Introduction
The idea that meritocratic recruitment (hereafter MR) in public administration is a quintessential feature of governments that perform best on government outputs and broader downstream outcomes is firmly established in the literature (for a recent review, see Oliveira et al. (Reference Oliveira, Abner, Lee, Suzuki, Hur and Perry2023)). Nevertheless, a lack of clarity persists concerning the mechanisms through which MR contributes to the realisation of such results. The extant literature attributes a wide range of effects to MR, including, but not limited to, heightened competence (Aneja and Xu, Reference Aneja and Xu2024; Boräng et al., Reference Boräng, Cornell, Grimes and Schuster2018; Cingolani, Reference Cingolani2023; Colonnelli et al., Reference Colonnelli, Prem and Teso2020; Cornell et al., Reference Cornell, Knutsen and Teorell2020; Evans and Rauch, Reference Evans and Rauch1999; Krause et al., Reference Krause, Lewis and Douglas2006; Lewis, Reference Lewis2007; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016; Rauch and Evans, Reference Rauch and Evans2000; Yu and Jennings, Reference Yu and Jennings2021), improved internal cohesion (Dahlström et al., Reference Dahlström, Lapuente and Teorell2012; Evans and Rauch, Reference Evans and Rauch1999; Henderson et al., Reference Henderson, Hulme, Jalilian and Phillips2007; Oliveros and Schuster, Reference Oliveros and Schuster2018; Rauch and Evans, Reference Rauch and Evans2000), the codification of behavioural norms consistent either with the efficiency metric (Evans and Rauch, Reference Evans and Rauch1999; Rauch and Evans, Reference Rauch and Evans2000) or professional standards and ethos (Boräng et al., Reference Boräng, Cornell, Grimes and Schuster2018; Carpenter, Reference Carpenter2002; Dahlström et al., Reference Dahlström, Lapuente and Teorell2012; Henderson et al., Reference Henderson, Hulme, Jalilian and Phillips2007; Miller, Reference Miller2000; Miller and Whitford, Reference Miller and Whitford2016), longer-term horizons (Lapuente and Nistotskaya, Reference Lapuente and Nistotskaya2009; Lewis, Reference Lewis2007; Rauch, Reference Rauch1995), improved predictability (Cornell et al., Reference Cornell, Knutsen and Teorell2020), and higher resistance to undue political influence (Boräng et al., Reference Boräng, Cornell, Grimes and Schuster2018; Cornell et al., Reference Cornell, Knutsen and Teorell2020; Dahlström and Lapuente, Reference Dahlström and Lapuente2017; Miller, Reference Miller2000; Miller and Whitford, Reference Miller and Whitford2016; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016). Although the existing literature acknowledges the possibility of several concurrent causal mechanisms (Boräng et al., Reference Boräng, Cornell, Grimes and Schuster2018; Cornell et al., Reference Cornell, Knutsen and Teorell2020; Evans and Rauch, Reference Evans and Rauch1999; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016), significant gaps remain in specifying the relationship between the independent variable (X) and the dependent variable (Y), as mediated by the intervening variable (M). First, there is a notable absence of a theoretically grounded and coherent rationale for why the X→M→Y causal path should exist, along with a detailed account of how it operates. Second, empirical evidence demonstrating associations between X and M, M and Y, or all three is lacking.
Understanding the underlying mechanisms is important for at least two reasons. First, it improves the clarity of knowledge about empirical phenomena, and second, it provides insights for informing policy and practice. Given the inherent complexity of social phenomena, grasping the underlying processes through which a cause leads to a specific effect is essential for unravelling these complexities. Mechanisms play ‘a crucial role in distinguishing true causal relations from spurious correlations’, especially in non-experimental settings, and are vital for the ‘extrapolation of causal findings from one setting to another’ (Hedström and Ylikoski, Reference Hedström and Ylikoski2010, 54). Considering the practical application of knowledge regarding mechanisms, the case of MR carries significant consequences for the organisational design of public administration.
This paper addresses these gaps in the literature by examining two mechanisms through which MR may affect developmental outcomes: competence and impartiality. Theoretically, we argue that both competence and impartiality independently influence developmental outcomes, given their roles in two distinct macro theories of development. Competence is integral to state capacity (SC) accounts of development, which underscore the need for a strong state capable of effectively implementing political decisions (Acemoglu et al., Reference Acemoglu, Moscona and Robinson2016; Andrews et al., Reference Andrews, Pritchett and Woolcock2017; D’Arcy and Nistotskaya, Reference D’Arcy, Nistotskaya, Bågenholm, Bauhr, Grimes and Rothstein2021). MR enhances the epistemic qualities of public administration, leading to improved developmental outcomes by bettering government outputs, such as regulatory quality. At the same time, impartiality is associated with theoretical perspectives that stress the importance of constraining power-holders from abusing the power and resources of the state. These perspectives include the credible commitment (CC) literature (Miller, Reference Miller2000; North and Weingast, Reference North and Weingast1989) and the Quality of Government (QoG) scholarship (Rothstein and Teorell, Reference Rothstein and Teorell2008), both of which suggest that meritocratically recruited bureaucrats are better positioned to resist undue political interference, thereby safeguarding the public interest in governmental processes and ultimately contributing to improved developmental outcomes. We propose that integrating these two markedly different meta-theories of development provides a coherent theoretical foundation that links MR to developmental outcomes through its dual impact on competence and impartiality.
Empirically, our study examines the relationship between MR (X), bureaucratic competence (M 1), bureaucratic impartiality (M 2), and our selected developmental outcome (Y). Specifically, we scrutinise the mediating pathways of competence and impartiality in the relationship between MR and entrepreneurship rates. Our analysis reveals a positive association between the institutionalisation of MR and the density of new businesses. Crucially, we find that both competence and impartiality act as pathways in this relationship. The results of the primary analysis, based on fourteen years of cross-national data, remain robust across various model specifications. These findings remain substantively the same when alternative measures of key concepts, available for only two-time points, are employed.
To our knowledge, this paper is the first explicit attempt to provide a mechanism-based explanation for the welfare-enhancing effects of MR in public administration. We undertake all three elements of mechanism-based explanations (Hedström and Ylikoski, Reference Hedström and Ylikoski2010): provide a theoretical rationale for why the X→M→Y path should exist, describe the ‘cogs and wheels’ of the causal processes that produce the effect of interest, and present empirical evidence of the associations between X and M, M and Y, and all three. By elucidating the distinct rationales underlying the competence and impartiality effects of MR, and by empirically demonstrating their simultaneous occurrence, our study deepens the understanding of the complexities inherent in recruitment practices in public administration and their implications for government quality and societal welfare.
This new knowledge may stimulate renewed interest in recruitment practices within public administration, in the contexts of both domestic reforms and international development cooperation. If MR strengthens SC and constrains political elites from abusing state power and resources, organisational reforms of government bureaucracies could be a key means to advance human development. This is particularly pressing given recent efforts by some political leaders to repoliticize bureaucracy (Bauer et al., Reference Bauer, Peters, Pierre, Yesilkagit and Becker2021), with President Trump’s ongoing attempt to upend the U.S. federal bureaucracy standing as a striking example (Moynihan, Reference Moynihan, Bouckaert, Hondeghem, Steen and Van de Walle2025).
Literature review
Public administration, or public bureaucracy, is an essential element of modern systems of governance.Footnote 1 The organisational design of bureaucracy has profound consequences for a large range of socio-economic and political outcomes, making it an important object of scientific inquiry. A large literature recognises the value of MR in bureaucracies for a host of development outcomes by drawing on Weber’s conception of bureaucracy as a fundamental part of legal-rational authority, which is more efficient in comparison to other types of authority (Weber, Reference Weber2019). Weber’s ideal type of modern bureaucracy is that of a hierarchical, specialised, meritocratically recruited, protected from politically-motivated dismissals, salaried and rule-bound organisation.
While Weberian bureaucracy is a multifaceted concept, merit-based recruitment – a method of personnel selection in public administration that emphasises expertise, including formal qualifications and relevant experience pertinent to the responsibilities of a position, alongside procedural safeguards against political interference in appointments – stands out as its hallmark feature. Numerous empirical studies operationalise Weberianism primarily through the extent of the institutionalisation of MR (Colonnelli et al., Reference Colonnelli, Prem and Teso2020; Lapuente and Nistotskaya, Reference Lapuente and Nistotskaya2009; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016; Oliveros and Schuster, Reference Oliveros and Schuster2018). The interchangeable use of the terms ‘Weberian’ and ‘meritocratic’ is common, thereby obscuring the understanding of whether it is MR or other features of Weberian bureaucracy that are driving the observed effects. To improve clarity, this paper narrows its focus to the defining attribute of Weberian bureaucracy: merit-based selection in public administration.Footnote 2
Despite extensive evidence suggesting that Weberian/meritocratic bureaucracy contributes to improved government outputs (Aneja and Xu, Reference Aneja and Xu2024; Boräng et al., Reference Boräng, Cornell, Grimes and Schuster2018; Charron et al., Reference Charron, Dahlström, Fazekas and Lapuente2017; Cingolani, Reference Cingolani2023; Dahlström et al., Reference Dahlström, Nistotskaya and Tyrberg2018; Krause et al., Reference Krause, Lewis and Douglas2006; Lewis, Reference Lewis2007) and broader downstream outcomes (Cingolani et al., Reference Cingolani, Thomsson and de Crombrugghe2015; Cornell et al., Reference Cornell, Knutsen and Teorell2020; Evans and Rauch, Reference Evans and Rauch1999; Henderson et al., Reference Henderson, Hulme, Jalilian and Phillips2007; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016; Patterson, Reference Patterson2024), the literature remains underdeveloped when it comes to theoretically elucidating and empirically testing the mechanisms through which these effects are realised. For example, in their seminal study, Evans and Rauch (Reference Evans and Rauch1999) argued that the positive impact of Weberian bureaucracy on economic growth operates through a ‘myriad [of] specific causal paths’ (p. 752). However, the study neither elaborates on these causal paths nor empirically tests them. Moreover, it attributes the positive effects to Weberian bureaucracy as a whole, without considering the contributions of its individual components. Similarly, Henderson et al. (Reference Henderson, Hulme, Jalilian and Phillips2007) posited that ‘the Weberianess of the state bureaucracy’ is likely to reduce poverty because it is more effective in designing and implementing government policies aimed at poverty alleviation (p. 520). A more recent study by Cornell et al. (Reference Cornell, Knutsen and Teorell2020) also employs a broad concept of Weberain bureaucracy and links it to economic growth through rule-following and impartial decision-making, increased competence, and the ability to curb ‘arbitrary interferences by political leaders’. However, neither of these papers delves into the cogs and wheels of the causal processes through which poverty reduction or economic growth is achieved, nor do they empirically test the proposed mechanisms.
The literature specifically focusing on MR in public administration does not offer any clearer understanding of the mechanisms at play. For example, both Rauch (Reference Rauch1995) and Cingolani (Reference Cingolani2023) examine the impact of MR on infrastructural development, but attribute its effects to different mechanisms: extended time horizons in Rauch (Reference Rauch1995) and enhanced bureaucratic professionalism in Cingolani (Reference Cingolani2023). Neither claim is accompanied by a detailed theoretical exploration of the relevant mechanism, nor are these mechanisms empirically tested.
Several studies argue that MR enhances personnel competence, whereas its opposite – ‘at will’ systems – undermine it (Boräng et al., Reference Boräng, Cornell, Grimes and Schuster2018; Colonnelli et al., Reference Colonnelli, Prem and Teso2020; Krause et al., Reference Krause, Lewis and Douglas2006; Lapuente and Suzuki, Reference Lapuente and Suzuki2020; Lewis, Reference Lewis2007; Yu and Jennings, Reference Yu and Jennings2021). Another strand of literature suggests that the merit of MR lies in its ability to reduce undue political interference on day-to-day bureaucratic functioning, thereby ensuring a more consistent application of norms (Dahlström et al., Reference Dahlström, Fazekas and Lewis2021; Dahlström and Lapuente, Reference Dahlström and Lapuente2017; Miller, Reference Miller2000; Oliveros, Reference Oliveros2016; Oliveros and Schuster, Reference Oliveros and Schuster2018). Yet, some scholars contend that both of these mechanisms operate simultaneously and independently (Boräng et al., Reference Boräng, Cornell, Grimes and Schuster2018; Cornell et al., Reference Cornell, Knutsen and Teorell2020; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016). However, the analytical narrative and empirical evidence supporting this view remain underdeveloped.
In summary, despite the growing interest in Weberian bureaucracy in general and in MR in particular, the existing literature has yet to adequately examine the mechanisms through which MR may produce welfare-enhancing effects. The mechanisms are more often assumed than rigorously theorised or empirically tested. Existing analytical narratives frequently lack grounding in a coherent, high-level theoretical framework and also a nuanced explanation of how X leads M and then to Y. Furthermore, the potential for multiple mechanisms is often overlooked, and even when acknowledged, their interrelationships and connections to existing theories of development are seldom explored. Finally, the literature has yet to empirically validate the suggested mechanisms through formal mediation analysis.
To advance existing knowledge, this paper focuses on one of the defining attributes of Weberian bureaucracy – merit-based selection in public administration – and systematically examines its two underlying mechanisms: competence and impartiality, which, we argue, are central to understanding its effects. Establishing a mechanism of change requires a plausible rationale for why the causal path should exist, a clear articulation of the underlying processes that bring about a meaningful difference to the outcome of interest (Hedström and Ylikoski, Reference Hedström and Ylikoski2010), and empirical evidence demonstrating the existence of the associations between X, M, and Y. In Section 3, we discuss the broader theoretical foundations of both mechanisms and detail how the relevant attributes of MR lead to the change in the outcome of interest. In Section 4, we empirically test the mechanisms, utilising a structural equation approach (SEM) for formal mediation analysis with multiple mediators (Coutts and Hayes, Reference Coutts and Hayes2022), drawing on fourteen years of cross-country data.
Argument
Consistent with previous studies, we argue that MR contributes to improved human development outcomes through two distinct mechanisms: first, by bolstering the competence of administrative personnel, and second, by promoting impartiality in bureaucratic behaviour. However, in contrast to prior research that has offered a broad range of rationales for the existence of these mechanisms without a clear and focused theoretical foundation, we posit that these mechanisms are grounded in two distinct meta-theories of development: SC and CC. These frameworks differ markedly in their perspectives on the state’s role in development. While SC emphasises the state’s capacity as a facilitator of development (Acemoglu et al., Reference Acemoglu, Moscona and Robinson2016), CC scholars contend that the state’s power and resources inherently carry a risk of abuse, thereby hindering sustained development (North, Reference North1981). Consequently, CC scholars focus on mechanisms to credibly commit power holders to refrain from such abuse, ensuring that state authority is exercised in ways that support development.
Viewed from this perspective, MR emerges as a bridge between these two markedly different meta-theories. It not only provides high-quality administrative personnel, a crucial element of SC, but also serves as an essential component of the institutional architecture that limits the abuse of state power, resonating with the CC argument. In other words, MR integrates these distinct theories by addressing the core development challenges each identifies, thereby merging SC and CC into a cohesive explanation of the determinants of development. Having established this integrated theoretical foundation for why these specific causal paths should exist, we now turn to detail how competence and impartiality contribute to developmental outcomes.Footnote 3
Competence
The prevailing view in the literature is that MR renders bureaucracy a more sophisticated administrative tool compared to alternative selection methods. The intellectual roots of this line of reasoning go back to Weber (Reference Weber2019), who argued that merit bureaucracy is a bureaucracy of higher technical quality (compared to alternatives such as patronage) and is, therefore, a more able implementation tool.
Several strands of literature underscore the significance of a competent public bureaucracy in shaping both government outputs and broader downstream outcomes. Empirical research consistently shows that a more competent bureaucracy is associated with improved government outputs, such as policy stability (Schnose, Reference Schnose2017), higher quality of public services (Dahlström et al., Reference Dahlström, Nistotskaya and Tyrberg2018), and more efficient procurement (Decarolis et al., Reference Decarolis, Giuffrida, Iossa, Mollisi and Spagnolo2020). Moreover, the developmental state literature has long recognised bureaucratic competence as a key determinant of positive downstream outcomes. For example, Johnson (Reference Johnson1982) showed that the analytical expertise of personnel at Japan’s Ministry of International Trade and Industry was instrumental to the country’s post-WWII economic revival. Similarly, variations in bureaucratic competence have been decisive in distinguishing the success or failure of state-led development initiatives in countries such as Brazil and Nigeria (Kohli, Reference Kohli2004).
Given the established link between bureaucratic competence and positive results, the question arises: how is such competence developed? There is empirical evidence suggesting that meritocratic bureaucracies underpin both competence-laden government outputs (Dahlström et al., Reference Dahlström, Fazekas and Lewis2021; Krause et al., Reference Krause, Lewis and Douglas2006; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016) and key developmental outcomes (Cingolani et al., Reference Cingolani, Thomsson and de Crombrugghe2015; Cornell et al., Reference Cornell, Knutsen and Teorell2020; Evans and Rauch, Reference Evans and Rauch1999; Patterson, Reference Patterson2024; Rauch, Reference Rauch1995). In this literature, bureaucratic competence is often viewed as the putative causal mechanism linking meritocratic bureaucracies to improved government outputs and developmental outcomes. However, this argument lacks robust theoretical grounding and has yet to be directly tested using measures that capture both MR and bureaucratic competence. Instead, existing studies often rely on broad proxies for these concepts and methods that assess associations rather than examining the underlying mechanisms. Moreover, the literature tends to treat meritocratic bureaucracy as a monolithic concept, failing to delineate its specific attributes or elucidate how these attributes contribute to bureaucratic competence.
To address this gap, we focus on one key attribute of Weberian bureaucracy: MR. We argue that MR (X) improves bureaucratic competence (M), which in turn leads to improved government outputs and developmental outcomes (Y). Three core elements of MR help ensure the selection of competent candidates.
The first is formal vacancy announcements that outline the tasks and responsibilities of the position and specify the qualifications and other attributes sought in a successful applicant. When a vacancy announcement is formal, information about the opportunity becomes accessible to a larger pool of potential applicants. Broadening the applicant pool increases the likelihood of finding the best-fitting candidate, thereby enhancing workforce competence (Keller, Reference Keller2018). This is true even in the context of internal hiring – such as in career-based public bureaucracies – as formal announcements reach employees across the organisation, including those in different departments or geographic locations, thereby expanding the pool of potential candidates (Keller, Reference Keller2018). Additionally, clear and detailed job descriptions, along with the specified attributes of a successful applicant, send a strong signal that candidates will be evaluated based on their ability to meet these criteria. This serves as a pull factor for highly competent candidates while discouraging less qualified ones (Mason and Belt, Reference Mason and Belt1986).
Second, vacancy announcements and other recruitment materials specify in advance the standards relevant to the job. For positions in public administration, this typically encompasses the necessary qualifications, analytical/cognitive and behavioural competences, skills, and specific experiences (OECD, 2021, 135–146). These criteria are objective and measurable, allowing candidates to be assessed based on observable evidence rather than subjective judgments. This objectivity and measurability foster consistent assessment across all applicants, thereby increasing the likelihood of identifying the most competent candidate for the position.
Third, bureaucratic competence is strengthened when hiring decisions are made collectively by a group rather than by a single individual. A hiring panel consisting of several members helps to mitigate individual biases, thereby increasing the likelihood of selecting the best-fit and most competent candidates. Together, these elements of MR foster competence within public administration.Footnote 4
Impartiality
Another pathway through which MR influences development outcomes is by fostering bureaucratic commitment to, and the impartial execution of, agreed-upon policies. This commitment is vulnerable to several risk factors, including illicit pressures from politicians, influence by societal and interest groups, and bureaucrats’ own self-serving motivations. Among these, illicit pressure from individual politicians is widely regarded as the most significant risk. The intellectual roots of this argument go back to Douglass North, who identified the main threat to development as emanating from those at the apex of the state (North, Reference North1981; North and Weingast, Reference North and Weingast1989). When left unconstrained, political elites may exploit state power and resources to serve narrow, short-term, or self-serving objectives. This raises critical questions: How can political elites be restrained from pursuing morally hazardous incentives? How can they be credibly committed to serving the public good?
The literature identifies several institutional solutions to the problem of CC, one of which involves limiting politicians’ powers over making and breaking bureaucratic careers through practices such as merit-based recruitment, protection from politically motivated dismissals, deferred compensation, and other measures (Dahlström and Lapuente, Reference Dahlström and Lapuente2017; Miller, Reference Miller2000; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016; Rothstein and Teorell, Reference Rothstein and Teorell2008).Footnote 5 MR introduces a degree of misalignment in incentives between politicians and bureaucrats, reducing bureaucratic commitment to advancing the agendas of individual politicians. Instead, it fosters a stronger commitment to, and the impartial execution of, agreed-upon policies.Footnote 6
To understand how MR fosters incentive misalignment and impartiality, it is useful to first examine the ‘at-will’ recruitment model, where political actors have considerable discretion in hiring and firing bureaucrats. Research suggests that such a recruitment modus operandi establishes a highly personalised and affective relationship between the appointee and the politician with appointing powers from the very beginning of the appointee’s career (Bearfield, Reference Bearfield2009; Lemarchand and Legg, Reference Lemarchand and Legg1972). The politician’s control over employment – a valuable resource as it confers substantive benefits upon the appointee – creates a pronounced imbalance in status and obligations between the two parties (Lemarchand and Legg, Reference Lemarchand and Legg1972, 152). This dynamic compels appointees to prioritise the interests and demands of their political patrons, including those that may be illegitimate or outright illicit. In essence, the deeply personalistic, hierarchical, and reciprocal nature of politician-bureaucrat relations under the ‘at-will’ model heightens bureaucrats’ sensitivity to the preferences – whether legitimate or otherwise – of the individual politicians who sponsor their jobs. These dynamics generate substantial pressures to conform to the expectations of political sponsors, leading bureaucrats to act in ways that align with the particularistic preferences of individual politicians rather than the broader, welfare-enhancing goals of public policies.
Furthermore, politicians’ discretionary appointment powers are typically accompanied by corresponding dismissal powers. The 2021 global expert survey of bureaucratic structures showed that countries scoring low on MR also tend to score low on tenure protection (Nistotskaya et al., Reference Nistotskaya, Dahlberg, Dahlström, Sundström, Axelsson, Dalli and Alvarado Pachon2021, 13),Footnote 7 highlighting a strong association between these two factors. In labor economics, it is well-established that the threat of dismissal fosters high responsiveness to employers’ preferences (Shapiro and Stiglitz, Reference Shapiro and Stiglitz1984). Similarly, in public administrations with weak tenure protection, the credible threat of dismissal has pressured bureaucrats to yield to political interference (Leucht, Reference Leucht2023; Miller, Reference Miller2000; Oliveros, Reference Oliveros2016; Van Riper, Reference Van Riper1958). Consequently, ‘at will’ recruitment, coupled with weak tenure protection, creates strong incentives for bureaucrats to be acutely attuned to the preferences of the political sponsors of their jobs and a willingness to act on those preferences in day-to-day bureaucratic operations.
In contrast, merit-based entry to public administration prevents the formation of the lopsided, personalised relationships characteristic of the ‘at-will’ model, which induce bureaucrats to prioritise the agendas of individual politicians. MR insulates bureaucrats from the sway of individual politicians, as no single politician can claim credit for their hiring or credibly threaten them with dismissal. Furthermore, robust tenure protection integral to MR allows bureaucrats to operate without the constant fear of losing their positions, substantially reducing their incentives to comply with informal or lateral demands from individual politicians. In summary, by introducing a degree of incentive misalignment, MR makes bureaucrats less responsive to undue political influence, mitigating the risk of political moral hazard infiltrating the state’s machinery of public goods production.Footnote 8
Building on the QoG literature (Dahlström and Lapuente, Reference Dahlström and Lapuente2017; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016; Rothstein and Teorell, Reference Rothstein and Teorell2008), we argue that bureaucratic impartiality is a robust – and also observable – indicator of the incentive misalignment-induced by MR. According to Rothstein and Teorell (Reference Rothstein and Teorell2008), impartiality entails that bureaucrats base their actions solely on what is explicitly stipulated in policies and laws, free from external influences as well as their own preferences and biases. The authors further link impartiality to MR, asserting that if recruitment to public administration were to be based on ‘clientelistic personal contacts, political leanings, bribes, or ethnic belongings, the impartiality principle would be violated’ (Rothstein and Teorell, Reference Rothstein and Teorell2008, 170). By eliminating these grounds for recruitment, MR reduces bureaucrats’ responsiveness to the preferences of influential figures within clientelistic, political, or ethnic networks, thereby introducing the incentive misalignment central to the impartiality mechanism.
In other words, MR (X) promotes greater bureaucratic impartiality (M), which, in turn, leads to improved government outputs and developmental outcomes (Y). To illustrate, consider the case of entrepreneurship (for a comprehensive discussion of why entrepreneurship is a suitable example of developmental outcomes, please refer to Appendix A). Bureaucratic impartiality stands as a cornerstone of what Acemoglu and Robinson (Reference Acemoglu and Robinson2012) define as growth-enabling institutions – institutions that ‘allow and encourage participation by the great mass of people in economic activities’ and ‘provide a level playing field that gives them the opportunity to do so. Those who have good ideas will be able to start business’ (p. 74, 76). Conversely, when partiality is the modus operandi of public administration, certain individuals enjoy preferential treatment over others in their interactions with the government. A vivid example of this is amiguismo (friendism) in Spain, a system where business success depends on who one knows in government (Sanchez, Reference Sanchez2024). In such environments, partiality acts as a barrier to the broad engagement in economic activities envisioned by Acemoglu and Robinson (Reference Acemoglu and Robinson2012). Individuals who experienced or even merely witness administrative partiality, such as biased government procurement (Dahlström et al., Reference Dahlström, Fazekas and Lewis2021) are more likely to refrain from pursuing legitimate business ventures, opting instead to either abstain from economic activities entirely or operate within informal economies (Nistotskaya et al., Reference Nistotskaya, Charron and Lapuente2015). Furthermore, bureaucratic partiality introduces uncertainty into the business environment, thereby negatively impacting individuals’ calculations on the expected utility of business venturing. This discourages the initiation of new businesses, as the risk of bias and unfair treatment diminishes the potential rewards of entrepreneurship.
Figure 1 depicts the relationships between MR and development outcomes through improved bureaucratic competence and impartiality in bureaucratic behaviour. Beyond competence and impartiality, the model allows for other mechanisms.
Meritocratic recruitment and development: mechanisms.
In summary, we argue that meritocratic entry to bureaucracy is a source of both bureaucratic competence and impartiality and that these mechanisms are consistent with two major, but opposing in their primary assumption, theoretical frameworks of human development. The SC framework assumes that the state is a benign Janus, whose powers are consequential for public goods provision, but the CC framework sees the state as a malevolent Janus and seeks to minimise the risk of the use of state power for particularistic ends. MR emerges as a bridge between these two markedly different meta-theories, integrating them into a cohesive explanation of the determinants of development.
Testing the mechanisms empirically
Method
To estimate the effects of MR on the developmental outcome of interest as transmitted through competence and impartiality, we employ a SEM approach for mediation analysis with multiple mediators (Coutts and Hayes, Reference Coutts and Hayes2022). We estimate the following equations:
$${M_1} = {d_{{M_1}}} + {a_1}X + \varepsilon_{M_1}$$
$${M_2} = {d_{{M_2}}} + {a_2}X + \varepsilon_{M_2}$$
$$Y{\rm{ }} = {d_y} + c'X + {b_1}{M_1} + {b_2}{M_2} + {\varepsilon _Y}$$
In models featuring two mediators, the effect of X on Y that is transmitted through the first mediator (M 1), while holding the second mediator (M 2) constant, is referred to as a specific indirect effect. The effect of X on Y through all other potential mechanisms not identified in the model is referred to as the direct effect. Together, the indirect and direct effects constitute the total effect. Equations (1) and (2) represent the effect of MR on the first mediator (competence) and the second mediator (impartiality), respectively. Equation (3) models the effect of MR on the outcome of interest, controlling for influence of both mediators. In this system of equations, a 1 b 1 represents the specific indirect effect transmitted through the first mediator, controlling for the second mediator, while a 2 b 2 represents the specific indirect effect transmitted through the second mediator, controlling for the first. The direct effect is represented as c ′. The expectation that M 1 and M 2 exert independent effects on the outcome of interest should be corroborated by their statistically significant estimates in the full model. We expect both estimates to be positively signed but remain agnostic with regards to the relative magnitudes of their effects. To determine which mediator exerts a stronger influence, we compare the raw difference between the indirect effects, as recommended by Coutts and Hayes (Reference Coutts and Hayes2022) when these effects are expected to share the same sign.
Operationalisation and measurement of key concepts
To test our theoretical argument, it is essential to identify appropriate indicators for the key concepts, particularly of bureaucratic competence and developmental outcomes. Translating these complex and multifaceted concepts into variables involves challenges and trade-offs. A primary challenge lies in ensuring substantive alignment between the chosen developmental outcome and the measure of bureaucratic competence. For instance, while the quality of civil registration and vital statistics (CRVS) data may serve as a reasonable proxy for bureaucratic competence (Mikkelsen et al., Reference Mikkelsen, Phillips, AbouZahr, Setel, De Savigny, Lozano and Lopez2015), its relevance to developmental outcomes such as infrastructure development or entrepreneurial activity is limited. These outcomes often depend on other dimensions of competence – such as technical expertise in engineering or regulatory capacity to foster business-friendly environments – that are not captured by CRVS data.
Many existing measures of bureaucratic competence are not finely tuned to the outcomes they aim to explain (Dahlström et al., Reference Dahlström, Nistotskaya and Tyrberg2018), limiting their utility for examining mechanisms. It is, therefore, imperative to ensure a stronger logical alignment between the measures of bureaucratic competence and developmental outcomes. Another significant challenge is the literature’s reliance on two problematic types of measure: (1) self-reported measures and (2) highly indirect proxies of bureaucratic competence. Bureaucrats may exaggerate their skills or experiences, or the quality of work done by their work unit (Decarolis et al., Reference Decarolis, Giuffrida, Iossa, Mollisi and Spagnolo2020), particularly when evaluations are linked to performance reviews or resource allocation. Both types of measures introduce noise into the data, compromising the validity and reliability of competence indicators.
To address these challenges, we operationalise developmental outcomes through the rate of incorporation of new local small and medium-sized enterprises (SMEs) per adult population (see Appendix A). Bureaucratic competence is operationalised through the quality of business entry regulation. This pairing is both logically coherent and empirically supported: research consistently demonstrates the critical role of regulatory environments in encouraging or deterring new firm creation (see Appendix B). Furthermore, there is a logical alignment between bureaucratic impartiality and entrepreneurship rates, supported by studies from diverse contexts settings (Nistotskaya et al., Reference Nistotskaya, Charron and Lapuente2015; Nistotskaya and Cingolani, Reference Nistotskaya and Cingolani2016).
Entrepreneurship, as measured by the number of newly registered SMEs, is a suitable proxy for developmental outcomes for several reasons. First, entrepreneurial activity is a key driver of economic growth, with extensive evidence linking higher entrepreneurship rates to higher economic growth (for a list of references, see Appendix A). Economic growth, in turn, plays a crucial role in human development by providing resources to the state to improve living standards, access to services, and the overall quality of life of citizens. This relationship is exemplified by comparisons of public expenditure on health and education in rapidly growing versus stagnant economies (Ranis et al., Reference Ranis, Stewart and Ramirez2000). Additionally, country-level studies find a positive relationship between entrepreneurship and human development indicators, particularly in developing countries (Dhahri and Omri, Reference Dhahri and Omri2018).
Second, SMEs are major contributors to private-sector employment, with young firms disproportionately driving net job creation. Employment fosters income generation and enhances social mobility, thereby supporting broader human development outcomes. Moreover, SMEs promote social inclusion, with enterprises led by women, young people, and ethnic minorities playing a critical role in reducing social exclusion and generating income in marginalised groups (for a detailed discussion, see Appendix A).
We operationalise bureaucratic competence through the quality of business entry regulation (see Appendix B). Firm creation depends on transparent and effective regulatory frameworks, which are shaped by bureaucratic quality. The quality of the regulatory environment provides a theoretically grounded, empirically validated, observable, outcome-oriented gauge of bureaucratic competence. It encapsulates the technical expertise, intellectual rigour, and analytical competence of bureaucratic regulatory agencies required to balance procedural simplicity with the protection of public interests.
We acknowledge that the selected operationalizations may not fully capture the multidimensional nature of bureaucratic competence or development; however, they effectively reflect key dimensions of these concepts and are sufficiently aligned to enable the empirical examination of the proposed causal chain.
Data
To estimate the relationships of interest (Figure 1), we use country-level panel data comprising almost 1,500 country-year observations for 137 countries over the period 2006–2019.Footnote 9 Appendix D provides detailed descriptions of the variables and data sources, while Appendix E discusses the sample characteristics. Entrepreneurship is measured as the (log) number of new formally registered companies per 1,000 adult population, drawn from the World Bank. To measure MR, we use an expert-based continuous indicator from Version 11 of the V-Dem dataset (Coppedge et al., Reference Coppedge, Gerring, Knutsen, Lindberg, Teorell, Alizada, Altman, Bernhard, Cornell, Fish, Gastaldi, Gjerløw, Glynn, Hicken, Hindle, Ilchenko, Krusell, Luhrmann, Maerz, Marquardt, McMann, Mechkova, Medzihorsky, Paxton, Pemstein, Pernes, von Romer, Seim, Sigman, Skaaning, Staton, Sundstrom, Tzelgov, Wang, Wig, Wilson and Ziblat2021), capturing the extent to which appointment decisions in public administration are based on skills and qualifications rather than personal or political connections. For ease of interpretation, we normalised Merit, where ‘1’ indicates higher levels of meritocracy.
To gauge bureaucratic competence, we utilise the Starting a Business variable from the World Bank’s Ease of Doing Business Index, which captures the number of procedures, time, and costs required to register a local limited liability company. To measure bureaucratic impartiality, we employ a V-Dem variable that quantifies the degree of impartiality in bureaucratic behaviour. While this variable has a broad spatial and temporal coverage, we recognise its potential limitations: democracy experts providing the data might implicitly associate higher levels of impartiality with more democratic regimes. To address this concern, we conduct robustness checks using data from the QoG Expert Survey (Nistotskaya et al., Reference Nistotskaya, Dahlberg, Dahlström, Sundström, Axelsson, Dalli and Alvarado Pachon2021), which relies on assessments by experts in public administration. However, the QoG data is more limited in scope, providing only 163 observations across two-time points.
The selection of control variables includes economic development level (measured by GDPpc in the main analysis and the Economic Complexity Index in the robustness checks), recent economic dynamics (such as GDP growth and unemployment rates), broader institutional quality indicators (including the country’s democratic status and property rights security), population size (urban population in the robustness checks), and government size. Additionally, we account for social and human capital determinants of entrepreneurship, captured through ethnic fractionalisation and the share of the population with secondary education, respectively. Due to missing data, the latter is included only in robustness checks.
Results
Figure 2 presents the results of a structural equation modelling (SEM) mediation analysis with two mediators (full results are reported in Table F1, Appendix F). After controlling for potential confounders, the coefficient for MR is statistically significant and positively signed for both competence and impartiality. In turn, these two factors exhibit statistically significant positive effects on the outcome not interest, while the direct effect of MR on the outcome becomes statistically non-significant.
Effects of meritocratic recruitment on new business density mediated by bureaucratic competence and impartiality. Note: the estimates are derived from Models 2, 4, and 6 in Table F1 (Appendix F). The dashed line connecting the mediators represents our assumption that they may be intercorrelated; P < 0.10, ** P < 0.05, *** P < 0.01.
Table 1 breaks down the total effect of MR on new business density into direct and indirect effects, transmitted through competence and impartiality. Models 1 and 2 report the results using V-Dem data for MR and impartiality, where Model 1 includes MR and two mediators, and Model 2 introduces a full set of control variables. In both models, the estimates for the indirect effects of competence and impartiality, which are the primary focus of this study, are statistically significant (at least at the 95% confidence level) and positively signed. This indicates that higher values of MR are associated with increased levels of competence and impartiality, which, in turn, contribute to greater new business registrations.
Direct, indirect, and total effects of meritocratic recruitment on new business density
Note: Direct, indirect, and total effects, as well as raw differences, are calculated using R’s lavaan package based on estimates from Table F1, Appendix F. The indirect effect of competence is calculated as a 1*b 1, and the indirect effect of impartiality as a 2*b 2. P < 0.10, ** P < 0.05, *** P < 0.01.
In Models 1 and 2, the coefficient for the indirect effect of impartiality is larger than that of competence. A test of raw difference indicates that this difference is statistically significant only in Model 1. The direct effect of MR (i.e., effects transmitted through all other potential mediators) is statistically non-significant. One interpretation of this is that the impact of MR is primarily mediated through competence and impartiality. However, it is also possible that causal pathways not captured by the model may offset each other, leading to a non-significant direct effect. The total effect – representing MR’s influence on new business density through all possible mediators – is statistically significant in both models, although only at the 90% confidence level in Model 2. These findings strongly support the notion that MR influences development outcomes through two distinct mechanisms of comparable magnitude.
To assess the robustness of these findings, we conduct additional analyses. First, we replicate the main analysis using MR and impartiality indicators from the QoG Expert Survey. While the full results are reported in Table F2 (Appendix F), Table 1 (Models 3 and 4) presents estimates for direct, indirect, and total effects. In Model 3, which includes MR and two mediators, the coefficients for competence and impartiality are statistically significant and positively signed. However, in Model 4, which incorporates control variables, the coefficient for impartiality remains statistically significant at the 90% confidence level, whereas competence is no longer statistically significant. Notably, while the direct effect of MR is statistically significant in the fully specified Model 4, the total effect is significant in both models. These results diverge slightly from the main analysis, particularly regarding the competence mechanism, where the coefficient is signed positively as expected but does not reach conventional significance thresholds. This discrepancy is likely due to the limited number of observations, which may reduce the precision of the estimates.
Second, we replace GDP per capita with the Economic Complexity Index and total population with urban population, while also controlling for the levels of education (full results in Table F3, Appendix F). Our key findings remain robust across these changes, except for the indirect effect transmitted through competence when controlling for urban population and education (Models 2 and 3, Table F4, Appendix F).
Third, we re-run the analysis excluding stock and flow measures of economic performance and employment, as these may be post-treatment variables for entrepreneurship. As shown in Tables F5 and F6 (Appendix F), this exclusion does not alter the key estimates. We also excluded several countries and years, potentially impacted by irregularities in the Doing Business Index calculations. Results in Tables F7 and F8 (Appendix F) indicate that our findings remain substantially unchanged. Finally, we incorporate two additional measures of institutional quality – V-Dem’s indices of legislative and judicial constraints on the executive – and find that the results remain robust to these inclusions (Tables F9 and F10, Appendix F).
Overall, these results support the idea that MR to public administration generates two welfare-enhancing effects: improved competence of bureaucratic personnel and increased impartiality of their actions, both of which positively influence developmental outcomes such as new business entry. However, the empirical evidence for the competence mechanism is not entirely consistent, highlighting the need for further examination.
Discussion
Several important considerations should be kept in mind when interpreting the results of this analysis. First, existing tools of SEM-based formal mediation analysis lack the capacity to handle the panel structure of our data. This limitation constrains our ability to fully account for unobserved unit-specific characteristics and temporal dynamics. To mitigate this problem, we cluster standard errors by country in the main analysis, which helps reduce bias arising from unobserved heterogeneity. Additionally, we conduct a regression analysis with two-way fixed effects (reported in Table F11, Appendix F). While the latter analysis yields results largely consistent with the main findings, it does not formally test mediation, precluding causal claims. Nevertheless, we present a robust theoretical framework for understanding the mechanisms of competence and impartiality and empirically substantiate their existence through a rigorous approach designed to test causal pathways.
Second, while our main analysis relies on a meritocratic bureaucracy measure sourced from the V-Dem dataset, known for its quality and extensive temporal and spatial coverage, it is essential to recognise that this measure is based on assessments by democracy experts. This raises the possibility that their evaluations of bureaucratic institutions may be influenced by their assessments of democracy. To address this concern, our robustness checks incorporate data from the QoG Expert Survey, which relies on assessments by bureaucracy experts. However, this dataset has more limited coverage.
Third, although our sample includes 137 countries, representing at least two-thirds of countries across each world region, only 58% of low-income countries are represented. An analysis of the missing data (Appendix E) suggests that the missingness in this income group is likely due to external factors, such as data collection constraints, rather than following a Missing Not at Random mechanism. However, the underrepresentation of low-income countries may limit the generalizability of our findings.
Finally, despite extensive efforts to assess robustness, some endogeneity problems are likely to have remained unaddressed, and this requires careful interpretation of the findings.
Conclusion
In this paper, we have explored the mechanisms through which MR in public administration influences development outcomes. Building on a large literature, we argue that MR generates at least two distinct welfare-enhancing effects: competence and impartiality. These encapsulate two distinct theoretical perspectives on the root causes of human development. The SC argument emphasises the ability of the state to effectively implement agreed-upon political decisions. Viewed through this lens, merit-based recruitment imbues bureaucracy with higher epistemic quality, making it a better administrative tool that helps politicians to ‘get things done’. On the other hand, the CC argument underlines the need to constrain power-holders from abusing state power and resources, including those within the bureaucracy. This perspective brings to light another welfare-enhancing feature of MR: it enables bureaucrats to resist undue political interference and act impartially when making and implementing policies. In this way, we reconcile two opposing views of the state: one as a benign Janus, where state powers are used for the provision of public goods, and the other as a malevolent Janus, where there is a constant temptation to exploit state power for particularistic ends. By offering a unified framework for understanding the mechanisms driving development outcomes, this study makes an important theoretical contribution.
The paper also makes a significant empirical contribution by providing the first formal empirical examination of MR’s mechanisms. Employing SEM-based mediation analysis on a large cross-country dataset, we establish the existence of two distinct pathways from MR to our selected developmental outcome: one through improved competence and another through increased impartiality of the bureaucracy. Our findings remain largely consistent across various robustness checks using alternative data sources, samples, and estimation techniques.
This said, our study is not without limitations, which opens avenues for further research. First, we focus on two mechanisms of MR, but future studies may explore other potential mechanisms, such as esprit de corps or the rise of professional standards and norms. Given that MR may facilitate development through several mechanisms, capturing these with robust measures will be crucial. Second, we focus only on one organisational characteristic of public administration, and future research should examine other organisational properties, such as tenure protection or openness/closedness. The development of indicators for these features, both nationally and sub-nationally, would greatly advance this research agenda. Third, while our operationalisation of key concepts captures important dimensions, alternative operationalizations and empirical indicators should be explored. Fourth, other types of data, such as experimental and individual-level data, should be employed to further test the mechanisms of MR. Finally, as methods of formal mediation analysis improve, they should be applied to further verify our findings.
Our study carries several significant implications for both research and policy. A key research implication is the need for social scientists across disciplines to investigate the causal pathways through which organisational features of public administration and other state institutions shape developmental outcomes. Moreover, our finding that competence and impartiality represent two distinct effects of MR highlights the need for caution when using MR indicators as proxies for SC. As this paper argues, MR generates two separate effects, only one of which aligns with the SC theory of development.
Our analysis suggests that in its emphasis on safeguarding the welfare-enhancing intent of public policies and eliminating potential sources of undue influence on policy implementation, the QoG framework aligns more closely with the CC perspective than with the SC argument. Consequently, efforts to empirically measure SC through bureaucratic impartiality risk undermining the concept-measurement congruence. Similarly, the widely used International Country Risk Guide measure of bureaucratic quality, capturing both ‘expertise’ and ‘autonomy from political pressure’ (The PCR Group, n/d, 15), should be employed with caution when examining causal pathways from bureaucratic quality to development outcomes.
Our study also holds significant policy implications. Given that both competence and impartiality emerged as key pathways to improved development outcomes, prioritising competence screening practices alongside measures to reduce political influence in bureaucratic appointments should be central to reform agendas. International development cooperation efforts to reform public administration, which peaked in the 1990s but have since waned, should be revitalised and strengthened. As this paper demonstrates, MR provides answer to core developmental challenges identified by both the SC and CC theories, positioning structural public administration reform as a highly promising avenue for advancing human development. Finally, this study sends a clarion call to political entrepreneurs who advocate for dismantling public administration systems marked by incentive misalignment between politicians and bureaucrats, warning that such actions risk jeopardising bureaucratic impartiality – a characteristic of the state that has long benefited many developed countries.
Acknowledgments
We thank the editors, five anonymous reviewers, Amy Alexander, Berkay Alica, Frida Boräng, Luciana Cingolani, Patrik Hall, Carl Henrik Knutsen, Bo Rothstein, Yves Steinebach, and the participants of the QoG Institute conferences, SWEPSA, the Norwegian National Political Science Conference, the Comparative Institutions and Regimes workshop at the University of Oslo, and Circle U. workshop at King’s College London, all of whom provided valuable feedback on earlier versions of the manuscript.
Open access
