TY - JOUR
AU - Rickne,, Johanna
AB - Abstract We exploit close elections in Swedish municipalities to test whether relatives of politicians who become mayors obtain economic benefits. We find no benefits for the siblings of new mayors, but the average earnings of children of newly appointed mayors rise by about 15%. Administrative information on occupational and residence status show that the higher earnings are unlikely reflect an illegitimate allocation of jobs. But the evidence suggests that children of election‐winning mayors are more likely to postpone tertiary education and remain in their parents’ municipality to work. In this article, we investigate if close relatives to top political executives receive economic benefits. We call these benefits dynastic political rents and study their existence in Sweden, a political environment with one the lowest corruption levels in the world (see transparency.org, Wångmar, 2013; Dahlström and Sundell, 2014). If close relatives benefit economically in Sweden, such benefits are likely to arise in environments with less transparent institutions. Anecdotal evidence suggests that economic benefits can trickle down through channels in legal grey‐zones, which only indirectly relate to the politicians themselves. Local governments in Sweden employ 20% of the country’s total labour force and allocate large sums via public‐procurement contracts. Relatives of politicians in power may be better informed about contract and employment opportunities, and bureaucrats who allocate jobs and contracts may have incentives to favour the relatives of their political superiors. In recent years, Swedish media has reported scandals of allocating publicly funded ‘summer jobs’ to the children of insiders. A recent government inquiry found that the public procurement system lacked adequate oversight (SOU, 2013) and investigative journalists have uncovered cases of questionable allocations of contracts to companies owned by relatives to top local politicians.1 We add to the insights from recent studies on dynastic political rents (Fafchamps and Labonne, 2014; Amore et al., 2015; Gagliarducci and Manacorda, 2015), in terms of both data quality and estimation methods. We use register data based on personal identification codes that automatically link children and parents at birth. This sidesteps the need to approximate family ties with shared last names. We have access to 20 years of panel data for the universe of municipal politicians and every one of their children and siblings. These data allow us to differentiate between different sources of income and different employment sectors and jobs, e.g., to test separately for effects on earnings from small businesses and public‐sector jobs. We can also test for impacts on the behaviour of relatives, most importantly – it turns out – the priorities of children between tertiary education and job opportunities in their top‐political parent’s municipality. Apart from capturing a wide variety of economic benefits and behaviours, the register data we use are of uniquely good quality, as they are collected by the government via tax returns and mandatory employer surveys. Our empirical strategy differs from some of the previous work. Amore et al. (2015) also rely on detailed administrative data, although for Denmark. However, they identify the effect of larger political power from a set of municipality mergers, comparing re‐elected politicians in merged (larger) municipalities with re‐elected politicians in non‐merged (smaller) municipalities. It is not entirely clear how to interpret and extrapolate this kind of treatment effect. Our approach instead relies on close elections in Swedish municipalities allocating positions of power to top‐ranked politicians in the party that appoints the top local executive. We compare the economic outcomes for children and siblings of close electoral winners to those of close electoral losers, i.e., top‐opposition politicians. Such a treatment effect of political power is more readily interpretable and more easily extended to other contexts than one that rests upon the relative size of the local assembly. Using municipal‐level data gives us a large number of observations without sacrificing either the economic importance of the position or the potential to extrapolate lessons outside of Sweden. Like elsewhere, Sweden’s local political assemblies have substantial economic and political powers, which are largely concentrated in top political executives (Jonsson, 2003; Montin, 2007). The municipal council boards, each led by a mayor, hold employer responsibilities for local public employees making up more than a fifth of total employment. They are also in charge of public spending making up about a fifth of Swedish GDP. Among these expenditures, public procurement accounts for between 10% and 15% (Swedish Competition Authority, 2015). Our analysis uncovers several new and interesting results. We find that the relatives of top‐ranked politicians in the local government obtain higher average earnings than relatives of top‐ranked politicians in the opposition in the period after a close election. Our identification strategy suggests that this is a causal effect, rather than a correlation based on pre‐existing differences. When we examine this effect in depth, it seems to be concentrated on relatives at the lower end of the income distribution. It also appears that the politician’s children, rather than their siblings, experience the largest increase in earnings. Further, the children of the highest ranked politician make the largest gains, conditional on living in the same municipality as their parent. Looking for the mechanisms behind the children’s additional earnings, we find little evidence that they run through shady channels. There is no effect on the probability of holding a job in the municipality’s public sector. Neither is there evidence that the children take over the job, e.g., at the family farm, held by their parent before taking up full‐time political office. We also analyse children’s behaviour in terms of study versus work and their municipality of residence. The evidence suggests that children to powerful politicians become more likely to live in the municipality where their parent holds power but less likely to be students. While these findings are not very precise, they suggest a possible explanation for the higher earnings found in the baseline estimation: children delay their entry into university and instead look for a job in their home municipality, when their parent is awarded the top political office. The finding that electoral victories of politicians may shape the behaviour and choices of their relatives – e.g., regarding residence – provides an interesting perspective on the small but growing body of research on dynastic political rents. For example, if family ties are inferred from shared last names between politicians and residents in the same geographical unit, geographical sorting of relatives will bias the results. Shifts in average earnings, or in occupation distributions, may as in our case, stem from mobility responses among relatives, rather from illegitimate favours extended to pre‐existing relatives in the same region. Our article contributes not only to the small literature on dynastic political rents but also to the wider literature on the roots and consequences of dynastic links. As for the consequences, a sizeable literature follows Fisman (2001) in investigating the value to firms of having political connections, often by conducting event studies of stock prices. A smaller number of papers examine if dynastic links between politicians generate differences in economic and social outcomes (Braganca et al., 2015; Labonne et al., 2015) and a growing body of work debates if relatives to politicians face a lower entry barrier into political office (Dal Bo et al., 2009; Querubin, 2013; Geys, 2017; Rossi, 2017; Van Coppenolle, 2015). Another strand of research examines the monetary returns from holding political office. A standard finding in this literature is that returns exist for political offices at the national and state levels (Eggers and Hainmueller, 2009; Kotakorpi et al., 2013; Fisman et al., 2014). The evidence is more mixed at the local level (Kotakorpi et al., 2013; Lundqvist, 2013; Amore et al., 2015). We contribute to this literature by investigating if monetary payoffs might also extend to the immediate relatives of politicians, extending the cost–benefit calculus of holding political office to advantages that may trickle down to the politician’s loved ones. The article is organised as follows. Section 1 gives some basic institutional information about Swedish municipalities. Section 2 presents our micro data, sample restrictions, outcome variables, as well as some summary statistics. Section 3 explains the estimation strategy. Section 4 presents the baseline results. Robustness checks and discussions of channels behind the results appear in Section 5. Section 6 concludes. 1. Municipal Assemblies and Politicians in Sweden Sweden has 290 municipal assemblies that vary in size between 31 and 101 seats, depending on the municipality’s population (the median municipality has about 15, 000 inhabitants). Municipal assemblies have substantial political and economic powers. Their right to local self‐government is granted by the Swedish Instrument of Government and, under the 1991 Local Government Act 2.1, local authorities are themselves responsible for matters of public interest, which relate to the municipal council and its inhabitants and are not the exclusive responsibility of the state. These responsibilities cover broad areas of public spending such as child and elderly care, schools and local infrastructure. Electoral institutions across all municipalities have list‐based proportional representation (PR) elections for the council every four years. Voters chose between party electoral ballots and seats shares in the municipal assembly are awarded to reflect the party vote shares.2 This way, almost all municipalities have some representation of all parties in the national parliament and often of unaffiliated local parties (the latter hold 2% of the seats). After the election, a governing coalition is formed and the largest party in this coalition appoints the head of the council board. Sweden has two stable political blocs: the left, i.e., the Social Democrats, Left party and Green party, and the Centre‐right, i.e. the Conservatives, Centre party, Liberal party and Christian Democrats.3 Both blocks strongly prefer within‐bloc coalitions. Data from 1982 to 2010 reveal that when either bloc obtained more than 50% of the seats, the largest party in that bloc had a 90% probability of appointing the mayor. Once a governing coalition has been formed, it appoints the most important executive positions and these usually go to the top politicians in the largest coalition party. The top executive positions include the chair of the Council Board, the chair of the council and the chairs of subcommittees in different domains. The most powerful position in the municipality is the council‐board chair: since this is the Swedish analogue to mayor, we use this label throughout the article. The political opposition usually appoints the vice chairs, with the most influential position going to the top‐ranked person in the largest opposition party (see the descriptive statistics in Section 2). Top politicians have disproportionate political influence compared to rank‐and‐file members of the municipal council. Since the establishment of the mayor’s office in the 1960s, power has been increasingly concentrated in these offices and individuals (Nilsson, 2001). Across municipalities, mayors have similar powers and obligations, most importantly to formulate and implement the agenda of the mayor’s party and political coalition (Jonsson, 2003). The mayor also represents the municipality in regional and national contexts. The second‐most important positions are the subcommittee chairs, who handle the implementation of policies in their respective areas. The chairperson of the council also has some influence but less than the other executives. 2. Data, Sample Restrictions, and Descriptive Statistics 2.1. Data Our dataset covers the relatives of all municipal politicians elected to a council in the six elections between 1991 and 2010. All politicians on all electoral ballots are linked to balanced panel data for their close relatives using personal ID codes. These codes are available for every politician, since parties are legally obliged to report to them to the electoral agency for every individual on the ballot. At birth, Swedish residents obtain a personal identification number, which is used for all government record keeping. For our purposes, we use data from the generational and sibling register to link every elected politician to his/her siblings and children. In a second step, we use balanced panel data, for each of the 21 years 1991–2011, registering the earnings, employment, municipality of residence, as well as other relevant variables for every relative. None of the data used in our article is thus self‐reported. Parenthood is registered at birth by Swedish public hospitals. Earnings are third‐party reported to the tax authorities by the employer. Other registers record everyone’s municipality of employment and residence. In the main analysis, we use three outcome variables for the earnings of the children and siblings of politicians. Earnings are measured during the election period, excluding election years (more on this below). Our first earnings variable, total earnings, sums employment earnings, business income and those social insurance benefits directly based on wages and conditional on employment (such as parental‐leave insurance or unemployment insurance). From this sum, we separate out the components of employment earnings (outcome variable number two) and business earnings (outcome variable number three). All earnings variables are deflated by the CPI. We conduct our empirical analysis for earnings measured both in thousands of Swedish Kronor and in logs. The main reason for showing both sets of results is that children of politicians have rather modest earnings. The log results can thus be misleading, as a small amount can translate into a large percentage deviation from a low average. By comparing estimates for monetary values and logs, we can better interpret the economic significance of our findings. In addition, the logged earnings variables are closer to normally distributed than monetary earnings. Online Appendix C shows graphically (Figure C1) the frequency distributions for our main outcome variable, total earnings, and their logarithm, separately for children and siblings in the full sample. In the logarithmic transformation, we use (1 + earnings) to avoid missing values.4 In the extended analysis, we add three more outcome variables. These are dummies for: (i) holding public‐sector employment; (ii) being a student; and (iii) living in the municipality of the relative who holds political office. 2.2. Sample Restrictions Our estimation sample consists of children and siblings of working age, i.e. people between 18 and 64 years old. The relatively high age of the top politicians precludes us from looking at labour‐market outcomes of their (mostly retired) parents. Spouses are excluded for two reasons. First, electoral victory causes non‐random divorces. As we show elsewhere (Folke and Rickne, 2016), younger female politicians are more likely to divorce when they win political power, while the divorce probability for male politicians goes the other way. Comparing the wages earned by spouses of winning and losing politicians is hence confounded with sorting in and out of marriage which alters average wages of partners, because relatively high earners (spouses of the female winners) drop out of the winner sample (Folke and Rickne, 2016). Second, electoral victory may induce income and substitution effects on the spouse’s labour supply.5 The sample for estimation includes only siblings and children of the top‐three politicians on the electoral ballots of the largest parties in the left and right blocs. We use the top three politicians as these have the most political clout potentially to affect the earnings of their relatives. Rank‐and‐file politicians have less influence and, indeed, running our analysis on the relatives of the rank‐and‐file members of the municipal council does not yield any significant findings. Another sample restriction is to drop observations of the election‐year outcomes of relatives. In election years, politicians’ (potential) rent extraction could be curtailed by heightened scrutiny from voters, other political parties, and the media.6 2.3. Descriptive Statistics The total dataset consists of 23,895 individual election‐period observations for earnings and socio‐economic background variables of children and siblings of top politicians. Table 1 shows summary statistics for this sample, split by governing status (majority or opposition) of the relative in political office, as well as by party rank (top name or top three). As expected, siblings are older than children, have higher earnings and shorter education. About half of children and a third of siblings live in the municipality where their relative holds office. Interestingly, we see few differences in the average earnings of relatives to majority and opposition politicians. There are also few noticeable differences between the relatives of top names versus second and third‐ranked politicians. When we present our empirical results, we also provide balance tests for personal traits and lagged outcomes. These tests are conducted for each specific sample in the empirical analysis, and all of them show balance in terms of both predetermined traits and lagged outcomes. Table 1 Summary Statistics for the Children and Siblings of Elected Politicians by Electoral Status, 1991–2011 Panel (a): children . Electoral status of parent . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 27.07 27.28 27.50 27.57 All earnings* 149.05 149.66 151.73 150.26 Employed (1/0)** 0.91 0.90 0.91 0.90 Employment earnings 138.06 139.09 141.55 139.20 Business owner (1/0) 0.03 0.03 0.03 0.04 Business income 2.96 2.33 2.29 3.11 Capital income −14.46 −6.99 17.75 −46.93 Student (1/0) 0.31 0.29 0.30 0.28 Lives in same municipality (1/0) 0.54 0.53 0.50 0.51 Years of education 13.31 13.19 13.32 13.08 Woman (1/0) 0.49 0.49 0.47 0.48 Panel (a): children . Electoral status of parent . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 27.07 27.28 27.50 27.57 All earnings* 149.05 149.66 151.73 150.26 Employed (1/0)** 0.91 0.90 0.91 0.90 Employment earnings 138.06 139.09 141.55 139.20 Business owner (1/0) 0.03 0.03 0.03 0.04 Business income 2.96 2.33 2.29 3.11 Capital income −14.46 −6.99 17.75 −46.93 Student (1/0) 0.31 0.29 0.30 0.28 Lives in same municipality (1/0) 0.54 0.53 0.50 0.51 Years of education 13.31 13.19 13.32 13.08 Woman (1/0) 0.49 0.49 0.47 0.48 Panel (b): siblings . Electoral status of sibling . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 49.51 48.17 49.95 48.29 Total earnings 236.83 221.21 230.17 208.96 Employed (1/0) 0.89 0.88 0.87 0.85 Labour earnings 222.22 205.05 217.15 192.75 Has business (1/0) 0.09 0.09 0.08 0.09 Business earnings 6.47 7.97 6.14 6.60 Capital income 162.39 10.36 23.24 26.03 Student 0.02 0.03 0.02 0.03 Lives in same municipality 0.33 0.36 0.30 0.31 Years of education 12.74 12.46 12.53 12.48 Woman (1/0) 0.49 0.49 0.49 0.49 Panel (b): siblings . Electoral status of sibling . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 49.51 48.17 49.95 48.29 Total earnings 236.83 221.21 230.17 208.96 Employed (1/0) 0.89 0.88 0.87 0.85 Labour earnings 222.22 205.05 217.15 192.75 Has business (1/0) 0.09 0.09 0.08 0.09 Business earnings 6.47 7.97 6.14 6.60 Capital income 162.39 10.36 23.24 26.03 Student 0.02 0.03 0.02 0.03 Lives in same municipality 0.33 0.36 0.30 0.31 Years of education 12.74 12.46 12.53 12.48 Woman (1/0) 0.49 0.49 0.49 0.49 *All earnings and income variables are reported in 100 SEK per year (10 SEK = 0.8 US$). **Employment is defined as having a nonzero annual wage income. Open in new tab Table 1 Summary Statistics for the Children and Siblings of Elected Politicians by Electoral Status, 1991–2011 Panel (a): children . Electoral status of parent . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 27.07 27.28 27.50 27.57 All earnings* 149.05 149.66 151.73 150.26 Employed (1/0)** 0.91 0.90 0.91 0.90 Employment earnings 138.06 139.09 141.55 139.20 Business owner (1/0) 0.03 0.03 0.03 0.04 Business income 2.96 2.33 2.29 3.11 Capital income −14.46 −6.99 17.75 −46.93 Student (1/0) 0.31 0.29 0.30 0.28 Lives in same municipality (1/0) 0.54 0.53 0.50 0.51 Years of education 13.31 13.19 13.32 13.08 Woman (1/0) 0.49 0.49 0.47 0.48 Panel (a): children . Electoral status of parent . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 27.07 27.28 27.50 27.57 All earnings* 149.05 149.66 151.73 150.26 Employed (1/0)** 0.91 0.90 0.91 0.90 Employment earnings 138.06 139.09 141.55 139.20 Business owner (1/0) 0.03 0.03 0.03 0.04 Business income 2.96 2.33 2.29 3.11 Capital income −14.46 −6.99 17.75 −46.93 Student (1/0) 0.31 0.29 0.30 0.28 Lives in same municipality (1/0) 0.54 0.53 0.50 0.51 Years of education 13.31 13.19 13.32 13.08 Woman (1/0) 0.49 0.49 0.47 0.48 Panel (b): siblings . Electoral status of sibling . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 49.51 48.17 49.95 48.29 Total earnings 236.83 221.21 230.17 208.96 Employed (1/0) 0.89 0.88 0.87 0.85 Labour earnings 222.22 205.05 217.15 192.75 Has business (1/0) 0.09 0.09 0.08 0.09 Business earnings 6.47 7.97 6.14 6.60 Capital income 162.39 10.36 23.24 26.03 Student 0.02 0.03 0.02 0.03 Lives in same municipality 0.33 0.36 0.30 0.31 Years of education 12.74 12.46 12.53 12.48 Woman (1/0) 0.49 0.49 0.49 0.49 Panel (b): siblings . Electoral status of sibling . In power . In opposition . Top name . Top 2 and 3 . Top name . Top 2 and 3 . Age 49.51 48.17 49.95 48.29 Total earnings 236.83 221.21 230.17 208.96 Employed (1/0) 0.89 0.88 0.87 0.85 Labour earnings 222.22 205.05 217.15 192.75 Has business (1/0) 0.09 0.09 0.08 0.09 Business earnings 6.47 7.97 6.14 6.60 Capital income 162.39 10.36 23.24 26.03 Student 0.02 0.03 0.02 0.03 Lives in same municipality 0.33 0.36 0.30 0.31 Years of education 12.74 12.46 12.53 12.48 Woman (1/0) 0.49 0.49 0.49 0.49 *All earnings and income variables are reported in 100 SEK per year (10 SEK = 0.8 US$). **Employment is defined as having a nonzero annual wage income. Open in new tab In Table 2, we turn our attention to the politicians themselves. The average politician has 1.6 children (above 18 years of age) and 1.5 siblings. Differences in average age and education between majority and opposition politicians are small. The only stark difference concerns gender: women are considerably less likely than men to be top ranked but almost equally likely to be second or third ranked. Table 2 Summary Statistics for Politicians by Electoral Status, 1991–2011 . Majority politicians . Opposition politicians . Top name . Top 2 and 3 . Top name . Top 2 and 3 . No. of children 1.68 1.61 1.66 1.61 No. of children in home municipality 0.90 0.87 0.82 0.82 No. of siblings 1.63 1.50 1.47 1.48 No. of siblings in home municipality 0.55 0.54 0.44 0.46 Age 52.44 51.31 52.52 51.20 Years of education 13.30 13.15 13.26 13.15 Woman (1/0) 0.26 0.48 0.28 0.47 No. of observations 2,979 5,742 3,004 5,847 . Majority politicians . Opposition politicians . Top name . Top 2 and 3 . Top name . Top 2 and 3 . No. of children 1.68 1.61 1.66 1.61 No. of children in home municipality 0.90 0.87 0.82 0.82 No. of siblings 1.63 1.50 1.47 1.48 No. of siblings in home municipality 0.55 0.54 0.44 0.46 Age 52.44 51.31 52.52 51.20 Years of education 13.30 13.15 13.26 13.15 Woman (1/0) 0.26 0.48 0.28 0.47 No. of observations 2,979 5,742 3,004 5,847 Open in new tab Table 2 Summary Statistics for Politicians by Electoral Status, 1991–2011 . Majority politicians . Opposition politicians . Top name . Top 2 and 3 . Top name . Top 2 and 3 . No. of children 1.68 1.61 1.66 1.61 No. of children in home municipality 0.90 0.87 0.82 0.82 No. of siblings 1.63 1.50 1.47 1.48 No. of siblings in home municipality 0.55 0.54 0.44 0.46 Age 52.44 51.31 52.52 51.20 Years of education 13.30 13.15 13.26 13.15 Woman (1/0) 0.26 0.48 0.28 0.47 No. of observations 2,979 5,742 3,004 5,847 . Majority politicians . Opposition politicians . Top name . Top 2 and 3 . Top name . Top 2 and 3 . No. of children 1.68 1.61 1.66 1.61 No. of children in home municipality 0.90 0.87 0.82 0.82 No. of siblings 1.63 1.50 1.47 1.48 No. of siblings in home municipality 0.55 0.54 0.44 0.46 Age 52.44 51.31 52.52 51.20 Years of education 13.30 13.15 13.26 13.15 Woman (1/0) 0.26 0.48 0.28 0.47 No. of observations 2,979 5,742 3,004 5,847 Open in new tab The next set of descriptive statistics in Table 3 verify a key assumption in our empirical analysis by comparing earnings of relatives to top politicians in the party that appointed the mayor with relatives to politicians in the largest opposition‐bloc party. For example, if the Social Democrats (a left‐bloc party) appoint the mayor, we call the centre‐right bloc party with the most seats the ‘opposition party’. For our design to capture dynastic political rents to political power, ruling‐party status should assign top executive posts to top politicians on this party’s ballot. Even if such appointments are a standard procedure in a parliamentary system, they should be validated in data. Table 3 Percentage of Top Ranked and 2nd or 3rd Rank Politicians that Hold Executive Positions, and the Average Yearly Earnings Among these Office Holders . Top politician . 2nd and 3rd rank politicians . Average yearly earnings (1,000s SEK) . Ruling party . Opposition party . Ruling party . Opposition party . Mayor (%) 86 0 4 0 539.2 Council chair (%) 5 2 14 0 346.1 Subcommittee chair (%) 27 1 35 3 429.7 Vice mayor (%) 0 57 10 5 443.9 Council vice chair (%) 0 4 6 12 310.4 Subcommittee vice chair (%) 3 18 10 27 368.3 . Top politician . 2nd and 3rd rank politicians . Average yearly earnings (1,000s SEK) . Ruling party . Opposition party . Ruling party . Opposition party . Mayor (%) 86 0 4 0 539.2 Council chair (%) 5 2 14 0 346.1 Subcommittee chair (%) 27 1 35 3 429.7 Vice mayor (%) 0 57 10 5 443.9 Council vice chair (%) 0 4 6 12 310.4 Subcommittee vice chair (%) 3 18 10 27 368.3 Notes This Table is based on pooled data for appointments in 2007 and 2011, collected by Statistics Sweden in a mandatory survey to municipalities. 1,000 SEK ≈ 120 US$. Open in new tab Table 3 Percentage of Top Ranked and 2nd or 3rd Rank Politicians that Hold Executive Positions, and the Average Yearly Earnings Among these Office Holders . Top politician . 2nd and 3rd rank politicians . Average yearly earnings (1,000s SEK) . Ruling party . Opposition party . Ruling party . Opposition party . Mayor (%) 86 0 4 0 539.2 Council chair (%) 5 2 14 0 346.1 Subcommittee chair (%) 27 1 35 3 429.7 Vice mayor (%) 0 57 10 5 443.9 Council vice chair (%) 0 4 6 12 310.4 Subcommittee vice chair (%) 3 18 10 27 368.3 . Top politician . 2nd and 3rd rank politicians . Average yearly earnings (1,000s SEK) . Ruling party . Opposition party . Ruling party . Opposition party . Mayor (%) 86 0 4 0 539.2 Council chair (%) 5 2 14 0 346.1 Subcommittee chair (%) 27 1 35 3 429.7 Vice mayor (%) 0 57 10 5 443.9 Council vice chair (%) 0 4 6 12 310.4 Subcommittee vice chair (%) 3 18 10 27 368.3 Notes This Table is based on pooled data for appointments in 2007 and 2011, collected by Statistics Sweden in a mandatory survey to municipalities. 1,000 SEK ≈ 120 US$. Open in new tab Table 3 relies on two waves of mandatory surveys to municipalities, following the 2006 and 2010 elections, to record the personal identification number of every individual appointed to an executive position. In our sample, we can thus compute the shares of the top‐three politicians on the electoral ballot of the ruling party and in the opposition party, who were indeed appointed to specific top positions. The Table also lists average earnings by executive position. Reading the Table, one should know that politicians can hold several positions at the same time, even though this is the exception rather than the rule. The summary statistics in Table 3 show several things. First, when a party appoints the mayor, the probability is 86% that the mayor is the individual at the very top of the ballot. Candidates ranked 2 or 3 on the ballot rarely became mayors (4%) but are more commonly appointed subcommittee chair (35%) or council chair (14%). For the largest party in the opposition bloc, the top‐ranked person is appointed vice mayor in 57 cases out of 100. The means of total earnings in the rightmost column, show that political hierarchy is reflected in average earnings; the mayor earns the most followed by the vice mayor. Of the municipality executives, the mayor is commonly the only one to get a full‐time political wage (the mayor’s annual wage ranks in the top two percentiles of the Swedish income distribution; authors’ own calculations). Lower executive posts, such as subcommittee chairs, usually receive piece‐rate compensation for their chair appointments, on top of the meeting compensation given to all committee members. Vice mayors commonly get a political part‐time wage. Most of the time, the lower office holders keep their regular job as a part‐time occupation. 3. Empirical Strategy 3.1. Basic Regression Specification Our empirical analysis compares average earnings among relatives to top‐ranked politicians in the largest majority party to average earnings among relatives to top‐ranked politicians in the largest opposition party. This comparison can be done in a basic regression specification: Yi,t=βPp,m,t+τt+ρm+εi,t,(1) where indexes i and t denote an individual relative i observed in election period t. The outcome variable, Y, will vary across applications; it may, for example, be the income from the relative’s own business. Our parameter of interest, β, captures the difference between relatives of leading politicians in the largest majority bloc party compared to relatives of leading politicians in the largest opposition‐bloc party. The key treatment variable, Pp,m,t, is a binary indicator that takes a value of 1 for relatives of politicians in party p that are appointed mayor (council‐board chair) in municipality m and election period t. The same indicator takes a value of 0 for relatives to leaders of the largest opposition‐bloc party. The specification also includes election‐period fixed effects, τt, and municipality fixed effects ρm. Because the treatment is decided at the level of the municipality and election period, standard errors are clustered at this level in all specifications. 3.2. Handling Systematic Selection by Close Elections A concern when estimating specification (1) is that politicians in the ruling party might differ systematically from those in the largest opposition party, such that their relatives have different earnings, irrespective of whether the politician is the mayor or the opposition leader. For example, Folke and Rickne (2016) find survey evidence that the position as mayor requires a great deal more work than that of the opposition leader. Perhaps then, some politicians might postpone seeking the position of mayor until their children have grown up or left home. Another possibility is that the internal competition for the mayor position in majority parties is greater than the competition for the position of opposition leader in minority parties. After all, the power and prestige of being a mayor, by far, exceed that of being an opposition leader. Fiercer internal competition for mayor could attract politicians whose children or siblings have personal traits that systematically differ from those of opposition leaders and which could also carry‐over to their earnings. This concern applies to situations where politicians can plan ahead to seek the position as mayor or opposition leader. A natural way to address the selection concerns would therefore be to use a regression discontinuity design (RDD) to assign the mayor position exogenously by exploiting narrow elections. As already mentioned in Section 1, Swedish political parties indeed form two stable blocs that compete for power. An electoral victory (50% of the seats or more) for a bloc often allows the largest party in that bloc to appoint the mayor. Previous studies, such as Pettersson‐Lidbom (2008), use left‐wing seat majority as the treatment variable in a sharp RDD. It turns out, however, that this treatment variable does not deterministically assign majority‐party status to the largest party in the left bloc. In about 10–15% of close elections, small parties form coalitions across bloc lines, leaving the largest party in the opposition. To amend this problem of systematic selection, one would need to exploit a fuzzy RDD. Furthermore, not only the threshold for a left‐bloc seat majority but also the threshold for a right‐bloc majority influences who is in power. Each of these two thresholds changes the probability of the largest party in the bloc appointing the mayor by about 40% (Folke and Rickne, 2016). In principle, one could thus implement a fuzzy RDD around the two thresholds. In this particular setting, however, we would be left with too few observations to estimate the treatment effect of political power with reasonable statistical precision (more detailed information and estimation results are available from the authors). 3.3. An Alternative Empirical Design However, we can use an alternative approach, which builds on the same intuition as an RDD but buys more statistical precision at the cost of stronger identifying assumptions. Specifically, we can select elections where the right bloc and the left bloc both receive close to 50% of the seats. We assume that the outcome of these elections is uncertain, such that the key political players in the largest parties of the bloc perceive a comparable chance of winning. By restricting our sample to close elections, we can meet the concerns about systematic selection. In the restricted sample, we can more convincingly argue that our estimates of the treatment effect in (1) indeed uncover a true causal effect. To explain the intuition for this approach in a different way, we can make the analogy with estimating a treatment effect on a matched sample. When the propensity scores for treatment – i.e. for winning the election and appointing the mayor – are comparable for the largest parties in the two blocs, we avoid some or all of the selection problems that would plague a regular OLS estimate. Defining ‘close’ elections is not as straightforward in a PR electoral system with many parties as in a majoritarian system with two parties. Using the realised vote share, as in Pettersson‐Lidbom (2008) comes at the cost of a systematic measurement error, because the seat share is not deterministically decided by the vote share.7 To overcome this measurement problem, we use the measure developed by Folke and Rickne (2016) and Fiva et al. (2016), which relies on simulation. Specifically, we use the simulated vote share each bloc needs to win, or lose, to gain or lose the seat majority in the municipal assembly. The details of this technique are explained in the online Appendix A, while we only give a brief graphical explanation here. Although Sweden is often described as a stable political environment with Social Democratic dominance in the post‐war period, this stereotype is misleading when it comes to the six elections during our sample period, 1991–2011. These elections were very competitive: in two‐thirds of our municipality election observations, either, or both, of the two blocs are within 10 percentage points of the vote share required to either win or lose the seat majority and 44% of our observations are within a margin of 5 percentage points.8 In this subsample with a 5% win margin or less, the ruling bloc changed from left to right, or vice versa, in 44% of all elections. To illustrate, Figure 1 plots binned averages of the dummy variable that the largest party in a bloc appoints the mayor against the simulated value of how close – in terms of vote share – the same bloc was to win (or lose) the seat majority. Clearly, the electoral outcome becomes much more difficult to predict close to the vertical zero line, where the scatters converge. Thus, more competitive elections indeed have more uncertain outcomes. Nevertheless, the event that a bloc receives more than half the votes (and seats) discontinuously shifts up the probability that the largest party in that bloc appoints the mayor. For both blocs, this upward shift is of about the same size, namely 40 percentage points. Fig. 1. Open in new tabDownload slide Probability that the Largest Party in a Political Bloc Appoints the Mayor Against the Simulated Margin to Seat Majority for the Left Bloc (Left Graph) and the Center‐right Bloc (Right Graph) Note. Colour figure can be viewed at wileyonlinelibrary.com Fig. 1. Open in new tabDownload slide Probability that the Largest Party in a Political Bloc Appoints the Mayor Against the Simulated Margin to Seat Majority for the Left Bloc (Left Graph) and the Center‐right Bloc (Right Graph) Note. Colour figure can be viewed at wileyonlinelibrary.com 4. Baseline Results To present our findings, we rely on a sequence of specifications. In particular, we start by a general specification for total earnings of a sample with all relatives. Then we consecutively zoom in on specific samples, types of incomes and treatments where we find clear evidence that political power has a clear positive effect on the earnings of relatives. The first set of estimates appears in Table 4. Table 4 Effects on Earnings of Having a Parent or Sibling in the Top Tier of the Largest Governing Party . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.08*** 0.06* 0.08* 0.06 0.11*** 0.12** 0.16** 0.17* (0.02) (0.04) (0.04) (0.06) (0.04) (0.06) (0.06) (0.10) DV: earnings Treatment 6.16*** 1.41 6.94 4.76 6.19** 6.69 8.92** 12.07 (2.21) (3.71) (4.28) (8.23) (2.63) (4.17) (4.47) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 190.07 193.77 196.23 199.85 162.34 165.13 166.08 164.49 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.07* 0.06 0.05 0.12*** 0.12** 0.14** 0.15 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.10) DV: earnings Treatment 6.02*** 1.32 5.96 3.55 6.49** 6.45 7.82* 9.59 (2.21) (3.71) (4.35) (8.29) (2.61) (4.15) (4.49) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 176.90 180.64 183.77 187.16 149.45 151.68 153.15 150.64 Panel (c): business earnings DV: log earnings Treatment −0.02 −0.01 0.02 0.01 −0.03 −0.04 −0.01 −0.04 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.56 −0.02 0.64 0.65 −0.61 0.28 −0.12 −0.07 (0.68) (0.78) (0.78) (1.19) (0.85) (1.40) (1.25) (1.79) Observations 23,596 10,387 8,254 3,596 10,104 4,516 3,481 1,551 Mean dep var 4.86 4.96 4.62 4.86 5.58 6.06 5.94 7.13 . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.08*** 0.06* 0.08* 0.06 0.11*** 0.12** 0.16** 0.17* (0.02) (0.04) (0.04) (0.06) (0.04) (0.06) (0.06) (0.10) DV: earnings Treatment 6.16*** 1.41 6.94 4.76 6.19** 6.69 8.92** 12.07 (2.21) (3.71) (4.28) (8.23) (2.63) (4.17) (4.47) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 190.07 193.77 196.23 199.85 162.34 165.13 166.08 164.49 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.07* 0.06 0.05 0.12*** 0.12** 0.14** 0.15 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.10) DV: earnings Treatment 6.02*** 1.32 5.96 3.55 6.49** 6.45 7.82* 9.59 (2.21) (3.71) (4.35) (8.29) (2.61) (4.15) (4.49) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 176.90 180.64 183.77 187.16 149.45 151.68 153.15 150.64 Panel (c): business earnings DV: log earnings Treatment −0.02 −0.01 0.02 0.01 −0.03 −0.04 −0.01 −0.04 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.56 −0.02 0.64 0.65 −0.61 0.28 −0.12 −0.07 (0.68) (0.78) (0.78) (1.19) (0.85) (1.40) (1.25) (1.79) Observations 23,596 10,387 8,254 3,596 10,104 4,516 3,481 1,551 Mean dep var 4.86 4.96 4.62 4.86 5.58 6.06 5.94 7.13 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 4 Effects on Earnings of Having a Parent or Sibling in the Top Tier of the Largest Governing Party . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.08*** 0.06* 0.08* 0.06 0.11*** 0.12** 0.16** 0.17* (0.02) (0.04) (0.04) (0.06) (0.04) (0.06) (0.06) (0.10) DV: earnings Treatment 6.16*** 1.41 6.94 4.76 6.19** 6.69 8.92** 12.07 (2.21) (3.71) (4.28) (8.23) (2.63) (4.17) (4.47) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 190.07 193.77 196.23 199.85 162.34 165.13 166.08 164.49 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.07* 0.06 0.05 0.12*** 0.12** 0.14** 0.15 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.10) DV: earnings Treatment 6.02*** 1.32 5.96 3.55 6.49** 6.45 7.82* 9.59 (2.21) (3.71) (4.35) (8.29) (2.61) (4.15) (4.49) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 176.90 180.64 183.77 187.16 149.45 151.68 153.15 150.64 Panel (c): business earnings DV: log earnings Treatment −0.02 −0.01 0.02 0.01 −0.03 −0.04 −0.01 −0.04 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.56 −0.02 0.64 0.65 −0.61 0.28 −0.12 −0.07 (0.68) (0.78) (0.78) (1.19) (0.85) (1.40) (1.25) (1.79) Observations 23,596 10,387 8,254 3,596 10,104 4,516 3,481 1,551 Mean dep var 4.86 4.96 4.62 4.86 5.58 6.06 5.94 7.13 . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.08*** 0.06* 0.08* 0.06 0.11*** 0.12** 0.16** 0.17* (0.02) (0.04) (0.04) (0.06) (0.04) (0.06) (0.06) (0.10) DV: earnings Treatment 6.16*** 1.41 6.94 4.76 6.19** 6.69 8.92** 12.07 (2.21) (3.71) (4.28) (8.23) (2.63) (4.17) (4.47) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 190.07 193.77 196.23 199.85 162.34 165.13 166.08 164.49 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.07* 0.06 0.05 0.12*** 0.12** 0.14** 0.15 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.10) DV: earnings Treatment 6.02*** 1.32 5.96 3.55 6.49** 6.45 7.82* 9.59 (2.21) (3.71) (4.35) (8.29) (2.61) (4.15) (4.49) (7.37) Observations 23,826 10,507 8,315 3,626 10,206 4,568 3,504 1,561 Mean dep var 176.90 180.64 183.77 187.16 149.45 151.68 153.15 150.64 Panel (c): business earnings DV: log earnings Treatment −0.02 −0.01 0.02 0.01 −0.03 −0.04 −0.01 −0.04 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.56 −0.02 0.64 0.65 −0.61 0.28 −0.12 −0.07 (0.68) (0.78) (0.78) (1.19) (0.85) (1.40) (1.25) (1.79) Observations 23,596 10,387 8,254 3,596 10,104 4,516 3,481 1,551 Mean dep var 4.86 4.96 4.62 4.86 5.58 6.06 5.94 7.13 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab These estimates pool siblings and children. The top panel (panel (a)) shows the effects on all earnings. As explained earlier, the treatment for the politician relatives is that their sibling or parent obtains large executive powers. The two lower panels in Table 4 separate out employment earnings (panel (b)) and business earnings (panel (c)). For every earnings type and specification, we use both the log outcome variable (the upper row of estimates) and the corresponding amount in thousands of Swedish Kronor (the lower row). To help interpret the economic significance of the results, we also report mean earnings within the sample in the row underneath the number of observations. In all three panels, the estimation sample is divided into two ways, by the identity of the powerful politician (top ranked, or top‐three on the ballot) and by residence of the relative (anywhere in Sweden, or in the same municipality as the politician). Finally, for every outcome and sample, we report results for the full sample of elections and parties (the columns marked ‘All’) and for elections and parties where the vote share of their bloc was within 5 percentage points from winning or losing the governing majority (the columns marked ‘5%’). The estimates for the full sample of relatives (panel (a)) suggest that dynastic political rents in Swedish local politics are large. The estimate for log earnings indicate that relatives of the top‐ranked politicians in the ruling party earn on average 8% – 6,000 SEK in the level estimates – more per year than the relatives of top‐ranked opposition politicians. As we narrow the sample to close uncertain elections, however, the estimates are considerably noisier and/or smaller in size. Turning to the relatives who live in the same municipality as their related politician, most estimates become larger and more precisely estimated in the narrow sample. On balance, the estimates now suggest large economic gains for close relatives of politicians with powerful political positions. For those related to the top‐ranked politician, the log estimates indicate 16% – and the level estimates, 8–10,000 SEK – higher earnings. While these estimated economic payoffs are quite large in relative terms, they are quite small in absolute terms. Although 8,000 SEK corresponds to an estimated increase of 15%, it amounts to only 3% of the median income of full‐time workers in Sweden in the year 2000 (234,000 SEK). This indicates that the economic gains are mainly distributed to relatives at the lower end of the income distribution. Examining separately the two different types of earnings, from employment and business ownership (panels (b) and (c)), we see clearly that earnings from employment rather than business ownership drive our results – if anything, the estimates for business income appear to be negative. Thus, economic gains are unlikely to stem from corporate contracts or other benefits that could befall business‐owning relatives. This is perhaps not surprising if the gains indeed accrue mainly to those at the bottom of the income distribution. 5. Robustness and Mechanisms 5.1. Robustness Tests To make sure that our results capture a causal effect of political power on the earnings of relatives, we re‐estimate all of the regressions from Table 4 replacing the outcome with the lagged‐dependent variable. Thus, we take the average of the outcome variable in the previous election period, excluding the election year. If there are no pre‐electoral differences in the outcome variable, as required by our identifying assumption, we should find no effect of our treatment on previous earnings once we focus on close elections. The results of this analysis appear in Table 5, which is structured in exactly the same way as in Table 4. Table 5 Effects on Lagged Earnings of Having a Parent or Sibling in the Top Tier of the Largest Governing Party . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: earnings Treatment 0.06*** 0.03 0.10** 0.10 0.06* −0.01 0.10* 0.06 (0.02) (0.03) (0.04) (0.06) (0.03) (0.05) (0.06) (0.09) DV: total earnings Treatment 3.31* −4.26 3.85 −1.15 1.24 −3.21 2.58 0.28 (1.96) (3.05) (3.73) (6.26) (2.23) (3.53) (4.14) (7.09) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 181.35 185.06 185.51 190.31 153.56 158.73 156.23 158.03 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.05 0.13*** 0.10 0.12*** 0.03 0.18*** 0.07 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.09) DV: earnings Treatment 3.86* −3.71 4.30 −2.72 3.47 −1.17 4.56 −1.70 (2.03) (3.16) (3.84) (6.36) (2.32) (3.67) (4.20) (6.92) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 168.55 172.04 173.06 178.28 140.94 145.45 142.88 144.35 Panel (c): business earnings DV: log earnings Treatment −0.03** −0.01 −0.02 −0.01 −0.08*** −0.06 −0.09** −0.05 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.04 0.23 −0.30 1.42 −1.82** −0.14 −2.00 2.24 (0.62) (0.78) (0.91) (1.57) (0.87) (1.46) (1.73) (3.23) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 5.27 5.40 5.42 4.98 5.85 6.45 6.84 7.24 . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: earnings Treatment 0.06*** 0.03 0.10** 0.10 0.06* −0.01 0.10* 0.06 (0.02) (0.03) (0.04) (0.06) (0.03) (0.05) (0.06) (0.09) DV: total earnings Treatment 3.31* −4.26 3.85 −1.15 1.24 −3.21 2.58 0.28 (1.96) (3.05) (3.73) (6.26) (2.23) (3.53) (4.14) (7.09) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 181.35 185.06 185.51 190.31 153.56 158.73 156.23 158.03 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.05 0.13*** 0.10 0.12*** 0.03 0.18*** 0.07 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.09) DV: earnings Treatment 3.86* −3.71 4.30 −2.72 3.47 −1.17 4.56 −1.70 (2.03) (3.16) (3.84) (6.36) (2.32) (3.67) (4.20) (6.92) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 168.55 172.04 173.06 178.28 140.94 145.45 142.88 144.35 Panel (c): business earnings DV: log earnings Treatment −0.03** −0.01 −0.02 −0.01 −0.08*** −0.06 −0.09** −0.05 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.04 0.23 −0.30 1.42 −1.82** −0.14 −2.00 2.24 (0.62) (0.78) (0.91) (1.57) (0.87) (1.46) (1.73) (3.23) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 5.27 5.40 5.42 4.98 5.85 6.45 6.84 7.24 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 5 Effects on Lagged Earnings of Having a Parent or Sibling in the Top Tier of the Largest Governing Party . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: earnings Treatment 0.06*** 0.03 0.10** 0.10 0.06* −0.01 0.10* 0.06 (0.02) (0.03) (0.04) (0.06) (0.03) (0.05) (0.06) (0.09) DV: total earnings Treatment 3.31* −4.26 3.85 −1.15 1.24 −3.21 2.58 0.28 (1.96) (3.05) (3.73) (6.26) (2.23) (3.53) (4.14) (7.09) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 181.35 185.06 185.51 190.31 153.56 158.73 156.23 158.03 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.05 0.13*** 0.10 0.12*** 0.03 0.18*** 0.07 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.09) DV: earnings Treatment 3.86* −3.71 4.30 −2.72 3.47 −1.17 4.56 −1.70 (2.03) (3.16) (3.84) (6.36) (2.32) (3.67) (4.20) (6.92) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 168.55 172.04 173.06 178.28 140.94 145.45 142.88 144.35 Panel (c): business earnings DV: log earnings Treatment −0.03** −0.01 −0.02 −0.01 −0.08*** −0.06 −0.09** −0.05 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.04 0.23 −0.30 1.42 −1.82** −0.14 −2.00 2.24 (0.62) (0.78) (0.91) (1.57) (0.87) (1.46) (1.73) (3.23) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 5.27 5.40 5.42 4.98 5.85 6.45 6.84 7.24 . Living in any municipality . Living in municipality of politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . Panel (a): total earnings DV: earnings Treatment 0.06*** 0.03 0.10** 0.10 0.06* −0.01 0.10* 0.06 (0.02) (0.03) (0.04) (0.06) (0.03) (0.05) (0.06) (0.09) DV: total earnings Treatment 3.31* −4.26 3.85 −1.15 1.24 −3.21 2.58 0.28 (1.96) (3.05) (3.73) (6.26) (2.23) (3.53) (4.14) (7.09) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 181.35 185.06 185.51 190.31 153.56 158.73 156.23 158.03 Panel (b): employment earnings DV: log earnings Treatment 0.09*** 0.05 0.13*** 0.10 0.12*** 0.03 0.18*** 0.07 (0.03) (0.04) (0.04) (0.07) (0.04) (0.06) (0.07) (0.09) DV: earnings Treatment 3.86* −3.71 4.30 −2.72 3.47 −1.17 4.56 −1.70 (2.03) (3.16) (3.84) (6.36) (2.32) (3.67) (4.20) (6.92) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 168.55 172.04 173.06 178.28 140.94 145.45 142.88 144.35 Panel (c): business earnings DV: log earnings Treatment −0.03** −0.01 −0.02 −0.01 −0.08*** −0.06 −0.09** −0.05 (0.01) (0.02) (0.02) (0.03) (0.02) (0.04) (0.04) (0.06) DV: earnings Treatment 0.04 0.23 −0.30 1.42 −1.82** −0.14 −2.00 2.24 (0.62) (0.78) (0.91) (1.57) (0.87) (1.46) (1.73) (3.23) Observations 19,694 8,716 6,998 3,062 8,418 3,751 2,940 1,298 Mean DV 5.27 5.40 5.42 4.98 5.85 6.45 6.84 7.24 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab The results do not show any significant differences in lagged outcome variables for the specifications in the close election samples. In the online Appendix (Table B1), we further test for balance between the treatment and control groups by running the same specifications with predetermined characteristics (age and years of education) as dependent variables. The differences in age and years of education are small in magnitude and far from statistically significant, which provide further evidence that characteristics are indeed balanced across relatives of top‐ranked politicians in the ruling and opposition parties. In the sample of close uncertain elections, we find no statistically significant differences in any predetermined characteristic for any of the estimation samples. In fact, the differences are quite small even in the full sample. In a third robustness check, we check whether our result hold up when a number of control variables are added to the regression. Importantly, the controls include a binary indicator for whether a party was the ruling party in the previous election. Robustness to this control is central for our identifying assumption. If close elections are indeed uncertain, our effects should not be driven by the party that appointed the mayor after the previous election. The other controls are particularly relevant to the parent‐child estimation, which follow below. Parents may choose to seek more demanding political offices once their children have reached a certain age or obtained a higher level of education. We therefore add binary indicators for age and education categories of the close relative in the appropriate regressions.9 The results (online Appendix, Table B2) show that once we limit ourselves to close, uncertain elections, the estimates are less sensitive to the controls. As in the case of Table 5, the estimates illustrate the importance of limiting the estimates to close elections. The full‐sample results are more sensitive to the controls, which is exactly what we would expect if politicians are able to select into top positions by planning ahead in predictable electoral environments. Altogether, these robustness tests strengthen our argument that the baseline estimates capture a treatment effect of political power, rather than a systematic difference in predetermined characteristics between relatives of politicians in and out of power. We also run the same battery of tests for each subsequent Table of results (Tables 6-8) but relegate all the robustness Tables to the online Appendix so as to save space. Table 6 Effects on Total Earnings of Having Either a Parent or a Sibling in the Top Tier of the Largest Governing Party . Children who live in the same municipality as the politician . Siblings who live in the same municipality as the politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . DV: log earnings Treatment 0.09* 0.06 0.22*** 0.22* 0.09 0.18* 0.01 0.02 (0.05) (0.07) (0.08) (0.13) (0.06) (0.09) (0.11) (0.14) DV: earnings Treatment 3.34 1.05 16.25*** 18.80** 6.99 10.43 −1.81 0.41 (3.07) (5.11) (5.12) (8.31) (4.64) (7.65) (8.27) (13.25) Observations 6,412 2,876 2,221 991 3,794 1,692 1,283 570 Mean dep var 135.94 137.15 135.13 133.06 206.97 212.69 219.64 219.13 . Children who live in the same municipality as the politician . Siblings who live in the same municipality as the politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . DV: log earnings Treatment 0.09* 0.06 0.22*** 0.22* 0.09 0.18* 0.01 0.02 (0.05) (0.07) (0.08) (0.13) (0.06) (0.09) (0.11) (0.14) DV: earnings Treatment 3.34 1.05 16.25*** 18.80** 6.99 10.43 −1.81 0.41 (3.07) (5.11) (5.12) (8.31) (4.64) (7.65) (8.27) (13.25) Observations 6,412 2,876 2,221 991 3,794 1,692 1,283 570 Mean dep var 135.94 137.15 135.13 133.06 206.97 212.69 219.64 219.13 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 6 Effects on Total Earnings of Having Either a Parent or a Sibling in the Top Tier of the Largest Governing Party . Children who live in the same municipality as the politician . Siblings who live in the same municipality as the politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . DV: log earnings Treatment 0.09* 0.06 0.22*** 0.22* 0.09 0.18* 0.01 0.02 (0.05) (0.07) (0.08) (0.13) (0.06) (0.09) (0.11) (0.14) DV: earnings Treatment 3.34 1.05 16.25*** 18.80** 6.99 10.43 −1.81 0.41 (3.07) (5.11) (5.12) (8.31) (4.64) (7.65) (8.27) (13.25) Observations 6,412 2,876 2,221 991 3,794 1,692 1,283 570 Mean dep var 135.94 137.15 135.13 133.06 206.97 212.69 219.64 219.13 . Children who live in the same municipality as the politician . Siblings who live in the same municipality as the politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . Relatives of top‐three politicians . Relatives of top‐ranked politician . All . 5% . All . 5% . All . 5% . All . 5% . DV: log earnings Treatment 0.09* 0.06 0.22*** 0.22* 0.09 0.18* 0.01 0.02 (0.05) (0.07) (0.08) (0.13) (0.06) (0.09) (0.11) (0.14) DV: earnings Treatment 3.34 1.05 16.25*** 18.80** 6.99 10.43 −1.81 0.41 (3.07) (5.11) (5.12) (8.31) (4.64) (7.65) (8.27) (13.25) Observations 6,412 2,876 2,221 991 3,794 1,692 1,283 570 Mean dep var 135.94 137.15 135.13 133.06 206.97 212.69 219.64 219.13 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 7 Effects on All Earnings from Having a Parent Who is Mayor, by Parent’s Incumbency Status . Children . Siblings . Relatives of first‐time chairs . Relatives of incumbent chairs . Relatives of first‐time chairs . Relatives of incumbent chairs . All . 5% . All . 5% . All . 5% . All . 5% . Log earnings Treatment 0.26*** 0.40** 0.14 0.09 0.13 0.23 −0.02 0.02 (0.10) (0.17) (0.19) (0.20) (0.13) (0.21) (0.18) (0.19) Earnings Treatment 15.74** 22.23** 6.78 7.70 7.00 19.16 −15.91 −21.24 (6.13) (10.07) (16.30) (22.19) (9.73) (18.50) (14.46) (22.74) Observations 1,617 698 604 293 924 399 359 171 Mean dep var 128.70 124.75 152.36 152.87 216.43 217.47 227.92 223.01 . Children . Siblings . Relatives of first‐time chairs . Relatives of incumbent chairs . Relatives of first‐time chairs . Relatives of incumbent chairs . All . 5% . All . 5% . All . 5% . All . 5% . Log earnings Treatment 0.26*** 0.40** 0.14 0.09 0.13 0.23 −0.02 0.02 (0.10) (0.17) (0.19) (0.20) (0.13) (0.21) (0.18) (0.19) Earnings Treatment 15.74** 22.23** 6.78 7.70 7.00 19.16 −15.91 −21.24 (6.13) (10.07) (16.30) (22.19) (9.73) (18.50) (14.46) (22.74) Observations 1,617 698 604 293 924 399 359 171 Mean dep var 128.70 124.75 152.36 152.87 216.43 217.47 227.92 223.01 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 7 Effects on All Earnings from Having a Parent Who is Mayor, by Parent’s Incumbency Status . Children . Siblings . Relatives of first‐time chairs . Relatives of incumbent chairs . Relatives of first‐time chairs . Relatives of incumbent chairs . All . 5% . All . 5% . All . 5% . All . 5% . Log earnings Treatment 0.26*** 0.40** 0.14 0.09 0.13 0.23 −0.02 0.02 (0.10) (0.17) (0.19) (0.20) (0.13) (0.21) (0.18) (0.19) Earnings Treatment 15.74** 22.23** 6.78 7.70 7.00 19.16 −15.91 −21.24 (6.13) (10.07) (16.30) (22.19) (9.73) (18.50) (14.46) (22.74) Observations 1,617 698 604 293 924 399 359 171 Mean dep var 128.70 124.75 152.36 152.87 216.43 217.47 227.92 223.01 . Children . Siblings . Relatives of first‐time chairs . Relatives of incumbent chairs . Relatives of first‐time chairs . Relatives of incumbent chairs . All . 5% . All . 5% . All . 5% . All . 5% . Log earnings Treatment 0.26*** 0.40** 0.14 0.09 0.13 0.23 −0.02 0.02 (0.10) (0.17) (0.19) (0.20) (0.13) (0.21) (0.18) (0.19) Earnings Treatment 15.74** 22.23** 6.78 7.70 7.00 19.16 −15.91 −21.24 (6.13) (10.07) (16.30) (22.19) (9.73) (18.50) (14.46) (22.74) Observations 1,617 698 604 293 924 399 359 171 Mean dep var 128.70 124.75 152.36 152.87 216.43 217.47 227.92 223.01 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections, and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 8 Effects on Occupational Status, Tertiary Studies and Geographical Residence from Having a Parent or Sibling in the Top Tier of the Largest Governing Party . Children of top‐three politicians . Children of top‐ranked politicians . All . 5% . All . 5% . Panel (a): municipal public sector employee Treatment 0.01 0.00 0.00 −0.04 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.26 0.24 0.26 0.26 Panel (b): employed in parent’s pre‐election employment sector Treatment −0.01 −0.01 −0.03** −0.03 (0.01) (0.01) (0.02) (0.02) Observations 6,115 2,739 2,139 957 Mean 0.10 0.11 0.12 0.14 Panel (c): being a student in tertiary education Treatment −0.00 0.01 −0.04* −0.03 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.29 0.30 0.30 0.31 Panel (d): living in the municipality of the parent politicians Treatment 0.03*** 0.01 0.04*** 0.02 (0.01) (0.01) (0.01) (0.02) Observations 10,207 4,443 3,546 1,561 Mean 0.52 0.52 0.52 0.52 . Children of top‐three politicians . Children of top‐ranked politicians . All . 5% . All . 5% . Panel (a): municipal public sector employee Treatment 0.01 0.00 0.00 −0.04 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.26 0.24 0.26 0.26 Panel (b): employed in parent’s pre‐election employment sector Treatment −0.01 −0.01 −0.03** −0.03 (0.01) (0.01) (0.02) (0.02) Observations 6,115 2,739 2,139 957 Mean 0.10 0.11 0.12 0.14 Panel (c): being a student in tertiary education Treatment −0.00 0.01 −0.04* −0.03 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.29 0.30 0.30 0.31 Panel (d): living in the municipality of the parent politicians Treatment 0.03*** 0.01 0.04*** 0.02 (0.01) (0.01) (0.01) (0.02) Observations 10,207 4,443 3,546 1,561 Mean 0.52 0.52 0.52 0.52 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 8 Effects on Occupational Status, Tertiary Studies and Geographical Residence from Having a Parent or Sibling in the Top Tier of the Largest Governing Party . Children of top‐three politicians . Children of top‐ranked politicians . All . 5% . All . 5% . Panel (a): municipal public sector employee Treatment 0.01 0.00 0.00 −0.04 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.26 0.24 0.26 0.26 Panel (b): employed in parent’s pre‐election employment sector Treatment −0.01 −0.01 −0.03** −0.03 (0.01) (0.01) (0.02) (0.02) Observations 6,115 2,739 2,139 957 Mean 0.10 0.11 0.12 0.14 Panel (c): being a student in tertiary education Treatment −0.00 0.01 −0.04* −0.03 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.29 0.30 0.30 0.31 Panel (d): living in the municipality of the parent politicians Treatment 0.03*** 0.01 0.04*** 0.02 (0.01) (0.01) (0.01) (0.02) Observations 10,207 4,443 3,546 1,561 Mean 0.52 0.52 0.52 0.52 . Children of top‐three politicians . Children of top‐ranked politicians . All . 5% . All . 5% . Panel (a): municipal public sector employee Treatment 0.01 0.00 0.00 −0.04 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.26 0.24 0.26 0.26 Panel (b): employed in parent’s pre‐election employment sector Treatment −0.01 −0.01 −0.03** −0.03 (0.01) (0.01) (0.02) (0.02) Observations 6,115 2,739 2,139 957 Mean 0.10 0.11 0.12 0.14 Panel (c): being a student in tertiary education Treatment −0.00 0.01 −0.04* −0.03 (0.01) (0.02) (0.02) (0.03) Observations 6,412 2,876 2,221 991 Mean 0.29 0.30 0.30 0.31 Panel (d): living in the municipality of the parent politicians Treatment 0.03*** 0.01 0.04*** 0.02 (0.01) (0.01) (0.01) (0.02) Observations 10,207 4,443 3,546 1,561 Mean 0.52 0.52 0.52 0.52 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab 5.2. Siblings Versus Children Next, we split the sample by kinship relations. The results are shown in Table 6, but – given the results in Table 4 – only for relatives who live in the same municipality as the powerful politician. We also restrict the analysis to the measure of all earnings. The results for the children are presented in the left part of the Table and those for siblings in the right part. The results for children convey a clear message. Children of top‐ranked politicians obtain a higher income when their parent’s party wins power in the election. On average, their earnings go up by about 20% (16,000 SEK in the level specification), relative to the earnings by children of opposition leaders – a result that holds up in the full sample as well as the sample of close, uncertain elections. To the extent that the child is still part of the leading politician’s household, this estimate may be a lower bound on the pure earnings effect – if we treat the household as one unit, we would expect the labour supply of the child to fall as the winning politicians income goes up, due to an income effect as well as a substitution effect.10 The estimated treatment effect corresponds to about a 10–15% of the average children earnings. The estimates are less consistent when we include children of politicians in the top three of the political hierarchy. We also find no clear effect for the politicians’ siblings, although the relatively large standard errors on the coefficients mean that the two sets of estimates – for children and siblings – are not significantly different from each other. For relatives of top three politicians, the precision of our estimates is thus low enough that we cannot pin‐point an effect neither for children nor for siblings. The robustness tests for Table 6 are contained in the online Appendix (Tables B3 and B4). They show that there is balance on predetermined characteristics and lagged outcome of children and siblings, and that the results are robust to the addition of control variables (the same ones as those listed above under subsection 5.1). 5.3. Winning Versus Losing the Position as Mayor Our baseline analysis jointly estimates the combination of two treatment effects: positive impacts on earnings by children of electoral winners and negative impacts on earnings by children of electoral losers. Some of the politicians in our sample have previously been mayors while others have not, which means that some of the politicians in the ‘opposition’ category actually lose the top executive post. In other words, they fail to get reselected because their party falls out of political power. Because we are interested in estimating the payoffs from a relative who wins office, we follow the approach suggested by Fafchamps and Labonne (2014) to separate out this particular effect. We thus divide the relatives of top‐ranked politicians into two subsamples, relatives of politicians who were never mayors (top ranked in the largest party in the governing majority in any election period since 1982) and those who previously held this post. In the first subsample, we estimate the effect of the relative becoming the mayor instead of becoming, or remaining, the opposition leader.11 In the second sample, we estimate the effect of the relative remaining the mayor compared to losing the position. Table 7 presents these results. They show that children of first‐time mayors do benefit from their parents gaining political power, whereas children of incumbent mayors do not appear to get anything close to the same benefits.12 Thus, the baseline findings seem to be driven by higher earnings when the parent becomes mayor for the first time, rather than remains at that post. The baseline results are also unlikely to stem from negative impacts on the earnings of children of politicians who lose the chair position. For siblings, the point estimates have opposite signs for the two groups but none of the estimates are close to being statistically significant. In the online Appendix (Tables B5 and B6), our battery of robustness tests for lagged outcomes, balance of socio‐economic traits and sensitivity to control variables, corroborate the statistical soundness of the findings in Table 7. 5.4. Mechanisms To explore further which mechanisms drive the higher earnings of children to politicians in power, we consider four additional outcomes. Children of politicians may make more money because they get government jobs through illegitimate channels, or get opportunistically hired by local government actors. It is also possible that children of ruling politicians have better information about available jobs than children of opposition politicians. To test these ideas, we use a binary indicator for having a job in the public sector in the municipality where one’s parent is a politician. A second reason for higher earnings could be that the child steps in to fill a vacancy left by the parent when s/he becomes a full‐time politician, for example, at a family firm or a family farm. Our best proxy for this phenomenon is a binary indicator that takes a value of 1 when the child has the same 5‐digit sector of employment as the one held by the parent in the year before becoming a full‐time politician. If the child steps in to fill a vacancy left by the politician, they should be more likely to work in the same sector as the one their parent used to work in. Children’s benefits also may arise from changes in their own behaviour. One such behavioural change is captured by a dummy for being a student: the child may choose to work rather than study depending on a higher expected relative return from working.13 For these three prospective sources of higher earnings, our sample includes only children living in the same municipality as the politician‐parent. Each of these outcome variables is measured as an average of the relevant binary indicator in the three years following their parent’s promotion or non‐promotion. Another behavioural change concerns location itself. We capture this by a dummy for the child living in the same municipality as their top‐politician parent: the higher political power of the parent may expand the labour market opportunities in that municipality. To evaluate the importance of this variable, we have to include all children in the sample. Their location is measured three years after the election. The estimates are presented in Table 8, with the results for being a municipal employee in panel (a), the results for being employed in the parent’s previous employment sector in panel (b), the results for taking part in tertiary education in panel (c) and the results for children co‐locating with their politician‐parent in panel (d).14 Tests of the identifying assumption, using lagged outcome variables, and sensitivity tests adding controls, are found in the online Appendix (Table B8). The results in Table 8 reject the hypothesis that children of mayors obtain rents because they are more likely than children of opposition leaders to get a job in the municipality’s public sector. Nor do we find that they are more likely to work in their parent’s previous employment sector. The negative estimates for this outcome variable suggest that, if anything, children of powerful politicians are less likely to work in the same sector as their parent did before entering political office. The point estimates in panels (c) and (d) suggest that the baseline findings of higher earnings for children may possibly reflect changes in their own behaviour. Children of mayors are 3–4 percentage points less likely to be students than children of opposition leaders, although the estimates are not significant in the close‐election sample. They are also 2–4 percentage points more likely to live in the same municipality as their parent than are children of opposition leaders – but again, this result is not significant in the close‐election sample. The robustness tests for Table 8 in the online Appendix (Tables B8 and B9) show that the lagged outcomes of the children are balanced in the sample of close elections. However, the results for being a student are sensitive to including the full set of control variables (listed above under subsection 5.1), which reinforces our qualification that these results are only suggestive of a possible mechanism. Although the results are inconclusive regarding the exact channel behind the higher earnings for children of mayors, they provide clear evidence against a few possible channels. It is thus unlikely that higher earnings reflect rents due to a public‐sector job. It is also unlikely that they reflect the children taking over their parents’ pre‐mayoral job. 5.5. Medium‐term Outcomes To extend our analysis of the earnings effects of a close relative in political office, we consider a longer time horizon than in the baseline analysis. Instead of examining the average of the outcome variable(s) in the three years following the relative’s election, we now consider average outcomes two election periods later, i.e. 9–11 years after the initial shift in power. One caveat of this method is that it greatly reduces our sample size. Due to end‐of sample time truncation, we can only use power shifts in the 1991–2002 elections. Another caveat is that closely losing politicians in our baseline specifications can gain office in a later period. Therefore, part of the control group is ‘treated’ with a later electoral wins: this applies to about 20% of the close losers in our baseline estimation. We examine two outcomes, total earnings and years of education. Our interest in years of education mirrors the earlier results on mechanisms. If the children of winning politicians opt to work in their home municipality for a few years, they may simply delay their university education to the proximate future. If so, we would not see any treatment effect on the amount of education cumulated two electoral periods later. Table 9 presents the estimates, for medium‐term earnings in the upper panel and education in the lower panel. All estimates are positive and for the children of top‐ranked politicians they are large. The estimates suggest that children of mayors still make about 15% more than the children of opposition leaders 9–11 years out from the original appointment, even though a fifth of the original opposition leaders become mayors in later periods. However, the estimates are imprecise, which means that they should only be interpreted as suggestive. In the online Appendix (Table B10), we show that this result, if anything, is strengthened by adding the control variables. Table 9 Effects on Long‐term Earnings and Years of Education from Having a Parent or Sibling in Political Power . Children of the top‐three politicians . Children of the top‐ranked politicians . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.10** 0.06 0.13 0.18 (0.05) (0.07) (0.08) (0.15) DV: earnings Treatment 10.46** 6.20 12.03 27.07* (4.54) (8.13) (8.02) (15.88) Observations 4,269 1,850 1,479 638 Mean dep var 226.34 236.09 230.92 237.71 Panel (b): years of education Treatment 0.14 0.13 0.02 −0.18 (0.09) (0.14) (0.16) (0.26) Observations 4,269 1,850 1,479 638 Mean dep var 13.53 13.57 13.67 13.64 . Children of the top‐three politicians . Children of the top‐ranked politicians . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.10** 0.06 0.13 0.18 (0.05) (0.07) (0.08) (0.15) DV: earnings Treatment 10.46** 6.20 12.03 27.07* (4.54) (8.13) (8.02) (15.88) Observations 4,269 1,850 1,479 638 Mean dep var 226.34 236.09 230.92 237.71 Panel (b): years of education Treatment 0.14 0.13 0.02 −0.18 (0.09) (0.14) (0.16) (0.26) Observations 4,269 1,850 1,479 638 Mean dep var 13.53 13.57 13.67 13.64 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab Table 9 Effects on Long‐term Earnings and Years of Education from Having a Parent or Sibling in Political Power . Children of the top‐three politicians . Children of the top‐ranked politicians . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.10** 0.06 0.13 0.18 (0.05) (0.07) (0.08) (0.15) DV: earnings Treatment 10.46** 6.20 12.03 27.07* (4.54) (8.13) (8.02) (15.88) Observations 4,269 1,850 1,479 638 Mean dep var 226.34 236.09 230.92 237.71 Panel (b): years of education Treatment 0.14 0.13 0.02 −0.18 (0.09) (0.14) (0.16) (0.26) Observations 4,269 1,850 1,479 638 Mean dep var 13.53 13.57 13.67 13.64 . Children of the top‐three politicians . Children of the top‐ranked politicians . All . 5% . All . 5% . Panel (a): total earnings DV: log earnings Treatment 0.10** 0.06 0.13 0.18 (0.05) (0.07) (0.08) (0.15) DV: earnings Treatment 10.46** 6.20 12.03 27.07* (4.54) (8.13) (8.02) (15.88) Observations 4,269 1,850 1,479 638 Mean dep var 226.34 236.09 230.92 237.71 Panel (b): years of education Treatment 0.14 0.13 0.02 −0.18 (0.09) (0.14) (0.16) (0.26) Observations 4,269 1,850 1,479 638 Mean dep var 13.53 13.57 13.67 13.64 Notes Robust standard errors clustered at the level of municipality and election period in parentheses. *Significant at 10%; **significant at 5%; ***significant at 1%. In the column titles, ‘all’ refers to the sample of all elections and ‘5%’ refers to the sample of elections where the win margin of the winning political bloc was 5% or less. Open in new tab The results for years of education do not support a permanent large difference between children of mayors and opposition leaders. The sign of the estimates depend on the specification and they are quantitatively small (the largest estimates suggests a difference of only 0.18 years). Thus, the estimates suggest that the short‐term impact on the probability of studying entails a postponed university education rather than an abandoned one. If this interpretation is correct, then the (weak) finding that children of mayors have higher medium‐term earnings could reflect an effect via labour‐market experience. 6. Conclusions We analyse dynastic political rents in a developed and stable democracy. Drawing on uniquely comprehensive register data for siblings and children of politicians, our baseline findings uncover a positive and large effect on the earnings of children from their parent’s ascension to the top position of local political power. We do not find any significant impact on earnings for the siblings of top political executives (although the noisy estimates for the latter renders the difference between children and siblings non‐significant). The hike in dynastic political rents for children is quite robust, passing test for balance in predetermined characteristics of parents that win and lose political power, and tests for addition of a number of control variables. Our extended analysis uncovers suggestive results on how children of mayors behave relative to children of opposition leaders. These results suggest an interpretation of our baseline findings. Earnings only go up for the children who reside in the same municipality as the parent ascending to power. This may appear intuitive, in that the clout of municipal politicians is regionally concentrated to their administrative unit. But when we use the residence of children as an outcome variable, we find that children sort geographically according to the electoral success of their parents. As indicated by the positive treatment effect on co‐residence and the negative treatment effect on being a student, electoral victories appear to encourage children to remain at home for work rather than to leave the municipality for a university education. As indicated by the null effect on employment of children in the municipality’s public sector, remaining in the municipality does not reflect better work opportunities in the sector under political control. Thus it is private employers who are more likely to hire the mayor’s son or daughter – possibly at better conditions – consistent with the findings by Gagliarducci and Manacorda (2015). The attraction of living in the municipality of an electorally successful parent could be viewed as a personal gain to parents from holding a top office (assuming they enjoy the company of their children). By revealed preference, location choices by children also indicate a benefit on their own behalf, quite possibly drawing on a higher status in the local community due to their parent’s hold on political power. However, these rents do not reach very far, as they do not extend to children of politicians who belong to the party elite but do not occupy the unique position as mayor. All in all, our study suggests that dynastic political rents are, at most, a marginal phenomenon in Sweden. This contrasts sharply to the findings for Italy by Gagliarducci and Manacorda (2015), who argue that jobs to the politician relatives make up 0.65% of private‐sector employment. Our results also contrast with the findings for the Phillipines by Fafchamps and Labonne (2014), who argue that politicians staff local bureaucracies with their own relatives. Footnotes 1 " For example, investigative journalists in the programme Uppdrag Granskning, Kommungranskarna (Municipal Reviewers) on Swedish State Television have uncovered public procurement contracts given to a small business owned by a brother‐in‐law of the local mayor, bypassing the rules of offering the contract in an open bidding process. The mayor argued that this large contract was not nepotistic, because the husband‐in‐law had had many previous contracts and handled them well (available on YouTube: https://www.youtube.com/watch?v=FmlRUQsEQRw). 2 " In 1998, Sweden also introduced the option of casting one preference vote per voter, for any candidate on their selected party ballot. This system has not affected party delegations in the municipal councils much because two‐thirds of the voters do not exercise their option of preference voting, most of those who do cast their vote for candidates at the top of party ballots and there is a high threshold for winning a seat by preference votes alone. 3 " There was also a short spree of representation for the populist centre‐right party New Democracy in the early 1990s, and a very recent rise in seats (in particular at the national level) for an anti‐immigration party the Sweden Democrats. In this article, we categorise both these parties as local. 4 " The results are not sensitive to adding smaller amounts. 5 " Indeed, running our analysis for politician spouses gives a negative estimate on the ‘political rents’ of a spouse winning political power. This is consistent with high‐earning spouses (mainly husbands of female politicians) being more likely to drop out of the sample due to a divorce, making the lower‐earning spouses (mainly wives of male politicians) over‐represented in the sample of politicians in power compared to politicians out of power. Folke and Rickne (2016), find no cross‐price effects on labour supply in the form of a reduction in market labour either among wives or among husbands of promoted politicians. 6 " See e.g. Persson and Tabellini (2000, ch. 4) on such electoral cycles in rents. 7 " In Pettersson‐Lidbom (2008), the left bloc is 20 percentage points more likely to win a close election (using the 2% vote share margin measure used by Petterson‐Lidbom to define close elections) when they held the seat majority in the previous election period than when they held a seat minority. We want to avoid such systematic sorting around the threshold. 8 " At the national level, the centre‐right coalition has held power in three of the eight most recent elections and in five of the ten most recent. 9 " We control for age for a dummy for each 10 year interval of age, i.e., under 20, 20–29, 30–39 etc. For education, we include dummies for seven different level of education. 10 " Cf. the discussion in Section 3 about why we do not include spouses among relatives. 11 " There is no difference in partisan make‐up of the two samples. In close elections, the left bloc accounted for 52.0% of the sample among the never chairs and 53.6% of the sample among incumbent chairs. In the full sample, the left bloc accounted for 52.9% of the sample among never chairs and 52.6% of the sample among incumbent chairs. Thus, any difference in estimated effects across the samples cannot be attributed to differences in partisanship. 12 " The careful reader may wonder how this finding squares with our findings in the online Appendix that our results are not sensitive to the mayor’s party being the incumbent. The answer is that those results merely control for incumbency status, whereas the present results look for a heterogeneous effect – stated in another way, the difference is between an interaction term and a control variable. 13 " Sweden’s system of higher education has a universal qualifying examination with the vast majority of schools using only high school grades to guide admissions. It is thus implausible, if not impossible, that an influential local politician would be able to lower their children’s entry threshold to an institute of higher learning. 14 " For completeness, the online Appendix (Table B7) shows the corresponding Table for siblings. We do not delve deeper into these results, though, because of the lack of a baseline effect on earnings in the sibling sample. References Amore , M. , Bennedsen , M. and Nielsen , K.M. ( 2015 ). ‘ Return to political power ’, mimeo, INSEAD . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Braganca , A. , Ferraz , C. and Rios , J. ( 2015 ). ‘ Political dynasties and the quality of government ’, mimeo, PUC . Dahlström , C. and Sundell , A. ( 2014 ). ‘Göteborgsandan, korruption och opartiskhet i svenska kommuner (‘The Gothenburg Spirit: corruption and Impartiality in Swedish Municipalities’)’, in ( D. Karlsson and M. Gilljam, eds.), Svenska politiker: om de folkvalda i Kommuner och Landsting , pp. 217 – 34 , Stockholm : Santerus . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Dal Bo , E. , Dal Bo , P. and Snyder , J. ( 2009 ). ‘ Political dynasties ’, Review of Economic Studies , vol. 76 ( 1 ), pp. 115 – 42 . Google Scholar Crossref Search ADS WorldCat Eggers , A. and Hainmueller , J. ( 2009 ). ‘ MPs for sale? Returns to office in postwar British politics ’, American Political Science Review , vol. 103 ( 4 ), pp. 513 – 33 . Google Scholar Crossref Search ADS WorldCat Fafchamps , M. and Labonne , J. ( 2014 ). ‘ ‘Do politicians’ relatives get better jobs? Evidence from municipal elections’, Centre for the Study of African Economies, Working Paper No. 2014–37. Fisman , R. ( 2001 ). ‘ Estimating the value of political connections ’, American Economic Review , vol. 91 ( 4 ), pp. 1095 – 102 . Google Scholar Crossref Search ADS WorldCat Fisman , R. , Schulz , F. and Vig , V. ( 2014 ). ‘ The private returns to public office ’, Journal of Political Economy , vol. 122 ( 4 ), pp. 806 – 62 . Google Scholar Crossref Search ADS WorldCat Fiva , J. , Folke , O. and Sørensen , R. ( 2016 ). ‘ The power of parties: evidence from close municipal elections in Norway ’, Scandinavian Journal of Economics , (forthcoming). OpenURL Placeholder Text WorldCat Folke , O. and Rickne , J. ( 2016 ). ‘ The price of promotion: gender differences in the impact of career success on divorce ’, mimeo, Uppsala University . Gagliarducci , S. and Manacorda , M. ( 2015 ). ‘ Politics in the family: nepotism and the hiring decisions of Italian firms ’, mimeo, Queen Mary University . Geys , B. ( 2017 ). ‘ Political dynasties, electoral institutions and politicians’ human capital ’, Economic Journal , vol. 127 ( 605 ), pp. F474 – 94 . Google Scholar Crossref Search ADS WorldCat Jonsson , L. ( 2003 ). Kommunstyrelseordförande: Kommunledare med politisk förankring (Municipal Council Board Chair: Municipal Leader with a Political Mandate) , Västra Frölunda : Nya Doxa . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Kotakorpi , K. , Poutvaara , P. and Terviö , M. ( 2013 ). ‘ Returns to office in national and local politics ’, mimeo, Aalto University . Labonne , J. , Parsa , S. and Querubin , P. ( 2015 ). ‘ Political dynasties, term limits and female political empowerment: evidence from the Philippines ’, mimeo, New York University . Lundqvist , H. ( 2013 ) ‘ Is it worth it? On the returns to holding political office ’, mimeo, Stockholm University . Montin , S. ( 2007 ). Moderna kommuner (Modern Municipalities) , Malmö : Liber . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Nilsson , T. ( 2001 ). ‘ Den lokalpolitiska karriären: En socialpsykologisk studie av 20 kommunalråd ’ (The local political career: a social psychology study of 20 full‐time local politicians), Doctoral dissertation, Lund University . Persson , T. and Tabellini , G. ( 2000 ). Political Economics , Cambridge, MA : MIT Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Pettersson‐Lidbom , P. ( 2008 ). ‘ Do parties matter for economic outcomes? A regression‐discontinuity approach ’, Journal of the European Economic Association , vol. 6 ( 5 ), pp. 1037 – 56 . Google Scholar Crossref Search ADS WorldCat Querubin , P. ( 2013 ). ‘ Family and politics: dynastic incumbency advantage in the Philippines ’, mimeo, New York University . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rossi , M.A. ( 2017 ). ‘ Self‐perpetuation of political power ’, Economic Journal , vol. 127 ( 605 ), pp. F455 – 73 . Google Scholar Crossref Search ADS WorldCat SOU ( 2013 ). ‘ Goda affärer – en strategi för hållbar offentlig upphandling ’ (Good business – A strategy for sustainable public procurement). Report of the Swedish Government SOU 2013:12. Fritzes : Stockholm . Swedish Competition Authority ( 2015 ). ‘ Siffror och fakta om offentlig upphandling Statistik om upphandlingar som genomförts under 2014 ’ (Facts and figures about public procurement in 2014). Swedish Competition Authority, Report 2015:9. Van Coppenolle , B. ( 2015 ). ‘ Political dynasties in the UK House of Commons: the null effect of narrow electoral selection ’, mimeo, Leiden University . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Wångmar , E. ( 2013 ). Tillit och korruption. Korruption – maktmissbruk och bristande tillit i svensk lokalpolitik 1963–2011 (Trust and Corruption. Corruption – Abuse of Power and Lack of Trust in Swedish Local Politics 1963–2011), Stockholm : Santérus . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Author notes " The authors thank the editor Frederic Vermeulen, two anonymous referees, Martin Rossi and seminar participants at the Empirical Political Economics and Political Science Workshop (Stockholm University) and the ‘Political Dynasties’ Workshop (Norwegian Business School BI) for helpful comments. We also thank Dany Kessel, Johan Arntyr, Elin Molin, Jonas Allerup, Roza Khoban and Sirus Dehdari for expert research assistance. Financial support from the Swedish Research Council and the Torsten and Ragnar Söderberg Foundations is gratefully acknowledged. © 2017 Royal Economic Society
TI - Dynastic Political Rents? Economic Benefits to Relatives of Top Politicians
JF - The Economic Journal
DO - 10.1111/ecoj.12494
DA - 2017-10-01
UR - https://www.deepdyve.com/lp/oxford-university-press/dynastic-political-rents-economic-benefits-to-relatives-of-top-e90pj1wCcB
SP - F495
VL - 127
IS - 605
DP - DeepDyve
ER -