MARCO TEÓRICO
Art 38. Objetivos de los programas de educación La educación básica y media asegurarán los conocimientos, valores y actitudes indispensables para:
Since the meaning of the concepts we use is now clear, we will proceed with finding a way to measure them. Normally, after the conceptualization, one should continue with the operationalization of the concepts into measurable variables and finally be engaged with the technical aspects of the data selection. We do not follow this path. Limited time and resources prohibit us from creating our own survey and dataset, based on our own operationalization. Relying on preexisting surveys means that we no longer have the luxury to operationalize as we wish. On the contrary, we have to look at what others have done and pick the operationalization that is closest to our needs. In a way, what we do is to reverse engineer others’ operationalization. When we look at what questions are available in surveys, we are actually looking at the results of others’ operationalization of the same or similar concepts. Of course, we have an idea of how we would have done the job ourselves but since this will not happen, we choose the questions that are closer to the ones we would have used if we had made our own survey. When the research topic is as narrow as ours is, and when the theoretical distinctions between the various concepts are so thin, the questions that qualify as appropriate are very specific. Even a slight divergence may compromise the results’ validity. Unfortunately, when what you need is not easy to find, it is tempting to settle for something similar that happens to be readily available. Many make this mistake, but we will not. Since we didn’t create our own survey, the questions we use are not tailored to our needs. However, we made sure that our data, albeit not optimal, is undeniably appropriate for the measurement of our variables. We rely on neither trust nor perceptions of knowledge. We want to measure civic knowledge and political support and this is what we try to do.
Since our subject of interest is a pan-European phenomenon, we decided that it is appropriate to conduct a study that uses data from the entirety of the EU. We thus need a big amount of reliable and accessible data that is collected from across the EU. Moreover, considering that both civic knowledge and political support are phenomena that only make
sense at the individual level, our level of analysis will be the individual. Hence, we are in need of reliable, accessible, EU-wide, individual level data. When it comes to sources of such data, few are as trustworthy as the Eurobarometer surveys. Firstly, we examined all the Standard, Special and Flash Eurobarometer surveys of the last 5 years and identified in each of them the questions that can help us measure the variables we are interested in. Having a good idea of what kind of information each dataset contains, we then chose to work with the dataset whose survey contained the best possible combination of questions. Our choice is the Standard Eurobarometer 84.1 that was conducted on September 2015 (European Commission and European Parliament, 2016).
Operationalization
Now let us turn to how we operationalized our two main variables, starting by support for the EP. As we said, by examining causality on the EP level, all we have to measure is support for the EP (which belongs to the object regime institutions), thus freeing ourselves from the need to measure the rest of the objects of support. We also know that support for regime institutions is mainly diffuse, which means that it comes in the form of a “reserve” of favourable or unfavourable attitudes about the EP. Instead of using questions about trust, we preferred to measure exactly this reserve of positive or negative attitudes towards this regime institution. On the third page of the 84.1 Eurobarometer survey, question 5 asks the respondents:
“In general, does the European Parliament conjure up for you a very positive, fairly positive, neutral, fairly negative or very negative image?”
1. very positive (1) 2. fairly positive (2) 3. neutral (3)
4. fairly negative (4) 5. very negative (5)
6. DK (6)
For our research, we will alter the dataset and change the values of the answers in a way that positive answers hold a positive value and negative answers hold a negative value. As for those who gave a “Don’t know” answer we are going to merge them with those that gave the "neutral" answer. We assume that a person who doesn't know if they support the EP or not or neither support nor oppose, is closer to those who know that they neither support nor oppose it. We said before that attitudes are only important because they always translate into actions. Here this part of our theory comes in handy. Those who answer “don’t know” can be expected to act similarly to those who were neutral. The following table shows how we code the data.
Answer Value in Eurobarometer’s dataset
Value in our dataset
very positive 1 +2 fairly positive 2 +1 neutral 3 0 fairly negative 4 -1 very negative 5 -2 DK 6 0 Table 1
When it comes to civic knowledge, we wanted to make sure that the survey we pick satisfies two criteria. One, that the questions included actually measure civic knowledge and not the very common political knowledge, issue knowledge or perceptions. Two, that the questions measure (to the extent that this is possible) the entire range of civic knowledge, from
very high to very low. In order to do that, some questions would have to be more difficult than others. It would be pointless to try to measure people’s mathematical skills by asking them to perform simple additions of the 1 + 1 = 2 type. Everybody with few exceptions would answer correctly and the peoples’ actual skills would be vastly overestimated. The opposite problem (underestimation) would arise if all the questions were so difficult that only a mathematician could give the right answer. Similarly, in order to measure civic knowledge, one would need questions of varying difficulty. We are doing this both to maximize the variance captured and also for the sake of proper operationalization in and by itself.
As we were going through the numerous Eurobarometer surveys, we realized that the same questions were being recycled and used repeatedly. We also noticed that while certain questions were very easy, some of them had a moderate difficulty and a few of them were more difficult to answer correctly, even for an informed citizen. We therefore divided all those questions in 3 broad categories: easy, average and difficult. Unfortunately, no survey was entirely focused on political or civic knowledge and therefore no survey contained all of the questions that appear periodically. That would have been very helpful and would have allowed us to track the levels of civic knowledge with great accuracy. Since this was not the case however, we needed a survey that contains enough questions to represent the entire spectrum of civic knowledge. If our survey contains an uneven amount of questions for each category (easy, average and difficult) it will be necessary for us to correct that and make sure that all 3 categories have the same impact on the final outcome. If for example we had 1 very easy question, 2 average ones and 7 very difficult questions, it is safe to assume that without controlling for that, most people would appear to be ignorant even if they had an average knowledge. The actual survey that we chose eventually has 5 questions, 1 easy, 2 average, and 2 difficult. Each category shall have equal gravity and shall count for 2 points. As a result the easy question will count for 2 points, and each of the average and difficult will count for 1 (see table 2). We should note here that a possible venue for criticism of our research is the way in which we assigned the degree of difficulty to each question. There may not be an objective way
of doing that, but the small number of questions and the big difference in their complexity makes our job easier and our decision in line with common sense.
The questions that we use to operationalize civic knowledge about the EP are question 11 on page 5 and question 12 on page 6 of the Eurobarometer survey. Here they are as they appear in the survey:
“For each of the following statements about the EU could you please tell me whether you think it is true or false?”
# Question True False DK
1 The members of the European Parliament are directly
elected by the citizens of each Member State. 1 2 3
2 The EU’s budget is determined jointly by the European
Parliament and the Member States 1 2 3
3
At the EU level, European laws (directives and
regulations) have to be agreed jointly by the European Parliament and the Member States
1 2 3
4 Each Member State has the same number of Members
of the European Parliament 1 2 3
Table 2
“And, in your opinion, do members of the European Parliament sit in the European Parliament according to…?
1. Their nationality (1) 2. Their political affinities (2)
3. DK (3)
We combine all of those questions in the following table.
Question Difficulty Respondent’s possible Answers Correct Answer The respondent’s answer is: Coding 1
The members of the European Parliament are directly elected by the citizens of each Member State
Easy 1 or 2 or 3 1 Right or
Wrong 1 or 0
2
The members of the European Parliament are directly elected by the citizens of each Member State
Easy 1 or 2 or 3 1 Right or
Wrong 1 or 0
3
Each Member State has the same number of Members of the European Parliament
Average 1 or 2 or3 2 Right or
Wrong 1 or 0
4
And, in your opinion, do members of the European Parliament sit in the European Parliament according to…
Average 1 or 2 or3 2 Right or
Wrong 1 or 0
5
At the EU level, European laws (directives and regulations) have to be agreed jointly by the European Parliament and the Member States
Difficult 1 or 2 or3 1 Right or
6
The EU’s budget is determined jointly by the European
Parliament and the Member States
Difficult 1 or 2 or3 1 Right or
Wrong 1 or 0
Table 3
For each individual respondent we can see how many questions out of 5 he or she has answered correctly, with question 1 counting twice (in order to keep the value of the different categories balanced) leading to a possible total score of zero to six. Zero signifies complete ignorance and six signifies very high knowledge. A “Don’t know” answer will be counted as a wrong answer since both of them indicate lack of knowledge.
Research Method
The ideal course of action would be to study the relationship for the EU in its entirety. Studying the phenomenon in the entirety of the EU would not only prove or disprove in a definitive manner the causal relationship between civic knowledge and diffuse political support; it would go a step further to actually measure the levels of support and knowledge in the Union, in a way that hasn’t been done to this day. Unfortunately, in our case this is not possible. Most of the surveys include at least a few questions that can be used to measure civic knowledge and political support but very few of them can measure the entire spectrum of civic knowledge on the one hand, and support on the other. Since testing the causal relationship on the EU as a whole is impossible, we opt to scale down and test it on one very specific EU institution, the European Parliament (EP). For the EP, the surveys allow us to measure the entire spectrum of knowledge and, since this institution does not represent the political regime as a whole, we avoid having to measure support for all 5 objects (political community, regime principles, regime performance, regime institutions and political actors). One could argue that the EP’s levels of support may differ from those of the EU in general, but for us this is not a problem.
The EP may enjoy more or less support than the EU as a whole and thus be a bad indicator for support (or knowledge) in the Union in general, but we do not try to measure the levels of knowledge (or support) in the EU by using the EP as a proxy. That would indeed, be problematic. What we try to do is to see if the relationship works for the EP, and if it does, then it should work for any institution or political regime that shares the EP’s democratic characteristics. More specifically, from this single institution causality should be generalizable to the object regime institutions and from there to the EU as a whole. If the variance in the levels of civic knowledge can explain the variance in the levels of diffuse political support for the EP, then (even though the actual levels may be different) the variance in the levels of civic knowledge for the EU in its entirety should explain the variance in the levels of support for the EU in its entirety (assuming that the EU as a whole shares in the eyes of the public some key democratic characteristics with the EP). Moreover, as the type of support for regime institutions, political community and regime principles is diffuse and therefore quite resilient, if our research discovers causality in one, this causal chain might hold for the other objects as well.
In the introduction we identify our research as positive, empirical and explanatory, with theory testing and theory generating goals. The previous paragraphs explained that we aim to study the link between knowledge about the EP and support for the EP because, if proved to be in place, this relationship would work for any similar institution or democratic regime. Now let us explain how exactly we are planning to do it. First of all, we will implement a large - N research design. We have in our disposal an enormous amount of individual level data, obtained through Likert-type scales and we will run regressions with the help of IBM's SPSS software.
The selection of research method was guided by the constraints we face due to the nature of our data and we concluded that the most appropriate method to use is the cumulative odds ordinal logistic regression with proportional odds, which is also one of the most popular types of ordinal regression. It makes the ordinal categories of our variable binary
and uses the cumulative probabilities to estimate how likely it is to be in a certain category or any category below it versus the likelihood of being in a higher category. But this method doesn’t stand alone; the ordinal regression model is one of the many models that comprise the generalized linear models which in turn are an extended version of the general linear models, so as to allow for non continuous or not normally distributed dependent variables, just like the one we are dealing with in our research, diffuse support for the EP. Having explained our level of analysis, data selection and research method, we shall now present the operationalization of our main variables.