Zeitschrift für Soziologie, Jg., Heft 5, 2015, Online- Anhang

Transcription

1 I Are Joiners Trusters? A Panel Analysis of Participation and Generalized Trust Online Appendix Katrin Botzen University of Bern, Institute of Sociology, Fabrikstrasse 8, 3012 Bern, Switzerland; katrin.botzen@soz.unibe.ch Content Table A1: Variables in the SHP Table A2: Descriptive Statistics of Variables A5.1: Additional Findings on Participation and Trust Analyzed with OLS and FE Regression Table A3: OLS and FE Models of Trust on Participation A5.2: Additional Statistical Specification Tests of the Dynamic Panel Models

2 2 Table A1: Variables in the SHP Variable General trust in people Participation in groups Yearly income (ln) Operationalization Would you say that most people can be trusted or that you can't be too careful in dealing with people, if 0 means "Can't be too careful" and 10 means "Most people can be trusted"? ( Würden Sie sagen, dass man den meisten Menschen vertrauen kann, oder kann man im Umgang mit anderen Menschen nie vorsichtig genug sein? 0 bedeutet, dass man im Umgang mit anderen Menschen nie vorsichtig genug sein kann, 10 bedeutet, dass man den meisten Leuten vertrauen kann. ) Do you take part in clubs' or other groups' activities, religious groups included? yes = 1 ( Machen Sie mit in Vereinen oder in anderen Gruppen, auch in kirchlichen? ) Yearly total personal income, net, logarithm Working Status Working status in year of interview: active occupied = 1 Education in years Highest level of education achieved, grid + individual 11 codes, converted in years Divorce Civil status in year of interview: separated/divorced = 1 Health How do you feel right now? very well/well = 1 Children Can you tell me if you have children, living with you or not? yes = 1 Source: Swiss Household Panel Table A2: Descriptive Statistics of Variables Variable Mean Std.Dev. Min Max Observations General trust in people N Observations = N Groups = 3276 Participation in groups N Observations = N Groups = 3277 Yearly income (ln) N Observations = N Groups = 3084 Working Status N Observations = N Groups = 3281 Education in years N Observations = N Groups = 3708 Divorce N Observations = N Groups = 3713 Health N Observations = N Groups = 3281 Children N Observations = N Groups = 3281 Source: Swiss Household Panel

3 3 A5.1: Additional Findings on Participation and Trust Analyzed with OLS and FE Regression This paragraph presents additional empirical findings in an OLS and FE regression context. To keep in mind, as OLS and FE suffers from endogeneity bias if lagged dependent variables are implied, these findings had to be taken under reserve. Nevertheless, they are provided to be put into context of former analyses in social capital research. As previously applied, time dummies are included in the analysis. The results in Table A3 demonstrate a significant and positive relationship between participation status and general trust. If a person participates in clubs or other groups in the current year, and participated in the previous year, this positively contributes to the level of generalized trust in people. The effect is stable and remains significant, even if the model controls for previous levels of trust. The dependent variable lagged twice is a significant explanatory variable, and shows that the trust level of previous years significantly correlates with the current trust level. As the FE approach is a within estimation, the negative coefficients of dependent variable lagged twice describe a mean reverting behavior of the individual in the long- term. The positive dummy variables of the years show an increase of the trust level with regard to the reference year 2004.

4 IV Table A3: OLS and FE Models of Trust on Participation OLS FE (1) FE (2) DV: Generalized trust in people Participation in groups *** * (0.000) (0.070) (0.012) L1.Participation in groups *** ** * (0.000) (0.009) (0.018) L1.General trust in people *** (0.000) L2.General trust in people *** (0.000) Dummy: Year *** *** *** Dummy: Year *** *** *** Dummy: Year *** *** *** Dummy: Year *** *** *** Dummy: Year *** *** *** Dummy: Year *** *** *** Dummy: Year *** *** *** Dummy: Year *** *** *** Constant *** *** *** N Observations Notes: Swiss Household Panel ; p- values in parentheses; + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001; standard errors are clustered

5 V A5.2: Additional Statistical Specification Tests of the Dynamic Panel Models To reach more confidence in the inference from the DPM in section 5, the models are re- estimated with different statistical modifications. With respect to statistical robustness, the first step provides the results of the Arellano- Bond- estimates, the so- called Difference- GMM (D- GMM) estimates in addition to the S- GMM results. The estimated coefficients for the L1.Participation in groups variable of the two- step D- GMM regressions are.133 *.144 * and.158 ** in the full models 2 to 4, which shows that the results remain similar in significance and meaning. Secondly, the regressions are estimated with the one- step S- GMM. Previously, researchers obtained estimates in a one- step estimation process, which assumed homoscedasticity of the error terms because even under heteroskedasticity the two- step estimator showed the disadvantage of downward biased standard errors (Bond 2002). But as Windmeijer (2005) proposes a finite sample correction, which is also used in the models presented here, the two- step GMM provides accurate estimates. Nevertheless, the one- step S- GMM estimators yield similar estimates as the two- step procedure (.127 +,.148 ** and.160 ** for the full models 2 to 4). Moreover, for the third adjustment, the DPMs number of instruments are truncated. In general, the use of more instruments than parameters estimated should lead to efficient estimates; however, if the requirements are not satisfied sufficiently, the number of instruments can be restrained which, in turn, should result in reliable estimates (Wawro 2002: 44). Therefore, the estimates robustness is tested with fewer instruments, as they are truncated to two and three lags. Likewise with the other tests, results confirm the interpretation of the main analyzes.

6 VI Bibliography Bond, S., 2002: Dynamic Panel Data Models: A Guide to Microdata Methods and Practice. CeMMap working papers No. CWP09/02, Centre for Microdata Methods and Practice, Institute for Fiscal Studies. Wawro, G., 2002: Estimating Dynamic Panel Data Models in Political Science. Political Analysis 10: Windmeijer, F., 2005: A Finite Sample Correction for the Variance of Linear Efficient Two- Step GMM Estimators. Journal of Econometrics 126: