTICLE IN PRESS

SOCSCI-1211; No. of Pages 9

M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

social capital the one that matters the most. In addition, we find strong evidence of misreporting in individual social capital. 3. Data and descriptive statistics The data used for these analyses were obtained from the sixth round of the World Values Survey (WVS, 2010–2012) for a sample of eighteen Middle East and North Africa (MENA). Table 1 summarizes the descriptive statistics for the countries: Algeria, Bahrain, Egypt, Iran, Iraq, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Qatar, Saudi Arabia, Sudan, Syria, Tunisia, Turkey and Yemen. In all, 105,166 individuals were studied (response rate 78.5%). Subjects reported information on sociodemographic variables such as gender, age, marital status, place of residence, religion, education, income and occupational status, with the latter three serving as proxy estimators of socioeconomic status. Results of descriptive analyses are expressed as mean and standard deviations (SD) for quantitative variables and as counts and percentages for categorical variables. The participants were mostly male with an average proportion of 69.5%. The mean age and formal education of the respondents were 46.75(SD 17.85) and 11.92 (SD 4.03) years, respectively. The respondent’s mother education level and

3

Table 1 Number of observation by country. Country

Frequency

Algeria Bahrain Egypt Iran Iraq Jordan Kuwait Lebanon Libya Morocco Oman Qatar Saudi Arabia Sudan Syria Tunisia Turkey Yemen Total

Percent

7253 2286 7694 7880 6968 5425 3685 4553 5398 6764 4263 3234 6643 7224 6227 5793 7423 6453

6.89 2.17 7.32 7.49 6.62 5.16 3.52 4.33 5.13 6.43 4.05 3.07 6.32 6.87 5.92 5.51 7.06 6.14

105,166

100.00

Cumulative 6.89 9.06 16.38 23.87 30.49 35.65 39.17 43.50 48.63 55.06 59.11 62.18 68.50 75.37 81.29 86.8 93.86 100.00

Source: WVS (2010–2012).

the respondent’s father education level were 1.49 (SD 1.34) and 1.93 (SD 1.51) respectively. The simple summary statistics of the variables included in the model for several social capital indicators related to health are shown in Table 2.

Table 2 Descriptive statistics of the variables used in the regression analysis. Variable

Description

Health

Indicator taking the value 1 if the individual reportsto be in very bad and value 5 if the individual reportsto be in very good Indicator taking the value between −5 and 5 if the individual agrees or quite agrees that the majority of people can be trusted Indicator taking the value one if the individual is a male, and zero otherwise Age in years Squared age Respondent born in the country of residence Birthplace of respondent’s mother Birthplace of respondent’s father Urban residence Population density at regional level Marital status Respondent’s education (in years) Respondent’s mother education (level) Respondent’s father education (level) Respondent’s mother employed Respondent’s father employed Respondent’s household income GDP per capita at regional level Growth rate at regional level Number of beds in the regional hospitals per 1000 inhabitants Number of heath personnel at regional level per 100,000 inhabitants Indicator taking the value 1 if the respondents reportsto be in very bad mental health and value 5 if the respondents reports to be in very good mental health Victim of a burglary or assault in the last 5 years Percentage of residents that reported having been victims of burglary/assault in last 5 years Self-reported rate of religiosity

Trust

Male Age Age2 Brncntr Motherbirth Fatherbirth Urban Density Married Educyrs Mothereduc Fathereduc Motheremploy Fatheremploy Income Gdp Gdpgrowth Hospital beds Hospital pers Mental health

Crime victim Mcrime Religious

Mean

Std dev

Min

Max

3.76

0.92

1

5

−0.35

2.43

−5

5

0.47

0.49

0

1

46.75 2563.58 0.72 0.83 0.89 0.62 318.46 0.53 11.92 1.49 1.93 0.31 0.69 5.46 9.76 3.27 1.53 328.62

17.85 1815.91 0.26 0.27 0.28 0.49 703.37 0.49 4.03 1.34 1.51 0.37 0.48 2.29 2.27 1.43 1.19 98.71

14 193 0 0 0 0 3.1 0 0 0 0 0 0 1 3.48 −2.00 1.04 122.8

101 9801 1 1 1 1 6073.5 1 30 6 6 1 1 12 21.66 9.35 2.39 1.75

3.74

0.35

3

5

0.20 0.19

0.39 0.08

0 0

1 0.624

4.96

2.98

0

10

Source: WVS (2010–2012) and MENASTATREGIO database (2012). Note. Regional dummies are Algeria, Bahrain, Egypt, Iran, Iraq, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Qatar, Saudi Arabia, Sudan, Syria, Tunisia, Turkey, Yemen.

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model SOCSCI-1211; No. of Pages 9

ARTICLE IN PRESS M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

4

The WVS provides information about individual social behavior and perception, such as political opinions, political participation, exposition to media and news, social relationships, trust in other people and institutions. In addition, the WVS network is particularly valuable because it provides detailed information about the respondents socioeconomic characteristics and parental background. Information on the region of residence at WVS network is, in most cases, also available. This feature has allowed us to supplement additional data about regional characteristics coming from the MENASTAT REGIO database. In particular, we have added regional indicators of development (GDP per capita, GDP growth and employment), of health supply (number of beds in hospitals per 1000 inhabitants and number of health personnel per 100,000 inhabitants) as well as population density (Table 2). As discussed above, social capital can be defined from many perspectives and there is no consensus in the literature on how social capital ought to be measured. Social capital can be a community, a social group or an individual asset, it can be informal or mediated by formal institutions, it can be inclusive or exclusive. Each dimension of social capital might have a specific impact of health. Regarding health, people are asked to rate their current health on a five-step ladder ranging from very bad (1) to very good (5). Community social capital associated to individual i is measured as the mean trust of the residents in the same region of individual i. Average measures of individual social capital are frequently used in the literature (Islam et al., 2006). For any individual i, community is defined as the set of individuals living in the same the town or village of i. However, our considering large geographical areas lowers the probability of having a variable affected by endogenous sorting into neighborhood. 3.1. The econometric model The structural model is composed of three equations that describe the impact of social capital on health (Eq. (1)), the misreporting in individual social capital (Eq. (2)) and the impact of health on social capital (Eq. (3)). We have three endogenous variables h, the individual health, S, the self-reported individual social capital and S* the unobservable individual social capital. We express S* the average true social capital in the MENA region, which we assume is not reported with error and S¯ ∗ the average reported one. In what follows, we describe and discuss the empirical methodology applied in more detail. h = ˛0 + ˛1 S ∗ + ˛2 S¯ ∗ + Xω1 + ε

(1)

S = S ∗ + (1 − )(S¯ − S ∗ ) + h + 

(2)

S ∗ = ˇh + Zω2 + ı

(3)

Individual health in Eq. (1) is a function of both the true individual social capital S* and the true communitarian one S¯ ∗ as well as of covariates X, including the regional fixed effects. We assume that the error terms are independent. The true individual social capital is reported with error and the misreporting is described in Eq. (2). Indeed, the selfreported individual social capital is not only a function of the true individual social capital S* , but also of the average

¯ In addition, it is declared social capital in the community S. worth noting that even the level of individual health h has an effect on the misreporting (see Eq. (2)). Finally, the true social capital S* in Eq. (3) depends on the individual health and on other controls Z. By introducing the health status in the social capital equation (Eq. (3)), we claim that the relationship between health and social capital is circular. In fact, the dynamics of interaction with other people (and the trust toward them) are shaped by the health status of the individual interviewed. The most striking case is represented by mental health, but also physical one plays a role in determining the individual trust. Taking the average of Eq. (2) and substituting it back into (2), we obtain:





S = S + (1 − )S¯ ∗ +

(1 − ) ¯ (1 − )¯ h+ + h +   

= S ∗ + (1 − )S¯ ∗ + ˝



(4)

¯ which we From Eq. (4), we can express both S* and S, substitute many times in Eq. (1) and we obtain: h = ˛0 + ˛1

S − (1 − )S¯ − ˝ S − (1 − )S ∗ − ˝ + ˛2  1−

+ Xω1 + ε

(5)

Eq. (5) is recursive in h, and it is just a function of the reported average and individual reported social capital S¯ and S, solving for h, the reduced form can be represented by the following estimation equation: h=

˛0 ˛1 / + S (1 − (˛1 /)) (1 − (˛1 /)) +

˛2 (/) (˛2 − (˛1 (1 − )/)) ¯ S+ h (1 − (˛1 /)) (1 − (˛1 /))



¯ Xω1 (˛1 /) ˛2 (/) + − +ε (1 − (˛1 /)) (1 − (˛1 /)) (1 − (˛1 /))

Defining ı the quantity (1/(1 − (˛1 /))), we obtain the reduced form which can be estimated as: h = ı˛0 + ı + ı˛2



˛1 ˛1 (1 − ) S + ı ˛2 −  





 S¯ + ı ˛2 

¯ ˛1 + ıXω1 + ı  + ε  



h (6)

Eq. (6) is the new first equation of the system obtained using (1) and (2), substituting Eq. (3) in (2), the reduced form equation to be estimated is then:



S = ˇ+

 



h + Zω2 + (1 − )S¯ +  + 

(7)

We are left with a simultaneous equation model composed by Eqs. (6) and (7), which are a function of on observed variables. From Eq. (7), we identify  and, plugging it into Eq. (6), we obtain the structural parameters ˛0 , ˛1 , ˛2 and /. In fact, to estimate the simultaneous equation probit model composed by Eqs. (6) and (7), we use a two-step

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model SOCSCI-1211; No. of Pages 9

ARTICLE IN PRESS M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

procedure following Stern (1989) to solve the endogeneity problem. Indeed, in the first stage, we transform the reduced form of Eqs. (6) and (7) in the two models, which can be defined as follows: h∗ = Xh + vh ∗

S = Xs + vs

(8) (9)

where X contains all the exogenous variables in (6) and (7), such as gender, age, age squared of the individual, place of residence, marital status, education, income, religion and occupational status. We consistently estimate the reduced form coefficients of models (8) and (9) by using probit specification equation by equation. After the estimation of the first stage, the predicted values hˆ ∗ and Sˆ ∗ will constructed in the second stage using the estimated coefficients of the

h and  S and substituting them into first stage, that is,  models (6) and (7). Our set of controls, common to all the equations, consists in individual variables such as gender, age (and its square), place of residence, marital. Parental characteristics such as place of birth of the parents, their education and occupational status (when alive) and a dummy indicating the parents died status, education, income, religion and occupational status and regional characteristics length of the unemployment rate, gdp, and density. 3.2. Regional and time dummies

In addition, in the health equation, in line with our structural model, we introduce the average value of health at the regional level. We claim that this variable can be considered exogenous since in our structural model does not enter in Eq. (7). Moreover, we rule out the possibility of endogenous sorting into region both because we consider comparatively large aggregates and because we control for unobserved regional characteristics using the regional dummies. As already mentioned in Section 3.1, in our setting three variables are considered endogenous: the self-reported health status, the individual social capital and the communitarian one (by construction), since one of its determinant,  enters in the health equation (see Eq. (6)). In order to solve the endogenity problem, we have to find proper instruments. Given that community social capital is computed as the average of individual social capital reported by the residents of a region, each instrument suitable for individual social capital is potentially relevant also for community social capital. All instruments must satisfy two requirements, they must be relevant, that is, correlated with the endogenous variables and must be exogenous, that is, they must affect individual health only through the instrumented variables, without independent and autonomous role. We first discuss the relevance of each instrument and next their exogeneity. Consistent with several authors (Alesina & La Ferrara, 2002; Easterly & Levine, 1997), we propose, as a determinant of trust not correlated with health, an exogenous negative shock the individual experienced: being a victim of a burglary. Having been recently a victim of a burglary or an assault is certainly related to the degree of trust in other

5

people. A similar shock is likely to induce a widespread feeling of fear and distrusts against people outside a relatively narrow circle of close friends and relatives. Actually, data show that the negative correlation is very strong. Exogeneity (i.e. the excludability condition of the proposed instruments) requires a more extensive and careful discussion. Exogeneity requires a more extensive and careful discussion. Having been victim of a crime or knowing a close person is certainly not an individual decision or under individual control. However, we cannot claim that being a victim of a crime is a truly random event (i.e. a completely exogenous accident, as people are able to modify the probability of these events by avoiding risky borough of their town, by installing security devices in their houses etc.). Moreover, it is likely that the risk of burglary increases with the people’s age and gender, income and with the crime intensity in the place of residence. Therefore, the instrument is likely to be correlated with age, gender, household income and with crime intensity in the region of residence. The latter variables have an impact on individual health, age and gender obviously, income by determining the opportunities of investment in health, criminality by reducing people’s mobility. Hence, we need to include them among individual and regional controls so what we can claim that the instrument for individual social capital has no independent effects on health. As discussed above, we instrument the average trust in the region using a transformation of the previous instrument, that is, the percentage of people in the region who have been victim of a crime. Since one of the regressions is the average of another one, the coefficient of the aggregate variable can be positive even if such a variable does not have an independent power in determining the dependent one (Acemoglu & Angrist, 2000). In order to have consistent estimates, we need a IV strategy that treats both regressors as endogenous and the instruments for the two regressors should generate the same coefficient when only one variable is considered endogenous. Our instruments meet such a condition. Let us start considering suitable instruments for the selfreported individual health: the instruments we consider are the number of hospital beds in the area in which the individual lives and the number of health personnel. There is evidence that the supply of health care does have a positive impact on the individual health, but we exclude the possibility that health care infrastructure directly affects individual social capital. Taken together, all our instruments could have an impact on individual health that is not mediated exclusively through social capital. However, this impact would not be an autonomous, independent one, but would rather be due to instruments’ correlation, with third variables: parental background, income, regional development. Hence, once we control for the latter, we claim that the exogeneity requirement is met. 4. Results and discussion Table 3 presents the estimates and the significance of the structural parameters of the health equations

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model

ARTICLE IN PRESS

SOCSCI-1211; No. of Pages 9

M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

6 Table 3 Structural parameters. Structural parameter

Coefficient

Std dev

t-Statistics

p-Value

˛1 ˛2   ˇ

0.312*** 0.021 0.962*** −2.415** 2.281**

0.068 0.016 0.009 1.113 1.139

4.59 1.33 99.85 −2.19 2.00

0.000 0.178 0.000 0.033 0.048

** ***

Significant at 5%. Significant at 1%.

(Eqs. (1)–(3)). Column 1 and 2 report the value estimates and standard deviations while the last 2 columns present the t-statistics and the p-value which test the significance of the instruments. Indeed, the estimated parameters of model (1) presented in column 1 and which has been the primary focus of the literature on the relationship between social capital and health, have both the positive expected sign. However, such parameters differ in significance and magnitude. In fact, individual trust has a positive and significant effect on health at the 1% level, and such an effect is greater than the impact of communitarian social capital, which is small and not significant. Model (2) which describes the relationship between true and reported social capital, gives us some interesting insights. As expected, our estimates reflect the presence of some level of misreporting of the value of social capital. Indeed, our results show that the individuals adjust the reported level of social capital according to the difference between individual trust and the average trust in the

region in which they live. In particular, if the average trust is greater than the individual one, people would report a value which is greater than the true one. It is worth noting that the entity of the adjustment is not huge (1− is around 3%), making the coefficient plausible. In addition, the estimated structural parameters of model (2) confirm the idea that health plays a significant role in the misreporting of social capital. It is worth noting that a negative sign of parameter  shows that healthy people tend to report a value of social capital lower than the true latent value. The result is not surprising considering the nature of the measure of social capital we are considering. An explanation of this is that people in good health are less dependent on other people’s help, they underestimate the level of trust that they have. In model (2), the negative sign of parameter  suggests that the weakness to consider misreporting in social capital gives misleading results, that is, the impact of health on the true social capital is underestimated. Finally, the estimate of coefficient ˇ in model (3) of our structural model gives us the positive and proper causal effect of health on social capital confirming the circular and positive association between the two variables. In summary, our findings confirm the evidence of a positive effect of social capital on self-reported health. It is worth noticing that we cannot compare the results in tables and with ones contained in other studies given that they are just the coefficients con the reduced form equations such that they do not have any causal interpretation and they just reflect the amount of social capital and health in equilibrium. However, a look at the first stages of the regressions in Table 4 can give us an assessment of the

Table 4 First stages regression for models (1)–(3). Variable

Health

Ind social capital

Coefficient ***

Male Age Age2 Brncntr Urban Density Married Educyrs Religious Mothereduc Fathereduc Motheremploy Fatheremploy Gdp Gdpgrowth Hospital beds Hospital pers Average health Crime victim Mcrime Household controls Individual controls Parental controls Regional dummies

0.0703 −0.0478*** 0.0000*** −0.0461** 0.0043 −0.0008* 0.0954*** 0.0312*** −0.0033** 0.0042** 0.0009** 0.0052*** 0.0193*** 0.0979** 0.00203 0.0000 0.0002 0.1948*** −0.147*** −0.465*** Yes Yes Yes Yes

Observations

105,166

* ** ***

t-Statistics 9.35 −7.42 7.96 −2.78 2.62 −0.96 8.71 6.85 −1.83 2.29 0.60 1.39 6.99 2.36 0.58 0.49 1.22 10.05 −10.59 −10.32

Coefficient ***

0.0470 −0.0075*** 0.0000*** −0.0242 −0.0205*** 0.0003 0.0216** 0.0288*** 0.0178*** 0.0382** 0.0013** 0.0071*** 0.0202*** −0.0000 −0.0033 0.0003* 0.0004** −0.0334 −0.1021*** −0.231* Yes Yes Yes Yes 105,166

Comm social capital t-Statistics

Coefficient

t-Statistics

7.62 −9.31 8.37 −2.14 −6.22 1.58 3.94 7.79 4.98 2.33 0.78 2.69 7.95 −2.63 −0.88 1.66 2.47 −1.96 −9.25 −1.89

−0.0048 −0.0025 0.0000 −0.0224 −0.0449*** 0.0006 0.0041 −0.0001 0.0004 0.0019** 0.0051** 0.0045** 0.0096** −0.0000 −0.0731*** 0.0096*** 0.0085*** 0.3342*** 0.0053 −4.358*** Yes Yes Yes Yes

−10.02 −0.74 2.41 −2.35 −8.74 2.98 2.88 −0.06 0.08 1.03 7.87 0.99 1.17 −0.37 −9.16 7.58 5.41 9.45 1.29 −10.01

105,166

Significant at 10%. Significant at 5%. Significant at 1%.

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model

ARTICLE IN PRESS

SOCSCI-1211; No. of Pages 9

M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

7

Table 5 Second stages regression for models (8) and (9). Variable

Social capital

Health

Coefficient

t-Statistics

Ind.health (predicted value) Ind.soc.cap (predicted value) Comm.soc.cap Male Age Age2 Brncntr Urban Density Married Educyrs Religious Mothereduc Fathereduc Motheremploy Fatheremploy Gdp Gdpgrowth Hospital beds Hospital pers Average health Crime victim Mcrime Household controls Individual controls Parental controls Regional dummies

−0.2319*

−2.03

0.0382*** 0.0624*** 0.0183*** 0.0001*** −0.0342 −0.0178** 0.0001 0.0424*** 0.0358*** 0.0164*** 0.0048** 0.0098** 0.0079* 0.0248*** −0.0000 −0.0002

3.95 6.26 3.42 4.75 −1.79 −2.63 0.58 3.42 9.73 9.88 3.53 2.98 2.70 4.49 −0.33 −0.22

−0.1345*** −0.1362 Yes Yes Yes Yes

−7.35 −2.18

Observations

105,166

* ** ***

Coefficient

t-Statistics

1.4096*** 0.0370 0.0042 −0.0373*** 0.0001*** −0.0118 0.0345** −0.0013* 0.0662*** 0.0096* 0.0282*** 0.0036** 0.0091** 0.0075* 0.0212*** 0.0000 0.0088* −0.0005 −0.0007* 0.2310***

9.98 1.13 0.49 −9.85 6.82 −0.44 3.10 −2.54 5.41 2.52 8.95 2.30 2.65 2.40 4.91 0.54 2.23 −1.76 −2.03 6.31

Yes Yes Yes Yes 105,166

Significant at 10%. Significant at 5%. Significant at 1%.

quality of our identification strategy and the validity of our exclusion restrictions. We think that our model is identified because at least one of our exogenous variables, the average health in the region, is really significant and with the expected sign in the first stage. In addition, also our instrument for the individual social capital, whether the respondent has been victim of a crime, is relevant. By comparison with the estimated coefficients in the first stages, the results reported in the second stages (Table 5) confirm those obtained in the first stages. Generally, we have obtained that individual social capital increases the probability of being in good health and the same sign is found in the inverse relationship, whereas, health status has a negative effect on the misreporting of individual social capital. However, the results suggest that community social capital affects health in a more shaded way once compared with the individual one, given that it turns out to be insignificant and very small. In particular, we confirm that in regions with a higher social capital, the marginal effect of individual social capital on health can be increases individual trust and increase the probability of being in good health. Moreover, we show that the impact of social capital depends on the community population. However, our results suggest that trusting the others should be easier in small communities, with higher opportunities of repeated interactions and lower costs of monitoring and information acquisition, than in large communities and organizations where the chance of dealing

more than once with the same partner is very little if not negligible. The impact of trust on health is expected to be stronger on individual health in smaller communities, as cooperation is more likely to be achieved and maintained over time, yielding therefore to considerable benefits. Cooperation and trust allows setting up informal institutions based on reciprocity, even among households or extended families rather than only among individuals, which can provide support in case of need. In coherence with the work of Subramanian, Kim, and Kawachi (2002), and Mellor and Milyo (2005), the results suggest that in smaller community levels, social capital has a larger impact although its size is small in absolute terms. On the other hand, the coefficient associated with the interaction between the instrument and the size of the community is negative and significant. Such results can be interpreted in the context of the relative deprivation thesis: the detrimental effect of social capital on health is inversely related to the social conditions of the neighborhoods or region of residence. In large cities, social activities are carpeted to be more developed implying that the expected average level of social capital is higher, and the perception of social capital and its negative effect on health is less accentuated. In both stages, it appears that a higher degree of education has affected the impact of trust on self-reported health. As education appears to be an essential determinant of social participation, policymakers may consider an

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model SOCSCI-1211; No. of Pages 9

ARTICLE IN PRESS M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

8

education policy targeting deprived groups. Increased level of education would lead to healthier lifestyles and greater use of preventive services, which would positively influence the overall community health. Education too has an indirect effect, because it increases the use of social capital which, in turn, improves health. The results confirm that social capital has a moderating effect on the relationship between socioeconomic status and health, leading to greater health improvements among the worse off, compared to the less educated and those in poorer socioeconomic circumstances. However, most aspects of social capital are positively correlated with the education and employment conditions, and social capital appears to partly mediate the effect of education and employment conditions on health. Many of these findings hold in a cross-country perspective, indicating that countries with more social capital tend to have on average better health, although the relationships are clearer in comparison in the countries with similar income levels. The fact that these relationships are found consistently across countries, using a range of alternative measures of health and social capital, confirms the relevance of the findings. Indeed, the consensus emerging from these studies is that individuals with more social capital enjoy a longer, happier and healthier life than their less socially integrated counterparts. While the overall message is that greater social capital goes hand in hand with better health and a healthier lifestyle, the strength of the association between social capital and health differs greatly across studies. Finally, acknowledging that community social capital alters the reporting of the individual, social capital proved to be crucial in producing reliable and sensible results. Actually, this effect is highly significant and large enough to determine the apparently negative independent impact of community social capital that is obtained in models (2) and (3), while community social capital plays no autonomous role. 5. Conclusions In this paper, we investigate the impact of social capital on individual self-reported health for a sample of eighteen countries from the MENA (Algeria, Bahrain, Egypt, Iran, Iraq, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Qatar, Saudi Arabia, Sudan, Syria, Tunisia, Turkey, Yemen), by using instrumental variable estimates the causality in the relationship between social capital and health. To the best of our knowledge, this paper is the first to assess the impact of social capital on health in some MENA countries in the way that it explicitly accounts for measurement error in self-reported trust and for the circular relationship between the two variables. Broadly speaking, we emphasized here the choice of adequate instrumental variables to break down the circular link between social capital and health, and on the simultaneity of the two processes, in order to identify the causal impact of social capital on health. The current study confirms that social capital cannot be overlooked by policy-makers seeking to improve health conditions. However, the findings suggest that policy interventions should be targeted at improving primarily individual social capital. In doing so, they would achieve a

double effect: on the one hand, directly improve individual health, and on the other hand, contribute to community social capital, which reinforces the beneficial role of individual social capital. Exploiting such reinforcing mechanism could improve the cost-effectiveness of policies: an intervention that succeeds in improving the social capital of a large number of individuals in one community would produce a larger health benefit than a one that targets the same number of individuals located in a number of different communities. The potential benefits of adopting a broader perspective appear particularly significant in the case of the MENA countries, where there is an obvious scope for improvement in social capital, compared to other countries in Europe and elsewhere. Further researches should be devoted to theoretically and formally model the role of social capital in influencing individual health. Empirical literature has so far proceeded without guidance from a clear theoretical framework. Additional attention should be devoted to other aspects, apparently unrelated to health, such as the availability of opportunities of social interactions and cooperation and the definition of the institutions able to promote social interaction, credibly enforce the law and order, reduce criminality, and discourage opportunistic behaviors. Authors’ contributions MY was responsible for analyzing and interpreting the data and drafting the manuscript. MC contributed to the study design, and revised the manuscript for intellectual content. All authors have read and approved the final manuscript. Acknowledgments We would like to thank Marta Lagos member in the Scientific Advisory Committee (SAC) for the World Values Survey (WVS) for facilitating access to the social indicators dataset for MENA countries. However, analysis of these data and the opinions expressed in this text are not those of statistics WVS. We would also like to thank the two anonymous reviewers for their input on the initial version of this text. References Acemoglu, D., & Angrist, J. (2000). How large are human-capital externalities? Evidence from compulsory schooling laws from NBER. Macroeconomic annual, 15. MIT Press. Alesina, A., & La Ferrara, E. (2002). Who trusts others? Journal of Public Economics, 85(2), 207–234. Berkman, L., & Syme, S. L. (1979). Social networks, host resistance and mortality. American Journal of Epidemiology, 109, 186–204. Bourdieu, P. (1986). The forms of capital in handbook of theory and research in the sociology of education. New York: Greenwald Press. Brown, G. W., & Harris, T. (1978). Social origins of depression: A study of psychiatric disorder in women. London: Tavistock Publications. Brown, T. T., Scheffler, R. M., Seo, S., & Reed, M. (2006). The empirical relationship between community social capital and the demand for cigarettes. Health Economics, 15(11), 1159–1172. Cobb, S. (1976). Social support as a moderator of life stress. Psychosomatic Medicine, 38, 300–314. Coleman, J. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, 95–120.

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008

G Model SOCSCI-1211; No. of Pages 9

ARTICLE IN PRESS M. Younsi, M. Chakroun / The Social Science Journal xxx (2014) xxx–xxx

D’Hombres, B., Rocco, L., Suhrcke, M., & McKee, M. (2010). Does social capital determine health? Evidence from eight transition countries. Health Economics, 19(1), 56–74. Easterly, W., & Levine, R. (1997). Africa’s growth tragedy: Policies and ethnic divisions. Quarterly Journal of Economics, 112(4), 1203–1250. Folland, S. (2007). Does “community social capital” contribute to population health? Social Science & Medicine, 64, 2342–2354. Ghosh, P., & Ray, D. (1996). Cooperation in community interaction without information flows. Review of Economic Studies, 63, 491–519. Islam, K., Merlo, J., Kawachi, I., Lindstrom, M., & Gerdtham, U. (2006). Social capital and health: Does egalitarianism matter? A literature review. International Journal for Equity in Health, 5(3) http://dx.doi.org/ 10.1186/1475-9276-5-3 Kawachi, I., & Berkman, L. (2000). Social cohesion, social capital and health. In Social epidemiology. New York: Oxford University Press. Kawachi, I., Subramanian, S. V., & Kim, D. (2008). Social capital and health. New York: Springer. Kranton, R. (1996). The formation of cooperative relationships. Journal of Law Economics and Organization, 12, 214–233. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. W. (1997). Trust in large organizations. American Economic Review, 87(2), 333–338. Lochner, K. A., Kawachi, I., Brennan, R. T., & Buka, S. L. (2003). Social capital and neighborhood mortality rates in Chicago. Social Science & Medicine, 56(8), 1797–1805. Lynch, J. J. (1977). The broken heart: The medical consequences of loneliness. Cambridge: Basic Books.

9

Mellor, J., & Milyo, J. (2005). State social capital and individual health status. Journal of Health Politics, Policy and Law, 30(6), 1101–1130. Mohan, J., Twigg, L., Barnard, S., & Jones, K. (2005). Social capital, geography and health: A small area analysis for England. Social Science & Medicine, 60(6), 1267–1283. Olsen, K. M., & Dahl, S. A. (2007). Health differences between European countries. Social Science & Medicine, 64, 1665–1678. Poortinga, W. (2006a). Social relations or social capital? Individual and community health effects of bonding social capital. Social Science & Medicine, 63(1), 255–270. Poortinga, W. (2006b). Social capital: An individual or collective resource for health? Social Science & Medicine, 62(2), 292–302. Putnam, R. D., Leonardi, R., & Nanetti, R. (1993). Making democracy work: Civic traditions in modern Italy. Princeton, NJ, Chichester: Princeton University Press. Scheffler, R., Brown, T., & Rice, J. (2007). The role of social capital in reducing non-specific psychological distress: The importance of controlling for omitted variable bias. Social Science & Medicine, 65, 842–854. Stern, S. (1989). Measuring the effect of disability on labor force participation. Journal of Human Resources, 24(3), 361–395. Subramanian, S. V., Kim, D. J., & Kawachi, I. (2002). Social trust and selfrated health in U.S. communities: A multilevel analysis. Journal of Urban Health-Bulletin of the New York Academy of Medicine, 79(4), 21–34. World Values Survey (2010–2012). http://www.worldvaluessurvey.org/ SebTest/wvs/articles/folder published/survey2010/files/WVS-waves. pdf

Please cite this article in press as: Younsi, M., & Chakroun, M. Does social capital determine health? Empirical evidence from MENA countries. The Social Science Journal (2014), http://dx.doi.org/10.1016/j.soscij.2014.08.008