Insulation or Patronage: Political Institutions and Bureaucratic Efficiency

3.1 Setup

Society consists of two groups of equal size identified by the parameter t∈A,B. Each of the two groups is politically represented by a politician of their type, A and B, respectively. In addition, two bureaucrats, one of each type, compete for a post in the state bureaucracy. Selection for the post allows the chosen bureaucrat to determine government policies which affect the whole of society. The full game has two periods and is structured as follows.

The politician of type A (incumbent) is assumed to be in power in period 1. He remains in power with probability p∈12,1 and is replaced by the politician of type B otherwise.^[16] Parameter p measures the level of political competition. High values of p can typically be associated with autocratic regimes in which the opposition has no access to political power. Values of p close to 12, on the other hand, can be associated with more competitive environments in which both politicians have a similar chance of being in power in the next period.^[17]

At the beginning of period 1 the incumbent chooses the bureaucratic institutions, S∈insu,patr, that govern bureaucratic recruitment in the second period. Patronage (S=patr) implies that the politician in power in period 2 can choose the bureaucrat. Under political insulation (S=insu) this power is ceded to a neutral commission that is only interested in bureaucratic competence.

Given S and p the two bureaucrats compete by specializing in one of two policy dimensions – public good provision or rent extraction. If a bureaucrat specialized on a policy dimension in period 1 she is able to provide a successful policy on that dimension with probability h∈12,1.^[18] However, specialization is never perfect in the sense that there is a positive likelihood of 1−h that she provides a successful policy on the other policy dimension in period 2. The successful provision of public goods by the bureaucrat pays one to every member of society regardless of their type. Policy success in rent-extraction pays τ>0 to the group of the bureaucrat and −τ to the other group. Assume for now that specialization costs on both dimensions are equal (to zero).

A bureaucrat is defined as competent if she specialized on the dimension of public good provision in period 1. Denote the level of competence by et∈h,1−h. The selection of a competent type A bureaucrat (eA=h) then implies, for example, that the public good is provided with probability h>12 while rents are extracted from group B with probability 1−h<12. A high level of et is best thought of as a high level of skill in the drafting and implementation of public policies. A bureaucrat with high levels of et has invested in a particular field of expertise, like the implementation of infrastructure projects or the setting of the central bank interest rate. Low levels of et reflect the development of strong social and professional ties with particular interest groups. Bureaucrats with low et therefore know how to favor or milk particular sectors of the economy through public policies. We treat both high and low et as equally costly only to simplify the analysis. Costly competence is discussed in the extensions.

The expected utility for all individuals (including the bureaucrat) of group t from a bureaucrat of type t and competence et is

where the first term represents the expected benefit from public good provision and the second term the expected benefit from rent extraction. Analogously, the expected utility for group t from a bureaucrat of type −t with competence e−t is

Equation [2] shows that the expected utility provided by a bureaucrat from the opposition is always increasing in her competence. This is because both increased public good provision and decreasing rent extraction by a bureaucrat of type −t benefit a member of group t. Expected utility in eq. [1], however, is only increasing in competence, et, if τ<1. Thus, the parameter τ measures the extent to which rent extraction implies higher benefits than public good provision. High values of τ could be either due to large rents (abundant natural resources, low cost of repression, etc.) or small public good gains (heterogeneous tastes, low demand for public goods, etc.). We call society more polarized if τ is larger. According to this definition, bureaucrats in a polarized society τ≥1 benefit less from public good provision than from extracting rents. When specializing, these bureaucrats are therefore tempted to focus on rent extraction.^[19]

The variable h is a measure for the level of specialization in public good provision and rent extraction. For h close to 12 the actual specialization of the bureaucrat does not matter very much. If public good provision requires very different skills than rent extraction h will be high. In this way the model is able to capture a situation where competence is general and not specific to either public good provision or rent extraction.

To end the description of the model, the sequencing of the actions is as follows. The incumbent first chooses the bureaucratic system S. Then, given S, the two bureaucrats simultaneously decide on their specialization et,t=A,B. The political process determines the politician for period 2. One of the two bureaucrats is recruited according to the rules of S and this bureaucrat implements policies that yield expected payoffs in eqs [1] and [2]. The next section shows that this game has a unique sub-game perfect equilibrium. Backward induction requires that bureaucratic recruitment in period 2 and the specialization decisions are discussed first. The combination of both then determines bureaucratic performance of patronage and insulation.

3.2 Bureaucratic Efficiency

A discussion of bureaucratic performance is most useful in the light of the first best. We assume that public goods are non-excludable, i.e. the politician does not internalize the overall benefits of public good provision. This drives a wedge between the politician’s utility in eqs [1] and [2] and overall welfare. Given that both groups in society are of equal size, a social planner who maximizes welfare would always want to select a competent bureaucrat in period 2 (et=h) regardless of the type t. Failure in recruiting a competent bureaucrat leads to a per capita welfare loss of h−1−h=2h−1. This is due to the fact that rent extraction is welfare neutral while public good provision improves overall welfare.

Political insulation is assumed to give all recruitment powers to a neutral commission that is only interested in competence. Assume that the commission chooses the candidate with the highest level of competence, et, and randomizes otherwise. We assume that if both bureaucrats have the same level of competence each one of them gets recruited with probability 12.

Patronage gives the right of selecting the bureaucrat to the politician in power in period 2. Clearly, politicians have an incentive to recruit a bureaucrat of their own type because she will distribute rents toward their group. This group-specific interest can clash with the selection by competence.

Lemma 1 explains why patronage can fail to recruit competent bureaucrats. The reason is that the politician might have to choose between a bureaucrat who would allocate rents to his group and a bureaucrat who is more competent. If rents are important (τ≥2h−1) bureaucratic competence is ignored.

An immediate effect of the bias described in Lemma 1 is that recruitment under patronage is fully determined by the type of politician in power. For all τ≥2h−1 politicians of type t always choose bureaucrats of type t. This can have consequences for bureaucratic competence under patronage.

Proposition 1 gives a possible explanation for why political insulation is often regarded as the better bureaucratic system. Patronage can imply that the bureaucrats specialize in extracting rents from the political opposition if these rents exceed the benefit from public good provision. From a welfare perspective this creates a loss of resources due to bureaucratic incompetence.

The political bias discussed in Lemma 1 is at the heart of this failure of patronage. Given that patronage recruits by type and not by competence there is effectively no competition between bureaucrats of different types. Polarization creates a labor market monopsony for bureaucrats with the correct type and, hence, bureaucrats know that they do not compete against bureaucrats with different policy preferences.

In a politically insulated system, competence can prevent the competing candidate from taking the bureaucratic post and implementing rent extraction in her favor. As both bureaucrats compete on an equal footing, polarization now works as an incentive for a specialization in competence.

Note that if τ<1 efficiency can be reached without giving bureaucrats additional incentives for specializing in competence. In order to focus on the interesting case we therefore make the assumption

This assumption implies two things. First, bureaucrats will specialize in rent extraction when left to their own devices. Second, politicians cannot commit to recruit by competence in period 2. As a result patronage is inefficient. We discuss the case of τ<1 at the end of the following section.

Given that patronage leads to a specialization in rent extraction the question arises whether the incumbent might still choose it as bureaucratic system. The next section shows that the incumbent’s decision is closely related to the level of political competition p and polarization τ.

3.3 Political Competition and Bureaucratic Institutions

In the first period the incumbent tries to affect outcomes in the second period by choosing the set of bureaucratic institutions, S∈insu,patr. By Proposition 1 and A1 his expected utility from patronage is

His expected utility from political insulation is

Note first that under patronage both governments always recruit a bureaucrat of their own type in period 2. This implies that if group A is more powerful (higher p) then the likelihood that bureaucrat A gets the post in period 2 is also higher. As p increases the incumbent can therefore expect more rents to be distributed toward the group he represents. Under political insulation both bureaucrats are selected with probability 12. At the same time bureaucratic efficiency is ensured. If expected political rents are low (because of low p and τ) the politician faces a commitment problem. He prefers competence to political recruitment but cannot commit the future government to recruit by competence under patronage. This leads to the following result.

Corollary 1 suggests a simple link between political and bureaucratic institutions (p and S, respectively). If political institutions exclude access to political power for the opposition (p=1) political insulation is impossible. An autocrat who knows that his absolute powers shield the bureaucracy from political influence by the opposition never introduces institutions like state exams or a civil service commission as these institutions would restrict his control over recruitment. Political insulation is only put in place if the incumbent is competing against other groups, i.e. if he knows there is a good chance that the opposition will come to power in the next period. Note that p˜<12, i.e. the incumbent insulates the bureaucracy even if that diminishes the likelihood that a bureaucrat from his group is recruited in period 2. In other words, the incumbent buys efficiency with a loss of political influence over the bureaucracy.

Corollary 2 follows immediately from condition (5). Higher polarization leads to a lower likelihood of insulation. In other words, the adoption of inferior bureaucratic institutions is more likely with the rising importance of rents.

Note that Corollary 2 relies on assumption A1. If τ<1 bureaucrats are not tempted to specialize in rent extraction – regardless of the incentives given by the bureaucracy. In other words, political insulation does not imply an improvement in bureaucratic efficiency in this case. As both bureaucrats always specialize in public good provision the incumbent insulates only if p<12 and polarization has no impact on the adoption of civil service reforms.^[20]

Corollary 1 implies that changes in political institutions can lead to an increase in bureaucratic efficiency because they lead to changes in bureaucratic institutions. A lot of the debate around the economic effects of democratization ignores this second-order effect. However, Corollary 2 suggests that the association between democratization and bureaucratic efficiency will depend on whether the implied reduction in p can compensate for lost rents. Here the level of polarization can play for or against an impact of democratization. To see this note, for example, that p˜ converges toward 12 as polarization increases. In very polarized societies a spill-over from democratization to political insulation is therefore unlikely as long as the incumbent enjoys at least some advantage over his competitor.

In summary, the model delivers a micro-foundation for the choice of inferior bureaucratic institutions and how this choice links to the political situation of a country. Absolutist political institutions and high potential for rent extraction lead to a choice of inefficient bureaucratic institutions which are geared toward the extraction of these rents. Political reforms can lead to an improvement of bureaucratic performance if they give a diverse set of groups access to political power. According to the model, political competition can imply that the political elite of a country adopts political insulation, bureaucratic efficiency increases and rent-extraction decreases.

4.1 Costly Investment in Competence

There is a recent literature on dictators which argues that stronger dictators might improve bureaucratic efficiency.^[21] One way to reconcile this view with our results is to assume that a specialization in competence is more costly than a specialization in rent extraction. This is particularly plausible where public good provision requires more expertise that is specific to the public sector.

Assume that competence costs the bureaucrat c regardless of whether she is selected for the post or not. Assume biased recruitment under patronage (τ≥2h−1). The candidate of type t=A now invests in competence if

while candidate B invests in competence if

which illustrates the fact that the strength of the patron (p or 1−p) now directly increases investment incentives. Political institutions now directly affect the level of competence under patronage because bureaucrats know that the return on their investment in competence depends on the probability of being hired in the next period.

Under insulation investment incentives are independent of p. We show in the appendix that for a range

there are multiple equilibria in investment. There are two asymmetric equilibria in which only one of the candidates invests with certainty. In these equilibria insulation always dominates patronage in terms of welfare. However, there is a third equilibrium in which the two candidates play a mixed strategy and invest with probability

In this equilibrium patronage can dominate insulation.

We illustrate the proposition in Figures 1 and 2. Figure 1 shows the level of competence under insulation (solid line) and patronage (dashed line) for different values of c and a monopoly of power (p=1). For investment costs below c_ both insulation and patronage lead to high competence. If investment costs are in the range c∈[c_,c¯] competence under insulation falls because the probability of investment in competence, given by eq. [6], falls whereas it remains high under patronage. This is a surprise given our earlier results. Competition for the post under insulation implies a degree of risk for the candidates which lowers investment incentives. Competence under patronage falls at c¯=2h−11−τ. At this point the type A candidate stops to invest despite guaranteed recruitment. For all c>c¯ both institutions lead to an incompetent bureaucracy.

Figure 1:

Competence with investment costs (no political competition).

Figure 2:

Competence with investment costs (strong political competition).

The situation of a weak incumbent (p=12) is illustrated in Figure 2. Now the point at which investment breaks down has shifted to the left under patronage but has remained unchanged under insulation. As a result, there is no value of c at which competence is higher under patronage than under insulation.

The main intuition behind Proposition 3 is that since specialization in competence is costly, a bureaucrat will only invest in competence if she is sufficiently confident that this competence will be put to use. Under patronage, a lack of political competition can therefore work as a guarantee that the candidate’s investment will pay off. In this way patronage can welfare dominate insulation under p=1. However, it only does so for strong incumbents. In other words, it is only in the combination between patronage and strong incumbent power that patronage performs well. When p=12 patronage does neither provide guaranteed recruitment nor competition between the candidates and is therefore always welfare dominated.^[22]

Proposition 3 is important because it reinforces the notion that political and bureaucratic institutions are complementary. An absolutist ruler can guarantee to hire his group and, thus, maximizes their incentives to invest into competence. This incentive can be so strong that welfare is higher under patronage than under insulation. However, if political power is shared, patrons are weak. Patronage then harms competence and insulation welfare dominates because it does not rely on the absence of political competition to provide investment incentives.

The incentives of the incumbent to adopt insulation are similar to the main model. Without political competition the incumbent always adopts patronage. If power is shared the incumbent always wants to insulate the bureaucracy. We discuss the adoption decision in more detail in the appendix.

4.2 Endogenous Political Power

Patronage is often referred to as the use of spoils to gather political support. Acemoglu, Ticchi, and Vindigni (2010), for example, use this view on patronage to provide a theory of inefficient states. This section incorporates this view by making p endogenous to the choice of bureaucratic system.

The political process is modeled as a voting game. There are three kinds of voters. The first kind is only interested in public good provision and votes for the politician that is expected to provide the highest benefit on that policy dimension. This group has a mass of 1−α.

The second group of voters has a mass of α and votes for the politician that promises most rents. Voters of this type can be thought of as forming the power-base of the respective politicians. They will always support their type of politician if this politician has control over spoils. This is because the opposing politician cannot commit to provide spoils after the election. For now assume that the political system is biased in the sense that a share λ∈12,1 of these voters are of type t=A.

Uncertainty in the political process is modeled through a third group of noise voters whose votes lead to an aggregate shock of votes ε for the politician of type B. Assume that this shock is distributed uniformly on the interval −1,1.

The game structure is as follows. First, the incumbent chooses the bureaucratic system. Second, the first two voter types support their preferred candidate. Finally, bureaucrats specialize before the uncertainty introduced by the third voter type gets resolved. Indifferent voters are assumed to randomize their vote.

Under insulation the identity of the politician in the second stage has no impact on outcomes and all decisions are made as before. In particular, both bureaucrats always invest into competence by Proposition 1 regardless of which group is more powerful. The incumbent’s expected utility from S=insu remains unchanged and is given by

We maintain assumption A1. Under patronage all voters interested in rents vote for a candidate of their own type. Competence between the two bureaucrats is identical. The incumbent therefore wins the election if

or

Given that ε is distributed uniformly on the interval −1,1 the probability of re-election under patronage can then be written as

while under insulation it is

The following result follows immediately.

The main novelty in Proposition 4 is that as spoils become more important for political outcomes (rising α), insulation is less likely. This is because the politician is not in control of spoils if the bureaucracy is insulated. If vested interests are the natural allies of the incumbent λcloseto1 this factor can prevent insulation. Inefficient bureaucratic institutions are then adopted to allow the political elite to stay in power.

Additional insights can be gained from the case in which λ<12. In this case the probability of re-election under patronage is

while it remains pinsu,λ,α=12 with an insulated bureaucracy. If the incumbent has strong vested interests against himself he can use bureaucratic insulation to dissipate political resistance. Insulation works because it ensures that bureaucratic decisions are not affected by the political leadership. Powerful interests that oppose the incumbent then know that rents might be distributed in their favor. This highlights a surprising mechanism in which bureaucratic reforms fill in for political change.

There are two important cases that might be explained by this reverse causality argument – Prussia and Imperial China. It has been argued that Prussian feudalism might have survived so long because of a relatively insulated bureaucracy. The rising classes were less willing to stage revolution because they were able to reach relatively influential positions in the powerful bureaucracy.^[23] Similarly, Northern Sung rulers chose civil service examinations to limit the development of alternative power centers by drawing into their government the sons of elites from newly emerging regions in South China.^[24] In both cases autocrats introduced bureaucratic reforms to prevent political opposition.

5.1 Political Interference and Bureaucratic Performance

In the nineteenth century Great Britain introduced state exams and a civil service commission; both of these institutional features reduced political control over recruitment with the declared aim of increasing competence. The arguments surrounding these reforms are well documented and can therefore give an insight into the motivation behind actual civil service reforms.

At the beginning of the nineteenth century patronage was the norm in Great Britain while objective criteria hardly played any role in recruitment into the bureaucracy.^[25] The struggle for the introduction of a merit system began in 1853 when Charles Trevelyan and Stafford Northcote were asked by the then Chancellor of the Exchequer, William Gladstone, to write a report on its recruitment and promotion.^[26] The resulting report, published in 1854, urges for insulation with the following argument:

It may safely be asserted that, as matters now stand, the Government of the country could not be carried on without the aid of an efficient body of permanent officers, occupying a position duly subordinate to that of the Ministers who are directly responsible to the Crown and to Parliament, yet possessing sufficient independence, character, ability, and experience to be able to advise, assist, and to some extent, influence, those who are from time to time set over them.^[27]

There is no doubt that Northcote and Trevelyan aimed at insulating the British bureaucracy from political influence. One of their contemporaries, John Stuart Mill, praised reforms in the East India Company as follows:^[28]

A second great advantage of the present system is, that those who are sent out as candidates to rise by degrees are generally unconnected with the influential classes in the country, and out of the range of Parliamentary influence. The consequence is, that those who have the disposal of offices in India have little or no motive to put unfit persons into important situations [...]^[29]

Thus, not only were reformers in Great Britain aware of political incentives to promote unfit personal, they saw political independence as a way to improve performance in this context. Two recent empirical studies lend support for this mechanism. Iyer and Mani (2012) study transfers in the top layers of the Indian civil service between 1980 and 2004. They show first that the reallocation of posts in the Indian bureaucracy is correlated with political change – i.e. even in a system geared toward selection by merit politicians use the leverage they have to alter careers. The clearest division line for political interests in India is caste membership. Consistent with the idea that this political division creates a labor market monopoly for members of the “right” caste, Iyer and Mani found that officers are more likely to be appointed to important positions when they belong to the same caste as the Chief Minister’s party base.

What are the efficiency costs of this politicization? While direct evidence is difficult to establish, Iyer and Mani provide some evidence that career incentives are indeed altered by the political allocation of posts. They show that bureaucrats who can expect to have more politicized careers (i.e. careers that depend more on political change) invest less in expertise. Their study also lends some support for the fact that the performance of government agencies suffers from more politicization.

In addition to this detailed study there are several cross-sectional empirical studies that lend support for the idea that political insulation improves performance. Lewis (2007) uses the scores of an evaluation program initiated by the Bush administration in 2004 to measure program performance. He finds that politically appointed managers perform significantly worse than career civil servants.^[30] He shows that the best proxy to explain this difference is the higher bureau experience that civil servants bring to the table. In a cross-country study Rauch and Evans (2000) find that meritocratic recruitment (recruitment by state exams) correlates closely and robustly with measures of bureaucratic performance.

5.2 When Do Politicians Introduce Insulation?

Our theory provides comparative statics regarding the effect of political competition on the use of patronage that are in conflict with existing theoretical work on the topic. In this section we therefore provide some evidence that political competition leads to insulation as Proposition 2 suggests. In addition, we relate our results on the role of polarization to recent discourses in the literature.

Barbara Gedde’s book Politician’s dilemma: Building state capacity in Latin America provides a detailed account of bureaucratic reforms in Latin America. She finds support for the idea that insulation was introduced in Latin America when new political groups gained access to political power. In Venezuela, for example, a civil service reform was drafted in 1960 after the dominant political party, Democratic Action (AD), had lost a large share of its seats in the Venezuelan chamber of deputies. Civil service reforms were implemented once the AD drew equal to its main political competitor in 1968,^[31] but then reversed by presidential decree when the AD returned to its dominant role in 1973. Such a reversal is a good indicator that political competition can be an important factor for changes in bureaucratic institutions.^[32]

Another example of the relationship between political and bureaucratic reform is the case of Great Britain discussed in the previous section. The Northcote and Trevelyan report was written at a time of radical political changes in the nineteenth century. The middle classes had gained the right to vote in the first reform act in 1832. In addition, the prerogative power of the crown had come to be increasingly taken over by ministers who could be held accountable by parliament. Thus, the executive had become more volatile in the interests it represented both because new groups had entered the political arena and because institutional changes facilitated their access to executive power. In 1854 Northcote and Trevelyan proposed to modify recruitment procedures through the introduction of an entry examination for all public servants, administered by an independent Board of Examiners. Reforms, however, did not take place because parliament resisted their implementation. State exams were not introduced until 1870 – in the middle of a wave of political reforms that would guarantee representatives of the working class access to parliament.^[33] One interpretation of this timing of events is that the established members of parliament feared that the control over political recruitment could fall into the hands of the middle and working classes. As their political power was fading, the members of parliament preferred to transfer the power of bureaucratic recruitment to a neutral body.

In addition to these detailed accounts there is empirical evidence from within-country studies that indicate that politicians prefer to insulate if their political power is challenged. Hanssen (2004), for example, studies the institutions governing the appointment of judges in the United States between 1950 and 1990. He shows that institutions that increase judicial independence from politics tend to be implemented in states that feature competitive politics and confirms this pattern with time variation available in the data. Similarly, Ruhil and Camões (2003) show that electoral competition was a key predictor of insulation at the state level in the United States. One standard deviation in closeness of elections more than doubled the chance that a state adopted competitive state examinations between 1900 and 1939.

In order to test whether these patterns appear more generally we need measures of both political competition and bureaucratic institutions which are consistent across countries. Sadly, there is little available data that quantifies specific bureaucratic institutions for a large set of countries.^[34] The most commonly used data on bureaucratic quality is provided by the Policy Risk Survey Group (PRSG) but the group does not state which institutions exactly are responsible for high bureaucratic quality. This makes the data less useful for testing the link between political competition and specific bureaucratic institutions.^[35] The concern can be avoided with data on CBI provided by Cukierman, Webb, and Neyapti (1992) and Crowe and Meade (2008). Their measure of CBI contains a category which seeks to measure the independence of the Central Bank CEO from political interference on a 0–1 scale.^[36] Most economists would agree that political selection might hinder efficiency in this bureaucratic post.

In order to show that political competition is related to bureaucratic independence it is important to find a relatively precise measure of the political institutions we have in mind.^[37] The measure most closely related to our concept of political competition is the index of competitiveness of executive recruitment provided by the Polity IV project. According to the Polity IV manual “Competitiveness refers to the extent that prevailing modes of advancement give subordinates equal opportunities to become superordinates.”^[38] The codebook distinguishes between fair elections of the chief executive (the top layer of government) and selection through hereditary succession, designation or rigged election. In Table 1, column (1) shows the correlation of CEO independence in 2003 with the average extent of political competition for the years 1998 till 2003.^[39] The coefficient suggests that an increase of political competition by 1 standard deviation implies half a standard deviation increase in bureaucratic independence. Column (2) shows that this simple correlation is robust to controlling for GDP per capita and past inflation. Column (3) reveals that running the same regression for the 1980s yields identical results. Interestingly, the extent of political competition in this earlier period does not predict CEO independence in 2003 (column (4)). The underlying reason is considerable change both in political competitiveness and bureaucratic independence between the 1980s and 2003. Column (5) confirms that an increase in political competition within this relatively small period of time went hand in hand with insulation even when controlling for changes in GDP per capita. This suggests a link between competitiveness and bureaucratic independence in the direction suggested by our theory.

Evidence on the connection between polarization and bureaucratic institutions is relatively sparse. While there is a wealth of work in development economics that studies the negative impact of variables like inequality, resource rents and ethnic divisions on economic outcomes our theory suggests that the effect should be coming through the capacity of the state to provide public goods as opposed to extracting rents. The most direct evidence here comes from Alesina, Baqir, and Easterly (1999) who study the impact of ethnic fractionalization on the provision of public goods in US cities, metropolitan areas and urban counties. They show that the share of spending on productive public goods declines with ethnic fractionalization and suggest that this might be driven by an increase in spending motivated by patronage in these jurisdictions.

Note that we modeled polarization as the benefits from rent extraction relative to the benefits of public goods provision. An intriguing application of this view is the military. To see this, note that the definition of polarization here captures the difference between external and internal wars. If an external war looms, all members of society would suffer a huge loss if national defense is not successful. Thus, the present government wants to recruit the military by merit because this maximizes defense (public good provision). In a civil war the government wants to recruit the military by patronage because it wants to be sure that the hired military is loyal, i.e. the group-specific control of rents is more important than maximizing the overall level of military competence. Here our theory closely relates to findings by Besley and Persson (2009) who show in a cross section of countries that measures of state capacity correlate positively with past external conflict but negatively with internal conflict.

Abundance of natural resources is another example of an environment with high rents. And indeed there are several case studies that link resource booms to increasing patronage. Robinson, Torvik, and Thierry (2006), for example, summarize several case studies to show how resource booms lead to politically motivated expansions of the public sector. A particularly revealing account comes from Ross (2001) who shows that countries in Southeast Asia dismantled existing forestry institutions (laws, regulations and departments) and increased patronage in the face of timber price shocks. He shows how institutions insulating the forestry bureaucracy from political pressures were dismantled in the Philippines and Malaysia by politicians in order to gain access to resource rents.