Intergenerational Transmission of Party Affiliation Within Political Families

We investigate the intergenerational transmission of political-party affiliation within families with at least two politicians. We use Swedish registry data that covers all nominated politicians for the years 1982 to 2014, as well as their family ties. First, we demonstrate there is a strong link between individuals and their parents concerning party affiliation. We also find that this intergenerational transmission persists over generations and across siblings. Our second aim is to investigate the mechanisms behind this result, which we do by first discussing two hypotheses: the one concerns a socialization pathway, the other a materialistic one. We then bring these hypotheses to the data, and we find that the socialization pathway matters more for intergenerational transmission.


Introduction
Individuals are affected by the context in which they grow up, where the family is expected to play a central role. A person's political participation and political engagement are likely to increase if he or she has a parent that has been nominated to political office. The impact is likely to be twofold: an individual whose father or mother is a politician for a given party is more likely to be politically active in general; in addition, he or she is more likely to sympathize with the specific party in question.
The focus of this paper is on families where both parents and children have been politicians. There are numerous examples from many different countries where the children of politicians themselves become politicians later on-often for the same party as their parents. In Sweden, for example, approximately 28% of those who are politicians and born after 1970 have a parent that was also a politician. There are many possible reasons for an intergenerational connection, with the children of politicians being socialized into specific political beliefs, behaviors, and opinions. They may also have similar political demands because they have a similar material standard, or because they possess the political skills and contacts necessary for success in the political arena.
The accumulation of power within families can have negative consequences for political equality. From the standpoint of equal opportunity, it is in itself problematic if the children of politicians have an advantage over children from non-political families. Furthermore, to the extent that children share the beliefs and opinions of their parents, centering power within certain families may reproduce the views and ideals currently held by the political elite, instead of giving room for candidates with new ideas and different perspectives.
In this paper, we study how common it is that individuals run for the same political party as their parents. Now then, while parents are important characters, they are not the solitary source of influence. A dynamic process takes place during childhood, in which one is also likely affected by the opinions, beliefs, and demands of other people in close proximity, such as siblings and grandparents. When we investigate transmission, therefore, we take the extended family into account.
Earlier empirical literature has provided us with important insights, but it is based above all on older survey data. Our main goal in this paper is instead to investigate the intergenerational transmission of political-party affiliation within the family by using Swedish registry data from more recent years. Swedish data is particularly suitable for addressing these questions, because population registers in Sweden provide information on family connections, which in turn may be linked to other registers with information on political candidacies and other background matters. Above all, registers have the advantage of covering the whole population and of being more accurate than surveys.
In this paper, we define a person as affiliated with a political party if he or she runs for that party in a general election. To study intergenerational transmission, we basically calculate the share of all candidates who are affiliated with the same party for which their parent, sibling, or grandparent has previously been nominated. The next questions we pose are: How can we explain this partisan transmission? What mechanisms are in play? By connecting our empirical analysis to the underlying theoretical mechanisms, our paper contributes to the literature on political socialization (which hitherto has been highly empirical, with a few exceptions). We highlight two main theoretical mechanisms, which we have chosen to denote as the socialization pathway and the materialistic pathway. The notion that an individual is socialized into a certain political affiliation reflects the premise that discussions at home are the main transmission mechanism. The materialistic view, on the other hand, is more connected to a rational-choice and political-economy approach to politics, where individuals are seen as utility-maximizers whose demands depend on their material standard, which in turn may be intergenerationally transmitted. We discuss these theoretical predictions one at a time, after which we bring the predictions to the data.

Earlier Literature
The literature analyzing how political affiliations are transmitted is vast and primarily empirical (see Jennings 2007, for a review). The literature review in this section will therefore focus on these findings; we shall return to the underlying theoretical arguments in the theoretical-framework section. The main focus of this earlier empirical literature has been, at least indirectly, on the socialization aspect. 1 Jennings and Niemi (1968) was one of the first papers investigating intergenerational transmission within families. Using American survey data from a sample consisting of high-school students and their parents, the authors find mixed results with regards to the transmission of political values. They do detect, however, a strong parent-child correspondence with regards to party identification. Jennings et al. (2009), expanding on this analysis by including additional cohorts, find that the transmission of opinions on a given topic is dependent on other contexts at the time of the survey. They find as well that the transmission is greater in families with a strong political commitment. Tedin (1974) points out that the intergenerational transmission of beliefs is dependent on how salient the issue is to the parents, as well on how correctly the child perceives the parents' attitudes. Beck and Jennings (1975) apply an analysis with three generations. They use a survey data set consisting of high-school students complemented with interviews on the stated political affiliation of parents and grandparents. One interesting conclusion they draw is that fathers' political affiliation seems to be the dominant source of political socialization in the parent-grandparent generation; but that, in the child-parent generation, both parents have an almost equal impact on the political affiliation of their children (when the parents disagree). Beck and Jennings (1991) focus instead on how the parental transmission of political beliefs has changed over time. They find that intergenerational transmission grew weaker at the time of the general anti-authority movements of the 1960s and 1970s; and that children originating from highly Republican or Democratic families were mostly affected by the general trend to challenge authority. Westholm (1991) studies Sweden just as we do, but with survey data from the first half of the 1980s which includes information on political opinions among parents and their children. His study shows, among many other things, that there is an intergenerational connection between parents' political views and their children's. Westholm (1999) focuses on the underlying transmission mechanisms, where the author argues that a child first need to perceive the opinions of the parents and then be persuaded by these opinions.
Later studies have emphasized the importance of the institutional setting for intergenerational transmission. Percheron and Jennings (1981) point out that France is dominated by a left-right dimension, and argue that the fluidity of the French party system means that the role of partisan identification is less pronounced than in the US. The authors perform their analysis with cross-country survey data which supports their view that intergenerational transmission finds expression either as party affiliation or as self-placement along the left-right dimension. Westholm and Niemi (1992), who revisits these results by adding Sweden and Finland to the data set, argues that the thesis of an opposite relationship between left-right transmission and party identification does not apply to these Nordic countries. Jennings (1984) also engage in a cross-country analysis and study the transmission of a left-right dimension, but they complement it with a materialistic-postmaterialistic dimension and an authority-antiauthority dimension. He finds that the transmission is primarily present in the left-right dimension. 2 These earlier papers use survey data in their measures of political attitudes and party identification, meaning that their authors focus on stated party preferences. The focus in our paper is instead on revealed party preferences, since we have access to registry data on nominated politicians. There are pros and cons connected to both approaches. Our approach has the advantage of using a more objective measure of political affiliation. Being nominated for a party is a clearer signal of political conviction than stating in a survey that you prefer a specific party. We also have the advantage of using population-wide data for all politicians in Sweden for a given time period. We may not focus, however, on a more fine-grained analysis with regards to political preferences, but instead settle on using political-party affiliation as a catch-all measure. The main disadvantage of using surveys is that people do not necessarily state their true preferences, and that parents and their children may influence each other when answering the same survey. There is also self-selection into generational surveys, where the propensity to answer may be correlated with how close the parents and children are in terms of beliefs (Connell 1972)-meaning that the researchers using the surveys are studying a selected sub-sample. We too have a selection problem, however, given that the registry data only allows us to observe the party affiliation of those who have taken the step of running for office. As a result, our study is restricted to families with at least two politicians. This selection problem means that we cannot interpret the intergenerational correlation as a causally identified ideological transmission, because it is possible that children who are ideologically close to their parents are more likely to run for office. We return to these issues in the theoretical-framework section. 3 Since we focus on families with multiple politicians, our paper is also linked to the literature on political dynasties, which analyzes how important family connections are for political careers. Dal Bó et al. (2009) demonstrate that the children of U.S. congressmen tend to run for and to be elected to the same congressional seat as their parents. Querubín (2016) finds evidence of a long-term dynastic effect, showing in a RD-framework that the probability of being elected to office in the Philippines increases if a relative (narrowly) wins an election. The existence of a dynastic element in politics has been further demonstrated in Japan (Asako et al. 2015), India (Chandra 2016), the U.S. (Feinstein 2010), and Ireland (Smith and Martin 2017). Fiva and Smith (2018) show there is an incumbency advantage in Norway, but they find no evidence that this in turn yields a higher probability that a family member will actually win an election. Nor does Van Coppenolle (2017) find any evidence that a dynastic effect results from having a relative serve in the UK House of Commons for a relatively long period of time. While our paper is somewhat related to the dynasty literature-given that we also study the impact of the family and of the extended family-our purpose is not to fasten on how political offices are inherited, but instead to focus on partisan affiliation in a broader sense.

Theoretical Framework
We focus on two potential theoretical channels through which the transmission of political-party affiliation may be mediated: (1) a socialization pathway, and (2) a materialistic pathway. These two conceptions spring from two different theoretical traditions. The idea of the socialization pathway has been presented verbally in previous literature, whereas that of the materialistic pathway originates from a very standard rational-choice (political-economy) framework. We present both pathways in words here in the main text, but the reader will find a formalized version of the materialistic pathway in the Online Appendix. We address selection issues below when we discuss the materialistic pathway, because these may be highlighted in a pedagogical way within rational-choice reasoning. However, such questions of empirical selection are also present when we test the socialization pathway.

The Socialization Pathway
Beginning with the socialization pathway, we may note that it formed the indirect starting point in many of the empirical papers discussed in the literature section above. The core idea is that discussions at home with parents and relatives function to socialize children. If a parent is a convinced liberal, then the child is likely to become more liberal than otherwise, simply because children spend so much time together with their parents and because parents act as role models. This point connects up with social learning theory, where the social context (the family) is understood as furnishing the frames for the learning process. According to social-learning theory, children observe the behavior of their parents (as well as that of other "models"), and they attempt to imitate it. If this results in a rewarding response from the parent, then the child will continue along the same path (and vice versa if his or her parents respond with negative reinforcement Bandura 1977). If political parents are more likely to reward certain forms of political behavior than others ("behavior" being defined here in the broadest way), and especially if such reinforcement is more likely to be positive for behavior that is in line with the parents' partisan alignment, then this may lead the children of politicians to participate politically to a greater extent than others, as well as creating an intergenerational correlation with respect to opinions and political affiliation. Social-learning theory connects up with Skinner's (1957) work on language development, according to which reinforcement and punishment are key elements.
As we indicated in our literature review, most papers that have studied intergenerational transmission of political affiliations are empirical, and the exact theoretical foundation in these papers is rarely explicit. The theory is mostly concerned with when the transmission takes place (Sears and Brown 2013). Does the socialization process start already at birth, or is an individual particularly susceptible during a certain time in life?
One central idea among scholars who stress the socialization pathway has been that political views crystalize as a person ages, meaning that a person's views and perceptions are more susceptible to change when he or she is younger (Campbell et al. 1980, Chap. 7;Krosnick and Alwin 1989). Given that a person spends most of the time together with his or her parents at a young age, it is natural to assume that parents play an important role for political socialization (McIntosh et al. 2007;Hagevi 2011). It has also been demonstrated by Fox and Lawless (2005) that people growing up in a politicized environment possess a stronger nascent political ambition, and thus a higher propensity to run for office as an adult. 4 A similar, yet distinct, socialization-based explanation is that an individual is particularly influenced by parents and peers during the impressionable years. The exact demarcation of when the impressionable years start and end is debated, but late adolescence and early adulthood are usually considered when this period in life is discussed. According to this hypothesis, a person during this period is placed for the first time in a socialization context where politics plays an important part, at the same time that he or she is defining him or herself as an independent individual. The argument is that the political views formed during this period have an abiding effect on adult political beliefs (Manheim 1952;Sears 1975;Sears and Brown 2013).
In brief, it seems there are somewhat conflicting views on when a person is most strongly influenced, although both explanations highlight political socialization. The two views are thus connected and may be interpreted in the light of parental influence. Clearly parents are important according to the crystallization hypothesis, but even the impressionable-years hypothesis is indirectly based on parental selection. A child that grows up in a particular area will go to school with a subset of children whose parents are likely to be similar to his or her own. It is also likely that political discussions initiated at home continue with peers. Socialization thus reinforces itself through various pathways in which parents play an important role (Andolina et al. 2003). When an individual reaches the impressionable years, he or she is put together with peers in a context that reflects many of the characteristics of his or her parents. It is likely that this context promotes similar views as those of the parents during childhood. However, this also means that it is not possible to isolate the causal effect that a parent has on its child from the possible effect that their shared environment, for example the neighborhood in which they reside, may have on them both.
Socialization pathways tend, then, to reflect the political beliefs of parents. These beliefs are transmitted during the impressionable years, in part directly and in part indirectly (through sorting into socialization contexts). 5 The focus in our paper is not primarily on distinguishing between the crystallization sub-hypothesis and the impressionable-years sub-hypothesis, but instead on contrasting both variants of the socialization hypothesis with the materialistic-pathway hypothesis that we discuss in the next subsection. 6 Before we continue, it is worth highlighting some implications of the socialization pathway. Given that parents' active presence is of the essence in this pathway, the absence of a parent-due to death or divorce, for example-should yield a lower parental transmission of political affiliation. According to the socialization-pathway hypothesis, children need to spend time with their parents. The impressionable-years sub-hypothesis too rests indirectly on the idea that this is necessary. If the child lives only with one parent, then it is likely that sorting into neighborhoods and schools is going to be based on the characteristics of this parent, and not primarily on the characteristics of the parent with whom the child does not live. We summarize our discussion in this subsection in one empirically testable prediction: Empirical testable statement 1: the intergenerational transmission of political affiliation will be less pronounced when a child did not live with the politically nominated parent during childhood. 5 It may also be the case that the parent has accumulated party-specific social capital-manifested as status, reputation or personal connections-which the child can benefit from if it chooses to join the same party. Such social capital could make children more likely to follow in the footsteps of their parents, as well as increase the probability of nomination if they do. 6 The reader can find an indirect test of the crystallization and impressionable-years sub-hypotheses in Fig. A2 in the Online Appendix.

The Materialistic Pathway
In contrast to the socialization view stands the materialistic one. This theory emphasizes that political demands are driven by economic conditions; thus the reason why, according to this view, parents and their children have similar political demands is that they share a similar material standard. In this subsection, we discuss the underpinning of the materialistic pathway in words. In the Online Appendix, by contrast, we sketch in a more formalized manner the rational-choice framework which underlies the reasoning here in the main text. We do not create a theoretical model where we solve for equilibria, and our theoretical reasoning consists mainly of modifications of earlier theoretical work. In particular, our simple theoretical framework follows Chaps. 1, 2, and 5 in Persson and Tabellini (2000), 7 but it is extended to take into account intergenerational aspects. It is also related to the work of Verba et al. (2003), who discuss the inequality in political participation that results from the intergenerational transmission of above all education. However, our focus is not on inequality in political activity in general, but instead on partisan intergenerational transmission.
Let us assume that individuals demand two things. They want to consume goods and services, and they also want the public sphere to provide them with a certain degree of redistribution and public goods. In order to buy goods and services, moreover, individuals need an income, a certain share of which they pay in taxes which are then used to finance redistribution and to provide public goods. A person with a high income thus has a better opportunity to consume more, and he or she is also less dependent on redistribution through the public purse. High-income persons, therefore, can be expected to demand less in the way of redistribution than their less well-off fellow citizens.
The second question is what determines the income of the individual. It is natural to hypothesize that the income of an individual in one generation is dependent on that of his or her parents. This may be because the persons in question inherit something from their parents, either economic resources or a shared genetic component, or it may be because the social environment during childhood-in particular the interaction between parent and child-provides opportunities for future income development. Income also depends on an individual component which is not linked to parents. The educational system is such an element, and it has accordingly been seen historically as a means by which to equalize opportunities for individuals from different socioeconomic backgrounds. The importance of the two different factors varies between individuals; in general, however, the higher the parental income, the higher the income of the next generation.
These two sub-hypotheses are cornerstones of the materialistic hypothesis. They are also easily linked to choice of political party. Right-wing parties, in general, favor lower taxes and less redistribution; left-wing parties favor higher taxes and more redistribution. 8 High-income earners are therefore more likely than their lowincome counterparts to prefer parties of the right. This also means a lower likelihood that the child and the parents will be affiliated with the same political party when they have different incomes in adulthood, due to the differing political demands arising from different income levels. Let us summarize our discussion in the statement below: Empirical testable statement 2: the intergenerational transmission of political affiliation will be less pronounced when the income differential in adulthood between the child and the parent is larger.
Let us turn now to selection issues. In democracies, the actual voting choice of a given individual is not public information. There is simply no register available from modern democracies with information on party choice. 9 We can observe, however, the political affiliation of those who run for political office. Our empirical analysis will be focused on these individuals. This means we have a selection issue, given that we can only observe those who have chosen to participate as politicians. The core question is whether political-party affiliation can function as a proxy variable for underlying political demands. Persons who are nominated politicians may certainly be affiliated with a political party because they have mapped their political demands on to the policy platform of that party. However, they may also have chosen to be a politician in order to please their parents. Or, conversely, they may have chosen not to become a politician in order not to anger their parents because of political disagreement.
The choice to run for office is a classic theme in political-economy models. In older models, politicians were seen as exogenous, office-motivated, and inclined to propose policy platforms in line with the views of the median voter. This is the central conclusion in the Hotelling-Downs model (Hotelling 1929;Downs 1957). More recent theoretical models, however, have portrayed the choice to become a politician as endogenous (Osborne and Slivinski 1996;Besley and Coate 1997). In this understanding, citizen-candidate politicians originate from the electorate, where all voters have political demands. For some individuals, however, it is beneficial also to run for office in line with such political demands when the choice to run is dependent on the cost of running. In the last part of the game, those who run for office and win election implement a policy in line with their true political demands, given that a different policy does not make it worthwhile for them to run for office in the first place.
If the citizen-candidate model is correct, then the individuals we observe running for office for a specific party do so because said party is a good representation for their political demands. For these candidates, we can also indirectly assess the intermediate theoretical steps in the materialistic pathway, such as individual income and the intergenerational transmission of income from parents, given that we have access to registry data.

Institutional Setting
Let us now review some basic information about the Swedish political system. Elections took place in Sweden every third year prior to 1994, and every fourth year after that. Elections for municipal councils, county councils, and the national parliament are held on a single day. Sweden has a system of proportional representation with closed lists, where position on a party list is important for one's prospects of election.
For the time period we study, Swedish political parties may be divided into two different blocs: the center-right and the center-left. The center-right bloc consists of the Moderate Party (conservatives), the Christian Democrats, the Liberal Party, and the Center Party. The center-left includes the Green Party, the Social Democrats, and the Left Party. There is also an eighth party, the Sweden Democrats, a rightwing populist party which is not normally considered part of either of the two blocs mentioned. The Sweden Democrats were small for a long time, but they entered parliament in 2010 and they have gained strength in recent years. Given that the Swedish political arena is clearly divided between a center-right and a center-left bloc, we will use these bloc variables in some of our empirical specifications, which we explain further in the section on results.
Each party has different nominating traditions. Some parties hold member primary elections to fix positions on a party list before the general election. In other parties, the local party board decides on the final list. Some political parties use a combination of primary elections and board decisions. While there can be fierce competition for the top positions on a ballot, active members of a party do not usually find it too difficult to gain a spot on the list.
In many countries, it is customary for politicians to originate from the upper classes. In view of our discussion of selection in the theoretical-framework section, it would be problematic if our data sample were highly selective in terms of social background and income-especially since we are interested in investigating the link between income and political affiliation. Dal Bó et al. (2017) investigate the background characteristics of Swedish politicians, and find them to be representative of the general population in terms of parental income and occupational background. Politicians appear, however, to be more intelligent on average than the population as a whole. 10 10 A question related to ours is whether the children of politicians get economic favors because their parents are politicians. This issue has been investigated by Folke et al. (2017), who actually finds that the income of children is increased if the parent is a top politician on the municipal level in Sweden. The authors conclude, however, that this finding seems to be driven by an increase in the probability that the child in question stays in the home municipality and works there, instead of moving in order to pursue higher education.
Another question is what kind of case Sweden constitutes. One the one hand, Sweden has traditionally exhibited a significant degree of class-based voting (Oskarson 2015). While class-based voting has declined over time, it may still be something which strengthens the intergenerational correlation compared to countries where class is less important. 11 On the other hand, Sweden has many different political parties, meaning that individuals can get involved in politics through a variety of platforms. This means in turn that the intergenerational transmission of party affiliation should be less pronounced than in a political system where there are two bigtent parties.

Data and Empirical Framework
The data material originates from Swedish population registers. What is unique with our data is that it covers the entire population of politicians at the national, county, and municipal levels. The data set begins in 1982, which is the starting year in our study; 2014 is the last year. In addition to our data on politicians, moreover, we have access to a multigenerational data set which enables us to link children to their parents and individuals to their siblings and grandparents. This is crucial given the purpose of this paper. Our sample consists of all politicians with a family connection.
Being a politician is relatively rare. In order to increase the sample size, therefore, we focus on all nominated politicians in our analysis, regardless of whether they won election or not. Focusing on nominated politicians is also natural if we are interested in political affiliation and not the intergenerational inheritance of elective office. We thus choose to add nominated politicians from all three national elections in Sweden together in our empirical analysis. 12 We should clarify our terminology on one point. We refer to individuals in our empirical analysis that belong to the last generations and who have been nominated between 1982 and 2014. These individuals have parents, siblings, and grandparents, who may in turn also have been nominated between 1982 and 2014. 13 This group of people constitutes our empirical sample when we investigate the transmission of political-party affiliation. In our empirical analysis, we restrict the sample to those cases where the individuals are nominated later in time than their parents, siblings, or grandparents, given that we are interested in the transmission of political-party affiliation to the individuals.
It is worth noting that the restriction of the analysis to the years to 1982-2014, means that we do not investigate the transmission of party affiliation of those 11 Note that this argument assumes that class mobility is lower than other sources of inter-generational mobility in party affiliation. Depending, however, on how strong the correlations that we find are, this assumption may not be true. 12 In those relatively rare cases where a person has been nominated for different political parties in different elections, we choose the party which the person has represented the largest number of times. 13 Another way to denote these individuals would be to call them children, since we are investigating intergenerational transmission between parents and their children. When we measure the outcome variables, however, these people are adults, which makes it more accurate simply to call them individuals.
individuals that are politicians during this period, but whose parents ran for office prior of 1982. This should above all include those with older parents or those parents that ran for office at a younger age.
To test the socialization pathway, we need to know whether individuals grew up with their politically nominated parents. We start by counting the number of years that a child and his or her parents lived together between the year the child was born and the year he or she turned 18. Because children with separated parents are registered in our data as living with either their mother or their father-even as it is actually quite common that they live part-time with the other parent as well-we count children as living with their parents as long as they reside within the same county (län). Due to this very broad definition, children not coded as living with one of their parents will be unlikely to meet them on a regular basis. Based on this number, we create one variable that measures the share of these years spent living together, as well as one binary indicator for whether this share amounts to a majority of the time. As a robustness check, we also create a similar variable that is based on a smaller geographical unit. 14 To test the materialistic hypothesis we need income data, which we obtain from the LISA register. 15 We standardize income for each cohort for the first year after the age of 33. Earlier literature has highlighted how problematic it is to use current income as a proxy for lifetime income (Engström and Hagen 2017;Böhlmark and Lindquist 2005). After reaching 30, most individuals who have enrolled in higher education have established themselves on the labor market; we therefore choose the age of 33.
Our empirical analysis for the benchmark results is very simple. Focusing on the party level, we calculate the share of nominated individuals who had a parent that was nominated for the same political party. We also calculate the equivalent share for grandparents and siblings in separate specifications. This descriptive analysis addresses the direct transmission from parents, siblings, and grandparents. 16 However, we need something to relate these shares to in order to assess the magnitudes. We therefore calculate the probability that two randomly drawn politicians belong to the same political party, which we interpret as what the number who run for the same party as their relatives would have been had there been no within-family transmission. We calculate these probabilities for three different scenarios: (i) where the two politicians are drawn from eight parties of equal size; (ii) where they are both from the universe of Swedish politicians between 1982 and 2014 (excluding parties without representation in the national parliament); and (iii) where one of the politicians is a randomly drawn parent from our sample, and the other is drawn from the pool of candidates who ran for office in the same municipality and year as that 14 These calculations are based on SAMS areas, which are sub-units within municipalities. A SAMS consists roughly of 1000 individuals. 15 LISA is a longitudinal register created for labor market research. 16 Throughout the paper, we cluster standard errors at the municipal level, because the probability of being nominated and indirectly running for the same political party as one's parents is clustered within municipalities. The reason for not clustering the standard errors at the family level is that most families yield only one observation, namely a dyad of one parent and one individual. parent's child. 17 We later turn to a regression framework in order to run robustness checks and mechanism analyses; we explain the details below.
It is worth noting that our main investigation is not a causal-effect analysis; that is, we do not pinpoint a treatment for which the effect can be estimated using a convincing identification strategy. Such an analysis would be interesting too, but it would answer a different and more local question. 18 The focus in this paper is instead Table 1 Proportions of individuals affiliated with the same party as family members The proportions refer to the share of candidates that ran for the same party as their parent, sibling, or grandparent. In the first row, the sample includes every candidate with either one or two parents who have also been nominated for one of the eight parties in the national parliament, regardless of whether any other relatives have run for office. However, candidates are excluded if the two parents ran for different parties. In rows 2-4, the sample is restricted to those who have either one or more parents, siblings, or grandparents who have run for office. The last three panels show the proportion that would be expected if candidates were randomly drawn (i) from a pool of candidates distributed equally between eight parties, (ii) from the universe of Swedish candidates between 1982 and 2014 (excluding parties without representation in the national parliament), or (iii) from a pool of candidates that ran in the same municipality and year as the candidates in our sample. Standard errors are clustered at the municipal level on overall interdependence across the generations, which by nature is a correlational question. However, given that we are restricting the analysis to cases where the child is nominated after the parent, we do not have the problem that the transmission in fact goes in the opposite direction. One strategy we could apply is to control for various background characteristics. However, by holding factors constant, we are essentially controlling away parts of the overall transmission. For example, if we controlled for parental characteristics as income or employment status, we would be left with the part of the transmission which is unrelated to economic conditions. To pinpoint a portion of the transmission is not our goal, which is instead to investigate the gross interdependence and the mechanisms behind it.

Benchmark Results
The benchmark results are presented in Table 1. Let us first look at the transmission from parents to their children. We restrict the sample to those nominated individuals who also have one parent who ran for office during the 1982-2014 period. We also put the restriction that, where there are two politician parents, they must belong to the same political party; however, we do not care whether they have other relatives in politics. Under these restrictions, we can conclude that as many as 76% of all candidates ran for the same party as their parent (row 1 in Table 1).
To put this into perspective, we must remember that some children would run for the same party as their parents even in the absence of any intergenerational transmission. If we relate the number above to the probability that two randomly drawn candidates belong to the same party, we can interpret the difference as representing how much larger the actual share who run for the same party as their parent is compared to what it would have been had the intergenerational correlation been zero. These probabilities are presented in rows 5-7, and they range from 12.5% (eight parties of equal size) to 19% (if we take into consideration where and when each individual runs for office). Taken together, this descriptive evidence points towards a strong transmission of political affiliation from parents to their children.
In rows 3-4 in Table 1, we turn our attention to the transmission from siblings and grandparents. In this case the sample restriction is slightly modified. Here we put a restriction that the individual had one or more siblings (row 3) or one or more grandparents (row 4), but no other relatives running for office. For comparison, we also calculate the transmission from parents under this modified sample restriction in row 2, where the individual had one or two parents nominated for office, but no other relatives. Beginning with row 2, the share of individuals that had one or two parents nominated for the same political party is, as expected, close to the share we calculated in row 1. What is interesting is that there is a transmission from siblings, and also to some degree from grandparents. For those individuals that had one or two siblings who were nominated for political office but no other politician relatives, the share of individuals that ran for the same political party is around 60%. 35% who had grandparents nominated for political office ran for the same political party as their grandparents. Both shares are higher than the probability that two randomly drawn candidates would belong to the same political party. It thus seems that intra-extended family transmission is present even in the absence of parents who were nominated politicians. In sum, parents seem to be the strongest source for the transmission of political-party affiliation; however, direct transmission is also present from siblings and to some degree from grandparents.
In the Online Appendix, we address potential objections to our baseline results. This robustness discussion and analysis, presented in Table A1, is primarily concerned with checking whether intergenerational transmission is affected when we focus on individuals who probably run for office because of a genuine will of getting elected, and not only as a favor to their parents. Our conclusion from this analysis is that the estimated transmission is still present under these robustness specifications. We also estimate the intergenerational transmission over time, in Fig. A1. We conclude that, while it has declined over time, it is still high for the later election years. In Fig. A2, we investigate whether the probability of running for the same political party as one's parents depends on the age when the parent was nominated. The conclusion is that the transmission is strong for all ages, but that it decreases somewhat when the parent is more advanced in years. In Tables A2, A3 and A4, we run the same analysis as in Table 1, but for those individuals nominated for the parliament, the county councils, and the municipal councils separately. We find that the intergenerational transmission of party affiliation is as strong among politicians in the national parliament as among their counterparts on the local level.

Mechanisms
We have now concluded that intergenerational transmission is strong with regards to political affiliation. How are we to explain this transmission? Let us return to the empirically testable statements that we made in the theoretical section. The first statement concerned the socialization pathway. In order for this mechanism to be confirmed empirically, the intergenerational correlation must be weaker if the child did not grow up with the nominated parent. Similarly, according to the materialistic pathway, the intergenerational correlation should be weaker for those children who differ from their parents in terms of their position in the income distribution.
To test these implications, we leave the descriptive statistics used in Table 1 and turn instead to a regression framework, where it is easy to include additional variables. To mimic the specification in the last subsection, we stick to the sample restriction in row 1 in Table 1, where the outcome variable equals 1 if the individual and the parent(s) were nominated for the same party, and 0 otherwise. Another way to express this share is simply to run a regression on the constant: SameParty i = i + u i . i is then the share of individuals that were nominated for the same political party as their parent. This specification is used in Column 1 (full sample) and Column 2 Table 2 Testing the socialization versus the materialistic pathway The dependent variable is a binary indicator for whether an individual was nominated for the same party as either or both of his or her parents. The sample includes candidates with either one or two parents who have also run for office; however, the candidate is excluded if the two parents ran for different parties. In Columns 2-5, the sample is restricted to observations without missing data on any of the variables used in these models. Children are coded as living with their parent if they reside within the same county (län). Standard errors in parentheses are clustered at the municipal level * p < 0.10, * * p < 0.05, * * * p < 0.01 (1) (3) Rest. Rest. Rest.
Adjusted  table) of Table 2, and the constant in Column 1 is identical to the share in the first row of Table 1. 19 Let us now test the socialization and materialistic hypotheses. To test the socialization hypothesis, we begin by creating a binary indicator for each year the individual and the parent lived together. Based on these binary indicators, we define one variable for the share of years between birth and age 18 that the individual lived together with the politician parent, and one binary variable for whether this amounts to more than or less than half the time.
Columns 3 and 4 in Table 2 concern the socialization pathway. Here the reader should pay attention to how the estimated constant is altered when the variables for living with the politician parent are included in the regression. What we find is that most of the intergenerational transmission is conditioned on living with the parent, and that the share of individuals who are nominated for the same political party as their parent decreases substantially if we focus on those who did not grow up in the same county as their politician parent [31% for those who never lived with their parent, compared to 76.6% (0.456 + 0.310) of those who lived with their parent between birth and age 18-assuming we trust the specification in Column 3]. This yields strong empirical support for the importance of the socialization pathway, and puts an upper limit on how much of the intergenerational correlation in party affiliation could potentially arise from a shared genetic heritage not dependent on social factors, or from some other factor which does not require that the parent and child live together.
It bears noting that the intergenerational transmission of political affiliation is still reasonably large even when the individual in question did not live with his or her politician parent, as illustrated by the estimated coefficients for the constant in Columns 3 and 4 in Table 2. However, even if we randomly drew pairs of politicians, 17.7% of them would belong to the same party (see Table 1), so the increase resulting from intergenerational transmission should not be overstated. As a robustness check, we have also based the socialization variables on SAMS instead of on counties. A SAMS is a much smaller geographical unit than a county. The results are presented in Table A6 in the Online Appendix. As expected, the constant is still reduced when the socialization-pathway variables are included in the regression, but the reduction is smaller.
Column 5 in Table 2 concerns the materialistic pathway. We include a variable that measures the distance between the child's and the parent's position in the income distribution, coded from 0 (same percentile group) to 1 (the maximum distance of 100 percentiles). For both the child and the parent, the income is measured at the age of 33, and the percentiles are calculated with data on other individuals within the same year and cohort. In contrast to Columns 3 and 4, we do not see a large change in the estimated constant when we include this variable. In fact, the constant remains virtually unchanged, indicating that the propensity to run for the 19 We use a linear probability model (LPM) in order to facilitate interpretation and because the means in Table 1 can then be translated into constants in Table 2. The corresponding models estimated with logistic regression are presented in the Online Appendix (Table A5), and the results are quite similar. same party as one's parent is not dependent on having a similar income. This is strong evidence against the materialistic hypothesis, which assumes that partisan similarity is caused by similar economic interests.
While we believe the analysis presented here is enough to refute the materialistic hypothesis, we have run some additional analyses, which can be found in the Online Appendix. In Table A7, we show there is an intergenerational correlation with regards to income, and that higher relative income increases the probability of an individual being affiliated with a center-right political party; but we also show that, taken together, these correlations are far too weak to have a non-negligible impact on the intergenerational transmission of party affiliation.

Supplementary Analysis
We have now demonstrated there is a large intergenerational transmission from parents to children with regards to party affiliation. The mechanisms behind these results seem to be in line with the socialization hypothesis rather than the materialistic one. We have also found there is a direct transmission from siblings and grandparents. Our data allows us, however, to expand from these findings by using additional information available in the register data, which further illustrates the transmission of political-party affiliation within the family.
The first thing we can do is to look at heterogeneity with regards to gender, and ask ourselves whether the transmission is different from males and females, and whether males and females incorporate the political affiliation from fathers and mothers to an equal extent. The results are presented in Table A8 in the Online Appendix. In essence, we find that the transmission is not heterogeneous with regards to the gender of the individual. What we do find, however, is that having the same gender as the nominated politician parent further increases the share of individuals running for the same political party as their parents. These results connect up with Folke et al. (2020), who use a cross-country data set on legislators as well as data on individual candidates from Ireland and Sweden. The authors find that women are more likely to enter politics if they have a relative who has been a politician.
The second thing we can do is to incorporate cases where there are two politician parents who belong to different parties-i.e., when there is disagreement within the family. In this case, we can investigate partisan transmission where we have variation across political parties. In order to keep this part of the analysis relatively contained, we focus on the two political blocs. 20 In this case, we are essentially using 20 Otherwise, all specifications would be run for all political parties. Even if we exclude all local parties, we still have eight political parties in all. In this part of the analysis, we cannot define a SameBloc variable, since parents may belong to different political blocs; moreover, the whole point of the analysis is to separate the transmission from the mother from that of the father. We thus need another regression specification. In this case we estimate CR i = 0 + 1 FamilyMembersCR i + e i , where CR stands for centerright. Given that we restrict the analysis to those political parties which are in either the center-right or the center-left bloc, the results will be equal for a specification with the center-left. the variation that arises when different family members belong to different political blocs in order to estimate the parameter value for the coefficient of interest.
The results are presented in Table A9. The sample in this case consists of those families where the individual, the mother, and the father have all been nominated for a political party. First we estimate the transmission from fathers and mothers respectively for the entire sample, and then for men and women separately. Overall, it appears that the political affiliation of mothers matters more than that of fathers when the mother and the father disagree (and this holds for both male and female children). This result-that mothers dominate when the parents disagree, but that the respective influence of the two is more equal when the other parent is neutralechoes the conclusions drawn in earlier literature (Beck and Jennings 1975). Our results also relate to Oberle et al. (2011), who find that transmission of political opinions goes from mothers to daughters and from mothers and fathers to sons. Our findings may also be viewed in light of Shulman and DeAndrea (2014), who argue that fathers and mothers discuss politics differently with their children-mothers being more influenced by the opinions of their children, and fathers employing a more downward type of communication. Another way of putting it is that mother discuss politics with their children, while fathers tell their children about politics and that discussing tends to be more effective. In our case, the mothers are unlikely to be affected by their children, since all of the parents in question were nominated as politicians before their children.

Discussion and Conclusion
Understanding how political beliefs are formed is one of the central themes of political science. Our focus has been on the intergenerational aspects of political-party affiliation. While families with multiple politicians are quite rare overall, they hold electoral and career advantages (Geys and Smith 2017;Smith and Martin 2017) that make them overrepresented among the world's top politicians. A common criticism of the Bushes, Clintons, and Kennedys has been that they take up space that might otherwise be available for candidates with new ideas and different perspectives (Pisani 2015). Understanding how and to what degree political beliefs are formed within the context of the family, therefore, is of utmost importance for understanding the larger question of the accumulation of political power.
In this paper, we have applied rich registry data from Sweden. Our finding from this empirical endeavor is that intergenerational transmission is very strong. These results might be thought unsurprising, given that parents are likely to play an important role in a person's life; however, they are surprising in that Sweden, according to the World Value Survey, is the country with the highest degree of self-expression values in the world (WVS 2018). Our paper has thus made a contribution in highlighting the persisting importance of family transmission in this age of individualism. The intergenerational link is likely to be even more pronounced in countries where traditional views on family and tradition dominate. Furthermore, we have demonstrated the presence of intergenerational transmission not just between children and their parents, but also between individuals and their grandparents as well as between siblings. This indicates that the transmission of political affiliation takes different pathways and that it reinforces itself through a variety of channels in proximity to the individual.
Our second aim in this paper has been to connect up with a theoretical discussion of how this intergenerational transmission manifests itself. We started out by discussing a socialization hypothesis, according to which discussions at home together with parents or with peers are the key drivers. We then discussed another theory, to the effect that relative differences in income are the key to understanding political demands. Because income in this view is inherited across the generations, political affiliation is as well. When we brought these empirical predictions to the data, we found that intergenerational transmission decreased substantially when an individual had not lived together with the politician parent in question. If the individual has had less opportunity to discuss politics at home with the politician parent, this is exactly what we would expect, according to the socialization hypothesis. When we tested the materialistic hypothesis, we did find partial support for some of its elements. Income is empirically correlated across generations, and individuals with higher relative incomes are more likely to be nominated for a center-right political party. When we add the components together, however, the benchmark intergenerational transmission is not affected by the inclusion of income. The materialistic pathway seems thus to play a much less pronounced role in explaining the intergenerational correlation.
While our results indicate that spending time together with parents during childhood is associated with a much higher probability of the individual's belonging to the same political party, we should underline that an element of intergenerational transmission remains even when the child did not live with the politician parent in question. We cannot determine whether this transmission is genetically inherited or if instead it reflects socialization that takes place even when the child is not living with his or her parent. However, we may interpret this as an upper bound for genetic transmission that is not dependent on social surroundings.
It bears stressing that we have focused exclusively on transmission within families in this paper, and that we have not considered other arenas of socialization such as neighborhoods. As discussed in the theoretical-framework section, selection into neighborhoods is another important factor to consider when piecing together the puzzle of how political beliefs and opinions are formed. Future research should take this context into account.