Satisfaction with standard of living influence on political trust

Analysis paper Social trust/ satisfaction with standard of living influence on political trust Summary This paper is issued with the trend to attribute political trust to interpersonal trust or satisfaction with the standard of living in the European countries specified later on. For this purpose, we use the variables drawn from Round 3 of European Social Survey (ESS) datasets. Our first step is defining the problem which is thoroughly explained in the introduction. Then we proceed with the data screening with the help of SPSS, search for missing values, outliers, non-normality and check that the responses are in a restricted range. We use factor analysis as our first multivariate technique in order to summarize our six variables concerning political trust so that they could be viewed for what they represent collectively. We get one factor which is later used in our regression analysis as a dependent variable. As we want to compare whether interpersonal trust or satisfaction with the standard of living has a greater impact on the political trust, we alternately use them as an independent variable. To have a better understanding, what else could influence political trust or has a stronger power in explaining the political trust than the previously chosen exogenous variables, we perform regression analysis with some other variables. The results tell us that both interpersonal trust and satisfaction with standard of living have a relatively low weight on the overall political trust (but the highest comparing to alternative variables), however without them, our dependent variable becomes negative which makes perfect sense. We also encounter a rather interesting case while looking at how much interest a person has in politics impacts his or her political trust. The more an individual follows the politics, the less he or she trusts in it. The research is being explained in details in the further text with assistance of graphs and our interpretations of the subject matter. Introduction Without question, political trust is crucial to democracy because it links ordinary citizens like us to the institutions, such as European Parliament, that are intended to represent our interests, thus, reinforcing both the legitimacy and the effectiveness of democratic government. Yet, measuring political trust is a controversial issue in the social sciences. In our analysis, we rest upon works of Luhmann, 1979; Paxton, 1999 and Nooteboom, 2002 stating that theories about the origins of trust can be distinguished along macro and micro theories. One of hypothesis of micro theories, called individual socialization in the scientific literature or, simply, social trust, applied in our project, says that political trust varies within and across countries according to individuals’ trust in others. People who trust each other are more likely to cooperate with each other in forming both formal and informal institutions such as music bands or community associations. In other words, interpersonal trust, learned by an individual early in life, much later is imposed on political institutions. On the contrary, macro theories emphasize that political trust, however, is based mainly upon second hand information about performance of political leaders. We decided to look at respondent’s satisfaction of standard of living in determining political trust, as it can be a wide measure of government performance. Standard of living is both material and intangible and varies from ones access and quality of health care, income growth and educational standards to ones spare time, safety, social life, physical health or environmental quality. If governmental policies tend to be successful, trust in politics is higher than if they fail, since people are more satisfied with the standard of living. As we already mentioned the controversy of political trust foundation, it is no surprise that the above theories are being challenged. We are going to provide the reader with an example. The scientist K. Newton claims that there is not a close and consistent link between social and political behavior. He states that there are minor fluctuations of social trust over the period, whereas, political trust, in most of the cases, can not be predictable in the longer period. In fact, when there are major dislocations in society with fundamental changes in social and political institutions, such as have occurred in post-Communist societies, then political trust will be relatively volatile, however, it does not imply changes in the trust among humans. They say in the theory everything is possible, still we will try to find the best solution being able to explain political trust using the methodological methods. Data Description The secondary data sets which will be used during our entire research project have been withdrawn from Round 3 of European Social Survey (ESS) in 2006/2007. It is a biennial multi-country social survey covering over 30 nations where the first round was fielded in 2002/2003, the second in 2004/2005. The core module of ESS aims to monitor change and continuity in a wide range of social variables, political interest and participation, socio-political orientations, governance and efficacy, moral, political and social values, social exclusion, national, ethnic and religious allegiances, well-being, health and security, demographics and socio-economics. The ESS project is directed by a Central Co-ordinating Team led by Roger Jowell at the Centre for Comparative Social Surveys, City University, London and it is designed and carried out to keep exceptionally high standards. Furthermore, the ESS data set of Round 3 is collected from at a single period of time and is called cross-sectional data. Since this kind of data is particularly suitable to test linear regression models, this analysis will be executed during the research. The data sets we will use contain the data from five countries: Austria, Ireland, Norway, Portugal and Ukraine. In our research project, we have decided to take those variables from the survey’s data set which the best reflect our chosen model and the ones which will be the most relevant during the linear regression analysis procedure. Since we want to make a research about how the social issues such as trust in people and satisfaction about the standard of living influence the individuals’ trust in various political institutions, we have chosen the variables from two offered data sets: Work and Politics and also Family Life and Society. The following are description of the variables: • Dependent variables. trstprl – a variable measuring to what extent each interviewed trusts in the country’s parliament. trstlgl – a variable measuring to what extent each interviewed trust in the legal system trstplt - a variable measuring to what extent each interviewed trust in politicians trstprt - a variable measuring to what extent each interviewed trust in political parties trstep - a variable measuring to what extent each interviewed trust in the European Parliament trstun - a variable measuring to what extent each interviewed trust in the United Nations All these six dependent variables can be described together because of their very similar characteristics. First of all, they are all metric variables measured on a scale that does not have a ‘zero’ point. Also they all have the same measurement scale – 11-point metric scale with anchors. More particularly, in the variables ‘00’ refers to ‘No trust at all’ and ‘10’ refers to ‘Complete trust’. Other possible values include: ‘77’ which corresponds to ‘refusal’, ‘88’ corresponding to ‘don’t know’, and ‘99’ corresponding to ‘no answer’. These three values are the codes assigned to missing values which are the same for all six variables. These dependent variables were chosen with similar characteristics in order to receive better results in the further stages of the research. In Factor analysis all six variables will be combined in order to create one factor which will be used as a dependent variable for the linear regression model. • Independent variables. ppltrst – a variable measuring to what extent each interviewed would say that in general most people can be trusted, or that he/she can't be too careful in dealing with people. stfsdlv – a variable measuring how satisfied each interviewed is with his/her present standard of living. It can be noticed that two independent variables also share a lot in common. Firstly, they both are metric variables measured on a scale that does not have a ‘zero’ point. They both have the same measurement scale – 11-point metric scale with anchors. The only difference found is the meanings: in variable ‘ppltrst’ ‘00’ refers to ‘You can't be too careful’ and ‘10’ refers to ‘Most people can be trusted’; in variable ‘stfsdlv’ ‘00’ refers to ‘Extremely dissatisfied’ and ‘10’ refers to ‘Extremely satisfied’. In both variables other possible values include ‘77’ which corresponds to ‘refusal’, ‘88’ corresponding to ‘don’t know’, and ‘99’ corresponding to ‘no answer’. These three values are the codes assigned to missing values for both variables. Finally, all the chosen variables share the same characteristics for the measurement scale with regard to the ‘direction’ of the answers’ meanings (‘00’ stands for very negative answer, whereas ‘10’ stands for very positive answer). As a result, no modifications in the variables measurement scales’ direction have to be made. The next step of the data description is data screening which will help to identify and remedy any problem with data before beginning the data analysis. First of all, we need to identify missing values in the variables and check for the systematic patterns in missing data. The best way to find out more about the pattern of missing values is to use missing values function. The summarizing results can be seen below (Table 1). The first impression you receive from the table is that all the missing values of each variable are summed up to one number and in such a way we lose an opportunity to specify the exact ‘source’ of missing values (like it was specified in all variables’ measurement scale: refusal, don’t know, no answer). From the table we can notice that there are two variables with high level of missing values: trust in the European Parliament (16.5%) and trust in the United Nations (14.5%). In order to find out what could cause such a high level of missing values we tried to go deeper into our research and find some explanations. First of all, we executed cross-tabulation of these two variables with the countries to find out which countries have the highest levels of missing values. The results revealed that Ukraine showed 31.8% missing values in the trust in the EP and 29.9% missing values in the trust in the UN and Portugal showed 18% and 18.9% missing values respectively. On the other hand, Norway showed less than 1% missing values in all the variables related to political trust except trust in the EP (15%). The conclusion we are able to make about this information is that high level of missing values in trust of the EP in Ukraine and Norway is simply caused by the fact that they both are not members of the EU and as a result show no high interest in its affairs. Furthermore, the Portuguese high level of missing values in EP can be explained by not very active participation in EU matters (for instance, only 38.6% voted in the last European Parliament elections). A large number of missing values in the UN (in all countries, except Norway, it was above 10%) may be because of the lack of the connection between individuals and the organization since the UN does have such an institution as the EP where people can vote for their candidates and can feel being part of the EU. To sum up, we have recognized two rather high levels of missing values. We decided not to take any action (no pairwise or listwise deletion because SPSS will do it automatically when we run the linear regression), but we will pay more attention to them in order to perform all other steps smoothly. Table 1. Univariate Statistics N Mean Std. Deviation Missing No. of Extremesa Count Percent Low High trstprl 9732 4.29 2.603 447 4.4 0 183 trstlgl 9745 4.80 2.753 434 4.3 0 0 trstplt 9835 3.19 2.335 344 3.4 0 353 trstprt 9819 3.26 2.316 360 3.5 0 332 trstep 8503 4.40 2.507 1676 16.5 0 125 trstun 8700 5.14 2.706 1479 14.5 0 0 polintr 10133 2.66 .909 46 .5 0 0 stfeco 9809 4.97 2.797 370 3.6 0 0 ppltrst 10073 5.00 2.576 106 1.0 0 0 stfsdlv 10091 6.30 2.521 88 .9 630 0 stfedu 9573 5.38 2.396 606 6.0 359 0 stfhlth 9986 4.55 2.734 193 1.9 0 0 a. Number of cases outside the range (Mean - 2*SD, Mean + 2*SD). Secondly, we make univariate data screening in order to check for impossible values, outliers, normality of distribution of metric variables, ceiling/floor effect and restricted range of values. As our almost all chosen variables have the same interval in which values can be possible (except polintr) it is very easy to check for impossible values: we only need to look to the minimum and maximum values. From the descriptive statistics table we find that all values are in range from 0 to 10 (for polintr from 1 to 4) and it proves the absence of impossible values. Univariate statistics table shows that six variables have quite low number of extreme values (less than 0.07%) which should not cause any trouble in our further analysis. The best way to check for normality is to draw histogram for each variable. After doing it, we find that all variables are approximately normally distributed. Also kurtosis and skewness are less than 1 for all the variables (for all variables skewness is less than 0.332 except stfsdlv is -0.717). In order to make clear that normality is at least approximate for the variable stfsdlv, we draw Q-Q Plot (below) where you can see that the points are rather close to the straight line and it means that the variable is approximately normal. Furthermore, 1 ceiling/floor effect and restricted range of values are not found in any variable. Next step is to perform bivariate data screening to check for the linear relationship between pairs of variables. In order to do that, we firstly check the Pearson correlation coefficient between variables ppltrst and stfsdlv. The value we receive is equal to 0.285 and it implies that there is no strong linear relationship between the variables (Pearson coefficient: the closer is to 1 the stronger linear relationship). This result is very important to our further regression analysis because of a multicollinearity problem: a situation when two or more regressors are strongly correlated among themselves and it can cause biased estimators of the coefficients of the explanatory variables in multivariate regressions. In order to support our statement about non-collinearity between two variables we also produce the variance inflator factor (VIF) which is another indicator of a multicollinearity problem. If the VIF is above 10 (rule of thumb) then there is a serious problem with respect to multicollinearity. As we execute this procedure through the linear regression function in SPSS we find out that the VIF is equal to 1.089 (this value can be seen in the Table 6 in the Regression Analysis section) and it again proves that there is no strong linear relationship between two independent variables. The last thing we can do regarding multicollinearity is to draw a scatter plot where linear relationship could be seen. In the scatter plot there is no sign of linear relationship found and it confirms the absence of the problem with regard to multicollinearity. Finally, it is relevant to specify the assumptions which are necessary in order to begin our further steps of research project. For the factor analysis, it is the most important that variables are normally distributed (no dummy variables chosen) and that the level of missing values is not significantly high. All these assumptions have been discussed above and with regard to missing values the attention will be paid in the next section. For the linear regression model the assumptions made are the following: 1. The mean of the error term must equal 0 (normal error term). 2. Omoskedasticity: the error variance must be constant across observations. 3. There is no correlation among the errors. 4. There is no correlation among the regressors and errors. Factor Analysis To continue our research, we want to identify a relatively small set of factors (alias, ‘latent variables’ or ‘dimensions’) that explain a larger set of observed variables (‘indicator variables’). We will do it by using one of the statistical techniques called Factor Analysis. We start FA, taking into account the rules of thumb for factor analysis design which says that at least five variables should be chosen to reveal factor structure; we include six variables which amount to political trust. Moreover, each variable must have at least 50 observations and ours have even several times more. Then, we need to check if data matrix has sufficient correlations to justify application of factor analysis, in other words, if data are suitable for factor analysis. Correlation matrix reveals that all the correlations are above 0,30 and all the partial correlations are bellow 0,70 in the anti-image correlation matrix, so we conclude that the model is appropriate. Another way of determining the suitability of FA is by using Kaiser- Meyer- Olkin measure of sampling adequacy which is an index that compares the sizes of observed correlation coefficients to the sizes of the partial correlation coefficients. As a general rule for evaluating KMO MSA the measures below 0,50 are unacceptable. In our case, it is 0,819 which is considered to be meritorious. We should also test the null hypothesis (‘Ho’) that the observed data are a sample from a multivariate normal population in which all correlation coefficient are 0 (‘identity matrix’) with Barlett’s test of sphercity. We look for values with p

Daugiau informacijos...

★ Klientai rekomenduoja

Šį rašto darbą rekomenduoja mūsų klientai. Ką tai reiškia?

Mūsų svetainėje pateikiama dešimtys tūkstančių skirtingų rašto darbų, kuriuos įkėlė daugybė moksleivių ir studentų su skirtingais gabumais. Būtent šis rašto darbas yra patikrintas specialistų ir rekomenduojamas kitų klientų, kurie po atsisiuntimo įvertino šį mokslo darbą teigiamai. Todėl galite būti tikri, kad šis pasirinkimas geriausias!