CAPITULO II. PROTECCION DEL DOMINIO PUBLICO MARITIMO-TERRESTRE ESPAÑOL, CON ESPECIAL
VIII. LOS EFECTOS DEL DESLINDE DEL DEMANIO MARÍTIMO TERRESTRE: MARCO GENERAL TERRESTRE: MARCO GENERAL
3. Alcance de la rectificación registral
In order to accord with Popper’s (1963) arguments regarding falsifiability, the following null hypotheses were also tested:
H01: Low satisfaction with life as a whole does not correlate with placement on the right side of the left right political scale.
112 H02: One’s opinion on whether immigrants make one’s country a worse or better place to live does not correlate with a right-side placement on the left-right political scale.
H03: One’s opinion about whether immigrants undermine cultural life does not correlate with a right-side placement on the left-right political scale.
H04: Low satisfaction with life as a whole does not predict far-right placement on the left-right political scale.
H05: One’s opinion on whether immigrants make one’s country a worse or better place to live does not predict far-right placement on the left-right political scale.
H06: One’s opinion about whether immigrants undermine cultural life does not predict far-right placement on the left-right political scale.
113 3.2 METHODOLOGY
3.2.1 Data
The data analysed in this chapter are drawn from European Social Survey (ESS), a major pan-European study of social and political attitudes collected every two years.
Analysis made use of round 7 (collected in the United Kingdom between September 2014 to February 2015 and October to December 2015, and in Hungary between April and June of 2015) and round 8 (collected from September 2016 to March 2017 in the United Kingdom and between May to September of 2017 in Hungary) data, details on the demographic spread of which are presented in table 3.1, below. Hungary is still considered as part of the 2016 survey round although the data was collected in 2017 and released in 2018. It must be noted that, longitudinally, these two data sets do not offer a perfect comparison as data was not collected at precisely the same times (see below) and different samples of respondents were surveyed in each Round. This analysis does, however, give a broader understanding of the social contexts of the two countries over several years.
It should also be noted that the data were collected across the United Kingdom, including Northern Ireland, and it is consequently extremely difficult to disaggregate data collected exclusively in Great Britain. Therefore, in this chapter the comparison will be between the political and social contexts of Hungary and the United Kingdom; the remainder of this thesis focuses on Great Britain, with the exclusion of Northern Ireland (as mentioned previously).
Half of the questionnaire is repeated every round of surveys. All Rounds of the surveys cover three broad categories: value and ideology (including religion, political views, and morality), cultural and national orientations (national and ethnic identity), and the social structure of their society (class, education, social exclusion) (Fitzgerald & Jowell,
114 2010). The 2014 survey covered more specific questions of immigration and the 2016 questions covered questions of climate change. For the purposes of this study, variables under the categories of socio demographics, politics, and the broad category of ‘subjective well-being, social exclusion, religion, national and ethnic identity’ were used.
Participants were all residents in private households, regardless of nationality, citizenship, or language, and were aged 15 years or older, with no upper age limit imposed on inclusion; all were selected by strict random probability methods. A fuller exploration of sampling is provided in the European Social Survey technical report (for the current sampling guidelines, see Lynn et al., 2018).
The Hungarian Round 7 data sample was made up of 1698 individuals. Several individuals were removed from the sample prior to statistical analysis as they were under 18 years of age. This resulted in a total of N = 1663 individuals for the analysis, with 704 males (42.33%) and 959 females (57.67%), with an average age of 46.9 (see Table 3.1).
The Hungarian Round 8 data sample was made up of 1614 individuals, with a total of N = 1576 individuals used in analysis; of these respondents, 662 were male (42.01%) and 914 were female (57.99%), with an average age of 49.
The UK Round 7 data sample was made up of 2264 individuals. Several individuals were removed from the sample prior to statistical analysis as they were under 18 years of age. This resulted in a total of N = 2206 individuals for the analysis, with 995 males (45.10%) and 1211 females (54.90%), with an average age of 47.3. The UK Round 8 data sample was made up of 1959 individuals, with a total of N = 1892 individuals used in analysis; of these respondents, 845 were male (44.7%) and 1047 were female (55.3%), with an average age of 48.
115 Table 3.1: Demographic description of all ESS Rounds in both Hungary and the United
Kingdom, showing distribution by gender, mean age of each sample, and standard distribution.
Analysis made use of previously-collected survey data, which is generally freely available online. This method of survey analysis has several advantages. It saves resources by eliminating the need for a large research team, as is generally needed in cross-national surveys such as the ones utilised here. It also saves time, as the surveys are already completed and available. Lastly, it also circumvents data collection problems as data is already computerised as machine-readable survey data (Kiecolt & Nathan, 2004).
However, there are also limitations. Researchers can have problems locating specific information that is needed, especially in vast data archives (Kiecolt & Nathan, 2004). Data may also not be available in the format that is needed for specific research, depending on the original intentions of those who conducted the survey. Also, errors made in the original survey are no longer visible; any typos and coding errors have disappeared into the survey and been forgotten (Kiecolt & Nathan, 2004). Lastly, surveys also rarely contain all the values of interest to a secondary researcher.
Keeping in mind these issues, secondary survey analysis was found to be the best method for this research as the data was already freely and readily available. Additionally, the same set of questions was asked of both the Hungarian and UK respondents. The ESS also covered the questions that were essential for this study, namely those surrounding
left-116 right political scale placement, satisfaction with life, views on immigrants, and the proper demographic information.
Three methods of analyses were used to explore the relationship between self-positioning on the left-right political scale, life satisfaction, and views on immigration:
bivariate correlation, linear regression, and binary logistic regression. In all cases positioning on the scale was the dependent variable. This analysis will begin with bivariate correlation, in order to explore whether a relationship exists before continuing to more complex analyses. Linear regression will then be applied, to gain a level of complexity by the addition of control measures. This allows to see whether demographic factors, such as age, gender, education, partnership, and employment, have any effect on the correlation, or if differences are purely due to the independent variables. Finally, this analysis will specifically explore far-right views, defined as values 9 and 10 on the political scale, through binary logistic regression. All analyses were conducted through SPSS and data was weighted using post-stratification weights in order to reduce the effects of sampling error and non-response bias.
Initial bivariate analyses were completed to find general relationships between dependent and independent variables. All correlations were completed using SPSS with Spearman’s rho, as questionnaire answers were completed on a Likert scale and are at an ordinal level of measurement. Some questions could have been interpreted at an interval level of measurement, but a nonparametric test was necessary as variables were not found to have normal distributions.
To gain levels of complexity, linear regression analysis was completed with the addition of control measures. This second level of analysis explored the same hypotheses as for the bivariate analysis (H1, H2, and H3). Predictors were gender, age, employment, partnership, and years in education. Gender and employment were transformed into binary
117 measures, while partnership was found by combining those individuals with legal marital status and those cohabiting with a partner. Age and years in education are scale measures.
Hierarchical linear regression was run using SPSS. Regressions were run in two steps: the first including only control measures, with the second step introducing the key independent variable. The results of the two models were then compared to measure the effect of adding the independent variable. Results were compared between the Hungarian and UK samples.
Bivariate analysis and linear regression explored the relationship between the independent variables and placement on the left-right political scale. In order to go one step further, binary logistic regression was employed to explore the relationship between the independent variables and far-right placement on the political scale (H4, H5, H6), while utilising control measures. Individuals were classed as far-right if they identified as a 9 or a 10 on the left-right political scale, which ranged from 0-10.
Several control measures were transformed to be binary in nature. For gender, males were transformed to ‘1’, while all others ‘0’. Similarly, for employment, those who were employed at the time of the survey were ‘1’, all others ‘0’. Partnership was found by combining those individuals with legal marital status and those cohabiting with a partner:
these were given ‘1’, all others ‘0’.
Logistic regression was run using SPSS in a stepwise fashion, with the first step testing for demographic predictors and the second step including one independent variable.
Goodness of fit was tested using the Hosmer-Lemeshow test, the Omnibus test showing the effect of the model, and Nagelkerke’s pseudo R2 showing the relationship between the predictors and the prediction of the model. Results were compared between the Hungarian and UK samples.
118 A 10 percent alpha level was used for all tests. It was decided that a significance level of p = .1 would be used mainly due to the preliminary nature of this analysis.
Conventionally there is a reliance on the 5%, or even 1%, significance level, but that is now thought to be merely arbitrary (Gerber & Malhotra, 2008). Since these research questions were chosen to examine a new theoretical perspective on these samples, it can be argued that the risk of a false positive outweighs the consequences of incorrectly identifying a relationship. Once this new theoretical perspective has been established, further testing is possible at lower alpha levels in the future.
3.2.3 Ethics
Some attention must now be turned to examining the ethical considerations of secondary survey analysis. The European Social Survey (ESS) data is freely available on the Internet, including an online analysis tool, hence permission for further use is implied.
However, the ESS does have one condition for the use of their data: they require a ‘deposit,’
meaning that users are required to register all bibliographic information in all forms of publication referring to ESS data to an online ESS bibliography database. This condition will be fulfilled once this research is submitted, and once any of this analysis is published.
The ESS have anonymised their data, so it is impossible to trace the data back to any individuals. However, even when data has been anonymised there can still be a risk that a participant may be identified (ESRC, 2012). While it is appreciated that this must be considered in any data dealing with individuals and possible risk, it would still be very difficult to trace a specific participant through ESS data. It would be especially impossible to trace a participant through the analysis in this study, as no specific participant numbers or correlated information is shared. Individual cases will not be discussed, and this research aims to find macro-level trends.
119 3.3 DATA DISTRIBUTION