1 Introduction

There has been a revival of interest in the role of the state in economic development (Centeno et al. 2017; Dincecco 2017). The analysis of state capacity, defined as the institutional capability of the state to carry out various policies that deliver benefits and services to households and firms (Besley and Persson 2011), has emerged as the cutting edge of research on the relationship between governance, institutions and long-term economic development.

The focus has been on two dimensions: fiscal capacity and legal capacity, defined as the capability of raising revenues from taxes, and of enforcing contracts and property rights, respectively. Besley and Person (2011) argue that such capacities are complementary and give rise to ‘development clusters’: groups of countries that are rich and have well-developed fiscal and legal capacities, or groups of countries that are ridden by poverty and have weak state capacity. Up to this point, the literature has mainly been concerned with the causal effect of state capacity on economic development (Dincecco and Katz 2016; Dincecco and Prado 2012). However, it has also emphasised that building fiscally capable states is at the heart of state formation and performance in providing public goods (e.g. Acemoglu 2005a).

This paper also looks at fiscal aspects, but it focuses on the financial planning side—the capability of states to design, implement and monitor their budget.Footnote 1 Strengthening such ability is strategically important to economic development, since achieving greater efficiency in public financial management implies more efficient public goods provision (see, for example, Martin-Vasquez 2014).Footnote 2 In doing this, the state needs a well-functioning bureaucracy with Weberian characteristics, which means autonomy from the government and professionalisation of its ranks (Evans and Rauch 1999).Footnote 3 This is particularly important for states in less developed countries, since they have much larger needs than high-income countries for investment in infrastructure and basic public goods and services.Footnote 4

In particular, this paper contributes to a thin literature on the long-run determinants of state capacity. Such literature has mainly been based on conditional correlations and has hitherto independently assessed the role of historical factors, such as the incidence of external and internal conflicts (Besley and Persson 2011), the experience of statehood (Bockstette et al. 2002), and the effects of geography, such as abundance of natural resources (Isham et al. 2005) or the conditions affecting population density (Herbst 2000). It has also considered the role of political systems, arguing that executive power subject to checks and balances will tend to promote common interests, rather than using the state to retain power (Besley and Persson 2011). But such political economy explanations remain an underexplored area in empirical research. We contribute to this literature by looking for the first time at the political determinants of administrative capacity for public financial management.

Our focus on constraints on the executive as a primary political determinant of administrative capacity is supported by a large literature in political science and political economy. This argues that the emergence of effective states in developing countries depends on how checks and balances on incumbents may help create a preference for more common-interest spending, primarily through two mechanisms: (1) a strong legislature which finds the need to generate broad-based coalitions, thereby offsetting the narrow focus of the executive, and (2) an independent judiciary which promotes broad-based access to public services through statutory service obligations or rights-based arguments and rulings (Besley and Persson 2011; Bardhan 2016; Centeno et al. 2017; Dincecco 2017).Footnote 5

To examine the determinants of administrative capacity, we use a recent data set on public financial management performance measures in developing countries constructed by the Public Expenditure and Financial Accountability (PEFA) project, a consortium of the World Bank and other donor agencies. For a sample of forty-seven developing countries, we find a substantial positive effect between institutions that place constraints on the executive power and current quality of budgetary institutions. Addressing identification concerns, we show that our results are robust to different specifications, controls and estimation methods. Our findings indicate that, to build financially capable states, it is important to build cohesive political institutions, providing strong checks and balances on the discretionary power of the executive.

The paper is organised as follows. Section 2 provides a conceptual framework on the relationship between political institutions and administrative capacity. Section 3 discusses our measures of administrative capacity. In Sect. 4, the empirical strategy, the data and the results are presented. Section 5 concludes.

2 The political determinants of administrative capacity

In the political science literature, there are four competing notions of state capacity (see Centeno et al. 2017). The first is the ability of the state to achieve its identified goals (implementation capacity). The second is the ability of the state to achieve an ideal set of goals usually determined by an outside party (state scope). The third is the ability of the state to impel citizens and other states to do what they may not have done otherwise (relational capacity). The fourth notion is the organisational capacity of the state (organisational or administrative capacity).

As Centeno et al. (2017) persuasively argue, the first three notions of state capacity conflate the concept of state capacity with its causes and consequences. Thus, the literature on implementation capacity, state scope and relational capacity is explicitly concerned with outcomes. In contrast, the concept of state capacity as administrative capacity is analytically neutral, and allows us to compare state capacity across politically distinct kinds of states (e.g. democratic versus authoritarian). For the purposes of this paper, we will focus on the concept of state capacity as administrative capacity, as we are interested in the causes of state capacity and not its consequences.

A large literature has studied the determinants of state capacity for developed countries, while the systematic quantitative analysis of state capacity in developing countries is still at its infancy (Dincecco 2017). For example, a well-known line of thinking associated with Tilly (1975) links the historical making of the fiscal–military state in early modern Europe with interstate wars (Bardhan 2016). State making in developed countries has also been associated with inducing citizen compliance in providing revenue (Levi 1988). In contrast, Bates (2014) and Reid (2014) argue that the ‘imperial peace’ implemented by European colonial powers cut short an indigenous process of war-related state development in Africa. Furthermore, Bates (2010) argues that ‘Cold War-era foreign aid hampered African political development by reducing the incentive of rulers to seek domestic political consensus for their policy choices’ (Dincecco 2017, p. 73).

An early contribution to the literature on the determinants of state capacity in developing countries, which focused on political factors, was Evans (1995), who argued that the ‘embedded autonomy’ of the Korean state was an important cause of the high levels of administrative capacity that one witnessed in the country, particularly in implementing the Korean political elite’s vision for long-term economic development. Here, ‘embedded autonomy’ was the combination of a Weberian bureaucratic structure autonomous of the business elites and of channels of deliberative processes involving important political stakeholders in the development process (Bardhan 2016).

More recent literature has highlighted the role of accountability mechanisms on state leadership—‘various kinds of checks and balances including constitutional constraints on executive power, separation of powers, electoral rules, independent judiciary, free media and other accountability mechanisms for the state leadership’ (Bardhan 2016, p. 871)—as necessary enabling conditions for capable states to emerge in developing countries. In particular, as Besley and Persson (2011) also argue, states with weak limitations on the executive have weak compulsions to supply common-interest services: ‘constraints on the executive will diminish the concern that the government is run in the interests of a narrow group’ (Bardhan 2016, p. 871).

Similarly, Dincecco (2017, pp. 21–22) argues that institutional impartiality is an essential precondition for effective statehood, understood as the presence of ‘an institutional player within the national government that has the formal political authority to regularly monitor state finances’. In parliamentary democracies, such a role will be played by an effective parliament, one which the political leader cannot call or disband at will. Here, an effective parliament can ‘regularly oversee the state’s budget, including authority over taxation, the right to audit previous government spending, and the right to veto new expenditures’ (Dincecco 2017, p. 22). In autocracies, such checks and balances on the executive may be weaker. However, as Besley and Kudamatsu (2008) point out, even in autocracies, there may be constraints on the autocrat placed by government insiders—the ‘selectorate’ (Bueno de Mesquita et al. 2003)—whose power does not depend on the current executive and who can remove poorly performing executives, if necessary. Thus, limits on executive power promote a common-interest environment, in which the ruling minority is unable to hand out favours to cronies or themselves (Besley and Persson 2011).

The literature emphasising the role of accountability mechanisms on state leadership has so far argued that they are important to build fiscal and legal capacity in developing economies (e.g. Besley and Persson 2009). What this literature has not investigated yet is whether such accountability mechanisms affect also administrative capacity for public financial management, which is the focus of this paper. In particular, we investigate the hypothesis that political institutions placing checks and balances on executive power foster the ability of states to deliver timely and effective financial planning. This is because, when subject to checks and balances, a ruler has less discretion over public spending decisions than one who is not. Hence, he or she may be more likely to promote an effective independent civil service (rather than one based on patronage, which may undermine the competence of the state bureaucracy) and so maintain or innovate the state’s public finance infrastructure and its ability to design, implement and monitor the budget. Similarly, with clear limitations on his or her powers, a ruler is more likely to follow the rule of law, so that the judicial system may counter rent seeking more effectively, and to have a more transparent policy process, thus reducing waste and corruption. The following testable proposition summarises our argument:

Proposition 1

Political institutions placing checks and balances on the executive power foster the ability of states to deliver effective and timely financial planning. The effect of higher constraints on the executive on the quality of budgetary institutions is positive.

The existing empirical literature has mainly been interested in the effects of state capacity, rather than in its origins, and has neither focused on the ability of states to deliver effective financial planning, nor satisfactorily measured bureaucratic quality. For example, Knack and Keefer (1995) used variables from the International Country Risk Guide data (ICRG 1997), which are among the oldest and perhaps the most commonly used measures of institutional quality in the empirical literature on institutions and economic development. The data come from subjective assessments of foreign investors and ‘business experts’. It includes government repudiation of contracts, the rule of law, the expropriation risk, corruption in government and bureaucratic quality indices. The first two capture the legal capacity of the state, the last two proxy for the level of administrative capacity, while the expropriation risk tries to express the likelihood of outright nationalisation. But what such variables capture remains somewhat methodologically and conceptually unclear (Savoia and Sen 2015).

Similarly, Rodrik et al. (2004) have used data from the Worldwide Governance Indicators (World Bank 2011) to estimate their effects on economic growth. Four such indicators may be seen as proxies for administrative capacity: rule of law, regulatory quality, government effectiveness and control of corruption. These are all subjective measures that try to improve on country coverage by aggregating the ratings from over 30 organisations (including ICRG 1997). Recently, the Quality of Government index assembled by Teorell et al. (2017) extends the temporal coverage of the ICRG database, but focuses only on three of its variables. It is calculated as the average of rule of law, corruption in government and bureaucratic quality indices from various editions of the International Country Risk Guide (the other two components seen above were discontinued after 1997).

State capacity indicators should focus on the competence and ability of the bureaucracy, since it is closer to the underlying concept. Evans and Rauch (1999) surveyed 35 developing economies to collect data with time-invariant values representing the period 1970–1990. Their ‘Weberianness Scale’ provides a measure of the degree to which bureaucratic agencies employ meritocratic recruitment and offer predictable, rewarding long-term careers.Footnote 6 They found that these ‘Weberian’ characteristics significantly enhanced economic growth after controlling for the initial levels of GDP per capita and human capital. Rauch and Evans (2000) used the same index to explain variations in the institutional quality indicators analysed by Knack and Keefer (1995). They found that these indicators were significant determinants of three out of five measures of bureaucratic performance. In particular, meritocratic recruitment is the element of bureaucratic structure that is most important for improving bureaucratic performance. Internal promotion and career stability are at best of secondary importance, whereas competitive salaries do not appear to have any effect on bureaucratic performance. Although Evans and Rauch’s approach is perhaps the most rigorous attempt to measure bureaucratic capacity (Centeno et al. 2017), it covers a fairly limited range of countries and, above all, does not focus on the quality of public financial management.

Our brief review of the empirical literature on state capacity highlights two limitations. First, while much of the literature has focused on the outcomes of high state capacity, less is known about how developing countries acquire the high levels of state capacity that are necessary for positive development outcomes. Second, existing empirical research has tended to look at state capacity as an aggregate (as expressed by catch-all measures of governance quality, commonly used in the empirical literature), without paying particular attention to the capacity of the bureaucracy to devise and implement efficient public financial management systems. Arguably, it is the administrative capacity of the state to deliver effective and timely public financial planning that matters the most for public goods and services delivery in the developing world. We address these two limitations here by examining the determinants of the state’s administrative capacity in developing countries, focusing in particular on political economy factors.

We next discuss how we measure administrative capacity with respect to public financial management in developing countries, using a novel database that is the first of its kind systematically to provide comparable measures of public financial management quality in developing countries.

3 Measuring administrative capacity

The survey in the previous section noted that, in the literature on state capacity, states’ ability to develop efficient financial planning has not received adequate scrutiny. Similarly, it showed that developing this type of capacity might rest on political institutions limiting the executive power. Here we use four indicators to capture the ability of states to develop high-quality financial planning, selected from the PEFA database, a unique source providing a granular view of public financial management performance in developing countries. The PEFA Program was founded in 2001 as a partnership between seven donor agencies and international financial institutions to assess the condition of countries’ public expenditure, procurement and financial accountability systems and develop a practical sequence for reform and capacity-building actions.Footnote 7

Using the selected indicators, we aim to give a more fine-grained picture of some activities performed by the bureaucracy, namely those related to the design, implementation and monitoring of the budget, compared to the assessment provided by generic indicators of bureaucratic quality. We see effective budgeting as a means to provide timely planning and to assess outlays after they have been disbursed with respect to revenues and expenditure, as well as debt management. Should expenditure not be in line with the budget, the competent bureaucracy would be able to reassess the process and to improve its behaviour in the future. Should the government try to interfere with the bureaucracy in the timing and allocation of resources, the independent civil servants would prevent them from extracting private or political benefits. The four indicators are described below (and their complete definition and coding are given in Appendix 1):

  1. 1.

    Aggregate expenditure out-turn compared with original approved budget. This captures the ability to implement budgeted expenditure by assessing the difference between actual primary expenditure and the originally budgeted primary expenditure.

  2. 2.

    Composition of expenditure out-turn compared with original approved budget. This variable assesses the extent to which the composition of expenditure varies from the original budget over the past 3 years, and so expresses whether the budget is a useful statement of policy intent.

  3. 3.

    Aggregate revenue out-turn compared with original approved budget. This captures the ability to produce accurate domestic revenue forecasts in the preparation of the budget by assessing the difference between actual revenues and those in the originally approved budget.

  4. 4.

    Recording and management of cash balances, debt and guarantees. This variable assesses the quality of debt management, by looking at the maintenance of a debt data system and regular reporting on the debt portfolio.

Table 1 presents the descriptive statistics for each variable, as well as a simple composite indicator created by taking their average.

Table 1 PEFA measures of administrative capacity

As these measures are based on surveys undertaken by public financial management experts and are subject to careful cross-checking and internal and external validity tests by PEFA, they provide an accurate picture of the quality of public financial management in the surveyed countries.Footnote 8 Further, the fact that the PEFA database provides scores—rather than just a summary of expert opinion of the country’s quality of public financial management—makes the data amenable for quantitative analysis. We next describe the methodology and data that inform the econometric analysis, as well as the results.

4 Methodology, data and results

This section first discusses our empirical strategy and the key variables. Then, we illustrate the results, accounting for potential endogeneity concerns.

4.1 Methodology and data

Since we look at the structural conditions under which countries develop capable states, regressions based on cross-sectional averages are a suitable approach, as they test relationships whose mechanisms have long-run characteristics.Footnote 9 The regression specification takes the form:

$$ {\text{AC}}_{i} = \, \beta_{0} + \beta W_{i}^{{}} + X_{i}^{'} \phi + \varepsilon_{i} $$

where ACi captures the quality of current administrative capacity as the average of the available observations of each PEFA indicator for country i between the end of the sample period, T = 2013, and T − 1 = 2005.

On the right-hand side, Wi is the determinant of interest, averaged between times t and t − 1, with t < T − 1, while β represents its long-run effect. It is measured as the average value of Constraints on the executive from the Polity IV dataset from 1965 (or independence year, if later) up to 2004 (Marshall et al., 2011), as we are interested in the long-run component of these constraints (and not in the annual fluctuations). This variable measures the extent of constitutional limits on the exercise of arbitrary power by the executive (on a scale from 1 to 7, where 1 indicates unlimited authority of the chief executive and 7 indicates executive parity or subordination, with intermediate values indicating moderate to substantial power limitations). Xi is a set of controls including Incidence of external conflict, Incidence of civil war, Total natural resource rents, Urban population (share) and Length of statehood.Footnote 10Appendix 2 reports definitions and sources. Finally, εi is the error term, capturing all other omitted factors. Appendix 3 provides the list of countries included in the analysis.

Figure 1 presents preliminary evidence, suggesting that there is a positive correlation between Constraints on the executive and each of the PEFA measures. However, while useful in illustrating the behaviour of key variables, one should not be tempted to place any causal interpretation on such correlations yet.Footnote 11 Before estimating (1), we should discuss whether estimating the impact of the political economy hypothesis is subject to identification problems. Although there are good reasons to expect a causal relationship between rulers’ accountability and high administrative capacity, Ordinary Least Squares (OLS) estimates are insufficient to document such a relationship. Building a political system is clearly an endogenous process, driven by a variety of social forces, including state actors. When estimating the relationship from the data, the effect of constraints on the executive could then be affected by reverse causality and hence subject to bias. Another concern is that the effect of political systems may be endogenous also in the statistical sense, namely correlated with the regression disturbances because of measurement error. Therefore, one might expect the coefficients on Constraints on the executive to be biased both away from zero and towards zero. The magnitude of the two types of bias, and their combined effect, is an open question, but here we attempt to address the problem using an instrumental variable (IV) approach.

Fig. 1
figure 1figure 1figure 1

Source: PEFA (2006) and Polity IV (Marshall et al. 2011)

a Aggregate expenditure out-turn compared with original approved budget and Constraints on the executive. b Composition of expenditure out-turn compared with original approved budget and Constraints on the executive. c Aggregate revenue out-turn compared with original approved budget and Constraints on the executive. d Recording and management of cash balances, debt and guarantees and Constraints on the executive. e Composite indicator of administrative capacity and Constraints on the executive

Our instrument has a prominent place in the literature: historical settler mortality, as captured by the (log of) mortality rate resulting from the disease environment at the time of colonisation. Acemoglu et al. (2001) documented the fact that such a variable picks up the exogenous variation in the type of institutions built in the former European colonies. Where Europeans settled en masse, life was organised around inclusive institutions, i.e. subjecting the ruling elite to binding limitations on their power. Where they could not settle, because of adverse sanitary conditions, institutions were extractive, i.e. subject to little or no constraints on the rulers. Compared with its alternatives, this instrument had a more plausible justification (see Acemoglu 2005b). Perhaps for this reason, it has proved to be resilient to criticism, which came on the grounds of data quality and associated historical records (Albouy, 2012). Since it was proposed, it has been successfully exploited to identify the effect of the Constraints on the executive variable (Acemoglu et al. 2001, 2003).Footnote 12 While we rely on Acemoglu et al.’s (2001) natural experiment argument that Settler mortality does not directly affect level of fiscal capacity (other than through its effect on Constraints on the executive), we also address exclusion restriction concerns through econometric testing.

4.2 Results

To account for potential instrument weakness, we estimate (1) by Limited Information Maximum Likelihood (LIML), using Fuller’s version (Fuller 1977; Baum et al. 2007).Footnote 13 This is more robust than Two-Stage Least Squares (2SLS) in the presence of weak instruments, as shown in the simulations carried out in Hahn et al. (2004), and appears to have lower small-sample variability. We set the user-specified constant (denoted by alpha in Fuller (1977)) to a value of four. While the Fuller 1 version yields the most unbiased estimator, the Fuller 4 version minimises the mean squared error of the estimator (Fuller 1977).

Table 2 regressions show that Constraints on the executive has a significant and consistently positive effect on administrative capacity, for three of its measures. This result holds also when using our simple composite indicator, averaging the four PEFA measures. Speculating on this, it may reflect possible complementarities existing among the different dimensions of administrative capacity, where investing in one of its dimensions may reinforce the others (see Besley and Persson 2011). The magnitude of the effect of IV estimates is comparable to OLS estimates, suggesting that perhaps the bias introduced by reverse causality and measurement error is such that they offset each other. However, constraints on the executive are irrelevant in IV estimates when it comes to predicting the level of Aggregate expenditure out-turn compared with original approved budget.

Table 2 Administrative capacity and constraints on the executive—OLS and IV estimates

How much does Constraints on the executive matter? One standard deviation increase (roughly 1.5 points) increases by over half standard deviations three of the four measures of administrative capacity (Table 3). Considering that in about 20% of developing economies Constraints on the executive is above one standard deviation, its effect seems economically meaningful, as well as statistically significant.

Table 3 Magnitude of effect of a change in Constraints on the executive

The key results are robust to checks for omitted variables. The literature on state capacity has proposed plausible alternatives (not exclusive) to the political institutions hypothesis. Some are historical in nature, i.e. experience of statehood and the incidence of external and internal conflicts. Others are geographical, i.e. the reliance of the economy as regards natural resource rents and population density. Following Besley and Persson (2009, 2011), we use the proportion of years at war from independence up to 2000 and the proportion of years in civil war over 1950–2000 to capture the incidence of external and internal conflict, respectively. Length of statehood is captured by the State antiquity index, proposed by Bockstette et al. (2002) and based on the intuition that longer histories of statehood lead to higher-quality administration thanks to ‘learning by doing’ effects. Introducing such variables leaves the significance of Constraints on the executive unchanged.

Geography-based robustness checks are particularly important, as the settler mortality rate could be proxying for ‘resource curse’ mechanisms or population density. For example, disease conditions may well be a determinant of where urban areas arise. So we can examine whether the Constraints on the executive results survive when we independently control for geographical variables. To capture such an effect, we use the share of urban population from World Bank (2013). And to capture ‘resource curse’ mechanisms, we use the 1970–2004 average share of GDP accruing from total resource rents (as the sum of oil, natural gas, coal, mineral and forest rents), from World Bank (2013). Introducing such controls does not greatly affect the significance and magnitude of the coefficient of interest. We experiment also with a number of other controls, including political democracy, legal origins, aid dependency, fractionalisation and regional dummies. Introducing such variables in our regressions does not significantly alter our findings (results are available on request).

Finally, we test whether the instrument meets the exclusion restriction by running a test of over-identification. Apart from a priori intuition, this is the other way of supporting the exclusion restriction. This approach is useful, since it is a direct test of our exclusion restriction. However, it is only partially satisfactory as such tests may have weak power (it may not lead to a rejection of the exclusion restriction if all instruments are invalid but still highly correlated with each other). Hence, the responses from these tests are not definitive, but could nonetheless give us additional confidence that settler mortality is a valid instrument. With this caveat in mind, and following Acemoglu et al. (2001), we choose distance from the equator as an additional instrument for Constraints on the executive. For such a variable to be valid here, its effects must also go through political institutions rather than through any other mechanism. This is potentially problematic, as it is not based on a natural experiment, but it is consistent with most arguments in the literature emphasising that geography affects development outcomes through political institutions, rather than directly (see Acemoglu 2005b).

The results of the over-identification tests are reported in Table 4. For each dependent variable, we rerun the third regression in Table 2 using both latitude and mortality rates as instruments. The first encouraging piece of evidence is that the new estimated coefficients are always quite close to those reported in Table 2. In addition, the results on the over-identification tests fail to reject the exclusion restriction at the conventional levels in all cases, and by a large margin in four out of five regressions. Hence, this exercise provides no evidence that sanitary conditions, as captured by Settler mortality, affect administrative capacity by any other channel than through political institutions.

Table 4 Accounting for instrument validity—over-identification tests with distance from the equator as an additional IV

5 Conclusions

Whatever benefit–cost case economists can make for investing in broad-based spending programmes—like infrastructure, health and education—in many low-income countries, there is little problem identifying the need for such public programmes. The problem comes in delivering them. This requires strengthening states’ ability to develop effective and timely public financial planning, to achieve more efficient provision of public goods and services. Starting from this premise, we have argued that placing checks and balances on the executive should provide a stronger basis for developing effective budgetary systems, as such political institutions may be more likely to develop bureaucracies on a meritocratic (rather than a patronage) basis and may be more effective (through an independent judiciary) at tackling corruption.

The ensuing empirical evidence, based on an IV approach to account for endogeneity, indicates that political institutions limiting the executive power tend to improve the ability of states in developing economies to design, implement and monitor their budget. Our findings have significant policy implications, as they suggest that the fundamental cause of weak public financial management institutions lies in the absence of political institutions promoting common interests. While much donor support for more efficient public financial management institutions tends to focus on technocratic solutions, such as reforming recruitment and promotion practices in the public sector, our paper suggests that an important component of donor support for Southern governments should also be building political institutions that place an effective system of checks and balances on executive power.