The Impact of Direct Democracy
- A quantitative study examining the effect of direct democracy on abortion
policy responsiveness regarding public opinion on abortion in Latin America
Political Science C
Department of Government Uppsala University
2018-05-17
2
Table of contents
1. Introduction ... 3
2. Theoretical framework ... 6
2.1 Direct democracy ... 7
2.1.1 Defining direct democracy ... 7
2.1.2 Direct democracy and policy responsiveness ... 8
3. Data and research design ... 13
3.1 Data ... 13
3.1.1 Measuring direct democracy ... 14
3.1.2 Measuring public opinion ... 17
3.1.3 Measuring abortion policy ... 18
3.2 Method ... 19
4. Analysis ... 21
4.1 Bivariate regression of policy responsiveness of public opinion ... 21
3
1. Introduction
The first form of democracy, the Athenian democracy, was established on a square in old Greece. The people gathered to vote directly on different issues and the concept was named democracy – rule by the people. It was direct democracy in its purest form. Today, democracy is more associated with representative democracy, where elected representatives take political decisions for the voters that elected them. Still, mechanisms of direct democracy (MDDs) exist to different extent around the world. MDDs are often described as mechanisms through which citizens can express their opinion in a direct way regarding specific issues, for examples through initiatives and referendums.1 But do these mechanisms move legislation closer to or further away from the public opinion? Previous research has examined this question with results supporting that MDDs can both enhance and distort the public will.2 The main discussion is about whether MDDs are a mean for the majority to express their opinion, creating policy better representing the will of the majority, or if they are exploit by a few, special interest leading to policies being pushed further away from the public opinion.3 There are several studies supporting both sides of the discussion, which will be presented under theoretical framework.
Before going further into the discussion, it is of importance to clarify the meaning of some terms commonly used in this thesis. ‘Policy responsiveness’ is used to describe the relationship between public opinion and policy, more precisely how well policy responds to the public opinion. This will be further explained under theoretical framework. Policy/public policy refers to national laws and legislation, and public opinion is defined as the people’s collective preference.
Alongside the discussion about whether direct democracy moves legislation closer to or further away from the public will, there is another methodological discussion about how direct democracy is best measured. A traditionally common way of measuring direct democracy is through a dummy variable measurement, which later has been criticized for its stiffness and for not considering other important indicators of direct democracy. For example, the dummy variable does not say anything about how often MDDs are used, how easy they
1 David Altman, Direct Democracy Worldwide, Cambridge: University Press, 2011, p. 33.
2 Kevin Arceneaux, “Direct Democracy and the Link between State Public Opinion and State Abortion Policy”, State Politics & Policy Quarterly, vol. 2, no. 4, 2002, p. 373.
4
are to initiate and the legal status of the MDDs, which might lead to under- or overestimating the effect of direct democracy on policy responsiveness of public opinion. 4
Accordingly, two conflict areas on the topic direct democracy have been identified. First, does direct democracy actually move public policy closer to the public opinion and second, does the relationship between public policy and public opinion change depending on the way we choose to measure direct democracy? With this as a background, the aim of this paper is to investigate if, and how, direct democracy affects the policy responsiveness of public opinion by striving to answer the following question:
How does the opportunity to direct democracy affect the abortion policy responsiveness of public opinion on abortion in Latin America?
The paper also strives to shed more light on the methodological discussion about measuring direct democracy by using two different measurements. First, a dummy variable measurement based on if the country permits initiatives and referendums or not, and second an index which is more nuanced, taking several indicators of direct democracy into account. Including these two measures of direct democracy makes it possible to compare the results and examine if they differ depending on which measurement is used.
Why is abortion a suiting policy area for examining direct democracy? Previous research has emphasized the importance of the policy being an ‘easy issue’, meaning it does not require neither cognitive ability nor attention to politics from the voters.5 That is to say, most people feel knowledgeable enough to understand and participate in the political debate.6 For example, the death penalty is an issue many people feel secure enough in their knowledge to express their opinion on. Thus, it can be characterized as an easy issue according to the definition above, even though there of course are several different discussions and arguments to consider for the interested person. An easy issue does not mean that an academic or philosophical discussion on the topic doesn’t exist, but that it is not necessary for the citizen to be a part of these discussions to understand the question.
4 Shaun Bowler and Todd Donovan, “Measuring the Effect of Direct Democracy on State Policy: Not All
Initiatives Are Created Equal”, State Politics & Policy Quarterly, vol. 4 no. 3, 2004, p. 345.
5 Edward G. Carmines et al., “The Two Faces of Issue Voting”, The American Political Science Review, vol. 74,
No 1, 1980, p. 88.
5
Abortion can also be categorized as an easy issue. Abortion does not require any political knowledge from the ordinary citizen, and is often conceptualized as a morality issue, characterized by conflicts over core values and beliefs.7 For example, an argument often used by abortion opponents is the foetus’ right to life. Proponents of this argument believe that life begins at conception, and often refer to different religious doctrines when stating this. On the other hand, the pro-choice movement often uses the woman’s right to decide over her own body as an argument in favour of abortion, which is based on an autonomic and feminist mindset. These are clear examples of how core values and beliefs can affect the public opinion on abortion, which makes it a suiting policy area for examining direct democracy.
The majority of previous research on direct democracy has been conducted in USA, with states as units of analysis. Even if this is a region with favourable circumstances for several reasons, for example good access to data and the many units of analysis, it is not the most interesting region to implement this study on. Mainly since a lot of similar studies on the region already exist, and to examine another geographical area would contribute more to existing research. If the results are different, it could be interesting for further research to seek explanatory factors explaining why results change depending on region. By studying Latin America this thesis applies existing framework, though modifying it slightly, on a new region and at the same time shifts the level of analysis from regional to national.
Latin American is also a particularly interesting region for studying abortion policy for several reasons. In general, the region’s abortion laws are very strict and abortion is only legal upon request in three areas: Uruguay, Cuba and the state Mexico City in Mexico.8 Latin America is also a region with a lively contemporary abortion debate, indicating a public interest regarding the question. Even though feminist movements in Latin America have been protesting these strict laws for a long time and demanding the right for women to decide over their own bodies, lately the mobilization seems to be even stronger. For example, attention has been drawn to the topic after several women have been accused of undergoing illegal abortions after having miscarriages. In El Salvador for example, where there is a total ban on abortion under all circumstances, women who go through stillbirths or miscarriages risk imprisonment accused of having voluntarily abortion. This has triggered mobilization within
6
the country, but also drawn attention from international organizations advocating for more liberal abortion laws.9 This confirms abortion being a salience issue in Latin America, which has been emphasized giving favourable circumstances for the correlation between public opinion and policy by previous research.10 More exactly, a salience issue is an issue that is of importance to the public and is prominent in public discourse, for example in media.11 This makes the probability of a correlation between public opinion on abortion and abortion policy greater in Latin America than other regions where the abortion issue is not as controversial or salient.
This thesis does not make a claim on explaining the casual mechanisms between direct democracy and the policy responsiveness of public opinion. Even though that would be interesting to examine, it calls for another research design and a deep study of the timeline between the variables. To examine this, a long-term process tracing-method would probably be more appropriate.12 Neither does the paper make any normative statements about direct democracy, i.e. if it is desirable or not. The purpose is, as been stated above, to simply investigate if, and in what way, direct democracy affects the policy responsiveness of public opinion and examining how the results might change using different measurements.
The outline of this thesis is as follows: First the theoretical framework is presented, including definitions, a review of previous research and the thesis hypotheses. This is followed by a data and method section, describing the data for each variable and the methods used to analyse this data. This is followed by the analysis and a discussion of the results. Lastly, the thesis finishes with some short conclusions.
2. Theoretical framework
This study is mainly inspired by Kevin Arceneaux’s study “Direct Democracy and the Link between State Public Opinion and State Abortion Policy”. Arceneaux studies how direct democracy influences the correlation between abortion policy and the public opinion on abortion in USA. He does this using a dummy variable for measuring direct democracy,
9 Elisabeth Malkin, ”They Were Jailed for Miscarriages. Now, Campaigns Aims to End Abortion Ban”, nytimes.com, April 9, 2018, https://www.nytimes.com/2018/04/09/world/americas/el-salvador-abortion.html, accessed May 2, 2018.
10 Jeffrey R. Lax and Justin H. Phillips, “The Democratic Deficit in the States”, American Journal of Political Science, vol. 56, no, 1, 2012, p. 157-158.
11 Ibid.
7
where 0 represents states that do not allow initiatives or referendums, and 1 represents states that do.13 As been mentioned in the introduction, this way of measuring direct democracy has been criticized for its stiffness. This thesis uses the main thoughts of Arceneaux’s study, but applies it on a different context and with other types of data. Most importantly, another definition of direct democracy is used, which opens for a more nuanced operationalization of the term.
2.1 Direct democracy
2.1.1 Defining direct democracy
Direct democracy lacks a common connotation and existing literature does not provide a unified definition of direct democracy. Often, the meaning shifts among different places, and direct democracy concepts such as popular initiatives, referendums and plebiscites have diverse normative undertones.14 Though, a collective view among the different interpretations is that MDDs allow citizens to in a direct way propose, veto or at least have a say in legislation.15
In this study, direct democracy is defined as:
“[…] Direct democracy refers to an institutionalized process by which citizens of a region or country register their choice or opinion on specific issues through a ballot. It encompasses initiatives, referendums, and plebiscites, but does not encompass recall elections, deliberative assemblies, or other settings in which the vote is not secret and the purview is restricted. It also does not apply to elections for representatives.”16
This is a suitable definition for measuring the opportunity to direct democracy. Mainly, because it includes three of the most common and frequently used MDDs: initiatives, referendums and plebiscites. At the same time, the definition makes it clear what circumstances are necessary for something to be called an MDD and emphasizes especially the importance of a secret voting procedure.
13 Arceneaux, “Direct Democracy and the Link between State Public Opinion and State Abortion Policy”, p. 379. 14 David Altman, “The Potential of Direct Democracy: A Global Measure (1900–2014)”, Soc Indic Res, vol. 33,
no. 4, 2017, p. 1208.
8
For understanding this definition, it is necessary to clarify the meaning of some terms used: Initiatives: Initiatives are initiated by citizens with the purpose of offering an alternative to the status quo. Initiatives are initiated by the citizenry through signature gathering, and serve as a proactive institution where the people can decide directly at the ballots on matters of concern to them. 17
Referendums: In contrast to initiatives, referendums serve as a defensive instrument for the citizens by allowing them to reject a law that has been passed by the legislature. Referendums are also initiated by the citizenry through gathering of signatures.18
Plebiscites: To plebiscites counts those MDDs that allow authorities to pose a question to the citizenry to answer. Some scholars have argued that plebiscites can’t be viewed as an MDD since they are not necessarily related to popular sovereignty in its traditional sense. According to the definition provided of direct democracy above, these mechanisms are to be viewed as MDDs since the citizens still exercise their sovereignty during the vote itself. In contrary to initiatives and referendums, plebiscites are initiated by the government.19
2.1.2 Direct democracy and policy responsiveness
When studying the relationship between public opinion and public policy, it is important to distinguish between two forms of policy representation: responsiveness and congruence. These concepts capture different dimensions of democratic performance. Responsiveness means that there is a positive correlation between opinion and policy, while congruence means that the policy actually matches the majority opinion.20 E.g. it is possible that countries with a liberal public view on abortion have liberal policies (suggesting high responsiveness) without the policy matching the exact public opinion on abortion (suggesting a lack of congruence). This thesis will only focus on the responsiveness between public opinion on abortion and abortion policy. The main reason for not studying the congruence is that the public opinion measurement does not perfectly match the policy measurement used in the study. To be able to study congruence it is necessary that the criteria for measuring policy and public opinion match each other. This is not the case in this study, which will be clear when presenting the data.
17 Ibid. 18 Ibid. 19 Ibid.
9
As mentioned in the introduction, scholars are divided regarding how direct democracy affects the policy responsiveness of public opinion. At first sight it might seem obvious that the presence of MDDs give public policies better reflecting the public opinion, since it is a way of giving voters a direct voice regarding different policy areas. One might also argue that they should erase the possibility of politicians choosing policies incongruent with the will of the majority, since referendums give the voters a way to reject laws already passed.21 This might not always be the case and different scholars, e.g. formal theories of social choice have found that direct democracy, and voting in general, is complicated.
A general critique of the voting procedure itself is the problem with preference aggregation. Preference aggregation is often brought forward as an obstacle for the will of the majority to get a breakthrough in democratic institutions. Historically the Condorcet Paradox, stating that the collective preference often is irrational even if it consists of transitive, rational individual preferences has been brought forward as problematic. The implications of this is that voting does not necessarily lead to an outcome which represents the will of the majority, which gives the voters incentives to strategic voting. 22 Second, Arrow’s Impossibility Theorem demonstrate that it is impossible to rank individual preferences into one societal preference and at the same time fulfil basic democratic criteria for democratic elections.23 Even though these theories first and foremost focus on general elections, they are also applicable on MDDs such as initiatives, referendums and plebiscites since these mechanisms provide the same relevant circumstances as general elections, i.e. multiple-choice questions (more than two alternatives) where the voters register their preference on a ballot. According to these theories, MDDs do not necessarily enhance the correlation between public opinion and public policy.
The main discussion in more recent literature on the topic direct democracy and its impact on the relationship between public opinion and public policy has focused more specifically on the following question: Is direct democracy a tool for a few, special interests or for the many to impose their will through voting on their preferred (majoritarian) desires.24 Studies have been conducted giving support for both sides of the debate. Among the opponents, i.e. those arguing that direct democracy distorts the relationship between public opinion and public
21 Arceneaux, “Direct Democracy and the Link between State Public Opinion and State Abortion Policy”, p. 373. 22 University of Michigan, Preference Aggregation, [video], Coursera,
https://www.coursera.org/learn/model-thinking/lecture/1kmXf/preference-aggregation, accessed May 2, 2018.
23 Kenneth J Arrow, ”A Difficulty in the Concept of Social Welfare”, Journal of Political Economy, vol. 58, no.
4, 1950, p. 345f.
10
policy, authors have argued that wealthy organized groups manipulate the MDD process to generate an outcome that will benefit their own narrow interest. A well-known example of an author advocating this view is David Broder who in his book Democracy Derailed: Initiative Campaigns and the Power of Money concluded:
“the experience with initiative process at the state level in the last two decades is that wealthy individual and special interest – the targets of the Populist and Progressives who brought us the initiative a century ago - have learned all too well how to subvert the process to their own purposes.”25
This quote aims at a phenomenon that in research has been called “the initiative industry”, which literally buys citizen’s signatures – the collector is paid for each valid signature.26 Money can also be used to shape and confuse voters’ attitudes through e.g. mass media, which is a possible canal to use for those wealthy enough to have the access.27 According to this view, direct democracy can’t be expected to enhance the relationship between public opinion and public policy, but instead give opportunity to an influential minority to shape politics.
It is important to note that this kind of initiative industry first and foremost has been found in USA, and differs from evidence gathered in other countries. In Switzerland and Uruguay for example, which both have a high level of direct democracy, research has not found any signs of a signature gathering industry. This appears to indicate that this kind of industry is isolated to USA, but research has shown that it is not that simple. For example, in Colombia the weekly magazine Semana reported that at least one company received approximately 3.50 American dollars per signature gathered to support a second constitutional re-election of the president during the winter 2008.28 It is hard to find any obvious motives for such industry regarding abortion policy. Even if there are institutions with strong opinions regarding abortion in Latin America, mainly the catholic church trying to uphold restrictive abortion
legislation, there has been no sign of an initiative industry connected to the policy area.
25 David Broder, Democracy Derailed: Initiative Campaigns and the Power of Money, New York: Harcourt Inc.,
2000, p. 243.
26 Altman, Direct Democracy Worldwide”, p. 54.
27 Tomas Cronin, Direct Democracy: The Politics of Initiatives, Referendum and Recall, Cambridge: Harvard
University Press, 1990, p. 90.
11
Quite a few studies have found that direct democracy has little or no impact on the relationship between policy and public opinion. E.g. Lax and Phillips concludes in their study The Democratic Deficit in the States that the citizen initiative does not enhance policy responsiveness.29 They examine the policy responsiveness and congruence on various issues in USA, and the results show that salience of the issue and policy-specific opinion has the largest impact, while the impact of initiatives are negligible. The same authors have conducted another study examining gay rights with similar results. The results show that direct democracy does not significantly affect the adoption of gay rights policy in one way or the other.30 Note that both of these studies measured direct democracy as the allowance of initiatives.
On the other hand, there are also several studies supporting that direct democracy enhance the policy responsiveness to public opinion. John G Matsusaka shows this in his book For the Many or the Few: The Initiative, Public Policy, and American Democracy. Matsusaka conducts a case study examining tax policies among the American States. The results show that states with initiatives spent and taxed less than the states without initiatives. Also, they raised more money from user fees than taxes and decentralized spending from state to local government. Opinion surveys conducted during this period show that the majority of people supported these policy changes, which gives support to the view that initiatives promote policies reflection the public opinion rather than reflecting a few special interests.31 Also Arceneaux’s study support this, confirming that direct democracy in the form of referendums and initiatives, gives better representation of the public opinion on abortion policy among American states.32
The most direct way in which direct democracy enhances policy responsiveness is through the use of MDDs, as initiatives, referendums and plebiscites. But there are also other indirect effects of direct democracy that has been found enhancing the correlation between the variables. For example, studies have shown that initiative votes may give legislators more
29 Lax and Phillips, “The Democratic Deficit in the States”, p. 164.
30 Jeffrey R. Lax and Justin H. Phillips, “Gay Rights in the States: Public Opinion and Policy Responsiveness”, American Political Science Review”, vol. 103, no. 3, 2009, p. 383.
31 John G Matsusaka, For the Many or the Few: The Initiative, Public Policy, and American Democracy,
Chicago: University of Chicago Press, 2008, p. 3.
12
accurate information about voter preferences than is available where initiatives are not used, leading to policy better matching the public opinion.33
Accordingly, there is a controversy between existing scholars. Previous research supports that direct democracy might make public policy more responsive to public opinion, but it also suggests that it might distort the relation, making policies representing only a few, special interests. With previous research being as divided, it is difficult to predict how direct democracy will affect the policy responsiveness of public opinion on abortion in Latin America. But based on research on the same issue, mainly Arceneaux’s study, the hypothesis of this paper is that direct democracy will enhance the abortion policy responsiveness of public opinion on abortion. Even though this study examines a different region, there are no obvious reasons that this will generate other results, and there is no sign of an initiative industry distorting the results.
2.1.3 Direct democracy measurements
When it comes to measuring direct democracy there is, as been pointed out, also a methodological discussion. The introduction mentioned the dummy variable measurement being criticized for being too stiff, not taking all important indicators of direct democracy into account. This section will give a deeper explanation on existing research on the topic.
The main critique towards the dummy variable measurement is that it can easily underestimate the effect of direct democracy, leading to validity problems. For example, Arceneaux’s study was later replicated by Shaun Bower and Todd Donovan, but with other measurements of direct democracy.34 They present 5 different measures of direct democracy, and then examines how the different measurements affect the results using the same data as Arceneaux. The measures used are:
1. A dummy variable where 1 = a direct democracy state (initiative or referendum) and 0 = otherwise.
2. A dummy variable representing just initiatives states (1 = state has initiatives, 0 = otherwise)
33 Shaun Bowler and Todd Donovan, “Measuring the Effect of Direct Democracy on State Policy: Not All
Initiatives Are Created Equal”, p. 246
34 Bowler and Donovan, “Measuring the Effect of Direct Democracy on State Policy: Not All Initiatives Are
13
3. A legislative insulation index, taking into account how much a legislature can change an initiative that has been approved by voters.
4. A qualification difficulty index, taking into account the difficulty of qualifying a measure for the ballot.
5. The average annual state use of initiatives.35
The results of the study indicate that the way direct democracy is measured play a crucial role in what impact it has on the relationship between public policy and public opinion. The hypothesis that direct democracy makes public policy more responsive to public opinion was supported when measuring direct democracy as qualification difficulty and the frequency of initiatives used, but not supported when the dummy variable or the legislative insulation measure of direct democracy was used. The authors predicted this effect, since the dummy variables does not capture all dimensions of direct democracy. Based in these results, the hypothesis in this thesis is that using the index measure will result in a better policy responsiveness of public opinion than using the dummy variable measurement, since the dummy variable does not capture all direct democracy effects. This might also be why some of the research presented above did not find any correlation between direct democracy and policy responsiveness, since it measured direct democracy as only the allowance of initiatives. This refers to the studies conducted by Lax and Phillips mentioned earlier.
3. Data and research design
3.1 Data
One of the biggest challenges when studying Latin America, as many other developing regions, is the lack of data. Even though the latest edition of the Open Data Barometer concludes that the overall score for open data in Latin America has improved, it is due to improvements in only a few countries: Mexico, Uruguay, Brazil and Colombia. The rest of the countries showed the opposite results, an alarming decrease in open data.36 In conducting this thesis, the lack of data was especially prominent when trying to find public opinion data on abortion. Since abortion have been, and still is, a very controversial issue in many parts of Latin America not many attitude studies have been conducted on the topic. The lack of data is
35 Op. cit., p. 349 - 352.
36 The World Wide Web Foundation, “Open Data Barometer: Regional Report Latin America”, 3rd edition,
14
also problematic due to the low number of units of analysis available. The data needed was available for only 18 countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay and Venezuela. For a quantitative study it is desirable to have more units of analysis, since it might provide a greater and more interesting variation among the units of analysis and makes it makes it possible to control the correlation for other possible explanatory factors.37 One solution to this is to include two or more years in the analysis which will generate more units of analysis, but unfortunately this was not possible since the public opinion data only was available for 2015. In fact, all data used in the report was conducted during 2015. One might also argue that it would be better using policy data from a later year, e.g. 2016 or 2017 since the policy is predicted to be an effect of public opinion. To prove evidence of time order is usually hard in quantitative analyses, this is the case even for this thesis. Even here the lack of data is an obstacle, since the latest year with available policy data is 2015. As been mentioned in the introduction, this thesis does not make any claims of finding causal mechanisms, and therefore using data from the same year is not a problem for this study.
The three variables in this study: direct democracy, public opinion on abortion and abortion policy are presented separately. Each section provides an explanation on how the data has been conducted, a graphical illustration of the data and a review of its strengths and weaknesses.
3.1.1 Measuring direct democracy
The direct democracy data used in the report is retrieved from the Varieties of Democracy dataset, hence called V-dem.38 V-dem is a relatively new dataset released in 2016 and focuses on seven high-level principles of democracy: electoral, liberal, participatory, deliberative, egalitarian, majoritarian and consensual. The direct democracy data used in this paper is summarized in Figure 1, where both the index and dummy variable measurements of direct democracy are shown.
37 Svensson och Teorell, ”Att fråga och att svara”, p. 183.
38 Varieties of Democracy, Country-Year:V-Dem, 8 edition, [dataset], Gothenburg: V-dem institute, 2018,
15 0, 01 0, 38 0, 13 0, 01 0, 08 0, 2 0, 03 0, 37 0, 01 0,03 0, 07 0, 07 0,08 0, 12 0, 17 0, 28 0, 58 0, 3
DIRECT DEMOCRACY IN LATIN AMERICA
Figure 1. Showing the 2 measures of direct democracy used in this paper. The height of the bars shows the Direct Popular Vote Index score for each country. The colour of the bars shows the dummy variable measure, blue representing countries allowing initiatives and/or referendums, red representing countries not allowing initiatives and/or referendums.
Measuring direct democracy as an index
V-dem provides several indexes measuring different parts of democracy. For this paper, it is suiting to use V-dem’s Direct Popular Vote Index, which falls under the participatory principle mentioned above. This is an index ranging from 0 – 1 and measures the potential to direct democracy by looking at the following indicators:
16
decision made by citizens (binding or merely consultative), and (b) the frequency with which direct popular votes have been approved in the past.”39
An index like this gives the opportunity to capture more dimensions of the measured concept than is possible with a dichotomous variable. When measuring a term as nuanced as direct democracy, the index measure should be more appropriate, not least from a validity point of view. Validity is the same thing as measuring what we intend to measure,40 and a well-designed index does this with better accuracy than the dummy variable when it comes to this type of variables. On the other hand, a critique towards the index is the risk of lower reliability. The more nuanced something is, there is a greater risk that random errors during the measure might occur, since there is more to take into account than when measuring a dummy variable, or another variable not as nuanced.41
Measuring direct democracy as a dummy variable
Data from V-dem is used to create a dummy variable for direct democracy as well. The material is retrieved from two indicators of direct democracy measured by v-dem: Referendums Permitted and Initiatives Permitted. V-dem does not measure these indicators as dummy variables, instead they have a five-step scale, operationalized as follows: 0: Not allowed, 1: Constitutionally allowed but never utilized, 2: Allowed but with intervening institutional veto, 3: Allowed but non-binding, 4: Allowed and binding. This study creates a dummy variable of this v-dem information, coded as 0 for countries who does not allow initiatives and/or referendums and 1 for countries that allow initiatives or referendums or both. The countries categorized as a 0 according to V-dem receives a 0 in the dummy variable measure, and countries categorized as 1-4 by V-dem are coded as 1 in the dummy variable measure. It is quite clear that this kind of measure might lead to validity problems, since direct democracy have many more dimensions than only the legal provision of referendums and initiatives. In this particular case, there are also obvious validity problems connected with operationalizing direct democracy to a dummy variable since it does not match the definition of direct democracy used. An appropriate definition for using this dummy variable would be “the legal permission of initiatives and/or referendums”. But since the discussion is about the dummy variable not capturing all important aspects of direct democracy, this definition would
39 Varieties of Democracy, Codebook, 2018, p. 45, retrieved from:
https://www.v-dem.net/en/reference/version-8-apr-2018/, accessed April 27, 2018.
17
be too thin which is the reason why the thesis doesn’t use such a definition. With that said, the reliability is probably higher since it does not provide much space for own interpretations by the researcher.
3.1.2 Measuring public opinion
For measuring public opinion on abortion, this study uses data from Latinobarómetro. Latinobarómetro is a public opinion survey in Latin American countries that involves some 20,000 interviews, representing more than 600 million inhabitants. The Latinobarómetro Corporation is a non-profit organization and solely responsible for carrying out the production and publication of the data.42
In the Latinobarómetro questionnaire for 2015 the interviewed were asked to answer the following question: “Please use this card to tell me whether you think abortion can always be justified, never be justified, or somewhere in between.” The interviewed then places him- or herself on a scale between 1 and 10, where 1 represents “abortion should never be justified”, and 10 that “abortion should always be justified”.43 Figure 3 illustrates the mean of each countries public opinion on abortion, together with the standard deviation.44
Figure 2. Showing the mean opinion on abortion and standard deviation for each mean.
42 Latinobarómetro, Latinobarómetro, http://www.latinobarometro.org/lat.jsp, accessed April 20, 2018. 43 Latinobarómetro, Questionnaire 2015, retrieved from: http://www.latinobarometro.org/latContents.jsp,
accessed April 20, 2018.
44 Latinobarómetro, Latinobarómetro 2015, [dataset], Chile: Corporación Latinobarómetro, retrieved from:
http://www.latinobarometro.org/latContents.jsp, accessed April 22, 2018.
0 1 2 3 4 5
PUBLIC OPINION ON ABORTION IN
LATIN AMERICA
18
To let the interviewed place him- or herself on a scale is a common way of measuring attitudes. One might argue that open questions, where the respondent gets the opportunity to formulate their answer openly, would be better at capturing attitudes. At the same time, open answers are often connected with a higher amount on non-response. Open questions are also very time consuming to categorize and in big surveys it is not practically possible.45 Another aspect of measuring attitudes is the risk of low reliability. For example, two persons who have the same attitude towards abortion might place themselves differently on the attitude scale leading to arbitrariness in the results. This is very hard to prevent when measuring attitudes, but is not a problem as long as this error exist in similar ways among all interviewed.46 In this case, there is no reason to believe that there is any pattern in the way the respondents place themselves on the scale, and this should therefore not affect the result.
3.1.3 Measuring abortion policy
The data used for measuring abortion policy is retrieved from the World Population Policies Database.47 They measure abortion policy by looking at the following 7 legal provisions under which the Government permits abortion in the country: 1. To save a woman’s life, 2. To preserve a woman’s physical health, 3. To preserve a woman’s mental health, 4. In case of rape or incest, 5. Because of foetal impairment, 6. For economic or social reasons, 7. On request. The amount of legal grounds under which abortion is permitted is operationalized according to the Comparative Abortion Index 1 (CAPI1).48 In the CAPI1 index, countries are given a score based on the number of legal criteria accepted as grounds for abortion. This means a country that does not permit abortion on any ground gets a CAPI1 score of 0, and a country where abortion is legal on request (and therefore all other circumstances) receive a CAPI1 score of 7.
Figure 3 shows CAPI1 scores for countries in Latin America. Three countries, Chile, El Salvador and Nicaragua does not allow abortion under any circumstances. Only in two countries, Mexico and Uruguay, is abortion legal upon request. The case of Mexico needs some further explanation. Since Mexico is a federation, public policy varies among the states
45 Eliasson, Annika, Kvantitativ metod från början, 2nd edition, Lund: Studentlitteratur, 2010, p. 36–37. 46 Svensson and Teorell, Att fråga och att svara, p. 56–57.
47 World Population Policy Database, 2015 revision, [Dataset], New York: The UN Policy Section, retrieved
from: https://esa.un.org/poppolicy/wpp_datasets.aspx, accessed April 24, 2018.
48 Aliza Forman-Rabinovici and Sommer, Udi, “Reproductive health policy-makers: Comparing the influences
19
within the country. Only in one Mexican state, Mexico City, is abortion permitted on request and because of this the results might be misleading.49 This is accounted for in the analysis by making two regression analyses, one including Mexico and one excluding the country. This decision will be discussed further in the method section.
Figure 3. Showing the CAPI1 score for countries in Latin America.
3.2 Method
To examine the impact direct democracy has on the policy responsiveness to public opinion, ordinary least square (OLS) regression analyses are used, due to its strength of finding correlation between two or more variables.50 First, bivariate regression analyses are used to investigate the effect of the independent variable public opinion on the dependent variable abortion policy. This will show if there is a policy responsiveness of the public opinion, meaning that countries with higher public support for abortion are also likely to have more liberal abortion policy. The bivariate regression will also be divided using the dummy variable, to compare how the bivariate relation of public opinion-policy change depending on the value of the dummy variable.
49 Allyn Gaestel and Allison Shelley, "Mexican women pay high price for country's rigid abortion laws", The Guardian, October 1, 2014, https://www.theguardian.com/global-development/2014/oct/01/mexican-women-high-price-abortion-laws, accessed May 7, 2018.
50 Svensson och Teorell, Att fråga och svara, p. 159.
20
Afterwards, multivariate regression analyses are used to investigate the interaction effects of direct democracy on the responsiveness for each direct democracy measurement. Interactions variables are suitable to use when examining a conditioned hypothesis, i.e. a hypothesis in which the relationship between two or more variables depends on the value of one or more other variables.51 In the case of this thesis, the hypothesis is that the relationship between public policy and public opinion depends on the third variable direct democracy. Thus, the interaction terms used in the multivariate analyses are Public Opinion*Direct Popular Vote Index and Public Opinion*Dummy Variable. As in the bivariate analysis abortion policy is the dependent variable. Lastly, some expected values are calculated and analysed based on the results from the multivariate analyses.
As been mentioned in the data section, Mexico is problematic due to that the high CAPI1 score only is true for the state Mexico City and not representable for the rest of the country. It is also an outlier that might generate misleading results by shifting the regression line, especially in small datasets like the one used in this thesis.52 This is accounted for in the bivariate analyses by calculating two versions of the bivariate relationship between abortion policy and public opinion on abortion, one including Mexico and one excluding the country to see if it generates different results.
Two other extreme values can also be observed among the units of analysis – Uruguay and Chile. Uruguay has a high CAPI1 score of 7 and Chile a low score of 0. These countries are outliers and it can be argued, as the case of Mexico, that they are exceptions and will generate misleading results by shifting the regression line. In small dataset as this one, it is not reasonable to exclude more countries since there will be too few units of analysis left to draw any conclusions on. Also, excluding Uruguay and Chile is not expected to make too big of a difference since they probably “play each other out”. This will be clearer when looking to Figure 1.
This thesis uses no control variables in the multivariate regression analyses. This is mainly because small datasets don’t allow many control variables, since this is depending on the
51 Thomas Brambor, William Roberts Clark and Golder, “Understanding interaction models: Improving
empirical analyses”, Oxford Journals, vol. 14, no. 1, 2006, p. 64.
52 Colin Lewis-Beck and Michael Lewis-Beck, Applied Regression: An Introduction, 2nd edition, United States
21
degrees of freedom. Common control variables when examining abortion policy is for example religion, ideology of government and the number of female legislators in the parliament.53 Even though there is a weakness not including these control variables, this has been accounted for in previous research on abortion, not having a significant effect on the policy responsiveness of public opinion.54
4. Analysis
In this section the results from the regression analyses are presented and discussed. To start with the bivariate regression analyses are presented, first the general policy responsiveness of public opinion without taking direct democracy into account, and then for countries that permits initiatives and referendums vs. countries that does not permit initiatives and referendums. Then, the analysis moves on to the multivariate regressions, including interaction variables of public opinion and the different direct democracy measurements.
4.1 Bivariate regression of policy responsiveness of public opinion
Figure 4. Showing the bivariate relationship between public opinion and abortion policy. The regression coefficient 1,12, p-value is 0,138.
The bivariate analysis in Figure 4 indicates that there is a positive correlation between public opinion on abortion and abortion policy, i.e. stronger support for abortion rights in the public opinion correlates with more liberal abortion policy. This means the abortion policies are responsive of the public opinion on abortion, even though the p-value indicates that the
22
correlation is not significant. A non-significant result means it is not entirely sure that the correlation exists. Even though this does not say anything about the impact of direct democracy on the policy responsiveness, it is important to understand the original correlation before examining how this relates to direct democracy.
One easy way to examine interaction effects is to look at the same bivariate regression in two different groups: one for countries allowing initiatives and referendums and another for countries not allowing these MDDs, which is done in Figure 5 and Figure 6. Comparing the graphical illustrations of the analyses, it becomes clear that the policy responsiveness of public opinion is much greater in countries allowing initiatives and referendums than in countries that don’t. The regression coefficient for Figure 5 is 2,30 and significant on 0,05 level of significance, indicating it is quite certain that this relationship exists. When looking to Figure 6 the correlation between the variables almost disappears, with the negligible regression coefficient 0,03 (not significant). According to these bivariate regression analyses it seems clear countries with direct democracy have better policy responsiveness of public opinion than countries that don’t.
Figure 5. Showing the bivariate relationship between public opinion on abortion and abortion policy for countries without initiatives and referendums. The regression coefficient equals 2,30 and p-value is 0,0149 (significant on 0,05 level of significance).
23 4.1.1 Excluding Mexico
Figure 7. Showing the bivariate relationship between public opinion and abortion policy excluding Mexico. The regression coefficient equals 0,95 and p-value is 0,157.
24 Figure 8. Showing the bivariate relationship between public opinion on abortion and abortion policy in countries with initiatives and referendums excluding Mexico. The regression coefficient equals 2,04 and p-value is 0,0174 (significant on 0,05 level of significance).
Figure 8 displays the policy responsiveness of public opinion in initiative and referendum countries, excluding Mexico. As in Figure 7, the correlation is a little bit weaker when excluding Mexico, the regression coefficient decreases from 2,30 to 2,04 compared to when including the country. There is still a strong policy responsiveness of public opinion in countries allowing initiatives and referendums, and the correlation continues to be significant on 0,05 level of significance.
25
4.2 Multivariate regressions
The results of the bivariate and multivariate regression are presented in Table 1, which mainly examines the different ways of measuring direct democracy. The dependent variable is, as in the bivariate analysis above, abortion policy. Each model will be presented followed by a discussion of its results.
Model 1 shows the bivariate regression of policy responsiveness already displayed in Figure 4. The bivariate results have already been discussed in the previous section, but are displayed as a point of comparison. As been mentioned, there is a general policy responsiveness of public opinion but the correlation is not significant.
Model 2 and 3 displays multivariate regressions including public opinion and the two direct democracy measurements as independent variables. The effects are very different, probably because the relationship between the public opinion and direct democracy variables differ. Model 2 shows that the effect of public opinion on policy decreases when including the index measurement in the analysis, even though public opinion still has a positive effect of policy. This effect is still not significant. At the same time, the direct democracy index has a large and significant effect on policy. As it is unlikely that direct democracy itself increase or decrease public support for abortion the correlation between public opinion and policy should not be spurious, meaning that direct democracy is an underlying variable affecting both public opinion and policy.55 It is more likely that there is an indirect correlation between the variables, where the effect of public opinion is mediated by direct democracy.
Model 3 shows, as in Model 2, that both public opinion and direct democracy have a positive effect on policy, though here both the effects are significant. The effect of public opinion increases compared to Model 1, suggesting that the way direct democracy is measured, either by a continuous variable (the index) or a dummy variable, changes the public opinion – policy relationship. The index seems to decrease the effect of public opinion on policy whereas the dummy variable strengthens the effect. In Model 4 and 5 this is studied further, by including interaction variables.
26
Multivariate regression of the effect of direct democracy and public opinion on abortion policy Model 1 2 3 4 5 Public Opinion 1,12 (0,72) 0,86 (0,66) 1,43** (0,62) 0,03 (0,93) 0,30 (0,98)
Direct Democracy (Index) 6,09*
(2,92) -2,42 (11,06) Direct Democracy (Dummy) 2,39** (0,91) -4,29 (3,57) Public Opinion* DD Dummy Variable 2,27* (1,18)
Public Opinion*DD Index 2,49
(3,12)
Intercept 0,71 -0,96 -3,18 1,08 0,83
R2 1,13 - - - -
Adj.R2 0,08 0,24 0,33 0,43 0,22
N 18 18 18 18 18
Table 1. * p < 0,05 level of significance, p < ** = 0,01 level of significance (two tailed test). Standard errors are shown in parenthesis.
27
is consistent with the results displayed in Figure 5 and 6. The regression coefficient for the constitutional variables shows that the effect of only direct democracy is negative on policy, in this case generating more restricted abortion policy. For public opinion on the other hand, the effect on abortion policy disappears.
Model 5 displays similar results, a positive regression coefficient for the interaction variable Public Opinion*Direct Popular Vote Index indicating that a higher index score (i.e. more direct democracy) gives a higher policy responsiveness of public opinion. This is in line with the hypothesis. As in Model 4, the effect of only direct democracy is negative, resulting in more restrictive abortion policy, but together with public opinion the interaction variable shows that the effect is positive on policy responsiveness. Public opinion maintains a little effect on policy even when direct democracy is zero, in contrary to model 4 where the correlation disappeared. The regression coefficient for the interactive variable is bigger than in Model 4 indicating that when using the index measurement, the policy responsiveness is greater. This indicates that the index measurement captures the public opinion better than the dummy variable. But since the effect is not significant it is not certain that the correlation exists, making Model 4 stronger. The adjusted R2 is greater in Model 4 than Model 5, indicating that this model best explains the variations in policy among the units of analysis.56 4.2.1 Predicted Values
Knowing the regression coefficients for the interaction variables it is possible to calculate the expected values of each country’s abortion policy according to the following equation:
Y = α + β₁x + β₂z + β₃xz Where Y = expected policy x = public opinion z = direct democracy α = Intercept
β₁ = Regression coefficient for public opinion β₂ = Regression coefficient for direct democracy
β₃ = Regression coefficient for the interactional variable
The expected values are calculated for four countries: Argentina, Chile, Peru and Uruguay. These countries are chosen due to similar public opinion on abortion but different level of
28
democracy. Argentina and Peru share almost the same public opinion on abortion, but Argentina has a low Direct Popular Vote score and does not permit initiatives and referendums. Peru on the other hand have a relative high index score and permits both initiatives and referendums. Chile and Uruguay have the highest public opinion scores among the examined countries, but their level of direct democracy variaes heavily. The CAPI1 score for each country is displayed in table 2. By calculating the expected abortion policy for both measurements based on the information in table 1 and comparing these to the actual policy, it is possible to analyse two things: First, how the differences varies within he measurements and second, if there is a pattern between the measurements indicating that one of the measurements is better at capturing the policy responsiveness of public opinion.
Actual policy vs. predicted policy value
Country CAPI1 Expected Policy:
Dummy Variable Diff. Dummy Variable Expected Policy: Index Diff. Index Argentina 3 1,177 1,823 1,852 1,148 Chile 0 1,211 - 1,211 2,222 - 2,222 Peru 3 4,219 - 1,219 3,373 - 0,373 Uruguay 7 7,163 - 0,163 7,293 - 0,293
Table 2. Displaying the difference between expected policy and actual policy for both direct democracy measurements for Argentina, Chile, Peru and Uruguay.
29
to the hypothesis. Comparing both measurements, the index generates smaller difference regarding all countries except for Chile. Also, when it comes to Argentina the predicted policy is lower than the actual, whereas the opposite happens for the other three countries where the predicted policy is higher than the actual.
To sum up the results from the multivariate regressions, Model 4 indicates that the interaction variable of public opinion and the dummy variable measurements have a positive, significant effect on policy responsiveness of the public opinion. This means that in countries with initiatives and referendums the abortion policy reflects the public opinion better, which is in line with the hypothesis. Looking to the interaction variable consisting of public opinion and the index measurement, the results are not significant. Even if the correlation is stronger in reference to the regression coefficient, suggesting a better policy responsiveness of public opinion the results can’t be interpreted as true without being significant. This is not in line with any of the hypotheses, suggesting that the policy responsiveness should be stronger when using the index measurement. The predicted values displayed in Table 2 are useful for examining the difference between the actual policy and the predicted policy according to the results in the multivariate regression. The index measurement generates predicted values better matching the actual for all countries except Chile, but since these results still not are significant they can’t be considered true, since there is a risk that this relationship does not exists.
30
5. Conclusions
The aim of this thesis was to investigate if, and how, direct democracy affects the policy responsiveness of public opinion. It did so by answering the question: How does the opportunity to direct democracy affect the abortion policy responsiveness to public opinion of abortion in Latin America? The paper also strived to shed more light on the methodological discussion about measuring direct democracy by using two different measurements.
The results show that direct democracy has a significant impact on policy responsiveness when using a dummy variable measurement. When using the index measurement, the results also indicate that direct democracy enhances the policy responsiveness of public opinion, even though these results are not significant and can’t be perceived as certain. This is in contrast with the hypothesis, which predicted that both direct democracy measurements were going to enhance the policy responsiveness, i.e. be significant. Due to the insignificant results, there is also no support for the hypothesis that the index measurement has a bigger impact on policy responsiveness. It is important to keep in mind that the dummy variable does not capture all dimensions of direct democracy, and even if the results are significant one should consider if this measurement captures what is intended to be measured before using it.
31
6. References
Altman, David, Direct Democracy Worldwide, Cambridge: University Press, 2011.
Altman, David, “The Potential of Direct Democracy: A Global Measure (1900-2014),” Soc Indic Res, vol. 33, no. 4, 2017, p. 1207-1227.
Arceneaux, Kevin “Direct Democracy and the Link between State Public Opinion and State Abortion Policy”, State Politics & Policy Quarterly, vol. 2, no. 4, 2002, p. 372–387. Arrow, Kenneth J., ”A Difficulty in the Concept of Social Welfare”, Journal of Political Economy, vol. 58, no. 4, 1950, p. 328-346.
Bowler, Shaun and Donovan, Todd, “Measuring the Effect of Direct Democracy on State Policy: Not All Initiatives Are Created Equal”, State Politics & Policy Quarterly, vol. 4 no. 3, 2004, p. 345–363.
Broder, David, Democracy Derailed: Initiative Campaigns and the Power of Money, New York: Harcourt Inc., 2000.
Brambor, Thomas, Roberts Clark, William and Golder, Matt, “Understanding interaction models: Improving empirical analyses”, Oxford Journals, vol. 14, no. 1, 2006, p. 63-82. Carmine, Edward G., and Stimson James A., “The American Political Science Review”, vol, 74, no. 1, 1980, p. 78-91.
Cronin, Tomas, Direct Democracy: The Politics of Initiatives, Referendum and Recall, Cambridge: Harvard University Press, 1990.
Eliasson, Annika, Kvantitativ metod från början, 2nd edition, Lund: Studentlitteratur, 2010. Forman-Rabinovici, Aliza and Sommer, Udi, “Reproductive health policy-makers:
Comparing the influences of international and domestic institutions on abortion policy”, Public Administration”, 2018, p. 185-199.
Gaestel, Allyn and Shelly, Allyn, "Mexican women pay high price for country's rigid abortion laws", The Guardian, October 1, 2014,
https://www.theguardian.com/global-development/2014/oct/01/mexican-women-high-price-abortion-laws, accessed May 7, 2018. Latinobarómetro, Latinobarómetro, http://www.latinobarometro.org/lat.jsp, accessed April 20, 2018.
Latinobarómetro, Latinobarómetro 2015, [dataset], Chile: Corporación Latinobarómetro, retrieved from: http://www.latinobarometro.org/latContents.jsp, accessed April 22, 2018. Latinobarómetro, Questionnaire 2015, retrieved from:
http://www.latinobarometro.org/latContents.jsp, accessed April 20, 2018.
32
Lax, Jeffrey R., and Phillips, Justin H., “The Democratic Deficit in the States”, American Journal of Political Science, vol. 56, no, 1, 2012, p. 148-166.
Lewis-Beck, Colin and Lewis-Beck, Michel, Applied Regression: An Introduction, 2nd edition, United States of America: Sage Publications, 2016.
Malkin, Elisabeth, ”They Were Jailed for Miscarriages. Now, Campaigns Aims to End Abortion Ban”, nytimes.com, April 9, 2018,
https://www.nytimes.com/2018/04/09/world/americas/el-salvador-abortion.html, accessed May 2, 2018.
Matsusaka, John G., For the Many or the Few: The Initiative, Public Policy, and American Democracy, Chicago: University of Chicago Press, 2008.
Mooney, Christopher Z., “The public clash of Private Values.”, In The Public Clash of Private Values: The politics of Morality Policy, edited Christopher Z Mooney. New York: Chatham House, 2001, page 3-2.
Svensson, Torsten och Teorell, Jan, “Att fråga och att svara”, Malmö: Liber, 2007.
The World Wide Web Foundation, “Open Data Barometer: Regional Report Latin America”, 3rd edition, 2016.
University of Michigan, Preference Aggregation, [video], Coursera,
https://www.coursera.org/learn/model-thinking/lecture/1kmXf/preference-aggregation, accessed May 2, 2018.
Varieties of Democracy, About V-dem, 2018, https://www.v-dem.net/en/about/, accessed April 27, 2018.
Varieties of Democracy, Codebook, 2018, retrieved from: https://www.v-dem.net/en/reference/version-8-apr-2018/, accessed April 27, 2018.
Varieties of Democracy, Country-Year: V-Dem, 8 edition, [dataset], Gothenburg: Varieties in Democracy Institute, 2018, retrieved from: https://www.v-dem.net/en/data/data-version-8/, accessed April 27, 2018.
33
Appendix: Dataset
Country Opinion. Policy. DD1 DD2
