Establishing Local Government in Fragile States: Experimental Evidence from Afghanistan

World Development Vol. 77, pp. 293–310, 2016 0305-750X/Ó 2015 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev

http://dx.doi.org/10.1016/j.worlddev.2015.08.025

Establishing Local Government in Fragile States: Experimental Evidence from Afghanistan TORSTEN JOCHEM a, ILIA MURTAZASHVILI b and JENNIFER MURTAZASHVILI b,* a University of Amsterdam, The Netherlands b University of Pittsburgh, USA Summary. — International and domestic policy makers often promote elections to establish village government in fragile states. However, two additional options are available in such countries: formalization of self-governing village councils and formalization of community development councils (CDCs). We designed a survey experiment in Afghanistan that compares the consequences of elections to establish village councils to each alternative. We find that elections, and to a lesser extent formalization of CDCs, improve support for democracy, while formalization of customary councils improves support for reconciliation with the Taliban. Moreover, the consequences of transplanting elections are contingent on social norms and institutions supportive of democracy. Ó 2015 Elsevier Ltd. All rights reserved. Key words — local elections, community-driven development, state-building, liberal peacebuilding, self-governance, transplant effect, Afghanistan

1. INTRODUCTION

and traditional power brokers undermine local representation of groups during efforts to mobilize communities (Sheely, 2015). Moreover, informal institutions may reinforce underlying structural inequality (Gomes, 2015). Thus, while the formalization of customary councils is an important option to consider, the consequences are far from obvious. The choice confronting fragile states seeking to establish local government is further complicated by the presence of community-level development councils created to facilitate both political representation and economic development. The donor-sponsored community-driven development (CDD) movement has led to a profusion of new development organizations (King & Samii, 2014; Mansuri & Rao, 2012). Formalization of local councils established as part of CDD projects, known in Afghanistan as Community Development Councils (CDCs), are yet another option in fragile states where such programs are active. We use an experiment embedded in a nationally representative survey of more than 8,000 Afghan households to consider the consequences of each of these ways of establishing village government. The constitution ratified in 2004 called for the establishment of village councils as the foundation of village government, but did not specify how these councils should be

Post-conflict reconstruction of fragile states routinely includes efforts to establish democracy. 1 Part of the reason is the perceived relationship between democracy at the national level and political order (Acemoglu & Robinson, 2012; Carbone & Memoli, 2015; North, Weingast, & Wallis, 2009). Yet others argue that elections at the local rather than national level are fundamental to political order (De Figueiredo & Weingast, 1997; Myerson, 2006, 2014; Weingast, 1997). For example, Myerson (2006, 2013, 2014) argues that elections for local councils strengthens the state and instills trust in the government. Despite the seemingly obvious appeal of elections to establish local government, the literature on the ‘‘transplant effect” cautions that new institutions may not work as intended in new contexts. 2 Rather, the consequences of new institutions are likely to depend on their fit with local norms and local capacity to implement new rules (Berkowitz, Pistor, & Richard, 2003, 2001; Boettke, Coyne, & Leeson, 2008). The transplant effect suggests that the consequences of institutional reform are likely to be contextually contingent, and therefore quite challenging to anticipate. Besides the transplant effect, a challenge in most fragile states is that elections are not the only way to establish local councils. Another possibility is to formalize self-governing councils. A large literature recognizes that such selfgoverning institutions are often quite effective in providing public goods locally (Leeson & Coyne, 2012; Leeson, 2013, 2006; Ostrom, 1990, 2005). For example, traditional systems of governance led by chiefs may promote economic development, provided informal institutions constrain them (Acemoglu, Johnson, & Robinson, 2003; Acemoglu, Reed, & Robinson, 2014). In some situations, there may even be benefits from the formalization of customary councils (Baldwin, 2013; Dı´az-Cayeros, Magaloni, & Euler, 2014). However, others find that customary forms of governance are often inconsistent with the demands of a modern state (Migdal, 2001; Scott, 1992), tied to patronage networks that undermine democracy (Joireman, 2008; Tripp, 2004), or that customary

* We thank three anonymous referees and the editor Arun Agrawal for thoughtful comments that improved the quality of the manuscript substantially. Daniel Berkowitz suggested using the transplant effect literature to understand the challenge of establishing democracy in fragile states and generously read a number of drafts of the manuscript. The manuscript also benefitted greatly from comments and discussions with Kate Baldwin, Daniel Bromley, Annemie Maertens, Anjana Rajamani and Nita Rudra. Glenn Cowan and Jed Ober at Democracy International/ USAID provided invaluable support for the survey and its experimental components. The project benefitted from many discussion with Terry Hoverter and Daniel Murphy at Democracy International. Min Zaw Oo and Sabawoon Ahmadzai went to extraordinary lengths to ensure the quality of the data. Fauzia Rahimi and M. Hasan Wafaey provided invaluable input into the design of the survey and trained survey enumerators. Final revision accepted: August 27, 2015. 293

294

WORLD DEVELOPMENT

selected. Despite holding several national-level elections, village elections have not been held and thus, village government has not yet been implemented. One option is to hold elections for local councils using the same method that is already in place to select members of provincial councils and the National Assembly. The formalization of customary councils, with members selected according to traditional norms, is an alternative to formal elections to establish village government. These customary village councils, known as shuras, jirgas, or simply ‘‘elders,” are often perceived as legitimate and effective at governing even though participation in these councils is determined by community standing rather than elections (Barfield, 2010; Murtazashvili, 2016; Roy, 1990). Another possibility is to formalize CDCs, as proposed by some factions within the Afghan government (National Solidarity Program, 2012). The survey experiment randomizes the method of formalizing village councils in a fictitious electoral setting in order to identify the causal consequences of different methods of establishing local government on voter perceptions of a range of outcomes related to democracy and security. Each respondent was randomly assigned to one of the selection methods and then told how their village council would be established (either through a formal election, the formalization of the customary council, or the formalization of a CDC). Respondents were then asked several questions, including how satisfied they were with the given method of selecting local representatives; how much importance they attached to holding elections to establish a formal village council; how likely it was that village council members would deliver on promises made; the extent to which those who were selected would represent the interests of the community to higher levels of government; and whether they would support reconciliation with the Taliban. In order to tie consideration of these policy options for establishing local government to theory, we characterize the literature in terms of several broad hypotheses. The first, which we call the ‘‘liberal peacebuilding” hypothesis, is that formal elections will lead to unambiguous improvements in perceptions of democracy. Second, a ‘‘self-governance” hypothesis suggests that there will be benefits from the formalization of preexisting institutions of local governance even though such institutions may not use formal elections to select members. Third, a CDD hypothesis suggests that there will be benefits from formalizing CDCs, which are the most recent example of a large-scale CDD project in Afghanistan, as the formal village government. 3 Fourth, as customary councils are often used to resolve conflict between local groups, we hypothesize that the formalization of customary councils increases individuals’ confidence that their interests will be protected and served by a formalized customary council and thus improve respondents’ security perceptions. Finally, we investigate the ‘‘transplant effect” hypothesis, which views formal elections as beneficial but expects that the effectiveness of transplanted institutions will depend on their fit with local conditions, such as social norms supportive of democracy. In our experiment, the relevant transplant is local elections. 4 Consistent with the liberal peacebuilding hypothesis, we find that formal elections for village councils improve individual perceptions of satisfaction with elections, importance of elections, and the expectation that politicians will keep promises compared to the formalization of customary councils. However, we find that individuals are more likely to believe that politicians will be able to represent interests to higher levels of government under both formalization of customary councils and formalization of CDCs compared to formal elections. In addition, the formalization of customary councils leads to substantially more support for reconciliation with the

Taliban than the alternatives, which supports the selfgovernance hypothesis. Finally, we find that formal elections have an even larger impact on perceptions of democracy when social norms and institutions are supportive of democracy, consistent with the transplant effect hypothesis. Besides providing insight into a key policy question in Afghanistan, this research makes several broader contributions. One is conceptualizing of local elections in fragile states as an institutional transplant. Although there is a large literature contemplating the transplant effect, much of it focuses on efforts to establish economic institutions in new contexts. Theoretically, one expects fit with local institutions to also influence the impact on efforts to establish democratic institutions. Second, we complement existing studies of CDD by explicitly considering the consequences of formalizing the development councils established as a component of these projects. An important theme in the literature on CDD is the challenges of elite capture and resource misallocation (Bernard, Collion, De Janvry, Rondot, & Sadoulet, 2008; Fritzen, 2007; Platteau & Gaspart, 2003). While these studies provide much insight into the politics within CDD projects, we know less about the broader political consequences of formalizing development councils. Third, this is the first study to compare the consequences of the formalization of CDCs in Afghanistan to the formalization of customary councils. Recent impact evaluations of CDCs in Afghanistan consider the consequences of different electoral rules within a CDC (Beath, Christia, Egorov, & Enikolopov, 2014) and how the distribution of foreign aid differs when decisions about spending are made by elected CDC councils compared to when these decisions are made by a village ‘‘headman” (Beath, Christia, & Enikolopov, 2013). Neither of these studies consider the consequences of formalizing customary councils even though such councils are the de facto government in the vast majority of Afghan villages. This study is organized as follows. Section 2 describes recent constitutional reforms in Afghanistan and key features of customary councils and CDCs. Section 3 describes the experimental design and theoretical predictions. Section 4 presents the findings of the survey experiment. Section 5 concludes. 2. FROM VILLAGE GOVERNANCE TO VILLAGE GOVERNMENT Afghanistan has a unitary, centralized system of government. This has been the case since 1923, when the first constitution of the country was adopted. Since then, all formal government officials at the subnational level have been appointed by the central government in Kabul. Furthermore, sub-national units have never had substantial budgetary authority, relying instead on directives from Kabul (Murtazashvili, 2015). Informally, there has often been power-sharing between lower levels of government, with authority dispersed between district governors and village representatives. Since these instances of power-sharing reflect informal relationships, rather than a formal delegation of authority to villages, the country can be described as an informal federation within a formally centralized constitutional system (Murtazashvili, 2014). De facto power-sharing within a heavily centralized system is not only true of contemporary politics, but was also characteristic of previous governments, such as under Zahir Shah (r. 1933–73), who allowed substantial local autonomy under a formally unitary government. Several constitutions have promised some degree of local democratic representation within this unitary system. This is

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

true not only of the 1964 constitution, which first established some local elections, but also of the current 2004 constitution. 5 The 1964 constitution called for elected provincial councils, the election of mayors, and municipal councils. Importantly, it also recognized village councils. Similar to the 1964 constitution, the 2004 constitution also allows for elected provincial councils, the election of mayors, as well as municipal councils. Yet the 1964 constitution, by simply declaring representation of village government without mention of an election, implies that there would be a de jure recognition of customary governance. In contrast, the 2004 constitution explicitly calls for elected village councils, but does not mention how these councils should be elected. 6 Both historically and today, these elected bodies had weak authority vis-a`-vis local officials appointed by Kabul. Thus, this experiment considers the consequences of formal ballot-box style elections, as well as the option implicit in the 1964 constitution, which was a recognition of customary councils. The remainder of this section describes customary village councils and CDCs. The evidence is based on a survey conducted in 2011 (further described in the experimental section) and fieldwork in Afghanistan from 2006 to 2008. The fieldwork was conducted by one of the authors in 32 villages across 17 districts and six provinces of rural Afghanistan. Informants included community-identified village representatives (typically customary leaders), local government officials, religious officials, and randomly selected villagers, including an equally sized sample of men and women. Between eight and twelve interviews and focus group discussions in three languages (Pashto, Dari, and Uzbek) were conducted in each village. The research team, which included six Afghans (male and female researchers from diverse ethnic backgrounds), also interviewed dozens of officials from local and international non-governmental organizations (NGOs), donors, diplomats, and government officials at several ministries at the national and sub-national level. (a) Customary village councils Beginning with the reign of Abdur Rahman from 1880 to 1901, who earned a reputation as the ‘‘Iron Amir” for his brutality in dealing with groups that opposed his efforts to consolidate state power, successive Afghan governments attempted to ‘‘modernize” the country by weakening customary governance. Despite these efforts, informal village councils, which are usually called shuras or jirgas, remain the fundamental source of governance in rural Afghanistan. Jirga is the term used by Pashtuns to describe these councils, although in nonPashtun parts of the country individuals refer to such ad hoc councils simply as shura, ‘‘meetings” (majlis or jalasa), or as a council of elders (rish safidan, lit. ‘‘white beards”). While these councils exhibit a diversity of names, they perform similar functions. For simplicity, we refer to them as ‘‘shuras” even though the names vary by region. Anthropological accounts written before the Afghan civil war that began in 1978 found village decision-making to be characterized by cooperation and consensus, rather than feudal rigidity and exploitation (e.g., Newell, 1972). Our fieldwork demonstrated that shuras remain a source of participatory governance. With very few exceptions, informants indicated that participation in shuras is voluntary and open to any male within the community, with deliberations typically presided over by so-called ‘‘elders” (who we found to be as young as their late-20s) who earned the respect of members of their community for their fairness. One of the central functions of these councils is dispute resolution. A Pashtun elder in Behsod District, Nangarhar

295

Province, described how these deliberative bodies work to resolve disputes, as well as their role in village governance in rural Afghanistan: People usually resolve their problems through discussion in the village. The rish safidan act as judges and resolve most of the disputes in villages. The people usually refer their conflicts to the elders who then try to resolve the problems. Elders talk with both sides and then they talk with each other. If both sides accept the authority of the elders to resolve their conflict, they will make a decision that both sides are bound to accept. 7

As the following account from a male informant illustrates, these customary councils are often a viable alternative to a costly and corrupt government: [The jirga] only meets when a problem arises. People are happy with this system. They don't like to go to the woluswali [district government office] because the courts will try to take lots of our money. 8

Issues brought to the attention of a shura are typically arbitrated by community members with a reputation for piety or fairness whom the disputants personally know and trust (Wardak, 2004). In this sense, shuras are similar to reputation-based systems in which one’s status in a community influences how its members behave toward one another (Ellickson, 1991; Greif, 2006, 1993). Since our experiment focuses on the formalization of shuras as formal village governments, we only briefly introduce the functions of maliks and mullahs, the other major actors in customary village governance. A malik is a village representative, typically selected by community members (frequently through a shura process), who serves as the face of the village to the outside world, especially to district governors. Maliks also serve as administrators who can help to resolve disputes and are also likely to participate in shuras. Rather than a chief of a village, a malik is typically a ‘‘first among equals.” 9 Although they are an integral component of village governance, maliks derive their ultimate authority from the consent of the village. Mullahs, in contrast, are religious arbiters who also participate in village governance. 10 Figure 1 depicts the system of customary governance. Although the fieldwork was conducted in a fairly large number of villages, there are between 20,000 and 40,000 villages in the country (estimates vary widely). The survey confirms the finding from fieldwork regarding the ongoing significance of customary governance in village life despite government efforts to weaken it. Table 1 shows that the majority of Afghans have such organizations in their communities.

Figure 1. Participation in customary councils.

296

WORLD DEVELOPMENT Table 1. Local governance in Afghanistan Do you have a malika?

Yes No Don’t know

Do you have a shura/jirga?

Total (%)

Villages (%)

Urban (%)

Total (%)

Villages (%)

Urban (%)

99.7 0.3 0.0

99.7 0.3 0.0

99.7 0.3 0.0

81.2 14.7 4.1

83.5 12.7 3.8

73.7 21.4 4.9

Do you have a mullah?

Yes No Don’t know

Do you have a CDC?

Total (%)

Villages (%)

Urban (%)

Total (%)

Villages (%)

Urban (%)

99.5 0.5 0.0

99.4 0.6 0.0

100.0 0.0 0.0

68.0 22.8 9.1

71.5 20.8 7.6

56.4 29.4 14.2

Full sample with 8,620 observations. a Alternatively: arbab, qaryadar, wakil, or namayenda.

Table 2. Confidence in customary and state institutions in Rural Afghanistan Confidence in . . . A lot of confidence Some confidence No confidence

Malika (%)

District governor (%)

CDC (%)

Provincial Council (%)

Police (%)

National Assembly (%)

38.8 48.3 10.6

28.8 37.7 30.7

17.8 45.6 28.9

17.0 43.4 34.7

14.8 29.0 53.3

7.5 45.1 43.5

Proportions based on 6,450 rural observations. a Alternatively: arbab, qaryadar, wakil, or namayenda.

The survey also indicates that customary governance is highly accessible compared to both the government and CDCs. For example, 40.3% said their shura/jirga was ‘‘very accessible,” compared to only 31.3% for the CDC. Moreover, as Table 2 shows, a higher percentage of people have ‘‘a lot of confidence” in their customary representatives than in the district governor (woluswal), police, National Assembly (Wolesi Jirga), and provincial councils. (b) Community development councils The post-2001 reconstruction effort has established many new formal governing organizations, although several major institutional reforms originated during Taliban times and before (Esser, 2013). Community development is an example of a reform process which has origins predating the current state-building process. The government-sponsored implementation of CDCs throughout Afghanistan since 2003 builds upon earlier efforts by NGOs to create ‘‘development shuras” in the 1980s in rural areas (Carter & Connor, 1989) and the establishment of similar councils by UN-HABITAT in the 1990s (Affolter, Sawayz, Noori, & Shrestha, 2006). Indeed, several informants mentioned that they were familiar with development shuras that predated the establishment of a CDC in their community. As one informant explained, ‘‘It [the earlier shura] was the same as a CDC. It was organized the same way. It had a head, a deputy, and other members.” 11 The main difference, according to this informant, was that the earlier development shuras were not supported by the government. While the concept of community development is not new, the scale of its implementation has been much more ambitious since 2001. The Afghan interim government established the National Solidarity Program (NSP) in 2002. Beginning in 2003, the Afghanistan Ministry of Rural Rehabilitation and Development (MRRD), with generous funding from the Afghanistan Reconstruction Trust Fund, administered by the World Bank, began to implement the NSP. The goals of the NSP include helping villages select donor-supported

projects for their communities and to serve as ‘‘a viable alternative to the traditional local governance structures that has provided a vehicle to re-building the social fabric and relationships at grassroots level” (National Solidarity Program, 2012). 12 As of this writing, MRRD has created more than 30,000 CDCs throughout rural Afghanistan with more than $1.5 billion in funding from the World Bank and other donors. According to the MRRD and the World Bank, a CDC should comprise no fewer than 50 but no more than 300 households. Each CDC initially receives $200 per household in block grant funds (between $10,000 and $60,000 per community) for community-selected projects. 13 Communities are required to provide a 10% contribution to each project, including through in-kind contributions, such as labor in construction projects. From field interviews and observations, individuals most commonly contribute through in-kind contributions of labor. Each CDC is supposed to have elections for members of an executive board consisting of a head, deputy, secretary, and treasurer. A major objective in the creation of CDCs was to increase the female participation in decisions involving distribution of funds since customary and other local organizations do not always include women in public deliberations. According to the initial NSP directives, CDCs should be equally representative of men and women and both sexes should meet jointly. Funds are to be distributed to the community only after these village elections have been held. These elections are not implemented by the government, but instead are typically presided over by the mostly foreign NGOs charged to implement the program. Although NSP is a government program, the funding, program design, and implementation relies almost entirely on foreign sources. To implement these councils, MRRD subcontracted with more than 30 organizations, mostly international NGOs. In most cases, it contracted with NGOs at the district level and afforded them substantial leeway in determining the size of CDCs and demarcating borders between communities.

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

This implementation strategy often caused confusion among rural Afghans. Many of the informants believed the CDCs were just another ‘‘development shura.” A teacher in Sorkhrod District, Nangarhar Province recounted growing tensions in the village that resulted from the presence of a foreign NGO implementing the NSP in the district. This account also clarifies a clear contrast between customary governance and CDCs: It is not clear who has what role and what responsibility. We need just one council here in our village, not all of these development councils. We have one qawmi [tribal] shura that meets with the woluswal. But a problem with these CDCs is that the woluswal cannot ask about their work in the villages. He has no authority over them. And there are all these foreigners working in our villages trying to replace our village system. If the government or villagers ask these musaseh [NGOs] how the money is being spent, they will tell you it is none of your business and that we have no right to interfere. I think this is a policy from other countries that want to remove our maliks from the villages. If the people do not like their malik, then they have the right to remove them and they can select another malik. If people cannot trust a malik, the shura can select another one. It is not the role of foreigners to remove maliks. 14

The evidence from the impact evaluations referenced earlier, which suggest that CDCs have improved outcomes such as representation of women and quality of public goods provision, alongside MRRD promoting them to be recognized as formal bodies of village government, suggests the importance of considering the consequences of formalizing these councils as village government. In addition, since donors played such an important role in establishing them with selection rules that differ from formal election rules in the country, the formalization of CDCs is an alternative to a formal election to establish village councils. However, from the accounts above, it is unclear whether the formalization of CDCs would improve perceptions of either democracy or security compared to either an election, or a formalization of customary councils.

3. EXPERIMENTAL DESIGN AND THEORETICAL PREDICTIONS (a) Survey design The experiment was embedded in a nationally representative, face-to-face survey of 8,620 individuals across all 34 provinces of Afghanistan, which was fielded in November 2011. The research was carried out by the Afghanistan Electoral Reform and Civic Advocacy (AERCA) project and Democracy International, an independent organization based in Washington, DC, with funding from the United States Agency for International Development. 15 The survey was conducted by the Opinion Research Center of Afghanistan (ORCA), one of the largest independent survey firms in Afghanistan. ORCA used a five-stage sampling plan that randomized selection of districts and within-district settlements. Within settlements, households were selected by randomizing walking directions, and within households, interviewers used a Kish grid for all above 18-year old household members to randomize the target respondent. A raking algorithm was used to compute post-stratification weights based on the 2011–12 population estimates of the Afghan Central Statistical Office on provincial population numbers, rural/urban ratios as well as gender ratios. Table 3 summarizes the survey design and provides additional information on quality control and response rates. The experiment considers three options to establish village councils in Afghanistan. One is a state-run election that would be conducted in the same way as presidential elections but at

297

the local level, with certification from the Independent Election Commission (IEC). A second option is to formalize customary councils, with members selected by consensustype procedures without the use of formal ballot boxes. A third option is to formalize a CDC as the village government. Although CDCs may be established with an election procedure, the selection of councils involved oversight and implementation by the mostly foreign NGOs contracted to create the councils with selection rules that differed from election procedures already established to select members of elected provincial councils and the National Assembly. Figure 2 summarizes each of these alternatives that we consider in the experiment. The manner by which a village council is established is the experimental treatment. The treatment was delivered to respondents as a vignette. Table 4 describes each alternative as delivered to respondents in the survey. After receiving one of the randomly determined treatments, each respondent was then asked the following questions:  How satisfied are you with this method of selecting local representatives?  How important are formal elections at the local level?  In your opinion, how likely is it that politicians will do what they say?  In your opinion, how likely is it that the people selected through this process will be able to represent the interests of the community to the district governor (woluswal)?  How likely would you be to support reconciliation with the Taliban or other insurgent groups in Afghanistan? Each of these questions addresses a potentially important aspect of the state-building process. Satisfaction with procedures to establish local councils is clearly an important consideration in the process of institutional transformation. It is also worthwhile to consider whether the procedures for establishing local government influence support for elections. Since politicians in fragile states often struggle to establish credibility with citizens (Coyne & Boettke, 2009; Flores & Thomas, 2012), we are further interested in the extent to which selection methods influence citizens’ expectation that candidates will keep their promises. The belief that members of the village council will be able to represent interests to higher levels of government provides insight into the quality of representative institutions. Respondents were also asked about support for reconciliation, which in a broad sense refers to bringing previously warring groups together in governance. Reconciliation is an important issue in rural Afghanistan where the Taliban insurgency continues to fight the government and foreign forces. However, the issue of reconciliation is a complicated one that involves many dimensions, such as relations between Afghanistan and Pakistan, resolution of grievances among political factions within the country (for example, between the Northern Alliance and the Taliban), and settling of other more localized grievances. While our experiment cannot address each of these issues, it provides insight into how the method of establishing local councils influences support for reconciliation with the groups directly tied to violence against the state. Under successful randomization, respondent characteristics should be independent from the assigned selection method, thereby overcoming the methodological shortcomings of non-randomized surveys or observational data. Importantly, survey experiments, including ours, do not introduce actual changes in institutions. Nonetheless, the observed differences across treatments remain unbiased. If subjects attached a lower importance to a survey experiment due to the lack of tangible payoffs, this would lead to more noise in all

298

WORLD DEVELOPMENT Table 3. Survey design

treatments, thus increased standard errors and reduced statistical significance on differences between treatments, while not affecting the magnitudes of treatment differences themselves (Mutz, 2011). The experiment also provides insight into the transplant of new institutions. Although the country has held presidential, provincial council, and national assembly elections, there have never been formal elections for village government. The challenges in implementing provincial council and national assembly elections illustrate that democracy is far from institutionalized. Note, however, that depicting elections as a transplant does not mean that there is no representation at the local level. Indeed, we expect that democracy will be more

likely to transplant successfully in the presence of since this provides an informal mechanism of representation of individuals in village governance. In addition, even though there may have been elections to select CDCs, NGOs (mostly foreign) presided over these elections, determining who was eligible to participate and even demarcating lines between communities based on contracts with the MRRD and the World Bank. As a result, these elections which were designed to help communities select community-development projects with generous donor funding, were perceived by rural Afghans during fieldwork as quite different from a state-run election. 16 For these reasons, we conceptualize of an election for a village council as an example of an institutional transplant.

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

299

Figure 2. Three policy options for selecting council members. Table 4. Alternative selection scenarios

(b) Theory Our hypotheses are derived from several broad theoretical perspectives. First, a large literature suggests that establishing formal representation at the local level improves prospects for political order (Myerson, 2014). Local elections are also viewed as important for improving the efficiency of public goods provision (Martinez-Bravo, 2014; Olken, 2010). A related literature focuses on the long-run links between political representation at lower levels of government, the emergence of rule of law, and economic development (De Figueiredo & Weingast, 1997; Weingast, 1997). These perspectives expect that elections for local councils will improve confidence in democratic institutions and security compared to alternative methods of selecting local representatives. A competing second perspective suggests that there may be benefits from formalizing preexisting governance institutions. Studies of self-governance recognize that institutions may arise spontaneously, without conscious design by the state (Greif, 2006; Leeson, 2007a; Williamson, 1991). Formalizing customary systems of representation may thus be an important option since existing institutions can represent a best response given contextual constraints and the costs of establishing new institutions (Leeson, 2006, 2007b, 2013).

Despite the pervasiveness of customary governing institutions in many fragile states, the literature is divided regarding the role of such institutions in post-conflict state-building processes. On the one hand, informal institutions may be effective, and perhaps even efficient, in providing local public goods such as dispute resolution (Dı´az-Cayeros et al., 2014; Leeson & Coyne, 2012; Leeson, 2007b; Menkhaus, 2006). For example, Baldwin (2013) finds a productive role for customary sources of order, showing that tribal chiefs connected to patronage networks are often quite effective in gaining access to state resources for constituents. On the other hand, Olken (2010) argues that customary and ‘‘traditional” forms of self-governance may also be more prone to elite capture, patronage, and rent-seeking. These diverging perspectives demonstrate the importance of evaluating the consequences of formalizing self-governing organizations such as shuras. 17 Third, an extensive literature considers the consequences of CDD in developing countries. Mansuri and Rao (2012) and Miguel and Gugerty (2005) explain that despite their popularity among international organizations, it is unclear how effective such projects are in facilitating democracy or development. It remains an open question whether formalizing such councils as formal village government would improve perceptions of the quality of democracy. The experimental research

300

WORLD DEVELOPMENT

evaluating NSP referenced above suggests that donor-assisted selection improves the provision of some public goods due to the infusion of massive donor funds, it is unclear whether the formalization of CDCs would generate higher satisfaction with democracy in comparison to the formalization of customary councils or a state-run election to establish local government. Finally, the ‘‘transplant effect” refers to the consequences of importing new institutions. Studies of the transplant effect counsel caution when establishing new institutions, as their effectiveness in new contexts are hard to predict (Acemoglu, Cantoni, J., & Robinson, 2011; Boettke et al., 2008; Berkowitz, Pistor, & Richard, 2001, 2003). Specifically, the transplant effect literature suggests that the presence of market intermediaries, such as judges and bureaucrats, will be important to whether new institutions take hold (Berkowitz et al., 2001; Seidler, 2014). Similarly, a literature in public administration maintains that the successful implementation of governance reform depends on the extent to which new policies are compatible with preexisting political and social norms (‘‘context”) and competence (Andrews, 2013; Grindle, 2007). This literature leads to three competing hypotheses regarding the relationship between selection methods and outcomes related to democracy. Hypothesis 1. (liberal peacebuilding): Formal elections will produce higher levels of satisfaction with selection methods, increased support for formal elections, increased perceptions that politicians will do as they say, and an increase in confidence that politicians will be able to represent members of the community to higher levels of government compared to alternative selection methods (henceforth, we refer to these outcomes as ‘‘democratic perceptions”).

Hypothesis 2. (self-governance): Formalization of customary councils will produce improved democratic perceptions. Hypothesis 3. (community-driven development): Formalization of CDCs will produce improved democratic perceptions. While competing, these hypotheses are not fully mutually exclusive: formal elections may perform better along some dimensions while self-governing councils and community

development councils may perform better along other dimensions. The hypothesis for reconciliation with warring groups in society is treated separately from the earlier ones since the issue of reconciliation concerns the competency of different types of representative institutions to resolve disputes. Since shuras are long-standing, traditional forums for resolving conflict between groups, including rival social groups (qawms), we hypothesize that the formalization of customary councils will yield increased support for reconciliation with the Taliban. 18 By formalizing a system of governance that has a proven record for resolving conflict between groups and has functioned for decades under conditions of conflict, we expect that individuals will be more confident that their interests will be protected and served by the village government, and thus more likely to support reconciliation. 19 Hypothesis 4. (security perceptions): Formalization of customary councils will improve security perceptions. The literature on liberal peacebuilding, to the extent that it associates elections with political order and stability, also suggests that elections may lead to improvements in security. The experiment also allows us to consider this possibility. There are two broad factors that are expected to influence the successful transplantation of new institutions to fragile states. One is the presence of customary systems of governance. Berkowitz et al. (2003) focus on the formal context, such as formal judges and bureaucrats who implement and interpret laws. Such bureaucratic capacity is viewed as important to understanding governance outcomes (Fukuyama, 2013). In a fragile state, informal or customary modes of governance often substitute for formal judges and bureaucrats and therefore become important for the presence of a transplant effect. Electoral institutions are more likely to transplant when political institutions are inclusive. One way to think about this is in terms of function and form. As Rodrik (2007) explains, it is critical to distinguish the functions performed by institutions from the specific form of institutions as different forms may perform similar functions. In a context where institutions are functionally similar to Western-style representative institutions, but differ in form, one expects less friction in the process of democracy promotion. We test the transplant effect hypothesis with respect to preexisting capacity in Hypothesis 5.

Figure 3. Theoretical predictions.

36.5*** 39.0*** 2.6*** 2.6** 4.7*** 2.1***

Full sample. Multivariate results shown in Table 9. *, **, *** represent statistical significance on the 10%, 5%, and 1% significance level.

2,155 4,310 Observations

2,155

37.2 53.1 9.7 Somewhat or very likely Not very or not at all likely Don’t know/Refused Panel E: Support for Reconciliation

34.6 57.7 7.6

73.7 14.1 12.3

39.0*** 43.7*** 4.7***

2.0*** 3.5*** 1.5*** 3.4*** 3.4*** 0.0 74.4 17.5 8.1 Somewhat or very likely Not very or not at all likely Don’t know/Refused Panel D: Representation of Local Interests

77.8 14.1 8.1

76.4 14.0 9.6

1.4* 0.1 1.5**

4.1*** 3.4*** 0.7** 3.1*** 3.4*** 0.3 83.2 11.6 5.2 Somewhat or very important Not very or not at all important Don’t know/Refused Panel C: Promise Keeping

80.1 14.9 5.0

79.1 15.0 5.9

1.0 0.0 1.0

5.0*** 2.9*** 2.1*** 5.6*** 3.6*** 2.0*** 0.6 0.7 0.1 86.3 11.8 1.9 Somewhat or very important Not very or not at all important Don’t know/Refused Panel B: Importance of Elections

85.7 12.5 1.8

80.7 15.3 4.0

2.2*** 1.9*** 0.3 2.5*** 2.5*** 0.0 83.8 13.4 2.9 Fairly or very satisfied Not very or not at all satisfied Don’t know/Refused

83.5 14.0 2.5

81.3 15.9 2.8

0.3 0.7 0.4

CDC to CUST. (%) (6) ELECT to CUST. (%) (5) ELECT to CDC (%) (4) CUST. (%) (3) CDC (%) (2) ELECT (%) (1)

Treatment

(a) Selection methods and perceptions of democracy and security Table 5 presents the experimental findings regarding the comparative consequences of each of the three methods of establishing village councils. Table A.1 in the internet appendix provides balance tables with key demographics of respondents by treatment group. While most differences in the balance tables are insignificant, we provide as a robustness check in Table 9 results from a regression analysis where we control for a large set of socio-economic observables, geographic fixed effects, and sampling point (or settlement) fixed effects to ensure that the results are not driven by anything other than the randomly assigned treatments. Since the inclusion of additional variables does not change any of the substantive findings, an interpretation of the statistically

301

Panel A: Voter Satisfaction

4. EXPERIMENTAL FINDINGS

Table 5. Univariate results by selection regime

Hypothesis 6. (social norms): Formal elections will be less likely to yield improvements in democratic perceptions in the presence of social norms that are inconsistent with Westernstyle elections. Figure 3 summarizes the theoretical predictions.

Response

Hypothesis 5. (preexisting capacity): Formal elections will be more likely to yield improvements in democratic perceptions in the presence of deliberative councils. Social norms are a second factor expected to influence the ability of institutions to transplant to new contexts. Acemoglu and Jackson (2015) and Boettke, Coyne, Leeson, and Sautet (2005) consider how fit with social norms explains the consequences of new institutions. These studies suggest that the presence of social norms that are consistent with democracy will influence how well it transplants. We test the transplant effect hypothesis with respect to social norms in Hypothesis 6 by considering whether the consequences of elections depend on affinity for different types of political regimes, experiences with preexisting deliberative institutions at the village level, and social norms that may conflict with elections. First, we expect the impact of elections on perceptions of democracy will be less favorable when an individual’s values are more aligned with authoritarian, theocratic or tribal rule. The reason is that support for the latter three political regimes suggests less of a fit with formal Western-style elections. Second, experiences with a local deliberative council such as a shura or CDC—particularly when those are more inclusive and accessible to locals—will improve perceptions of democracy under the election treatment. Third, we consider whether Pashtuns’ social norms and support for Taliban rule affect perceptions about the expected performance of democracy. Some scholars suggest that the Taliban insurgency is a Pashtun insurgency that is fundamentally at odds with the kinds of formal democratic institutions that the Taliban views as imposed by the West (Johnson & Mason, 2008). Anthropologists have maintained that Pashtun norms are largely egalitarian (Ahmed, 2013; Barth, 1959). To the extent that Pashtun norms are egalitarian, they may be more receptive to democratic institutions. These contrasting perspectives suggest that Pashtun norms may influence the transplantation of formal elections, although the literature is divided regarding whether such norms will inhibit or facilitate the transplantation of formal elections. 20

Change from. . .

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

302

WORLD DEVELOPMENT

significant findings accompanies the presentation of the univariate results. Panel A of Table 5 presents the results for satisfaction with the method of selecting village councils. Although formal elections are statistically indistinguishable from the formalization of CDCs, both elections and formalization of CDCs leads to significantly more satisfaction with the selection method than the formalization of customary councils. The finding that formal elections and a formalization of CDCs significantly increases voter satisfaction is consistent with the liberal peacebuilding and CDD hypotheses, yet inconsistent with the self-governance hypothesis. The results on the importance of elections also illustrate that formal elections lead to increased support for elections relative to the formalization of customary councils (Panel B of Table 5). Formal elections lead to 5.6% greater support as a selection process than the formalization of customary councils, while there is 5.0% greater support when CDCs rather than customary councils are formalized. The experiment yields no statistical difference between formal elections and the formalization of CDCs. As such, these results are also consistent with the liberal peacebuilding and CDD hypotheses. The belief that candidates are more likely to keep their promises increases by 3.1% (relative to a formalization of CDCs) and 4.1% (relative to a formalization of customary councils) when formal elections are used to establish village councils (Table 5, Panel C). There is no statistically significant difference between formalizing CDCs or customary councils regarding promise keeping. Although there is a common belief by the Afghan government and donor community that CDCs generate more accountability than customary governance, we find no evidence that CDCs improve expectations that politicians will keep promises compared to customary councils. Panel D of Table 5 presents the findings regarding beliefs that local representatives will be able to represent community interests to the government. Both the formalization of customary councils and a formalization of CDCs increase the belief that members of their village council will be able to represent the interests of their community compared to formal elections, with statistically significant increases of 2.0% and 3.4%, respectively. 21 In contrast to the prior three results, this finding provides some evidence in support of the self-governance hypothesis. It also supports the CDD hypothesis, while we find no support for the liberal peacebuilding hypothesis. Panel E of Table 5 shows the strongest finding in terms of magnitude: the formalization of customary councils produces much larger support for reconciliation with the Taliban or other insurgent groups than either alternative. In fact, support for reconciliation nearly doubles from 37.2% (elections) and 34.6% (formalization of CDCs) to 73.7% under the formalization of customary councils. This is a potentially important finding in terms of the peacebuilding process because it suggests that customary methods of selecting local representatives increases the extent to which citizens support reconciliation with groups tied to violence, which is a potentially relevant finding regarding how to attain long-term peace. A plausible explanation for this finding is that the effectiveness of customary councils as a source of dispute resolution leads to a greater belief that reconciliation will be successful, or will have lasting benefits. This explanation is also supported from our fieldwork that indicated that customary organizations have important advantages in resolving disputes compared to formal political institutions such as courts, and that people rarely turn to CDCs to resolve disputes. An alternative explanation is that respondents believe customary authorities are better positioned to protect their interests vis-a`-vis a government that includes

Taliban interests than if members are selected using formal elections or CDC methods. Finally, some may believe that the Taliban may be more willing to negotiate and deal with customary authorities on local governance matters than those selected through a top-down, government-induced process (CDCs or formal elections). While we cannot fully separate the channel, the formalization of customary councils increases in either case respondents’ support for reconciliation and may in turn enhance overall chances for a political compromise. The results in Panels A–D of Table 5 indicate statistically significant differences across different methods of establishing local councils on democratic perceptions in the range of 2–6 percentage points. As Table 9 shows, the differences between treatments are substantively unchanged (and most with p-values below 0.01) after including a large set of respondent- and geographic-specific factors. Thus, the robustness checks exclude the possibility that these finding can be explained by chance. On the one hand, several of these findings support the broad contention that elections are an important way to establish local government. On the other hand, the benefits of elections is not uniform. For some outcomes, formalization of CDCs would yield outcomes that are statistically indistinguishable from elections, and for representation to higher levels of government, a case can be made to formalize customary councils or CDCs. As far as the security-related outcome, we find a very sizable effect under a formalization of customary councils. Specifically, we find a doubling of the support for reconciliation in the treatment in which customary councils are formalized. This finding, which is also robust to the inclusion of a large number of additional variables, suggests that the formalization of customary councils is an important option in situations in which reconciliation is viewed as an important path to political peace. (b) Testing the transplant effect We next test the transplant effect hypothesis by considering whether the consequences of transplanted institutions (formal elections and the formalization of CDCs) vary with several variables, including attitudes toward different types of political regimes (Table 6), the experience with preexisting deliberative institutions at the village level (Table 7), and social norms that may be opposed to formal, Western-style elections (Table 8). We use evidence from the non-experimental part of the survey, which include questions about norms, beliefs, and ideology, as well as on the presence of and access to different local institutions. Since our interest is in the effect of transplanted institutions, we only interpret how the effect of elections or formalization of CDCs vary with social norms and preexisting institutional capacity (customary councils are not transplants). Moreover, as the transplant effect hypothesis most clearly applies to democracy, we focus on the link between transplanted institutions, context, and perceptions of democracy, rather than reconciliation. The first test of the transplant effect considers whether the consequences of establishing local councils through elections or formalization of CDCs vary with attitudes toward different political regimes. According to the transplant effect hypothesis, beliefs supportive of democracy should influence individual perceptions of different outcomes and thus the consequences of transplanted institutions. The first way we measure these beliefs is through support for elections. One of the survey questions asked, ‘‘Would you say that having elections to choose national and local leaders is a very good, a fairly good, neither good nor bad, a fairly bad, or

Table 6. Results by support for political regime

B: Authoritarian Democracy 4 H0: Autocratic & ELECT – Liberal & ELECT = 0 5 H0: Autocratic & CDC – Liberal & CDC = 0 6 H0: Autocratic & ELECT – Autocratic & CDC = 0 C: Islamic Democracy 7 H0: Religious & ELECT – Secular & ELECT = 0 8 H0: Religious & CDC – Secular & CDC = 0 9 H0: Religious & ELECT – Religious & CDC = 0 D: Decentralized Democracy 10 H0: Tribal & ELECT – Federal & ELECT = 0 11 H0: Tribal & CDC – Federal & CDC = 0 12 H0: Tribal & ELECT – Tribal & CDC = 0

Exp.

Voter Satisfaction

Election Importance

Promise Keeping

Representation of Interests

+

0.220*** (p < 0.001)

0.154*** (p < 0.001)

0.188*** (p < 0.001)

0.119*** (p = 0.001)

+

0.145** (p = 0.038)

0.086 (p = 0.168)

0.112 (p = 0.153)

0.086 (p = 0.168)

n.a.

0.038 (p = 0.230)

0.038 (p = 0.166)

0.037** (p = 0.012)

Obs.

2,479

2,508

2,447

0.010 (p = 0.286)

2,333

0.048** (p = 0.012)

0.044*** (p = 0.006)

0.021 (p = 0.314)

0.045* (p = 0.064)

0.083*** (p < 0.001)

0.049** (p = 0.011)

0.050** (p = 0.040)

0.066** (p = 0.036)

n.a.

0.030* (p = 0.092)

0.021 (p = 0.179)

0.054 (p = 0.476)

Obs.

3,092

3,116

2,980

0.016 (p = 0.698)

0.026 (p = 0.438)

0.033 (p = 0.533)

0.064 (p = 0.179)

0.068 (p = 0.107)

0.069 (p = 0.234)

n.a.

0.007 (p = 0.799)

0.022 (p = 0.520)

0.085* (p = 0.091)

0.006 (p = 0.845)

Obs.

2,479

2,508

2,447

2,333

0.009 (p = 0.775)

0.006 (p = 0.856)

0.081 (p = 0.195)

0.035 (p = 0.420)

0.004 (p = 0.914)

0.007 (p = 0.829) 0.002 (p = 0.943)

0.014 (p = 0.478) 2,858 0.044 (p = 0.262)

0.020 (p = 0.529) 0.022 (p = 0.604)

n.a.

0.040 (p = 0.167)

0.015 (p = 0.470)

0.044** (p = 0.043)

0.003 (p = 0.942)

Obs.

2,719

2,676

2,686

2,582

Each cell contains the results of a hypothesis test of equality after the marginal effects from a Logistic regression have been computed. The full model, joint effects and hypothesis tests are shown in the appendix in Table A.2. The Logistic probability model has a 0 (‘‘not very or not at all important/satisfied/likely”) or 1 (”somewhat, fairly or very important/satisfied/likely”) as the dependent variable. The model includes the treatment indicators (CDC and CUST) as well as extensive vectors of subject and geographic controls. The Subject controls include indicator variables for gender and head of household as well as fixed effects for age groups, ethnicities, income and education categories, religion, marital status and job-type categories. Geographic controls include fixed effects for regions, provinces, and degrees of urbanization (village, town, city, metro) as well as a provincial security measure based on ANSO data. Column 3 shows the expected sign for coefficients under the transplant hypothesis. All specifications include heteroskedasticity-robust standard errors which are clustered at the primary sampling unit (settlement). The sample *, **, *** indicate statistical significance at the 10%, 5%, and 1% level.

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

A: Supportive of Elections 1 H0: Support Elections & ELECT – No Support Elections & ELECT = 0 2 H0: Support Elections & CDC – No Support Elections & CDC = 0 3 H0: Support Elections & ELECT – No Support Elections & CDC = 0

303

304

WORLD DEVELOPMENT Table 7. Results by local governance institutions

Dependent variable

Exp.

Variable of Interest

Panel A: Existence of shura or CDC in respondent’s settlement Voter Satisfaction + d(shura or CDC): Election Importance + d(shura or CDC): Promise Keeping + d(shura or CDC): Representation of Interests + d(shura or CDC): Panel B: Accessibility of local shura or CDC to respondents Voter Satisfaction + Accessibility: Election Importance + Accessibility: Promise Keeping + Accessibility: Representation of Interests + Accessibility:

0.074*** (0.014) 0.050*** (0.013) 0.045*** (0.013) 0.030** (0.015)

0.028*** (0.008) 0.017** (0.008) 0.031*** (0.009) 0.038*** (0.011)

Subject Controls

Geographic Controls

Obs.

Yes Yes Yes Yes

Yes Yes Yes Yes

8,314 8,336 8,099 7,787

Yes Yes Yes Yes

Yes Yes Yes Yes

5,537 5,568 5,438 5,298

Each row corresponds to a Logistic probability model where the dependent variable is either 0 (‘‘not very or not at all important/satisfied/likely”) or 1 (‘‘somewhat, fairly or very important/satisfied/likely”). The two panels make use of information on the existence and accessibility of local deliberative institutions in respondents’ settlements. d(shura or CDC) equals 1 if at least one of the institutions is present, else 0. Accessibility ranks between 1 (‘‘not accessible at all”) and 4 (‘‘very accessible”). The expected sign of the indicator variables under the transplant hypothesis is shown in column 2. Subject controls include indicator variables for gender and head of household as well as fixed effects for age groups, ethnicities, income and education categories, religion, marital status, work status and job-type categories. Geographic controls include fixed effects for regions, provinces, and degrees of urbanization (village, town, city, metro). All specifications include heteroskedasticity-robust standard errors which are clustered at the primary sampling unit (settlement). *, **, *** indicate statistical significance at the 10%, 5%, and 1% level.

Table 8. Pashtun ethnicity and support for the Taliban Exp. A: Pashtun vs. Non-Pashtun Ethnicity 1 H0: Pashtun & ELECT – nonPashtun & ELECT = 0 2 H0: Pashtun & CDC – non-Pashtun & CDC = 0 3 H0: Pashtun & ELECT – Pashtun & n.a. CDC = 0 Obs. B: Pashtun Pr. vs. Non-Pashtun Province 4 H0: Pashtun Pr. & ELECT – nonPashtun Pr. & ELECT = 0 5 H0: Pashtun Pr. & CDC – nonPashtun Pr. & CDC = 0 6 H0: Pashtun Pr. & ELECT – n.a. Pashtun Pr. & CDC = 0 Obs. C: Support vs. Opposition to Taliban 7 H0: Sympath. & ELECT – Advers. & ELECT = 0 8 H0: Sympath. & CDC – Advers. & CDC = 0 9 H0: Sympath. & ELECT – Sympath. n.a. & CDC = 0 Obs.

Voter Satisfaction

Election Importance

Promise Keeping

Representation of Interests

0.043*** (p = 0.004)

0.040*** (p = 0.005)

0.008 (p = 0.639)

0.026 (p = 0.175)

0.036** (p = 0.039)

0.016 (p = 0.412)

0.004 (p = 0.877) 0.019 (p = 0.137)

0.008 (p = 0.519)

0.011 (p = 0.249)

0.043*** (p < 0.001)

8,334

8,358

8,120

0.014 (p = 0.423)

7,806

0.116*** (p < 0.001)

0.170*** (p < 0.001)

0.114*** (p < 0.001)

0.137*** (p < 0.001)

0.126*** (p < 0.001)

0.168*** (p < 0.001)

0.133*** (p < 0.001)

0.146*** (p < 0.001)

0.017 (p = 0.139)

0.011 (p = 0.283)

0.050*** (p < 0.001)

8,314

8,336

8,099

0.024 (p = 0.109) 7,787

0.137*** (p < 0.001)

0.134*** (p < 0.001)

0.109*** (p = 0.001)

0.075** (p = 0.016)

0.132*** (p < 0.001)

0.114*** (p < 0.001)

0.095*** (p = 0.003)

0.077** (p = 0.032)

0.013 (p = 0.558) 6,491

0.002 (p = 0.784) 6,508

0.034 (p = 0.239) 6,330

0.013 (p = 0.682) 6,120

Each cell contains the result of a test of equality of the marginal effects from a Logistic probability model. The dependent variable in the Logistic probability model is either 0 (‘‘not very or not at all important/satisfied/likely”) or 1 (”somewhat, fairly or very important/satisfied/likely”). (Table A.3 in the internet appendix shows the full marginal effects.) Panel A separates respondents by ethnicity; Panel B by place of living, and Panel C by whether respondents expressed ‘‘a lot of confidence” versus ‘‘no confidence” in the Taliban. Column 3 shows the expected sign for coefficients under the transplant hypothesis. ELECT, CDC, and CUST represent the assigned treatments. All models include the treatment indicators and subject and geographic controls. Subject controls include indicator variables for gender and head of household as well as fixed effects for age groups, ethnicities, income and education categories, religion, marital status and job-type categories. Geographic controls include fixed effects for regions, provinces, and degrees of urbanization (village, town, city, metro) as well as a provincial security measure based on ANSO data. In Panel A, ethnicity fixed effects are omitted, while in Panel B province fixed effects have been omitted. All specifications include heteroskedasticity-robust standard errors which are clustered at the primary sampling unit (settlement). The sample *, **, *** indicate statistical significance at the 10%, 5%, and 1% level.

a bad way of governing Afghanistan?” We separate subjects by whether they responded ‘‘a very good way” or ‘‘a bad way” and include an indicator variable based on this separation into a Logistic model. We then computed the marginal effects and

test for equality between the joint marginal effects. To keep the tables readable, we only report the differences in the joint effects between groups.22 Our expectation is that there will be a larger positive impact of formal elections on the outcomes

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

305

Table 9. Different selection regimes in a regression framework Response

Subject Controls

Geographic Controls

Settlement fixed effects

Voter Satisfaction

No Yes No Yes Yes

No No Yes Yes Yes

No No No No Yes

0.007 0.012 0.006 0.013 0.014

(0.010) (0.010) (0.009) (0.009) (0.010)

0.026 (0.011)** 0.023 (0.010)** 0.024 (0.010)** 0.018 (0.009)* 0.021 (0.011)**

Importance of Elections

No Yes No Yes Yes

No No Yes Yes Yes

No No No No Yes

0.007 0.011 0.006 0.012 0.012

(0.009) (0.009) (0.008) (0.008) (0.009)

0.040 0.037 0.036 0.030 0.036

(0.010)*** (0.010)*** (0.009)*** (0.009)*** (0.011)***

No Yes No Yes Yes

No No Yes Yes Yes

No No No No Yes

(0.012)*** (0.012)*** (0.011)*** (0.011)*** (0.011)***

0.039 0.037 0.036 0.031 0.032

(0.012)*** (0.011)*** (0.011)*** (0.011)*** (0.011)***

No Yes No Yes Yes

No No Yes Yes Yes

No No No No Yes

No Yes No Yes Yes

No No Yes Yes Yes

No No No No Yes

Promise Keeping

Representation of Interests

Support for Reconciliation

CDC

0.037 0.041 0.035 0.041 0.041

0.036 (0.011)*** 0.031 (0.011)*** 0.034 (0.010)*** 0.028 (0.011)*** 0.031 (0.012)** 0.039 (0.016)** 0.046 (0.016)*** 0.040 (0.016)** 0.048 (0.016)*** 0.042 (0.016)***

Customary

0.034 0.037 0.034 0.040 0.045

(0.012)*** (0.011)*** (0.011)*** (0.011)*** (0.013)***

0.432 0.438 0.442 0.449 0.432

(0.017)*** (0.017)*** (0.016)*** (0.016)*** (0.018)***

Obs. 8,314

8,336

8,099

7,787

7,724

Each row correspondents to a Logistic probability model where 0 is ‘‘not very or not at all important/satisfied/likely” and 1 is ‘‘somewhat, fairly or very important/satisfied/likely.” Due to the non-convergence of Logistic regressions with 862 fixed effects, the specifications that include settlement fixed effects are linear probability models (while still included, larger geographic fixed effects are subsumed in the settlement fixed effects). The omitted category is ELECT. Subject controls include indicator variables for gender, head of household, six age categories, 11 ethnicity categories, nine income categories, eight education categories, five religion categories, four marital status categories, seven work status categories, and 14 job-type categories. Geographic controls include individual indicator variables for seven regions, 34 provinces, and four degrees of urbanization (village, town, city, metro) as well as a provincial security measure based on ANSO data. Settlement fixed effects are separate fixed effects for each sampling point (village/neighborhood) from which respondents were selected; each sampling point/settlement has 10 interviews. All specifications include heteroskedasticity-robust standard errors which are clustered at the primary sampling unit (settlement). *, **, *** represent statistical significance on the 10%, 5%, and 1% significance level.

of interest when individuals are more supportive of elections in general. Panel A of Table 6 shows the results. Consistent with the transplant effect hypothesis, the consequences of using formal elections to establish local councils vary with norms and beliefs supportive of democracy. In particular, respondents who indicated that they support elections as a means to select politicians are significantly more likely to be satisfied with voting, more likely to believe that elections are important, that elected representatives will keep their promises and that politicians will be able to represent local interests when assigned to the formal elections treatment. The second way we measure these beliefs is by asking respondents to rate different ways of governing Afghanistan. Respondents could answer for each of the following questions ‘‘very good”, ‘‘fairly good”, ‘‘neither good nor bad”, ‘‘fairly bad”, or ‘‘bad”:  ‘‘How would you rank ‘strong leadership without holding elections’ as a way of governing Afghanistan?”  ‘‘How would you rank ‘having religious leaders rule’ as a way of governing Afghanistan?”  ‘‘How would you rank ‘having elders and tribal leaders rule’ as a way of governing Afghanistan?” These questions, respectively, gauge support for authoritarian rule, theocratic rule, and customary/traditional rule. The transplant effect suggests that preexisting beliefs supportive of these alternative modes of governance will influence the consequences of elections. Panels B, C, and D of Table 6

classifies respondents into whether they thought a specific way of governance was a ‘‘very good” or a ‘‘bad” approach for governance in Afghanistan. Our results give additional evidence in support of the transplant effect hypothesis for autocratic-leaning individuals but not for those that support a role for religious and elderly/tribal leaders. Panel B shows that respondents sympathetic to autocratic rule have significantly lower democratic perceptions under elections or formalization of CDCs (rows 4, 5) relative to respondents who dislike autocratic rule. Panels C and D, however, do not show any significant changes in democratic perceptions for religious or customary governance preferences. Table 7 investigates the possibility that the presence of customary institutions or a CDC influences the consequences of elections. Our hypothesis is that the presence of a shura or CDC will increase the positive effects of elections since the shura is a deliberative institution and CDCs are supposed to cultivate democratic norms and values. These institutions may thus serve important representative functions that increase the ability to transplant elections to rural villages even though the current form of local representation differs. In Panel A of Table 7, subjects are grouped by whether their settlement has a local shura or a CDC, while Panel B uses a ranking that subjects assigned to the accessibility of those local institutions. Since we are including CDCs as an explanatory variable, we only consider how perceptions of democracy change under the elections treatment. Panel A shows that past

306

WORLD DEVELOPMENT

Pashtun-majority provinces (in which much of the Taliban insurgency occurs). (c) Robustness

Figure 4. Sampling points and local violence. Violent incidences by district and sampling points. The shading indicates the number of violent incidences on the district level with a darker shading indicating greater levels of violence. Points indicate the locations of sampling points (villages or urban neighborhoods). The district violence measure was obtained by aggregating the GPS locations of 28,415 reported violent incidents during 2004–09 by district as obtained from the Afghan War Diary.

experiences with local councils significantly increases voter satisfaction (7.4%), the belief that elections are important (5.0%), that elected representatives will keep their promises (4.5%), and that they will be able to represent local interests (3.0%). These results hold while including a large set of subject- and geographic-specific controls. If experience with deliberative councils at the local level indeed improves receptivity to elections, then we would further expect to see greater improvements in democratic perceptions whenever individuals had more possibilities to interact with those councils. This is what we find in Panel B, where accessibility to local councils is ranked by subjects between 1 (not accessible at all) and 4 (very accessible). 23 Substantively, these findings suggest that the presence of and access to deliberative local institutions may influence whether elections are more likely to ‘‘stick.”. Table 8 further assesses the transplant effect hypothesis by separating respondents by social norms and beliefs that may influence the consequences of elections. To keep the tables readable, we report the differences in the joint effects between groups. 24 In Panels A and B of Table 8, respondents are separated by whether they are Pashtun or not, and whether they are likely to live under local Pashtun institutions or not. One possibility is that Pashtun tribal norms are inconsistent with democracy. Another is that they reinforce democracy due to their egalitarian nature. Rather than offer a firm hypothesis, which would require making strong generalizations about groups, we will simply use the statistical test to understand if there is an empirical relationship between these variables and support for democracy. The hypothesis regarding support for Taliban is less controversial, as we would expect more support for the Taliban to make transplanting either elections or CDCs to Afghanistan more challenging. Panel C groups respondents by how much confidence they have in the Taliban. The transplant effect suggests that social norms among Pashtuns and support for the Taliban influences the consequences of transplants such as Western-style elections. Consistent with the transplant effect hypothesis, we find that Pashtun respondents, respondents living in Pashtun-majority areas and respondents expressing confidence in the Taliban are less likely to favor elections. The effect becomes stronger for those who arguably have greater grievances with the state, in particular respondents sympathetic to the Taliban and for

(i) Multivariate results A concern for any experimental finding is whether respondents’ unobservables that are correlated to the outcome variable are orthogonal to the randomly assigned treatment. For example, if older respondents had a preference for customary selection methods over elections and if the average age of respondents was higher in the customary treatment, then this would lead to an upward bias in the preference for customary selection methods in the panels of Table 5. On the other hand, significant differences in unobservables would not be worrisome if they were uncorrelated to the treatment response. 25 To reduce the likelihood that our results in Table 5 are driven by any unobservables, we make use of a rich set of socioeconomic and geographic characteristics that we obtain from the demographic section of the survey. Specifically, each row in Table 9 represents a Logistic specification where the dependent variable is either 0 (representing ‘‘not very or not at all important/satisfied/likely”) or 1 (‘‘somewhat, fairly or very i mportant/satisfied/likely”). 26 We successively add subjectand geographic-related controls and present the marginal effects of the coefficients on the treatment indicator variables CDC and Customary (the omitted/baseline category is an election for the village council). In the specification with settlement fixed effects, each of the 862 sampling points (villages/ neighborhoods) in which interviews took place has its own separate intercept term. With those fixed effects included, the treatment coefficients are identified from within-settlement variation only (each consisting of 10 interviews) that is due to random treatment assignment. Thus any shared local characteristics of respondents in the same settlement is controlled for. A comparison of Tables 5 and 9 shows that the multivariate results are qualitatively identical to the univariate ones when including a large set of controls. 27 (ii) Response rate, safety concerns and other potential issues A related concern may arise from a non-random refusal rate to participate in the survey which would be problematic if the reason for refusal was correlated to the preference for a specific selection regime. 28 Importantly, while such a non-random refusal rate may lead to a level effect in the sample proportions of the treatments (Columns 1–3 in Table 5), the reason for refusal is likely orthogonal to the randomization of treatments, and thus there would be fewer of such households in all treatments equally. Consequently, the differences between treatments in Table 5 (columns 4–6) and the results in Table 9 remain unaffected. While we are foremost interested in the differences in democratic perceptions across selection methods, we also offer several responses to concerns about a level effect. First, our participation rate of 86.5% (cf., Table 3) is similar to the participation rates of other comparable studies and thus not suspiciously low. 29 Second, as Figure 4 illustrates, our sampling points are drawn from both secure and insecure areas and the refusal rate of 3.5% is uncorrelated to the provincial safety classification. Third, among those that did participate in the survey, we find that the refusal rates to a set of politically sensitive questions (in the non-experimental part of the survey) is not significantly correlated with several security measures. 30 This is also reflected in that relatively safe provinces such as Parwan (‘‘negligible” and ‘‘medium” risk) had high (>25%) non-response rates to some questions of the

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

non-experimental survey part while very unsafe provinces like Ghazni (‘‘extreme” and ‘‘very high” risk) or Zabul (‘‘high” and ‘‘very high” risk) had very low (below 5%) non-response rates to sensitive questions. 31 Finally, Table 9 also includes settlement (or sampling point) fixed effects; as a result, any shared level of security on the settlement level is controlled for and results remain largely unchanged. Finally, we applied a series of consistency checks in which conflicting responses may raise a flag about the interview, the interviewer or the sampling point. 32 Importantly, if the occurrence of such issues was orthogonal to the randomized treatment assignments, then they would occur in all treatments equally and again not affect the differences between treatments. We re-estimated our results after excluding questionable observations, and after excluding all observations from sampling points which had at least three such respondents; the results remain unchanged. 5. CONCLUSION State-building is not simply an effort to improve government capacity. Rather, these are often concerted attempts by the international community to also establish democracy in post-conflict states (Barnett & Zu¨rcher, 2009; Zu¨rcher, 2012). Yet in a crowded institutional environment populated by customary councils and development councils, elections are not the only, and not necessarily the first-best, option to establish representative institutions at the local level. One of the central policy questions in Afghanistan concerns customary councils. Despite their ongoing significance, we find that introducing formal village elections would improve outcomes associated with democracy and elections, including increasing satisfaction with voting, importance attached to elections, and the belief that elected representatives will keep their promises. However, formalizing customary councils would increase the belief that representatives will represent local interests and increase support for reconciliation with the Taliban. These contrasting findings caution against uncritical support for elections (as is sometimes the case in the literature on liberal peacebuilding), but also question whether formalization of selfgoverning institutions is the best option for Afghanistan. These findings underscore enduring tensions in peacebuilding literature and practice between the rush to establish procedural

307

democracy or a more gradual approach to institutional transformation as a pathway to peace. To the extent that policymakers are more concerned with improving support for democratic institutions, elections appear to be a better option than the formalization of customary councils. If establishing order is the most pressing concern, as is the case in Afghanistan, then customary councils may be the optimal choice. Another possibility in Afghanistan is to forgo elections for the foreseeable future. To the extent that self-governing councils are perceived as legitimate and effective, it may be better to wait to establish village government. The well-documented challenges of fraud associated with recent provincial- and national-level elections in Afghanistan call into question the feasibility of conducting village-level elections (whose numbers and boundaries have yet to be determined). Nor is it entirely clear the government could provide financial support to so many new local units of government. Thus, while elections may be the best option conditional upon proceeding with plans to establish village government, it is not entirely clear if the state will be able to implement these elections. While our experiment provides insight into the consequences to democratic and security perceptions of several methods for local council elections, more research is necessary to consider the benefits and costs of establishing village government. Our study also provides further reason to consider context in efforts to promote democracy. Social norms which are inconsistent with democracy often sidetrack state-building processes (Coyne, 2008). Broadly speaking, we find that democracy will face less friction in areas where individual beliefs and social norms support Western-style elections and where there are deliberative local institutions already in place. A more general implication is that efforts to promote democracy at the local level in post-conflict settings may introduce support for democracy, but may be relatively ineffective in building support for peace and reconciliation. The results from this experiment highlight potential trade-offs between democracy and reconciliation in Afghan villages. They also suggest possible sequencing of reforms. If building peace is a pathway to democracy, it may make sense to rely on customary councils to provide governance in the short term as the evidence here shows they may help facilitate peace, while planning on establishing formal elections in the long term when or if reconciliation between parties occurs.

NOTES 1. This survey experiment was funded by the United States Agency for International Development and implemented jointly with the authors by Democracy International, an American non-government organization. 2. ‘‘Government” refers to formal political institutions, while ‘‘governance” refers to the ‘‘study of good order and workable arrangements” (Williamson, 2005). 3. Donor-supported development councils such as the CDCs are not new to Afghanistan, as they have been implemented under various donor programs since the mid-1980s. However, the most recent CDCs are the first ones that are formally endorsed by the Afghan government and it is the first time that they are created on a nation-wide scale. 4. Despite three rounds of elections, democracy is far from institutionalized in the country. While Zahir Shah, the longest ruling Afghan monarch (1933–73), held parliamentary elections in 1964, his elections were largely pretense (Dupree, 1973). The country did not meaningfully

experience democracy until after 2001 (Coburn & Larson, 2014). However, each round was mired in controversy, including intimidation of voters and allegations of corruption (Wilder, 2005). The questionable quality of democracy at the national level, and the fact that there have never been elections for village councils, is why we view local elections as a transplant despite some experience with democracy in the country both in the post-2001 period and, to a much lesser degree, under Zahir Shah’s constitution. 5. The discussion of subnational governance in the 2004 constitution was based upon the 1964 model. 6. Article 140, Constitution of the Government of the Islamic Republic of Afghanistan of 2004. 7. Interview, Behsod District, Nangarhar Province, 2007. 8. Interview, Balkh Province, Nahri Shahi District, 2007.

308

WORLD DEVELOPMENT

9. We use the term ‘‘malik” generically even though this individual has the title of arbab, qaryadar, wakil, or namayenda depending on where the village is located in the country. Our survey accounted for this diverse terminology. 10. The Taliban attempted to increase the role of mullahs in village governance (Crews & Tarzi, 2009; Rashid, 2002). However, the decline of Taliban rule brought a return to the prominent role of the shura. During fieldwork, many villagers explained that they reconstituted their shura or began to select again a malik as Taliban influence subsided. 11. Interview, Bamiyan Center, Bamiyan Province, 2007. 12. Our discussion of CDCs focuses at the village level rather than at the district level even though MRRD also sought to cluster CDCs together in the hopes of scaling up their activities. These clusters of CDCs would then serve as the basis of district councils. As this particular vision of district government has not gained much traction in policy spheres, the experiment focuses on CDCs as an option for village government. 13. National Solidarity Program Operations Manual. 14. Interview, female teacher, Sorkhrod District, Nangarhar Province, 2007. 15. For a full description of the program see http://democracyinternational.com/projects/afghanistan-electoral-reform-and-civic-advocacy-project-aerca. 16. Indeed, informants rarely mentioned that they were aware of how CDC members were selected. Likewise, very few informants said that they had experienced an election for these councils. 17. It is important to note that the formalization of self-governing councils is state intervention in community affairs (Murtazashvili & Murtazashvili, 2015). Anarchy, which is defined by the absence of state intervention in community governance, may be better than state intervention when self-governance works well Leeson, 2006, 2013. In Afghanistan, another option—which we do not consider explicitly—is to simply allow villages to remain self-governed, rather than establishing a formal level of government at the village level. 18. Qawm, which is the fundamental category of social identity in Afghanistan, refers to one’s solidarity group in Afghan society (Roy, 1990). 19. Murtazashvili (2015) suggests that customary councils in Afghanistan have historically provided a source of security against predatory governments. 20. The hypothesis just mentioned comes with a caveat. Generalizing about ‘‘Pashtun norms” can be challenging in light of substantial variation among Pashtuns and across Pashtun communities Marten, Johnson, and Mason, 2009. Nonetheless, as there is a large literature that considers social identity as important variable in Afghanistan (Barfield, 2010; Rubin, 2002; Roy, 1990), we would be remiss if we did not at least consider the influence of Pashtun ethnicity on the impact of elections on individual perceptions. Even discussions of the influence of Taliban presence on transplantation of democracy are complicated in terms of expectations. Ruttig (2010) explains that it is often unclear if the Taliban are a Pashtun movement, a tribal movement, or a national movement.

Consequently, the same caveats for Pashtuns could apply to the Taliban regarding the potential conflict between Taliban norms and Western-style elections. 21. While there is borderline significant evidence that formalizing CDCs lead to improved beliefs relative to customary councils in Table 5, the significance is no longer present in later results (cf., Table 9). 22. The complete output is shown in Table A.2 in the internet appendix. 23. The exact magnitudes in Panel B should be interpreted with caution as the Logistic regression implicitly assumes rankings to be equidistant. 24. The complete output is provided in Table A.3. 25. With many observables, pure chance may still lead to significant differences among a few demographics despite randomization (type 1 error). Table A.1 provides balance tables with respondents’ demographics by treatment group. 26. The specifications with settlement point fixed effects is based on a linear probability model (OLS) as the Logistic regressions with 862 fixed effects fail to converge. 27. Table A.4 in the appendix provides results without post-stratification weights and Table A.5 provides multivariate results separated by ethnicity. 28. Selection issues can also be understood as an omitted variable bias problem Heckman, 1979. 29. For example, Lyall, Blair, and Imai (2013) obtain a 89% participation rate in their field survey in Afghanistan. 30. The measures include a security measure based on the province’s quartile among violent incidents in Afghanistan between 2004 and 2009 as disclosed in and geocoded from the Afghan War Diaries, a province-level security risk assessment by the U.N. from October 2010, and a provincelevel security risk measure from January 2012 based on the number of attacks on NGO workers throughout 2011 compiled by the Afghan NGO Safety Office. Those questions included the likelihood of continued fighting with the Taliban, the Taliban’s likely influence in five years, and the survival of the state. 31. There were several measures taken by the survey firm that helps explain the absence of a correlation between responses to sensitive questions and security, in particular that interviewers were careful to put interviewees at ease and autonomy to answer honestly, such as through same-gender interviewing, conducting interviews in separate rooms without family present, and promises of anonymity. 32. As an example, a household earning more than 40,000 Afghani per month (about 800 U.S. dollars as of November 2011) also claimed that it did not have enough food at least once in the prior year. Or, a respondent said it was ‘‘very good” to have strong leadership without holding elections and later also stated it was ‘‘very good” to have elections to choose national and local leaders. Questionable replies may occur due to a social desirability effect so to please the interviewer (e.g., claiming to vote in the next election while not doing so), or due to a subject-expectancy effect (e.g., claiming to be very dissatisfied with democracy in the hopes of receiving governmental or international aid).

ESTABLISHING LOCAL GOVERNMENT IN FRAGILE STATES: EXPERIMENTAL EVIDENCE FROM AFGHANISTAN

309

REFERENCES Acemoglu, D., Cantoni, D., J., S., & Robinson, J. A. (2011). The Consequences of radical reform: The French revolution. American Economic Review, 101(7), 3286–3307. Acemoglu, Daron, & Robinson, James A (2012). Why nations fail: The origins of power, prosperity, and poverty. New York: Crown. Acemoglu, D., & Jackson, M. O. (2015). History, expectations, and leadership in the evolution of social norms. The Review of Economic Studies, 82(2), 423–456. Acemoglu, D., Johnson, S., & Robinson, J. A. (2003). An African success story: Botswana. In D. Rodrik (Ed.), In search of prosperity: Analytical narratives on economic growth. Princeton, NJ: Princeton University Press. Acemoglu, D., Reed, T., & Robinson, J. A. (2014). Chiefs: Economic development and elite control of civil society in Sierra Leone. Journal of Political Economy, 122(2), 319–368. Affolter, F. W., Sawayz, S. Z., Noori, H., & Shrestha, B. S. (2006). Community elections in Rural Afghanistan. Central Asia and the Caucasus, 2, 75–90. Ahmed, A. S. (2013). The thistle and the drone: How America’s war on terror became a global war on Tribal Islam. Wathington, DC: Brookings Institution Press. Andrews, M. (2013). The limits of institutional reform in development. New York: Cambridge University Press. Baldwin, K. (2013). Why vote with the chief? Political connections and public goods provision in Zambia. American Journal of Political Science, 57(4), 794–809. Barfield, T. (2010). Afghanistan: A cultural and political history. Princeton, NJ: Princeton University Press. Barnett, M., & Zu¨rcher, C. (2009). The Peacebuilder’s contract: How external state-building reinforces weak statehood. In R. Paris, & T. D. Sisk (Eds.), The dilemmas of statebuilding: Confronting the contradictions of postwar peace operations (pp. 23–52). New York: Routledge. Barth, F. (1959). Political leadership among the Swat Pathans. New York: Athlone. Beath, A., Christia, F., Egorov, G., & Enikolopov, R. (2014). Electoral rules and political selection: Theory and evidence from a field experiment in Afghanistan. World Bank Policy Research Working Paper 7361. Beath, A., Christia, F., & Enikolopov, R. (2013). Do elected councils improve governance outcomes? Experimental evidence on local institutions in Afghanistan. World Bank Policy Research Working Paper 6510. Berkowitz, D., Pistor, K., & Richard, J.-F. (2001). Economic development, legality, and the transplant effect. European Economic Review, 47 (1), 165–195. Berkowitz, D., Pistor, K., & Richard, J.-F. (2003). The transplant effect. American Journal of Comparative Law, 51, 163–204. Bernard, T., Collion, M.-H., De Janvry, A., Rondot, P., & Sadoulet, E. (2008). Do village organizations make a difference in African rural development? A study for Senegal and Burkina Faso. World Development, 36(11), 2188–2204. Boettke, P. J., Coyne, C. J., & Leeson, P. T. (2008). Institutional stickiness and the new development economics. American Journal of Economics and Sociology, 67(2), 331–358. Boettke, P. J., Coyne, C. J., Leeson, P. T., & Sautet, F. (2005). The new comparative political economy. The Review of Austrian Economics, 18 (3–4), 281–304. Carbone, G., & Memoli, V. (2015). Does democratization foster state consolidation? Democratic rule, political order, and administrative capacity. Governance, 28, 5–24. Carter, L., & Connor, K. (1989). A preliminary investigation of contemporary Afghan councils. Peshawar: ACBAR. Coburn, N., & Larson, A. (2014). Derailing democracy in Afghanistan: Elections in an unstable political landscape. New York: Columbia University Press. Coyne, C. J. (2008). After war: The political economy of exporting democracy. Palo Alto, CA: Stanford University Press. Coyne, C. J., & Boettke, P. J. (2009). The problem of credible commitment in reconstruction. Journal of Institutional Economics, 5(1), 1–23. Crews, R. D., & Tarzi, A. (2009). The Taliban and the crisis of Afghanistan. Cambridge, MA: Harvard University Press.

De Figueiredo, R. J., & Weingast, B. R. (1997). Self-enforcing federalism: Solving the two fundamental dilemmas. Journal of Law, Economics, Organization, 21(1), 103–135. Dı´az-Cayeros, A., Magaloni, B., & Euler, A. R. (2014). Traditional government, citizen engagement, and the distribution of public goods: Local evidence from Mexico. World Development, 53(1), 80–93. Dupree, L. (1973). Afghanistan. New York: Oxford University Press. Ellickson, R. C. (1991). Order without law: How neighbors settle disputes. Cambridge, MA: Harvard University Press. Esser, D. (2013). The political economy of post-invasion Kabul, Afghanistan: Urban restructuring beyond the north–south divide. Urban Studies, 50(15), 3084–3098. Flores, T. E., & Nooruddin, I. (2012). The effect of elections on postconflict peace and reconstruction. The Journal of Politics, 1(1), 1–13. Fritzen, S. A. (2007). Can the design of community-driven development reduce the risk of elite capture? Evidence from Indonesia. World Development, 35(8), 1359–1375. Fukuyama, F. (2013). What is governance?. Governance, 26(3), 347–368. Gomes, J. F. (2015). The political economy of the Maoist conflict in India: an empirical analysis. World Development, 68, 96–123. Greif, A. (1993). Contract enforceability and economic institutions in early trade: The Maghribi Traders’ coalition. The American Economic Review, 83(3), 525–548. Greif, A. (2006). History lessons: The birth of impersonal exchange: The community responsibility system and impartial justice. The Journal of Economic Perspectives, 20(2), 221–236. Grindle, M. S. (2007). Good enough governance revisited. Development Policy Review, 25(5), 533–574. Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica, 47(1), 153–161. Izrael, D., Hoaglin, D. C. & Battaglia, M. P. (2004). To rake or not to rake is not the question anymore with the enhanced raking macro. Proceedings of the twenty-ninth annual SAS users group international conference. Johnson, T. H., & Mason, M. C. (2008). No sign until the burst of fire: Understanding the Pakistan–Afghanistan frontier. International Security, 32(4), 41–77. Joireman, S. F. (2008). The mystery of capital formation in Sub-Saharan Africa: Women, property rights and customary law. World Development, 36(7), 1233–1246. King, E., & Samii, C. (2014). Fast-track institution building in conflictaffected countries? Insights from recent field experiments. World Development, 64, 740–754. Leeson, P. T. (2006). Efficient anarchy. Public Choice, 130(1), 41–53. Leeson, P. T. (2007a). An-arrgh-chy: The law and economics of pirate organization. Journal of Political Economy, 115(6), 1049–1094. Leeson, P. T. (2007b). Better off stateless: Somalia before and after government collapse. Journal of Comparative Economics, 35(4), 689–710. Leeson, P. T. (2013). Anarchy unbound. New York: Cambridge University Press. Leeson, P. T., & Coyne, C. J. (2012). Sassywood. Journal of Comparative Economics, 40(4), 608–620. Lyall, J., Blair, G., & Imai, K. (2013). Explaining support for combatants during wartime: A survey experiment in Afghanistan. American Journal of Political Science, 107(4), 679–705. Mansuri, G., & Rao, V. (2012). Localizing development: Does participation work?. Washington, DC: World Bank. Marten, K., Johnson, T. H., & Mason, M. C. (2009). Misunderstanding Pakistan’s federally administered tribal area?. International Security, 33 (3), 180–189. Martinez-Bravo, M. (2014). The role of local officials in new democracies: Evidence from Indonesia. The American Economic Review, 104(4), 1244–1287. Menkhaus, Ken (2006). Governance without Government in Somalia spoilers, state building, and the politics of coping. International Security, 31(3), 74–106. Migdal, J. S. (2001). State in society: Studying how states and societies transform and constitute one another. New York: Cambridge University Press.

310

WORLD DEVELOPMENT

Miguel, E., & Gugerty, M. K. (2005). Ethnic diversity, social sanctions, and public goods in Kenya. Journal of Public Economics, 89(11), 2325–2368. Murtazashvili, I., & Murtazashvili, J. (2015). Anarchy, self-governance, and legal titling. Public Choice, 162, 287–305. Murtazashvili, J. (2015). Bad medicine: Explaining the failure of state building in Afghanistan. Central Asian Affairs, 2(1). Murtazashvili, J. (2016). Informal Order and the State in Afghanistan. New York: Cambridge University Press. Murtazashvili, J. B. (2014). Informal federalism: Self-governance and power sharing in Afghanistan. Publius: The Journal of Federalism, 44 (2), 324–343. Mutz, D. C. (2011). Population-based survey experiments. Princeton, NJ: Princeton University Press. Myerson, R. (2006). Federalism and incentives for success in democracy. Quarterly Journal of Political Science, 1(1), 3–23. Myerson, R., (2013). Standards for state-building. World Bank DEC Lecture. Myerson, R. (2014). Constitutional structures for a strong democracy: Considerations on the Government of Pakistan. World Development, 53, 46–54. National Solidarity Program. (2012). The National Solidarity Program (website). Newell, R. S. (1972). The politics of Afghanistan. Ithaca, NY: Cornell University Press. North, D. C., Weingast, B. R., & Wallis, J. (2009). Violence and social orders: A conceptual framework for interpreting recorded human history. New York: Cambridge University Press. Olken, B. A. (2010). Direct democracy and local public goods: Evidence from a field experiment in Indonesia. American Political Science Review, 104(2), 243–267. Ostrom, E. (1990). Governing the commons: The evoluation of institutions for collective action. New York: Cambridge University Press. Ostrom, E. (2005). Understanding institutional diversity. Princeton, NJ: Princeton University Press. Platteau, J.-P., & Gaspart, F. (2003). The risk of resource misappropriation in community-driven development. World Development, 31(10), 1687–1703. Rashid, A. (2002). Taliban: Militant Islam, oil and fundamentalism in Central Asia. New Haven, CT: Yale University Press. Rodrik, D. (2007). One economics, many recipes: Globalization, institutions and economic growth. Princeton, NJ: Princeton University Press. Roy, O. (1990). Islam and resistance in Afghanistan. New York: Cambridge University Press.

Rubin, B. R. (2002). The fragmentation of Afghanistan: State formation and collapse in the international system. New Haven, CT: Yale University Press. Ruttig, T. (2010). How Tribal are the Taleban. Kabul: Afghanistan Analysts Network. Scott, J. C. (1992). Domination and the arts of resistance: Hidden transcripts. New Haven, CT: Yale University Press. Seidler, V. (2014). When do institutional transfers work? The relation between institutions, culture and the transplant effect: The case of Bornu in North-Eastern Nigeria. Journal of Institutional Economics, 10 (3), 371–397. Sheely, R. (2015). Mobilization, participatory planning institutions, and elite capture: Evidence from a field experiment in Rural Kenya. World Development, 67, 251–266. Tripp, A. M. (2004). Women’s movements, customary law, and land rights in Africa: The case of Uganda. African Studies Quarterly, 7(4), 1–19. Wardak, A. (2004). Building a post-war justice system in Afghanistan. Crime, Law and Social Change, 41(4), 319–341. Weingast, B. R. (1997). The political foundations of democracy and the rule of law. American Political Science Review, 91(2), 245–263. Wilder, A. R. (2005). A house divided? Analysing the 2005 Afghan elections. Kabul: Afghanistan Research and Evaluation Unit. Williamson, O. E. (1991). Economic institutions: Spontaneous and intentional governance. Journal of Law, Economics, & Organization, 7, 159–187. Williamson, O. E. (2005). The economics of governance. American Economic Review, 95(2), 1–18. Zu¨rcher, C. (2012). Conflict, state fragility and aid effectiveness: Insights from Afghanistan. Conflict, Security and Development, 12(5), 461–480.

APPENDIX A. SUPPLEMENTARY DATA Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.worlddev.2015.08.025.

Available online at www.sciencedirect.com

ScienceDirect