Fiscal weakness,the (under-) provision of public services,and institutional reform

Abstract

The fact that many developing countries spend little on growth-promoting public services is often blamed on “fiscal weakness”. We propose a novel framework to study the consequences of an exogenous improvement in public revenues. We demonstrate that higher revenues may induce the group in power to use a dual repression approach, one pillar of which is to aggravate the “underspending” on growth-promoting public services. However, we also identify circumstances in which the group in power will abstain from repression and instead opt for institutional reform. This result contrasts with the view that exogenous revenue improvements necessarily undercut growth-promoting institutions.

JEL classification: D70; O11; F53; H50; Q34

Keywords

Fiscal weakness public services reform repression resource extraction

Introduction

The state plays an important role in the process of economic development. In an environment that offers little protection against expropriation by parasitic elites, private investment is low (e.g. Acemoglu and Johnson, 2005). Without adequate funding of public infrastructure (such as transport, electricity, and telecommunications), economic growth can hardly be sustained (e.g. Röller and Waverman, 2001). It is therefore reasonable to think of “fiscal weakness” as a drag on economic growth: if the state’s capacity to extract resources from the economy is low, as is the case in many developing countries, the government may lack the resources to assume a key developmental role (e.g. Herbst, 2000).¹ So an increase in the resources available to a fiscally weak government may spur economic growth by permitting additional infrastructure investment and better public services (e.g. Sachs, 2005). This notion also informs the international post-2015 development framework. The United Nations estimates the need for infrastructure investment in developing countries to be about $$ 1.5$ trillion per year and calls on rich countries to contribute towards the funding of these needs (DESA, United Nations Department of Economic and Social Affairs, 2015). The World Bank (2013) suggests strengthening domestic resource mobilization, not least by imposing higher taxes on multinationals engaged in natural resource extraction.

However, recent experiences with sudden improvements in fiscal strength sound a note of caution. Consider the case of Chad, a country that has experienced a huge increase in public revenues after the completion of an oil pipeline that now connects the country with the Gulf of Guinea. The pipeline project was supported by the World Bank and approved in 2000 when Chad passed a law to curtail the government’s discretion over the allocation of the new revenues. The law required that the lion’s share be used to fight poverty and spur economic development. During the following 10-year period, the project generated $$ 6.3$ billion (current US dollars) in cumulative public revenues.² Yet, despite this improvement, there is little evidence of a marked increase in the spending on priority sectors, i.e. sectors identified to be key for poverty alleviation and development (such as public health and education). In fact, the contrary seems to be the case. Annual spending on public health, one of the few “priorities” for which time-series data is available, fell from $$ 26$ p.c. (constant 2011 international dollars) in 2000 to $$ 23$ p.c. in 2010. During the same period, military spending rose from $$ 15$ to $$ 111$ p.c. (constant 2011 international dollars), arguably in response to coup attempts and rebel attacks. Moreover, there is evidence that some of the additional resources were diverted into private pockets and hence must be regarded as transfers. As we will discuss below, the Chadian experience is not unique. There are parallels with the situation in Azerbaijan, a country whose public finances have benefited from an even larger oil windfall.

This paper develops a novel model to study the consequences of exogenous (e.g. natural resource-related) increases in public revenues in places where the population is fractionalized and the executive only weakly constrained. There are two actors, the group in power and the group out of power. A key feature of the model is that the group in power has a rich set of choice variables, broadly those appearing in the accounts of the events in Chad and Azerbaijan. In the basic version of the model, the choice set consists of the following spending options: spending on public services, which—like health services—improve average individual productivity and hence lift incomes (as in Barro, 1990); spending on the military, whose purpose is to address internal threats, i.e. to help the group in power prevail against the other group in a power contest with sequential input choice (as in Leininger and Yang, 1994); and spending on transfers, which can be group-specific to the extent the political institutions permit (as in Besley and Persson, 2011a, b). In the extended version of the model, the group in power can also choose to strengthen the cohesiveness of political institutions, i.e. to commit to an institutional reform that reduces its discretion in making group-specific transfers in the future.³ In this setup, the group out of power may have an incentive to take political power in order to gain privileged access to transfers—something the group in power may want to prevent by means of repression and/or institutional reform.

Our framework offers two main insights. The first insight concerns the way the group in power uses its spending options for repression once public revenues have grown beyond some “lower” threshold. The optimal approach to repression is a dual one, combining a rise in military spending with underspending on productivity-improving public services. Importantly, such underspending is not just a mechanical by-product of a tighter budget constraint owing to rising military spending. To prevail in the power contest, the group in power must increase the cost that the other group would need to incur to successfully get a hold on power. The former group does so by simultaneously strengthening the military and reducing the provision of public services below the level it would prefer if there were no internal threats to its political power. Strengthening the military increases the financial effort the group out of power must incur to successfully take power. A reduction in the provision of public services, on the other hand, increases the marginal cost of funding such an effort. Why? Reducing public services impairs private-sector productivity and hence reduces the non-transfer incomes of all individuals, including those who belong to the group out of power. Since the members of that group receive few transfers, they experience a large increase in the marginal utility of consumption. As a result, even if the military were not strengthened, a successful attempt to take power would involve a larger loss in utility terms.⁴ In other words, the marginal cost of the resources required to prevail in a contest against the group in power is endogenous. This distinctive element of our model (e.g. as compared with Besley and Persson, 2011a, b) implies that underspending on productive public services is an integral aspect of repression.

The second insight stems from the extended version of the model, in which the group in power additionally can choose to strengthen the cohesiveness of political institutions. The insight concerns the circumstances under which the group will abstain from repression and instead commit to strengthening institutional cohesiveness to avert a power contest. Such an institutional reform weakens the incentives by the group out of power to take over, possibly rendering the use of repression unnecessary. If, in fact, institutional reform were the only response by the group in power to a rise in public revenues beyond the “lower” threshold, the consequences of the rise would be benign: the funding of public services would significantly improve in the short term and further increase over time. From the perspective of the group in power, the relative cost-effectiveness of the two alternative measures—repression vs reform—depends on the level of public revenues. We show that the group in power will indeed restrict itself to institutional reform as long as public revenues remain below some “higher” threshold. However, there will be a switch to repression once this threshold is crossed.

In combination, the two insights imply that an exogenous increase in public revenues can have widely differing consequences. A negative outcome is possible but not inevitable. In an initial situation with low revenues, a small increase, leaving total revenues below the “lower” threshold, induces an equally small immediate improvement in public services. An intermediate increase, pushing total revenues to a level between the “lower” and the “higher” threshold, leads to a significant immediate improvement and, by inducing institutional reform, to an even larger improvement in the future. Finally, an increase that lifts revenues beyond the “higher” threshold provokes repression—a combination of heightened military spending and underspending on public services. As a consequence, public services may deteriorate immediately. The locations of the two thresholds are context-specific and depend, among other things, on the initial level of institutional cohesiveness and the productivity of the private sector.

While informing the policy debate on fiscal weakness and exogenous revenues in general, our framework also offers a coherent perspective on the events in Chad and Azerbaijan. As to the Chadian pipeline project, it appears that the unexpectedly large increase in public revenues made the government rescind its initial measures to increase institutional cohesiveness and switch to (more intense) repression instead. The fact that the Chadian government impaired productivity by reducing public-health spending despite rising revenues can be explained as being part of the government’s repression approach. In a similar vein, our framework helps make sense of reports suggesting that the Azeri government stifled smaller businesses and blocked investment outside the oil sector for years after the start of the oil boom.

In setting up our theoretical framework, we draw on a number of existing papers in the literature on economic growth and the political economy of development. On the one hand, by assuming that the provision of public services raises private-sector productivity, we build on Barro (1990), Acemoglu (2005), Besley and Persson (2009), Caselli and Cunningham (2009), and Oechslin (2010), among others. On the other hand, we follow Acemoglu and Robinson (2000), Besley and Persson (2011a, b), Acemoglu et al. (2013b) and Besley et al. (2016), among others, by endogenizing the parameter representing political institutions, where here political institutions are understood to be provisions that constrain the discretionary power of the executive. Because of this particular combination of elements, we are in a position to discuss within one single framework factors that influence whether an exogenous rise in public revenues will lead to reforms (better public services, increase in institutional cohesiveness) or to repression (worse public services, increase in military spending).⁵

As our setup includes competing groups that contest the access to public revenues, this paper also relates to the literature on rent seeking and the resource curse (e.g. Dunning, 2005; Hodler, 2006; Mehlum et al., 2006; Torvik, 2002; Tornell and Lane, 1999; Wick and Bulte, 2006). Models in that literature, surveyed by Torvik (2009) and van der Ploeg (2011), approach the political-economy problem as one between politically symmetrical groups (Kolstad and Wiig, 2009). Instead, we primarily focus on the role of the government and the incentives of the group in power to retain access to rents (as do, e.g. Robinson et al., 2006; Robinson and Torvik, 2005). We depart from the usual assumptions of a contest game with simultaneous input choice and an exogenous marginal cost of contest participation. Instead, we assume a contest with sequential input choice (as, e.g. in Leininger and Yang, 1994). Sequential input choice, together with endogenous marginal participation costs, allows us to cleanly identify underspending on productivity-improving public services as one of two pillars of repression. As a result, we can offer a novel explanation for observations of an inverse relationship between public revenues and public services: a sufficiently large increase in public revenues induces the group in power to use repression, relying on an approach that combines a rise in military spending with purposeful underspending on productive public services.

The remainder of this paper is organized as follows. The next section sets out our basic theoretical framework. The two subsequent sections solve for the equilibrium. While in the third section the cohesiveness of the political institutions is exogenously fixed, cohesiveness is treated as a choice variable in the fourth section. The fifth section interprets the recent experiences of two different countries, Chad and Azerbaijan, through the lens of our framework. The final section concludes.

The model: assumptions

Actors, preferences, and incomes

We develop a political-economy model with two periods, $t \in {1, 2}$ . The population size is normalized to $1$ , and each individual belongs to one of two equal-sized groups, $A$ and $B$ . There is no within-group heterogeneity. As a consequence, we can refer to groups $A$ and $B$ as the two actors of the economy.⁶ In each period, one of the two groups holds political power, while the other one is out of power. In period 1, the allocation of political power is exogenous; in period 2, the allocation will be endogenous. We use the state variable $P_{t} \in {A, B}$ to denote the group that holds political power in period $t$ . Holding political power means control of the government choice variables.

The economy produces a unique non-storable consumption good (which is the numéraire). Preferences are represented by the intertemporal utility function:

U (c_{1, i}, c_{2, i}) = \ln (c_{1, i}) + β \ln (c_{2, i}),

(1)

where $c_{t, i}$ refers to the period $t$ level of individual consumption in group $i \in {A, B}$ and $β \in (0, 1]$ denotes the discount factor.⁷ In what follows, we set $β$ equal to $1$ to simplify the analysis. In the two-period setup under consideration, this particular choice does not distort the substance of the model’s implications (also see footnote 7).

Individual income from the private sector in period $t$ is described by the production function:

y_{t} = χ H_{t}^{α},

(2)

where $0 < α < 1$ and $H_{t} ⩾ 0$ refer to the level of public services provided by the government in period $t$ . Following Barro (1990), production function (2) captures in a simple way that the government plays an important role in promoting the productivity of the private sector, for instance by running a public-health system, maintaining public infrastructure, or enforcing private contracts. The factor $χ > 0$ stands for the extent to which the private sector has access to things like advanced technologies and physical capital. In what follows, we will refer to $χ$ simply as the productivity of private-sector technologies. Note that aggregate private-sector income, denoted by $Y_{t}$ , is equal to $y_{t}$ as the population size is normalized to $1$ .

The assumption that $H_{t}$ serves as an intermediate good rather than a final consumption good (as, e.g. in Besley and Persson, 2011a, b) is important. As we will discuss below, in the present setup, public services have an impact on individual budget constraints. Therefore, other things equal, reductions in $H_{t}$ limit the affordability of consumption goods.

Public revenues, policies, and institutions

In each period, public revenues are given by $R > 0$ units of the consumption good. For simplicity, we assume that this revenue stream is exogenous, i.e. unrelated to private-sector output. It is natural to think that $R$ reflects the recurrent income from the extraction of publicly owned natural resources.

The set of government choice variables includes four spending options: spending on public services, $H_{t} ⩾ 0$ ; spending on the military, $M_{t}^{G} ⩾ 0$ ; and transfers to each of the two groups, denoted by $T_{t, A} ⩾ 0$ and $T_{t, B} ⩾ 0$ . We further assume that the government does not have access to the capital market. Since the consumption good is non-storable, this implies that the government’s flow budget constraint must be satisfied in each period:

H_{t} + M_{t}^{G} + T_{t, A} + T_{t, B} ⩽ R .

(3)

The allocation of transfers is subject to an institutional constraint. The transfer that goes to the group out of power must not be less than a share $I_{t} \in [0, 1]$ of the transfer that is allocated to the group in power. In formal terms, we impose

T_{t, j} ⩾ I_{t} T_{t, i},

(4)

where $i \in {A, B}$ refers to the group in power and $j \neq i$ to the group out of power. The larger $I_{t}$ is, the smaller the extent to which one group can be favored at the expense of the other. Following Besley and Persson (2011a, b), we say that $I_{t}$ reflects the degree to which the political institutions are cohesive. If $I_{t} = 0$ , the group in power enjoys maximally privileged access to government transfers; on the other hand, if $I_{t} = 1$ , holding political power does not entail any material privileges. In what follows, we assume that in period $1$ the degree of institutional cohesiveness is exogenously given: $I_{1} = λ^{l}$ , where

0 ⩽ λ^{l} < 1 .

(R1)

The next section treats cohesiveness as a deeply ingrained, and hence invariable, institutional characteristic: $I_{1} = I_{2} = λ^{l}$ . In the fourth section, we consider the case of an endogenous $I_{2}$ and assume that $I_{2}$ is determined in period 1 (simultaneously with the spending options).

The group that holds power in period $t$ determines the spending vector ${H_{t}, M_{t}^{G}, T_{t, A}, T_{t, B}}$ , taking into account the budget constraint as well as the institutional constraint. If institutional cohesiveness is invariable (next section), the set of government choice variables in period $t$ is identical to the period $t$ spending vector. If $I_{2}$ is treated as an endogenous variable (fourth section), the set of government choice variables in period 1 includes the first-period spending vector as well as second-period institutional cohesiveness, $I_{2}$ .

Political power and its transition

While in period 1 the allocation of political power is exogenous, the group that initially is out of power can get hold of the government in period 2. A power transition takes place if at the end of period 1 the group out of power prevails in a contest with the group in power. The direct contestants are the regular military (on the part of the group in power) and a private militia (on the part of the other group). While the funding of the military, $M_{1}^{G}$ , comes from the government budget, the funding of the private militia, $M_{1}^{O}$ , is provided by the group out of power. The outcome of the contest is given by

P_{2} {\begin{matrix} = P_{1} & : & M_{1}^{G} ⩾ ζ M_{1}^{O} \\ \neq P_{1} & : & M_{1}^{G} < ζ M_{1}^{O} \end{matrix},

(5)

where $ζ$ captures the productivity of any militia spending relative to the productivity of the spending on the regular military.⁸ As the regular military tends to have better access to advanced equipment, we assume that militia spending is less productive than military spending:

0 < ζ < 1 .

(R2)

We further assume that the period $t$ decisions by the group in power are observed by the group out of power before the latter decides on $M_{t}^{O}$ . Finally, we note that our model focuses on power contests among internal actors. As a result, in the model, the sole purpose of military spending is to confront internal threats (rather than external ones).

Summary

This is a dynamic game of complete and perfect information and we use backward induction to find the Subgame Perfect Nash Equilibrium. There are two actors, group $A$ and group $B$ . In period 1, the allocation of political power, $P_{1}$ , is exogenous. Without loss of generality, we henceforth assume $P_{1} = A$ . The timing is as follows. In period 1, (i) group $A$ determines the spending vector ${H_{1}, M_{1}^{G}, T_{1, A}, T_{1, B}}$ , thereby observing constraints (3) and (4); in the fourth section, group $A$ additionally and simultaneously determines second-period institutional cohesiveness, $I_{2}$ ; (ii) observing these choices, group $B$ decides on how much to spend on its militia, $M_{1}^{O}$ ; and (iii) the future power allocation, $P_{2}$ , is determined according to equation (5) and the individual consumption levels $c_{1, A}$ and $c_{1, B}$ materialize. In period 2, (i) the group in power determines the spending vector ${H_{2}, M_{2}^{G}, T_{2, A}, T_{2, B}}$ , thereby observing constraints (3) and (4); (ii) the individual consumption levels $c_{2, A}$ and $c_{2, B}$ materialize and the game ends. The individual payoffs in groups $A$ and $B$ are given by $U (c_{1, A}, c_{2, A})$ and $U (c_{1, B}, c_{2, B})$ , respectively, where the individual overall utility $U (\cdot)$ is specified in equation (1).

Exogenous institutional cohesiveness

Decisions in period 2

Suppose that group $i \in {A, B}$ holds political power in period 2 (i.e. $P_{2} = i$ ). When deciding on the spending vector ${H_{2}, M_{2}^{G}, T_{2, A}, T_{2, B}}$ , group $i$ maximizes

c_{2, i} = χ H_{2}^{α} + \frac{T_{2, i}}{1 / 2},

(6)

where $T_{2, i} ⩾ 0$ is the total transfer received by the group and the division by 1/2 accounts for the fact that the group size is 1/2. Although period 2 is the final period, the institutional constraint regarding transfers to the group out of power still binds; yet, being exclusively interested in $c_{2, i}$ , group $i$ does not provide group $j$ with a transfer larger than strictly required: $T_{2, j} = I_{2} T_{2, i} ⩾ 0$ . At the same time, in the final period, military spending would not serve any purpose. As a result, $M_{2}^{G} = 0$ . Considering this, and given that it is clearly optimal to spend all available resources, constraint (3) will hold with equality, implying $T_{2, i} = (1 + I_{2})^{- 1} (R - H_{2})$ . So group $i$ solves the following optimization problem:⁹

\max_{{H_{2}}} {χ H_{2}^{α} + \frac{{(1 + I_{2})}^{- 1} (R - H_{2})}{1 / 2}} s . t . H_{2} - R ⩽ 0 .

(7)

Proposition 1: The solution to optimization problem (7) is given by $H_{2} = R$ if $R ⩽ H^{*} (I_{2})$ and by $H_{2} = H_{2}^{*} (I_{2})$ if $R > H_{2}^{*} (I_{2})$ , where

H_{2}^{*} (I_{2}) \equiv {[\frac{α χ (1 + I_{2})}{2}]}^{1 / (1 - α)} .

(8)

Proof. See Online Appendix B. ■

So group $i$ appropriates public resources through transfers only if public revenues, $R$ , exceed a particular threshold, $H_{2}^{*}$ . Otherwise, if $R ⩽ H_{2}^{*}$ , transfers are zero as public revenues have a larger marginal effect on $c_{2, i}$ if spent on public services. However, as long as $I_{2} < 1,$ $H_{2}^{*}$ is strictly less than ${(α χ)}^{1 / (1 - α)}$ , the level that maximizes aggregate consumption. This is because only half of the return to public services accrues to group $i$ (the other half goes to group $j$ ), while the group is able to capture a share $(1 + I_{2})^{- 1} > 1 / 2$ of total transfers.

Given the results above, one can calculate the period $2$ levels of individual consumption as

c_{2, i} (i, I_{2}, R) = {\begin{matrix} χ R^{α} & : & R ⩽ H_{2}^{*} (I_{2}) \\ 2 (1 + I_{2})^{- 1} α^{- 1} [(1 - α) H_{2}^{*} (I_{2}) + α R] & : & R > H_{2}^{*} (I_{2}) \end{matrix}

(9)

and

c_{2, j} (i, I_{2}, R) = {\begin{matrix} χ R^{α} & : & R ⩽ H_{2}^{*} (I_{2}) \\ 2 (1 + I_{2})^{- 1} α^{- 1} [(1 - α I_{2}) H_{2}^{*} (I_{2}) + α I_{2} R] & : & R > H_{2}^{*} (I_{2}) \end{matrix},

(10)

where the “ $i$ ” in $c_{2, \cdot} (i, I_{2}, R)$ indicates that group $i$ holds political power in period 2. Since public revenues are spent entirely on public services if $R ⩽ H_{2}^{*} (I_{2})$ , there is no difference in consumption levels across the two groups in that case. However, if $R > H_{2}^{*} (I_{2})$ , the level of individual consumption in group $i$ exceeds the corresponding level in group $j$ if the political institutions are not fully cohesive (i.e. if $I_{2} < 1$ ).

Decision by the group out of power in period 1

Moving one step backwards, the final decision in period $1$ concerns group $B$ ’s spending on its private militia, $M_{1}^{O}$ . To inform this decision, group $B$ derives the maximum amount it is prepared to spend in order to take political power in period $2$ . The maximum amount, denoted by ${\bar{M}}_{1}^{O},$ is implicitly determined by

\ln [χ H_{1}^{α} + 2 T_{1, B}] + \ln [c_{2, B} (A, I_{2}, R)] = \ln [χ H_{1}^{α} + 2 (T_{1, B} - {\bar{M}}_{1}^{O})] + \ln [c_{2, B} (B, I_{2}, R)],

(11)

where $c_{2, B} (A, I_{2}, R)$ and $c_{2, B} (B, I_{2}, R)$ come from equations (10) and (9), respectively. The left-hand side of equation (11) represents individual overall utility in group $B$ under the assumption that $M_{1}^{O} = 0$ and $P_{2} = A .$ On the right-hand side, it is assumed that group $B$ spends ${\bar{M}}_{1}^{O}$ on its militia and that political power shifts from group $A$ in period $1$ to group $B$ in period $2$ . By rearranging terms in equation (11), we obtain

{\bar{M}}_{1}^{O} = \frac{1}{2} \frac{c_{2, B} (B, I_{2}, R) - c_{2, B} (A, I_{2}, R)}{c_{2, B} (B, I_{2}, R)} [χ H_{1}^{α} + 2 T_{1, B}] ⩾ 0 .

(12)

Now two cases must be distinguished. First, if $R ⩽ H_{2}^{*} (I_{2})$ , holding political power in period $2$ does not entail any rents. Therefore, equations (9) and (10) imply $c_{2, B} (B, I_{2}, R) = c_{2, B} (A, I_{2}, R) .$ So group $B$ does not spend any positive amount on taking power ( ${\bar{M}}_{1}^{O} = 0)$ . Second, if $R > H_{2}^{*} (I_{2})$ , holding power in period $2$ involves privileged access to rents. As a result, $c_{2, B} (B, I_{2}, R) > c_{2, B} (A, I_{2}, R)$ , where explicit expressions for the two consumption levels can be inferred from equations (9) and (10). By using those expressions in equation (12), and then rearranging terms, we obtain

{\bar{M}}_{1}^{O} = f (I_{2}, R) \cdot [χ H_{1}^{α} + 2 T_{1, B}],

(13)

where

f (I_{2}, R) \equiv \frac{(1 - I_{2})}{2} \frac{- α H_{2}^{*} (I_{2}) + α R}{(1 - α) H_{2}^{*} (I_{2}) + α R}

(14)

is proportional to the maximum share of income group $B$ is willing to spend in order to take power. For any $R ⩾ H_{2}^{*} (I_{2})$ , $f (I_{2}, R)$ is strictly increasing in $R$ and strictly decreasing in $I_{2}$ (see Lemma 1 in Online Appendix A). If public revenues increase, the rents that come with political power rise, too; as a result, group $B$ is willing to spend a larger income share on its militia. On the other hand, an increase in the degree of institutional cohesiveness, $I_{2}$ , implies a rise in the provision of public services in period 2 and hence a reduction in the importance of transfers for second-period utility.

To determine the actual spending on its militia, group $B$ considers $M_{1}^{G}$ , the funding of the regular military forces. If $M_{1}^{G} < ζ {\bar{M}}_{1}^{O},$ the maximum willingness to spend on taking political power exceeds the amount that is actually necessary (equation 5). In this case, $M_{1}^{O} = ζ^{- 1} M_{1}^{G} + ε$ , where $ε \to 0^{+}$ ensures that $ζ M_{1}^{O}$ is marginally greater than $M_{1}^{G}$ such that power moves from $A$ to $B$ . Otherwise, if $M_{1}^{G} ⩾ ζ {\bar{M}}_{1}^{O}$ , group $B$ does not want to take power, implying that militia spending is zero. To summarize:

Proposition 2: Suppose $R > H_{2}^{*} (I_{2}) .$ Then, the spending by group $B$ on its militia is given by

M_{1}^{O} = {\begin{matrix} 0 & : & M_{1}^{G} ⩾ ζ {\bar{M}}_{1}^{O} \\ ζ^{- 1} M_{1}^{G} + ε & : & M_{1}^{G} < ζ {\bar{M}}_{1}^{O} \end{matrix},

(15)

where $ε \to 0^{+}$ and ${\bar{M}}_{1}^{O}$ rises in the period 1 individual income in group $B$ .

So there are two variables that can be used to influence group $B$ ’s spending on its militia: military spending, $M_{1}^{G},$ and the period 1 income of the group members.

Proof. In the text. ■

Reducing the level of individual income in group $B$ raises the utility loss associated with a given level of militia spending (i.e. increases the marginal cost of funding the militia). This, in turn, reduces the maximum willingness to spend on taking power.

Decisions by the group in power in period 1

In this section, second-period institutional cohesiveness is considered exogenous (and equal to the first-period level). So we henceforth use $I_{2} = I_{1} = λ^{l}$ . When it comes to the decision by the group in power, it is again necessary to distinguish between the cases $R ⩽ H_{2}^{*} (λ^{l})$ and $R > H_{2}^{*} (λ^{l}) .$ In the first case, as discussed above, holding political power in period $2$ does not entail any advantage. So group $A$ does not want to spend any positive amount on the military: $M_{1}^{G} = 0 .$ Yet, because group $B$ will not spend any positive amount on its militia either ( $M_{1}^{O} = 0$ ), group $A$ continues to hold power in period $2$ (equation 5). Since military spending is equal to zero, as is the case in period $2$ , group $A$ solves an optimization problem that is equivalent to the second-period optimization problem described by equation (7). As a result, the period 1 spending vector is identical to that in the second period. This means $H_{1} = R$ and $M_{1}^{G} = T_{1, A} = T_{1, B} = 0$ .

Now suppose that $R > H_{2}^{*} (λ^{l})$ . In this case, holding political power in period 2 carries an advantage. As a result, group $B$ is willing to spend a strictly positive amount on taking power (equation 13). Given this, group $A$ has two a priori reasonable options when it comes to military spending. One option, to which we will return further below, is to set $M_{1}^{G} = 0$ and accept the loss of power. The alternative option is to spend at least the minimum amount necessary to retain political power in period 2:

M_{1}^{G} ⩾ ζ {\bar{M}}_{1}^{O} \equiv {\underline{M}}_{1}^{G} .

(16)

Any level of military spending that is strictly between zero and ${\underline{M}}_{1}^{G}$ is sub-optimal: reducing $M_{1}^{G}$ to zero would free up resources for public services and transfers, thereby increasing the period 1 individual income in group $A$ (with the second-period income unchanged). Using the explicit expression for ${\bar{M}}_{1}^{O}$ (equation 13), we obtain

{\underline{M}}_{1}^{G} = ζ f (λ^{l}, R) [χ H_{1}^{α} + 2 T_{1, B}] .

(17)

Moreover, considering that there are no incentives to make transfers to group $B$ that exceed the minimum mandated by the institutional constraint (4), i.e. that $T_{1, B} = (R - H_{1} - M_{1}^{G}) λ^{l}$ $(1 + λ^{l})^{- 1}$ , equation (17) can be rewritten as

{\underline{M}}_{1}^{G} = ζ f (λ^{l}, R) [χ H_{1}^{α} + 2 \frac{λ^{l}}{1 + λ^{l}} (R - H_{1} - M_{1}^{G})] .

(18)

Now suppose that group $A$ wants to retain political power in period $2$ . Then, when it comes to the maximization of the period 1 level of individual consumption,

c_{1, A} = χ H_{1}^{α} + 2 T_{1, A} = χ H_{1}^{α} + 2 \frac{1}{1 + λ^{l}} (R - H_{1} - M_{1}^{G}),

(19)

two constraints must be observed. First, total transfers must not be negative: $R - H_{1} - M_{1}^{G} ⩾ 0 .$ Second, actual military spending must not be less than the minimum necessary to discourage group $B$ from taking power: $M_{1}^{G} ⩾$ ${\underline{M}}_{1}^{G}$ . As a result, group $A$ solves the following constrained optimization problem:

\max_{{H_{1}, M_{1}^{G}}} {c_{1, A}} s . t . H_{1} + M_{1}^{G} - R ⩽ 0 and - M_{1}^{G} + {\underline{M}}_{1}^{G} ⩽ 0,

(20)

where $c_{1, A}$ and ${\underline{M}}_{1}^{G}$ are stated in equations (19) and (18), respectively.

Proposition 3: Suppose $R > H_{2}^{*} (λ^{l})$ . If group $A$ wants to retain political power in period 2, it solves optimization problem (20). The solution to this optimization problem is

H_{1} = {\frac{α χ (1 + λ^{l})}{2} [1 - \frac{1 - λ^{l}}{1 + λ^{l}} 2 ζ f (λ^{l}, R)]}^{1 / (1 - α)}

(21)

and

M_{1}^{G} = {\underline{M}}_{1}^{G} = \frac{2 ζ f (λ^{l}, R)}{1 + 2 ζ f (λ^{l}, R) λ^{l} / (1 + λ^{l})} [\frac{1}{2} χ H_{1}^{α} + \frac{λ^{l}}{1 + λ^{l}} (R - H_{1})]

(22)

and involves strictly positive total transfers: $R - H_{1} - M_{1}^{G} > 0$ .

Since $f (λ^{l}, R) > 0$ (Lemma 1 in Online Appendix A), it follows that first-period spending on public services is strictly less than second-period spending: $H_{1} < H_{2}^{*} (λ^{l}) .$

Proof. See Online Appendix B. ■

If group $A$ wants to retain political power in period $2$ , why does the group spend less on public services in the first period? The answer lies in the following economic logic. An increase in first-period spending on public services raises the period 1 individual income in group $B$ and so lifts the maximum amount the group is willing to spend on taking power (equation 13). Therefore, to retain power, group $A$ must ensure that an increase in the spending on public services is accompanied by an appropriate increase in military spending. However, this raises the perceived marginal cost of providing public services and therefore reduces $H_{1}$ below the second-period level. Returning to Proposition 2, a different way of framing this result is to say that group $A$ makes use of both instruments at its disposal to discourage group $B$ from taking power. Besides relying on the military, group $A$ also reduces the period 1 individual income in group $B$ in order to lower the group’s willingness to spend on its militia.¹⁰

Still assuming that group $A$ wants to retain political power, we calculate the corresponding period 1 level of individual consumption. By using the explicit expression for ${\underline{M}}_{1}^{G}$ (Proposition 3) in equation (19), and then rearranging terms, we obtain

c_{1, A} |_{P_{2} = A} = {[1 + 2 ζ f (λ^{l}, R) \frac{λ^{l}}{1 + λ^{l}}]}^{- 1} \frac{2}{(1 + λ^{l}) α} [(1 - α) H_{1} + α R] .

(23)

A comparison with equation (9) shows that in period 1 the level of individual consumption is lower than in period 2. This reflects that in period 1 resources are spent on the military and spending on public services is lower than in period 2.

Now let us return to the option $M_{1}^{G} = 0$ . If chosen, the first-period optimization problem to be solved by group $A$ simplifies and becomes equivalent to the second-period problem (7). The period 1 level of individual consumption in group $A$ is thus given by equation (9):

c_{1, A} |_{P_{2} = B} = c_{2, A} (A, λ^{l}, R) .

(24)

As discussed above, this level exceeds the level reached if group $A$ retains political power: $c_{1, A} |_{P_{2} = B} > c_{1, A} |_{P_{2} = A}$ . However, if $M_{1}^{G} = 0,$ group $B$ will take power in period $2$ , implying that the period 2 level of individual consumption in group $A$ is given by equation (10). So a higher period 1 level of individual consumption comes at the cost of a lower level in period $2$ . To see which of the two options— $M_{1}^{G} =$ ${\underline{M}}_{1}^{G}$ or $M_{1}^{G} = 0$ —group $A$ prefers, one must compare individual overall utilities. Retaining power is strictly preferred if

\ln [c_{1, A} |_{P_{2} = A}] + \ln [c_{2, A} (A, λ^{l}, R)] > \ln [c_{1, A} |_{P_{2} = B}] + \ln [c_{2, A} (B, λ^{l}, R)],

(25)

where the four consumption levels are specified in equations (23), (9), (24), and (10). The left-hand side of condition (25) represents individual overall utility if group $A$ retains political power. The right-hand side represents overall utility if power is ceded.

Proposition 4: Suppose $R > H_{2}^{*} (λ^{l})$ . Then, condition (25) is satisfied, implying that group $A$ retains political power in period 2 by adopting the spending vector described in Proposition 3.

Proof. See Online Appendix B. ■

Proposition 4 implies that group $B$ emerges as the loser of the power contest. In fact, it refrains from entering the contest. Relative to group $A$ , group $B$ is in a weak position because any militia spending is less productive than military spending. Moreover, group $B$ would have to bear the full cost of the militia, while the resources for the military come from the government budget and hence are only partly borne by group $A$ .

Proposition 4 completes the description of the equilibrium outcome. Here is a brief summary:

Proposition 5: In the Subgame Perfect Nash Equilibrium, we observe the following decisions:

(a) If $R ⩽ H_{2}^{*} (λ^{l}),$ in both periods $t \in {1, 2}$ , there are no transfers and public revenues are entirely spent on public services: $H_{t} = R$ and $T_{t, A} = T_{t, B} = 0 .$ Since holding political power does not entail any rents, military spending is zero and the group out of power does not attempt to take power in period 2: $M_{1}^{G} = M_{1}^{O} = 0 .$

(b) If $R > H_{2}^{*} (λ^{l}),$ in both periods $t \in {1, 2}$ , there are positive transfers: $T_{t, A} > 0$ and $T_{t, B} ⩾ 0 .$ In period 1, the spending on public services and the military is described by Proposition 3: $H_{1} < H_{2}^{*} (λ^{l}) = H_{2}$ and $M_{1}^{G} = {\underline{M}}_{1}^{G}$ , implying that the group out of power does not attempt to take power in period 2 (i.e. $M_{1}^{O} = 0$ ).

Proof. The proposition follows from the discussion in the text. ■

To secure political power if $R > H_{2}^{*}$ , group $A$ uses a repression approach that rests on two pillars: military force and underspending on public services—a term that refers to the purposeful reduction in the provision of public services below the level the group would prefer if the group’s power were not threatened. In formal terms, underspending is the difference between $H_{1}$ and $H_{2}^{*}$ , where $H_{2}^{*}$ is the preferred level in absence of a threat. Also note that the group in power does not try to influence the other group by making transfers that would go beyond the level mandated by the political institutions. In fact, such “appeasement” would be counterproductive as it would lower the cost (in terms of period 1 individual utility) the other group would have to bear in an attempt to take power.

Comparative-static properties

We now describe the first-period relationship between public revenues, public services, and military spending, and also examine the influence of the parameters.

Proposition 6: In period 1, the relationship between public revenues and public services is hump-shaped (Figure 1a): $H_{1}$ first increases one-to-one in $R$ , reaches the global maximum at $R = H_{2}^{*} (λ^{l})$ , and then monotonically falls towards the limit

lim_{R \to \infty} H_{1} = H_{2}^{*} (λ^{l}) {[1 - \frac{{(1 - λ^{l})}^{2}}{(1 + λ^{l})} ζ]}^{1 / (1 - α)} < H_{2}^{*} (λ^{l}) .

(26)

In period 1, the relationship between public revenues and the ratio of military spending to the spending on public services is convex (Figure 1b): $M_{1}^{G} / H_{1}$ is zero as long as $R ⩽ H_{2}^{*} (λ^{l})$ and then monotonically rises towards infinity as $R \to \infty$ .

Figure 1.

Public revenues and public spending. (a) Relationship between $R$ and $H_{1} .$ (b) Relationship between $R$ and $M_{1}^{G} / H_{1}$ .

Proof. The proposition follows from Proposition 5, equations (21) and (22), and the properties of f stated in Lemma 1 (in Online Appendix A).■

At low levels of public revenues, the marginal effect of public services on private-sector output is large and prevents the group in power from making transfers. As a result, there is a positive association between $R$ and $H_{1} .$ However, as soon as $R$ exceeds $H_{2}^{*}$ , the group in power spends a positive share of public revenues on transfers, implying that additional revenues increase the rents associated with holding power. All else equal, the maximum amount the group out of power is prepared to spend on taking power rises with the rents, and so the group in power increasingly uses underspending. At the same time, the latter tends to spend more on the military, with the result that the ratio $M_{1}^{G} / H_{1}$ rises without limit.

Proposition 7: An improvement in institutional cohesiveness, $λ^{l}$ , and/or the productivity of private-sector technologies, $χ$ , leads to an increase in the threshold $H_{2}^{*} (λ^{l})$ and, in the case of $R > H_{2}^{*} (λ^{l})$ , to an increase in $H_{1}$ (Figure 2a).

If $R > H_{2}^{*} (λ^{l}),$ a reduction in the relative productivity of the spending on the private militia, $ζ$ , leads to an increase in $H_{1}$ (Figure 2b).

Proof. The proposition follows from Proposition 5, equations (8) and (21), and the properties of f stated in Lemma 1 (in Online Appendix A). ■

Figure 2.

Comparative statics. (a) Rise in $λ^{l} (λ^{l, 2} > λ^{l, 1})$ . (b) Fall in $ς (ς^{2} < ς^{1})$ .

$H_{2}^{*}$ is an increasing function of $λ^{l}$ because individual marginal benefits from transfers fall in $λ^{l} .$ Similarly, $H_{2}^{*}$ is an increasing function of $χ$ because the opportunity cost of transfers rises in $χ$ . If $R > H_{2}^{*}$ , $λ^{l}$ and $χ$ exert a positive effect on $H_{1}$ because both variables increase the period 2 individual income in the group out of power: a higher period 2 individual income implies a lower marginal utility of consumption and hence a lower benefit from taking power; as a result, the willingness to pay for power falls, allowing the group in power to cut the extent of underspending on public services without losing political control.

Finally, for the case $R > H_{2}^{*}$ , Proposition 7 establishes that a rise in $ζ$ impairs the provision of public services.¹¹ Why? With fixed values of $H_{1}$ and $M_{1}^{G}$ , a rise in $ζ$ would lead to a power transition. To prevent this, the group in power lowers $H_{1}$ , with the aim of reducing the period 1 individual income in the group out of power—and therefore the maximum amount the latter group is willing to spend on its militia.

Endogenous institutional cohesiveness

Repression vs reform

If political power offers privileged access to transfers, holding on to it demands costly repression in the form of military spending and underspending on public services. The intensity of repression, and therefore its cost, could be reduced if in period 1 the group in power found a way to commit to a more equal distribution of transfers in period 2: if it becomes clear that holding political power will offer fewer benefits in the future, the incentives to take power are weaker. This section explores under what circumstances the group in power would commit to a more equal distribution of transfers in period $2$ if it had the option to do so. The Chad–Cameroon Pipeline Project contained elements—such as foreign-based escrow accounts in combination with external monitoring of Chad’s compliance with spending priorities—that could help governments make such commitments.

In the model, $I_{t}$ determines the extent to which transfers can differ between the two groups. A higher value means that the political institutions are more cohesive in the sense that the group out of power must receive a larger share of aggregate transfers. In the previous section, we maintained the assumption that the degree of institutional cohesiveness is exogenous and unalterable: $I_{1} = I_{2} = λ^{l}$ . In this section, we continue to assume that $I_{1} = λ^{l}$ , but we give the group that initially holds power the opportunity to commit to an increase in the degree of institutional cohesiveness in period 2: $I_{2} \in {λ^{l}, λ^{h}}$ , where $λ^{l} < λ^{h} ⩽ 1 .$ $I_{2}$ is determined together with the first-period spending vector. If $I_{2} = λ^{h}$ , we say that the group in power commits to an institutional reform. As a large part of the analysis above continues to apply, we can focus on the first-period choice problem faced by the group in power.

Decisions by the group in power in period 1

If $R ⩽ H_{2}^{*} (λ^{l})$ , group $A$ , i.e. the group in power, is not interested in repression as holding power does not involve any privileges. In what follows, we therefore focus on the case $R > H_{2}^{*} (λ^{l})$ . In this situation, if $I_{2} = I_{1} = λ^{l}$ , the previous section implies that in period $2$ the group in power is better off than the other group. As a result, if the degree of institutional cohesiveness is left unchanged, retaining power means repression (as described in the previous section). The same holds if $I_{2} = λ^{h} > λ^{l}$ but $R > H_{2}^{*} (λ^{h}) .$ In this situation, again, holding power in period 2 involves rents and only repression will allow group $A$ to retain power. However, if $H_{2}^{*} (λ^{l}) < R ⩽ H_{2}^{*} (λ^{h})$ , group $A$ can avoid repression altogether by raising the degree of institutional cohesiveness: $I_{2} = λ^{h}$ ensures that in period 2 public revenues will be used for public services only; as a result, repression in the form of military spending and underspending on public services would be useless.

Yet will the group in power in fact abstain from repression and improve cohesiveness if $H_{2}^{*} (λ^{l}) < R ⩽ H_{2}^{*} (λ^{h})$ ? The group prefers this option if

\ln [c_{1, A} |_{P_{2} = A}] + \ln [c_{2, A} (A, λ^{l}, R)] ⩽ \ln [c_{1, A} |_{ref}] + \ln [χ R^{α}],

(27)

where the left-hand side of condition (27) represents individual overall utility under repression (identical to the left-hand side of condition 25). The period 1 and 2 levels of individual consumption under institutional reform are given by $c_{1, A} |_{ref}$ and $χ R^{α}$ , respectively.¹² If, in fact, group $A$ chooses reform, military spending is zero ( $M_{1}^{G} = 0$ ) and the group’s optimization problem becomes equivalent to the second-period optimization problem (7). Therefore,

H_{1} |_{ref} = {[\frac{α χ (1 + λ^{l})}{2}]}^{1 / (1 - α)} = H_{2}^{*} (λ^{l})

(28)

and

c_{1, A} |_{ref} = \frac{2}{(1 + λ^{l}) α} [(1 - α) H_{1} |_{ref} + α R] .

(29)

Using equations (23), (9), and (29), condition (27) can be rewritten as

D \equiv χ R^{α} [1 + \frac{λ^{l}}{1 + λ^{l}} 2 ζ f (λ^{l}, R)] - \frac{2}{(1 + λ^{l}) α} [(1 - α) H_{1} |_{no ref} + α R] ⩾ 0,

(30)

where $H_{1} |_{no ref}$ is given by equation (21). If $H_{2}^{*} (λ^{l}) < R ⩽ H_{2}^{*} (λ^{h})$ , $D$ gives the gain in individual overall utility that results if group $A$ switches from repression to institutional reform. As we show in Lemma 2 in Online Appendix A, the properties of $D$ imply that there exists a unique $\hat{R}$ such that $D (\hat{R}) = 0$ , $D (R) > 0$ for all $R \in$ $(H_{2}^{*} (λ^{l}), \hat{R}),$ and $D (R) < 0$ for all $R > \hat{R}$ . To complete the characterization of the equilibrium, we introduce the following definition:

R^{*} \equiv \min {H_{2}^{*} (λ^{h}), \hat{R}} .

(31)

Proposition 8: There exists a nonempty interval $(H_{2}^{*} (λ^{l}), R^{*}]$ such that for all $R \in (H_{2}^{*} (λ^{l}), R^{*}]$ the group in power opts for institutional reform in period 1: $I_{2} = λ^{h} > λ^{l} = I_{1} .$

As a result, if $R \in (H_{2}^{*} (λ^{l}), R^{*}]$ , in period 1 we have $M_{1}^{G} = 0$ and $H_{1} = H_{1} |_{ref} = H_{2}^{*} (λ^{l})$ —which implies positive transfers: $T_{1, A} > 0,$ $T_{1, B} ⩾ 0 .$ In period 2, there are no transfers and public revenues are entirely spent on public services: $H_{2} = R,$ $T_{2, A} = T_{2, B} = 0 .$

Proof. The proposition follows from the discussion in the text, in particular the definition of $D$ (equation 30), Lemma 2 (in Online Appendix A), and the fact that $H_{2}^{*} (λ^{h}) > H_{2}^{*} (λ^{l}) .$ ■

If $R$ is in the neighborhood of $H_{2}^{*} (λ^{l})$ , the marginal product of spending on public services is comparatively large; so, in period $2$ , the individual utility loss is just small if group $A$ forgoes a small rent in favor of better public services. Therefore, when it comes to the approach to defense, the group prefers institutional reform over repression.¹³ However, as the distance between $R$ and $H_{2}^{*} (λ^{l})$ increases, reform becomes a comparatively costly approach; thus, at some point, there will be a switch to repression.

The following proposition describes the equilibrium outcome for any level of public revenues:

Proposition 9: In the Subgame Perfect Nash Equilibrium, we observe the following decisions:

(a) If $R ⩽ H_{2}^{*} (λ^{l}) < R^{*}$ , political power does not offer privileged access to transfers, implying that the group in power (out of power) is not interested in repression (taking power). As a result, the two groups act as described in Proposition 5 a.

(b) If $H_{2}^{*} (λ^{l}) < R ⩽ R^{*}$ , the group in power responds to the threat by the group out of power by relying on the institutional reform approach, thereby avoiding repression altogether. As a result, the two groups act as described in Proposition 8.

(c) If $H_{2}^{*} (λ^{l}) < R^{*} < R$ , the group in power responds to the threat by the group out of power by relying on repression as described in Proposition 5 b, either without or with institutional reform. In the latter case, $f (I_{2}, R)$ is evaluated at $I_{2} =$ $λ^{h}$ and $H_{2} = H_{2}^{*} (λ^{h})$ .

Proof. See Online Appendix B. ■

Figure 3 illustrates the relationship between public revenues and the spending on public services. In Figure 3a, the reform-only range is relatively small (i.e. $R^{*} < H_{2}^{*} (λ^{h})$ ); in Figure 3b, it is larger and coincides with the range $(H_{2}^{*} (λ^{l}), H_{2}^{*} (λ^{h})] .$

Figure 3.

Public revenues and institutional reform. (a) Narrow reform-only range. (b) Broad reform-only range.

What determines whether the economy is in situation 3a or 3b? An important factor in this regard is the relative productivity of the spending on the militia, $ζ$ :

Proposition 10: For given other parameter values, $R^{*} < H_{2}^{*} (λ^{h})$ holds if the relative productivity of the spending on the private militia, $ζ$ , is sufficiently small. If $R^{*} < H_{2}^{*} (λ^{h}),$ an increase in $ζ$ raises $R^{*} .$

Proof. See Online Appendix B. ■

So the group in power is more inclined to abstain from repression if it has a smaller advantage when it comes to military force. Together, Propositions 7 and 10 imply that $ζ$ plays an ambiguous role. On the one hand, a decrease in the relative productivity of the spending on the militia improves the provision of public services if the group in power uses repression. On the other hand, such a decrease narrows the range in which $R$ must lie so that the group in power abstains from repression.

Finally, even in the situation described by Figure 3b, the group in power will abstain from institutional reform if public revenues are sufficiently large:

Proposition 11: There exists a level of public revenues $\bar{R} \in [R^{*}, \infty]$ such that for any $R > \bar{R}$ the group in power relies on repression without institutional reform.

Proof. See Online Appendix B. ■

From the perspective of the group in power, complementing repression with institutional reform brings benefits in terms of both lower military spending and less underspending on public services. However, institutional reform results in a higher second-period transfer to the group out of power. As public revenues rise, this “redistribution” becomes more costly. If $R$ is sufficiently large, the cost of institutional reform must outweigh the benefit.

Two cases: Chad and Azerbaijan

This section examines two recent cases to which our framework can be applied: Chad and Azerbaijan. Both countries experienced a surge in public revenues against the backdrop of significant needs for increased spending on public health, education, and infrastructure; in both countries, this surge was driven by a natural resource windfall, in particular a rapid development of the oil sector; finally, in both countries, the power of the executive has only been weakly constrained (see, e.g. the Polity IV project). We document that there are parallels in the two countries’ experiences with surging public revenues and discuss how those shared experiences can be interpreted through the lens of our framework.

The Chad–Cameroon pipeline project

Ever since its independence from France in 1960, the Central African country of Chad has been among the poorest in the world. Since 2003, however, the government of Chad has benefited financially from a new oil pipeline—the Chad–Cameroon pipeline—that connects Chad’s vast oil reserves in the south of the country with offshore export facilities in the Gulf of Guinea (see, e.g. World Bank, 2009). The project involved many parties, among them the government of Chad and the World Bank. The intention of the World Bank was to transform the riches in the ground into funds that could be used to alleviate poverty and spur economic development. This idea prompted Chad to pass the 1999 Revenue Management Law, which required the allocation of 85% of the oil revenues into priority sectors (among them education, health services, rural development, infrastructure, and environmental and water resources).

The new pipeline is generally considered a financial success. In the year 2000, total annual rents generated by Chad’s natural-resource sector amounted to just $$ 105$ p.c. (constant 2011 international dollars), according to data from the World Development Indicators. By 2010, 10 years after the final approval of the project, the rents had surged to $$ 471$ p.c.—with positive consequences for the public purse: between the years 2000 and 2010, the project had generated $$ 6.3$ billion (current US dollars) in cumulative government revenues.¹⁴ However, this surge in public revenues is hardly reflected in higher spending on poverty alleviation and economic development. For instance, public-health spending fell from $$ 26$ p.c. (constant 2011 international dollars) in 2000 to $$ 23$ p.c. in 2010 and has surpassed the earlier threshold only in recent years. Moreover, numerous reports of inflated prices in public procurement suggest that part of the spending should be regarded as transfers rather than investments in infrastructure and public services. It is therefore no surprise to find Chad still among the least developed countries in the world. So, despite the safeguards put in place on the World Bank’s request, the hopes associated with the pipeline project have been dashed.

What went wrong? Detailed accounts of the pipeline project (e.g. Pegg, 2006, 2009; van Dijk, 2007) suggest that growing political instability in the years 2004–2006 played a significant role. Over this period, the government of President Idriss Déby faced several coup attempts and a growing rebellion involving armed opposition groups. As for the motives of these opposition groups, van Dijk (2007, p. 701) notes that it appears that

they are after the goose with the golden eggs: the Chad–Cameroon pipeline project and its revenues.

As a result, President Déby wanted to step up military spending. In December 2005, the Chadian parliament amended the 1999 Revenue Management Law, adding the security sector (among others) to the list of priority sectors (in violation of its Loan Agreement with the World Bank—see Pegg, 2009, p. 313). In the ensuing conflict with the World Bank, President Déby succeeded in weakening the conditions of the original pipeline deal further. By mid-2006, the Chadian government was no longer bound by the stringent external conditions of the original project and could step up military spending in order to secure power. Between the years 2000 and 2010, military spending per capita rose by a factor of $7.4$ (from $$ 15$ to $$ 111$ in constant 2011 international dollars), according to data from the World Development Indicators.

For Chad, near complete data on natural-resource rents, public-health spending, and military spending are available from the World Development Indicators for about two decades, from the mid-1990s to the mid-2010s. Over this period, as illustrated by Figure 4a, a clear pattern emerges: the ratio of military spending to public-health spending shows a positive relationship with natural-resource rents p.c.¹⁵ Taking public-health spending as a proxy for spending on poverty alleviation and development more generally, Figure 4a suggests that in periods of high government revenues the goal of strengthening the military gains in importance relative to development goals. Given that the vertical axis of Figure 4a uses a logarithmic scale, the association between the spending ratio and natural-resource rents p.c. is in effect convex. In quantitative terms, the relationship is relatively strong: using the estimated semi-elasticity, we find that an increase in natural-resource rents p.c. by $$ 100$ is associated with a 30.8% rise in the ratio of military spending to public-health spending.

Figure 4.

Natural resource rents and the ratio military to public-health spending, 1994 to 2013. (a) Chad. (b) Azerbaijan.

The Azeri–Chirag–Gunashli Development Project

Azerbaijan, a middle-income country in Southwestern Asia, gained independence from the Soviet Union in 1991. In 1994, the country signed the so-called “Contract of the Century”, an agreement with an international consortium of oil companies that would develop three large offshore oil fields in Azerbaijan’s section of the Caspian Sea (Guliyev, 2014). Combined, the Azeri, Chirag, and Gunashli (ACG) oil fields hold proved reserves of about five billion barrels, a magnitude that is sizable even by international standards (Guliyev, 2014). So it is no surprise that the Contract of the Century has had a large impact on the development of the Azerbaijani natural-resource sector. Data from the World Development Indicators suggest that in 2004, 10 years after the conclusion of the contract, annual natural-resource rents reached $1,875 p.c. (constant 2011 international dollars), up from just $$ 573$ p.c. in 1994. To the best of our knowledge, there is no publicly available information on the impact of this development on Azerbaijan’s public revenues. However, as in the case of the Chad–Cameroon pipeline project, the impact is expected to be large: Guliyev (2014), for instance, estimates that the Azerbaijani government receives 70–75% of profits generated by the ACG project.

Regarding consequences, there are at least two parallels to what happened in Chad. First, there is little evidence suggesting that Azerbaijan’s government significantly stepped up public spending on poverty alleviation and economic development. De Waal (2016) notes that much of the new revenue has been spent on prestige projects and the military, while sectors important for poverty alleviation and economic development have received little.¹⁶ This is partly reflected in the data on public health spending. The World Development Indicators suggest that during the first 10 years after the conclusion of the contract, public-health spending p.c. was roughly stagnant and increased markedly only after 2004, when natural-resource rents exploded (reaching $5,481 p.c. in 2008, 10 times the level of 1994).

The second parallel to the case of the Chad–Cameroon pipeline project is that the ACG development project also appeared to amplify pre-existing political tensions during the period following the conclusion of the contract. Referring to a coup attempt against the government of President Heydar Aliev—happening in October 1994, about one month after the conclusion of the Contract of the Century—Cornell (2001, p. 120) notes:

Interestingly, this occurred immediately after the signing of a $7 billion oil deal between Azerbaijan and foreign, mainly Western, oil companies to develop three Azerbaijani offshore oilfields in the Caspian Sea.

However, as from 1996, Aliev was able to consolidate his power (Cornell, 2001), a process that took place against the backdrop of rising military spending. The World Development Indicators suggest that military spending rose from $$ 75$ p.c. (constant 2011 international dollars) in 1996 to $$ 169$ p.c. in 2004 to $$ 762$ p.c. in 2014.¹⁷ While the volatile regional situation had certainly been one driving force behind this military build-up, it is plausible that another important motivation was to strengthen the capacity to fight internal opposition.

How did the relative importance of the goal of strengthening the military respond to changes in public revenues? Figure 4b, which parallels Figure 4a and also covers the period from the mid-1990s to the mid-2010s, suggests a similar pattern to that in Chad: there is a significant positive association between the ratio of military spending to public-health spending and natural-resource rents p.c. As compared with Chad, the relationship is somewhat less strong in quantitative terms: using the estimated semi-elasticity, we find that an increase in natural-resource rents p.c. by $$ 100$ is associated with a rise in the ratio by about 0.9%.

In summary, neither in Chad nor in Azerbaijan did the natural resource windfall, and the associated relaxation of fiscal constraints, have the hoped-for consequences for public spending on poverty alleviation and economic development. Yet both countries experienced a steep increase in military spending, along with rising political instability.

Discussion

Our framework offers a coherent perspective on the experiences shared by Chad and Azerbaijan. In particular, it provides a possible explanation for the relationship between public revenues and the composition of public spending documented in Figure 4. In both countries, there is a strong convex relationship between, on the one hand, the ratio of military to public-health spending and, on the other hand, natural-resource rents p.c. Figure 1b illustrates that the framework implies a similar pattern. In the framework, the relationship arises because—if future public revenues exceed a certain threshold—additional revenues induce the group in power to increase military spending and reduce the provision of public services in an attempt to weaken the incentives of the other group to take power. Strengthening the military increases the total financial effort the group out of power must incur to take power. Reducing the provision of public services increases the marginal cost of funding such effort. Anecdotal evidence from both countries suggests that, in fact, it was the anticipation of higher public revenues in the future that strengthened the inclination of opposition groups to attack the government. We further note that the framework offers an explanation for government measures (like stifling smaller businesses) intended to harm the broader economy: such measures can be rationalized as an integral part of what we call the repression approach to internal threats. It is not obvious how existing models, such as those emphasizing that internal threats weaken governments’ incentives to invest by shortening their time horizons (e.g. Besley and Persson, 2011a; Oechslin, 2010), could account for such intentionally harmful measures.

Our framework may further shed light on the adoption and eventual collapse of the safeguards that constituted an integral part of the Chad–Cameroon pipeline project. As described above, the original pipeline deal included provisions that should ensure that the lion’s share of the additional revenues are used for the purpose of poverty alleviation and economic development. Shortly after the project had started to generate revenues in 2003, it became clear that the rise in public revenues would be much larger than initially anticipated.¹⁸ In the following years, political tensions rose and, in the fall of 2005, the Chadian government started to dismantle the safeguards. In the model, the introduction of such safeguards amounts to a commitment to raise the cohesiveness of political institutions, a commitment the group in power may be interested in to reduce the cost of repression. However, the model also suggests that institutional reform is an ineffective defense in the case of a very large increase in public revenues. The model thereby offers an interpretation of the sequence of events in Chad: when the pipeline deal was concluded, the actors assumed that the new level of public revenues, R, would fall on the flat portion of the curve in Figure 3b and so the safeguards were adopted; however, shortly after the oil revenues had started to flow in, it became clear that, in terms of Figure 3b, the new R would lie more to the right—with the result that political tensions rose and the government wanted to drop the safeguards.

In Azerbaijan, too, evidence of a sustained institutional reform is lacking. Looking through the lens of our framework, this is no surprise as the impact of the “Contract of the Century” on public revenues is thought to be very large. Nor is it a surprise from the wider perspective of the literature on foreign aid and natural resource windfalls, which often emphasizes that increases in public revenues harm political institutions in the long run (see, e.g. Deaton, 2013 and van der Ploeg, 2011, for overviews).¹⁹ However, in comparison to that literature, our analysis offers a less unequivocal conclusion. Relying on a framework in which the group in power can respond to internal threats with repression and/or institutional reform, we identify circumstances in which an exogenous increase in public revenues improves political institutions and economic outcomes. This finding suggests that safeguards such as those envisaged as part of the Chad–Cameroon pipeline project may be helpful when it comes to smaller projects, in the context of which governments may be interested in a credible commitment to improve institutional cohesiveness in order to avoid costly repression.

Conclusion

The Chad–Cameroon pipeline project and the Azeri–Chirag–Gunashli development project are examples of how in fiscally weak states large exogenous increases in public revenues can make matters worse. By establishing that the use of measures to impair private-sector productivity can be understood as an integral part of repression, our analysis offers a novel perspective on the experiences of Chad and Azerbaijan. However, we do not suggest that these experiences are unavoidable. Our framework implies that additional resources are used to promote economic development if private-sector technologies are sufficiently productive or political institutions sufficiently cohesive. The framework further identifies circumstances in which governments may even find it optimal to start institutional reforms in an attempt to ease rising tensions caused by higher public revenues. Yet by suggesting that the consequences of an increase in public revenues are highly country-specific, the framework cautions against calls to indiscriminately increase the flow of resources towards fiscally weak countries.

Supplemental Material

CMP894735_supp_mater – Supplemental material for Fiscal weakness, the (under-)provision of public services, and institutional reform

Supplemental material, CMP894735_supp_mater for Fiscal weakness, the (under-)provision of public services, and institutional reform by Manuel Oechslin and Mauricio Rodriguez in Conflict Management and Peace Science

Footnotes

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Mauricio Rodriguez

Supplemental material

The Online Appendix is available online through the SAGE CMPS website:

Notes

References

Acemoglu

(2005) Politics and economics in weak and strong states. Journal of Monetary Economics 52(7): 1199–1226.

Acemoglu

Johnson

(2005) Unbundling institutions. Journal of Political Economy 113(5): 949–995.

Acemoglu

Robinson

(2000) Why did the West extend the franchise? Democracy, inequality and growth in historical perspective. Quarterly Journal of Economics 115(4): 1167–1199.

Acemoglu

Robinson

Santos

(2013a) The monopoly of violence: Evidence from Colombia. Journal of the European Economic Association 11(s1): 5–44.

Acemoglu

Robinson

Torvik

(2013b) Why do voters dismantle checks and balances? Review of Economic Studies 80(3): 845–875.

Barro

(1990) Government spending in a simple model of endogenous growth. Journal of Political Economy 98(5): S103–S125.

Besley

Persson

(2009) The origins of state capacity: Property rights, taxation, and politics. American Economic Review 99(4): 1218–1244.

Besley

Persson

(2011a) Fragile states and development policy. Journal of the European Economic Association 9(3): 371–398.

Besley

Persson

(2011b) Pillars of Prosperity. The Political Economics of Development Clusters. Princeton, NJ: Princeton University Press.

10.

Besley

Persson

(2013) Taxation and development. In: Auerbach

Chetty

Feldstein

Saez

(eds) Handbook of Public Economics. Amsterdam: Elsevier.

11.

Besley

Persson

Reynal-Querol

(2016) Resilient leaders and institutional reform: Theory and evidence. Economica 83(332): 584–623.

12.

Caselli

Cunningham

(2009) Leader behaviour and the natural resource curse. Oxford Economic Papers 61(4): 628–650.

13.

Cornell

(2001) Democratization falters in Azerbaijan. Journal of Democracy 12(2): 118–131.

14.

Deaton

(2013) The Great Escape. Health, Wealth, and the Origins of Inequality. Princeton, NJ: Princeton University Press.

15.

DESA, United Nations Department of Economic and Social Affairs ( 2015) The Addis Ababa Action Agenda. DESA Briefing Note.

16.

De Waal

(2016) Azerbaijan at 25: A new era of change and turbulence. Carnegie Endowment For International Peace.

17.

Diamond

Koszegi

(2003) Quasi-hyperbolic discounting and retirement. Journal of Public Economics 87, 1839–1872.

18.

Dunning

(2005) Resource dependence, economic performance, and political stability. Journal of Conflict Resolution 49(4): 451–482.

19.

Fergusson

Larreguy

Riano

(2015) Political competition and state capacity. CAF Working Papers, 2015/03.

20.

Fergusson

Robinson

Torvik

Vargas

(2016) The need for enemies. Economic Journal 126(June): 1018–1054.

21.

Guliyev

(2014) The “Contract of the Century” 20 years after: A mixed blessing. Mimeo, Jacobs University Bremen.

22.

Herbst

(2000) States and Power in Africa: Comparative Lessons in Authority and Control. Princeton, NJ: Princeton University Press.

23.

Hodler

(2006) The curse of natural resources in fractionalized countries. European Economic Review 50(6): 1367–1386.

24.

Jones

Tarp

(2016) Does foreign aid harm political institutions? Journal of Development Economics 118: 266–281.

25.

Kolstad

Wiig

(2009) It’s the rents, stupid! The political economy of the resource curse. Energy Policy 37(12): 5317–5325.

26.

Leininger

Yang

C-L

(1994) Dynamic rent-seeking games. Games and Economic Behavior 7(3): 406–427.

27.

Mehlum

Moene

Torvik

(2006) Institutions and the resource curse. Economic Journal 116(January): 1–20.

28.

Oechslin

(2010) Government revenues and economic growth in weakly institutionalised states. Economic Journal 120(June): 631–650.

29.

Oechslin

(2014) Targeting autocrats: Economic sanctions and regime change. European Journal of Political Economy 36: 24–40.

30.

Pegg

(2006) Can policy intervention beat the resource curse? Evidence from the Chad–Cameroon pipeline project. African Affairs 105(418): 1–25.

31.

Pegg

(2009) Briefing: Chronicle of a death foretold: The collapse of the Chad–Cameroon pipeline project. African Affairs 108(431): 311–320.

32.

Rasizade

(2011) Azerbaijan’s chances in the Karabakh conflict. Harvard International Review, 18 January.

33.

Robinson

Torvik

(2005). White elephants. Journal of Public Economics 89(2–3): 197–210.

34.

Robinson

Torvik

Verdier

(2006) Political foundations of the resource curse. Journal of Development Economics 79(2): 447–468.

35.

Röller

L-H

Waverman

(2001) Telecommunications infrastructure and economic development: A simultaneous approach. American Economic Review 91(4): 909–923.

36.

Sachs

(2005) The End of Poverty: Economic Possibilities for Our Time. New York: Penguin Books.

37.

Saint-Paul

Ticchi

Vindigni

(2016) A theory of political entrenchment. Economic Journal 126(June): 1238–1263.

38.

Tornell

Lane

(1999) The voracity effect. American Economic Review 89(1): 22–46.

39.

Torvik

(2002) Natural resources, rent seeking and welfare. Journal of Development Economics, 67(2): 455–470.

40.

Torvik

(2009) Why do some resource-abundant countries succeed while others do not? Oxford Review of Economic Policy 25(2): 241–256.

41.

Van der Ploeg

(2011) Natural resources: Curse or blessing? Journal of Economic Literature 49(2): 366–420.

42.

Van Dijk

(2007) Political deadlock in Chad. African Affairs, 106(425): 697–703.

43.

Wick

Bulte

(2006) Contesting resources—Rent seeking, conflict and the natural resource curse. Public Choice 128(3–4): 457–476.

44.

World Bank (2009) Program Performance Assessment Report No. 50315. Washington, DC: Independent Evaluation Group—The World Bank Group.

45.

World Bank (2013) Financing for Development Post-2015. Washington, DC: The World Bank Group.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.59 MB