26
Is Social Capital a Factor? Keming Yang Department of Sociology University of Reading United Kingdom Email: [email protected] QMSS Conference Prague, June 2007

Is Social Capital a Factor? Keming Yang Department of Sociology University of Reading United Kingdom Email: [email protected]@reading.ac.uk QMSS

Embed Size (px)

Citation preview

Is Social Capital a Factor?

Keming YangDepartment of Sociology

University of ReadingUnited Kingdom

Email: [email protected]

QMSS Conference

Prague, June 2007

Overview

1. Tackling social capital’s multi-dimensionality: contrasting the factor approach and the index approach;

2. Confirmatory factor analysis;

3. An index of social capital;

4. The distribution of index scores across European countries;

5. Discussions.

Social Capital: Two Issues

1. Individual or aggregate? Dealt with in another paper;

2. Multiple dimensions: trust, social networks, civic engagement, etc.

• Question: How should we deal with the multi-dimensionality in statistical analysis? Factor, or index?

Multidimensionality and Quantification

Putnam: social capital ‘is stubbornly resistant to quantification’ because ‘social capital is multidimensional, and some of those dimensions themselves are subject to different understandings’ (2002: 11-12).

But, in his books, there are many indices, statistics, and graphs.

Clarifications

1. Statistics (quantification) are allowed, but on proxy measures, not directly on social capital.

2. Each proxy indicator is quantifiable, but it does not make sense to sum them up (oranges and apples).

Explication

• Social capital is not directly measurable but can be assessed and inferred via the measurement of its observable proxies social capital is a factor (latent construct).

• Dimensions of social capital are qualitatively different and therefore not subject to mathematical operation it is not legitimate to create any index for social capital.

Demands for Index

• Practical: governments, international organizations, charities, etc., need a simple measure of social capital to see the effects of policies and programmes;

• Theoretical: the amount of social capital: how much does an individual (community, nation, etc.) have?

The Factor Approach:The One-Factor Confirmatory Model

Social Capital

V1 V2 V3 V4 V5 V6 V9V8V7

e1 e1e1e1e1e1 e1e1 e1

The Factor Approach:A Second-Order Confirmatory Model

Social Capital

Trust Networks Civic Orgs

V1 V2 V3 V5 V6 V7V4 V8 V9

e1 e2 e3 e4 e5 e6 e7 e8 e9

The Index Model

Social Capital

V1 V8V7V6V5V4V3V2 V9

Kenneth Bollen: ‘cause indicators’ vs. ‘effect indicators’ (Babbie, 2007: 158)

Data and Variables• Data: ESS (round1, 2002)• Variables:

– Trust1. People trusted? (11 point scale)2. People fair? (11 point scale)3. People helpful? (11 point scale)

– Social network1. Being social (7 point scale)2. Relative social participation (5 point scale)3. Importance of friends in life (11 point scale)

– Civic engagement1. Index of level of engagement in interest-based voluntary

organizations (transformed into 11 point scale)2. Index of level of engagement in influence-based voluntary

organizations (transformed into 11 point scale)3. Importance of VO in life (11 point scale)4. Importance of being active in VO for a good citizen

Polychoric Correlation Matrix(Total effective n=36589)

Trust Fair Help Social Activity Friends ImpVO VOCtzn IntrstVO InflVO 1.0000.595 1.0000.488 0.517 1.0000.135 0.137 0.117 1.0000.146 0.128 0.107 0.357 1.0000.129 0.154 0.152 0.247 0.215 1.0000.062 0.042 0.048 0.102 0.175 0.206 1.0000.050 0.035 0.036 0.073 0.126 0.180 0.565 1.0000.236 0.212 0.196 0.215 0.313 0.131 0.273 0.190 1.0000.246 0.202 0.180 0.122 0.210 0.076 0.163 0.123 0.514 1.000

One-Factor: All countries

Chi-square=40365.76 (df=35);

P<0.00001;

RMSEA = 0.177

One-Factor Model: Austria

Chi-square=3449.38 (df=35);

P<0.00001;

RMSEA = 0.215

Second-Order Factor Model: All countries

Second-Order Factor Model: Austria

The Second-Order Model: Results by country

Country Effective n chi-square p RMSEA

Austria 2103 851.22 <0.00001 0.110

Belgium 1899 824.95 <0.00001 0.114

Denmark 1506 257.93 <0.00001 0.068

Finland 1956 592.69 <0.00001 0.095

France 1461 259.50 <0.00001 0.070

Germany 2873 493.89 <0.00001 0.071

Greece 2403 570.13 <0.00001 0.084

Hungary 1513 356.36 <0.00001 0.083

Ireland 1899 699.43 <0.00001 0.105

Israel 2309 1002.16 <0.00001 0.115

Note: Czech and Switzerland VO information missing; df=32.

The Three-Factor Model: Results by country

Country Effective n chi-square p RMSEA

Italy 1109 543.69 <0.00001 0.120

Luxembourg 1359 558.01 <0.00001 0.110

Netherlands 2289 437.85 <0.00001 0.074

Norway 2020 550.10 <0.00001 0.090

Poland 1913 412.66 <0.00001 0.079

Portugal 1300 547.22 <0.00001 0.111

Slovenia 1440 416.48 <0.00001 0.091

Spain 1494 597.86 <0.00001 0.109

Sweden 1911 505.38 <0.00001 0.088

United Kingdom 2024 668.42 <0.00001 0.098

From Factor to Index1. Neither the one-factor model nor the

second-order model fits the data; 2. Three forms of social capital are weakly

correlated (not reported here);3. No evidence supports the existence of

social capital as a latent construct; an umbrella abstraction;

4. Creating an index by treating observed variables as cause indicators rather than effect indicators.

Rationales of Making an Index of Social Capital

1. It may not make sense to add apples and oranges, but it does make sense to measure fruit intake.

2. ‘Individual social capital’ lends conceptual support.

3. Allow analysis of the distribution of social capital across individuals, communities and nations.

4. Governments and organizations need a simple and cumulative index.

A Simple Cumulative Index

Sum of weighted (w) total of each respondent’s scores of i components (c).

1. Ad hoc due to the lack of commonly accepted theories; for this study, items in ESS (i=10);

2. No weights applied except for level of engagement in voluntary organizations;

3. Scales with more points more likely to have higher scores, so all scores with fewer points (‘meet socially’ (7) and ‘social activities’ (5)) are transformed into a scale of ten.

4. Range: 0.34 to 100.

i iSC wc

Polyserial Correlations (Index and each indicator)

Indicator PS Correlation

People trusted? 0.53

People fair? 0.54

People helpful? 0.50

Socially meeting 0.38

Social activities 0.50

Importance of friends in life 0.44

Importance of VO in life 0.62

VO for good citizen 0.54

Interest-based VO 0.61

Influence-based VO 0.59

Ranking of Social Capital in EuropeCountry Mean SE

Denmark 56.63 9.85

Netherlands 56.63 9.85

Finland 52.96 9.57

Norway 52.96 9.57

Ireland 52.94 10.56

Sweden 52.94 10.56

Austria 50.3 11.83

Italy 50.3 11.83

Israel 50.1 11.18

Ranking of Social Capital in EuropeCountry Mean SE

United Kingdom 50.1 11.18

Belgium 49.55 11.26

Luxembourg 49.55 11.26

France 48.17 10.94

Poland 48.17 10.94

Germany 47.97 11.12

Portugal 47.97 11.12

Greece 41.89 10.53

Slovenia 41.89 10.53

Hungary 38.71 11.73

Spain 38.71 11.73

Observations• Overall and on average, people in Northern

Europe have higher level of social capital, followed by the West, the South and the East;

• Northern Europe also has the smallest variation of social capital;

• Van Oorschot, Arts and Gelissen (2006) found a similar pattern without making an index;

• Ignore specific components;• Reliability of the index needs to be established,

perhaps by accuracy of prediction.

What’s Next?

1. Consensus on the components and their corresponding weights in the composite index of individual social capital;

2. Theories and empirical research for explaining the level of social capital at the individual level;

3. Relative distribution analysis of social capital index across groups;