Confidence in Institutions and Social Trust in Spain - European ...

2 downloads 0 Views 309KB Size Report
major role that institutions play as creators of social trust (Paxton, 1999; Hardin, 2002;. Rothstein ..... compounded by the impact of vote in last national elections, as individuals who voted ...... Princeton, New Jersey: Princeton University Press.
Confidence in Institutions and Social Trust in Spain (1980-2005)

Paper prepared for the ECPR Joint Session, May 2007, Helsinki, Finland. Workshop on “Social Trust, the State and Diversity”.

Martiño Rubal Carmen Voces Mónica Ferrín Miguel Caínzos Contact: Martiño Rubal investigacion04.egap@xunta,es 0034 981 546 334 Escola Galega de Administración Pública Rúa de Madrid, 2-4, 15707 Santiago de Compostela, Galicia, Spain.

1

1. INTRODUCTION Recent literature on social capital has emphasized that social capital can not be considered apart from institutions (Levi, 1998; Hall, 1999; Stolle, 2002; Rothstein & Stolle, 2002; Rothstein, 2005). More specifically, this so-called institutional approach has highlighted that institutions (governments, parliaments, politicians, political parties agencies which implement some policies, i. e. health system, education system) play an important role in promoting social trust. This paper deals exclusively with what is commonly regarded as one of the several components of social capital: social trust1. Many factors have been stressed in the literature as determinants of social trust (see Freitag, 2003, and also Welch and others, 2005). It is said that personal characteristics of individuals, either personality traits, particular individual moral orientations acquired early in life (Uslaner, 2002), or other beliefs as religion and stable political orientations (as political interest or ideology) influence trust and confidence. Other determinants refer to social environment: belonging to a certain community with certain characteristics (Paxton, 2002, Uslaner, 2002) or being part of informal social networks. Other important factors commonly emphasized by literature are political participation and civic engagement, mainly associational membership (Brehm y Rahn, 1997; Putnam, [2000] 2002; Stolle, 2001; Newton, 2001; Uslaner, 2002; Herreros, 2004; but see Zmerli et al., 2007). Other factors are related to personal resources of individuals (such as education) (Putnam, [2000] 2002; Freitag, 2003). Finally, most recent research has put the accent on the major role that institutions play as creators of social trust (Paxton, 1999; Hardin, 2002; Rothstein & Stolle, 2002; Herreros, 2004; Rothstein, 2005). This presumed importance of institutions invites us to guess that short-term political factors could have a non negligible indirect influence in social trust through their impact on institutional confidence2. Recent research has identified a handful of institutions as cornerstones—together with other factors—of the creation of social trust. Institutions held responsible for the implementation of universal policies are said to contribute most to the feeling that “most people can be trusted” (Rothstein & Stolle, 2002). According to this view, Police, Army,

1

Anyway, as Welch and others have stated, it is difficult to distinguish those factors which contribute to the formation of social capital from those that influence the formation of social trust (Welch and others, 2005). 2 In this paper we will not intend to review in a systematic way the literature about social trust or about confidence in institutions. For this purpose, see Rubal et al. (forthcoming).

2

Courts, Health care System and Educational System are part of these institutions, and differ from partisan “political/biased” institutions (Parliament, Government, Political Parties) or “checking/control” institutions (Newspapers, T.V., Civil Service)3. This group of authors assumes that implementation institutions have socialization effects, as people in contact with them internalize messages of fairness, equality etc. that increase trust in other people (generalized trust). There is, from this point of view, a positive relation among confidence in certain institutions and social trust.

2. OBJECTIVES, DATA AND MEASUREMENT We think that, if we take it seriously, the argument that institutions promote social trust must be supplemented by two additional ideas. First, whether institutions in general have this capacity or not should depend on their own perceived trustworthiness. Secondly and closely related to the former, if institutions do really differ in their trustgenerating power, they should also differ in their trustworthiness. With these ideas as a starting point we will set up an analysis of the relationship between –and the determinants of- institutional confidence and social trust that will be developed in three steps. Firstly, we contrast the hypothesis that citizens can distinguish among different types of institutions when expressing their confidence, for a country (Spain) and a period of 25 years (from the beginning of the democratic period until recent time). Secondly, we try to examine the determinants of the institutional confidence in Spain, considering the dimensionality of the confidence (whether there is more than one group of institutions). Finally, we extend our analysis to another data set, incorporating more variables to our initial model, in order to explore the direction of the relationships among variables. In addition, we test the model in other five European countries to examine the similarities and differences with Spain. Obviously, the extent to which we will be able fulfil these objectives will be conditioned by the availability of data. We will attempt to overcome the shortcomings of the available data sources by using three different data sets. A. Do citizens distinguish among institutions when they express their confidence? Our first objective in this paper is to answer this question, for the Spanish case.

3

Although other typologies of institutions have been proposed in the literature (see for example Denters et al., 2007), we will take this taxonomy as a point of departure, and compare it with the results from previous work on the Spanish case.

3

Objective1: to contrast the hypothesis that citizens discriminate between different types of institutions when expressing their confidence. For this purpose, the study of the dimensionality of institutional confidence during the last 25 years in Spain, we carry out an individual level analysis of longitudinal data, using information from 1980 to 20054. We compare the three dimensions of institutional confidence established by Stolle and Rothstein (Rothstein & Stolle, 2002), with a twodimensional distinction we have obtained in a previous work, through a Confirmatory Factor Analysis for each year (1980, 1990, 1996, 2001 and 2005), in order to observe which of these models fits better to data. We think that the case of Spain is especially interesting in this regard, because, at the beginnings of the period we study, Spain completed a successful transition to democracy. This gives special relevance to the longitudinal perspective, as it will allow us not only to verify the existence of a distinction between the institutions, but also to detect changes between a time when democracy was still emerging (1980) and later periods, in which political life took place in a fully consolidated democratic setting. B. The second objective we come through is the study of the determinants of institutional confidence in Spain at the individual level during the period 1996-2005. The results of our first objective will be conditioning the number of dependent variables which will be considered in this second stage. Objective2: to study some of the determinants of confidence in institutions for the Spanish case, taking into account its dimensionality. In this second step of our work, we were forced to restrict the data set previously used, because the surveys of 1980 and 1990 did not contain enough information about the alleged

determinants

of

institutional

confidence.

Therefore,

we

use

three

Latinobarometer surveys (1996, 2001 and 2005) from the Centro de Investigaciones Sociológicas. In this second step, we run multivariate regression analyses addressed to identify the determinants of confidence in institutions.

4

We use data from the World Values Survey (1981 and 1990) and from Centro de Investigaciones Sociológicas (Barometer May 1990, code number 1871; Latinobarometer July 1996, code number 2218; Latinobarometer May 2001, code number 2417 and Latinobarometer October 2005, code number 2620).

4

C. There has been much quarrel about the existence of a relation between social trust and institutional confidence in the literature. Some authors detect a relation at the aggregate level (Newton, 2001) or both at the aggregate and individual levels of analysis (Hall 1999; Rothstein & Stolle, 2002). Others notice a reciprocal relation among these two types of trust (Brehm and Rahn, 1997) or even do not find any relation at all. Our hypothesis establishes a relation among confidence and trust, although the direction is not clear. One of the main purposes at this stage of our work is to elucidate the direction of this relationship. Besides, we will investigate the relation of institutional confidence with other determinants stressed in the literature. At the same time, in this phase, we try to broaden the scope of the analysis by increasing the number of independent variables formerly employed in the regression model. Thus, we will compare our model for Spain with other countries, in order to discover similarities and differences between them, being Spain the reference5. Objective3: to examine the relation among institutional confidence and other variables (including social trust) and the direction of the relation between confidence in institutions and social trust for the Spanish case, proposing a model which we replicate for other European countries. For this purpose, we will use data from the first wave of the European Social Survey (2002/3). The methodology is based on the Structural Equation Model, and we make an analysis of invariance of the model, comparing Spain with five countries: France, Germany, Italy, Great Britain and Sweden. In steps two and three of our work we have included five different types of variables. The first one encompasses variables with a political component. Within this block we can distinguish four subgroups. The first one concerns evaluation of -and satisfaction with- the political and economic context (satisfaction with the government, satisfaction with the present state of the economy and evaluation of current working of some welfare policies). A second subgroup is formed by variables linked with orientations which are usually assumed to be relatively stable, either value orientations (i. e. political ideology) or motivational orientations (i. e. interest in politics). The third subgroup refers to exposure to information about politics and links the two previous groups of variables. Finally, there is a set of political variables associated with political 5

There are contextual factors that influence both social trust and institutional confidence. Our intention is not to create a model for each country, but just to find differences with the Spanish model.

5

behaviour (voting and political participation). The second block of variables that we have taken into account measures different socio-demographic features of individuals (age, sex, education). The third block includes a single variable: satisfaction with life. Beyond its undeniable dependence on the external context, we assume that this variable has a stable component related to optimism and a feeling of control over one’s own life; hence, we understand that it measures not only a transient state of mind or mood but also a steady personality trait. The fourth block is one particularly relevant according to the main thrust of the literature about social capital and social trust, and has to do with sociability, both of a formal (participation in associations) and an informal kind (relations with friends, relatives and workmates). The last block of variables is related to religion6.

3. EMPIRICAL RESULTS

3.1 Study of the dimensionality of confidence in institutions Recent literature has stressed the necessity of further elaboration of the concept of institutional confidence. One of the most influential formulations distinguishes among representative, implementation and control institutions. According to this view, institutions with implementation functions are the most important in promoting social trust, due to some of the characteristics they share (transparency, fairness, universalism, impartiality) and their ability to exert a socialisation role. We assume as a working hypothesis the possibility of the existence of different kinds of institutions. But, instead of taking for granted any typology, our first goal in this paper is to empirically adjudicate between two different classifications built upon the same set of institutions

(Political

parties,

Parliament,

Government,

Civil

Service,

Courts,

Newspapers, TV, Church, Army and Police). On the one hand, the typology suggested by Rothstein and Stolle differentiates three sets of institutions (Model A, Figure 1). On the other hand, an alternative distinction, based on our previous work, separates two groups of institutions (Model B, Figure 1). For this purpose, we have compared the whole fit of both models to survey data in Spain through confirmatory factor analysis.

6

For more details, see Appendix 1.

6

1 Political/ Biased Institutions

Parties

Parliament

Government

1

1

1

e1

Parties

e2

1

e3

Government

Polítical Institutions 1

Courts

e6

1

Army

Police

1

1

e6 1

e7

e8

T. V.

1

e10

Civil Service

Newspapers

e5

1

e8

e9

Church

Control Institutions

e3

e7

1

1

e2

1

Newspapers 1

e1

1

Civil Service

Courts

1 Neutral/ Order Institutions

Parliament

1

1

T. V.

Model A. Stolle & Rothstein, 2002

1

e11

e12

Social Order Institutions

Army

Police

Church

1

1

1

e9

e10

e11

Model B. Caínzos et al., 2006

Figure 1. Two models of the dimensionality of confidence in public institutions.

The model proposed by Rothstein and Stolle7 differentiates three dimensions. The first one is formed by political/biased institutions, that is, representative institutions with elected officials: Political Parties, the Parliament and the Government. The second group is formed by “neutral/order” institutions (the Courts, the Army, the Police; we have added a fourth institution to this group – the Church – because, at least in the Spanish case, it appears to be similar to the institutions belonging to this dimension). And the last dimension, made up by actors with a control power, is formed by Civil Service, Newspapers and Television. Instead, our model distinguishes two different dimensions of the confidence in institutions8. On the one side are those institutions which play an important role in democratic political life, either as representatives of citizens (political parties, the 7

This model dropped out of two different analyses: one at the macro level with 50 countries, and the other at the micro level for Sweden. In both cases, the method used was Principal Component Analysis and results were quite similar (Rothstein & Stolle, 2002). 8 In a previous paper (Caínzos et al., 2006), as a preliminary step to an age-cohort-period analysis of the evolution of social trust and political confidence in Spain, we executed a Principal Component Analysis based on a pooled data set which merged data from several surveys covering the period 1980 to 2005. Two distinct components resulted: one with a political character, and the other with a social order character.

7

Parliament and the Government), as external control agents (newspapers and television), or as internal control agents (Civil Service and Courts). On the other side, the second dimension is made up of social order institutions: the Army, the police and the Church. χ2

df

χ2/df

GFI

AGFI

NFI

CFI

RMSEA

Model A

613.65

12

51.14

.917

.805

.906

.908

.153

Model B

332.41

13

25.57

.958

.910

.949

.951

.107

Model A

772.03

24

32.17

.865

.748

.758

.762

.162

Model B

792.22

26

30.47

.860

.758

.751

.757

.157

223.81

7

31.97

.974

.921

.986

.987

.104

69.17

8

8.65

.992

.978

.996

.996

.052

Model A

880.18

32

27.51

.920

.862

.919

.922

.103

Model B

675.67

34

19.87

.945

.910

.938

.941

.087

Model A

628.55

32

19.64

.948

.911

.945

.947

.086

Model B

1447.58

34

42.58

.890

.823

.873

.875

.129

Model A

697.88

32

21.81

.954

.920

.928

.931

.083

Model B

867.84

34

25.53

.948

.916

.911

.914

.090

1980

1990

1990





2

Model A Model B 1996

2001

2005

3

3

3

1

2

Sources: World Values Surveys (measure: 1 to 4); Barometer, Centro de 3 Latinobarometer, Centro de Investigaciones Sociológicas (measure: 0 to 10); Investigaciones Sociológicas (measure: 1 to 4).

Table 1. Goodness of fit statistics for two models of the dimensionality of confidence in public institutions. We tested the two models for 1980, 1990, 1996, 2001 and 2006 in order to encompass a wide period9. Surveys employed were mainly chosen by two reasons: the availability of the questions about institutional confidence, which appear for the first time in 1980 – it reappears again in 1990; and the selection of a sufficient distance between the surveys, which would allow to trace possible changes over time (to examine the available institutions for each of the selected years, with their means and typical deviations, see appendix 2). Although there are differences in the results obtained for each year, we can find a similar pattern for some of them. If we focus our attention on global indexes of goodness of fit, as displayed in table 1, we could say that our model achieves better values for three surveys (1980, 1996, and the second of the surveys we have used for 9

See footnote 4 for more details about the surveys used in this analysis.

8

1990), whereas Rothstein and Stolle’s model is more successful for the 2001 and 2005 surveys. Nevertheless, this general assessment of the models must be nuanced when we also examine the values of the estimates parameters for the different models, and not only to the goodness of fit statistics (see appendix 3). Survey data collected in 1980 are more clearly characterized by the existence of two dimensions (political and social order institutions), as shows the fact that in the two indexes of global goodness of fit considered here (GFI and AGFI) model B achieves values above 0.90 and the Root Mean Square Error of Approximation (RMSEA) borders its admitted value (0.10). Better results for this model are also observed in the Normed Fit Index (NFI) and the Comparative Fit Index (CFI), very close to 0.95. Model A performs quite worse for all these indexes. Summing up, it seems to be more appropriate for this year to maintain a two-dimensional solution. Both models proved to fit very poorly to the data of the first survey that we used to contrast their validity in 1990. According to the values from table 1, none of the models achieves fit values within the recommended range of acceptability. Therefore, to solve this weakness, which could be due to some peculiarity or anomaly of this survey, we resorted to a different data source for this year10. Although this new survey has its own deficiencies — it includes only 6 from the initial sample of 10 institutions —, both models exhibit an excellent global fit, even though results are better for model B. Nevertheless, if we check the correlation among the latent variables (see appendix 3), we can observe an extraordinary overlap between factors, what points towards a onedimensional solution as the best description of this sample data. For the 1996 data, both the goodness of fit statistics and the adequacy of parameters estimates indicate the convenience of maintaining model B as the more valid approximation to the components of confidence in institutions. Once again, data show that the dimensionality is lower than that proposed by Rothstein and Stolle. Data from 2001 reveal Model A to be the one with best fit. Nevertheless, correlation between the political/biased and neutral/order factors is over 0.90, what reflects that the distinction among both types of institutions is quite blurry. This makes us think that

10

In this survey (a barometer of the Centro de Investigaciones Sociológicas), confidence in institutions was measured using an eleven points scale (from 0 to 10), instead of the scale of 1 - 4 of WVS and Latinobarometer. Moreover, the sample is larger in this study.

9

it is convenient to be cautious about the viability of the three-dimensional solution, even if it offers a good fit to data. Finally, Model A successes with data from 2005. However, in this case we must notice that correlation between political/biased institutions and neutral/order institutions falls outside the admissible range, surpassing value 1.0. These results lead us again to reject the existence of three dimensions and throw us back to two-dimensionality. Concluding, dimensionality of confidence in institutions in Spain seems to be restricted to one or two dimensions for all the years we analysed. Even if, at first sight, a model of higher dimensionality seems to have a better fit for some of the samples, this view is disconfirmed by an inspection of the parameters of that model, which point towards the preferability of a simpler solution. It is worth to draw attention to the fact that this low dimensionality of institutional confidence is present all along a time-period through which Spain suffered dramatic political and social changes. More specifically, we find one- or, more commonly, two-dimensionality both in a moment of early development of democracy and at a time when citizenship had a relatively long experience of living under normal democratic political conditions. Hence, it does not seem that low dimensionality of political confidence should be seen as an ephemeral phenomenon, but rather as a stable feature of the structure of Spaniards’ political attitudes. The preferred solution reduces institutional confidence to two factors. On the one hand, confidence in social order institutions, which encompasses attitudes towards three institutions: the army, the police and the church; the loadings of the army and the police are consistently high along the whole time period, while the church presents lower values after 1990. On the other hand, confidence in political institutions; under this heading are subsumed government, parliament, political parties, courts, civil service, newspapers and TV. Three of them have persistently high scores (government, parliament and political parties), while other two have high loadings on this dimension in most years (civil service, with the only deviation of 2001; and courts, with the exception of 2005); newspapers and TV are the items which work worse in all years.

3.2 The determinants of confidence in institutions Once we have knowledge of the existence of two types of institutions through factor analysis, we are now able to study the elements that influence levels of institutional

10

confidence in Spain. Moreover, considering both dimensions – social order institutions and political institutions – will permit us to find out whether there are determinants that are more relevant for one or another type of institutional confidence. That is, we will try to establish which are the elements that promote confidence in social order institutions as well as those that explain confidence in political institutions. In this step of our work, we have been forced to restrict the scope of our data set, because several of the surveys we used in our dimensionality analysis do not include the kind of explanatory variables which have been given saliency in the literature on the determinants of institutional confidence. For this reason, we will only use data from 1996 (Latinobarometer, CIS 2.218), 2001 (Latinobarometer, CIS 2.417) and 2005 (Latinobarometer, CIS 2620) surveys. They incorporate some important variables for the analysis of social and institutional confidence, even if they are not wholly satisfactory, because they lack information on such relevant aspects as political participation and associational involvement. By analysing data from these years, we cover a time period characterized by consolidated democratic institutions, after almost twenty years of democracy. Nevertheless, this period presents substantial diversity in terms of the prevailing economic and political conditions, which may be of interest in order to check whether changes in political climate alter the level of institutional confidence and, above all, whether they modify the relationships between attitudes towards institutions and our independent variables.

Year

N

Mean

St. Dev.

Stat. sign. of differences between the means Year

P