This blog analyses the Role of Government IV data set from the International Social Survey Programme in 2006 in New Zealand. The full report was conducted for a politics paper at the University of Auckland. This is a brief summary to share my findings. A disclaimer to everyone that I have not done any formal statistics at university, so the analyses may contain (accidental) errors and unfortunately it will not be at a super advanced level (partially self-taught). Hopefully the results will still be useful and interesting.
Politicians and academics often talk about the importance of education, and it is often linked to developing a more tolerant, pluralistic and liberal society. If you want to be a bit more cynical, the conservative media also portrays higher education as a liberal indoctrination tool. This piece examines that relationship in New Zealand.
The question is also interesting for the field of political philosophy (my main background). Western political thought has historically emphasised the need for educating its citizens to recognise and respect political rights. In particular, Aristotle and Jean-Jacques Rousseau have been such proponents and influential in the motivation and design of the modern democratic state. There is also a wealth of academic literature supporting this, including the work of Hochschild, Jerit, Delli Carpin, and Nie.
Hochschild goes as far as saying
‘formal education, political knowledge, and political sophistication are not synonymous, and at one point later in the discussion I distinguish among them. [But] for most of this article, I treat them as interchangeable.’
More specifically, Nie also argues
‘the more educated citizen is more tolerant of the freedom of expression of unpopular political views, more knowledgeable of the fundamental principals (sic) of democracy, and maintains more information on other current political facts. … Education [also] creates engaged citizens by cutting the cost of pursuing and protecting interests in politics.’
There is also an assumption that people who are older tend to be more conservative. However, people who are older also tend to have higher education, so that effect might cancel each other out.
Is this all true? This is a question I have asked myself and wanted to find out. This piece investigates the relationship using quantitative analysis (I understand quantitative analysis has its limitations, but that is a completely different discussion).
Null hypothesis: The level of education has no effect on the individual’s support for civil liberties.
Alternative hypothesis: A higher level of education has a moderate and positive correlation with higher support for civil liberties.
Variables and Design
The independent variable will be the level of education (x). The dependent variable will be the level of support for civil liberties (y). The third variable age (z) is selected to further interpret the relationship between the variables.
Very quickly – what do we mean by ‘liberal’?‘
Civil liberties’ is a useful but ambiguous concept. Civil liberties are supported by the social-left and the libertarian-right. This is extremely important, because this is not a research question on whether education makes an individual more left or right wing, but whether it made them more liberal. ‘Liberal’, in this paper, has a precise meaning which we will contrast with ‘conservative’. It does not denote a specific identity on the political spectrum, nor some of its other popular usages. Instead, it denotes the general overall support for the set of individual rights against arbitrary power, such as freedom of speech, democratic freedoms, civil rights, and rule of law.
The dependent variable y is best produced by weighting and merging the responses to multiple relevant questions. This is because civil liberties can be polarizing and often operate as trade-offs, thus the best way to capture overall degree of support is by examining and averaging multiple scores. For this we need to look at the relevant questions used in the survey.
Raw data and Relevant Questions from the Survey
This survey had some fascinating questions. I encourage you to look at the questions I used for variable y (support of civil liberties), which ultimately required a subjective selection and weighting process because civil liberty is a subjective concept.
For the level of education, x:
Q36: Which of these categories best describes your highest formal qualification?
The answers range from no formal education to tertiary education and above.
For the level of support for civil liberties, y:
There is a total of eight questions (Q2.1, Q2.2, Q2.3, Q3.1, Q3.2, Q9.1, Q9.2, Q9.3) used for this variable. For all of these questions, the respondent used a scale to indicate their level of agreement (i.e. ‘definitely’ to ‘definitely not’).
Q2: There are many ways people or organisations can protest against a government action they strongly oppose. Please show which you think should be allowed and which should not be allowed by ticking a box on each line.
- Organisation public meetings to protest against the government.
- Organising protest marches and demonstrations.
- Organising nationwide strike of all workers against the government.
Q3: There are some people whose views are considered extreme by the majority. Consider people who want to overthrow the government by revolution. Do you think such people should be allowed to:
- Hold public meetings to express their views?
- Publish books expressing their views?
Q9: Suppose the government suspected that a terrorist act was about to happen. Do you think the authorities should have the right to:
- Detain people for as long as they want without putting them on trial?
- Tap people’s telephone conversations?
- Stop and search people in the street at random?
For the age level, z:
Q27: Please indicate the year in which you were born.
The whole process of cleaning, processing and merging the data to create a scale of civil liberties can be found in the full report so I will not go into detail here.
Results I – Analysis (Graphs)
The data set has 1263 responses. After filtering out entries with unanswered responses, the sample size is n = 1102. The report has more details on error. In short, this is a postal survey of 2250 people aged 18 and over who are randomly selected, with 1200 valid responses and an effective response rate of 60%. The sample size has a confidence interval of 95% with a margin error of 3%, and the survey data was weighted to correspond to the age-sex distribution of the New Zealand population. The sampling frame is the New Zealand electoral roll, about 3.3 million people.
Firstly, my apologies about the graph having strange headings and labels, those are the name of the variables I used in SPSS. Hopefully they are self-explanatory; I will explain them briefly below each one. The report contains cross-tabs and correlation tables. I have only inserted the graphs.
x – level of education of sample This shows the frequency of the level of education attained by the respondents. The untransformed data for the level of education is sufficiently distributed across each category for each category to have a large enough sample size. The data is slightly skewed to the right with a skewness of 0.212, which means the sample contains slightly more respondents with only high school diploma equivalent or below, but this is small and does not substantially distort the data.
y – level of support for civil liberties This shows the frequency of the level of support for civil liberties for the respondents. The score is a merged/weighted score of the eight questions on civil liberties, where individuals’ responses were assigned 1-4 based on their answer. 8 (minimum score) shows the lowest possible support to all questions to 32 (maximum score) shows the highest possible support to all questions. The result is normally distributed, and has a skewness of -0.111. This shows that the sample is very slightly more liberal than illiberal, but not by much.
z – age level This shows the frequency of the age of the respondents. It should be noted, that there are some data which are outside two standard deviations from the mean, those under 20, and those above the age of 82. They are not outliers, but they fall outside the 95% confidence interval range.
For bivariate and multivariate analyses, the univariate data has been transformed into categories to better visualise the data.
Low (Support) = Conservative, Moderate (Support) = Moderate, High (Support) = Liberal.
This graph breaks down the level of support for civil liberties based on level of education (horizontal axis). This graph breaks down the level of support for civil liberties based on age (horizontal axis).
The pie chart breaks down the level of support for civil liberties based on education (columns) and age (rows). Note that the pie chart is good at visualising the proportion (percentage) whereas the bar charts for bivariate analysis above showed the amount.
You can see clearly that generally, as education increases (going across each row) the proportion of liberals (grey) increases and conservatives (blue) decreases. As age increases (going down each column), the proportion of liberals (grey) shrinks and conservatives (blue) increases.
Results II – Interpretation
The chi-square test indicates that the relationship is statistically significant. Cramer’s V value indicates that there is a low to moderate association (= 0.181) between the level of education and support for civil liberties. The results support the alternative hypothesis of moderate and positive correlation when only looking at these two variables.
Effects of Age?
It is not enough to simply control for age, we need to look at the specific effects for each age group. The cross tabulation (see pie chart; full crosstab in report) and Cramer’s V values show age level is a moderating variable and not a spurious or intervening relationship. More specifically, if you compare the multivariate analysis with the original bivariate analysis between education level and support for civil liberties, the largest group of people at each education level are moderates except tertiary graduates, which have more liberals.
When the data is broken down by age, the percentage of liberals substantially increased for 18-34 year-olds. For individuals with trade certificates and diplomas, 18-34 year-olds became the only age group within that education level where there were more liberals than moderates. The opposite happened for 65+. There are significantly less liberals. In fact, 65+ was the only age group for tertiary graduates where there were actually less liberals than moderates.
Relationships: controlling the third variable
A partial correlation was conducted to identify the correlation while controlling for the ‘other’ variable.
Controlling for age, the correlation between education and perception of civil liberties is positive and becomes 0.194.
Controlling for education, the correlation between age and perception of civil liberties is negative and becomes -0.252 (moderate negative relationship).
Controlling for civil liberties, the correlation between education and age is negative and becomes -0.143 (small negative relationship)
None of the Cramer’s V values for each bivariate analysis is sufficiently similar to suspect a spurious relationship. The controlled partial correlations show that there is a direct correlation between the level of education and level of support for civil liberties, a direct (but negative) correlation between the level of education and age level, and a direct (but negative) correlation between age level and the level of support for civil liberties. We already established that age level also has a moderating relationship. It is interesting to note that because of this, the dependent variable and moderating variable essentially form a negative feedback loop on the independent variable, though this is beyond the scope of this discussion. The negative correlation is important, otherwise it would be an intervening relationship alongside the direct relationship.
The relationship between the three variables can be summarised by the diagram below.
There is a low to moderate direct relationship between the level of education and support for civil liberties, even when controlling for age. Age has a moderating relationship on the relationship between education and support for civil liberties. Younger people (18-34) increases the strength of the correlation between higher education and supporting civil rights, whereas this effect dissipates for older, retired people (65+). However, all relationships are only low to moderate strength, which suggests there are many other uncontrolled variables affecting the relationships (perhaps family upbringing, religion, ethnicity and/or gender). Nonetheless, the method supports the design, and the data supports the academic literature.
The null hypothesis was disproven. The alternative hypothesis is correct.
References for Quotes
 Hochschild, “If Democracies Need Informed Voters, How Can They Thrive While Expanding Enfranchisement?.”
 Nie, Education and Democratic Citizenship in America.
This entry was posted in Democracy, Political Philosophy, Social Science and tagged Age, Civil Liberties, Democracy, Education, Freedom, New Zealand, Politics, Quantitative Analysis, Social Research, Social Science, Statistics.