A New Political Divide?
Laméris, Maite Dina
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2019
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Laméris, M. D. (2019). A New Political Divide? Political ideology and its economic implications. University of Groningen, SOM research school.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Political Divide?
Political ideology and its
economic implications
Layout and cover design by
Design Your Thesis, www.designyourthesis.com Printed by Ridderprint B.V., www.ridderprint.nl ISBN 978-94-034-1324-2 978-94-034-1323-5 (ebook) © 2019 Maite D. Laméris
A New Political Divide?
Political ideology and its economic implications
PhD thesis
to obtain the degree of PhD at the University of Groningen
on the authority of the Rector Magnificus Prof. E. Sterken
and in accordance with the decision by the College of Deans This thesis will be defended in public on Thursday 14 February 2019 at 16:15 hours
by
Maite Dina Laméris
born on 15 October 1990 in Groningen
Co-supervisor Dr. R.M. Jong-A-Pin Assessment Committee Prof. C. Bjørnskov Prof. J. de Haan Prof. K. Kis-Katos
I remember that about four years ago I had to decide whether or not to apply for a PhD. Somehow I found this a very difficult decision. On the one hand doing a PhD seemed to be a super challenging next step, on the other hand I was not sure whether working on an individual project for three years was really something for me. I could imagine that it might be lonely to do research, it seemed to be mostly individual work. And for the people who know me, it probably does not come as a surprise that this is not really how I would describe my ideal job. But, the opportunity to work on the topic of my choice proved to be decisive. I took the chance and went for it.
Now a bit more than three years after starting my PhD, I am very happy that I did. Time flew by. Time in which I learned quite a bit about myself, in which I was challenged to think outside the box, and in which I worked harder than I did before. I would be lying if I would say that none of my worries came true. As many others who did a PhD probably recognise, I went through some deep lows, although these were almost always followed by amazing highs. And sometimes it was indeed a bit lonely. But the people around me made this feeling go away quickly and made my PhD an experience I would not have missed for the world! This is why I want to take this opportunity to thank them.
First of all, I want to thank my supervisors, Harry Garretsen and Richard Jong-A-Pin. From the beginning to the end, I had the feeling you believed in my project, even when others did not. This gave me the confidence I needed to also believe I could do this. I very much enjoyed our regular meetings. You always gave me helpful and constructive comments, from which I could continue my work. I liked that you pushed me to write papers, not necessarily chapters for a thesis. This definitely led to one of these papers being published, and the others being either under review or ready to be send out for reviewing. It was also reassuring that you never seemed to worry about whether I would be able to merge these papers into a thesis. This felt as a vote of confidence, which I needed at times. Harry, your experience and ability to see the bigger picture made my thesis into what it is now. Richard, your critical eye definitely improved my thesis, and your enthusiasm about my projects and passion about doing research in general is contagious. It was an enormous pleasure working with you both. I am also very grateful for all other academics that contributed to my thesis. Firstly, I want to thank professor Bjørnskov, professor de Haan and professor Kis-Katos for
Guillaume Méon and Anne-Marie van Prooijen for their co-authorship. I really enjoyed working with you, exchanging thoughts and creating papers that are worth reading, if I may say so myself. I was inclined to think that doing research is individual work, but you showed me that it does not have to be. Thank you for making it teamwork. Pierre-Guillaume, I consider myself lucky to have crossed paths with you in Gargnano, which led to us working together. I appreciate that you invited me to Brussels and gave me the opportunity to present my work in a seminar at ULB. As a junior researcher, it was really exciting and I very much enjoyed my stay. I would also like to thank Adriaan Soetevent and Noémi Péter for their advice and guidance surrounding the experiment, and Oliver Herrmann and Nannette Stoffers for their excellent research assistance.
Next to those that academically contributed to my thesis, I also want to thank those that contributed to it in other ways. During my PhD, a tight-knit little community of GEM PhDs (and a few that were adopted as such) formed. Without these fellow PhDs I would definitely not have enjoyed these past three years as much as I have. Aobo, Bingqian, Joeri and Kailan, I really enjoyed working alongside you and hopping in your offices now and then to ask for your advice (and gossip a bit). And of course, thank you for always eating my home-baked cakes and saying they were tasty (even when I knew they were not). Charlotte, Jiasi and Romina, you were (and are) wonderful office-mates. Daan, Ferdinand, Johannes, Nikos, Stefan and Timon, I find you amazing people. Thank you for letting me cry in your office (Ferdinand, that’s you), for telling me off when I am too curious for my own good (Nikos, this one’s for you), for pretending to listen while I am explaining things you probably understand better (Stefan, that’s for you), for sharing your stories with me and listening to mine, and for not only being my colleagues, but also my friends.
There are also people outside academia that I want to thank for being there for me during my PhD. Sanne, you always see the bright side of things, which made any obstacles I was facing seem smaller. The way you handle what life throws at you is admirable and helped me put my PhD-‘problems’ into perspective. Juliette, your
about my next step. Amanda, I love how you always listened to me talk about models and variables without hesitation, and tried to make sense of what I was talking about. Thank you for having my back and being in my corner when I needed it most. And to the other Liefjes, your support and encouragement during these years were wonderful. The last people I want to thank are my family, without whom I would not be where I am now. First my two big sisters, Lodi and Joran. You are an incredible support to me and without you these past years would have been a lot harder and a lot less fun. Your sisterly advice is always just a phone call away, and knowing this is sometimes already enough. I love how you reminded me to celebrate the ups, which I do not always do, and of course, how you then celebrated these with me. And I love how you helped me through the lows by simply listening and being there for me. Ger, your calm and collective attitude really helped me to stay grounded throughout this adventure. Your dry sense of humour always put a smile on my face, also when I did not really feel like smiling. And Anna, even though you will never be able to read this, I want to thank you for your part in making me become the person I am today. The support I get from all of you feels unconditional to me, which is an amazing feeling. Knowing that you are proud of me, makes me feel proud of myself.
Enough now with the mushy talk, doing a PhD and writing my thesis was really awesome. And even though I am a bit sad that I can no longer call myself a student, I am excited for what else is coming my way!
Groningen, November 2018
Maite
11
1. Introduction27
2. Political ideology and theintragenerational prospect of upward mobility
57
3. An experimental test of the validity of survey-measured political ideology97
4. On the measurementof voter ideology
133
5. How students’ beliefsand values vary across and within disciplines
187
6. Conclusion197
References213
Appendix229
Dutch summary / Nederlandse samenvattingThe traditional debate on political ideology has been dominated by the view that political preferences are either left or right. Not only scholars, but also politicians and journalists commonly refer to left-wing versus right-wing when discussing voters’ preferences, political parties and policies. Moreover, the terms left and right are still often used in political and academic debates. However, in light of recent events and evolutions in the political landscape, this traditional view of ideology has increasingly come under attack. Does this, however, mean that the left-right divide is ill-suited for the contemporary political environment? And if so, what would be a better-fitting, alternative ideological divide?
Events such as the election of Donald Trump, the vote in favour of Brexit and the European migrant crisis have shown that contemporary politics is no longer only focused on traditional left versus right topics. Issues as redistribution and the amount of government involvement in the market, along which the left versus right divide has conventionally been classified, lost their prominent position in the political debate. It is even claimed in popular media that the left versus right political classification has been replaced by one along open versus closed lines.1 According to these media,
the political landscape is nowadays divided along issues as migration, protectionism, (anti-)establishment and cultural change, instead of redistribution, economic equality and the level of government involvement in the economy. Moreover, the former topics have received abundant attention during recent elections in Europe, such as the 2017 Dutch and the 2018 Italian general elections.2 This apparent new political divide even
led to the formation of an Italian government by the extreme ‘right-wing’ Lega and extreme ‘left-wing’ Five-Star Movement. This suggests that traditional left versus right topics have been placed on the back burner.
The emergence and electoral successes of many contemporary populist parties in the European political landscape is an additional evolution suggesting that left and right might be outdated as political dividers. Moreover, it suggests an electorate that is confused about what the concepts ‘left’ and ‘right’ encompass. Nowadays some of the largest political forces in their countries, parties like the Dutch PVV, the French Front National, the Italian Lega, and the Austrian FPÖ have been gaining momentum. In 1. See for example these articles in the Economist of 30 July 2016: ‘The new political divide’ and ‘Drawbridges up’, 30 July 2016.
2. See the following articles published in the Guardian for an overview: ‘Dutch elections: all you need to know’ (2 March 2017) and ‘Italy’s election: who will win and why does it matter?’ (4 March 2018).
the academic as well as popular debate, these parties and their constituents are referred to as right-wing. Yet, they either find economic issues to be inferior to their social and cultural goals or do not support traditional right-wing economic policies (Mudde (2007)). Furthermore, voters seem to interpret the left-right scale nowadays on cultural and immigration grounds (de Vries, et al. (2013)). This raises the question of whether the left-right political distinction is still suited to evaluate partisan differences in the contemporary political landscape.
In academics this issue has also received considerable attention. However, whereas it seems to be receiving scrutiny by the media and popular debate only in recent years, research on the use(fulness) of the left-right distinction in political beliefs goes back to at least the fifties. Eysenck (1954) was one of the first to identify multiple dimensions of political beliefs, one of which is similar to the traditional, i.e. economic, left-right distinction. His work was followed by Converse (1964), who argued that the mass of the electorate does not have political attitudes, which follow the logical and coherent structure that is assumed by left-right ideology. Furthermore, Lipset (1960) claimed that the conflicts needed for voters to manifest themselves along ideological lines declined to such an extent that there were no real differences anymore between the left and the right.
The work of authors as Eysenck (1954), Converse (1964) and Lipset (1960) spurred research on the structure of political beliefs. If a one dimensional, left-right representation of ideology was not able to coherently structure political attitudes of individuals, perhaps multiple dimensions could. Conover and Feldman (1984) showed that individuals structure the same political information in different ways that cannot be simplified to one dimension. Moreover, Carmines and D’Amico (2015) claimed that individuals’ ideology is the result of personal values and beliefs, and as such, is not constrained by a framework such as the left-right one. This claim is supported by Feldman (2013), who asserted that there exists a diverse set of values. As political attitudes find their origin in such values (see also Feldman (1988); Rokeach (1973)), there is no reason to believe that a single dimension can structure them. Many more studies examining the structure of political attitudes followed (see e.g. Feldman & Johnston (2014); Layman & Carsey (2002); Otjes (2017); Treier & Hillygus (2009)).
number of studies (…) have examined the dimensionality of political beliefs and issue preferences among people in many different countries. In virtually no case is a single factor (left-right) model an adequate fit to the data.’
Despite these insights from political science and political psychology, research in economics has continued to rely on left-right distinctions of political ideology when studying its economic implications. Think, for example, of the median voter theorem of Downs (1957) and the partisan models of Hibbs (1977) and Alesina (1987), or the theory on social mobility and redistributive preferences by Piketty (1995). These all rely on left-right distinctions of political beliefs. Not only is the left-right classification used in theoretical work, many scholars in economics also use it to empirically evaluate the impact of political preferences on the economy. Left-right ideology is, for example, used to evaluate happiness and well-being (Bjornskov, et al. (2013); Dreher & Ohler (2011)), voting behaviour (Ansolabehere & Socorro Puy (2016); Garcia-Vinuela, et al. (2018)), and preferences for economic policy (Boeri, et al. (2001); Giesenow & de Haan (2018); Pitlik, et al. (2011)). Moreover, a complete strand of literature in political economy is dedicated to study how left-right political attitudes affect voter’s preferences for redistribution (e.g. Alesina & Angeletos (2005); Alesina & Giuliano (2011); Alesina, et al. (2018); Buscha (2012); Olivera (2015); Page & Goldstein (2016)).
In this thesis, I study the role of ideology in political economy research, taking into account the evolutions in the contemporary political environment, as well as research on the structure of political beliefs. I operate in three different areas that have political ideology as common theme. Firstly, I examine how left-right political beliefs impact redistributive preferences by considering its indirect effects - following the convention regarding the measurement of political ideology. By taking into account indirect effects of political ideology, I am improving upon existing work in the field. Secondly, I focus on measurement issues surrounding political beliefs taking into consideration the findings on ideology from political science and political psychology. I start with a test of the validity of the conventionally used left-right measure of political ideology. Then, I examine the dimensionality of political ideology and propose an alternative measure of political beliefs. Thirdly, I study individual heterogeneity in underlying
sources of ideology by examining beliefs and values to get better insight into what drives political attitudes. I discuss each of these parts in more detail in the remainder of this chapter.3
1.1 INDIRECT EFFECTS OF POLITICAL IDEOLOGY
Many scholars investigate how political preferences affect economic outcomes. This is either done on a governmental-level, in which (the outcome of) economic policy is often the variable of interest (e.g. Belke & Potrafke (2012); Bjornskov (2005); Cukierman & Tomassi (1998); Tavares (2004)), or on an individual-level, in which preferences for such outcomes and policies are generally under investigation (e.g. Bodenstein & Faust (2017)), Boeri, et al. (2017); Pitlik, et al. (2011); Scully, et al. (2012)). One particular subject has been the focus of attention of many: (preferences for) redistribution. As major shares of government budgets are spent on redistributive transfers, it is essential for policy-makers and researchers to know what determines public support for it.4 In Chapter 2, I add to the existing literature on redistribution
by studying political ideology, income mobility and redistributive preferences. There is a long history of studies into the determinants of redistribution and the influence of political attitudes. Since the median voter theory of redistribution by Meltzer and Richard (1981), research into what factors influence redistribution has taken off. In Meltzer and Richard’s (1981) model, however, people were modelled as self-interested and there was no role for political beliefs. One of the first to incorporate political effects into models of redistribution were Dixit and Londregan (1998), who modelled voters motivated by a concern for inequality. This started a new strand of literature, theoretical and empirical, into the effects of political attitudes on redistribution (see e.g. Case (2001); Feld (2000); Roemer (1998); Roemer (1999)). More recently, the focus has been on explaining voter’s preferences for redistribution
using political ideology as one of the motives for redistribution (e.g. Alesina & Angeletos (2005); Alesina & Giuliano (2011); Olivera (2015); Page and Goldstein (2016)).
An additional factor influencing redistributive preferences is the prospect-of-upward-mobility (POUM), established by the work of Benabou and Ok (2001). According to the POUM-hypothesis, individuals expecting future upward income movements might rationally demand lower levels of redistribution, even though, based on their current income they would benefit from it. Since the seminal paper by Benabou and Ok (2001), several authors have studied the effect of upward income mobility expectations on redistributive preferences. Ravaillon and Lokshin (2000), for example, find a substantial effect of both upward and downward mobility on support for redistribution among families that are currently already well-off. Checchi and Filipin (2004) find experimental evidence for the POUM-hypothesis. Furthermore, Alesina and La Ferrara (2004) find that both subjective expectations and objective measures of income mobility are able to explain why poor individuals might demand a low rate of redistribution. Using only subjective measures of income expectations, Corneo and Gruener (2002), Rainer and Siedler (2008) and Cojocaru (2014) draw similar conclusions.
Adding to these findings, students of the POUM-hypothesis have focussed their attention to the role of political beliefs when examining mobility expectations. Buscha (2012) finds that individuals that expect their income to increase in the future are more likely to be right-wing, whereas those that expect their income to decrease are more likely to be left-wing. Furthermore, expecting upward income mobility increases the likelihood of individuals to vote for a conservative party. Buscha (2012), however, does not consider redistributive preferences, even though his results suggest an indirect link between mobility and ideology. Alesina, et al. (2018) do take this indirect effect into account in their paper on the effect of inter-generational mobility on preferences for redistribution. They study the POUM-effect, while allowing for differences between left-wing and right-wing individuals and find a robust effect of mobility on redistributive preferences that is conditional on political beliefs.
In Chapter 2, I add to this literature by studying political ideology, the POUM-effect and preferences for redistribution in an intra-generational setting. Different from Alesina, et al. (2018), I consider the influence of life-cycle earnings by focussing on the individuals for which the POUM-effect is most relevant. This enables me to give
a more accurate picture of the conditionality of the POUM-effect. Moreover, where Alesina, et al. (2018) examine perceptions of country-level opportunities for income mobility, I consider individual expectations. That is, I examine how political attitudes affect the relation between an individual’s personal mobility expectations and his/her support for redistribution.
The data I use consists of a cross-section of individuals and is gathered using a survey set out among the Dutch population. Political preferences are measured using self-reports of ideology on a left-right scale. The findings show that there is a statistically significant POUM-effect on preferences for redistribution. That is, individuals who expect their income to increase over time have preferences for significantly lower redistribution. However, I find that this effect is conditional on political ideology. Only right-wing individuals’ demand for redistribution is negatively affected by expectations of upward income mobility. For left-wing individuals, it holds that preferences for redistribution are independent of their expected income mobility. Regardless of what they expect to earn in the future, redistribution is a preferred outcome for them.
1.2 MEASUREMENT ISSUES IN POLITICAL IDEOLOGY
In Chapters 3 and 4, I address two issues related to the measurement of political ideology. In Chapter 3, I start by testing whether the conventionally used left-right measure of ideology is a valid predictor for beliefs along traditional, i.e. economic, ideological lines. After finding evidence opposing such predictive validity, I challenge the assumption that ideology can be measured along a linear, left-right dimension in Chapter 4. I argue that more dimensions are needed to accurately depict political preferences and propose an alternative measure of political beliefs.
In Chapter 3, I test the predictive validity of left-right political ideology. It is conventional in political economy to rely on left-right measures of political beliefs. Yet, in doing so you not only have to assume that ideology is one-dimensional, you also assume that the measure captures underlying preferences and is able to predict behaviour. Moreover, the majority of research on left-right ideology uses survey measures. Survey-measurement, however, can be subject to self-serving biases,
et al. (2011)). A natural question that follows is: how can we be sure that the left-right survey-measure captures what we believe it does? In Chapter 3, I aim to answer this question by testing the validity of survey-measured political ideology using an incentivised real-effort distribution experiment.
Using an incentivised experiment to examine the predictive validity of survey-measures is an accepted method (see e.g. Armantier, et al. (2015); Dohmen, et al. (2011); Falk, et al. (2016); Fehr, et al. (2003); Glaeser, et al. (2000); Vischer, et al. (2013)). However, to the best of my knowledge, there is no study yet that specifically focusses on validating left-right political ideology. I conduct an incentivised real-effort distribution experiment that is designed to capture preferences regarding the equality-efficiency trade-off, which is at the core of the left-right divide (Jost (2009)). Within the context of this experiment, I test whether self-reported left-right ideology can predict these preferences. Moreover, by including a real-effort stage, I am able to take potential differences in behaviour due to entitlement concerns into account.
From existing studies we know that behaviour in experiments in which subjects decide over earned wealth differ from behaviour when they decide over given wealth (e.g. Barr, et al. (2015); Cappelen, et al. (2013); Cherry, et al. (2002); Durante, et al. (2014); Engel (2011); Erkal, et al. (2011); Gee, et al. (2017); Krawczyk (2010)). However, not many studies go into detail about what could explain this difference (exceptions being Barr, et al. (2015) and Cappelen, et al. (2013)).5 I consider left-right ideology
as a driver of these behavioural differences. It has been shown in existing work that views regarding (in)equality, redistribution and efficient outcomes are influenced by beliefs about (the role of) effort and luck, both on an individual and societal level (e.g. Alesina & Angeletos (2005); Alesina & Giuliano (2011), Benabou & Tirole (2006); Fong (2001); Lefgren, et al. (2016); Piketty (1995); Varian (1980)). Simultaneously, it is either explicitly stated or implicitly assumed that there are differences between left and right-wing individuals in how they think about effort and luck, and consequently, their role in determining success or income (see e.g. Alesina & Angeletos (2005); Benabou & Tirole (2006); Jost, et al. (2009); Piketty (1995)). Hence, I expect differences in behaviour of left-wingers and right-wingers. This difference will depend on earnings being determined by luck or by effort, in which entitlement concerns come into play.
5. Barr, et al. (2015) consider economic status as an explanation, whereas Cappelen, et al. (2013) take into account needs considerations.
The experiment I conduct is a simple distribution experiment, which captures equality versus efficiency preferences. It relates to the study of Engelmann and Strobel (2004). This study examines whether behaviour is motivated by inequality aversion, efficiency considerations or maximin preferences. I add to their work by including a real-effort stage. Accordingly, the experiment is an incentivised two-stage real-effort experiment. In this experiment, a decision-maker distributes earnings over two anonymous recipients and him/herself. In the first stage, (s)he either receives these earnings as ‘manna-from-heaven’ or earns them during a task. In the second stage, the decision-maker distributes these earnings over the group. (S)he has a choice between two distributions: an equal, but inefficient distribution, or an unequal, but efficient one. Within this stage, I vary the earnings that the decision-makers receive when choosing a distributive outcome. This allows me to test whether my results are robust to small monetary incentives.
I expect behaviour to be in line with self-reported ideology, i.e. I expect left-wing decision-makers to prefer the equal distribution over the efficient one; and vice versa for right-wing decision-makers. However, taking into account ideological differences regarding the role of effort and luck, I expect left-wing decision-makers to do so under luck and right-wing decision-makers under effort. I find that self-reported right-wing ideology significantly predicts preferences for efficiency when entitlement concerns play a role. Self-reported left-wing ideology does not have predictive value in explaining preferences for equality; neither under luck, nor under effort. Therefore, I conclude that only self-reported right-wing ideology has predictive value. This finding suggests that, while right-wing ideology is still related to the traditional interpretation of left versus right, left-wing ideology does not represent this aspect of the conventional political divide anymore.
I study the dimensionality of political ideology in Chapter 4. Most common in political science and political psychology is to assume that ideology has two dimensions to represent economic and social preferences. Feldman (2013) argues that these dimensions are needed, since both find their origin in distinctly different values and personality characteristics. Furthermore, Feldman and Johnston (2014) demonstrate that ideology is better represented by at least two dimensions compared to just one. As
two dimensions, i.e. an economic and social one, and argue that they are theoretically different and only weakly correlated. This opens up the possibility for an individual to be more right-wing on one dimension, but left-leaning on the other. Based on similar findings, Treier and Hillygus (2009) argue that this leads individuals to be, so-called, ‘cross-pressured’, even though they have coherent political beliefs on each separate dimension. Achterberg and Houtman (2009) find additional evidence in favour of a two-dimensional representation of political beliefs.
However, whereas most assume a two-dimensional economic and social structure, Otjes (2017) shows in a recent study that, even when considering only economic preferences, a left-right structure is not an adequate fit. The take-away from these studies is, as Carmines and D’Amico (2015, p. 206) state: ‘If the basic measurement of ideology is flawed, it is likely that insights from research into both the ideological character of the public and the consequences of ideological thinking cannot be trusted.’ Yet, the use of one-dimensional measures of ideology in political economy has been persistent (e.g. Alesina, et al. (2018); Bjornskov, et al. (2013); Olivera (2015); Pitlik, et al. (2011)).
Relying on a one-dimensional scale means strict assumptions are needed. Firstly, you need to assume that left-wing and right-wing ideology are opposites of each other. This implies that individuals that are right-wing on economic issues, should also be right-wing on for example social issues. Secondly, you need to assume that beliefs are mutually exclusive and that individuals label their political beliefs along a left-right scale. Moreover, using such a measure suggests that the meaning of left and right is the same across individuals and relatively stable over time. The validity of these assumptions is challenged by earlier work in political science and political psychology (Conover & Feldman (1981); Feldman and Johnston (2014) and references therein; Jost, et al. (2009); Maynard (2013); Treier & Hillygus (2009); de Vries, et al. (2013)). Even though allowing for a second dimension is an improvement on one dimension, these studies force political beliefs to follow a two-dimensional structure by conducting only confirmatory analyses (e.g. Achterberg & Houtman (2009); Feldman & Johnston (2014); Treier & Hillygus (2009)).
In Chapter 4, I improve upon these studies by having no a priori assumptions about the number of dimensions of ideology. I study the dimensionality of voter ideology using an exploratory factor analytical approach and use accepted statistical methods to decide on the appropriate number of dimensions. Moreover, I do not restrict
dimensions to be uncorrelated with each other, and I validate the structure of political beliefs using a separate subset of the dataset. I identify and validate four relevant dimensions that capture preferences for economic equality, preferences for markets and efficiency, preferences for personal and cultural freedom, and nationalist, protectionist and populist preferences. These dimensions are correlated with each other, meaning they are not mutually exclusive. However, correlations are relatively low, indicating that each dimension captures a separate element of individual voter ideology.
Using the Dutch party space to further examine the dimensions, I find that there is much heterogeneity in preferences between voters of parties that remains hidden when relying on a left-right measure. Moreover, I find that voters interpret left and right on the basis of different ideological dimensions. A right-wing score for one party based on the ideology of their constituents is, thus, not necessarily directly comparable with a similar right-wing score for another party. I continue with an analysis of the determinants of multidimensional ideology and compare these to the determinants of the traditional left-right measure of ideology. I find that there is substantial heterogeneity in these determinants and using a one-dimensional left-right representation of voter ideology conceals most of it.
1.3 SOURCES OF HETEROGENEITY IN POLITICAL IDEOLOGY
After considering measurement issues, I continue with underlying sources of ideology as I study values and beliefs in Chapter 5. There is consensus among scholars that political attitudes originate in individuals’ values (e.g., Carmines & D’Amico (2015); Feldman (1988); Rokeach (1973); Schwartz, et al. (2010)). Rokeach (1973) argues, for example, that differences in political beliefs can be explained by heterogeneity in values of individuals. Inglehart (1971) goes a step further and shows that values do not only impact political ideology, but also affect long-term changes in partisanship. Relating the effect of values to ideology and political sophistication, Goren (2004) claims that all individuals have core beliefs and values, on which they depend when taking a position regarding political issues. Similar conclusions are drawn by Jacoby (2006). As such, these authors argue that the electorate relies on ideology to make sense of the political world. Moreover, Piurko, et al. (2011) find that people’s values are
Taking into account that differences in political ideology find their origin in values, I am interested in examining individual heterogeneity in such values. To do so, I study differences in values between students of different disciplines in Chapter 5. As such, I examine how study choice, both as a self-selection mechanism and over time, affects values that underlie differences in political ideology. Aside from values, I also examine beliefs. As the goal of higher education is to provide students with knowledge and information about how the world works, there is agreement on the notion that studying a certain discipline affects beliefs (see Hastie (2007) and Mayhew, et al. (2016) for reviews). However, there is not yet a consensus on whether this is also the case for values. I, therefore, study both in Chapter 5, as this allows me to compare the effect of field-of-study on values with that on beliefs.
The focus is on business students, as the curriculum of business schools is claimed to take a positive perspective, whereas in practice many of the subjects taught in business are value-laden. The most apparent example of this are courses in business ethics, that inherently contain value judgements. Moreover, business students often end up in leadership and managerial positions, in which they make decisions with potential major impact. After corporate scandals, in which business managers were accused of ethical misconduct (e.g. Enron), part of the blame has subsequently been put on these managers’ education (Goshal (2005); Haski-Leventhal, et al. (2017); Hummel, et al. (2018); Matten & Moon (2004)). Therefore, I focus on the effect that studying business has on students’ values and beliefs, and compare business students with those enrolled in other disciplines.
Regarding ideological differences between students of business and economics and students of other disciplines, Stigler (1959) was the first to conclude that studying economics makes for individuals with more politically conservative (i.e., right-wing) attitudes. This finding has been corroborated by some (Allgood, et al. (2012); Fischer, et al. (2017); Luker & Proctor (1981)). However, others argue that, while students of economics and business might seem to be more right-wing, the effect disappears when controlling for unobserved individual characteristics (Delis, et al. (2018)). These findings suggest that there are variables related to both ideology and studying business/economics, which could explain the initial relationship between the latter two; potential candidates being the values and beliefs of students. Fischer, et al. (2017)
hint, as well, that differences in values between students from different disciplines might be underlying their results, as they could be influencing both study choice and ideology.
Existing research examining differences between business students and students of other disciplines differentiates between selection effects and, what I call, socialisation effects. Selection effects refer to the initial differences between students that prompt them to self-select into a specific discipline. Socialisation effects refer to changes over the course of education, which are the result of (a combination of) learning, social interactions and/or students dropping-out. Sidanius, et al. (2003), for example, find significant selection-effects among students on the basis of attitudes with regard to social equality, but no effects of socialisation over time. Frey and Meier (2005) also only find evidence for selection, not socialisation, of students of business economics when studying selfish behaviour. Cipriani, et al. (2009) and Haucap and Just (2010) find both selection and socialisation effects when studying situations in which there is a trade-off between efficiency and ethical behaviour, and differential price and allocation mechanisms in situations of scarcity.
I add to this literature by examining both beliefs and values, while at the same time distinguishing between economics and business students. As exposure to economics courses is often blamed for the harmful effects of business education (e.g. Ghoshal (2005); Racko (2017)), differentiating between the two groups of students allows me to evaluate the role of economics. Moreover, I add to the literature on field-of-study differences by investigating a set of beliefs and values. This chapter also adds to the literature on the origins of ideology by examining education (in a certain field) as one of the determinants explaining differences in the values underlying political attitudes. By comparing values and beliefs of business students with those of students from four other disciplines (i.e. economics, law, psychology and social sciences) at the start and the end of their first year, I am able to test for selection and socialisation effects. I find that business students at the start of their academic education differ significantly from students of other disciplines in terms of their values and beliefs. Moreover, I find that some of these differences are persistent over time. Looking at changes within, instead
field on the basis of values and beliefs. Furthermore, compared to those at the start of their first year, business students at the end of the year show significant changes in beliefs and values. This indicates that there are socialisation effects from studying business.
In Chapter 6, I bring the three parts of this thesis together by discussing how the findings in this thesis impact the study of political ideology. I also use this concluding chapter to relate the findings in this thesis to each other and I explore exciting avenues for future research.
Political ideology and the
intragenerational prospect
of upward mobility
This chapter is based on Laméris, et al. (2018a) and is currently under review at an international journal
2.1 INTRODUCTION
Governments spend major shares of GDP on redistribution and social transfers.6 This
explains the long history of studies into the determinants of redistribution, and the influence of political ideology and inequality aversion on it. Seminal contributions by Meltzer and Richard (1981) and Dixit and Londregan (1998) brought forward an entire literature on political attitudes and redistribution or redistributive preferences (e.g., Alesina & Angeletos (2005); Alesina & Giuliano (2011); Case (2001); Feld (2000); Olivera (2015); Page & Goldstein (2016); Roemer (1998), (1999)).
Another factor influencing redistributive preferences is the so-called prospect of upward mobility (POUM) hypothesis that has been pioneered by Benabou and Ok (2001). According to this POUM hypothesis, individuals expecting future upward income movements might rationally demand lower levels of redistribution. Even though these individuals would benefit from it based on their current income. The POUM hypothesis has generated a number of studies searching for evidence (e.g. Alesina & La Ferrara (2004); Checchi & Filipin (2004); Cojocaru (2014); Corneo & Gruner (2002); Rainer & Siedler (2008); Ravaillon & Lokshin (2000)).7 The
consensus among these studies is that an increase in income mobility (whether actual or perceived) leads to less support for redistribution.
Recent studies focusing on POUM-effects aim to take the role of political beliefs and attitudes into account when studying income mobility. Buscha (2012) finds that individuals who expect their financial situation to improve over time are more right-wing, whereas those that expect a deterioration are more left-wing. Furthermore, he finds that individuals are more likely to support a conservative party if they expect upward income mobility and if they have right-wing political preferences. Whereas these findings suggest an indirect link between expectations of upward mobility and redistributive preferences through political beliefs, Buscha (2012) does not examine such preferences. Alesina, et al. (2018) do consider preferences for redistribution and political beliefs by studying how perceptions of mobility affect support for redistributive policies distinguishing between left-wing and right-wing individuals. 6. For example, public social expenditures totals 22% of Dutch GDP (OECD average: 21%, 2016) and over 50% of total expenditure of the Dutch government is dedicated to social expenditures (OECD average: 45%, 2013). Source: OECD.Stat.
In an intergenerational context, these authors find a strong link between support for redistributive policies and perceptions of income mobility. They also find that this link is conditional on political ideology.
In this chapter, we study the role of ideology in the relation between mobility expectations and preferences for redistribution from an intragenerational perspective. Unlike Alesina, et al. (2018), we take into consideration the influence of life-cycle earnings by focusing on those individuals for which the POUM-effect is most relevant. As such, we aim to give a more precise account of the conditional effect of expected upward income mobility on the preferred level of redistribution. Apart from the difference between intergenerational mobility and intragenerational mobility, our research makes another important contribution. Whereas Alesina, et al. (2018) study perceptions of individuals about mobility opportunities on a country-level, we consider expected income mobility on an individual-level. In other words, we look at how an individual’s expectation of own income mobility relates to his/her preference for redistribution, and how this relation is affected by political ideology.
To study the relation between political ideology and the prospect of intragenerational upward mobility, we use survey data obtained from the CentERdata panel that consists of a representative sample of Dutch households. Previewing our results, we find a statistically significant POUM-effect on redistributive preferences: individuals who expect upward income movements have a lower preference for redistribution compared to those not expecting upward mobility. However, we find that this POUM-effect runs through political beliefs. Expected upward income mobility only affects preferences when respondents have right-wing political beliefs. For those with centre or left-wing political beliefs, expected upward income mobility has no effect on preferences. Regardless of what these individuals expect to earn in the future, they prefer a society with redistribution over one without.
This chapter continues as follows. In the next section, we describe our data. In section 2.3 we present our main results, as well as sensitivity checks using different measures to capture redistributive preferences. In section 2.4, we discuss our findings and conclude.
2.2 DATA AND MODEL
Our dataset consists of 2453 observations and was gathered by CentERdata.8 This
institute has access to over 2000 households, which together form a representative sample of the Dutch population. In March 2016 an invitation to participate in our survey was sent to all panel-members, of which 79.8 percent responded. The survey included questions on political preferences, current income position, future income expectations and beliefs regarding the desired level of redistribution. Additionally, we asked respondents a broad set of questions concerning their socio-economic background.9
To examine whether there is a POUM-effect that is conditional on political ideology, we focus on respondents aged between 25 and 54. We focus on this age group for four reasons. Firstly, we concentrate on intragenerational mobility, which means we should consider the influence of life-cycle earnings profiles. As argued by Benabou & Ok (2001), the heterogeneity of a person’s earnings over the course of his or her life could be an influential factor in how mobility expectations affect preferences for redistribution. We take into account this heterogeneity by focussing our identification on individuals that are of working age and have a prospect of climbing the income ladder in the remainder of their careers. In other words, by considering the concavity of life-cycle earnings (see e.g. Blundell, et al. (2015); Mincer (1974); Polachek (2008)) our identification rests on those individuals for which upward income mobility over time is possible. Secondly, earlier studies find that POUM-effects are generally found among individuals that are younger, more educated and less likely to be employed (Cojocaru (2014)). As such, our focus is on those individuals for which the theory is most relevant. Thirdly, individuals at later stages in their life are more likely to be in or go into retirement, and thus, more likely to consider intergenerational factors. Given the substantial literature on the relation between pension schemes, social security programs, retirement decisions and labour force participation (see Gruber & Wise (1999), (2004)), we exclude those respondents for which pension considerations are relevant. Fourthly and related to the latter argument, the survey questions we
8. CentERdata is a Dutch institute for data collection and research. This institute sets out surveys on request of academic, public and private institutions.
9. The survey (in Dutch (original) and in English (translation by authors)) can be found in the appendix to this thesis. The corresponding dataset is also used in Chapter 4 of this thesis.
use to measure upward mobility expectations ask about expected income 10 years from today. We, thus, also exclude respondents aged between 55 and 64, who are considering pension income when asked about their future income.10
Table 2.1 shows summary statistics of respondents’ characteristics; the second column for the full sample and the third column for the respondents aged between 25 and 54. As would be expected, net household income and the level of employment is higher for the age group we consider for identification.
Table 2.1 Summary statistics of respondent’s characteristics - full sample and sample restricted to ages
25-54
Variable
Full Sample Ages 25-54
Mean S.D. N Mean S.D. N
Age 54 17 2,453 40 8 1093
Household income (monthly; net) 2820 1391 2,449 3180 1427 724
Women (in percentages) 49 - 2,453 56 - 1,093
Employed (in percentages) 51 - 2,453 82 - 1,093
Married (in percentages) 77 - 2,453 80 - 1,093
Religious (Christian; in percentages) 17 - 2,453 15 - 1,087
Note: Average Dutch net household income in 2014, the most recently available year, was 35,000 euro. This results in 2917 euro on a monthly basis. Source: Central Bureau of Statistics Netherlands. Religiosity is based on whether a respondent votes for a Christian political party.
In line with the literature, we measure respondents’ redistributive preferences using statements that ask about beliefs regarding redistribution. Most studies use one statement to capture these preferences (e.g. Alesina & La Ferrara (2004); Corneo & Gruner (2002)). We use three statements: (1) ‘The government should tax the rich and subsidise the poor’, (2) ‘Everyone should be rewarded by effort and performance, even when this leads to income differences’ and (3) ‘Income differences between the rich and the poor should be reduced as much as possible’. The first statement mentions a means
for the government to achieve redistribution. The second statement touches upon beliefs about reasons that might justify income differences. The third statement deals with feelings towards income differences more generally and more explicitly: should there be any differences in income at all? All three statements, thus, capture different aspects of redistributive preferences. We asked the respondents to what extent they agree with these statements on a 5-point Likert scale ranging from completely disagree to completely agree. A high score on the first and the third statement and a low score on the second statement indicates a strong preference for redistribution.
Figure 2.1 shows the distribution of responses on the redistribution statements for respondents aged 25-54. The majority chooses the neutral option when it comes to taxing the rich and subsidising the poor, and about the same amount of respondents agree (35 percent) with the statement as disagree (33 percent) with it. Considering the second statement, more than half of the respondents believe that some income differences are allowed, as long as rewards are based on effort and performance. Still, the majority of respondents believe income differences should be reduced as much as possible (statement 3, 42 percent). 28 percent disagrees with this statement. The correlation between the redistribution statements ranges from -0.26 (statements 1 and 2) and -0.38 (statements 2 and 3) to +0.60 (statements 1 and 3). To capture preferences for redistribution in one variable, we conduct a factor analysis using the three statements. Results show that the statements are well-represented by one factor, which we interpret as measuring redistributive preferences. Factor loadings can be found in the appendix to this chapter, A2. We predict factor scores for each respondent in the sample and label the corresponding factor ‘preferences for redistribution’. We measure respondents’ subjective views towards the prospect of upward income mobility with three survey questions. We use these to create two measures of upward income mobility.11 The first, which we refer to as the ‘absolute’ question is posed as
follows: ‘Would you say your income position in about ten years will be worse, the same or better than now?’ The resulting dummy variable is equal to 1 if the respondent answered that he/she expects his/her income position to be better in the future and 0 otherwise. The second question captures what we refer to as ‘relative’ expectations regarding future income: ‘How high do you expect your income to be in comparison to
11. We choose to focus on subjective measures based on empirical results, see Alesina & La Ferrara (2004), Ravaillon & Lokshin (2000) and Rainer & Siedler (2008).
2 3 4 5 S tat ement 1 0 10 20 30 40 50 Percent 1 2 3 4 5 S tat ement 2 0 10 20 30 40 50 Percent 1 2 3 4 5 S tat ement 3
edistribution statements (in per
centages)
w the distribution of opinions on the r
edistribution statements for r
espondents aged betw
een 25 and 54. The left panel sho
ws
The go
ver
nment should tax the rich and subsidise the poor
). The centr e panel sho ws r edistribution statement 2 ( Ev er yone should be t and per for mance, ev
en when this leads to income differ
ences
). The right panel sho
ws r
edistribution statement 3 (
Income differ
ences betw
een the
educed as much as possible
). The scale ranges fr
om 1 (completely disagr
ee) to 5 (completely agr
ee). F
or statements 1 and 3, a high
efer
ence for r
edistribution; for statement 2 this is indicated b
y a lo
w scor
others in about ten years?’12 Here, respondents answer on a 5-point Likert scale ranging
from considerably below average to considerably above average. To create our relative measure of upward mobility, we combine this with respondents’ answers to the following question: ‘Compared to others, how high do you think your current income is?’ Again, respondents answer on a 5-point Likert scale ranging from considerably below to considerably above average. Combining these questions, our ‘relative’ measure of mobility is a dummy equal to 1 when respondents judge their income in ten years to be higher than their current income (compared to others), and 0 otherwise. For example, a respondent that views his/her current income as below average, but expect his future income to be either average, above average, or considerably above average is considered to expect upward income mobility.
To investigate the prevalence of expected upward income movements, we relate respondents’ views regarding their future income to their views of their current income relative to others. Table 2.2 shows a cross-tabulation of current income and future expectations (relative to others). As with our relative measure of mobility, we define expected upward movements as believing income to be higher in the future than today (compared to others). These cells are marked light-grey. Expected downward mobility is defined as expecting future income to be lower than today’s income (relative to others). These cells are marked dark-grey. In our sample, the majority (73 percent) expects no income movements in the upcoming 10 years. 9 percent (101 respondents) expect downward mobility whilst 18 percent (191 respondents) expect upward mobility. Furthermore, out of those expecting upward mobility, 48 percent (92 respondents) expects their income to be above average in the future.
We measure respondents’ left-right political ideology on a linear scale that ranges from 1 (left-wing) to 10 (right-wing) using the question: ‘In politics people usually speak of the left and the right. Where would you place your own political ideas?’ The mean of this self-reported score is 5.3 (std. dev. 1.9).13 Figure 2.2 shows the corresponding
distribution of left-right ideology. In all subsequent analyses, we distinguish between respondents with left-wing ideology, centre ideology and right-wing ideology. Subjects with self-reports smaller than or equal to 4 are considered ‘left’. Those with self-reports
12. With relative we mean expected income in comparison to something else, here: other people’s income. We do not mean relative in the sense of connectedness, i.e. affiliated or associated.
centages (of the total amount) of self-indicated curr
ent income and expected futur
e income of r espondents r elativ e to others e E xpec ted F utur e In com e C onsider ably belo w a ver age Belo w a ver age A ver age Abo ve a ver age C onsider ably abo ve a ver age Total w av er age 22 / 2 .0% 11 / 1. 0% 10 / 0 .9% 3 / 0 .2 % 2 / 0 .2 % 4 8 / 4 .4 % 5 / 0 .5 % 82 / 7 .5 % 78 / 7 .1% 10 / 0 .9% 0 / 0 .0% 17 5 / 15.9% 1 / 0 .1% 38 / 3.5 % 4 23 / 38 .5 % 57 / 5. 2% 2 / 0 .2 % 521 / 4 7. 4% 1 / 0 .1% 5 / 0 .5 % 40 / 3. 6% 246 / 22 .4 % 18 / 1. 6% 310 / 28 .2 % ve av er age 0 / 0 .0% 0 / 0 .0% 0 / 0 .0% 11 / 1. 0% 23 / 2 .1% 34 / 3. 1% 29 / 2 .7 % 136 / 12 .4 % 551 / 50 .2 % 32 7 / 29 .8 % 4 5 / 4 .1% 1088 / 100% centages) of curr
ent income and expected futur
e income of r
espondents aged betw
een 25-54 is sho wn. ’ vie ws on their curr ent income r elativ e to others is sho wn. This is cr oss-tabulated with r espondents
’ expectation of their futur
larger than or equal to 7 are consider ‘right’. Respondents with a self-reported score of 5 or 6 are in the centre of the political spectrum. For each of the 3 categories we construct dummies.14 0 5 10 15 20 P ercent 1 2 3 4 5 6 7 8 9 10
Left-right ideology: 1 (left) - 10 (right)
Figure 2.2 Distribution of left-right ideology (in percentages)
Note: This graph shows the distribution of left-right ideology for respondents aged between 25 and 54. The left-right scale ranges from 1 (left) to 10 (right). We asked respondents: ‘In politics people usually speak of the left and the right. Where would you place your own political ideas?’
To see if and how ideology affects the relation between expected upward income mobility and redistributive preferences, we relate political beliefs to mobility expectations using our absolute measure of expected upward mobility. In table 2.3 we show the prevalence of respondents expecting upward mobility split according to self-reported left-wing, centre and right-wing ideology. There are 1091 respondents in our sample, for which we have information on both their (absolute) expected mobility and their political beliefs. 34 percent has left-wing ideology, 36 percent considers themselves to be in the centre of the political spectrum and 30 percent has right-wing ideology. Table 2.3 tells us that 33 percent of all respondents in the sample expect upward 14. If we consider self-reports from 1-3 to be left-wing; 4-7 to be centre; and 8-10 to be right-wing ideology, and redo the analyses, it does not affect our main results and conclusions.
income movements versus 67 expecting no or downward movements (based on the absolute measure). What happens when we consider differences in political beliefs? The table shows that 31 percent of left-wingers and 28 percent of the respondents with centre beliefs expect upward income movements. Right-wingers expect the most upward income mobility: 40 percent versus 60 percent that expect no or downward mobility. Based on a Chi-squared test of association on the cross tabulation, we reject the null hypothesis that mobility and political ideology are independent (test-statistic = 13.11, p-value = 0.001).
Table 2.3 Counts and percentages of respondents expecting upward mobility by political ideology
Left-wing Centre Right-wing Total
No expected upward mobility 255 / 69% 285 / 72% 194 / 60% 734 / 67%
Expected upward mobility 116 / 31% 110 / 28% 131 / 40% 357 / 33%
Total 371 / 100 % 395 / 100% 325 / 100% 1091 / 100%
Note: The absolute measure is used to measure expected upward mobility. Political ideology is split out according to left-centre-right ideology. A self-report between 1-4 is considered left-wing and a self-report between 7-10 right-wing. Self-reports of 5 and 6 indicate centre ideology.
These descriptive findings suggest that there is a relation between upward income mobility expectations and political beliefs. As such, our expectations regarding a POUM-effect that is conditional on ideology are reinforced. We estimate the following model that is designed to capture this:
𝑃𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒𝑠 𝑓𝑜𝑟 𝑟𝑒𝑑𝑖𝑠 𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛𝑖= 𝛽0 + 𝛽1 𝑈𝑝𝑤𝑎𝑟𝑑 𝑚𝑜𝑏𝑖𝑙𝑖𝑡𝑦𝑖 + 𝛽2 𝐶𝑒𝑛𝑡𝑟𝑒 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦𝑖 + 𝛽3 𝑈𝑝𝑤𝑎𝑟𝑑 𝑚𝑜𝑏𝑖𝑙𝑖𝑡𝑦𝑖 ∗ 𝐶𝑒𝑛𝑡𝑟𝑒 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦𝑖 + 𝛽4 𝑅𝑖𝑔ℎ𝑡
𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦𝑖 + 𝛽5 𝑈𝑝𝑤𝑎𝑟𝑑 𝑚𝑜𝑏𝑖𝑙𝑖𝑡𝑦𝑖 ∗𝑅𝑖𝑔ℎ𝑡 𝑖𝑑𝑒𝑜𝑙𝑜𝑔𝑦𝑖 + 𝛾 𝑍𝑖 + 𝜀𝑖
(2.1)
where 𝑍𝑖 is a vector containing our control variables and 𝜀𝑖 the error term. As main dependent variable we use the factor ‘preferences for redistribution’. Factor scores are standardised and continuous, which allows us to estimate the model with
The focus is, however, on the included interactions between mobility and political beliefs.15 These interactions allow us to test for any conditional effects, and as such,
we can answer our main research question: is the effect of intragenerational prospect of upward mobility on redistributive preferences conditional on political ideology? We follow existing literature and control for a range of individual characteristics, including subjective (i.e. how easy it is to make ends meet) and objective (i.e. net household income) measures of current income position, education levels, gender, age, marital status, employment status, the number of children living at home and religiosity. Additionally, we control for the degree of risk-aversion.16 We present
estimation output as well as marginal effects of income mobility on redistribution for the three (i.e. left, centre and right) ideological groups.17
We expect pro-redistributive beliefs among left-wing individuals and vice versa for right-wing individuals. Furthermore, in line with existing research we expect expectations of upward mobility to negatively affect redistributive preferences. However, this effect is believed to (partly) run through political ideology.
15. Note that respondents that indicated to be left-wing are the reference category in our model and estimations.
16. See e.g. Alesina et al. (2018), Alesina & Giuliano (2011), Fong (2001), Guillaud (2013), and Olivera (2015). Additionally, race is one of the standard controls in research on redistributive preferences. Unfortunately, our dataset does not contain information on the race or origin of our respondents. With regards to risk aversion, as Benabou & Ok (2001) argue, only individuals that are not too risk-averse can be affected by a POUM-effect, as it is risk-averse individuals that also view redistribution as insurance against downward income movements (for empirics see Cojocaru (2014)).
17. The education variable is denoted in the amount of years needed (on average) to obtain a specific educational degree, i.e. the higher this variable, the higher level of obtained education. In the Dutch education system, this leads to the following scoring: 6 years (elementary school) / 8 years (low-level secondary education) / 10.5 years (vocational education) / 11.5 years (high-level secondary education) / 14 (low-level (applied) university education) / 16.5 (high-level university education). The religion dummy is a proxy based on whether a respondent votes for a Christian political party. The subjective measure of household income asks respondents how easy it is for them to make ends meet. The corresponding scale ranges from 1 (very difficult) to 5 (very easy). The monthly household income categories are 1) lower than 1150 euro, 2) between 1151-1800 euro, 3) between 1801-2600 and 4) more than 2600 euro.
2.3 RESULTS
Table 2.4 shows the estimation results using the absolute measure of upward mobility as independent variable in column (1) and the relative measure in column (2). As to our main research question, we first focus on the signs and significance of the estimated coefficient of the interaction terms. Considering centre ideology, the interaction effect with upward mobility captured with the absolute measure is insignificant. The interaction between centre ideology and upward mobility captured with the relative measure is marginally significant (at the 10% level). This suggests that for this ideological group expecting upward income movements has a negative effect on preferences for redistribution compared to when no or downward income mobility is expected. However, this result is dependent on the measure of mobility that is used. For right-wing respondents we find negative and significant (at the 1% and 5% level) coefficients of the interaction terms for both mobility measures. Thus, for right-wingers we find a conditional effect of mobility expectations on preferences for redistribution (relative to the reference category consisting of left-wing individuals). We find that, while right-wing respondents have a lower preference for redistribution to begin with, those also expecting upward income movements prefer even less redistribution. Furthermore, we find that both ideology dummies are negative and significant at the 1% level. When no upward mobility is expected, both centre and right-wing respondents have a lower preference for redistribution compared to left-wing respondents. This effect of ideology is an established outcome (e.g. Alesina, et al. (2018); Alesina & Giuliano (2011); Olivera (2015)). Table 2.4 also shows that both dummies measuring expected upward mobility are insignificant. We, thus, find no effect of upward mobility on redistributive preferences for our left-wing respondents. Additionally, we can infer from table 2.4 that an increase in the (subjective) current income position of respondents leads to less support for redistribution. The easier it is for people to make ends meet, the less redistribution is preferred. Being more risk-loving also reduces the support for redistribution. Moreover, employed individuals and individuals with higher education prefer less redistribution as well. These findings confirm earlier research on redistributive preferences (e.g. Alesina & Giuliano (2011);
Table 2.4 OLS estimation results using ‘preferences for redistribution’ as dependent variable Dependent variable: Preferences for redistribution
(reference: left-wing)
(1) Absolute Measure
(2) Relative Measure
Dummy expectation of upward income mobility -0.091 (0.130) 0.151 (0.142)
Centre ideology -0.654*** (0.102) -0.657*** (0.095)
Dummy expectation of upward income mobility x centre
-0.166 (0.175) -0.347* (0.206)
Right-wing ideology -0.894*** (0.122) -1.013*** (0.109)
Dummy expectation of upward income mobility x right-wing
-0.488*** (0.189) -0.515** (0.227)
Risk averse - risk loving -0.042* (0.021) -0.043** (0.022)
Education -0.073*** (0.015) -0.079*** (0.015)
Dummy female 0.008 (0.072) 0.044 (0.072)
Age -0.002 (0.005) 0.002 (0.005)
Dummy married 0.004 (0.109) 0.006 (0.111)
Dummy employed -0.263*** (0.098) -0.236** (0.099)
It is difficult to make ends meet (ref: very difficult) -0.219 (0.296) -0.148 (0.290)
We exactly make ends meet (ref: very difficult) -0.371 (0.290) -0.291 (0.285)
It is easy to make ends meet (ref: very difficult) -0.676** (0.295) -0.598** (0.290)
It is very easy to make ends meet (ref: very difficult) -1.040*** (0.316) -0.957*** (0.310)
Household income category 2 (ref: category 1) 0.239 (0.173) 0.222 (0.174)
Household income category 3 (ref: category 1) 0.105 (0.168) 0.092 (0.169)
Household income category 4 (ref: category 1) -0.204 (0.171) -0.233 (0.172)
Dummy religious -0.044 (0.110) -0.019 (0.110) Children -0.008 (0.035) -0.004 (0.034) Constant 2.470*** (0.460) 2.224*** (0.456) Observations 1068 1065 Adj. R-squared 0.267 0.255 F-statistic 20.330 20.075 (p-value) 0.000 0.000
Note: OLS regression results are displayed with robust standard errors clustered at the household level (in parentheses). Significance is indicated as follows: *** p<0.01, ** p<0.05, * p<0.1. The dependent variable is the factor ‘preferences for redistribution’. The sample consists of respondents aged between 25-54. In column 1, the dummy indicating expected upward mobility is created using the absolute measure; in column 2, using the relative measure. Political ideology is captured with left-centre-right dummies. The reference group consists of respondents with left-wing ideology. Please refer to the footnotes for descriptions of the education, religion and household income variables. A high score on ‘risk averse - risk loving’ indicates risk-loving; ‘children’ is the number of children living at home.
Table 2.5 Marginal effects of expecting upward income mobility on preferences for redistribution for left-wing, centre and right-wing ideology
Dependent variable: Preferences for redistribution
(1) Absolute Measure (2) Relative Measure Left-wing ideology -0.091 (0.130) 0.151 (0.142) Centre ideology -0.257* (0.132) -0.196 (0.161) Right-wing ideology -0.579*** (0.145) -0.364* (0.186) Observations 1068 1065
Note: This table shows marginal effects of expected upward mobility for left-wing, centre and right-wing ideology on preferences for redistribution. Standard errors are in parentheses. Significance is indicated as follows: *** p<0.01, ** p<0.05, * p<0.1. The marginal effects are calculated from coefficients estimated using an OLS specification. The dependent variable is the factor ‘preferences for redistribution’. In column 1 mobility is measured with the absolute measure. In column 2 the relative measure is used. In order to gain more insight into the conditionality of the POUM-effect, we calculate average marginal effects of upward mobility expectations for left-wing, centre and right-wing respondents. These marginal effects, which can be found in table 2.5, confirm our earlier findings. There is no significant effect of upward mobility on redistributive preferences for left-wing respondents. For these individuals, expecting to earn more in the future does not affect their preferred level of redistribution today. However, for both measures of mobility expectations, we find a negative and significant marginal effect of mobility expectations for wing respondents. Identifying with right-wing ideology and expecting upward income movement leads to lower support for redistribution. Considering respondents with centre ideology, we find a negative and significant (at the 10% level) effect of upward mobility when relying on the absolute measure. However, there is no statistical difference between the marginal effects of the left-wing and centre respondents (as shown by the insignificance of the interaction term in table 2.4). The marginal effect of upward mobility expectation on preferences for right-wingers, however, is statistically different from that of left-wingers. For the relative measure, we find no significant marginal effects for individuals in the centre of the political spectrum.18
