DISCUSSION PAPER SERIES

IZA DP No. 13086

Jumping the Queue:

Nepotism and Public-Sector Pay

Andri Chassamboulli
Pedro Gomes

MARCH 2020
DISCUSSION PAPER SERIES

IZA DP No. 13086

Jumping the Queue:

Nepotism and Public-Sector Pay

Andri Chassamboulli
University of Cyprus
Pedro Gomes
Birkbeck, University of London, Universidad Carlos III de Madrid, CFM and IZA

MARCH 2020

Any opinions expressed in this paper are those of the author(s) and not those of IZA. Research published in this series may
include views on policy, but IZA takes no institutional policy positions. The IZA research network is committed to the IZA
Guiding Principles of Research Integrity.
The IZA Institute of Labor Economics is an independent economic research institute that conducts research in labor economics
and offers evidence-based policy advice on labor market issues. Supported by the Deutsche Post Foundation, IZA runs the
world’s largest network of economists, whose research aims to provide answers to the global labor market challenges of our
time. Our key objective is to build bridges between academic research, policymakers and society.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper
should account for its provisional character. A revised version may be available directly from the author.

ISSN: 2365-9793

IZA – Institute of Labor Economics

Schaumburg-Lippe-Straße 5–9 Phone: +49-228-3894-0
53113 Bonn, Germany Email: publications@iza.org www.iza.org
IZA DP No. 13086 MARCH 2020

ABSTRACT
Jumping the Queue:
Nepotism and Public-Sector Pay*
We set up a model with search and matching frictions to understand the effects of
employment and wage policies, as well as nepotism in hiring in the public sector, on
unemployment and rent seeking. Conditional on inefficiently high public-sector wages,
more nepotism in public-sector hiring lowers the unemployment rate because it limits the
size of queues for public-sector jobs. Wage and employment policies impose an endogenous
constraint on the number of workers the government can hire through connections.

JEL Classification: E24, J31, J45, J64

Keywords: public-sector employment, nepotism, public-sector wages,
unemployment, queues

Corresponding author:
Pedro Gomes
Department of Economics, Mathematics and Statistics
Birkbeck, University of London
Malet Street
WC1E 7HX London
United Kingdom
E-mail: p.gomes@bbk.ac.uk

* We would like to thank the participants at the Employment in Europe Conference, Aix-en-Province SaM
workshop, University of Essex SaM workshop, University of Kent Macroeconomics workshop, 26th ENSAI Economic
Day workshop, T2M Conference, 1st NuCamp Oxford Conference, Lubramacro conference, SAM annual conference,
and seminars at the Universidad Carlos III, INSPER and FGV São Paulo, for comments and suggestions. In particular,
we would like to thank Chris Pissarides, Andrey Launov, Thepthida Sopraseuth, Christian Merkl, Christian Haefke and
Espen Moen. An earlier version of this paper circulated under the title “Non-Meritocracy and Public Sector Pay”.
1 Introduction
Governments hire workers to produce public goods, but they do not face the same competitive
forces as private firms. As a result, governments use their employment and wage policies to
accomplish a multitude of goals: to attain budgetary targets [Gyourko and Tracy (1989)]; to
implement a macroeconomic stabilization policy [Keynes (1936)]; to redistribute resources
[Alesina et al. (2000)]; or to satisfy interest groups for electoral gains [Gelb et al. (1991)].
This paper builds on the observation that, in several countries, government hiring practices
are sometimes based on nepotism.
We define nepotism as the restriction that some jobs in the public sector are reserved for a
subset of workers that have political or personal connections. By having access to this subset
of jobs, some workers can use their connections to “jump the queue” and find jobs in the
public sector faster. One dimension that is common to all countries is political appointments.
Whenever there is a change in government, there is a subsequent turnover of jobs. The report
Government at a Glance by OECD (2017) highlights the cross-country differences in staff
turnover following a change of government. In countries such as Germany and the UK,
there is little turnover, mainly in advisory posts. In countries such as Greece and Spain,
the turnover extends to layers of senior and middle management. A second dimension is the
influence that politicians or civil servants use to hire friends or family members. Besides
vast anecdotal evidence of such practices, it is also backed by survey evidence.1 In Section
2 we analyse data from the Quality of Government Survey and the European Quality of
Government Index and find that these practices are present in the public sector, more than
in the private sector, and that they vary widely across European countries. In particular,
they are more prevailing in countries where the public-private wage differential is larger.
Our objective is to study the interaction between public-sector policies, nepotism and
unemployment. First, we want to understand the effects of nepotism in the public sector
on unemployment. We find a silver lining to nepotistic hiring. Although it is inefficient
and is absent in the first-best equilibrium, conditional on inefficiently high public-sector
wages, more nepotism lowers the unemployment rate by shortening the queues for these
jobs and increasing employment in the private sector. Second, we want to understand how
employment and wage policies influence incentives to use political and personal connections
to get a job. We show that nepotism only exists if public(-sector) wages are too high and
1
The anecdotal evidence is particularly widespread in Southern European or developing countries, but
not exclusively. The current US president hired his daughter and son-in-law, and a leading French presidential
candidate was found to have put his wife, son and daughter employed on the public payroll. In Spain, the
press exposed that in the “Tribunal de Cuentas”, the institution in charge of invigilating economic and
financial irregularities in the public sector, close to 100 of its 700 workers were family members or friends of
the directors or of important politicians.

2
that it can be restricted if the government sets an efficient wage.
Given the amount of anecdotal and survey evidence of such practices, it is perhaps
surprising that research documenting evidence of nepotism or cronyism in the public sector
is limited. Scoppa (2009) finds that the probability of working in the public sector in Italy
is 44 percent higher for individuals who have a parent also working there. Durante et al.
(2011) find a higher concentration of last names in universities in Italy relative to the overall
population, and that this concentration increased in regions with low civic capital, after
a reform decentralizing the university hiring choices in 1998. Martins (2010) finds that
in Portugal, between 1980 and 2008, over the months preceding an election, appointments
in state-owned firms increased significantly compared to private-sector firms. Hiring also
increased after elections, but only if a new government took office.
On top of these papers that provide suggestive evidence of nepotism and cronyism in
the public sector, two recent papers by Fafchamps and Labonne (2017) and Colonnelli et al.
(2018) have a better identification strategy. Fafchamps and Labonne (2017) find that, fol-
lowing the 2007 and 2010 municipal elections in Philippines, individuals who shared one or
more family names with a local elected official were more likely to be employed in better-
paying occupations, compared to individuals with the loosing candidates’ family names. The
magnitude of the effect is consistent with preferential treatment of relatives as managers in
the public sector. Colonnelli et al. (2018) apply a regression discontinuity design in close
electoral races in Brazil to matched employer-employee data on the universe of public em-
ployees. They find that politically connected individuals enjoy easier access to public-sector
jobs, but are less competent. Despite these empirical efforts to identify nepotism, given the
nature of this activity, it is difficult to empirically measure its aggregate effects.
We study the conditions that allow for nepotism in hiring in the public sector, and its
consequences, from a theoretical angle. In Section 3 we set up a search model in which
workers can search for jobs in either the private or the public sector. Employment and
wages in the private sector are determined through the usual channels of free entry and
Nash bargaining. This ensures that job-finding rates reflect nothing but match surplus so
that identical workers have equal chances of finding a job. In the public sector, by contrast,
employment and wages are exogenous. We account for the possibility of nepotism or cronyism
by assuming that job seekers can use their personal relationships and connections to find a
public-sector job. We assume that prior to entering the labor market, workers can pay a
cost to get “connections” that is drawn from an exogenous distribution across workers. In
our setting, nepotism means that the government reserves some of its jobs for workers with
those connections. Under such practices, in equilibrium, workers with connections can more
easily find public-sector jobs than similar workers that do not have connections.

3
 This paper contributes to the recent labor market search literature that analyzes the role
and effects of public-sector employment and wages. Burdett (2012) includes the public sector
in a job-ladder framework where firms post wages. Bradley et al. (2017) further introduce
on-the-job search and transitions between the two sectors to study the effects of public-sector
policies on the distribution of private-sector wages. Albrecht et al. (2018) consider hetero-
geneous human capital and match specific productivity in a Diamond-Mortensen-Pissarides
model. Michaillat (2014) shows that the crowding-out effect of public employment is lower
during recessions, giving rise to higher government spending multipliers. Navarro and Tejada
(2018) analyse the interaction between public employment and the minimum wage. These
papers’ objective is to determine how employment and wage policies affect private employ-
ment and wages, as well as the unemployment rate. They assume that the unemployed
randomly search across sectors, and, hence, policies affect the equilibrium only by affecting
the outside option of the unemployed and their reservation wage.
Hörner et al. (2007) study the effect of turbulence on unemployment when wages in the
public sector are insulated from this volatility. Quadrini and Trigari (2007) analyze the effects
of exogenous business cycle rules on unemployment volatility. Gomes (2015) emphasizes the
role of the wage policy in achieving the efficient allocation, while Afonso and Gomes (2014)
highlight the interactions between private and public wages. Gomes (2018) examines the
heterogeneity of public-sector workers in terms of education. These papers assume that the
two sectors’s labor markets are segmented, and that the unemployed choose which of the
sectors to search in, depending on the government’s hiring, separation and wage policies.
We add to this literature by considering the choice of finding a public-sector job through
connections and by analyzing how government policies affect this rent-seeking activity. More-
over, while we assume segmented markets, in Section 7 we contrast the transmission mecha-
nism and our results with those from a model with random search. We prefer the assumption
of segmented markets because it portrays a realistic mechanism of selection into the public
sector, documented empirically by Krueger (1988), Nickell and Quintini (2002) or exper-
imentally by Bó et al. (2013), lying at the heart of current policy discussions. High pay
attracts many unemployed to queue for public-sector jobs. Conversely, if pay is too low, few
unemployed search in the public sector, which then faces recruitment problems.
Our first main finding is perhaps surprising. Conditional on inefficiently high wages, more
nepotism in the public sector lowers the unemployment rate. When the value of a public-
sector job is higher than that of a private-sector job (because of either high wages or a low
separation rate), more of the unemployed queue for these jobs, moving away from the private
sector. If most of these jobs are available only through connections, fewer unconnected
unemployed are going to queue and will search in the private sector instead. Although it

4
fosters an inefficient rent-seeking activity, nepotism mitigates the adverse effects that high
public wages have on employment. The evidence from survey data, shown in Section 2,
is consistent with this result. A corollary of this first result, shown in Section 5, is that,
although it entails itself a cost, nepotism might reduce the welfare losses of inefficiently high
public-sector wages.
Although the mechanism is different, this result echoes those found in papers studying
referrals – e.g., Horvath (2014), Galenianos (2014) or Bello and Morchio (2017) – which
have focused exclusively on the private sector. These papers argue that social networks
can improve the matching process by working as an information channel or increasing the
efficiency of search. We argue that hiring through connections works differently in the
public sector. In the private sector, free entry of firms ensures that the gains of alternative
hiring channels translate into job creation, and wage bargaining guarantees that the surplus
generated is shared between firms and workers. On the contrary, we view the public sector as
having a fixed number of jobs that are safer and better paid, which induces workers to find
alternative ways to get them. The mechanism does not involve a better search technology
or better information about vacancies, but the knowledge that some vacancies are reserved
for a subset of workers with connections, which shortens the queues for public-sector jobs.
Focusing on the public rather than the private sector allows us to understand how poli-
cies affect nepotistic hiring. In our setting, the government can hire through connections,
provided that it pays high enough wages to attract enough searchers. In other words, gov-
ernment employment and wage policies impose an endogenous limit on how many workers
it can hire through connections. The constrained-efficient allocation can be achieved with
an optimal wage that simultaneously limits the queues for public-sector jobs and makes it
impossible to hire through connections. This second result is supported by the evidence
from the survey data that non-meritocracy in the public sector is associated with higher
wage premium. It rationalizes why evidence of nepotism in the public sector is common in
Southern European countries, in which public sectors pay substantial premia relative to the
private sector, while it is absent in Nordic countries, which tend to pay a negative premium.
Given the common perception that workers hired through political connections are less
competent, supported by Colonnelli et al. (2018), we consider in Section 7 an extension where
workers have heterogeneous ability. The importance of wage and human resource policies on
the quality of public-sector workers has attracted much interested from an empirical micro
literature, summarized by Finan et al. (2015). It has also been recently studied theoretically
by Geromichalos and Kospentaris (2020). We find that, when the government prioritizes
high-ability workers in the recruitment, low-ability workers face lower changes of getting a
job, so they have stronger incentives to get connections to jump the queue.

5
 In our model we take the government policies as being exogenous in order to isolate the
effects of each of the policies. In Section 6 we provide one possible microeconomic founda-
tion for the government’s policy choices. The government’s employment, wages and use of
nepotism in hiring workers are chosen to maximize an objective function that includes the
production of government services, the preferences of a union, a benefit of nepotism, which
could reflect general corruption or vote buying and a cost of nepotism in terms of possible me-
dia backlash. The simple model of government choices highlights possible interdependencies
of policies and generates the different particular cases that we study.

2 Survey evidence
While the economics literature on nepotism in the public sector is limited, there is a com-
pelling survey evidence suggesting that the hiring practices of the government are non-
meritocratic in several countries. This survey evidence is commonly used in the political
science literature focussing on corruption, such as Charron et al. (2017). We use data from
two of such surveys.

2.1 Quality of Government Survey

The first one is the Quality of Government Survey (QoG). This is a survey of 1294 public
sector experts in 159 countries. They ask experts on the structure and behavior of public
administration, such as, hiring practices, politicization, professionalization, and impartiality.
See Dahlström et al. (2015) for a description of the dataset. We use three questions in a
section on recruitment and careers of public employees. The survey asks the experts whether
when recruiting public-sector workers, the (a) skills and merits of the applicants decide who
gets the job, (b) political connections decide who gets the job, or (c) personal connections
of the applicants (for example kinship or friendship) decide who gets the job. The experts
are asked to rate from 1 (hardly ever) to 7 (almost always).
We focus on 30 European countries. A table with the scores by country is in Appendix F.
The average score for “skills and merits” is 4.9, varying from 2.7 to 6.6. The average scores for
“political” or “personal” connections” are around 3.5, varying from 1.57 to 5.5. As expected,
skills matter in hiring workers in the public sector, but what is perhaps more noteworthy is
that experts consider political and personal connections to be also important in deciding who
gets hired in the public sector. There is, however, a large variation in recruitment practices.
In seven countries - Italy, Portugal, Cyprus, Bulgaria, Hungary, Romania and Slovakia - the
score for “skills and merit” is lower than both other scores. The 8 countries where the score

6
 Figure 1: Recruitment practices and the public-private wage ratio
Skills and merit Political connections Personal connections
7

6
5
6

5
Index (1 to 7)

Index (1 to 7)

Index (1 to 7)
4
5

4
3
4

3
2
3

2
1
1 1.5 2 2.5 1 1.5 2 2.5 1 1.5 2 2.5
Aggregate Public−private wage ratio Aggregate Public−private wage ratio Aggregate Public−private wage ratio

(a) (b) (c)

Source: Indexes of recruitment practices are takes from the Quality of Government Survey. Data on gov-
ernment and private sector employment is from EUROSTAT and OECD. Data on government wage bill and
private sector wage bill is from AMECO. See data in Appendix F.

of skills and merits is highest includes the Nordic countries (Denmark, Finland, Sweden and
Norway) plus Luxembourg, Switzerland, Netherlands and Ireland. In those countries, the
average index for political or personal connections is lower than 2.5.2
These differences in the role of political or personal connections are related to public-
sector wages. We use aggregate data to calculate an average public-private wage ratio.
Using OECD and AMECO data, we calculate the government’s wage bill over the size of
government employment relative to the private-sector wage bill over the size of private-sector
employment. Figure 1 shows the relation between the three indexes of recruitment practices
and the public-private wage ratio. Higher average wages in the public sector is associated
with recruitment practices less based on merit (a correlation coefficient of -0.4) and more
based on political and personal connections (correlation of 0.4).

2.2 European Quality of Government Index

The second survey is based in an EU regional level governance survey, used to construct the
European Quality of Government Index (EQI). The survey was first ran in 2010 and then
repeated in 2013 and 2017. The index focusses on both perceptions and experiences with
public-sector corruption, along with the extent to which citizens believe various public-sector
services are well allocated and of good quality. See Charron et al. (2014).
An advantage of this survey is the more disaggregated level of information at a regional
2
While there is a substantial variation in recruitment practices across European countries, there are
even larger variation across the other 129 countries. As shown in Appendix F, only in Western Europe and
North America, East Asia and The Caribbean the score for “skills and merits” is higher than “personal
connections”. In the remaining 100 countries, personal and political connections matter more than skills
and merits when recruiting public-sector workers. Non-meritocratic practices seem to be more widespread
Sub-Saharan Africa, South Asia and Latin America with average scores above 4.5.

7
level - NUTS 1 and 2 - albeit for only 21 countries. The disadvantage is the absence of a
specific question about recruitment. Instead, the survey asks a more general question on
whether workers in the public sector can succeed, varying from 1 (“most people can succeed
if they are willing to work hard”) to 10 (“Hard work is no guarantee of success – it’s more
a matter of luck and connections”). Interestingly, it also asks the same question about the
private sector where the score also varies between 1 to 10.
The average score at country level is 5.6 for the private sector and 6.4 for the public
sector, suggesting non-meritocracy is a more relevant problem there. The six countries
with lower score for the public sector (more meritocratic) are Austria, United Kingdom,
Germany, Denmark, Finland and Sweden. The six countries with higher score for the public
sector (less meritocratic) are Bulgaria, Greece, Croatia, Slovakia, Romania and Portugal.
The correlation between the scores of the public and private sector is high (0.8), suggesting
the behavior in the two sectors go in parallel. As such, we create a new relative index of
non-meritocracy, which is the ratio of the score of the public relative to the private sector.
We correlate this index with a measure of public-sector wage premium and of unemploy-
ment rate. The wage premium at a regional level is estimated using microdata from the 2010
Structure of Earnings Survey. Relative to the aggregate measures of public-sector premium,
these regressions allow us to control for worker’s characteristics. Following the literature, i.e.
Katz and Krueger (1991), Disney and Gosling (1998) or Christofides and Michael (2013), we
regress the log of the gross hourly wage on a gender dummy, region, age, education, occupa-
tion and a part-time dummy. To measure the premium, we include a public-sector dummy
interacted with the available region. Some small countries do not have the regional (NUTS
1) identifiers, while for other countries the NUTS 1 are aggregated into larger regions. We

Figure 2: Non-meritocracy in the public sector relative to the private sector

Relation with public−sector wages Relation with unemployment rate
1.4

1.4
1.3
1.3
Ratio of indexes

Ratio of indexes
1.2
1.2

1.1
1.1

1
1

−.1 0 .1 .2 .3 5 10 15 20 25 30
Public−private wage premium Unemployment rate, 2010

(a) (b)
Source: The y-axis has the ratio of the index for the public over the index for the private. A number larger
than 1 means the public sector is perceived to be less meritocratic than the private sector. Both indexes are
taken from European Quality of Government Index dataset. The public-sector wage premium is estimated
with microdata from the 2010 Structure of Earnings Survey. Unemployment rate is taken from Eurostat.

8
end up with 70 observations.
The first graph in Figure 2 shows that the positive association between non-meritocracy
and wages is also present at a regional level. Notice that the non-meritocracy index is larger
than 1 in all but two observations, meaning that it is perceived as more widespread in the
public sector. The second graph shows a negative association between non-meritocracy in
the public sector and unemployment that is predicted with our model. We claim that hiring
based on connections limits the negative effect on unemployment by reducing the queues for
public-sector jobs, particularly when they offer high wages.
We show this association more rigorously in Table 1. Columns (1) and (2) reflect the
association shown in Figure 2. When regressing our index on both variables, both are
statistically significant at 1 percent, as shown in column (3). In column (4) we interact
unemployment rate with dummies for countries above and below the median public-sector
wage premium. The negative relation with unemployment rate is only present in countries
with a high premium. In columns (5) and (6) we calculate our index in a different way. In
column (5) we calculate the index in differences of public and private sector scores, rather
than in ratio. In column (6) we calculate the public-sector score based only on respondents
working in the public sector and the private sector score based on respondents working in the
private sector only, before computing the ratio. In both alternatives we have similar results.
In Appendix G, we show the regressions of the unemployment rate on public wages and the

Table 1: Regression of the ratio of indexes of non-meritocracy

Baseline variables Alternative variables
(1) (2) (3) (4) (5) (6)
Public-sector wage premium 0.265** 0.384*** 0.723*** 3.484*** 0.459**
(2.62) (3.74) (4.48) (3.81) (2.22)
Unemployment rate -0.003* -0.006***
(-1.71) (-3.12)
× High public wage -0.009*** -0.038*** -0.010***
(-4.16) (-2.98) (-3.51)
× Low public wage -0.002 0.000 -0.006*
(-0.95) (0.02) (-1.86)
Observations 70 70 70 70 70 70
R-squared 0.09 0.041 0.207 0.283 0.194 0.169
Notes: The t-statistics are shown in brackets.*** indicates significance at the 1% level, ** at 5% level, and * at the
10% level. The dependent variable is the ratio of the non-meritocracy index for the public sector over the index for
the private sector. It increases when the public sector is perceived to be less meritocratic than the private sector.
The index is constructed with data taken from European Quality of Government Index dataset. The public-sector
wage premium is estimated with microdata from the 2010 Structure of Earnings Survey. Unemployment rate is
taken from Eurostat. In column (5) we use an alternative index which is the difference between the index for the
public over the index for the private. In column (6) we use an alternative index which is the ratio between the
index for the public sector (answer by only public-sector workers) over the index for the private sector (answered
by only private-sector workers).

9
index of non-meritocracy. Mirroring these results, unemployment is negatively associated
with non-meritocracy in the public-sector, particularly when its wages are high.
The evidence in this section, based on survey data, finds a positive association between
public-sector wages and nepotism and a negative association with between nepotism in the
public sector and the unemployment rate, stronger in regions with a higher premium. Clearly,
the association between these variables can have several explanations. Given the problems
to design an empirical strategy that identifies nepotism in the public sector and its effects,
we develop a model that provides one interpretation of these associations.

3 Model with nepotism in the public sector

3.1 Preliminary considerations
The defining characteristic of the public sector is that it does not sell its goods or services
- it supplies them directly to the population. There is no market price. Governments
finance employment, not by selling goods, but by using the power of taxation. As such, the
public sector does not have shareholders and it does not maximize profits. The decisions
regarding employment reflect different government objectives. Even in determining wages
(or wage growth) there is a discretionary component that can create widely documented
wage differentials vis-à-vis the private sector. As such, the usual mechanisms that drive the
private sector adjustments studied by economists do not map into the public sector.
Our modeling choices reflect this view. We discuss two particular assumptions. As in
Bradley et al. (2017) or Albrecht et al. (2018), we assume that the public-sector wage is
exogenous. We view it not as an equilibrium outcome (i.e. private wages, which may reflect
match productivity and outside option) but a policy variable (i.e. unemployment benefits or
government spending, which may reflect various objectives). Notice that public wages are
a payment in units of the private good (financed with taxation), not in units of the public
goods, hence they are not necessarily associated to the productivity of the worker. Wage
and employment policies might be influenced by several factors, such as unions, inequality
or elections. In our model we take the government’s choices of wages and employment as
given, in order to characterize the labour market effects of changes in policies. The only role
of the government is to maintain its employment constant by hiring in enough workers to
replace those that separate. Still, in Section 6 we provide one possible microfoundation for
the government’s choices, which helps understand our modeling choices.
The second assumption is that we consider homogeneous workers in terms of education,
ability and productivity. Given the role of the public sector as a large-scale employer, our

10
focus is to study the effects of nepotism and wages on the labour market and unemployment,
which we believe are of first-order importance. Still, we study the common argument that
workers that get jobs though connections are of worse quality in subsection 7.2, where we
extend the model to include workers of heterogeneous ability.

3.2 General setup

We consider a search and matching model with firms and a public sector. Workers can be
either employed and producing or unemployed and searching for a job. Each firm is endowed
with a single vacancy that can be vacant or filled (job). At each instant, τ individuals are
born (enter the labor market) and die (retire) so that the working population is constant
and normalized to unity. All agents are risk-neutral and discount the future at rate r > 0,
and time is continuous.
All individuals, prior to entering the labor market, can obtain connections by paying a
cost c. The cost is distributed across individuals according to the cumulative distribution
function Ξ(·) on [0, c̄]. If a family member works in the public sector, the cost of connections
is low. If getting connections requires the affiliation with a political party, it is more costly.
Some jobs in the public sector are reserved for workers with connections. By obtaining
connections workers can gain access to these “connected” jobs and thus have priority – a
higher job-finding rate – for public-sector jobs.
An endogenous proportion of the population (those whose connection cost is sufficiently
low) become “connected”. For a connected individual, using his/her connections to find
a job in the public sector job strictly dominates all other options. But, if an individual is
unconnected, then she has a further decision of whether to search for jobs in the private or in
the public sector through standard search.3 The two sectors are segmented. In Section 7.1,
we consider the case in which workers without connections search randomly for jobs in both
sectors. Figure 3 depicts these choices. In total, there are three active markets: the private
sector and the two public-sector submarkets, one for connected and one for unconnected
workers. Variables are indexed by the superscript x = [g, p], where g refers to the public
(government) sector and p to the private sector, and the subscript j = [c, u], where c refers
to connected and u to unconnected. A searching (unemployed) worker receives a flow of
income b, which can be considered the opportunity cost of employment.
3
Throughout the paper, we use the terms “connected jobs/vacancies” to refer to the jobs that the
government reserves for job seekers with connections. We use the term “unconnected jobs/vacancies” to
refer to the remaining government jobs that are filled by workers without connections. We use the terms
“connections sector” and “no-connections sector” to refer to these two public sub-sectors.

11
 Figure 3: Decision of newborn

Cost Type Sub-market

c Connected Public sector Ucg

Newborn
Public sector Uug
Unconnected
Private sector Uup

3.3 The Private sector

The private and public sectors differ in two aspects: hiring practices and wage-setting. The
rate at which workers are hired into firms is endogenous. In particular, firms open vacancies
and search for suitable workers until all rents are exhausted. The rate at which workers find
p
private-sector jobs depends positively on the tightness, θ = uv p , where v p is the measure of
private-sector vacancies, and up is the number of unemployed searching for those vacancies.
They are hired at Poisson rate m(θ), while firms fill vacancies at rate q(θ) = m(θ)θ
.
The output of a match between a worker and a firm is y (we consider worker heterogeneity
in subsection 7.2). Wages in the private sector, denoted as wp , are determined by Nash
bargaining, such that the worker gets a share β of match surplus. With higher match
surplus, firms expect larger profits from creating jobs; firm entry is higher; and workers can
more easily find jobs and earn higher wages. Hence, the private-sector hiring and wage-
setting procedures are, in a sense, meritocratic. Individuals have equal chances of finding
jobs.
A vacant firm bears a recruitment cost κ, related to the expenses of keeping a vacancy
open and looking for a worker. When a vacancy and a worker are matched, they bargain
over the division of the produced surplus. The surplus that results from a match is known
to both parties. After an agreement has been reached, production commences immediately.
Matches in the private sector will dissolve at the rate sp . Following a job destruction, the
worker and the vacancy enter the market and search for a new match.

3.4 Government
In the public sector, by contrast, policies are taken to be exogenous. To produce some
services, the government employs an exogenous number of workers. In each period, it has to
hire enough workers to replace the workers that exogenously separate or retire. That means

12
hiring (sg + τ )eg workers, where sg is the separation rate. A fraction µ of jobs are reserved
for workers who have connections.
The matching function in the public sector is Mjg = min{vjg , ugj }. To maintain its employ-
ment level, the government must be able to attract a number of searchers in each segment,
ugj , at least equal to the number of job openings, vjg , meaning that Mjg = vjg . Otherwise
public-sector services break down. As we show in Lemma 2, this imposes a condition on
public wages to be high enough to attract at least the same number of searchers as of vacan-
cies. We choose this particular functional form for the matching technology for simplicity
and clarity. First, it makes the concept of queues in the public sector clearer. When there
are more unemployed than vacancies, the vacancy filling rate for the government is 1, and
all the unemployed in excess are queuing. As we will show, this makes the efficient wage a
very clear and intuitive object, easy to calculate. Second, such assumption has been used
in other papers, i.e. Quadrini and Trigari (2007) and there is evidence that the elasticity of
matches with respect to unemployed is much lower in the public sector than in the private
(Gomes (2015)). This does not mean that there are no matching frictions, only that they
are one-sided. Nothing substantial would change in the model if the matching function in
the public sector were Cobb Douglas: Mjg = (vjg )η (ugj )1−η . In this case, the vacancy filling
probability of the government would no longer be 1, and it would need to set the vacancies
endogenously such that the total number of matches would equate exactly the number of

1 1−η
workers that retire or separate. Solving for vjg we would obtain v g = Mjg η /(ugj ) η , but the
job-finding of unemployed would be the same. Hence, our main results, detailed in Section
4, still go through if the matching technologies are identical across the two sectors.
We assume that the recruitment is part of the role of the government and is done by its
workforce. Since the government’s objective is to maintain employment level (eg ) by hiring
enough workers to replace those that separate or retire, it follows that vug = (1 − µ)(sg + τ )eg
and vcg = µ(sg + τ )eg . Connected and unconnected workers find public-sector jobs at rate
g )eg g +τ )eg
mgc = µ(s u+τg
c
and mgu = (1−µ)(s
ugu
, respectively.4 For the moment, we set µ = µ̄, where µ̄
is an exogenous parameter reflecting the target fraction of jobs the government aims to fill
through connections. In Section 4.2, we analyze the case in which the government cannot
reach its target because there are not enough workers with connections.
As will become clearer below, because public employment is exogenous, the productivity
4
We considered a less segmented way to model the connected and unconnected market. In particular, we
considered a setting with only one public-sector market in which both connected and unconnected workers
were searching randomly for jobs. Connected workers have higher efficiency of search (which would be the
exogenous variable reflecting nepotism) and hence a higher job-finding rate. In such setting, the composition
of the public employment of connected and unconnected workers would be endogenous. All the results and
intuition would be similar to our baseline setting, but at a cost of more mathematical complexity.

13
of workers in the public sector is not important for our results. We assume that the separation
rates, as well as other labor market friction parameters, are exogenous. Notice that in this
setting, where the government has a fixed employment level, the separation rates sg play a
double role: they reflect the expected duration of the match but also determine the number
of new hires. Lower separations increase the value of employment but, at the same time,
reduce the vacancies and make an unemployed worker less likely to find a job there. Finally,
the public wage, wg is the other exogenous policy variable. We ignore the issue of how the
government finances its wage bill and assume that it can tax its citizens in a non-distortionary
lump-sum tax.

3.5 Value functions, Free entry, Wages

Let Uup and Eup be the values (expected discounted lifetime incomes) associated with unem-
ployment (searching for a job) and employment in the private sector, defined by:

(r + τ )Uup = b + m(θ) [Eup − Uup ] , (1)

(r + τ )Eup = wp − sp [Eup − Uup ] . (2)

The values associated with unemployment in the public sector with and without connec-
tions are given, respectively, by:

(r + τ )Uug = b + mgu [Eug − Uug ] , (3)

(r + τ )Ucg = b + mgc [Ecg − Ucg ] . (4)

We assume the wage in the public sector does not depend on connections. In Appendix F
we consider a case where workers with connections also have a wage premium. Despite equal
wages, the values of being employed are different for workers with and without connections:

(r + τ )Eug = wg − sg [Eug − Uug ] , (5)

(r + τ )Ecg = wg − sg [Ecg − Ucg ] . (6)

On the firm’s side, let Jup be the value associated with a job and Vup be the value associated
with posting a vacancy and searching for a worker to fill it, given by:

rJup = y − wp − (sp + τ ) [Jup − Vup ] , (7)

rVup = −κ + q(θ) [Jup − Vup ] . (8)

14
 In equilibrium, free entry drives the value of a private vacancy to zero:

Vup = 0. (9)

Wages are determined by Nash bargaining between the matched firm and worker. Their
outside options are the value of a vacancy and the value of being unemployed. Let Sup ≡
Jup − Vup + Eup − Uup denote the surplus of a match. With Nash bargaining, the wage wp is
set to a level such that the worker gets a share β of the surplus, and the share (1 − β) goes
to the firm. This implies two equilibrium conditions:

βSup = Eup − Uup (1 − β)Sup = Jup − Vup . (10)

Setting Vup = 0 in (8) and imposing the Nash bargaining condition in (10) gives:

κ
= (1 − β)Sup . (11)
q(θ)

Using (1)-(7) together with (10) and the free-entry condition Vup = 0, we can write:

y−b
Sup = , (12)
r + τ + sp + βm(θ)

and the free-entry condition as

κ (y − b)(1 − β)
= . (13)
q(θ) r + τ + sp + βm(θ)

This job-creation condition sets the expected costs of having a vacancy equal to the expected
gain from a job. It can be used to determine the equilibrium market tightness θ and, in turn,
the rates at which workers find jobs in the private sector, m(θ). Imposing the free-entry
condition (11) for private-sector vacancy creation, the Nash bargaining solution implies that

wp = b + β(y − b + κθ). (14)

Lemma 1 Tightness and wages in the private sector are independent of the government
employment and wage policies (eg , wg , sg and µ).

This lemma is a useful intermediate result and follows directly from equations (13) and
(14). Government employment and wage policies do not affect wages and tightness in the
private sector. It implies that they affect the equilibrium only by affecting the connections
decision of the newborn or the scale of the private sector through the number of unemployed

15
directing their search towards the private sector. Given a constant tightness, policies that
make the public sector more attractive will drain workers from the private sector and reduce,
one-to-one, the number of vacancies, leaving private wages unchanged.

3.6 Newborn‘s Decisions

We can summarize the three options of the newborn as

m(θ)
(r + τ )Uup = b + [wp − b], (15)
r + τ + sp + m(θ)
mgu
(r + τ )Uug = b+ [wg − b], (16)
r + τ + sg + mgu
mgc
(r + τ )Ucg = b+ [wg − b]. (17)
r + τ + sg + mgc

These three options were depicted in Figure 3. Workers without connections can search in
either the public or the private sector. In equilibrium, the two values have to equate:

Uu = Uug = Uup . (18)

This condition determines the number of unconnected searchers in the public sector, ugu ,
which is the variable that compensates any asymmetry in the value of the job in the two
sectors. An increase of the value of a public-sector job, Eug , (driven by either higher wages or
lower separations) raises the number of unemployed searching for openings and lowers their
job-finding probability (mgu ), such that its effect on Uug is neutralized.
Alternatively, workers can use connections to find jobs only in the public sector. In what
follows, we drop the superscript g in Ucg and set Uc ≡ Ucg . The newborn chooses the option
that, given her c, gives the highest value between:

M ax{Uu , Uc − c}. (19)

A worker with a cost c chooses to obtain connections only if the benefit, Uc − Uu , exceeds
the cost, that is, only if c ≤ Uc − Uu . The threshold level of c at which a worker is indifferent
between using and not using connections to find a job is, therefore, given by:

c̃ = Uc − Uu . (20)

Lemma 2 There exists a public-sector unconnected market with employment level eg , pro-
vided that it pays a sufficiently high wage wg ≥ wgu . There exists a public sector of size eg with
a connected market where µ = µ̄, provided that it pays a sufficiently high wage wg ≥ wgc > wgu

16
 Figure 4: Lemma 2: the role of public-sector wage

wg
Public sector breakdown wgu Connections sector constrained wgc Connections sector µ̄ exists

The exact expressions for wgu and wgc are in Appendix A. This lemma, depicted graphically
in Figure 4, states that the public sector needs to pay a sufficiently high wage in order to
attract enough job seekers to fill its vacancies and maintain a constant employment level.5 If
the wage is above this threshold, wgu , some unemployed will prefer to queue for public-sector
jobs. This threshold depends positively on private-sector wages, wp , and unemployment
benefits, b. However, for the government to be able to fill µ̄ of its vacancies with workers
that have connections, this wage has to be higher. The wage paid to a public employees
is independent of how he/she was hired (with or without connections). Nevertheless, the
benefit from using connections to jump the queue is larger when wages are higher, because
then, more workers are searching for public-sector jobs and getting one of them without
connections takes much longer. For the government to be able to attract enough workers
with connections to fill µ̄ of its vacancies with such workers, the wage must be high enough,
so that the benefit from having connections compensates the costs of acquiring them. This
second threshold wage, wgc , depends positively on wgu and on the size of the connections
sector µ̄. In what follows, we assume that the wages are always above wgc , meaning that
the government can fill any target fraction µ̄ of its vacancies through connections. Note,
however, that if the wage is lower but still above wgu , the government is able to attract some
connected job searchers, and fill some of its vacancies through connections, but not enough
to fill its target fraction µ̄. We analyze this case in Section 4.2.

Lemma 3 If a connections sector exists (c̃ > 0), the job-finding rate in the connections
sector is higher than in the unconnected sector (mgc > mgu ).

This lemma follows directly from equations (18) and (20). They imply that the value of
searching for a job in the public sector is higher for connected than for unconnected workers
which, given that wages and separation rates are the same for both types of workers, can
only be achieved with a smaller queue for connected workers.
5
Notice that what matters for the existence of a public sector unconnected and connected market is not
the size of the wage per se, but the size of the surplus that jobs in the public sector generate to the worker.
While the surplus increases with the wage, a higher public-sector surplus could also reflect higher job security
and other fridge benefits.

17
3.7 Equilibrium Allocations
Workers’ cutoff c̃ determines their selection into two groups: those who use connections to
find public-sector jobs (Lgc ) and those who do not have connections (Lu ). We can measure
each of these two groups’ share in the labor force as:

Lgc = Ξ(c̃) (21)

Lu = 1 − Ξ(c̃) (22)

Among the workers who do not have connections, some will be attached to the private
sector (Lpu ) and some to the public sector (Lgu ). Hence, Lu = Lpu + Lgu .
Using (10)-(13) and (15)-(17), we can write the cutoff as:

µ(sg +τ )eg
" #
1 ugc βκθ
c̃ = g )eg
[wg − b] − . (23)
r+τ r+τ + sg + µ(s u+τ g
c
(1 − β)

Definition 1 A steady-state equilibrium consists of a cut-off cost {c̃}, private sector tight-
ness {θ}, and unemployed searching in each market {up , ugc , ugu }, such that, given some ex-
ogenous government policies {wg , eg , µ̄}, the following apply.

1. Private-sector firms satisfy the free-entry condition (13).

2. Private-sector wages are the outcome of Nash Bargaining (14).

3. Newborns decide optimally their investments in connections (equation 19), and the
population shares are determined by equations (21)-(22).

4. The search between the public and private sectors by the unconnected unemployed sat-
isfies equation (18).

5. Flows between private employment and unemployment are constant:

(sp + τ )ep = m(θ)up . (24)

6. Population add-up constraints are satisfied:

Lgu = (1 − µ)eg + ugu , (25)

Lgc = µeg + ugc , (26)
Lpu + Lgu + Lgc = 1 (27)

18
 7. The government fills its target fraction of vacancies through connections µ = µ̄, that
is, wg ≥ wgc .

4 Main results
This section details the main results, under three propositions. All the derivations and proofs
are shown in Appendix A, including the proof that the equilibrium exists and is unique.

4.1 Nepotism, public-sector wages and unemployment

Proposition 1 An increase in µ̄ decreases the number of workers searching for public-sector
jobs (decreases ug = ugu + ugc ), increases the number of workers in the private sector (i.e.,
increases Lpu = 1 − Lgu − Lgc ) and increases the employment rate.

This result is perhaps surprising but is quite logical and consistent with evidence in Figure
2. As shown in Lemma 2, the existence of a connections sector requires that the public-sector
wage is high enough. Under this condition, there are large queues of unconnected workers
for public-sector jobs. With a higher fraction of these jobs being reserved for workers with
connections, the value of searching for one without connections decreases. Workers have
more incentive to direct their search towards the private sector or to obtain connections.
Since it is costly to obtain connections, some of them – those whose connection cost is high –
abandon search in the public sector and search in the private sector instead. With a constant
tightness in the private sector, job creation goes up one-to-one as the number of searchers
and overall employment increases.

Proposition 2 An increase in wg increases the number of workers searching for public-

sector jobs (increases ug = ugu + ugc ) decreases the number of workers in the private sector
(i.e., decreases Lpu = 1 − Lgu − Lgc ) and decreases the employment rate. These negative effects
are smaller when µ̄ > 0 than when µ̄ = 0.

A higher wage in the public sector makes the value of searching for a job there higher
and shifts workers away from the private sector, thereby lowering the employment rate.
When a fraction of jobs in the public sector are reserved for workers with connections, the
number of unconnected workers that queue for public-sector jobs is smaller. Some choose
to use connections in order to get in. But because obtaining connections is costly, the total
increase in the number of workers queuing up for public-sector jobs is smaller. The number
of workers that abandon search in the private sector in response to the increase in wg is

19
therefore smaller. Hence, the recruitment through connections mitigates the negative effects
of more generous compensation policies on employment. This proposition is consistent with
evidence from Table 1.

4.2 When nepotism is bounded: a limit to µ

We now relax the assumption that µ is isolated from labor market conditions. We show that
in situations in which the wage premium is large enough to maintain public employment,
but not large enough to generate queues of connected jobs searchers, changes in wages can
influence the size of the connected sector.
We interpreted µ̄ as the government’s target fraction of vacancies to be filled through
connections. The government is able to meet its target – fill a fraction µ = µ̄ of jobs through
connections – if it pays a sufficiently high wage. According to Lemma 2, there exists a
wage, wgc , at which the government is able to attract exactly ugc = µ̄(sg + τ )eg connected job
searchers. Hence, for any wage wg ≥ wgc , the government is able to attract an even larger
number of connected job searchers so that ugc ≥ µ̄(sg + τ )eg . Consequently, some of the
connected searchers also queue up waiting for jobs.
If the government wage is lower, i.e. wgc > wg > wgu , the number of connected job
searchers, ugc , is lower, but still positive: 0 < ugc < µ̄(sg +τ )eg . In this case also, a connections
sector exists, but the government is restricted to fill only a fraction µ < µ̄ of vacancies
through connections, where µ is such that ugc = µ(sg + τ )eg and there are no connected
workers queuing for jobs. The remaining vacancies (1 − µ) are filled by unconnected workers.
Using (26), we can solve for µ and write:

Lgc
µ= . (28)
eg (sg + τ + 1)

This equation states that the total number of connected workers Lgc equals the new hires
µ(sg + τ )eg plus those already employed in the public sector µeg and no connected worker is
left searching for a job.
In the limiting case, where wg = wgu , no worker has the incentive to use connections to
find a public-sector job; hence, ugc = 0, which means that µ = 0. To sum up, we generalise
Condition 7 in Definition 1, by replacing it with



 µ̄ if wg ≥ wgc

Lgc
µ = e g (sg +τ +1) if wgc > wg > wgu (29)


if wg = wgu .

0

20
Proposition 3 Provided that the public wage is high enough to attract some connected job
searchers, but not high enough to generate queues of connected job searchers i.e. wgc > wg >
wgu , the fraction of vacancies that the government fills through connections, µ, is smaller,
the smaller the public wage wg and the larger the size of public employment, eg . If wg = wgu ,
there is no nepotism (µ = 0).

The government can fill a higher fraction of jobs through connections when the public
wage is higher because the supply of connected job searchers is larger. Larger public employ-
ment means that the number of workers that the government needs to hire each period, to
replace those that separate due to retirement or other reasons, is also larger, while the num-
ber of connected workers searching for jobs is smaller. Hence, the proportion of government
jobs filled by connected job searchers is smaller.
This proposition tells us how government policies place a constraint on the level of nepo-
tism and is consistent with evidence in Figure 1. Governments that have large employment
levels but offer low premia to their workers – such as those in Nordic countries – will have
endogenous limits on hiring through connections.

5 Efficiency
5.1 Efficient allocation
The social planner’s problem and the first-order conditions are shown in Appendix B. There
are three types of inefficiencies in this model: i) the existence of a connections sector that
propels newborns to take on rent-seeking activities; ii) the existence of queues for public-
sector jobs; and iii) the usual thick-market and congestion externalities.
Inefficiencies i) and ii) are both solved by setting the optimal wage. To avoid queues
and given the assumption of the min matching function in the public sector, the government
should set a public-sector wage such that ugu = vug = (sg + τ )eg . In other words, at any
instant both the job-finding rate for government jobs and its vacancy-filling rate should be
1, which implies setting wgu . This same wage, according to equation (29), eliminates the
connections sector (ugc = vcg = 0). This shows that the connections sector is inefficient only
when the public wage is set optimally.
We then show that the inefficiency iii) is solved with the Hosios condition. The Hosios
condition in private-sector bargaining guarantees that the thick market and the congestion
externalities are internalized.

21
5.2 Optimal µ conditional on inefficient public-sector wage
Suppose, now, that the government sets a high enough wage to fill its target fraction µ̄ of
vacancies through connections; that is, wg > wgc . In this case, a connections sector exists, as
some workers find it optimal to use connections. The question that arises is whether or not
the existence of a connections sector, under inefficient government policies, improves welfare.
To address this question, we discuss the impact of increasing µ̄ (µ = µ̄) on net surplus. Net
surplus is total private output net of vacancy posting costs, plus unemployment income,
minus the resources spent in connections. Since public employment is fixed, an increase in
total output can be achieved by an increase in private employment.
As summarized in Proposition 1, an increase in µ̄ raises employment in the private sector
and, thus, increases output and net surplus. However, we cannot conclude that a larger
connections sector means higher net surplus overall, because an increase in µ̄ also induces
some workers to use connections, thus increasing the total resources wasted. If obtaining
connections is difficult and costly for most workers, relative to the benefit of being employed
in the public sector, then an increase in µ̄ is more likely to drive workers away from the
public sector and cause a large shift in workers’ search towards the private sector, resulting
in a large increase in private employment, but naturally also a larger waste of resources with
connection costs. If, on the other hand, obtaining connections is easy and the benefit of a
public-sector job large, then an increase in µ̄ will have a small impact on private employment
and will, instead, cause a larger shift towards forming connections.
We cannot establish that an increase of µ̄ is optimal, given an inefficient wage policy.
As discussed above, the connections costs, the size of public wages, and other benefits are
important. However, the interesting point here is that we cannot rule out that nepotism in
the public sector can increase welfare when wages are inefficient, because it raises output
production and shortens public-sector queues.

6 A microfoundation of public-sector policies

In the preceding analysis we considered the effects of changes in government policies with-
out taking a stance on how governments decide on these policies. We also studied three
different cases: (i) the case in which the government sets the efficient wage and there is no
hiring through connections (µ = 0); (ii) the case in which the government targets to fill a
certain fraction of jobs through connections (µ̄) and faces no restrictions in achieving this
target (µ = µ̄); and (iii) the case in which the government cannot achieve its target and fills a
smaller fraction of jobs through connections (µ < µ̄). We now provide one possible microeco-

22
nomic foundation for the public-sector policies, which highlights possible interdependencies
of policies and can generate the different particular cases analysed so far.
Consider a government that is limited in its amount of spending to ω̄, exogenous, that
arises from budgetary constraints. The government has an objective function with three
components. The first, log(eg ) is the preference for government services that are produced
using its workforce. The second, is the preferences of a union represented by ς log(a). Here ς
represents the weight of the union in the government preferences and a is the extra payment
to public-sector workers on top of the minimum required wage for the existence of the public
sector (wg = wgu +a). The union knows what the minimum required wage is and tries to push
for wages above it. The third element, ϕ log(egc ) − ϑegc , reflects nepotism and has two parts.
ϕ represents the weight attributed to hiring connected workers, egc , that could reflect general
corruption, cronyism or vote buying. In other words, the government can offer jobs in order
to favor certain groups, gain influence or buy votes. ϑ represents the cost of nepotism for
the government, for instance the public backlash when cases are denounced by the media.
Stronger media in the country should raise the cost of such practices, i.e. raise ϑ.
The government’s problem can be written as:

max log(eg ) + ς log(a) + ϕ log(egc ) − ϑegc

eg ,egc ,a

s.t.
(wgu + a)eg = ω̄,
χegc ≤ a

where χegc ≤ a is the restriction that the wage is high enough for a connections sector to
exist. It is basically the restriction wg ≥ wgc and is derived using the expression wgc =
wgu + Ξc,−1 (µ(sg + τ )eg )(r + τ + sg + 1), in Appendix A. Assuming that the distribution of
connections is uniform we get a linear relation between a and the number of connected public
sector workers, represented by the parameter χ. The three first-order conditions determining
government policies are given by:
1
= Λ1 wg , (30)
eg
ς
= Λ1 eg − Λ2 , (31)
a
ϕ
− ϑ = Λ2 χ, (32)
egc
plus the complementary-slackness condition:

Λ2 (a − χegc ) = 0, (33)

23
where Λ1 and Λ2 are the multipliers in both constraints. These first-order conditions show
the possible interdependence between the government policies.
We are going to distinguish the three cases that mimic the special cases discussed in
the paper. In the absent of unions or vote buying (ϕ = ς = 0), the government sets
the minimum wage that would guarantee hiring enough workers and maximize government
production. This would be the efficient solution discussed in Section 5.1, where there are no
distortions in the labour market and the government maximizes the provision of its services,
given its budget constraint. This is the outcome of a benevolent government that never hires
through connections.
Consider a second case where there is no intrinsic cost of nepotism, ϑ = 0, i.e. the gov-
ernment has a tight control over the media. In such case, the second constraint always holds
with equality, generating the constrained case in Section 4.2. In this case the government
wants to use connections at the maximum. In other words, since there is no cost of nepotism,
it wants to set µ = µ̄ = 1, but budgetary constraints prevent the government from doing so.
The government cannot set the wage high enough to attract enough connected job searcher
and is restricted to fill only a smaller fraction of jobs through connections (µ < µ̄ = 1).
Substituting out the multipliers, we get a the solution for the three variables:

ω̄(1 − ϕ − ς)
eg∗ = (34)
wgu

ϕ+ς
a∗ = wgu (35)
1−ϕ−ς
wgu ϕ + ς
eg∗
c = (36)
χ 1−ϕ−ς
Both ϕ and ς raise wages and nepotism, and lower employment, relative to the efficient case.
High wages and nepotism in the public sector can therefore reflect two different situations.
Consider first a scenario where ς is low, so unions do not have much power, but where ϕ
is high - the government has a strong interest in nepotism. The government wants to hire
a larger number of connected workers, so it lowers employment and sets the wage higher
in order to attract a higher number of connected job searchers. Consider an alternative
scenario, where ϕ is low so there is no intrinsic interest in cronyism, but the weight of unions
is high (represented by an increase in ς). This induces the government to free up resources for
raising wages by lowering employment. Government jobs are now fewer and better paying.
This relaxes the constraint on the nepotism, lowering the multiplier which raises egc even if
ϕ is very close to zero (because there is no other cost of nepotism). This reflects the case
in which nepotism exists in the public sector mainly because union pressures set the wage

24
high, which in turn, generates large queues of unemployed seeking to get public jobs and
induces workers to find alternative ways to get them. In the two scenarios, both wages and
nepotism would be high, but for different reasons.
If there is an additional cost of nepotism ϑ > 0, there are two solutions depending on
whether the second constraint holds with equality or not. The third case, which mimics the
baseline version of the model, exists when the second constraint holds with strict inequality.
The interior solution is given by:
ω̄(1 − ς)
eg∗ = (37)
wg
ς
a∗ = w g (38)
1−ς
eg∗ g
c = ēc (39)

where ēgc = ϕϑ is the unconstrained choice, which we assume it is smaller than optimal
choice of eg∗ . In this case, the pressure from the media constrains nepotism, more than the
wages. The government’s targeted fraction of connected jobs is small enough so that the
restriction that wages are high enough never binds. The government is able to get its target
g
number of connected workers, given by µ̄ = eēg∗c , and nepotism does not affect government’s
choice of the number of workers nor their wage. An increase in µ̄ could reflect an increase
in governments’ intrinsic interest in nepotism (an increase in ϕ) or stronger control over
the media (a decrease in ϑ). Such changes would increase µ̄ but would not affect public
employment or wages. Wages could increase because of an increase in union power, which
would drain resources from the production of services. This would increase µ̄, as ēgc would
be the same but employment lower.6
We think there could be alternative ways of modeling the government that could generate
different interactions between policies. For instance, one could consider that the government
incorporates the effects of policies on the labour market, for instance on unemployment,
when deciding. In the absence of a consensual theory, we prefer to analyse the effects of each
policy variable on the labour market in isolation.

7 Extensions
In this section, we discuss and compare the effects of nepotism and government policies on
employment under two alternative model assumptions: (i) random search in the unconnected
6
The constrained solution when ϑ > 0, involves solving a quadratic equation in a: wg ϕ 2
ϑ a + (1 − ϕ −
g
ς)a − w (ϕ + ς). Beside a more complicated algebra it is conceptually similar to the constrained case with
ϑ = 0.

25
market; (ii) heterogeneous workers. The assumption of random search generates different
transmission mechanisms of employment and wage policies. While it is also used in the
literature, there is no clear evidence whether it is a more or less realistic assumption when
compared to segmented markets. Introducing worker heterogeneity allows us to characterize
how the selection of the workers into the public and private sector interacts with wages and
the existence and use of nepotism in the public sector. It allows us to capture the common
notion/belief that nepotism is associated with non-meritocratic hiring, i.e. not selecting the
best workers but those who have connections.

7.1 Random search between the private sector and the uncon-
nected public sector
We now analyze the case in which the workers without connections cannot direct their job
search exclusively towards the public or the private sector. We assume that these workers
search randomly for jobs in the two sectors. A matching function m(vu , uu ) determines the
total number of matches between unconnected workers and jobs and m(θ), where θ = uvuu ,
gives the rate at which unconnected workers match with a given vacancy. Since they search
randomly for jobs, the total number of vacancies available to them, consists of both private-
sector v p and government unconnected vacancies vug , vu = v p + vug . They find jobs in the
p
private sector at rate m(θ)γ p and in the public sector at rate m(θ)(1 − γ p ), where γ p = vvu is
the fraction of private-sector vacancies in the total number of vacancies available to workers
without connections.
The key difference between the model with random search and segmented markets is the
value of unemployment for unconnected workers. It changes because they now randomly
search for jobs in both sectors. Specifically,

(r + τ )Uu = b + m(θ)γ p [Eup − Uu ] + m(θ)(1 − γ p ) [Eug − Uu ] . (40)

Under segmented markets, tightness in the private sector is independent of any government
policy (see Lemma 1) because the outside option (unemployment value) of workers searching
in the private sector is independent of government policy. Under random search, by contrast,
the outside option of unconnected workers is a convex combination of the value a public-
sector job (Eug ) and the value of a private-sector job (Eup ) with weights reflecting the relative
number of vacancies in the two sectors, as seen in equation (40). Thus, public-sector wages,
employment, separation rate and nepotism affect private-sector wages, that are given by

wp = b + β [y − b + γ p θκ] + (1 − β)D(wg − b), (41)

26
 p
(1−γ )m(θ)
where D = r+τ +s g +(1−γ p )m(θ) measures how much public wages influence private-sector wage

bargaining. A free-entry condition as in (11) determines the number of vacancies posted

p−wp
by firms. But now the match surplus, Sup = r+s p +τ , depends also on public-sector policies

and nepotism. In addition, the cutoff connection cost, c̃ = Ucg − Uu , changes to reflect that
the value of unemployment to unconnected workers is now given by (40). The full set of
equations describing the model with random search, a formal definition of a steady-state
equilibrium and conditions for existence of a steady-state equilibrium are in Appendix C.
Under random search, the effects of government policies work through: i) the selection
into connected and non-connected workers (as in segmented markets); and ii) the outside
option of unconnected workers and its impact on private wages. We show in Appendix C
that:

Proposition 4 An increase in wg lowers job creation (lowers θ), induces more workers to
obtain connections and queue for connected public-sector jobs (i.e., increases Lgc and lowers
Lu ) and lowers the employment rate.

The increase in the public wage improves a worker’s payoff from getting a job in the
public sector. This improves the outside option of searching workers pushing their wage
in the private sector up and reducing firm’s incentives to create jobs. At the same time,
it induces more workers to obtain connections and queue for connected jobs. Both the
decrease in θ (m(θ)) and the decrease in Lu lower the employment rate. If µ = 0, meaning
that no connections sector exists, then all effects work only through the outside option.
In the other extreme case, where µ = 1 (meaning that vug = 0, γ p = 1, D = 0, and all
government vacancies are for connected workers), tightness and wages in the private sector
become identical to those obtained under segmented markets, and all effects work through
the selection into connected and non-connected workers, as in segmented markets.

7.1.1 The effect of nepotism on job creation and employment

Under random search, public-sector policies work not only through the selection into con-
nected and unconnected workers, but also through their impact on private-sector wages and
in turn, tightness. For this reason, the effect of nepotism on employment can be either
positive or negative. With a higher fraction of public-sector jobs being retained for workers
with connections, a larger fraction of workers who do not have connections end up in private-
instead of public-sector jobs (i.e., γ p increases, shifting weights in (40) from Eug to Eup ). As-
suming that government jobs are more valuable to workers than private jobs are (that is,
Eug > Eup ), the presence of nepotism in the public sector worsens the outside option of uncon-
nected workers; private wages decrease; and job creation of firms increases with a positive

27
impact on employment. In addition to this job-creating effect, an increase of nepotism makes
the option of investing in connections more attractive. More workers seek public-sector jobs
through their connections, queuing up, with a negative impact on employment.
In segmented markets, a decrease in the fraction of government jobs available to non-
connected workers has a positive impact on employment because some workers, those whose
cost of obtaining connections is large, will direct their search towards the private sector.
Under the assumption of random search this positive effect is not present, because workers
cannot direct their search towards the private sector. On the other hand, under random
search there is an additional positive effect on employment, which is not present under
segmented markets: nepotism hurts the outside option of workers thereby increasing private-
sector job creation.

7.2 Heterogeneous Workers

We assume that a fraction Xh of the labor force has high ability and the remaining has low
ability. Since the total labor forced is normalized to 1 we can write Xl = 1 − Xh , where Xl is
the share of low-ability workers in the labor force. High-ability workers are more productive
than low-ability workers, that is, yh > yl . To avoid problems of adverse selection, we assume
that ability is observable. In the private sector, there are two separate submarkets, one for
high- and one for low-ability workers. In the public sector we assume only one submarket
in which the government opens vacancies and both types of workers can apply. We further
assume that it pays the same wage to both types of workers, reflecting the fact that wages in
the public sector are more compressed (even within education categories) than in the private
sector. But the chances of being hired differ between the two types. Since ability is observ-
able, the government can perfectly screen candidates and gives priority to the high-types.
Perfect screening and unique wages in the public sector were also assumed by Geromichalos
and Kospentaris (2020). We differ on the assumption on the private sector. While they
assume random search from the firm side, which creates more intricate interactions between
composition of ability and job creation, given our assumption that ability is observed, we
have segmented markets, meaning that firms can direct their search towards either high- or
low-type workers.
Both high- and low-ability workers prior to entering the labor force, they decide which
sector to join and whether or not to obtain connections. In the connected public sector the
job finding rate is independent of ability. As above, by obtaining connections workers can
gain access to a subset of jobs reserved for connected workers, but in addition, can avoid
being screened based on ability. As it will be further discussed below, the low-types have

28
an additional incentive to use connections, which is to avoid screening, whereas, the only
reason why the high ability workers might use connections is to jump the queue when the
queue for public jobs is too long.
Since the two types differ in terms of productivity, the value of searching for a job as
well as the value of being employed in the private sector also differ. On the firm size, there
are now two free-entry conditions, two different surpluses, and hence different tightness and
wages for high- and low-ability workers, given by:

κ (yi − b)(1 − β)
= . (42)
q(θi ) r + τ + sp + βm(θi )
wip = b + β(yi − b + κθi ). (43)

where i = [h, l] denotes high- and low-ability. It can be verified that θh > θl because yh > yl .
It follows also that whp > wlp and high-ability workers enjoy higher values in the private sector
p p p p
than lower ability workers. That is, Eu,h > Eu,l and Uu,h > Uu,l . Moreover, following Lemma
1, θh and θl are both independent of employment and wage policies.
In the public sector also, the values of being employed or unemployed for workers that
do not have connections differ by ability. Despite wages being the same for all workers
irrespective of ability, because the government screens candidates and gives preference to
high-ability workers they have different job-finding rates, mgu,i . In the connection market,
we assume that ability does not affect your probability of being hired, given by mgc . Since
the wages are also equal for the two types, the values of unemployment and employment for
connected workers are the same irrespective of ability and remain as in (4) and (6).

7.2.1 Workers’ selection and public-sector wages

High- and low-ability workers without connections can search in either the public or the
private sector. In an equilibrium where both markets are active, the values of these two
options would have to equate:
p g
Uu,h = Uu,h = Uu,h (44)
p g
Uu,l = Uu,l = Uu,l (45)

These two conditions would determine the numbers of high- and low-ability unconnected
searchers in the public sector, ugu,h and ugu,l . Given our setup, only one of the two types of
workers will be active in the unconnected public-sector market. We show that the presence
of high- and low-ability workers in each of the two segments of the public sector depends
on the wage. If wages are low, only low-ability workers are interested in the public-sector
and no high-type will apply. Thus, ugu,h = 0 and ugu,l is pinned down by condition (45). If

29
wages are high enough to attract high-ability workers, as they have priority, given the perfect
screening, the low-ability have no chance to get a job. Hence, ugu,l = 0 and ugu,h is pinned
down by condition (44).
Alternatively all workers can use connections to get into the public sector after paying
the cost c drawn from Ξ(·). The threshold cost at which a worker of type-i is indifferent
between using and not using connections to find a government job is c̃i = Uc −Uu,i . These two
thresholds determine the allocation of each type of worker into connected or unconnected.
The following lemma, depicted graphically in Figure 5, summarizes the results on selection.

Lemma 4 The (unconnected) public sector will attract high-ability workers only if wg >
wgu,h . If wg > wgu,h there are no low-ability workers in the unconnected public-sector and all
low ability workers attached to the public sector have connections. Some high-ability workers
will use connections to get a public-sector job only if wg > wgc,h (> wgu,h ).

The exact expressions for wgu,h and wgc,h are in Appendix D. Since high-ability workers
enjoy higher wages and can find jobs in the private sector faster than low-ability workers
there is a cutoff wage below which high-ability workers ignore the public sector. Below this
cutoff, wgu,h , the wage is too low to attract better workers, even when their chances of getting
one are at the maximum (mgu,h = 1). It follows that if wg ≤ wgu,h then only the low types
are attached to the public sector, while all high-type workers are in the private sector. This
case corresponds to our benchmark model where all workers searching in the public sector
are identical. The only difference here is that an additional “high-ability” market exists
in the private sector, which is, however, completely isolated from the public sector or the
“low-ability”market of the private sector. So results in Propositions 1-3 carry through to
this case and, following Lemma 2, two additional cutoffs, wgu and wgc , below wgu,h , determine
the existence of a public sector of size eg and the existence of a connections sector of size
µ̄eg , given that the public sector attracts only low-ability workers.
If wg > wgu,h , the number of high-type unconnected workers wanting a public-sector job is
greater than the number of vacancies (mgu,h < 1) and some of them will be queuing up. This
means that if wg > wgu,h a low-type can never get a public job without connections, because
the high-types will be given priority. Low-ability workers will either go to the private sector

Figure 5: Lemma 4: The role of public-sector wage

Only low-type in public sector (benchmark) Only high-type in unconnected public sector
wg
wgu wgc wgu,h wgc,h
Public sector Connections Connections No high-type in High-type in
breakdown constrained unconstrained connected market connected market

30
or try to get a public-sector job through connections. If wages are just above this cut-off,
no high-type would get connections and all workers hired through connections are of low-
type. But if the wage is much higher then large queues of high-ability workers waiting to get
jobs in the public sector may make it worthwhile for some of them to invest in connections
to jump the queue. In particular, there is a wage wgc,h (> wgu,h ), such that if wg > wgc,h ,
then Uc > Uu,h (c̃h > 0) and some high-type workers will try to get public jobs through
connections (Lgc,h = Xh Ξ(c̃h ) > 0), whereas, if wg ≤ wgc,h , then Uc ≤ Uu,h (c̃h ≤ 0) and no
high-ability worker obtains connections (Lgc,h = 0).

7.2.2 The effects of public-sector wages and nepotism

When the wage is below wgu,h the model looks very much like the baseline model. An increase
in µ̄ increases total employment by shortening public sector queues, while more generous
wages have the opposite effect, as summarized in Propositions 1 and 2. The only difference
here is that these changes affect only the low-ability workers since for wg ≤ wgu,h there are
no high-ability workers in the public sector.
In Appendix D we also derive results for the effects of an increase in µ̄ on employment
in the case where wg > wgu,h , and both high- and low-ability workers get connections. We
explore, in addition, the effect of changes in µ̄ on the composition (in terms of ability)
of private- and public-sector labor force and employment. Results are summarized in the
following two Propositions.

Proposition 5 If wg > wgc,h an increase in µ̄ decreases the number of workers searching for
public-sector jobs (decreases ug = ugu + ugc ), increases the number of workers in the private
sector (i.e., increases Lpu = 1−Lgu −Lgc ) and increases the employment rate (as in Proposition
1). But, if wgc,h ≥ wg ≥ wgu,h these effects are ambiguous.

At higher µ̄, more workers use connections to get jobs in the public sector, but also more
workers go to the private sector. To be worthwhile for some high-ability workers to use
connections to get public jobs (wg > wgc,h ), it must be the case that the public sector is too
crowded; queues for government jobs are so long so that incentives to use connections to
get one are strong. If queues are long, an increase in µ̄, will drive more high-ability workers
away from the (unconnected) public sector and into the private sector than workers from
the private towards the (connections) public sector. As a consequence, the total labor force
attached to the private sector increases.
In the intermediate case where high-ability workers have no incentive to use connections
(wc,h ≥ wg ≥ wgu,h ), it is not clear if an increase in µ̄ will drive more high-ability workers
g

from the (unconnected) public sector towards the private sector or more low-ability workers

31
from the private sector towards the connections sector. Additionally, an increase in µ̄ effects
the two types of workers differently through its impact of the relative size of the public sector
(decreases it for high-ability and raises it for low ability). As the public sector has lower
separation rate, there is an additional effect through frictional unemployment. Hence, the
impact on private-sector labor force in this case is ambiguous.

Proposition 6 If wg > wgu,h an increase in µ̄ increases (decreases) the average ability of

workers in the private (public) sector.

When the government attracts both low- and high-ability workers (wg > wgu,h ), those
discouraged by the presence of nepotism are the high-ability ones. The low types benefit the
most from the use of connections because they can avoid screening and competition from the
high types and in addition, their wage premium is greater than that of high-type workers.
Low-types have more incentives to get connections so that an increase in the fraction of
jobs available through connections will attract more of them. Although through a different
mechanism, this result echoes the findings in Geromichalos and Kospentaris (2020) that
meritocratic hiring of public employees decreases private sector’s productivity. This raises
another question of where is it better to have the best workers: in the private or the public
sector? An open and difficult question to be answered with future research.

7.3 Further analysis

We consider also three additional exercises presented in Appendixes E, F and H. The first
is to consider a version with competitive search in the private sector. We show that the
equilibrium conditions are identical to those obtained in the benchmark model when the
Hosios condition holds. The second is to assume that the newborn pay connections costs
to current connected public-sector workers so that current workers will help fast-track them
into the public sector. These payments are a “connections premium”, which will further raise
the value of working in the public sector for connected workers. Finally, we parameterize the
model to the Spanish economy and perform some numerical exercises. Besides confirming
the results of propositions, the main finding regards the welfare effects of nepotism. In
the model with segmented markets more nepotism raises welfare, while in a random search
model, although it also lowers unemployment, the effects on welfare are sometimes negative.

8 Conclusion
This paper provides a benchmark model to understand how public-sector hiring and wage
policies affect rent-seeking decisions and employment. The model takes in account one perva-

32
sive characteristic in many public sectors - hiring practices are sometimes based on nepotism.
Our results provide insights that can explain some European cross-country facts. Previous
literature has highlighted the problems of setting high public-sector wages. For example,
Gomes (2015) and Afonso and Gomes (2014) shown that they generate higher unemploy-
ment. Cavalcanti and Santos (2017) argue that higher wages might lead to misallocation
of resources with a lower entrepreneurship rate. We highlight an additional negative effect.
Higher public-sector wages might lead workers to pursue rent-seeking activities.
We have shown that the existence of a “connections” market for public jobs requires that
public-wages are very high compared to those in the private sector. This result is consistent
with evidence that Southern European countries, known for having non-meritocratic hiring,
have a higher public-sector wage premium, while Nordic countries, in which governments
follow more meritocratic hiring, tend to have a lower or a negative premium. The results
also suggest why Southern European governments might maintain the status quo of the hiring
process. Conditional on high wages and long queues for public-sector jobs, the existence of
nepotism actually lowers unemployment.
The connections market that we have emphasized could not exist in the private sector
in the same form. We have shown that in the public-sector connected workers are given
priority for jobs even if the surplus they generate is not larger than that of workers without
connections. In the private sector this is not possible. Wage bargaining and free-entry of
firms would ensure that job-finding rates would reflect nothing but match surplus. Obtaining
connections would help find jobs in the private sector faster only if connections could help
improve the match surplus or only if employers could somehow benefit more from hiring
through connections than through standard search. If not, the endogeneity of job-creation
– that is absent in the public sector – would eliminate any incentive to become connected.
We have also shown that, if low-ability workers face lower chances of entering the public
sector, the existence of nepotism induces the high-ability workers to search for private jobs
and low-productivity workers to invest in connections, reflecting another common perception
of nepotism: that connected workers are of worse quality. To further evaluate the welfare
effects in a model with heterogeneous workers, we would need a metric on the productivity
of the public sector and of workers in the public sector, as well as the value of public-sector
services, for which there is little empirical evidence. Note that if nepotism drains lower
quality workers away from the private and into the public sector, the effect on welfare is
not necessarily negative if high-quality workers are more important in the private- instead
of public-sector production.
While this paper was motivated by differences across European countries, several of the
results are useful to think more widely about public sectors in developing countries. Finan

33
et al. (2015) describe a growing body of field experiments in developing countries exploring
the personnel economics of the state. Our model can provide a theoretical foundation to
help designing field experiments. The literature commonly argues that higher wages for civil
servants are necessary to avoid corruption in the public sector. We show that, on the other
hand, higher wages for civil servants creates an asymmetry with the private sector, which
might itself create an incentive for a different type of corruption.
Although we have emphasized the role of nepotism in recruiting government employees,
our model is very general, and some of the results can be extrapolated to other country-
specific public-sector characteristics. Dickson et al. (2014) find that countries with a positive
lifetime premium of the public sector, France and Spain in their sample, are also the countries
that require costly entry procedures, such as national exams. We could reinterpreted the
model, considering the cost of connections as the cost of preparing for an exam, and µ the
fraction of civil servants hired through an exam. In this case, high ability workers would have
more incentive to take the exam, and national exams would further discourage low-ability
workers from applying to the public sector. We would conclude that, although this channel
could be inefficient, conditional on an inefficient wage policy, it might be one way to not only
screen candidates and improve the quality of public-sector labor force, but also to minimize
the effects on unemployment.

References
Afonso, A. and P. Gomes (2014): “Interactions between private and public wages,”
Journal of Macroeconomics, 39, 97–112.

Albrecht, J., M. Robayo-Abril, and S. Vroman (2018): “Public-Sector Employment

in an Equilibrium Search and Matching Model,” Economic Journal, 129, 35–61.

Alesina, A., R. Baqir, and W. Easterly (2000): “Redistributive Public Employment,”

Journal of Urban Economics, 48, 219–241.

Bello, S. L. and I. Morchio (2017): “Like Father, Like Son: Occupational Choice,
Intergenerational Persistence and Misallocation,” Tech. rep.

Bó, E. D., F. Finan, and M. A. Rossi (2013): “Strengthening State Capabilities: The
Role of Financial Incentives in the Call to Public Service,” The Quarterly Journal of
Economics, 128, 1169–1218.

Bradley, J., F. Postel-Vinay, and H. Turon (2017): “Public sector wage policy
and labour market equilibrium: a structural model,” Journal of the European Economic
Association, 15, 1214–1257.

34
Burdett, K. (2012): “Towards a theory of the labor market with a public sector,” Labour
Economics, 19, 68–75.

Cavalcanti, T. and M. Santos (2017): “(Mis)Allocation Effects of an Overpaid Public

Sector,” Tech. rep.

Charron, N., C. Dahlstrm, M. Fazekas, and V. Lapuente (2017): “Careers, Con-

nections, and Corruption Risks: Investigating the Impact of Bureaucratic Meritocracy on
Public Procurement Processes,” The Journal of Politics, 79, 89–104.

Charron, N., L. Dijkstra, and V. Lapuente (2014): “Regional governance matters:

quality of government within European Union member states,” Regional Studies, 48, 68–
190.

Christofides, L. and M. Michael (2013): “Exploring the public-private sector wage

gap in European countries,” IZA Journal of European Labor Studies, 2.

Christofides, L. N. and P. Pashardes (2002): “Self/paid-employment, public/private

sector selection, and wage differentials,” Labour Economics, 9, 737–762.

Colonnelli, E., P. Mounu, and E. Teso (2018): “Patronage and Selection in Public
Sector Organizations,” Documentos de Trabajo 016723, Universidad del Rosario.

Dahlström, C., J. Teorell, S. Dahlberg, F. Hartman, A. Lindberg, and M. Nis-

totskaya (2015): “The Quality of Government Expert Survey II A Report,” Tech. rep.

Dickson, M., F. Postel-Vinay, and H. Turon (2014): “The lifetime earnings premium
in the public sector: The view from Europe,” Labour Economics, 31, 141–161.

Disney, R. and A. Gosling (1998): “Does it pay to work in the public sector?” Fiscal
Studies, 19, 347–374.

Durante, R., G. Labartino, and R. Perotti (2011): “Academic Dynasties: Decen-

tralization and Familism in the Italian Academia,” NBER Working Papers 17572, National
Bureau of Economic Research, Inc.

Fafchamps, M. and J. Labonne (2017): “Do Politicians’ Relatives Get Better Jobs?
Evidence from Municipal Elections,” Tech. Rep. 2.

Finan, F., B. A. Olken, and R. Pande (2015): “The Personnel Economics of the State,”
NBER Working Papers 21825, National Bureau of Economic Research, Inc.

Fontaine, I., I. Galvez-Iniesta, P. Gomes, and D. Vila-Martin (2020): “Labour

market flows: Accounting for the public sector,” Labour Economics, 62, 101770.

Galenianos, M. (2014): “Hiring through referrals,” Journal of Economic Theory, 152,

304–323.

Gelb, A., J. B. Knight, and R. H. Sabot (1991): “Public Sector Employment, Rent
Seeking and Economic Growth,” Economic Journal, 101, 1186–1199.

35
Geromichalos, A. and I. Kospentaris (2020): “The Unintended Consequences of Mer-
itocratic Government Hiring,” Working Papers 335, University of California, Davis, De-
partment of Economics.
Gomes, P. (2015): “Optimal public sector wages,” Economic Journal, 125, 1425–1451.
——— (2018): “Heterogeneity and the public sector wage policy,” International Economic
Review, 59, 1469–1489.
Gyourko, J. and J. Tracy (1989): “The Importance of Local Fiscal Conditions in
Analyzing Local Labor Markets,” Journal of Political Economy, 97, 1208–31.
Hörner, J., L. R. Ngai, and C. Olivetti (2007): “Public Enterprises And Labor
Market Performance,” International Economic Review, 48, 363–384.
Horvath, G. (2014): “Occupational mismatch and social networks,” Journal of Economic
Behavior and Organization, 106, 442–468.
Katz, L. F. and A. B. Krueger (1991): “Changes in the Structure of Wages in the
Public and Private Sectors,” NBER Working Papers 3667.
Keynes, J. M. (1936): The General Theory of Employment, Interest and Money, Macmil-
lan, 14th edition, 1973.
Krueger, A. B. (1988): “The Determinants of Queues for Federal Jobs,” ILR Review, 41,
567–581.
Martins, P. S. (2010): “Cronyism,” IZA Discussion Papers 5349.
Michaillat, P. (2014): “A theory of countercyclical government multiplier,” American
Economic Journal: Macroeconomics, 6.
Moen, E. R. (1997): “Competitive Search Equilibrium,” Journal of Political Economy,
105, 385–411.
Navarro, L. and M. Tejada (2018): “On the Interaction between Public Sector Em-
ployment and Minimum Wage in a Search and Matching Model,” Tech. rep.
Nickell, S. and G. Quintini (2002): “The consequences of the decline in public sector
pay in Britain: a little bit of evidence,” Economic Journal, 112, F107–F118.
OECD (2017): “Government at a Glance 2017,” Tech. rep.
Quadrini, V. and A. Trigari (2007): “Public Employment and the Business Cycle,”
Scandinavian Journal of Economics, 109, 723–742.
Salomäki, A. and T. Munzi (1999): “Net Replacement Rates of the Unemployed. Com-
parison of various approaches,” European Commission Economic Papers 133.
Scoppa, V. (2009): “Intergenerational transfers of public sector jobs: a shred of evidence
on nepotism,” Public Choice, 141, 167–188.

36
 COMPANION APPENDIX

Jumping the queue: nepotism and public-sector pay

Andri Chassamboulli and Pedro Gomes

Appendix A: Proofs of propositions

• A.1 Lemma 2
• A.2 Proof of existence and uniqueness
• A.3 Proposition 1
• A.4 Proposition 2

Appendix B: Efficiency

Appendix C: Random search

• C.1 Setup
• C.2 The case µ = 1
• C.3 Definition of equilibrium
• C.4 Proof of existence and uniqueness
• C.5 Proof of proposition 4

Appendix D: Heterogeneous workers

• D.1 Setup
• D.2 Wage cutoffs
• D.3 Proof of proposition 5
• D.4 Effects of increasing wg when wg > wgu,h
• D.5 Proof of proposition 6

Appendix E: Competitive search in the private sector

Appendix F: Connections premium

Appendix G: Survey data

• Figure G1: Quality of government survey - European countries
• Figure G2: Quality of government survey - World regions
• Figure G3: Regression of the unemployment rate

Appendix H: Numerical Analysis

• Figure H.1: 4-state stocks and flows, Spain
• Table H.1: Parameterization of segmented markets model
• Figure H.2: Effects of nepotism, role of public-sector wages
• Figure H.3: Effects of nepotism, role of connections costs
• Figure H.4: Effects of public-sector wages
• Figure H.5: Effects of public-sector employment
• Table H.2: Effects of policies under different models
• Figure H.6: Effects of nepotism (segmented markets and random search)

i
A Proofs of propositions
A.1 Lemma 2
We consider that the public-sector unconnected labour market breaks down if the government
is not able to hire enough workers to replace the workers that have lost their job. At the
limit, it means the government needs to post a wage, defined as wgu , such that it attracts at
least (1 − µ)(sg + τ )eg job searches. This means ugu = (1 − µ)(sg + τ )eg and the job-finding
rate is 1 (mgu = 1). Applying this to (16) and then setting Uup = Uug gives
1
b+ [wg − b] = (r + τ )Uup
r + τ + sg + 1 u
Substituting the (r + τ )Uup by equation (15) we get

(r + τ + sg + 1)m(θ∗ ) p,∗
wgu = [w − b] + b
r + τ + sp + m(θ∗ )

where θ∗ and wp,∗ are the equilibrium tightness and wages in the private sector.

If µ = 0 then no connections sector exists and all workers hired into the public sector are
unconnected. If, on the other hand, a connections sector exists then a share µ of public-sector
workers are hired through connections. For the existence of a connections sector, through
which the government is able to hire a fraction µ of its employees the government needs to
attract at least µ(sg + τ )eg connected job searchers. This occurs when the government pays
a higher wage, wgc , so that queues in the public sector are long enough to induce enough job
searchers to use connections to get government jobs.

wgc = wgu + Ξc,−1 (µ(sg + τ )eg )(r + τ + sg + 1)

where Ξc,−1 is the inverse of the distribution of connection cost. What it means is that, at
the margin, the government has to pay high enough wages such that public-sector queues
are long enough and a sufficiently high mass of newborns decide to pay the cost and obtain
connections.
Notice that wgc is increasing in µ, while wgu is independent of µ. If µ = 0 then we get
wc = wgu , whereas if µ = µ̄ then wgc = wgc where
g

wgc = wgu + Ξc,−1 (µ̄(sg + τ )eg )(r + τ + sg + 1)

A.2 Proof of Existence and Uniqueness of a Steady-State Equi-

librium
It can be easily verified that the free-entry condition in (13) pins down a unique equilibrium
value for tightness in the private sector θ∗ . To complete the proof of existence and uniqueness

ii
we need to show that with θ∗ substituted in, the threshold condition in (23) gives a unique
equilibrium value for c̃.
Let us write (23) as:
µ(sg +τ )eg
" #
1 g
Lc −µe g
g 1 βκθ
µ(s g +τ )eg (w − b) − c̃ = (A.1)
r + τ r + τ + sg + g g r + τ (1 − β)
Lc −µe

where Lgc = Ξ(c̃). The left-hand-side of (A.1) decreases with c̃. This means that with
θ∗ substituted in we can use (A.1) to solve for the equilibrium value of c̃. The equilibrium
conditions (13) and (23) thus give a unique set of equilibrium values c̃∗ and θ∗ . This completes
the proof of existence and uniqueness.

A.3 Proof of Proposition 1

p
First, we show that dLdµ
u
> 0 (where µ = µ̄):
g g g
Let L = Lu + Lc denote the total number of workers that are either employed or are
searching in the public sector. Using conditions (18) and (20) to solve, respectively, for Lgu
and Lgc , and then adding them up gives:
" " ##
µ 1 − µ
Lg = eg λ + (1 − λ)(wg − b) β
+ β (A.2)
c̃(r + τ ) + 1−β κθ 1−β
κθ

where λ = r+srg +τ . Recall that the equilibrium value of θ is given by equation (13) and is
dθ
independent of µ; thus dµ = 0. Given this, we can write:

dLg ∂Lg ∂Lg ∂c̃

= + (A.3)
dµ ∂µ ∂c̃ ∂µ

and using (A.1) we can derive that

∂c̃
>0 (A.4)
∂µ
g ∂Lg
Moreover, it can be easily verified from (A.2) that ∂L
∂µ
< 0 and ∂c̃
< 0, implying from (A.3)
that
dLg
< 0.
dµ
p
Given that Lpu = 1 − Lg , it follows that dL
dµ
u
> 0.
dug
Next we show that dµ < 0. The number of workers searching in the public sector with
and without connections are given, respectively, by ugc = Lgc − µeg and ugu = Lgu − (1 − µ)eg .
By adding them up we get ug = ugu + ugc = Lg − eg . The number of workers employed in
g
the public sector, eg , is exogenous and independent of µ, while, as shown above, dL
dµ
< 0. It
dug
follows that dµ < 0.

iii
 de
Finally, we show that the employment rate (e) increases. That is, dµ > 0. The total
g p g
employment rate is given by e = e + e , where e is exogenously set by the government,
p p m(θ)Lpu
while e can be derived from (24) and (27): e = sp +τ +m(θ) . Adding them up gives:

m(θ)(1 − Lg )
e = eg + (A.5)
sp + τ + m(θ)
de dLpu
Evidently, dµ
> 0, since θ is independent of µ, while dµ
> 0.

A.4 Proof of Proposition 2

First, let us show that the number of workers searching in the public sector increases as the
dLg dLp
public-sector wage increases; that is dw g > 0, which ultimately implies that dw g < 0, since

Lpu = 1 − Lg .
Using condition (18) we can solve for Lgu and obtain:
" !#
g
w − b
Lgu = (1 − µ)eg λ + (1 − λ) β
(A.6)
1−β
κθ

where λ is as defined above (in Proposition 1).

The total number of workers attached to the public sector is given by Lg = Lgu +Lgc where
Lgu is as derived above and Lgc = Ξ(c̃). Taking the derivative with respect to wg gives:

dLg dLgu dLgc dc̃

= + (A.7)
dwg dwg dc̃ dwg
dLgu dLgc
It is straightforward to verify from (A.6) that dwg
> 0. Moreover, dc̃
= ξ(c̃) > 0 and using
(A.1) we can derive that:

dc̃ M
= M (1−M )(wg −b) dLgc
>0 (A.8)
dwg r+τ + Lgc −µeg dc̃

µ(sg +τ )eg
g
Lc −µeg
where M = µ(sg +τ )eg . It follows from (A.7) that:
r+τ +sg + g
Lc −µeg

dLg
>0 (A.9)
dwg
Using (A.1), (A.2) and (A.8) we can further show that:
   
g
dL
= eg (1−λ) 
µ 1 − r+τ + 1−µ
β
g
  g β
dw c̃(r + τ ) + 1−β κθ β
r + τ + c̃(r + τ ) + 1−β (1−M )
κθ Lgc −µeg dL c
1−β
κθ
dc̃
(A.10)

iv
Note that if µ = 0 then " #
g
dL 1
g
= eg (1 − λ) β
(A.11)
dw 1−β
κθ
Comparing (A.10) to (A.11) shows:

dLg dLg
> (A.12)
dwg µ=0 dwg µ>0

and the increase in the number of workers searching in the public sector due to an increase
in the public-sector wage is larger when µ = 0 than when µ > 0.
Since Lpu = 1 − Lg and ug = Lg − eg , the decrease and increase, respectively, in Lpu and
ug , is also larger when µ = 0 than when µ > 0.

v
B Efficiency
As also mentioned in the text, the existence of a connections sector and of queues for public-
sector jobs are both inefficient. These two types of inefficiencies can be eliminated by setting
µ = 0, which implies Lgc = 0, and wg = wg , which ensures that ugu = (sg + τ )eg and the
job-finding rate for government jobs is 1. We next compare the central planner’s solution to
the decentralized one, described in the text, and show that the remaining inefficiency, the
congestion externalities can be eliminated with the Hosios condition.
We follow Hosios (1990), Charlot and Decreause (2005), among others, and set r = 0,
so that the central planner maximizes the steady-state surplus. The planner’s problem is to
choose θ, up to maximize total output, plus unemployment income, minus job creation costs.
Given that public sector employment is fixed. The planner’s objective is to

max(1 − Lg ) [(1 − up )y + up b − θκup ]

s.t
sp + τ
up =
sp + τ + m(θ)
We set the Langrangian

sp + τ
 
g p p p p
L = (1 − L ) [(1 − u )y + u b − θκu ] + φ u − p (B.1)
s + τ + m(θ)

The three optimality conditions are

∂L m0 (θ)
=0 ⇒ φ p = (1 − Lg )κ (B.2)
∂θ s + τ + m(θ)
∂L
= 0 ⇒ φ = (1 − Lg ) [y − b + κθ] (B.3)
∂up
∂L sp + τ
= 0 ⇒ up = p (B.4)
∂φ s + τ + m(θ)

Substituting (B.3) into (B.2) gives:

κ η(y − b)
= p (B.5)
q(θ) s + τ + m(θ)(1 − η)

where it may be useful to recall that m(θ) = θη and m0 (θ) = ηq(θ). It is easy to verify by
comparing (B.5) to (13), that given r = 0, if β = (1 − η), then the decentralized equilibrium
is identical to the central planner’s solution.

vi
C Random search
C.1 Setup
In this appendix we give the full set of equations of the model with random search and
characterize its steady-state equilibrium. Further, we show that in the limiting case where
µ = 1 the model with random search becomes identical to the model with segmented markets
and we provide proofs of Proposition 4.
The values of being unemployed and employed for connected workers remain as in the
Benchmark model; given by (4) and (6). The same holds for the values of being employed in
either the private or the public sector for unconnected workers (equations 2 and 5), and the
values of a private-sector filled jobs and vacancies (equations 7 and 8). The cutoff connection
cost as well as the selection of workers into the two groups, Lgc , Lu , also remain as given in
equations (21) and (22). As discussed in the text, only the value of unemployment for
unconnected workers changes. It is now given by equation (40). The Nash bargaining wage
of the private sector changes accordingly and is as given in (41).
Both government and private firms that seek to hire workers p g
through regular search in the
market meet with workers at rate q(θ) = m(θ) θ
, where θ = vu +vu
uu
. The number of vacancies in
the private sector is determined endogenously by free entry that drives the value of a private-
sector vacancy to zero, Vup = 0. The government needs to post enough vacancies for workers
without connections to ensure that the total number of matches with such workers, q(θ)vug ,
equals the number of unconnected workers that it needs to hire. Hence, the government
posts vug vacancies to ensure q(θ)vug = (1 − µ) (sg + τ ) eg .
Setting Vup = 0 and using the Nash bargaining conditions in (10), we can write the surplus
of a private-sector match as

y − b − D(wg − b)
Sup = (C.1)
r + τ + sp + (1 − D)βm(θ)γ p

and the zero-profit condition that determines job creation in the private sector becomes:

κ (1 − β)(y − b − D(wg − b))

= (C.2)
q(θ) r + τ + sp + (1 − D)βm(θ)γ p

We can write the threshold level of connection costs, c̃ = Ucg − Uu , as:

µ(sg +τ )eg
" #
p
1 ucg
g g βκθγ
c̃ = µ(s g +τ )eg (w − b) − D(w − b) − (1 − D) (C.3)
r+τ r+τ +s + g g (1 − β)
uc

As in the benchmark model we treat public sector employment as an exogenous policy

variable. There are eg workers employed in the public sector. Among these workers, µg are
workers who were hired through connections (egc ) and the remaining (1 − µ)eg are workers
hired through regular search in the market (egu ). The number of workers employed in the
private sector is endogenous and depends on job creation in the private sector as well as
conditions in the public sector. The labor force of workers without connections consists
of those employed in the public sector, those employed in the private sector (epu ), and the

vii
unemployed (uu ). Hence, uu = Lu − (1 − µ)eg − epu . By equating the flows in, m(θ)γ p uu , to
the flows out of the state where a worker is employed in the private sector, epu (sp + τ ) we
obtain:
m(θ)γ p [Lu − (1 − µ)eg ]
epu = (C.4)
m(θ)γ p + τ + sp
(τ + sp ) [Lu − (1 − µ)eg ]
uu = (C.5)
m(θ)γ p + τ + sp
p g
vu +vu
Given θ = uu
and q(θ)vug = (1 − µ) (sg + τ ) eg , we can use (C.5) to write:

sp + τ m(θ) [Lu − (1 − µ)eg ] − (1 − µ)(sg + τ )eg

 
p
γ = (C.6)
m(θ) (sp + τ ) [Lu − (1 − µ)eg ] + (1 − µ)eg (sg + τ )
Using (C.4) and (C.6) we can write the total employment of workers without connections,
eu = epu + (1 − µ)eg as:

m(θ)Lu + (1 − µ)eg (sp − sg )

eu = (C.7)
sp + τ + m(θ)

C.2 The case µ = 1

If µ = 1, then, as can be seen from (C.6), γ p = 1, which implies D = 0. Setting γ p = 1 and
D = 0 in (41), (C.2) and (C.3) gives:

wp = b + β [y − b + θκ] (C.8)
 
κ y−b
= (1 − β) (C.9)
q(θ) r + sp + τ + βm(θ)
µ(sg +τ )eg
" #
1 g
Lc −µe g
g βκθ
c̃ = µ(s g +τ )eg (w − b) − (C.10)
r+τ r+τ +s + g g g (1 − β)
Lc −µe

The zero-profit condition in (C.9) gives a unique equilibrium values for θ which is indepen-
dent of government policy or conditions in the government sector. Moreover, the zero-profit
condition in (C.9), private-sector wages in (C.8) and the cut-off cost in (C.10) are identical
to those obtained under segmented markets (equations 13, 14 and 23, respectively). Hence,
if µ = 1, private-sector job creation and tightness, wages, as well as the composition of the
labor force in terms of connections in the model with random search are identical to those
obtained under segmented markets.

C.3 Definition of Equilibrium

A steady state equilibrium consists of a cut-off cost {c̃}, tightness {θ}, and unemployed
{uu , ugc }, such that, given some exogenous government policies {wg , eg , µ}, the following
apply.

viii
 1. Private-sector firms satisfy the free-entry condition (C.2).
2. Private-sector wages are the outcome of Nash Bargaining (41).
3. Newborns decide optimally their investments in connections and the population shares
are determined by equations (21)-(22).
4. Flows between private employment and unemployment are constant

(sp + τ )ep = m(θ)γ p upu ,

5. Population add up constraints are satisfied:

Lu = ep + (1 − µ)eg + uu (C.11)
Lgc = µeg + ugc (C.12)
Lu + Lgc = 1 (C.13)

C.4 Proof of Existence and Uniqueness

To prove the existence and uniqueness of a steady state equilibrium under random search
we show below that the free-entry condition in (C.2) gives a unique equilibrium value for
θ. The equilibrium values of the cut-off costs can then be determined by substituting the
equilibrium value of θ in equation (C.3). Then using (21) and (22) we can determine Lu and
Lgc , which in turn, together with the equilibrium value of θ can be substituted in equations
(41), (C.4), (C.5), (C.11) and (C.12) to determine wages and employment in the private
sector.
The job creation condition in (C.2) and the cut-off connection cost in (C.3) can be written
as:
κ (y − b − OO)
= (C.14)
q(θ) r + τ + sp
1
c̃ = [Ac − OO] (C.15)
r+τ
µ(sg +τ )eg
g
Lc −µeg
where Ac ≡ µ(sg +τ )eg (wg − b),
r+τ +sg + g
Lc −µeg

β
OO = D(wg − b) + (1 − D) κθγ p (C.16)
1−β
(1−γ )m(θ) p
is the expression for the outside option of workers, and D = r+τ +s g +(1−γ p )m(θ) .

Taking the derivative of (C.16) and (C.15) with respect to θ we obtain:

dOO ∂OO ∂OO dc̃

= + (C.17)
dθ ∂θ  ∂c̃ dθ
∂Ac dLgc dc̃ dOO

dc̃ 1
= − (C.18)
dθ r + τ ∂Lgc dc̃ dθ dθ

ix
where

∂OO (1 − D)m(θ) (Eug − Eup ) (sp + τ )(1 − γ p ) dLu

= − p (C.19)
∂c̃ (s + τ ) [Lu − (1 − µ)eg ] + (1 − µ)eg (sg + τ ) dc̃
 g
Eu m(θ) + E P (sp + τ )
  
∂OO p p p
= (1 − D)q(θ) (1 − γ )η − Uu + γ (Eu − Uu )(C.20)
∂θ sp + τ + m(θ)
g
Recall that Lgc = Ξ(c̃) and Lu = 1 − Ξ(c̃) so that dL c
dc̃
= ξ(c̃) > 0 and dL u
dc̃
= −ξ(c̃) < 0
(where ξ is the pdf of the distribution of connection costs). It can also be easily verified that
∂Ac
∂Lgc
< 0.
dc̃
Equations (C.17) and (C.18) can be used to solve for dθ :

dc̃ − ∂OO∂θ
= <0 (C.21)
dθ r + τ − ∂A c
∂Lgc
ξ(c̃) + ∂OO
∂c̃

Plugging the above into (C.17) we get:

" #
dOO ∂OO r + τ − ∂A
∂Lgc
c

= >0 (C.22)
dθ ∂θ r + τ − ∂A c
∂Lg
ξ(c̃) + c
∂OO
∂c̃

For private and unconnected public-sector jobs to exist it must be the case that Eup − Uu > 0
and Eup −Uu > 0, respectively, which ensures that they generate positive profits. This ensures
that ∂OO
∂θ
> 0. As can be seen from (C.19) sufficient (but not necessary) condition to ensure
also that ∂OO
∂c̃
> 0 is Eug − Eup > 0 meaning that the value to an unconnected worker is
higher when that worker is working for the government than in the private sector. If this
condition holds then we know for sure that the term in the bracket of (C.22) is positive and
thus, dOO
dθ
> 0, while as shown in (C.21) dθdc̃
< 0. If dOO
dθ
> 0 holds, then the right-hand-side
of (C.14) is decreasing while its left-hand-side is increasing in θ. Equation (C.14) thus pins
down a unique equilibrium value for θ, which can then be used to solve for c̃ and the rest of
the endogenous variables.

C.5 Proof of Proposition 4

From (C.14) and (C.15) we get:

   
∂OO ∂Ac ∂OO ∂Ac
dθ ∂wg
r+τ − ∂Lgc
ξ(c̃) + ∂c̃ ∂wg
= −    (C.23)

dwg r+τ − ∂Ac
ξ(c̃) ∂OO
− κq 0 (θ)
(r + sp + τ) r + τ − ∂Ac
ξ(c̃) + ∂OO
∂Lgc ∂θ q 2 (θ) ∂Lgc ∂c̃
∂Ac
dc̃ ∂wg
− ∂OO
∂wg
(1 − B)
= (C.24)
dwg r + τ + ∂OO
∂c̃
(1 − B) − ∂A c
∂Lgc
ξ(c̃)

x
 ∂OO
where B ≡ ∂OO
∂θ
κq 0 (θ) .
∂θ
− 2 (r+sp +τ )
q (θ)

As shown above (in Section C.4) ∂OO ∂θ

> 0 and ∂OO∂c̃
> 0 while q 0 (θ) < 0. These imply
that the denominators in the above expressions are always positive. We know in addition
∂Ac ∂Ac ∂OO
that ∂w g > 0, ∂Lg < 0 and ∂w g > 0, meaning that the numerator in the bracket of (C.23)
c
is positive also. Further, from (C.15) we can verify that for c̃ > 0 it must be the case that
∂Ac
∂wg
− ∂OO
∂wg
> 0, which ensures, also, that the numerator of (C.24) is positive. It follows,
then, that
dθ
< 0 (C.25)
dwg
dc̃
> 0 (C.26)
dwg
Using (C.7) we can easily verify that total employment e = eu + µeg will decrease with
an in increase in wg since:
de de dc̃ de dθ
g
= g
+
dw dc̃ dw dθ dwg
and
de
< 0
dc̃
de
> 0 (C.27)
dθ
dc̃ dLgc dLu
Since Lgc = Ξ(c̃), Lu = 1 − Ξ(c̃) and dwg
> 0, it follows that dwg
> 0 and dwg
< 0.

xi
D Heterogenous Workers
D.1 Setup
Since the two types differ in terms of productivity, the value of searching for a job as well
as the value of being employed in the private sector both differ depending on the worker’s
ability. Thus,
p
 p p

(r + τ )Uu,i = b + m(θi ) Eu,i − Uu,i (D.1)
p p
= wip − sp Eu,i p
 
(r + τ )Eu,i − Uu,i (D.2)

where i = [h, l] denotes high- and low-ability.

The value of a private-sector vacancy and job, respectively, of type-i is given by

rVip = −κ + q(θi ) [Jip − Vip ] (D.3)

rJip = yi − (sp + τ ) [Jip − Vip ] (D.4)
p p p
Using the Bellman equations and the Nash bargaining conditions βSu,i = Eu,i − Uu,i ,
p p p
(1 − β)Su,i = Ju,i − Vu,i we can write the surplus of a private job of type i as:

p yi − b
Su,i = (D.5)
r + τ + sp + βm(θi )

There are now two free-entry conditions one for each private-sector market, Vip = 0.
Imposing the free-entry and Nash bargaining conditions gives the conditions shown in the
main text:
κ (yi − b)(1 − β)
= . (D.6)
q(θi ) r + τ + sp + βm(θi )
wip = b + β(yi − b + κθi ). (D.7)
It can be easily verified that θh > θl because yh > yl . It follows also that whp > wlp and
high-ability workers enjoy a higher value when applying to the private sector than lower
p p p p
ability workers. That is, Eu,h > Eu,l and Uu,h > Uu,l . Moreover, following Lemma 1, θh and
θl are both independent of public-sector employment and wage policies.
In the public sector, despite wages being the same for all workers irrespective of ability,
the values of being employed or unemployed for workers that do not have connections differ
by ability, because the government screens candidates and gives preference to high-ability
workers.
g
= b + mgu,i Eu,i
 g g

(r + τ )Uu,i − Uu,i (D.8)
g g g
 g g

(r + τ )Eu,i = w − s Eu,i − Uu,i . (D.9)

In the connection market, we assume that ability does not affect your probability of being
hired, as long as workers have connections. Since in the connections sector not only both
types receive the same wage, but have also equal chances of getting a job, given by mgc .
As such, the values of unemployment and employment for connected workers are the same

xii
irrespective of ability and remain as in (4) and (6).
High- and low-ability workers without connections can search in either the public or the
private sector. In an equilibrium where both markets are active, the values of these two
options would have to equate:
p g
Uu,i = Uu,i = Uu,i (D.10)
These two conditions would determine the numbers of high- and low-ability unconnected
searchers in the public sector, ugu,h and ugu,l . Using the Nash bargaining conditions and (42)
we can write
βθi κ
(r + τ )Uu,i = b + (D.11)
(1 − β)
Alternatively they can use connections to get into the public sector after paying the cost
c drawn from Ξ(·). The threshold level of cost at which a worker of type-i is indifferent
between using and not using connections to find a government job is

c̃i = Uc − Uu,i (D.12)

These two thresholds determine the allocation of each type of worker into those who use
connections to get government jobs and those who search either in the private or in the
public sector without connections. We can measure each of these four groups’ share in the
labor force as:

Lgc,i = Xi Ξ(c̃i )
Lu,i = Xi (1 − Ξ(c̃i )) (D.13)

where it may be recalled that Lu,i = Lgu,i + Lpi and Xi is the fraction of type-i workers in the
labor force. Finally, population add up constraints are satisfied so that

Lgu,i = (1 − µ)eg + ugu,i

Lgc,i = µeg + uc,i
Lpu,h + Lgu,h + Lgc,h = Xh
Lpu,l + Lgu,l + Lgc,l = Xl (D.14)

and by equating the flows in to the flows out of unemployment in the private sector we obtain

sp + τ
 
p
ui = Lpu,i (D.15)
sp + τ + m(θi )

D.2 Wage Cutoffs

The cutoff wage wgu,h is such that a high-ability (unconnected) worker is indifferent between
searching for a public or a private job, given mgu,h = 1. Setting mgu,h = 1 in (D.8) and using
(D.10), (D.11) gives
βκθh
wgu,h = (r + sg + τ + 1) + b (D.16)
(1 − β)

xiii
 g
If the public sector wage equals wgc,h then Uu,h g
= Uc,h and a high-ability worker is in-
different between using connections or not. Setting (D.11) equal to (4) and solving for the
wage gives:
r + sg + τ
 
g βκθh
wc,h = +1 +b (D.17)
(1 − β) mgc
Comparing (D.16) to (D.17) reveals that wgc,h > wgu,h since mgc < 1. We know that at any
wage above wgc connected workers will be queuing up for government jobs, meaning that
mgc < 1.
Note that for wgc,h to exist it must be the case the µ̄ is large enough. Incentives for
high ability workers to use connections are at the maximum when queues for government
jobs through the regular (unconnected) channel are as long as possible. This occurs when
Lgu,h = Xh , meaning that all high ability workers search in the public sector. The threshold
wage wgc,h will not exist, if, even in this case, where the unconnected public market is as
g g
crowded as possible, Uu,h > Uc,h , meaning that no high type wants to obtain connections.
Lgc
g g
Notice that Uu,h > Uc,h implies mgu,h > mgc , which gives µ̄ < Lgc +L g . Setting Lgu,h = Xh gives
u,h
g g
µ̄ < LgcL+X
c
h
. If µ̄ ≤ LgcL+X
c
h
then no matter how large wg is, the high-type workers will never
opt for connections; in other words, wgc,h does not exist. This threshold wage exists only if
g
µ̄ > LgcL+X
c
h
.

D.3 Proof of Proposition 5

Using (D.10) (with (D.8) and (D.11) substituted in) to solve for Lgu,i we obtain:
" !#
wg − b
Lgu,i = (1 − µ̄)eg λ + (1 − λ) β
(D.18)
1−β
κθi

where λ = r+srg +τ . Using (D.12) with either i = l or i = h and (4) and (D.11) substituted in
to solve for Lgc , we get:
" !#
g
w − b
Lgc = µ̄eg λ + (1 − λ) β
(D.19)
c̃l (r + τ ) + 1−β κθl
" !#
g
w − b
Lgc = µ̄eg λ + (1 − λ) β
(D.20)
c̃h (r + τ ) + 1−β κθh

which implies:
βκ βκ
c̃l (r + τ ) + θl = c̃h (r + τ ) + θh (D.21)
(1 − β) (1 − β)
Given wg > wgu,h , meaning that all unconnected workers attached to the public sector are
of high ability (Lgu = Lgu,h ), the total number of workers in the public sector is Lg = Lgu,h +Lgc .
g dLg g
Thus, dL
dµ̄
= dµ̄u,h + dL c
dµ̄
.
Notice that Lc = Lc,l = Xl Ξ(c̃l ) (Lgc,h = 0) if wg ≤ wgc,h (c̃h ≤ 0) and Lgc = Lgc,l + Lgc,h =
g g

xiv
Xl Ξ(c̃l ) + Xh Ξ(c̃h ) if wg > wgc,h (c̃h > 0). As can be verified from (42) dθ i
dc̃i
= 0, i = [h, l] so
dc̃l dc̃h
that we know from (D.21) that dµ̄ = dµ̄ . We can then use either (D.20) or (D.19) to get:

dLgc Lgc
   
Xl ξ(c̃l ) + Xh ξ(c̃h ) Xl Ξ(c̃l ) + Xh Ξ(c̃h ) Xl ξ(c̃l ) + Xh ξ(c̃h )
= = > 0 if wg > wgc,h
dµ̄ µ̄ Xl ξ(c̃l ) + Xh ξ(c̃h ) + ∆ µ̄ Xl ξ(c̃l ) + Xh ξ(c̃h ) + ∆
dLgc Lgc
   
Xl ξ(c̃l ) Xl Ξ(c̃l ) Xl ξ(c̃l )
= = > 0 if wg ≤ wgc,h (D.22)
dµ̄ µ̄ Xl ξ(c̃l ) + ∆ µ̄ Xl ξ(c̃l ) + ∆
µ̄eg (1−λ)(wg −b)(r+τ ) µ̄eg (1−λ)(wg −b)(r+τ )
where ∆ = β
(c̃l (r+τ )+ 1−β κθl )2
= β
(c̃h (r+τ )+ 1−β κθh )2
>0
From (D.18) we get:
dLgu,h Lgu,h
=− <0 (D.23)
dµ̄ 1 − µ̄
Thus,

 g
dLg Lu,h Lgc
 
Xl ξ(c̃l )
= − − if wg ≤ wgc,h
dµ̄ 1 − µ̄ µ̄ Xl ξ(c̃l ) + ∆
g
 g
Lu,h Lgc
 
dL Xl ξ(c̃l ) + Xh ξ(c̃h )
= − − if wg > wgc,h (D.24)
dµ̄ 1 − µ̄ µ̄ Xl ξ(c̃l ) + Xh ξ(c̃h ) + ∆

The terms in the parentheses of the above expressions are less than 1. Further, it can be
Lgu,h g
easily verified from (D.18) (with i = h) and (D.20) that 1−µ̄ T Lµ̄c when c̃h T 0 (wg T wgc,h ).
Therefore,
dLg g
dµ̄
< 0, du
dµ̄
< 0, if wg > wgc,h
dLg g
dµ̄
S 0, du
dµ̄
S 0, if wg ≤ wgc,h

where it may be recalled that ug = ugu + ugc = Lg − eg .

The total employment rate in the model with worker heterogeneity is given by e =
eg + eph + epl , where eg is exogenously set by the government, while eph and epl can be derived
m(θi )Lpi
from (D.14) and (D.15): epi = sp +τ +m(θ i)
, i = [h, l] so that

m(θh )(Xh − Lgh ) m(θl )(Xl − Lgl )

g
e=e + p + p (D.25)
s + τ + m(θh ) s + τ + m(θl )

We know that for wg > wgu,h all low-ability workers in the public sector are connected.
dLg
Hence, Lgl = Lgc,l = Xl Ξ(c̃l ), and clearly dµ̄l > 0, since dc̃
dµ̄
l
> 0. But the composition of the
g
high-ability workers attached to the public sector (Lh ) in terms of connections depends on
whether wages are above or below wgc,h .
If wg > wgc,h (> wgu,h ), then Lgh = Lgc,h + Lgu,h , where Lgc,h = Xh Ξ(c̃h ). Since dc̃
dµ̄
h
> 0 then
dLgc,h dLgu,h
dµ̄
> 0, while, as shown above, dµ̄
< 0. Moreover, as shown above, for wg > wgc,h ,

xv
dLg
dµ̄
< 0 which means that the decrease in Lgu,h , dominates over increase in Lgc,h and Lgc,l .
It follows, then, that the decrease in Lgh dominates over the increase in Lgl . This, together
with the fact that m(θh ) > m(θl ) implies dµ̄ de
> 0. That is, epl decreases and eph increases
with the increase in µ̄ but the increase in eph is larger than the decrease in epl and thus total
employment increases.
g
If wg ≤ wgc,h we can not show that dµ̄
de
> 0 because the sign of dL dµ̄
is ambiguous. In this
case Ll = Lc,l and Lh = Lu,h and we have an increase in Lc,l and decrease in Lgu,h , but we
g g g g g

do not know if the decrease in Lgu,h dominates over the increase in Lgc,l . To sum up:

deph depl
dµ̄
> 0, dµ̄
de
< 0, dµ̄ S 0 if wg ≤ wgc,h
deph depl
dµ̄
> 0, dµ̄
de
< 0, dµ̄ > 0 if wg > wgc,h

D.4 Effects of increasing wg when wg > wgu,h

First, let us show that the number of workers searching in the public sector increases as the
dLg dLp
public-sector wage increases; that is dw g > 0, which ultimately implies that dw g < 0, since

Lp = 1 − Lg . Recall also that for wg > wgu,h all workers searching for public jobs without
connections are of high ability. The total number of workers attached to the public sector
is given by Lg = Lgu + Lgc where Lgu = Lgu,h and Lgu,h is as given in (D.18) (with i = h),
while Lgc = Lgc,h + Lgc,l = Xh Ξ(c̃h ) + Xl Ξ(c̃l ) if wg > wgc,h (> wgu,h ) and Lgc = Lgc,l = Xl Ξ(c̃l ) if
wg ≤ wgc,h .

dLg dLgu,h dc̃h dc̃l

= + X h ξ(c̃ h ) + X l ξ(c̃ l ) , if wg > wgc,h
dwg dwg dwg dwg
dLg dLgu,h dc̃l
g
= g
+ Xl ξ(c̃l ) g , if wg ≤ wgc,h (D.26)
dw dw dw
dLg dL g
It is straightforward to verify from (D.18) that dwu,h g > 0. To show that dw g > 0 we need to
dc̃i
show further that dw g > 0, i = [h, l].

We can use (D.12), together with (4), (6) and (D.11) to derive:
µ̄(sg +τ )eg
" #
1 Lgc −µ̄eg g 1 βκθi
c˜i − µ̄(s g +τ )eg (w − b) = , i = [h, l] (D.27)
r + τ r + τ + sg + g g r + τ (1 − β)
Lc −µ̄e

and obtain:

dc̃l M
= M (1−M )(wg −b)
≥ 0, if wg ≤ wgc,h
dwg r+τ + Xl ξ(c̃l )
Lgc −µ̄eg
dc̃l dc̃h M
g
= = M (1−M )(wg −b)
≥ 0, if wg > wgc,h (D.28)
dw dwg r+τ + (Xl ξ(c̃h ) + Xh ξ(c̃l ))
Lgc −µ̄eg

xvi
 µ̄(sg +τ )eg
g
Lc −µ̄eg dc̃i
where M = g µ̄(sg +τ )eg . It is evident from the above equations that dwg
> 0, i = [h, l]
r+τ +s + g
Lc −µ̄eg
dc̃i
only if µ̄ > 0, while dwg
= 0 if µ̄ = 0.
It follows that:
dLg dLgu,h
dwg
> 0, dwg
>0
dLgc
dwg
> 0, if µ̄ > 0
dLgc
dwg
= 0, if µ̄ = 0
dLg dLg dLgl
Therefore dwhg > 0 and dwlg > 0 if µ̄ > 0 and dwg
= 0 if µ̄ = 0. From (D.25) and given
ug = Lg − eg we can also write:
de du g dep
dwg
< 0, dw g > 0, dw g < 0
h

depl
dwg
< 0, if µ̄ > 0
depl
dwg
= 0, if µ̄ = 0 (D.29)

D.5 Proof of Proposition 6

Let us firs consider the case wgc < wg ≤ wgu,h where there are only low-type workers in the
public sector. As summarized in Proposition 1, the private-sector labor force and employment
increase with an increase in µ̄ due to some workers who would otherwise search without
connections for public jobs choosing to search for private jobs instead. Since all workers in
the public sector are of low ability the additional workers entering the private sector are of
low ability, thereby lowering average ability in the private sector, while all workers in the
public sector remain of low ability.
Consider next the case where wgc,h ≥ wg > wgu,h where all low ability workers attached to
the public sector are connected while all high ability workers attached to the public sector
have no connections. In this case, as shown above (see Section D.3) an increase in µ̄ will
dLgc,l
attract more low-ability workers into the public sector ( dµ̄
> 0) and will drive high ability
dLgu,h
workers away from the public and into the private sector ( dµ̄ < 0). This means that the
skill composition of employment/labor force in the public sector deteriorates, while those in
the private sector improve.
When wg > wgc,h , there are also some high-ability workers in the connections sector. An
increase in µ̄ in this case will induce more of the high-ability workers to get connections and
move into the public sector. However, as shown above (Section D.3), the decrease in the
number of high-ability unconnected workers attached to the public sector dominates over
the increase in the number of connected (high or low ability workers) attached to the public
sector. Hence, in this case also, the skill composition in the public sector deteriorates while
that in the private sector improves.

xvii
E Competitive Search in the Private Sector
Suppose now that, as in the benchmark model, the two sectors, private and public, are
segmented. However, we depart from the assumptions of Nash bargaining and random
search in the private sector. Instead, as in Moen (1997), we introduce a competitive search
equilibrium in the private sector. To this end, we assume that the private-sector market
consists of submarkets with different posted wages and equilibrium tightness.
In each submarket, there is a subset of unemployed workers and firms with vacant jobs
that are searching for each other. A matching function determines the number of matches
in each submarket. The number of matches in submarket n is m(vn , un ) = (vn )η (upn )(1−η) ,
m(θn ) is the job finding rate and q(θn ) the job filling rate. Unemployed workers are free to
move between submarkets. They choose to search for a job in the submarket that yields
the highest expected income. Since workers are ex-ante identical, and movement across
submarkets is free, in equilibrium, the value of search is equal across submarkets. A market
maker determines the number of submarkets in each market and the wage in each submarket.
The wage is chosen to maximize the value of a vacancy, and since all vacancies in the same
submarket are identical, they offer the same wage. There is free entry of vacancies in each
submarket, which drives the value of a vacancy to zero, and determines the number of
vacancies posted in each submarket.
We present next the full set of Bellman equations describing the optimal behavior of
workers and firms, the equilibrium conditions and the model solution. For a worker in
submarket n
p
 p p

(r + τ )Uu,n = b + m(θn ) Eu,n − Uu,n (E.1)
p
= wnp − sp Eu,n p p
 
(r + τ )Eu,n − Uu,n (E.2)

Unemployed workers are free to move between submarkets. They will choose to search
for a job in the submarket that yields the highest expected income. Since workers are ex-ante
p
identical and movement across submarkets is free, this means that Uu,n = Uup . Using (E.1)
and (E.2) we can write:

(r + τ )Uup − b
 
m(θn ) = p p (r + τ + sp ) (E.3)
wn − (r + τ )Uu

The values of vacancies and filled jobs in submarket n satisfy

p
= −κ + q(θn ) Jup (wnp ) − Vu,n
p
 
rVu,n (E.4)
rJup (wnp ) = yn − wnp + (sp + τ ) Vu,n
 p
− Jup (wnp )

(E.5)
p
Using (E.4) and (E.5) to solve for Vu,n gives

p −κ(r + sp + τ ) + q(θn )(yn − wp )

rVu,n = (E.6)
r + q(θn ) + sp + τ
In a competitive search equilibrium a market maker determines the number of submarkets
in each market and the wage in each submarket. The wage is chosen to maximize the value of

xviii
a vacancy. All vacancies in the same submarket offer the same wage. Setting the derivative
of (E.6) with respect to wnp equal to 0 we get the first order condition for optimal wages:

dθn
−(1 − η)(r + sp + τ ) [yn − wnp + κ] = θn (r + sp + τ ) + m(θn ) (E.7)
dwnp
There is free entry of vacancies in each submarket, which drives the value of a vacancy to
p
zero. Setting Vu,n = 0 in (E.6) gives:

κ yn − w p
= (E.8)
q(θn ) r + sp + τ

Taking the derivative of (E.3) with respect to wnp we obtain

 
dθn θn 1
p = − p p (E.9)
dwn wn − (r + τ )Uu η
dθn
Using (E.8) and (E.9) to substitute for κ and dwn
p , respectively, in (E.7) and then solving for
p
wn we get
wnp = (1 − η)yn + η(r + τ )Uup (E.10)
Using (E.3) and (E.8) we can substitute for (r + τ )Uup in (E.10) and obtain

wnp = b + (1 − η) (yn − b + θn κ) (E.11)

Substituting wnp from (E.11) into (E.8) we get the job creation condition in each submarket

κ η(yn − b)
= p
(E.12)
q(θn ) r + s + τ + (1 − η)m(θn )

Notice that if yn = y, meaning that productivity is the same across all submarkets then
θn = θ and wnp = wp . All submarkets offer the same wage and job finding rate. If in addition
the Hosios condition holds, i.e. 1 − η = β, then job creation, market tightness and the Nash
bargaining wage in the Benchmark model described in the text (see equations 13 and 14) are
identical to those derived under competitive search. Hence, the results discussed in Sections
4 and 5 carry over to this alternative assumption of competitive search in the private sector.

xix
F Connections Premium
In the benchmark model, we consider that connected and unconnected workers enjoy the
same benefits of working in the public sector. We also assume that the costs incurred by the
newborns to get connections were wasted. We now assume that newborn pay connections
costs to current connected public-sector workers so that current workers will help fast-track
them into the public sector. These payments are the “connections premium”, Υ, which will
further raise the value of working in the public sector for connected workers. We describe
here the basic set up and in Section H we compare quantitative results in this alternative
setup to those obtained in the benchmark model, in which no such connections premium
exists.

(r + τ )Ecg = wg + Υ − sg [Ecg − Ucg ] . (F.1)

In equilibrium, this connections premium depends on the threshold R c̃ of connections costs,

Υ = Υ(c̃). The total connections cost paid by newborns is τ 0 cξ(c)dc, where ξ is the
pdf of the distribution of connection costs. To avoid creating further interactions between
sectors, we assume that newborns’ total connections cost is divided equally among connected
workers:
R c̃
τ 0 cξ(c)dc
Υ(c̃) = . (F.2)
µ̄eg
In principle, this extension could create multiple equilibria, with people expecting high re-
turns of connections investing in connections (creating a lot of side payments) or people
expecting low returns of connections not investing in connections (generating few side pay-
ments). We show, below, that provided some regularity conditions on the distribution of
connections costs are satisfied, there are no multiple equilibria.
With the introduction of a connection premium all other Bellman equations but the
value of being employed in the public sector for a connected worker (Ecg ) remain as in the
Benchmark model described in Section 3. It follows that all equilibrium conditions remain the
same, but equations (23) that determine the cut-off connection costs. The cut-off connection
cost now change to take into account that the existence of a connection premium increases
the value of being a connected and employed public employee. In particular, equation (23)
becomes:
µ(sg +τ )eg
" R c̃ ! #
1 Lgc −µeg τ cξ(c)dc βκθ
c̃ = wg − b + 0 g − (F.3)
r + τ r + τ + sg + µ(sgg +τ )eg g µe (1 − β)
Lc −µe

As shown in Appendix A.2, equations (13) gives unique equilibrium value for θ. To guarantee
the existence and uniqueness of a steady-state condition we need to show that with the
equilibrium value of θ substituted in, equation (F.3) gives a unique equilibrium value for c̃.

xx
Rearranging terms in (F.3) we can write:
µ(sg +τ )eg
" R c̃ !#
1 Lgc −µeg τ cξ(c)dc 1 βκθ
c̃ − µ(s g +τ )eg wg − b + 0 g = (F.4)
r + τ r + τ + sg + g g µe r + τ (1 − β)
Lc −µe

Since the right-hand-side of the equation above is independent of c̃ a unique equilibrium

value of c̃ exists if the left-hand-side of the equation above is increasing in c̃. Sufficient (but
not necessary) condition for the left-hand-side of (F.4) to be increasing in c̃ is:

mgc τ c̃ξ(c̃)
1− g >0
r + τ + s + mc r + τ µeg
g

Sufficient but not necessary condition for the above inequality to be always satisfied is

c̄ξ(c̄) ≤ µeg

xxi
G Survey data

Table G1: Quality of government survey - European Countries

Country QoG Indexes Aggregate public-
Skills and Merit Political connections Personal connections private wage ratio
Austria 5.00 5.00 4.67 1.72
Belgium 5.71 3.00 2.14 1.40
Bulgaria 3.23 5.38 5.00 1.96
Croatia 4.20 4.10 3.80
Cyprus 3.20 5.40 5.40 2.41
Czech Republic 5.00 4.11 4.10 1.33
Denmark 6.29 1.57 2.60 1.08
Estonia 4.67 3.33 3.56
Finland 6.00 3.33 2.50 1.16
France 5.67 2.64 3.00 1.16
Germany 5.89 2.62 2.40 1.38
Greece 4.13 3.87 3.73 2.43
Hungary 3.67 5.07 4.60 1.36
Iceland 4.83 2.83 2.83 1.61
Ireland 6.64 1.82 2.45 2.31
Italy 3.25 4.25 4.20 2.05
Latvia 4.60 3.40 3.60 1.36
Lithuania 4.88 3.56 3.44 1.14
Luxembourg 0.88
Malta 3.75 4.50 3.50 1.48
Netherlands 6.08 2.46 2.68 1.92
Norway 6.40 2.07 1.87 0.98
Poland 5.50 2.80 3.20
Portugal 4.25 4.63 4.56 2.23
Romania 4.24 4.94 4.33
Slovakia 2.67 4.78 5.56 1.26
Slovenia 4.38 3.63 4.13 1.51
Spain 5.04 3.17 3.17 2.05
Sweden 5.92 1.92 2.69 0.93
Switzerland 6.20 3.00 2.80
United Kingdom 5.79 2.82 2.88 1.09
Note: Indexes of recruitment practices are takes from the Quality of Government Survey. Data on govern-
ment and private sector employment is from EUROSTAT and OECD. Data on government wage bill and
private sector wage bill is from AMECO.

xxii
 Table G2: Quality of government survey - World regions
Region QoG Indexes
(no of countries) Skills and Merit Political connections Personal connections
Eastern Europe and post Soviet Union (25) 3.71 4.54 4.28
Latin America (16) 3.44 4.86 4.51
North Africa and the Middle East (11) 3.32 4.71 4.14
Sub-Saharan Africa (25) 3.60 4.92 5.11
Western Europe and North America (22) 5.51 2.92 1.99
East Asia (4) 5.32 2.87 3.08
South-East Asia (7) 4.35 4.44 4.57
South Asia (6) 3.78 4.78 5.67
The Pacific (1) 3.66 5.00 4.83
The Caribbean (3) 4.00 4.08 3.75
Note: Indexes of recruitment practices are takes from the Quality of Government Survey. Average for
different world regions.

Table G3: Regression of the unemployment rate

Baseline variables Alternative variables
(1) (2) (3) (4) (5) (6)
(1) (2) (3) (4) (5) (6)
Public-sector wage premium 19.2*** 24.6*** 49.0*** 43.0*** 41.8***
(3.31) (4.29) (4.93) (5.78) (4.23)
Ratio of indexes of non-meritocracy -20.4*** -20.4***
(-1.71) (-3.12)
× High public wage -23.2*** -6.51*** -18.1***
(-3.70) (-4.24) (-3.61)
× Low public wage -18.8*** -0.21 -14.1*
(-3.02) (-0.17) (-2.73)
Observations 70 70 70 70 70 70
R-squared 0.14 0.041 0.248 0.335 0.343 0.325
Notes: The t-statistics are shown in brackets.*** indicates significance at the 1% level, ** at 5% level, and * at the
10% level. The dependent variable is the unemployment rate. The ratio of the non-meritocracy index for the public
sector over the index for the private sector, increases when the public sector is perceived to be less meritocratic
than the private sector. The index is constructed with data taken from European Quality of Government Index
dataset. The public-sector wage premium is estimated with microdata from the 2010 Structure of Earnings
Survey. Unemployment rate is taken from Eurostat. In column (5) we use an alternative index which is the
difference between the index for the public over the index for the private. In column (6) we use an alternative
index which is the ratio between the index for the public sector (answer by only public sector workers) over the
index for the private sector (answered by only private sector workers).

xxiii
H Numerical exercise
It is not our objective in this section to do a full quantitative exercise. Such an exercise
would face the same problems as the empirical work on non-meritocratic hiring discussed in
the introduction – the identification of the fraction of public-sector workers hired through
connections and the distribution of connection costs would be problematic. However, we
think that a simple numerical exercise can help us improve our understanding of the model.
The objective of our numerical exercise is threefold. First, we want to inspect whether under
a reasonable parametrization, conditional on an inefficient wage policy, hiring through con-
nections increases or decreases welfare. Second, given the endogenous limits that government
policies place on µ discussed in Section 4.2, and given a set of parameters, we might be in a
region where: i) µ in not constrained and is equal to µ̄ or ii) µ is constrained. Changes in
government policy may switch the economy from one region to the other, making it difficult
to solve for their effect in the full model analytically. In our quantitative exercise we account
for such switches and are able to characterize the full effect of policy changes. Finally, we
want to compare the benchmark model with the alternative models proposed in Section 7 –
in particular, to compare the transmission mechanisms under the assumptions of segmented
markets and random search. We have also done simulations changing the deep parameters of
the model is Section 6, but the effects simply boil down to combinations of different policies.

H.1 Parametrization
We parameterize our benchmark model with segmented markets to match the Spanish econ-
omy at a quarterly frequency, drawing largely on the Spanish Labour Force Survey (SLFS)
and the Structure of Earnings Survey (SES) microdata for the period 2005-2015. A set of
parameters is directly fixed to values taken from the data, while a second set of parameters
targets steady-state values. We chose Spain because it is one of the countries where there is
widespread anecdotal evidence of nepotism and chronism.7 Table H1 lists all the parameters,
their values and the data sources.
From the Spanish Labour Force Survey, we calculate the stocks and flows of public- and
private-sector workers and the unemployed. These are shown in Figure H1. Around 13.2
percent of the labour force works in the public sector (eg = 0.132). Following Fontaine et al.
(2020), we construct data on worker flows to calibrate the separation rates by sector. The
numbers are sg = 0.022 and sp = 0.044, imply that the private sector has a higher separation
rate than the public sector.
We consider, in the private sector, a Cobb-Douglas matching function with matching
efficiency ζ and matching elasticity with respect to the unemployment of η. As the matching
efficiency and the cost of posting vacancies are not separable, we normalize the matching
efficiencies ζ = 1. The costs of posting vacancies, κ is set to target the unemployment rate
of 18 percent, the average of the sample. The matching elasticity is set to the common value
of 0.5, and the Hosios condition is assumed to hold (η = 0.5).
7
Recently the press exposed that in the “Tribunal de Cuentas”, the Spanish institution in charge of
invigilating economic and financial irregularities in the public sector, close to 100 of its 700 workers were
family members or friends of the directors or of important politicians in Spain.

xxiv
 Figure H1: 4-state stocks and flows, Spain
0.07 (p)
Private 0.18 (h) Public
Employment Employment
15521 (t) 2975 (t)
40.9 (p) 0.04 (p) 7.8 (p)
0.51 (h) 0.17 (p)
2.22 (h)
1.41 (p)
3.40 (h)
1.77 (p) 0.18 (p)
4.38 (h) 0.44 (h)

1.96 (p)
21.19 (h) 1.25 (p) 0.19 (p)
3.04 (h) 2.47 (h)
0.22 (p)
2.35 (h)
1.67 (p)
Unemployment 17.60 (h) Inactivity
4034 (t) 15447 (t)
10.6 (p) 40.7 (p)
1.95 (p)
4.80 (h)

Source: Spanish Labour Force Survey, average 2005-2015. The worker stocks and flows are expressed as total
number of people in thousands (t), as a percentage of the working-age population (p) or as a hazard rate (h).
See Fontaine et al. (2020) for details. For the calibration, we excluded the flows from and to inactivity.

We use microdata from the Structure of Earnings Survey, for the waves of 2002, 2006,
2010 and 2014, to calculate the public-sector wage premium. We run regressions of the
log gross hourly earnings on a dummy for the public sector, controlling for region, gender,
age, occupation, year and part-time and find that the premium is 13.9 percent. We set the
g
public-sector wages such that w wp
= 1.139. A recent paper by Dickson et al. (2014) argues
that the lifetime premium in the public sector is lower than that measured by standard
cross-section methods. They report that, in Spain, it is 7.17 percent. We report exercises
using their numbers. We also report the equilibrium under the efficient public-sector wage
g
premium: w wp
= 0.91. The fact that the optimal wage premium is negative reflects mainly
the facts that the expected duration of a job in the public sector is longer.
Salomäki and Munzi (1999) find that the unemployment benefit net replacement rate is
44 percent in Spain. We set b = 0.398 to target this number. Additionally, r = 0.012 and
τ = 0.006 target a yearly interest rate of about four percent and an average working life of
40 years.
The most relevant parameters are the fraction of jobs reserved for people with connec-
tions, µ̄, and the distribution of connections costs, Ξ(·), but identifying them is subject to the
difficulties that prompted us to approach this question from a theoretical angle. Regarding
µ̄, we proxy it with data from the Quality of Government Survey. For Spain, the index for
“skills and merit” is 5 while for both “political” and “personal connections” is 3.2. Dividing
one by the sum of the two, we get µ̄ = 0.4.The distribution of connections costs is assumed to
be uniformly distributed between 0 and 55, set exogenously. This distribution implies that
the deadweight cost of corruption is 0.1 percent of the total consumption of private-sector
goods. Most of the exercises consist of varying these parameters. We vary the parameter
µ̄ from 0 to 1 and consider high and low values for the upper bound of the distribution of
connections costs of c̄ = 10 and c̄ = 100.

xxv
 Table H1: Parameterization of segmented markets model
Fixed parameters Source Values
Government employment Spanish LFS eg = 0.132
Job-separation rate (private) Spanish LFS sp = 0.044
Job-separation rate (public) Spanish LFS sg = 0.022
Matching elasticity Standard η = 0.5
Bargaining power of workers Hosios Condition β = 0.5
Discount rate Standard r = 0.012
Retirement rate Standard τ = 0.006
Matching efficiency Normalization ζ=1
Productivity Normalization y=1
Fraction of connected government jobs Quality of government survey µ̄ = 0.40
Connections costs upper bound Set exogenously c̄ = 55

Other parameters Target (Source) Values

Public-sector wage Public-sector wage premium (SES ) wg = 1.027
Cost of posting vacancies Unemployment rates (LFS ) κ = 6.31
Unemployment benefit Replacement rate (EC ) b = 0.398

In the baseline steady-state, the government can achieve its target fraction µ̄ of jobs,
meaning that µ is unconstrained.

H.2 Effects of nepotism

We start by analyzing the effects of nepotism in public-sector hiring for different combinations
of public-sector wages and connections costs. We take into account that changes in policies
or parameters might trigger the endogenous limit of µ to bind, as determined by equation 29.
Sometimes the government might not be able to fill its targeted µ̄ fraction of jobs through
connections. Figure H2 shows how different variables vary with µ̄ for three different wage
policies: the benchmark policy with premia of 13.9 percent; an intermediate wage policy with
premia of 7 percent; and the efficient wage policy consisting of premia of -9 percent. We
examine the effects on unemployment rates, the fractions of connected workers, and welfare,
calculated as private-sector production net of the connections costs (as in Section 5), relative
to the efficient allocation. As in Gomes (2015), the optimal policy is a negative public-sector
wage premium in order to compensate for the higher relative job security.
Under the efficient wage policy, µ is constrained to be zero. There are no queues for
public-sector jobs and no connections sector. Unemployment rate is roughly 3 percentage
points lower. The higher public-sector wages are responsible for the higher unemployment
and a 2.5 percent lower welfare relative to the efficient scenario.
The graphs reveal that the effects of nepotism seem to be larger the more inefficient the

xxvi
 Figure H2: Effects of nepotism, role of public-sector wages
Welfare Fraction of connected workers Connection cost cutoff

15
1

8
6
.99

10
Index

4
.98

2
.97

0
.05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95
µ µ µ

Unemployment rate Unconnected unemployed in public sector Connected unemployed in public sector

.015
19

.05
18

.04

.01
17

.03
%

.02
16

.005
.01
15
14

0
.05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95
µ µ µ

Note: We vary µ̄ along the x-axis. The dark blue line is the benchmark calibration (wg /wp = 1.139).
The light green line is the scenario with efficient public-sector wages (whg /whp = 0.908). The bright blue
dashed line is the scenario with an intermediate public-sector wage premium (wg /wp = 1.072). Welfare is
expressed as a fraction of the efficient steady state. In the scenario with efficient public-sector wages, µ is
constrained to zero. In all the other scenarios µ is unconstrained. Tightness and wages in the private sector
are constant and independent of public-sector wages or nepotism (θ = 0.06, wp = 0.901).

public-sector wage is. In this numerical exercise, hiring through connections indeed raises
welfare. As shown in Proposition 1, it lowers the unemployment rate. By restricting access
to public-sector jobs to those with connections, workers are discouraged from searching for
unconnected vacancies in the public sector, and turn to the private sector. As tightness
is constant, there is a one-to-one effect on private vacancies. While, indeed, the fraction
of connected workers increases - with the respective increase in deadweight loss - this is
outweighed by the increase in private-sector employment. Thus, welfare increases.
Figure H3 reproduces the same exercise for three levels of connections costs. Again, for
this set of parameters, an increase in µ̄ increases welfare. The increase is larger for high levels
of connection costs. When the connections costs are higher, the connections market becomes
more exclusive. When increasing µ̄, more workers are pushed into the private sector, which
implies larger decreases in unemployment and larger increases in welfare.
In Figure H3, the kink observed for high connections costs reflects the fact that, because
it is so costly to get connections, the endogenous limit binds for µ. As shown in Lemma
2, the minimum wage for the government to be able to fill a fraction µ̄ of jobs through
connections – wgc – is increasing in µ̄. If the public-sector wage is not high enough to sustain
a large connections sector (that is wg < wgc ), the endogenous limits bind and µ is determined
by equation (28), and hence changes in µ̄ do not affect the equilibrium.

xxvii
 Figure H3: Effects of nepotism, role of connections costs
Welfare Fraction of connected workers Connection cost cutoff
.99

20

10
.985

15
Index
.98

%
10

5
.975

5
.97

0
.05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95
µ µ µ

Unemployment rate Unconnected unemployed in public sector Connected unemployed in public sector
19

.05

.04
.04
18

.03
.03
%
17

.02
.02

.01
16

.01
15

0
.05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95
µ µ µ

Note: We vary µ̄ along the x-axis. The dark blue line is the benchmark calibration (c̄ = 55). The light
green line is the scenario with low connections costs (c̄ = 10). The bright blue dashed line is the scenario
with high connections costs (c̄ = 100). Welfare is expressed as a fraction of the efficient steady state. In the
scenario with high connections costs, µ becomes constrained. Tightness and wages in the private sector are
constant and independent of public-sector wages or nepotism (θ = 0.06, wp = 0.901).

H.3 Effects of policies

Figure H4 shows the effect of public wages, for three levels of target µ̄: 0.2, 0.40 and 0.8. In
general, decreasing public wages raises welfare, since, as outlined in Proposition 2, cutting
them has a positive effect on the employment rate. A 10 percent cut in the wages of public-
sector workers lowers the unemployment rate by 1.5 percentage points. However, for some
combination of parameters (high µ̄), there is a region in which µ becomes constrained. In
that region welfare declines with wage cuts. This happens because, as shown in Proposition
3, in the constrained case µ decreases with wage cuts. Decreasing µ means freeing up public
jobs for job searchers that do not have connections. This pushes more unemployed workers
to queue for public-sector jobs, and increases the unemployment rate.
Figure H5 shows the effects of increasing public-sector employment. The effect of in-
creasing public-sector employment on the selection of workers into the two sectors resembles
those of increasing public-sector wages. In both cases the value of searching in the public
sector goes up and this drains workers from the private to the public sector. The fact that
higher public-sector employment lowers welfare follows trivially from the lack of assumption
on the value of public-sector production. What is interesting to notice is that it can increase
or decrease unemployment, depending on the level of nepotism. In line with the results out-
lined in Proposition 2 for the case of increasing wg , high nepotism prevents large increases in
the queues for public-sector jobs, which helps reduce unemployment. Conversely, when most
public-sector jobs are available to unconnected workers, more job openings at high wages,
attract a disproportionate number of searchers raising unemployment.

xxviii
 Figure H4: Effects of public-sector wages
Welfare Fraction of connected workers Connection cost cutoff

15
1

8
.99

6
10
Index
.98

4
5
.97

2
.96

0
.9 .95 1 1.05 1.1 1.15 1.2 .9 .95 1 1.05 1.1 1.15 1.2 .9 .95 1 1.05 1.1 1.15 1.2
g p g p g p
w /w w /w w /w

Unemployment rate Unconnected unemployed in public sector Connected unemployed in public sector
.06
20

.02
.015
.04
18
%

.01
.02
16

.005
14

0
.9 .95 1 1.05 1.1 1.15 1.2 .9 .95 1 1.05 1.1 1.15 1.2 .9 .95 1 1.05 1.1 1.15 1.2
g p g p g p
w /w w /w w /w

Note: The dark blue line is the benchmark calibration (µ̄ = 0.4). The light green line is the scenario with
low nepotism (µ̄ = 0.2). The bright blue dashed line is the scenario with high nepotism (µ̄ = 0.8). Welfare
is expressed as a fraction of the efficient steady state. In all scenarios, when skilled public-sector wages are
low, µ becomes constrained. Tightness and wages in the private sector are constant and independent of
public-sector wages or nepotism (θ = 0.06, wp = 0.901).

Figure H5: Effects of public-sector employment

Welfare Fraction of connected workers Connection cost cutoff
1.15

20

10
1.1

8
15
1.05

6
Index

%
10
1

4
.95

2
.9

0 .05 .1 .15 .2 0 .05 .1 .15 .2 0 .05 .1 .15 .2

g g g
e e e

Unemployment rate Unconnected unemployed in public sector Connected unemployed in public sector
.015
.08
20
19

.06

.01
18

.04
%
17

.005
.02
16
15

0 .05 .1 .15 .2 0 .05 .1 .15 .2 0 .05 .1 .15 .2

g g g
e e e

Note: The dark blue line is the benchmark calibration (µ̄ = 0.4). The light green line is the scenario with
low nepotism (µ̄ = 0.2). The bright blue dashed line is the scenario with high nepotism (µ̄ = 0.8). Welfare
is expressed as a fraction of the efficient steady state. In all scenarios, µ is never constrained. Tightness
and wages in the private sector are constant and independent of public-sector wages or nepotism (θ = 0.06,
wp = 0.901).

xxix
H.4 Comparing different models
We now compare the results from the baseline segmented market model with those from
the alternative models discussed in Section 7. For the model in which search in the public
and private sectors is random, we reparameterize the cost of posting vacancies to target the
steady-state unemployment rate (κ = 7.31). We follow the same procedure for the model
with a connections premium (κ = 6.29). Once recalibrated, the steady state of the remaining
variables is very close to that of the benchmark model.
Table H2 shows the effects of three different policies: i) a decrease in µ̄ from 0.4 to 0.2;
ii) an increase in µ̄ from 0.4 to 0.6; and iii) a ten-percent decrease in public-sector wages.
We start by comparing the model with segmented markets with the model of random
search. Graphs with a more detailed comparison are shown in Figure H6. We can see in the
table that random search in the labor market weakens the effects of policies on unemploy-
ment. Although the effects go in the same direction, the mechanisms at work are different.
Under random search, nepotism affects tightness (θ) positively and private wages negatively.
By having fewer unconnected vacancies, the outside option of an unemployed worker bar-
gaining with a firm is weaker, pushing wages down and raising job creation. This effect on
private wages raises the public-sector wage premium endogenously.
As discussed above, the effect of µ̄ on welfare are ambiguous. As Figure H6 shows, under
this parametrization, and in contrast with segmented markets, the effect is negative. When
we move from µ̄ = 0.40 to µ̄ = 0.20, welfare increases by 0.22 percent. When moving to
µ̄ = 0.60, welfare also increases but marginally.
Turning, now, to the model with connections premium, it tends to amplify the effects
of policies on the number of connected workers, but because the premium represents only
1.3 percent of public-sector wages, the effects are quantitatively similar to those in the
benchmark model.

Table H2: Effects of policies under different models

Policy Segmented Random Connections
Policy markets search premium
Reduction of nepotism to µ̄ = 0.20
%∆ welfare -0.28% 0.22% -0.31%
∆ fraction of connected -3.10 p.p. -3.12 p.p. -3.16 p.p.
∆ unemployment rate 0.48 p.p. 0.12 p.p. 0.42 p.p.
Increase of nepotism to µ̄ = 0.60
%∆ welfare 0.40% 0.00% 0.45%
∆ fraction of connected 2.85 p.p. 2.89 p.p. 2.94 p.p.
∆ unemployment rate -0.68 p.p. -0.30 p.p. -0.61 p.p.
Reduction of public-sector wages by 10 percent
%∆ welfare 1.36% 0.88% 1.35%
∆ fraction of connected -0.74 p.p. -0.68 p.p. -0.76 p.p.
∆ unemployment rate -1.80 p.p. -1.03 p.p. -1.82 p.p.
Note: The random search and connections premium models are recalibrated (κ = 7.31) and (κ = 6.29).

xxx
 To sum up, we can draw three main conclusions from this section. First, under the
baseline model, parameterized to a country with a large public-sector wage premia, welfare
is increasing in µ̄, but this is not always true in the random search model. Second, public-
sector wage cuts have a large quantitative effect on reducing the unemployment rate. Third,
in the random search model, the effects of policies on unemployment are qualitatively similar
but quantitatively smaller than in the model with segmented markets. The same holds for
the “connections premium” model.

Figure H6: Effects of nepotism

Welfare Fraction of connected workers Connection cost cutoff
.99

15

8
.985

6
10
Index
.98

4
5
.975

2
.97

0
.05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95
µ µ µ

Unemployment rate Unconnected unemployed in public sector Connected unemployed in public sector
19

.05

.015
.04
18

.01
.03
%
17

.02

.005
16

.01
15

.05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95
µ µ µ

Tightness Private wages Public sector premium

.065

.904

1.146
1.144
.902
.06

1.142
.9
.055

1.14
.898

1.138
.05

1.136
.896
.045

.05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95 .05 .15 .25 .35 .45 .55 .65 .75 .85 .95
µ µ µ

Note: The dark blue line is the economy with segmented markets. The light green line is the economy
with random search.

xxxi