Born in the Purple: Political Dynasties and Politicians’ Human Capital · 2014. 6. 30. · 2 the onset of this article, we evaluate whether political dynasties induce a political

Born in the Purple:

Political Dynasties and Politicians’ Human Capital*

Gianmarco Daniele a and Benny Geys

b

a Vrije Universiteit Brussel (VUB), Department of Applied Economics, Pleinlaan 2, B-1050 Brussel, Belgium

b Norwegian Business School BI, Department of Economics, Nydalsveien 37, N-0442 Oslo, Norway

Email: [email protected]; [email protected]; [email protected]

This version :

June 2014

Abstract:

Recent academic interest in political dynasties is predominantly occupied with why such

dynasties arise and/or persist, rather than what their political and socio-economic

consequences are. In contrast, this article focuses on the effects of political dynasties in terms

of the human capital of elected politicians. Specifically, we argue – and empirically

substantiate using data from Italian municipal governments over the period 1985-2012 – that

political dynasties are likely to induce a political ‘Carnegie Effect’: i.e. a reduction of elected

politicians’ human capital due to the inheritance of electoral advantages associated with

political dynasties.

Keywords: Political dynasty, Local government, Elections, Human capital.

Word count: 10161 words

* The authors are grateful to Tommaso Aquilante, Lisa Blaydes, Andrea Colombo, Benoit S.Y. Crutzen, Nicola

Coniglio, Ernesto Dal Bó, Robert Dur, Zac Greene, Joshua Holm, Marc Jegers, Neil Malhotra, Francisco Pino,

Erik Wibbels, as well as participants at the ‘Political Economy Workshop’ at Erasmus University Rotterdam,

and research seminars at University of Bari, Université Libre de Bruxelles and Vrije Universiteit Brussel for

helpful comments and discussions. The first author also gratefully acknowledges the hospitality of the

Department of Political Science at Stanford University during part of the research underlying this article. Both

authors thank FWO Vlaanderen (grant number G.0022.12) for financial support.

1

“The political talent and policy depth so evident in the first

generation isn’t always present in the second generation, in

part because it’s not as necessary to fuel the rise [to power].”

Jeffrey Smith (former Missouri State Senator)

1. Introduction

Family firms have in recent years attracted substantial academic interest. Although empirical

results are not unambiguous, most studies suggest that family firms tend to under-perform

relative to firms with professional managers (Villalonga and Amit, 2006; Bennedsen et al.,

2007; Bloom and Van Reenen, 2010). One commonly expressed explanation for this finding

is that nepotism undermines firms’ performance by limiting labor market competition

(Burkart et al., 2003; Perez-Gonzales, 2006). This is known as the ‘Carnegie Effect’, after

Carnegie’s (1962 [1891]) famous argument that parents bequeathing large wealth to their

heirs will depress their talents and effort.

Even though positions in the political arena are generally awarded via elections, families

often continue to play a significant role also in politics. In fact, political dynasties are

common in as diverse settings as Argentina (Rossi, 2009), India (Bjolken and Chandra, 2012),

Japan (Fukui and Fukui, 1992; Asako et al., 2013), Mexico (Camp, 1995), the Philippines

(Querubin, 2013) and the United States (Dal Bó et al., 2009; Feinstein, 2010). In contrast to

research on family firms, however, this literature is predominantly interested in why political

dynasties arise and/or persist (Mosca, 1896; Fukui and Fukui, 1992; Dal Bó et al., 2009;

Rossi, 2009; Chhiber, 2011; Querubin, 2013), rather than what the consequences of their

existence are.

In this article, we instead focus on the relation between political dynasties and the human

capital of elected politicians. Particularly, and in line with Jeffrey Smith’s intuition cited at

2

the onset of this article, we evaluate whether political dynasties induce a political ‘Carnegie

Effect’: i.e. a reduction of elected politicians’ human capital resulting from the inheritance of

electoral advantages associated with political dynasties.1 This is an important question

because politicians’ human capital has been shown to play a central role in determining

socio-economic outcomes (Besley et al., 2010, 2011; Congleton and Zhang, 2013;

Gagliarducci and Nannicini, 2013). As such, our analysis concentrates on the critical –

though as yet overlooked – link between studies illustrating why political dynasties exist

(such as Dal Bó et al., 2009; see also above) and those concerned with their societal impact

(such as Asako et al., 2013).

From a theoretical perspective, we set up a simple probabilistic voting model where voters

cast their votes based on candidates’ human capital (reflected in their innate level of ability or

competence), political ideology and family connections. The model highlights that political

dynasties have a dampening effect on the human capital of elected dynastic politicians to the

extent that they inherit an electoral advantage from their political forebears. The intuition is

that such an electoral advantage (for evidence of such advantage, see section 4.1 below, as

well as Dal Bó et al., 2009; Rossi, 2009; Feinstein, 2010; Querubin, 2013) buttresses dynastic

politicians’ election odds, implying that they are able to win an election against a non-

dynastic politician even when their other characteristics (including human capital or ability)

are less favourable.

Our empirical analysis is based on a novel dataset including nearly 550.000 local politicians

in more than 8.000 Italian municipalities in the period 1985-2012. Following recent work on

1 Such electoral advantages may derive from the inheritance of political connections and networks (Fukui and

Fukui, 1992; Dal Bó et al., 2009; Asako et al., 2013; Querubin, 2013), clientilism and patronage ties that imply

the inheritance of actual vote(r)s (Mosca, 1896; Querubin, 2013), or name recognition effects (Feinstein, 2010;

Kam and Zechmeister, 2013). We return to this in more detail below.

3

nepotism in Italian academia (Allesina, 2011; Durante et al., 2011), political dynasties in the

Philippines (Querubin, 2013) and inter-generational social mobility (Clark, 2014; Clark and

Cummins, 2014), we operationalize political dynasties based on elected politicians sharing

the same surname. To deal with the imperfections in this measure and increase the validity of

our inferences, we not only assess this measure against detailed data from one randomly

drawn large city in our sample, but also implement robustness checks taking into account the

national and regional distribution of family names. Our key dependent variable is elected

politicians’ education level, which has previously been employed as a proxy for politicians’

human capital and quality by, for instance, Besley et al. (2011), Besley and Reynal-Querol

(2011), De Paola and Scoppa (2011), Galasso and Nannicini (2011), Geys and Daniele (2013)

and Pinotti (2013). The results illustrate that, all else equal, politicians appearing later within

a given dynasty indeed tend to be characterised by a lower level of education. This is in line

with a political ‘Carnegie Effect’, because we also observe that such dynastic politicians

benefit from significant electoral advantages. That is, they are, compared to non-dynastic

politicians, significantly more likely to achieve election into important political offices (i.e.

mayor or alderman) at their first election. Moreover, using a subsample of dynastic and non-

dynastic candidates running for mayor, we show that the former are significantly more likely

to be elected (controlling for their personal characteristics, including their education levels).

The next section sets out a simple theoretical framework that prepares the ground for the

empirical analysis. Section 3 presents the data and operationalization of our key variables,

while section 4 summarizes the empirical approach and key results. Section 5 concludes.

2. Theoretical model and hypotheses

4

To more thoroughly attend to the possible effect of political dynasties on elected politicians’

human capital, and derive empirically testable hypotheses, we introduce political dynasties

into a simple probabilistic voting model in the spirit of Persson and Tabellini (2002). In the

model, two politicians (i = A,B) run for election in a jurisdiction with a continuum of citizens

(represented by superscript j). To abstract from complications when allowing politicians to

gain experience from multiple terms in office, we assume they have not been in office before

and no historical information is available about them.2 For simplicity, we also normalize the

size of the population to 1, and assume that all individuals initially have one unit of a private

good c. Production of a public good by the government is financed entirely through taxes,

and requires that a uniform tax τ is imposed on voters (Besley and Case, 1995; Persson and

Tabellini, 2002), such that agent j’s budget constraint is .

While the level of taxation τ is assumed to be given (e.g., inherited from a previous period),

the level of the public good achieved with this tax revenue depends on the competence (or

ability) of the elected politician.3 In particular, we assume that high-ability, competent

politicians can achieve more public good (or higher-quality public goods) at any level of

taxation τ. This may reflect differences in rent-seeking behaviour (Besley and Case, 1995;

Ashworth et al., 2014), or varying ability to cater public good provision to the desires of the

2 The only agents making a decision in the model are voters, who have to decide which candidate to support

during an election. One could extend the model by including a (self-)selection stage, which occurs prior to the

actual election we here focus on. Such selection stage could involve two parties each choosing one candidate

to run in the election, or individuals deciding whether or not to stand for election. While such extension would

arguable enrich the model, we refrain from it here for two reasons. First, using backwards induction to solve

such extended model would imply working out the election stage, and then building on this to solve for the

(self-)selection stage equilibrium. The first step in this process would simply generate the outcome for the

election stage described below. Second, our dataset unfortunately does not allow addressing additional

hypotheses that can be derived with respect to equilibrium outcomes in the (self-)selection stage, such that

extending the model in this way appears unwarranted. 3 This is functionally equivalent to assuming – as in, for instance, Besley and Case (1995, 30) – that the

government “provides one unit of a public service of a given quality, financed entirely by taxes”, and sets a

level of taxation τ that depends on the ability level of the politician. We prefer to view public good provision

as the more flexible policy instrument since politicians can generally influence its amount, composition or

quality within a given budget constraint. Allowing the government to set the tax rate τ rather than the amount

(or quality) of public good does not affect our main inferences below.

5

electorate. Either way, it makes competent politicians more desirable to voters. Politicians’

competence is thereby modelled as a direct function of their ability ai [ ], which is

distributed according to a distribution function F(a) with associated density function f(a).4

Hence, the policy platform – in terms of the proposed level of public good (or, equivalently,

the level taxation τ) – of politician i (=A,B) is given by . Hence, the quasi-linear

preferences of any agent j over public good and private good c are given by:

( ) + (1)

where (.) is a concave, well-behaved function translating politician i’s competence (which

directly depends on her ability level ai; see above) into a given amount of public goods.

Politicians A and B not only differ in terms of their ability ai, but also in two other respects.

First, both politicians have a certain ideological make-up, which implies that voter j may have

an ideological bias towards one of them (j, which is uniformly distributed on [

] and,

without loss of generality, indicated from the perspective of politician B). While this

distribution is common knowledge, only voter j observes his parameter j. Second, each

candidate is characterised by an additional ‘popularity’ endowment, which for politician B

can be written as . In this popularity term, is the same for all voters

and is drawn from a uniform distribution on [

], Di

is an indicator variable that takes

value 1 if the politician is a dynastic candidate (0 otherwise), and h(.) measures the

effectiveness of the dynastic (dis)advantage. Hence, when one politician (i.e. B) is part of a

political dynasty while the other is not, reflects B’s popularity benefit/cost from

being part of a political dynasty.

4 Throughout the article, we assume that ability is an innate characteristic that politicians cannot alter; they are

simply born with it. This reflects the idea that politician’s ability level is given at the time they run for election.

While they may develop additional political skills in office, voters’ best available proxy for this future

development at the time of politicians’ first election is their innate ability level.

6

While, in principle, h(.) can be a positive or negative function depending on dynasties’

electoral (dis)advantage, previous work strongly suggests that political dynasties provide

substantial electoral benefits (Dal Bó et al., 2009; Rossi, 2009; Feinstein, 2010; Querubin,

2013). Such advantage may derive from a number of sources. First, dynastic politicians may

inherit the political connections and networks of their political ancestors, which provide a

substantial benefit during election campaigns (Fukui and Fukui, 1992; Dal Bó et al., 2009;

Asako et al., 2013; Querubin, 2013). Second, when at least some citizens base their vote

choice on personal ties, clientilism and patronage rather than public policy preferences or

ideology, dynastic politicians may also more directly inherit actual vote(r)s from their

political ancestors (Mosca, 1896; Querubin, 2013). Finally, dynastic politicians may benefit

from voters’ recognition of their name (Feinstein, 2010), since “name recognition, as a

heuristic for decision making, can play an important role in electoral politics” (Kam and

Zechmeister, 2013: 971). We summarize these various effects in the following assumption,

which we maintain throughout the remainder of the theoretical framework and test explicitly

in the empirical analysis.

Ass 1. Political dynasties confer an electoral advantage to their members.

The timing of the model follows Persson and Tabellini (2002): first, each voter observes his

own ideological bias j as well as politician i’s ability a

i. Second, is realized and all

uncertainty is resolved. Third, elections take place, and voters cast a ballot in favour of their

most preferred candidate. Given this framework, it is clear that voter j is indifferent between

politicians A and B under given ability levels aA and a

B whenever:

(2)

7

Then, for all aA and a

B there exists a voter with intrinsic parameter , ,δ) who is

indifferent between the two politicians. This level of ideological bias can be written as:

(3)

Given the distributional assumptions imposed above, this implies that the equilibrium vote

share of politician A equals:

[

] (4)

Assuming that elections take place under simple majority rule, politician A’s probability of

winning the election then is:

, or:

[ ] (5a)

where ∑ . Candidate B wins with probability , or:

[ ] (5b)

The important thing to observe from equations (5a) and (5b) is that any candidate’s

probability of election is increasing in both her ability level ai, but may be increasing or

decreasing in her family connection Di depending on the sign of h(.). The latter implies,

however, that any electoral advantage of dynastic politicians (i.e. when Ass. 1 holds) would

act as a substitute for ability in the electoral battlefield. Hence, a larger dynastic benefit

reduces the need to have a higher ability level ai than one’s opponent, which leads to our

central hypothesis.

H1. To the extent that political dynasties confer an electoral advantage to their members,

elected dynastic politicians have, on average, a lower level of ability than elected

non-dynastic politicians.

8

The intuition is that any electoral benefit obtained from being a dynastic politician increases

one’s probability of election for any given level of ability ai. Hence, relative to non-dynastic

politicians, dynastic politicians can win an election with a lower ai, all else equal. Since the

key mechanism underlying the negative dynasty-human capital relation in hypothesis H1

therefore lies in the electoral benefit conveyed by political dynasties on their members,

empirical verification of hypothesis H1 requires the concomitant substantiation of

Assumption 1.

3. Data and operationalization

3.1. Institutional setting: Italian municipalities

The Italian political system is characterized by four levels of governance, where

municipalities (about 8000) represent the lowest level, followed by provinces (110), regions

(20) and the national level. Nonetheless, as in most other West-European countries,

municipal governments retain important responsibilities with respect to education, social

welfare, culture and recreation, city planning, transport, economic development, waste

management and local police. They also have important fiscal powers, whereby setting the

local property tax rate represents the central annual financial decision (Bordignon et al.,

2003).

Local elections are held every five years to (s)elect council members and the (directly elected)

mayor. The exact electoral system depends on the size of the municipality. Below 15000

inhabitants, voters effectively have only one vote, which they cast for a candidate mayor and

her supportive list of candidates for the municipal council (though additional ‘preference

votes’ for candidates within this list of council candidates are possible). Elections take place

in a single round, in which the mayoral candidate obtaining most votes is elected as mayor

9

and her supportive list is allocated at least 66% of the council seats. The remaining seats are

allocated proportional to the vote share of all remaining mayoral candidates’ supportive lists.

Above 15000 inhabitants, voters are faced with parties (or coalitions thereof) that present a

list of candidates for the municipal council and support a candidate mayor. They cast one

vote for a candidate mayor and one vote for a list of candidates for the municipal council

(which can, but need not be, the list supporting a voter’s preferred mayoral candidate).

Elections for mayor here follow a run-off system, whereby the two top candidates run in a

second round whenever no candidate obtained an outright majority in round one. The list(s)

supporting the winning mayor are allocated at least 60% of the council seats, though now

there is a 3% threshold for the proportional allocation of the remaining seats (see Bordignon

et al., 2013, for more details).5

Finally, when elected, the mayor selects 2 to 12 aldermen (depending on the size of the

municipality) to enter into the Giunta, the governing body of the council. In practice, the

mayor selects aldermen from among the candidates enlisted in her supportive list(s). Indeed,

in cities below 15000 inhabitants, the mayor can only select aldermen among elected

councillors. In cities above 15000 inhabitants, however, she can select aldermen among all

eligible citizens (we return to this below).

3.2. Dataset

To empirically assess Assumption 1 and Hypothesis 1, we construct a novel dataset that

includes 549.966 politicians elected in the period 1985-2012 to any of the just over 8.000

Italian municipal councils. The original dataset is available from the website of the Italian

Ministry of Interior (http://amministratori.interno.it/AmmIndex5.htm; in Italian), and

5 Note that, independent of the size of the municipality, politicians defeated in their campaign for mayor are

generally elected as council members as long as their ‘supportive list(s)’ obtains at least one seat in the

municipal council.

10

contains information on politicians’ socio-demographic background (e.g., age, gender,

education level), their position within the local government (e.g., mayor, alderman,

councillor), and the time at which they were (s)elected into their current position. Clearly,

politicians are normally in office for more than one year, which implies that politicians

appear multiple times in the complete dataset we assembled. Politicians might, however, be

accumulating political capital over time (and thus presumably achieve higher quality as a

politician), or invest in additional education after becoming elected the first time. To avoid

any bias induced by such effects, we only include one unique observation per municipality

for each politician in our estimations: namely, their first year of election with the education

level recorded at that time. This restriction also reflects the assumption in the theoretical

model that politicians have not been in office before (see above), and avoids bias from any

incumbency advantages that might arise during later elections (Erikson, 1971; Ansolabehere

et al., 2000; Carson et al., 2007; Hirano and Snyder, 2009). Note, however, that this does not

exclude politicians from appearing twice in the dataset when they move to another

municipality within the period of observation. In that case, they do appear multiple times, but

only once per municipal council they become elected into (i.e. in the first year of their

respective elections). Nonetheless, such multiple reappearances are very rare as politicians’

electoral support at the local level is not easily transferable from one municipality to another.

For politicians appearing in the dataset near the onset of our observation period (i.e. 1985),

we cannot be certain that this is really their first election. While these politicians are retained

throughout the analysis below, we experimented with the exclusion of all politicians first

appearing into the dataset before 1990 or 1995. These cut-offs are inspired by the length of

the local-level electoral cycle in Italy (i.e. five years), and the fact that Italian local politicians

generally have very short tenures. In fact, approximately 90% of local politicians in southern

11

Italy are replaced within ten years (Geys and Daniele, 2013), which implies that politicians

appearing after 1995 are highly unlikely to have been elected before in the same municipality.

While not reported to preserve space, excluding politicians first appearing before 1990, or

before 1995, leaves all results below qualitatively unchanged (details upon request).

3.3. Measuring political dynasties

The central explanatory variables throughout our analysis reflect the position of each

politician within a political dynasty. Similar to recent studies on academic and political

dynasties (Allesina, 2011; Durante et al., 2011; Querubin, 2013) as well as inter-generational

social mobility (Clark, 2014; Clark and Cummins, 2014), we thereby look at individuals with

the same surname to identify (presumed) family ties. This operationalization proceeds in two

steps. In the first step, we locate all politicians with the same surname in the same municipal

council throughout the period of observation. This indicates that in 69 percent of all cases,

only one politician has a certain surname in a given municipality in our sample period.

Approximately 18 percent of all surnames appear twice and 12 percent appear three times or

more in a given municipality (note that it is extremely rare to observe more than five

politicians with the same surname in the same municipal council throughout the period of

observation). Then, as a second step, we determine the position of each politician within a

dynasty and define a set of indicator variables (i.e. Dynasty=2 and Dynasty 3) equal to 1

whenever a politician is, respectively, a second or third (or later) generation politician.6 The

reference category in the estimations below thus is made up of politicians whom are either

not a member of a political dynasty, or the first generation in a political dynasty. ‘First-

generation’ dynastic politicians are included in the control group because they were not (yet)

part of a political dynasty at the time we first observe them. As such, we have no theoretical

6 Given the observed distribution of surnames among politicians in any given municipality (see above), we

impose a cut-off at three surname occurrences. Still, all results reported below remain valid when imposing

alternative cut-offs (details upon request).

12

reason to view them as different from non-dynastic politicians at that point in time. Moreover,

the key mechanism highlighted in the theoretical model relates to inherited electoral benefits

within political dynasties (see Ass. 1), which is evidently only possible as of the second

generation.

Using surnames to operationalize political dynasties is a valid approximation in our Italian

setting since children receive the surname of their father, while women commonly take the

surname of their husband upon marriage. Even so, one can wonder about the precision of a

dynastic variable based on surnames, since people can have the same surname without having

any kinship ties (i.e. surname homonymy). To assess this, we checked whether presumed

dynastic politicians in our dataset were effectively part of a political dynasty. As we fail the

necessary resources to do this for all 549.966 politicians in our sample, we arbitrarily chose

the city of Bari (the second biggest city in southern Italy, with about 315.000 citizens) for this

analysis. This choice was not only driven by the need to have a relevant number of

observations and real-world information availability, but also by the fact that surname

homonymy without kinship ties is likely to be a larger concern in big cities. Hence, a large

city presents a harder test for our operationalization. The results indicate that 47 out of 271

politicians elected into Bari’s municipal government in the period 1985-2012 are presumed to

be dynastic politicians based on shared surnames. After meticulous fieldwork, we count 18

real cases of dynastic politicians within this group (mostly father/son and father/daughter

connections). Dropping the most common surnames from the dataset (using a cut-off at 50

occurrences, though results are similar using different cut-offs), the sample of presumed

dynastic politicians is reduced to 34, while the group of real dynastic politicians still counts

18 observations. Given the hard case provided by a large city, we evaluate this result as very

encouraging. Moreover, these results indicate that we can reduce any bias in our estimates

13

due to surname homonymy by dropping the most commonly observed surnames from the

estimation sample. Hence, we will experiment extensively with different cut-offs to verify the

robustness of our results.

There are two additional reasons why surnames provide an imperfect approximation of actual

family connections. First, it only catches ties between family members when they have the

same surname. While these reflect the closest family ties that are likely to generate the

strongest effects (e.g. spouse, children, grandchildren), it may overlook more distant kinship

ties (e.g. cousins, nephews, son-in-law). Second, as our observation window starts in 1985,

we may overlook family connections to politicians that were active before this period.

Although data availability prevents us from directly addressing both issues, it is important to

observe that they bias our estimates towards zero. Both issues indeed imply that we fail to

define a certain number of dynastic politicians as part of a dynasty, such that these remain in

the control group. Since elected dynastic politicians are expected to have a lower level of

ability compared to elected non-dynastic politicians (see hypothesis H1), this ‘mis-allocation’

pushes the average ability level in the ‘control’ group (i.e. non-dynastic politicians) closer to

the average ability level in the ‘treatment’ group (i.e. dynastic politicians) – inducing a bias in

our estimates towards zero. This not only stacks the deck against us (in terms of the

likelihood to find supportive evidence of H1), but also implies that our findings below reflect

a lower bound of the true effect of political dynasties.

A final potential concern about using surnames to operationalize political dynasties is that

there may exist a correlation between surnames and individuals’ characteristics. For instance,

it might be that more common last names are associated with a lower socio-economic

background. As such social background is likely to be correlated with individuals’

14

educational attainment (Ermisch and Francesconi, 2001; Dustmann, 2004), this could bias our

estimates. Still, we consider this unlikely in our setting for two reasons. First, hereditary

surnames started to be generally used in Italy in the 12th

century (Marcato, 2010). Even if

their distribution originally might have been partially tied to social class, several centuries of

up- and downward social mobility is unlikely to have sustained this to any significant extent.

Second, the main drivers of Italian surname distributions are geography and history, and not

social categories. This is due to substantial “linguistic fragmentation and (…) the late process

of standardization following the late and slow diffusion of a national language”

(http://www.treccani.it/enciclopedia/cognomi_(Enciclopedia-dell'Italiano)/, accessed on 29

January 2014, own translation). While we implicitly control for the distribution of last names

at the local level in the specification of our central dependent variable (see below), our

robustness checks further address this concern by excluding the most common regional

surnames from the estimation sample.

4. Empirical analysis

Our empirical analysis proceeds in two steps. First, in section 4.1, we analyse whether, and to

what extent, dynastic politicians demonstrate an electoral advantage over non-dynastic

politicians (Ass. 1). Then, in section 4.2, we turn to our assessment of the effect of political

dynasties on the human capital of elected dynastic politicians (hypothesis H1).

4.1. Political dynasties and politicians’ careers

To evaluate whether dynastic politicians benefit electorally from their political ancestry, we

follow two strategies. First, we assess to what extent dynastic politicians are more likely to be

elected into decision-making positions within the municipal government (i.e. mayor or

alderman) in their first election. This approach is based on the idea that, particularly during

15

their first election, politicians’ success arguably depends less on their own political capital or

experience (which is by definition minimal at that time), and more on any electoral or

political benefits inherited from political forebears. Hence, the electoral advantages of

political dynasties are likely to play a relatively larger role during politicians’ first election.

The empirical model takes the following form (subscript i refers to politicians, c to

municipalities, and t to time):

(6)

Our dependent variable ( ) is an indicator variable equal to one if politician i

in municipality c at time t was elected into an important decision-making position within the

municipal government at his/her first election (0 otherwise). We thereby treat positions as

mayor or alderman separately such as to assess possible heterogeneity of the studied effects

across different political positions. Beside our key dynastic variables (i.e. a set of indicator

variables equal to one whenever a politician is, respectively, a second or third-or-later

generation politician; see section 3.2), the set of control variables at the individual level

( ) includes politicians’ gender (1 if female), age, and the time of their first

election.7 We also control for the effect of politicians’ level of education as members of

political dynasties might exhibit higher levels of education (or other aspects of political

quality) thanks to a privileged upbringing as children of politicians. Including education as a

control implies that we look at the effect of political dynasties independent of its potential

effect on politicians’ education levels. The education variable is thereby based on information

concerning the highest degree a politician obtained: i.e. Elementary School (or less), Lower

Secondary School, Higher Secondary School and University Degree or Higher Degree. For

7 We do not include politicians’ partisan attachment because local-level politics in Italy is often characterised by

purely local lists based on specific policy issues or political personalities.

16

ease of interpretation, we follow De Paola and Scoppa (2010), Baltrunaite et al. (2012) and

Geys and Daniele (2013) in translating degrees into the minimum number of years necessary

to obtain them (assuming an uninterrupted education path): i.e. elementary education = 5

years; lower secondary = 8 years; higher secondary = 13 years; university or more = 18 years.

Control variables at the municipal level ( ) include population size and

density, average number of individuals per family, share of people with at least secondary

school degree, school dropout rate, share of elderly (over 65 years), unemployment rate,

altitude and indicator variables for being a seaside town and provincial capital (summary

statistics and sources for all variables are in Table A1 in the appendix). All estimations also

include a full set of region fixed effects to control for possible regional heterogeneity.8

Finally, since politicians are by construction embedded in municipalities, we cluster error

terms at the municipality level to account for the possible lack of independence of

observations from the same municipality.

The results are summarized in table 1 (for election as mayor) and table 2 (for election as

alderman).9 To mitigate the potential influence of surname homonymy on our findings (see

above), we not only report results using the entire sample (column 1), but also using reduced

samples retaining only surnames with less than 100 (column 2) or 50 occurrences (column 3)

in the overall nation-wide sample – or less than 10 occurrences within a given region (column

4). Since these cut-offs naturally are somewhat arbitrary, it is important to observe that our

results remain robust also when imposing different cut-offs.

__________________

8 We rely on region – rather than municipality – fixed effects merely for computational reasons. Including

municipality fixed effects increases the size of the estimated variance-covariance matrix beyond the point

where our statistical software (i.e. Stata 12) can readily handle it. However, randomly splitting the sample in

two equal parts and estimating both subsamples with municipality fixed effects leaves our findings unaffected

(details available upon request). 9 Remember that the mayor can in some circumstances also appoint non-elected individuals as aldermen. We

exclude those observations here, such as to only focus on politicians who effectively ran for election.

17

Tables 1 and 2 about here

__________________

A brief look at the individual-level controls in tables 1 and 2 suggests that women are

substantially less likely to be elected as mayor in their first election (table 1), while the

reverse holds for women’s likelihood to obtain a position as alderman (table 2). This

difference might be due to the fact that aldermen are appointed by the mayor, and women

represent a minority among Italian politicians (17% in our sample). To preserve sufficient

female representation in decision-making positions, a mayor might then be tempted to give

preference to women when selecting aldermen. The year of politicians’ first election has a

negative effect in table 1, but a positive one in table 2. This suggests that it has become

harder over time to become mayor during one’s first election, but becoming alderman has

become somewhat easier. Finally, older and better educated politicians are more likely to

achieve an important political office in their first election. The latter is particularly important

because it reflects an important ‘signalling’ value of education (which we rely on below).

Turning to our key dynastic variables, we find, in line with the central precondition for a

political ‘Carnegie effect’ (i.e. Ass. 1 in section 2), that dynastic politicians have a

significantly higher probability to become mayor in their first election. Specifically, being a

second generation politician increases such probability by 21 to 23 percent, while being a

third (or later) generation dynastic politician increases it by 30 to 36 percent. Table 2 reports

roughly equivalent – though generally weaker – results regarding dynastic politicians’

probability of becoming alderman in their first election. Specifically, being a second

generation politician increases this probability by 6 to 8 percent, while being a third (or later)

generation politician increases it by 10 to 13 percent. These results arise despite direct

controls for politicians’ education, which implies that political dynasties have a strong effect

18

over and above any influence deriving from a potentially privileged upbringing. Note also

that the size of the estimated effect in both tables 1 and 2 tends to increase when excluding

the most common surnames (i.e. moving towards the right-hand side of both tables). This

confirms a slight bias towards zero induced by surname homonymy in the full-sample

estimates.

Second, as an alternative strategy to assess the electoral advantage of dynastic politicians, we

make use of the fact that information about candidates running for mayor is publicly available

starting with the year 2000. Using this information for the period 2000-2012 (N=42,982), we

are able to test whether dynastic candidates are more likely to win the race for mayor

compared to non-dynastic candidates, controlling for a number of background characteristics.

The benefit of this alternative approach is that we not only observe actually elected mayors

(as in the analyses underlying tables 1 and 2), but also candidates that failed to win the

mayoral election. This allows us to explicitly rule out potential self-selection effects arising

from the fact that dynastic politicians might be more likely to run for mayor compared to

non-dynastic politicians.

The estimation model is comparable to that in equation (6), except that the dependent

variable is now an indicator variable equal to 1 if a politician runs as mayor and wins the

election (0 if (s)he runs as mayor, but loses the election). As the dataset here includes all

mayoral candidates independent of the number of previous elections they contested

(remember that thus far we only employed politicians in their first election), our set of

individual-level control variables is extended with an indicator variable equal to one for

candidates without any previous political experience (i.e. First Election). This is important to

19

control for any possible incumbency advantage (Erikson, 1971; Ansolabehere et al., 2000;

Carson et al., 2007; Hirano and Snyder, 2009). The results are reported in Table 3.

__________________

Table 3 about here

__________________

Starting the discussion again from our control variables, we find that inexperienced mayoral

candidates are considerably less likely to be elected as mayor. Candidates with higher

education are again more likely to be elected as mayor. Gender and age have negative effects.

Crucially, the results in table 3 again confirm a positive effect of political dynasties on the

probability to win (mayoral) elections. The effect is substantial, ranging from 16%-22% for

second generation politicians to 32%-39% for third-or-later generation politicians.

Finally, one might wonder whether the electoral advantage of dynastic politicians is larger for

inexperienced politicians (who might need to rely more heavily on their political forebears;

see above). To assess this, we split the sample between politicians who run for mayor for the

first time (column 5) and politicians who have previously already been elected into the

municipal council (column 6).10

The results indicate that the dynastic advantage is indeed

substantially larger for candidates without previous political experience (in column 5) as

compared to previously elected politicians (in column 6). This is in line with our theoretical

argument as it suggests that the ‘signalling’ value of membership in a political dynasty is

larger for new politicians.

10

Including interaction terms between First Election and the dynastic variables provides qualitatively similar

results, and also highlights that the difference in the dynastic effect between experienced and inexperienced

politicians is statistically significant beyond the 99% confidence level.

20

4.2. Political dynasties and politicians’ ability

Overall, the results in section 4.1 indicate that being a dynastic politician is a strong predictor

of political success in Italian local politics. This is in line with previous work suggesting that

political dynasties provide substantial electoral benefits (Dal Bó et al., 2009; Rossi, 2009;

Feinstein, 2010; Querubin, 2013), and supports Ass. 1 in the theoretical model. Having thus

established the electoral benefit provided by political dynasties, we can now turn to testing

whether this invokes a political ‘Carnegie Effect’ (as stated in hypothesis H1). The empirical

model takes the following form (where subscript i refers to politicians, c to municipalities,

and t to time):

(7)

Our dependent variable (Yict) operationalizes the ability level of politician i in municipality c

at time t via her educational attainment (measured, again, by translating degrees into the

minimum number of years necessary to obtain them; see above). This follows recent research

that has employed elected politicians’ human capital as a proxy for their political quality

(Besley et al., 2011; De Paola and Scoppa, 2010; Galasso and Nannicini, 2011; Geys and

Daniele, 2013; Pinotti, 2013). It rests on the idea that human capital is a valid proxy for

individuals’ skills, which has been widely recognized in research on the role of human capital

in the private sector (Black and Lynch, 1996; Hitt et al., 2001) and is gaining increasing

prominence also with respect to public sector activities (Besley and Reynal-Querol, 2011).

Importantly, unlike alternative possible measures of politicians’ ability (such as, for instance,

political tenure), politicians’ education level has the benefit of being unaffected by potential

reverse causality issues. The reason is that a politician’s final education level cannot

influence her being born into a political family (for obvious reasons), and is highly unlikely

21

to induce other family members to enter politics (which would be required for membership of

a political dynasty to follow – rather than precede – obtaining one’s highest degree). This

allows us to interpret our findings as reflecting a causal connection running from (the

electoral advantage conferred by) political dynasties to elected politician’s human capital.

We should note here that we define our dependent variable (Yict) in relative terms. That is, we

specify it as the difference between the education level of politician i in municipality c at time

t and the average level of education among all politicians in municipality c at time t. This

transformation is important because it allows us to directly control for the possible effects of

idiosyncratic characteristics of the municipalities – including, for instance, last name

distribution. The distribution of the final dependent variable (Yict) is provided in figure 1.

__________________

Figure 1 about here

__________________

At the individual level ( ), we again control for politicians’ gender (1 if

female), age, and the time of their first election. Given the specification of our dependent

variable in relative terms, there is no need to include our earlier municipal-level controls

(though doing so anyway leaves our results unaffected; details upon request). The results –

using OLS due to the continuous nature of our dependent variable (see figure 1) – are

presented in table 4. As before, we first display a set of baseline results using the full sample

of politicians. Then, we present robustness checks using reduced samples retaining only

surnames with less than 100 (column 2) and 50 occurrences (column 3) in the overall nation-

wide sample – or less than 10 occurrences within a given region (column 4).

__________________

Table 4 about here

22

__________________

Table 4 provides substantial evidence in favour of a political Carnegie Effect. Indeed, all

coefficient estimates for variables related to political dynasties are significantly negative,

which indicates that elected dynastic politicians – independent of their position in the dynasty

– are characterised by lower education levels on average. In terms of size, our findings

suggest that being a second generation politician is linked to a reduced level of education

equivalent to four months (seven months for third-or-later generation politicians) or 8% of

one standard deviation of the relative education variable (15% for third-or-later generation

politicians). From the results in tables 1 to 3, we can infer that this effect is at least in part

caused by the electoral advantages provided within political dynasties. Moreover, the

negative dynasty-education relation tends to strengthen substantially when moving from

second to third (or later) generations. In fact, the coefficient estimates roughly double in size,

and this difference between second and third (or later) generations is statistically significant

at conventional levels. This finding bears striking similarity to the downward trend in ruler

duration within dynasties in the Islamic World between 700CE and 1500CE (Blaydes and

Chaney, 2013). While providing no direct evidence to substantiate this, Blaydes and Chaney

(2013, 28) suggest that this negative trend may well derive from differences in human capital

between the founders and subsequent members of dynasties: i.e. they “expect sovereign

human capital to mean-revert over time”. Our results are clearly in line with such

supposition.11

11

Since the key mechanism underlying our theoretical argument consists of a substitution effect between the

electoral advantages provided by political dynasties and politicians’ human capital (see section 2), it follows

that anything undermining such dynastic electoral advantage should weaken the negative dynasty-human

capital relation observed here. As a supplementary test of this corollary, we collected information on the

closeness of all mayoral election outcomes (defined as one minus the difference between the vote shares of the

winner and runner-up in the mayoral race) over the period 2000-2012. Close races tend to entail more

vigorous electoral campaigns (Cox and Munger, 1989) and make that politicians are more responsive in terms

of “policy promises (speeches)” (Hobolt and Klemmensen, 2008: 309), which could undermine any dynastic

advantage. Adding interaction terms between the closeness of the election and our two dynastic variables to

the analyses in table 3 and 4, we find that the electoral advantage of political dynasties indeed weakens in

elections characterised by higher political competition. This, in turn, translates into a weaker dynasty-human

23

As a final test, tables 5 and 6 replicate the estimations in table 4 on subsamples of mayors

(table 5) and aldermen (table 6). This allows investigating whether the negative link between

political dynasties and elected politicians’ human capital is heterogeneous across political

offices. The results indicate that the dynasty-human capital relation observed in table 4 is

much stronger for aldermen than for mayors. Indeed, in the mayor subsample (table 5), the

estimated effect ranges from 5% to 10% of one standard deviation, and fails to reach

significance at conventional levels in the most restrictive subsample. In the aldermen

subsample (table 6), the effect of political dynasties on politicians’ human capital ranges from

six to twelve months of reduced education, or 14% to 25% of one standard deviation.12

As

mayors and aldermen are de facto the decision makers in the municipal council, these results

can have important implications for the performance of the local council.

__________________

Tables 5 and 6 about here

__________________

Note that we also learn from tables 4, 5 and 6 that older politicians have lower education

levels relative to the average of all other politicians in their municipality at the time of their

initial election. This is unsurprising given rising education levels over time. Also, female

politicians have, ceteris paribus, a significantly higher level of education relative to the

average of all other politicians in their municipality at the time of their initial election. This is

in line with recent evidence from German MPs illustrating that female politicians require

stronger quality signals – in terms of education level or terms in office – to obtain extra-

parliamentary jobs once they become elected (Geys and Mause, 2014).

capital relation when electoral competition is higher (full details upon request). These results further

substantiate our central claim that a negative dynasty-human capital relation (hypothesis H1) depends at least

to some extent upon political dynasties providing an electoral advantage (Ass. 1). 12

These findings are unchanged if we introduce interaction terms (i.e. dynasty*mayor and dynasty*aldermen) in

the analysis underlying table 4 – rather than splitting the sample in different groups (as in tables 5 and 6).

24

5. Conclusion

Although recent work on political dynasties has improved our understanding of why such

dynasties arise and/or persist (Mosca, 1896; Fukui and Fukui, 1992; Dal Bó et al., 2009;

Rossi, 2009; Chhiber, 2011; Querubin, 2013), analyses of their economic and political

consequences have been slow to emerge (for a recent exception, see Asako et al., 2013). This

article provided a first step towards bridging this gap by evaluating the relation between

political dynasties and politicians’ human capital. From a theoretical perspective, we argued

that any electoral advantages inherited by dynastic politicians (see also Dal Bó et al., 2009;

Feinstein, 2010; Querubin, 2013) are likely to positively affect their election odds

independent of their inherent ability, which has a dampening effect on the human capital of

elected dynastic politicians.

Our main empirical results – using data about approximately 550.000 Italian local politicians

– first of all confirmed that dynastic politicians obtain a substantial advantage at the start of

their political career. Specifically, they are significantly more likely to be elected as mayor or

appointed as alderman in their first election (compared to non-dynastic politicians), and have

a significantly higher probability to win mayoral elections (compared to non-dynastic

candidates in the mayoral race). Secondly, and in line with theoretical predictions, we find

that dynastic politicians are characterised by a lower ability level (as approximated by their

highest level of educational attainment), and that this negative dynasty-human capital relation

increases with one’s rank in the dynasty. Although the estimated effects in our analyses are

likely to reflect a lower bound, they are nonetheless substantively meaningful. For instance,

the reduction in the observed education level among second-generation (third-or-later

25

generation) dynastic politicians reflects approximately 8% (15%) of one standard deviation in

observed education levels. This is indicative of an important political ‘Carnegie Effect’.

Overall, our analysis suggests that political dynasties may have important implications in

terms of policy decisions as well as socio-economic outcomes, since the latter are in part

determined by politicians’ human capital (Besley et al., 2010, 2011; Besley and Reynal-

Querol, 2011; Congleton and Zhang, 2013; Gagliarducci and Nannicini, 2013). Nonetheless,

our results should not be interpreted as suggesting that dynastic politicians are necessarily

worse politicians. On the one side, dynastic politicians benefit from social networks,

campaigning skills and other inherited competences that might be beneficial in terms of

policy-making; on the other side, clientilism voting and name recognition are likely to just

foster inefficiencies in the political arena. Our analysis cannot exclude that the dampening

effect of political dynasties on elected politicians’ human capital could be compensated by

other political skills (e.g. electoral campaign organization, social networking) that dynastic

politicians might inherit from their ancestors. Therefore, one important area for future

research is to evaluate whether, and to what extent, the educational implications uncovered

here carry over into policy-making. In similar vein, and consistent with the rapidly expanding

literature on family firms (Villalonga and Amit, 2006; Bennedsen et al., 2007; Bloom and

Van Reenen, 2010), future research should assess whether jurisdictions with a larger number

(or share) of dynastic politicians tend to document different performance (e.g., in terms of

economic growth or government efficiency) relative to jurisdictions where political dynasties

are less prevalent.

26

References

Allesina, S. (2011). Measuring nepotism through shared last names: The case of Italian

academia. PloS one 6(8): e21160.

Ansolabehere, S., Snyder, J.M. and Stewart C. (2000). Old voters, new voters, and the

personal vote: Using redistricting to measure the incumbency advantage. American

Journal of Political Science, 44: 17-34.

Asako, Y., Iida, T., Matsubayashi, T., and Ueda, M. (2013). Dynastic politicians: Theory and

evidence from Japan. Mimeo.

Ashworth, J., Geys, B. Heyndels, B. and Wille, F. (2014). Competition in the Political Arena

and Local Government Performance. Applied Economics 46(19): 2264-2276.

Bennedsen, M., Nielsen, K., Perez-Gonzales, F., and Wolfenzon D. (2007). Inside the family

firm: The role of families in succession decisions and performance. Quarterly Journal of

Economics 122(2): 647-691.

Besley, T. and A. Case (1995). Incumbent Behavior: Vote Seeking, Tax Setting and

Yardstick Competition. American Economic Review 85: 25-45.

Besley, T. and Reynal-Querol, M. (2011). Do democracies select more educated leaders?

American Political Science Review 105(3), 552-566.

Besley, T., Montalvo, J., and Reynal-Querol, M. (2011). Do educated leaders matter?

Economic Journal 121(554): 205-227.

Besley, T., Persson, T., and Sturm, D. (2010). Political competition, policy and growth:

Theory and evidence from the United States. Review of Economic Studies 77(4): 1329-

1352.

Bjolken, A., and Chandra, K. (2012). Dynastic politics and party organizations: Why family

ties improve electoral performance in India. Mimeo.

Black, S.E., and Lynch, L.M. (1996). Human capital investments and productivity. American

Economic Review 86(2): 263-267.

Blaydes, L., and Chaney, E. (2013). The feudal revolution and Europe’s rise: Political

divergence of the Christian West and the Muslim world before CE1500. American

Political Science Review 107(1): 16-34.

Bloom, N., and Van Reenen, J. (2010). Why do management practices differ across firms and

countries? Journal of Economic Perspectives 24(1): 203-224.

Bordignon, M., Cerniglia, F., and Revelli, F. (2003). In search of yardstick competition: a

spatial analysis of Italian municipality property tax setting. Journal of Urban Economics,

54(2), 199-217.

Bordignon, M., Nannicini, T., and Tabellini, G. (2013). Moderating political extremism:

Single round vs runoff elections under plurality rule. IZA Discussion Paper, No. 7561

Burkart, M., Panunzi, F., and Shleifer, A. (2003). Family firms. Journal of Finance 58: 2167-

2202.

Camp, R.A. (1995). Political Recruitment across Two Centuries: Mexico, 1884-1991. Austin:

University of Texas Press.

Carnegie, A. (1962 [1891]). The advantages of poverty. In E.C. Kirkland (ed.), The Gospel of

Wealth and Other Timely Essays. Cambridge: The Belknap Press.

Carson, J.L., Engstrom, E.J. and Roberts, J.M. (2007). Candidate Quality, the Personal Vote,

and the Incumbency Advantage in Congress. American Political Science Review 101(2):

289-301.

Cartocci, R. (2007). Mappe del Tesoro. Bologna: il Mulino.

Chhibber, P. (2011). Dynastic parties: Organization, finance and impact. Party Politics 19(2):

277-295.

Clark, G. (2014). The Son also Rises: Surnames and the History of Social Mobility. Princeton:

Princeton University Press.

27

Clark, G. and Cummins, N. (2014) Intergenerational wealth mobility in England, 1858-2012:

Surnames and social mobility. Economic Journal, forthcoming.

Congleton, R. D. and Zhang, Y. (2013). Is it all about competence? The human capital of US

presidents and economic performance. Constitutional Political Economy, 1-17.

Cox, G.W. and M.C. Munger (1989), Closeness, expenditures and turnout in the 1982 US

house elections. American Political Science Review, 83(1): 217-230.

Dal Bó, E., Dal Bó, P., and Snyder, J. (2009). Political dynasties. Review of Economic Studies

76(1): 115-142.

De Paola, M., and Scoppa, V. (2011). Political competition and politician quality: Evidence

from Italian municipalities. Public Choice 148(3-4): 547-559.

Durante, R., Labartino, G., and Perotti, R. (2011). Academic dynasties: Decentralization and

familism in the Italian academia. NBER Working Paper, no. w17572.

Dustmann, C. (2004). Parental background, secondary school track choice, and wages.

Oxford Economic Papers 56(2): 209-230.

Erikson, R.S. (1971). The advantage of incumbency in congressional elections. Polity, 3:

395-405.

Ermisch, J. and Francesconi, M. (2001). Family matters: Impacts of family background on

educational attainments, Economica, 68, 137-156.

Feinstein, B.D. (2010). The dynasty advantage: Family ties in congressional elections.

Legislative Studies Quarterly 35: 571-598.

Fukui, S.N., and Fukui, H. (1992). Elite recruitment and political leadership. Political Science

and Politics 25: 25-36.

Gagliarducci, S., and T. Nannicini (2013). Do better paid politicians perform better?

Disentangling incentives from selection. Journal of the European Economic Association,

11(2), 369-398.

Galasso, V., and Nannicini, T. (2011). Competing on good politicians. American Political

Science Review 105(1): 79-99.

Gelman, A. and G. King (1990). Estimating Incumbency Advantage Without Bias. American

Journal of Political Science 34(4): 1142-1164.

Geys, B. and G. Daniele (2013). Organized crime and political quality: Evidence from Italian

municipalities. Mimeo.

Geys, B., and Mause, K. (2014). Are female legislators different? Exploring sex differences

in German MPs’ outside interests. Parliamentary Affairs forthcoming.

Hirano, S. and Snyder, J.M. (2009). Using multi-member districts to decompose the

incumbency advantage. American Journal of Political Science, 53: 292-306.

Hitt, M.A., Bierman, L., Shimizu, K., and Kochhar, R. (2001). Direct and moderating effects

of human capital on the strategy and performance in professional service firms: A resource

based perspective. Academy of Management Journal 44: 13-28.

Hobolt, S.B. and R. Klemmensen (2008), Government Responsiveness and Political

Competition in Comparative Perspective. Comparative Political Studies, 41: 309-337.

Kam, C.D. and Zechmeister, E.J. (2013). Name Recognition and Candidate Support.

American Journal of Political Science, 57(4): 971-986.

Marcato, C. (2010). Enciclopedia dell’Italiano. Treccani.it.

Mosca, G. (1966 [1896]). La Classe Politica. Bari: Laterza Editori.

Perez-Gonzales, F. (2006). Inherited control and firm performance. American Economic

Review 96(5): 1559-1588.

Persson, T. and Tabellini, G. (2002). Political Economics: Explaining Economic Policy.

Cambridge: MIT Press.

28

Pinotti, P. (2013). Organized Crime, Violence and the Quality of Politicians: Evidence from

southern Italy. In P. Cook, S. Machin, O. Marie and G. Mastrobuoni (eds.), Lessons from

the Economics of Crime: What Works in Reducing Offending?, Cambridge: MIT Press.

Querubin, P. (2013). Families and politics: Dynastic incumbency advantage in the Philippines.

Mimeo.

Rossi, M.A. (2009). Political dynasties: Evidence from a natural experiment in Argentina.

Mimeo.

Villalonga, B., and Amit, R. (2006). How do family ownership, control and management

affect firm value? Journal of Financial Economics 80(2): 385-417.

29

Table 1: Electoral benefits of political dynasties: Election as mayor in first election

All Cut-off=100 Cut-off=50 Region Cut-off=10

Dynasty=2 1.210*** 1.226*** 1.229*** 1.215***

(0.027) (0.031) (0.034) (0.039)

Dynasty 3 1.298*** 1.303*** 1.325*** 1.362***

(0.035) (0.040) (0.044) (0.059)

Woman 0.531*** 0.547*** 0.563*** 0.580***

(0.017) (0.019) (0.021) (0.024)

Age 1.059*** 1.058*** 1.058*** 1.056***

(0.001) (0.001) (0.001) (0.001)

Time First Election 0.963*** 0.963*** 0.963*** 0.963***

(0.001) (0.001) (0.001) (0.001)

Education 1.224*** 1.222*** 1.221*** 1.216***

(0.003) (0.003) (0.003) (0.004)

City Controls YES YES YES YES

Region Fixed Effects YES YES YES YES

R2 0.09 0.07 0.09 0.09

N 534,450 432,442 363,316 296,785

Note: Dependent variable is an indicator variable equal to 1 if a politician was elected as mayor at his/her first

election (0 otherwise). Coefficient estimates are represented as odds ratios from a Logit model (robust

standard errors clustered at the level of the municipality in parentheses). The dynasty variables are

indicator variables for politicians’ position within their political dynasty (reference category is the first of a

political dynasty and non-dynastic politicians). City Controls refers to a full set of control variables at the

city level: Population (log), Average number of members per family, unemployment rate, elderly index,

population density, school drop-out rate, secondary school rate, province capital, altitude and seaside town. + p<0.1; * p<0.05; ** p<0.01; *** p<0.001.

30

Table 2: Electoral benefits of political dynasties: Election as alderman in first election

All Cut-off=100 Cut-off=50 Region Cut-off=10

Dynasty=2 1.061*** 1.060** 1.069*** 1.083***

(0.014) (0.016) (0.017) (0.020)

Dynasty 3 1.101*** 1.110*** 1.119*** 1.130***

(0.018) (0.020) (0.023) (0.029)

Woman 1.147*** 1.145*** 1.140*** 1.168***

(0.016) (0.018) (0.019) (0.021)

Age 1.014*** 1.014*** 1.013*** 1.013***

(0.000) (0.001) (0.001) (0.001)

Time First Election 1.041*** 1.041*** 1.042*** 1.041***

(0.001) (0.001) (0.001) (0.001)

Education 1.055*** 1.055*** 1.055*** 1.056***

(0.002) (0.002) (0.002) (0.002)

City Controls YES YES YES YES

Region Fixed Effects YES YES YES YES

R2 0.09 0.07 0.09 0.09

N 534,450 432,442 363,316 296,808

Note: Dependent variable is an indicator variable equal to 1 if a politician was selected as alderman after his/her

first election (0 otherwise). Coefficient estimates are represented as odds ratios from a Logit model (robust

standard errors clustered at the level of the municipality in parentheses). The dynasty variables are

indicator variables for politicians’ position within their political dynasty (reference category is the first of a

political dynasty and non-dynastic politicians). City Controls refers to a full set of control variables at the

city level: Population (log), Average number of members per family, unemployment rate, elderly index,

population density, school drop-out rate, secondary school rate, province capital, altitude and seaside town. + p<0.1; * p<0.05; ** p<0.01; *** p<0.001.

31

Table 3: Electoral benefits of political dynasties: Mayor elections 2000-2012

All Cut-off=100 Cut-off=50 Region Cut-

off=10

First

Election

Later

Election

Dynasty=2 1.161*** 1.191*** 1.170*** 1.223*** 1.511*** 1.062+

(0.033) (0.039) (0.041) (0.049) (0.082) (0.035)

Dynasty 3 1.318*** 1.353*** 1.382*** 1.391*** 1.737*** 1.152***

(0.044) (0.051) (0.059) (0.076) (0.098) (0.045)

Woman 0.742*** 0.752*** 0.763*** 0.748*** 0.721*** 0.766***

(0.027) (0.030) (0.033) (0.037) (0.046) (0.033)

Age 0.998* 0.998* 0.997* 0.996* 1.018*** 0.991***

(0.001) (0.001) (0.001) (0.001) (0.002) (0.001)

First Election 0.418*** 0.428*** 0.425*** 0.421*** - -

(0.011) (0.012) (0.013) (0.015)

Education 1.057*** 1.054*** 1.054*** 1.058*** 1.087*** 1.051***

(0.004) (0.004) (0.004) (0.005) (0.008) (0.004)

Year Fixed Effects YES YES YES YES YES YES

City Controls YES YES YES YES YES YES

Region Fixed Effects YES YES YES YES YES YES

R2 0.05 0.05 0.05 0.05 0.05 0.04

N 41,997 33,463 27,966 22,675 11,559 30,467

Note: Estimation sample includes only politicians running for mayor. Dependent variable is an indicator variable

equal to 1 if a politician was elected as mayor at his/her first election (0 otherwise). Coefficient estimates

are represented as odds ratios from a Logit model (robust standard errors clustered at the level of the

municipality in parentheses). The dynasty variables are indicator variables for politicians’ position within

their political dynasty (reference category is the first of a political dynasty and non-dynastic politicians. +

p<0.1; * p<0.05; ** p<0.01; *** p<0.001.

32

Table 4: Political dynasties and elected politicians’ education level

All Cut-off=100 Cut-off=50 Region Cut-off=10

Dynasty=2 -0.325*** -0.351*** -0.364*** -0.352***

(0.020) (0.022) (0.024) (0.026)

Dynasty 3 -0.676*** -0.712*** -0.740*** -0.669***

(0.030) (0.033) (0.035) (0.041)

Woman 0.800*** 0.806*** 0.804*** 0.796***

(0.019) (0.020) (0.022) (0.023)

Age -0.052*** -0.053*** -0.053*** -0.048***

(0.001) (0.001) (0.001) (0.001)

Time First Election -0.000 0.000 0.000 -0.005 *

(0.002) (0.002) (0.002) (0.002)

R2 0.03 0.03 0.03 0.03

N 549,935 445,120 374,010 305,636

Note: Dependent variable is the difference between politicians’ level of education (measured in years) and the

average education level of other politicians in the same city in the year of their first election. Coefficient

estimates derive from OLS models, with robust standard errors clustered at the level of the municipality in

parentheses. The dynasty variables are indicator variables for politicians’ position within their political

dynasty (reference category is the first of a political dynasty and non-dynastic politicians). + p<0.1; *

p<0.05; ** p<0.01; *** p<0.001.

33

Table 5: Political dynasties and elected politicians’ education level: Mayors

All Cut-off=100 Cut-off=50 Region Cut-off=10

Dynasty=2 -0.246** -0.242 * -0.297** -0.151

(0.084) (0.094) (0.102) (0.118)

Dynasty 3 -0.305** -0.321** -0.411*** -0.228

(0.098) (0.111) (0.122) (0.154)

Woman 0.156 0.191+ 0.187 0.109

(0.105) (0.115) (0.124) (0.138)

Age -0.062*** -0.064*** -0.065*** -0.057***

(0.003) (0.004) (0.004) (0.005)

Time First Election -0.008 * -0.006 -0.003 -0.016 **

(0.004) (0.005) (0.005) (0.006)

R2 0.03 0.03 0.03 0.03

N 14,561 11,738 9,903 7,807

Note: Dependent variable is the difference between politicians’ level of education (measured in years) and the

average education level of other politicians in the same city (on the same scale) in the year of their first

election. Coefficient estimates derive from OLS models, with robust standard errors clustered at the level of

the municipality in parentheses. The dynasty variables are indicator variables for politicians’ position

within their political dynasty (reference category is the first of a political dynasty and non-dynastic

politicians). + p<0.1; * p<0.05; ** p<0.01; *** p<0.001.

34

Table 6: Political dynasties and elected politicians’ education level: Aldermen

All Cut-off=100 Cut-off=50 Region Cut-off=10

Dynasty=2 -0.521*** -0.520*** -0.578*** -0.521***

(0.052) (0.058) (0.063) (0.072)

Dynasty 3 -1.028*** -1.061*** -1.058*** -1.092***

(0.064) (0.072) (0.079) (0.099)

Woman 0.886*** 0.932*** 0.927*** 0.893***

(0.047) (0.052) (0.057) (0.061)

Age -0.051*** -0.052*** -0.051*** -0.045***

(0.002) (0.002) (0.003) (0.003)

Time First Election -0.010** -0.011** -0.013** -0.019***

(0.004) (0.004) (0.004) (0.004)

R2 0.04 0.04 0.04 0.03

N 46,419 37,493 31,522 26,308

Note: Dependent variable is the difference between politicians’ level of education (measured in years) and the

average education level of other politicians in the same city (on the same scale) in the year of their first

election. Coefficient estimates derive from OLS models, with robust standard errors clustered at the level of

the municipality in parentheses. The dynasty variables are indicator variables for politicians’ position

within their political dynasty (reference category is the first of a political dynasty and non-dynastic

politicians). + p<0.1; * p<0.05; ** p<0.01; *** p<0.001.

35

Figure 1: Distribution of relative education level Italian local politicians

Note: The figure presents the distribution of the relative education level Italian local politicians over

the period 1985-2012. The x-axis reflects the difference between politicians’ level of education

(measured in years) and the average education level of other politicians in the same municipality

in the year of their first election.

36

Appendix A

Table A1: Descriptive Statistics

Variable Obs. Mean Std. Dev. Min Max Source

Individual Variables

Years of Education 549966 12.57722 4.228997 5 18

Own

calculations

based on

Italian

Ministry of

Interior

Relative Education 549966 0.539404 4.501231 -13 13

Dynasty_position 549966 0.764867 1.156689 0 3

Woman 549966 0.169568 0.375253 0 1

Age 549935 42.61702 11.51286 18 93

Time First Election 549966 1996.371 7.903749 1985 2012

Mayor 549966 0.026478 0.160552 0 1

Alderman 549966 0.084410 0.278003 0 1

Subsample of Mayor Elections

Years of Education 40924 14.64999 3.526249 5 18

Own

calculations

based on

Italian

Ministry of

Interior

Relative Education 40924 2.435001 3.944615 -12.25 13

Dynasty_position 40924 0.709365 1.13141 0 3

Woman 40924 0.119954 0.324912 0 1

Age 40923 45.8158 11.20198 18 91

Year 40924 2006.497 3.515034 2000 2012

First Election 40924 0.281009 0.449497 0 1

Closeness 39987 0.804930 0.170999 0 0.999830

City Variables

Log Population 538160 8.312174 1.513339 3.401197 14.77761

ISTAT (2001);

data are

available at

www.istat.it

Average Number of Individuals per

Family

546888 2.571793 0.306711 1.24 3.88

Unemployment Rate 546888 10.91192 9.29503 0 51.32

Elderly Rate 546418 4.50752 4.042613 0.51 136

Population Density 546888 412.9516 889.8287 1 13323

School Dropout Rate 546888 10.80747 4.17772 0 43.75

Secondary School Degree Rate 546888 26.82322 7.367255 3.85 72.83

Provincial Capital 546026 0.0453202 0.208005 0 1

Altitude 546026 317.0598 283.5223 0 2035

Seaside Town 546026 0.1161337 0.320385 0 1