TY - JOUR
AU1 - Malik, Adeel
AU2 - Mirza, Rinchan Ali
AB - Abstract This paper offers a novel illustration of the political economy of religion by examining the impact of religious elites on development. We compile a unique database on holy Muslim shrines across Pakistani Punjab and construct a historical panel of literacy spanning over a century (1901–2011). Using the 1977 military takeover as a universal shock that gave control over public goods to politicians, our difference-in-differences analysis shows that areas with a greater concentration of shrines experienced a substantially retarded growth in literacy after the coup. Our results suggest that the increase in average literacy rate would have been higher by 13% in the post-coup period in the absence of shrine influence. We directly address the selection concern that shrines might be situated in areas predisposed to lower literacy expansion. Finally, we argue that the coup devolved control over public goods to local politicians, and shrine elites, being more averse to education since it undermines their power, suppressed its expansion in shrine-dense areas. 1. Introduction Inspired by Max Weber’s seminal contribution on the relationship between Protestantism and prosperity, a niche literature has emerged on the impact of religion on economic outcomes, including human capital. While still described as a relatively “nascent” area of enquiry, the dominant focus of this expanding literature on the economics of religion has been on studying the broader impact of religiosity—see Iannaccone (1998) and Iyer (2016) for an exhaustive review. In this paper, we tackle a relatively understudied dimension: the impact of religious elites on economic development. The role of elites is central to understanding the evolution of institutions and their impact on development. Elite attitudes can determine whether growth promoting policies are permitted or hindered (North et al. 2009; Robinson 2012).1 Specifically, elites can oppose education or “block initiatives to increase the human capital of the masses” when it risks eroding their future political power (Acemoglu and Robinson 2006, p. 124). Despite the salient role accorded to elites in the dominant literature on political economy, we have little systematic knowledge of how the incentives and capacities of religious elites might shape development. This paper seeks to fill this gap by examining the impact of religious leaders on literacy in the world’s second most populous Muslim country, Pakistan. Our analysis highlights the role of religious elites whose power is historically embedded and derived from their association with holy Muslim shrines, which historians have described as the “symbolic cultural outposts of the power of Islam and the Muslim state” and are originally credited with the spread of Islam in South Asia (Gilmartin 1988). While largely organized as informal institutions and funded by the communities they serve, shrines dispense important welfare functions, which include free kitchens for the poor, healing for the sick, and resolution of local disputes. Together with their sacred lineage, this allows shrine elites to enjoy tremendous religious legitimacy and unquestionable obedience and devotion of the vast networks of followers. With their religious authority, shrine leaders have become important controllers of economic and political power and serve as crucial intermediaries between the ruler and the ruled. Courted and rewarded by every ruler from the pre-colonial period to present-day Pakistan, shrine elites are structurally positioned within the prevailing political economy structures and can therefore profoundly influence local development. In this paper, we focus on the impact of shrine elites on literacy. This is inspired by a wealth of qualitative literature that shows how the religious authority of shrine leaders is directly threatened by the spread of mass schooling. Education strikes at the very heart of the regime of voluntary obedience that shrine guardians seek to maintain, and can compromise their ability to control the underlying economic and political structure. An advantage of focusing on shrine elites relative to other elites (e.g. landed elites) who might be similarly incentivized to obstruct schooling is that the former have a greater ability to shape cognitive processes, moral perceptions, and beliefs, and thus exert a deeper and more durable form of control over local populations. Compared with elites whose power solely rests on land, shrine elites have both greater aversion and capacity to oppose education. To investigate the impact of shrine elites on literacy, we study a rare encounter between historically embedded religious power and formal institutional change. Specifically, we use a major policy shock induced by the military coup in 1977 that brought General Zia-ul-Haq to power. After assuming power, General Zia radically altered the system of public goods provision by reassigning the control over public development spending from centralized bureaucrats to elected representatives. While this was a universal policy shock that affected all regions, we hypothesize that, given the greater aversion of shrine elites to educational expansion, the effect of this shock should be more pronounced in regions where shrine elites have historically enjoyed greater influence. To conduct the empirical analysis, we assembled a rich dataset on the number and presence of shrines using hitherto untapped primary archival sources, principally the colonial-era British District Gazetteers and the Department of Religious Endowments at the Government of Punjab. This is the first and most comprehensive data collection effort on Muslim religious shrines to date. We then use the number of historically recognized shrines per thousand persons to proxy for the religious power of shrine elites, measured at the tehsil level, the lowest administrative unit for which literacy data are available. Our empirical strategy exploits two important sources of variation: cross-sectional variation in the number of historically important shrines per thousand persons and temporal variation arising from the coup-induced administrative shock to public goods provision. Combining these two sources of variation in a difference-in-differences (DID) framework allows us to investigate whether the impact of the coup-induced administrative shock on literacy was modulated by the strength of historically embedded religious power.2 We show that growth in literacy was slower in regions where the number of historically important shrines per thousand persons was higher (relative to regions where it was lower) after the coup-induced administrative shock (relative to before).3 Concretely, our results show that the increase in average literacy rate in Punjab would have been higher by 13% in the post-coup period in the absence of shrine influence. In interpreting these results, we emphasize that the impact of historically embedded religious elites is only activated after the coup-induced shock to public goods provision in 1985 when elected politicians, including shrine elites, were given direct control over development funds earmarked for their constituencies. With their greater aversion to education, shrine elites were now better able to thwart literacy expansion. We shed light on the precise mechanisms through which this happened and investigate both supply- and demand-side explanations. We rule out the obvious possibility that shrine elites might have opposed educational expansion by restraining public spending on education or opposing school construction. Using a dataset on the dispersal of development funds in national electoral constituencies of Punjab, we find no evidence that shrine-linked legislators systematically under-utilized development funds in their constituencies. We also present evidence that shrine-dominated regions did not have systematically lower supply of schools or teachers. Instead, shrine elites undermined education by making school provision defective. Drawing on several auxiliary datasets, we show that shrine elites manipulated school provision in three ways: limiting physical access to schools, compromising the quality of school infrastructure, and inefficiently dispersing school resources. Finally, testing for potential demand-side explanations, we show that neither urbanization nor economic migration appear as primary mechanisms for our results. In carrying out our empirical analysis, we take several steps that are standard in the literature. First, we control for unobserved time-invariant heterogeneity and period-specific shocks by including in our specifications both tehsil and year fixed effects. Next, the historical literature on the diffusion of Islam in the subcontinent suggests certain geographic characteristics that were crucial to the locational decision of shrines and could have been correlated with long-run development (Eaton 2009). Related work has identified the large-scale displacement caused by the Partition of British India as another key determinant of long-term trends in literacy across the subcontinent (Bharadwaj and Mirza 2019; Mirza 2020). Taking cue from these literatures, we use distance to Delhi (the capital of historical Delhi Sultanate, 1206–1526), Mughal-era land classifications, and partition-induced displacement as baseline controls in our analysis. Our identification strategy is based on the assumption that, conditional on our baseline controls, the interaction between the timing of the coup-induced administrative reforms and shrine presence is exogenous with respect to literacy. However, we recognize potential concerns with such an assumption. First, regions with historically recognized shrines could have been suffering from lower literacy levels even before the introduction of coup-induced reforms and, at the same time, be the primary drivers of the coup itself. If this were the case, then our identification assumption would be violated and the negative effect of shrines on literacy that we document cannot be attributed to the coup-induced policy shock activating the power of shrine elites. To rule out this concern, we show that there were no systematic trends in literacy prior to the reforms between high- and low-shrine regions. Even if shrine-dominated regions were not on a differential trajectory in terms of literacy prior to the coup-induced administrative restructuring, they could still have differed along other dimensions that matter for literacy. Whilst limited in our ability to perform a similar test for parallel trends in other relevant dimensions due to data constraints, we are at least able to show that there are no levels differences between high- and low-shrine regions for a range of public infrastructure variables that are important for literacy prior to the coup. Finally, the presence of shrines could still somehow be generally related to other tehsil-level characteristics that might have a bearing on how the coup-induced reforms influenced literacy. Accordingly, we account for other powerful historical explanations for Punjab’s under-development, including colonial-era land revenue, initial urbanization rate, and the presence of Christian missions. Our results remain highly robust to the inclusion of these additional controls. To demonstrate that our results highlight the impact of religious elites rather than the generalized impact of religiosity, we carry out a placebo exercise by replacing our historic shrines measure with a variable covering the universe of shrines in Punjab. We fail to find any significant impact of this all-shrine measure. Besides clarifying the impact of religious elites, this result helps to rule out any remaining selection concern over and above the dimensions, we already account for through our baseline controls. This is because if there was a systemic selection on location then it should equally apply to all shrines, not just shrines recognized by the colonial administration. Finally, we carry out a battery of robustness checks and show that our main findings remain robust to including alternative measures of shrine presence, removing outliers, using alternative fixed effects, and restricting the sample to the post-partition period. We also conduct sensitivity analysis for any remaining concern of omitted variables. Implementing the tests proposed by Altonji, Elder, and Taber (2005) and Oster (2019), we show that the selection on unobservables needs to be more than two and a half times stronger than the selection on observable characteristics to drive our estimated shrine effect to zero. Given its focus on the impact of religious elites on literacy, this paper contributes most directly to a growingly influential literature on the economics and politics of religion. An exhaustive review of this literature is offered by Iannaccone (1998), Iyer (2016), Carvalho, Iyer, and Rubin (2019), and Kuran (2018). Specifically, our work is situated at the intersection of two important fields of enquiry that Iyer (2016) described in her wide-ranging review as warranting further scholarly attention. The first pertains to the relationship between religion and human capital and the second to “the whole field of religion, political processes, and their interaction with economic processes” (Iyer 2016, p. 432). While a growing body of scholarship has studied the first dimension by studying the long-run impact of religious missions on human capital (Woodberry 2004; Becker and Woessmann 2009; Gallego and Woodberry 2010; Nunn 2010; Wantchekon, Klašnja, and Novta 2014; Valencia Caicedo 2019), this literature has primarily focused on non-Muslim contexts. A partial exception is Chaudhary and Rubin (2011), who show how early religious and political institutions shaped subsequent differences in human capital formation in colonial India. By contrast, our paper focuses on the attitudes of Muslim religious elites towards education and offers one of the first empirical attempts at uncovering the impact of historically embedded religious power on the evolution of literacy.4 The second area, the relationship between religion, politics, and economic processes, is particularly under-studied in the context of Muslim societies. While past studies have examined the impact of the Hajj, Islam’s global congregation, on social attitudes of Muslim believers (Clingingsmith, Khwaja, and Kremer 2009), the analysis of veiling by female Muslims (Carvalho 2013), and the impact of Ramadan fasting on economic outcomes and subjective well-being (Campante and Yanagizawa-Drott 2015), the relationship between Islam and political economy has only recently gained prominence (Platteau 2017; Rubin 2017). Using centuries of flood data on river Nile, Chaney (2013) showed that Egypt’s central religious authority assumed greater political power in periods of economic shocks. Similarly, Bazzi, Koehler-Derrick, and Marx (2020) utilize the Indonesian land reforms of 1960, which exempted religious lands from expropriation, and show that regions that faced a greater threat of land expropriation witnessed higher transfers of land by rural elites into Islamic charitable endowments and subsequently resulted in stronger electoral support for Islamic political parties. This literature has primarily focused on two aspects of organized religion: religious clergy (Ulama) and Islamic political parties. Our paper broaches a third important yet neglected dimension: the impact of Islamic mystical tradition (Sufism) on long-run development. This article also relates to a broader literature that has studied elite opposition to education in the context of high land inequality (Sokoloff and Engerman 2000; Galor, Moav, and Vollrath 2009; Cinnirella and Hornung 2016). Relative to this literature on land inequality and education, we emphasize the religious as opposed to an economic motive and offer more precise evidence on the strategy used by these elites to restrain literacy. Our work also connects with the literature on informal institutions and development (Platteau 2000; Casson, Della Giusta, and Kambhampati 2010). While past work focuses on culture, norms and social capital (Guiso, Sapienza, and Zingales 2006; Tabellini 2010; Qian et al. 2015), the role of informal religious authority and its interplay with formal power structures has been largely neglected.5 Finally, our paper complements an expanding literature on the role of history for explaining economic development (Nunn 2009). The challenge of “persistence studies is to figure out why and how history matters” (Fouka 2020). We do not merely claim that shrines, which were set-up centuries ago and have historically shaped religious authority of their guardians, still exert an impact on literacy but clearly identify “when” such an effect should have occurred and offer a plausible conceptual explanation for “why” it occurred. This brings our work in line with an emerging strand of scholarship on “time-varying persistence”, the idea that the impact of a historical variable can remain latent for a long time until it interacts with other factors to shape contemporary outcomes (Fouka and Voth 2013; Belmonte and Rochlitz 2019; Ochsner and Roesel 2019). The rest of this paper proceeds as follows. Section 2 sets out the essential background for our empirical analysis. Section 3 outlines the empirical strategy, Section 4 describes the data, and results are presented in Section 5. Finally, Section 6 discusses plausible mechanisms for our results, and Section 7 concludes the paper. 2. Historical Background This section provides a background on three critical dimensions of our analysis: the historical placement of shrines and the persistent power of their guardians, their relevance for literacy, and the coup-induced administrative restructuring of public goods provision. 2.1. The Placement and Power of Shrines The Placement of Shrines The spread of Islam in South Asia occurred between the 13th and 18th centuries. Islamic mystics (Sufis) served as the primary conduits of religious transmission. As Gilmartin (1988, p. 40) notes, “many rural Punjabi tribes have traced their conversion [to Islam] to these medieval saints”. Travelling from central Asia, Afghanistan, and the Middle East, Sufis settled in all regions of Punjab, whether prosperous or poor. While the historical literature suggests that the spiritual diktats and demands of a mystical order typically determined where these saints finally settled (Nizami 1953), the gradual diffusion of Islam was “embedded in larger socio-economic, indeed environmental, transformations” (Eaton 2009, p. 217). Studying the patterns of early Muslim conversions in Punjab, eminent historian Richard Eaton argues that the spread of Islam coincided with “the transition to, or intensification of, settled agriculture” and “this transformation was conceptually linked with the pioneers or shrines through whose agency it had occurred” (Eaton 2009, p. 216). In Eaton’s argument, shrines and cultivators were both crucial aspects of Islamization. Aiding this process of Islamization and the associated shift from pastoral to settled agriculture were two crucial factors: political geography and hydrology. In terms of political geography, areas initially considered as frontier or “marginal” to Delhi Sultanate provided an important base for Sufi preaching and early religious conversions. As Eaton (2009, p. 208) notes: “The Islamization of western Punjab between the thirteenth and eighteenth centuries” took place in regions that were “located along the margins of expanding, Delhi-based empires”. An important reason why frontier regions were more propitious for the diffusion of Islam lied in the fact that they were home to large “indigenous populations that were not well-incorporated into a Hindu socio-religious world at the time of their earliest contact with Muslims” (Eaton 2009, p. 208). These were typically clans consisting of pastoral nomads who moved northwards along the Indus valley to sparsely populated areas that were agriculturally fertile but hitherto uncultivated. These areas were effectively outer spaces separating populated areas and wild forests that lied between rivers forming “a natural barrier between two different human settlements” (Soofi 2014).6 As these nomadic clans “adopted field agriculture, they also generally acquired a Muslim identity” (Eaton 2009, p. 211). An enabling factor for both religious conversions and the feasibility of agriculture was hydrology. In Punjab water has long been recognized as the principal constraining factor for agriculture. As the 1921 Census of India observed: “Cultivation in the Punjab is affected more by rainfall and irrigation than by difference in the soil....”7 In this milieu, riverine tracts had some of the most conducive agricultural conditions, especially “when wells were few and canals not at all, the low-lying lands along the river were best, and greedily seized by the invader” (Lyall 1928, p. 63). It was partly for their superior agricultural potential that the riverine regions attracted many earlier Muslim saints who preferred to depend on independent income streams that agriculture afforded. Access to river also expanded the spiritual constituency for saints by affording greater mobility of people, especially in an age when other means of communication were under-developed, and key river crossing points served as logistic nodes. It is therefore unsurprising that some of the oldest and most prominent shrines are “often located by or near the great rivers of Punjab” (Eaton 2009, p. 208). As Lyall (1928, p. 62) notes, “it is a curious feature of riverain tracts that they are mainly inhabited by Muhammedans”. In her famous travelogue on the Indus river, whose northern parts and tributaries lie in Punjab, Alice Albinia dedicated a chapter on “River Saints” in which she describes how, in the popular imaginary, the “power of Islam” coexists with the “power of the Indus” in such profound ways that the river serves as a “silent protagonist” of the teachings of proselytising saints (Albinia 2008, pp. 89, 107). To summarize, shrines of historic significance were more likely to be situated in frontier regions that were home to pastoral nomads who were yet unattached to the Hindu cultural system and lacked an established social hierarchy. These were also regions where the potential for agriculture was higher either through proximity to the river or through other means of access to water. Our empirical analysis will systematically account for these dimensions that could have a bearing on shrine location and at the same time influence patterns of long-run development. The Power of Shrine Elites The religious authority of shrine elites is primarily derived from their guardianship of the shrine and their sacred lineage with founding saints. Shrine elites are distinguished by their control over symbolic power that helps to create patterns of voluntary obedience and compliance among local populations.8 This has important implications for power relations. In their landmark work, Character and Social Structure: The Psychology of Social Institutions (Mills and Gerth 1953, pp. 193–5), they emphasize the role of such obedience in power relations by arguing that: the crux of the problem of power rests in understanding the origin, constitution, and maintenance of voluntary obedience ... An adequate understanding of power relations thus involves a knowledge of the grounds on which a power holder claims obedience, and the terms in which the obedient feels an obligation to obey. Shrine elites thus control the social structure at a much deeper level through their influence on mental constructs, cognitive capacities, and the popular understanding of religious beliefs. The religious authority that such elites derive from their affiliation with the shrine and their control over local populations has made them important allies for all centralized rulers who have turned to them for legitimacy and support and offered material rewards, in return. In fact, “[N]o major ruler passed by the area without showing deference to their “spiritual power” (Eaton 1984, p. 347). However such deference and rewards became more systematized under British rule when Muslim religious elites, such as Syeds, Sheikhs, and Qureshis, were considered as “agrarian castes” and became fit to receive landed gentry grants. While land grants by Mughal rulers could easily revert to the throne upon the saint’s death they were preserved under British rule, thanks to the establishment of formal property rights.9 This made the hereditary succession of shrine-based religious authority a powerful economic proposition and converted shrine elites into both “spiritual and feudal masters”, appropriately termed pir-zamindars. Instances of colonial patronage to shrine guardians, which included both permanent land grants and other forms of appeasement (e.g. honours, titles, and appointments) are extensively documented in the historical literature (see Online Appendix A). Over time, shrine elites were able to use their religious authority to harness, not only economic but also political power. In fact, historians have long argued that the religious and political authority of shrine elites is co-constituted (Gilmartin 1979; Ansari 1992). A vast network of devotees, considerable wealth, and a superior ability to mobilize meant that Punjab’s shrine elites possessed the three crucial elements of political power: number, resources, and organization. It is thus unsurprising that when the British introduced electoral politics in early 20th century shrine elites became natural contenders for political power and have since protected their political turf, surviving through both military regimes and civilian governments in Pakistan.10Online Appendix Table A.1 provides a snapshot of the political persistence of leading shrine families. To illustrate the consolidated power base of shrine elites, Figure 1 charts the genealogy of Syed Yusuf Raza Gilani, former Prime Minister of Pakistan and a leading shrine aristocrat. As Figure 1 shows, the power of the Gilani clan is historically embedded. Several members of the Gilani clan were appointed to key positions during the British era, including membership of district boards, municipal committees, judiciary, and legislative assemblies. Since then, the Gilanis have enjoyed an enduring electoral presence. The Gilanis are, by no means, an isolated example but broadly reflective of the persistent power of historic shrine families (see Online Appendix Table A.1).11 Figure 1. Open in new tabDownload slide Genealogy of a shrine elite: Syed Yusuf Raza Gilani. Difference from Landed Elites and Implications for Literacy While political genealogies of shrine elites might be similar to those of landed elites (see Online Appendix Figure A.1), their power is differentially configured in ways that translates into their stronger aversion to mass education and greater capacity to oppose literacy expansion. Relative to landed elites, shrine elites have an additional source of power, which we have described earlier as symbolic power. At the very heart of this power are hierarchical structures based on obedience, unquestioned loyalty, and superstition. Mass education can undermine these structures and challenge the very foundation of the religious authority upon which the political and economic power of shrine leaders is based. Any improvement in human capital thus poses a more fundamental threat to shrine elites compared to other elites whose opposition to education derives mainly from economic concerns.12 Plentiful references in both historical and contemporary literatures highlight the particularly strong aversion of shrine elites to education. As Sir Malcolm Darling observed in The Punjab Peasant, the Pirs are “instinctively opposed to the two movements from which the ordinary cultivator has most to hope. Neither education nor cooperation has their sympathy, for both strike at the regime [of voluntary obedience] which it is their one object to maintain” (Lyall 1928, p. 100). Echoing the same concern, Anatol Lieven highlights how educational progress could “strike directly at the cultural and social bases” of shrine power (Lieven 2012, p. 138). To historian, K. K. Aziz, shrine opposition to education is unsurprising. “How could a pir”, he asks, “countenance any prospect for the education of the masses when his supremacy, status and income depended on their ignorance and superstition?” (Aziz 2001, p. 27). Besides their greater aversion, shrine elites also possess stronger capacity to oppose schooling. First, shrine elites are able to combine instruments of control typically deployed by traditional elites—coercion and co-option—with obedience and compliance. This can result in a voluntary suspension of agency that makes collective action around common interest resources, such as education, even less likely in a milieu where shrine leaders enjoy greater bargaining power with respect to their constituents and are subjected to fewer sanctions given their religious status. Second, shrine elites enjoy distinct political advantages over landed elites. The religious network sustained through the relationship of piri-mureedi (master–disciple) and the unconditional allegiance of devotees allows shrine leaders to enjoy a more stable constituency of followers—a captive vote bank of sorts—that makes them electable even in an uncertain political game. Compared to a typical landed aristocrat, shrine elites are also more accessible to ordinary people. Third, as a fixed institution offering a “semi-permanent family seat” and durable religious legitimacy, shrines provide their guardians a more stabilizing source of power. Shrine families are less prone to inheritance-related subdivision of property, since the gaddi (seat of religious power) can provide an important “safeguard against dilution of landed power” (Malik and Malik 2017). Fourth, shrine leaders operate as a more cohesive group as the ties of shrine brotherhood (pir-bhai) provide a stronger basis for the emergence and consolidation of collective interests.13 Finally, given their religious authority, shrine elites are also more effective legitimizing agents for political rule. This makes them more eligible to receive state patronage and affords them greater ability to mobilize other elite segments, including landlords, who often seek their divine intercession and are usually dependent on the shrine guardian’s support for winning an election. Thus, even when they do not hold legislative power, shrine elites enjoy an indirect influence over other landed, political, and bureaucratic elites. Such differences mean that shrine elites can be more effective at restraining literacy. 2.2. The 1977 Coup and the Policy Shock Governing Public Goods A third key element of our analysis is the coup-induced policy shock that brought a radical change to the administrative regime governing public goods provision. Following civil disorder, General Muhammad Zia-ul Haq deposed the democratically elected government through a military coup in 1977. Having safely ensconced himself in power, General Zia succumbed to the international demands for restoring parliamentary democracy and held nationwide elections under a non-party based system. In a radical shift from the past, elected representatives were given direct control over development funds for public goods in 1985, which were previously determined largely by bureaucrats. In principle, these funds could be allocated to each member of the legislature in equal amounts. While the resources are not directly placed in the hands of elected representatives, the latter could identify development schemes in their constituencies and determine the corresponding resource allocation.14 The overall spending mandate is broad, covering such diverse items as water supply and electrification, with education being a central component. The administrative shake-up in 1985 gave elected representatives a controlling stake over development spending. Apart from indirectly deriving rents through collusion with contractors, elected members now exerted greater influence on shaping the composition of spending (e.g. how much is to be spent on schools versus roads or school construction and salaries versus quality) and flow of resources in their respective constituencies. Importantly, they could influence the quality of infrastructure provision, appointments of school teachers, and decide where schools were to be built. The decade of 1980s witnessed an unprecedented number of new educational initiatives and a noticeable increase in public spending on education (see Figure 2 and Online Appendix B). Consequently, there was a major expansion in school construction and education departments became the single largest employer at the provincial level. Figure 2. Open in new tabDownload slide Educational policies and spending in Pakistan, 1952–1998. The figure, obtained from Mirza (2020), displays the evolution of public expenditure per reference population for Pakistan during the period, 1960–1990. Reference population consists of individuals that fall in the relevant school-going age. Each bullet point represents a new educational initiative. The Figure shows a sharp spike in expenditures during the 1980s and a larger number of educational policies (as evidenced by the clustering of bullet points) in 1980s. The policy shift in mid-1980s converted education into an important arena of patronage for politicians and led to a significant “institutionalization of patron–client relationship”, especially with regard to public goods provision (Wilder 1999). As Gazdar (2000) notes, “the entry of MNAs and MPAs as local power brokers in the mid-1980s” marked a qualitative shift for public education.15 While previously the district education officer (DEO) enjoyed greater “discretionary (and rent-earning) powers, his or her political influence and power was much smaller than that wielded, subsequently, by the local MPA or MNA” and there were greater constraints on the “abuse of discretionary powers” (Gazdar 2000, p. 30). An insider account of the decline of Pakistan’s bureaucracy by Mufti (2020) clearly suggests that bureaucrats were relatively insulated from local politicians in the pre-Zia period. Major decisions on public goods provision were made more centrally under the direction of the Planning Commission of Pakistan through its Five-Year Plans (see Online Appendix B). With politicians gaining direct control over pubic resources, the educational machinery became susceptible to elite capture, especially in a milieu where, according to Martin (2015, p.4), “control over the state apparatus” is crucial for traditional elites’ “strategies of accumulation and dominance”. In short, the coup-induced administrative shock to the system of public goods provision represented a sharp policy discontinuity. Three aspects of this policy shock are especially relevant for our empirical strategy. First, both the 1977 military coup and the resulting policy change were exogenous to the power of shrine elites. The coup was the outcome of a complex interplay between geo-strategic influences and domestic power struggles. Neither did shrine families drive the coup nor was the coup influenced by economic characteristics or the potential for literacy expansion in constituencies that later elected shrine families into politics. The coup was preceded by public mobilization organized by Pakistan National Alliance (PNA), a coalition of ethnic and religious parties who were united in their opposition to Mr Zulfiqar Ali Bhutto, the elected Prime Minister deposed through the military coup. The PNA mobilized urban street power and was principally supported by urban traders. Shrine guardians were not the principal driving force for this protest movement. Second, institutional changes in the organization of public goods provision were universal and not targeted towards shrine families alone. Third, the new administrative regime has continued despite General Zia’s departure and the resumption of democracy in 1988.16 3. Data 3.1. Informal Religious Power Historically Recognized Shrines To construct our proxy for religious power, we gathered information on historically important shrines from the colonial district gazetteers. Issued periodically during the colonial era, the gazetteers contained vital information on the name and location of major shrines, their guardians, and the spiritual and material influence of shrine families.17 While shrines are a ubiquituous feature of Punjab’s sacred geography, their influence varies considerably across regions with some areas known for highly influential shrines, while others have shrines “dedicated to lesser known saints” that “had only the most localized significance” (Gilmartin 1988, p. 41). Online Appendix C.1 provides further evidence from colonial-era district gazetteers of the varying significance of shrines across regions. The Gazetteers thus offer a critical historical resource as no prominent shrine with pre-colonial influence would have missed a mention. To construct a tehsil-level measure of the concentration of historic shrines, we use the number of shrines (mentioned in District Gazetteers) per thousand persons in 1931, the census year for which consistent and comprehensive information was available across all tehsils: $$\begin{eqnarray} \rm{Shrines}_{i}^{\rm{hist}}=\frac{\rm{No. of shrines in tehsil ({\textit{i}}) in 1931}}{\rm{No. of persons in tehsil ({\textit{i}}) in 1931}} \times 1,000. \end{eqnarray}$$(1) In our robustness analysis, we will use two alternative measures: the number of historic shrines per square mile and a high shrine dummy. The former is defined as the number of historic shrines per square miles in 1931, while the high shrine dummy is coded as one for all tehsils where historic shrines per thousand persons is above the median of the overall sample distribution in 1931. Auqaf and Political Shrines We supplement the Gazetteer data with an exhaustive compilation of the names and locations of the universe of shrines across Punjab. In this regard, we gathered shrine lists maintained by the Punjab Auqaf Department and conducted specialized interviews with informed respondents in each district circle to verify the names and locations of shrines. Correspondingly, we construct an all-shrines measure that is defined as the number of all shrines (whether historically recognised or not) per thousand persons in a tehsil. We use this as an alternative measure of shrine concentration in our auxiliary regressions. Finally, we compiled a third database that mapped the direct participation of shrine families in electoral politics using multiple waves of election results since 1937. To develop this fine-grained database of shrines–politics linkage, we draw on electoral records from several sources (see Online Appendix C.2 for further details). Taken together, these three complementary databases provide a comprehensive coverage of shrines across all tehsils of Punjab and constitute the most systematic data effort on shrines to date. Overall, there are a total of 598 shrines in our database. We identify 146 of these as historically significant (i.e. mentioned in Punjab District Gazetteers). Our main variable of interest will be the measure of historically recognized shrines per thousand persons. A quick glance at Figure 3 shows substantial variation in our measure. While Ahmadpur East has the highest share of historic shrines, other tehsils have fewer or no shrines recognized in Gazetteers (e.g. Gujrat and Narowal). We will exploit this spatial variation in historic shrine density to learn about the evolution of literacy before and after the coup-induced shock. Figure 3. Open in new tabDownload slide Proportion of historically recognized shrines (by tehsil). The figure plots the number of historically recognized shrines per thousand persons for tehsils of Punjab. Tehsil names are mentioned on the horizontal axis. 3.2. Literacy Data on literacy for our long panel come from the following three sources: (a) census reports for colonial Punjab; (b) district census reports for Pakistan; and (c) Multiple Indicator Cluster Survey (MICS). Our study includes ten time periods spanning over a century and cover the period, 1901–2011. The pre-independence data on literacy are obtained from decadal census reports of colonial Punjab, covering the years 1901, 1911, 1921, and 1931. Colonial census reports contain information on literacy by sex and religion at the tehsil level. Data on post-independence literacy are obtained from the district census reports of Pakistan and the MICS, and covers the following years: 1961, 1972, 1981, 1998, 2008, and 2011.18 Data for 1961–1998 are obtained from district census reports and are supplemented with tehsil-level data on literacy from MICS for 2008 and 2011. All of the above data sources use a minimum agreeable definition of literacy. Literacy is defined as being able to read and write in any language.19 Although the definition of a literate person did not change during the colonial period, slight modifications were made after independence. These minor definitional changes should be captured by the inclusion of year fixed effects in our specifications. To arrive at a consistent balanced panel, we impute literacy data for a few tehsil-years (17 data points in total) that were missing in the first three census years. Our procedure is inspired by Bharadwaj, Khwaja, and Mian (2008, 2015) and described in Online Appendix D. Later on in the empirical analysis, we will show the robustness of our results to excluding these imputed observations. An empirical challenge in using historical data on literacy is that such data are available at the level of colonial tehsils. Over time, some tehsil boundaries have undergone periodic changes due to administrative restructuring. We therefore map current tehsils onto historical boundaries. To do so, we fix tehsil boundaries at 1931, the census year for which comprehensive historical data are consistently made available for the largest number of tehsils. In the spirit of Bharadwaj, Khwaja, and Mian (2008), we then follow a two-step mapping procedure that is described in detail in Online Appendix D. In our empirical analysis, we also draw on supplementary sources to construct measures of gross enrolment, school size, physical access to schools, and school infrastructure. Using the annual reports of Punjab Development Statistics, we constructed a panel of gross enrolment rates and school size for the period 1982–98 (annual data are only available for this period). The latter is defined as the number of pupils normalized by number of schools in a tehsil.20 Next, we used household data from MICS 2007-8 to construct tehsil-level indicators of distance to primary schools. We construct three indicators based on the proportion of households in a tehsil reporting their nearest primary school as being less than 2 km away, within the 2–5 km range, and greater than 5 km. Finally, using recent monitoring data on school facilities, we construct tehsil-level measures for the proportion of schools without boundary walls and electricity.21 3.3. Descriptive Statistics Table 1 presents basic summary statistics for key variables used in our analysis. Our main sample consists of 68 tehsils (defined as per colonial administrative boundaries). Average total literacy rate during the full period of our panel, 1901–2011, is 21.4%, with female literacy significantly lower at 7%. Noticeably, there is significant temporal variation in literacy, especially after 1980s. The major literacy expansion since the 1980s provides an important backdrop for this paper. The total literacy rate in the post-shock period is 52% compared to 8% during the 1901–1981 period (see Table 1). Exploiting these patterns of variation over time and across tehsils, we will explore whether high-shrine regions witnessed a differential increase in literacy after the coup-induced shock. Figure 4 shows the spatial distribution of historic shrines per thousand persons and provides a parallel view of average literacy rate for the post-shock period (1998–2011). A quick glance at Punjab’s sacred geography shows that shrines are more concentrated in the south, with significant inter-tehsil variation (also see Online Appendix Figure A.3). The two maps considered together suggest that districts with the highest shrines per thousand persons (darker shades in top panel) also have the lowest average post-shock literacy rate (lighter shades in bottom panel). Figure 4. Open in new tabDownload slide Historic shrines and literacy: heterogeneity. The top and bottom panels of the above figure show the spatial variation in historic shrines per capita and average post-coup literacy, respectively. The Quantile Method in QGIS was used for classifying the data into different bins. This method sorts the data from the smallest to the largest value and then determines bins such that number of values in each bin is the same. Table 1. Summary statistics. . (1) . (2) . (3) . (4) . (5) . . Mean . Standard deviation . Min . Max . N . Shrine concentration (1931) Number of shrines |$\times$| 1,000/tehsil population 0.015 0.024 0 0.110 68 Number of shrines/tehsil area 0.002 0.003 0 0.017 68 High shrine dummy 0.500 0.504 0 1 68 Number of auqaf shrines |$\times$| 1,000/tehsil population 0.003 0.005 0 0.028 68 School size (1982–1999) Average pupils per primary school 77.168 27.746 24.169 187.5 924 Average pupils per middle school 233.160 64.263 62.5 529.851 924 Average pupils per high school 562.668 132.278 181.818 1580.873 924 School access (2007) Proportion of boys public schools less than 2 km 0.928 0.078 0.497 1 123 Proportion of boys public schools from 2 to 5 km 0.051 0.049 0 0.216 123 Proportion of boys public schools greater than 5 km 0.021 0.044 0 0.379 123 Proportion of girls public schools less than 2 km 0.910 0.105 0.421 1 123 Proportion of girls public schools from 2 to 5 km 0.056 0.059 0 0.273 123 Proportion of girls public schools greater than 5 km 0.034 0.065 0 0.486 123 Proportion of boys private schools less than 2 km 0.779 0.238 0.102 1 123 Proportion of boys private schools from 2 to 5 km 0.079 0.089 0 0.394 123 Proportion of boys private schools greater than 5 km 0.143 0.177 0 0.780 123 Proportion of girls private schools less than 2 km 0.778 0.240 0.102 1 123 Proportion of girls private schools from 2 to 5 km 0.078 0.089 0 0.389 123 Proportion of girls private schools greater than 5 km 0.145 0.180 0 0.780 123 Literacy rates Average total literacy rate (1901–2011) 0.214 0.225 0.013 0.814 680 Average total literacy rate (1901–1981) 0.082 0.069 0.013 0.365 476 Average total literacy rate (1998–2011) 0.522 0.149 0.157 0.814 204 Average male literacy rate (1901–2011) 0.180 0.181 0.024 0.841 457 Average male literacy rate (1901–1981) 0.117 0.091 0.024 0.449 389 Average male literacy rate (1998–2011) 0.539 0.143 0.224 0.841 68 Average female literacy rate (1901–2011) 0.069 0.117 0.000 0.638 457 Average female literacy rate (1901–1981) 0.030 0.045 0.000 0.284 389 Average female literacy rate (1998–2011) 0.297 0.142 0.075 0.638 68 Baseline controls High rainfall dummy (1931) 0.426 0.498 0 1.000 68 Proportion nahri (1931) 0.360 0.337 0 0.993 68 Proportion abi (1931) 0.006 0.010 0 0.043 68 Proportion chahi (1931) 0.231 0.226 0 0.804 68 Proportion sailabi (1931) 0.091 0.101 0 0.509 68 Proportion barani (1931) 0.312 0.374 0 0.997 68 Proportion minorities (1931) 0.121 0.047 0.023 0.276 68 Distance to Delhi (km) 414.146 93.963 249.166 629.725 68 Additional controls Distance to river (km) 20.706 18.686 1 77 68 Urbanization (1931) 0.095 0.109 0 0.647 68 Land revenue per capita (1901) 1.859 0.854 0.306 4.801 68 . (1) . (2) . (3) . (4) . (5) . . Mean . Standard deviation . Min . Max . N . Shrine concentration (1931) Number of shrines |$\times$| 1,000/tehsil population 0.015 0.024 0 0.110 68 Number of shrines/tehsil area 0.002 0.003 0 0.017 68 High shrine dummy 0.500 0.504 0 1 68 Number of auqaf shrines |$\times$| 1,000/tehsil population 0.003 0.005 0 0.028 68 School size (1982–1999) Average pupils per primary school 77.168 27.746 24.169 187.5 924 Average pupils per middle school 233.160 64.263 62.5 529.851 924 Average pupils per high school 562.668 132.278 181.818 1580.873 924 School access (2007) Proportion of boys public schools less than 2 km 0.928 0.078 0.497 1 123 Proportion of boys public schools from 2 to 5 km 0.051 0.049 0 0.216 123 Proportion of boys public schools greater than 5 km 0.021 0.044 0 0.379 123 Proportion of girls public schools less than 2 km 0.910 0.105 0.421 1 123 Proportion of girls public schools from 2 to 5 km 0.056 0.059 0 0.273 123 Proportion of girls public schools greater than 5 km 0.034 0.065 0 0.486 123 Proportion of boys private schools less than 2 km 0.779 0.238 0.102 1 123 Proportion of boys private schools from 2 to 5 km 0.079 0.089 0 0.394 123 Proportion of boys private schools greater than 5 km 0.143 0.177 0 0.780 123 Proportion of girls private schools less than 2 km 0.778 0.240 0.102 1 123 Proportion of girls private schools from 2 to 5 km 0.078 0.089 0 0.389 123 Proportion of girls private schools greater than 5 km 0.145 0.180 0 0.780 123 Literacy rates Average total literacy rate (1901–2011) 0.214 0.225 0.013 0.814 680 Average total literacy rate (1901–1981) 0.082 0.069 0.013 0.365 476 Average total literacy rate (1998–2011) 0.522 0.149 0.157 0.814 204 Average male literacy rate (1901–2011) 0.180 0.181 0.024 0.841 457 Average male literacy rate (1901–1981) 0.117 0.091 0.024 0.449 389 Average male literacy rate (1998–2011) 0.539 0.143 0.224 0.841 68 Average female literacy rate (1901–2011) 0.069 0.117 0.000 0.638 457 Average female literacy rate (1901–1981) 0.030 0.045 0.000 0.284 389 Average female literacy rate (1998–2011) 0.297 0.142 0.075 0.638 68 Baseline controls High rainfall dummy (1931) 0.426 0.498 0 1.000 68 Proportion nahri (1931) 0.360 0.337 0 0.993 68 Proportion abi (1931) 0.006 0.010 0 0.043 68 Proportion chahi (1931) 0.231 0.226 0 0.804 68 Proportion sailabi (1931) 0.091 0.101 0 0.509 68 Proportion barani (1931) 0.312 0.374 0 0.997 68 Proportion minorities (1931) 0.121 0.047 0.023 0.276 68 Distance to Delhi (km) 414.146 93.963 249.166 629.725 68 Additional controls Distance to river (km) 20.706 18.686 1 77 68 Urbanization (1931) 0.095 0.109 0 0.647 68 Land revenue per capita (1901) 1.859 0.854 0.306 4.801 68 Open in new tab Table 1. Summary statistics. . (1) . (2) . (3) . (4) . (5) . . Mean . Standard deviation . Min . Max . N . Shrine concentration (1931) Number of shrines |$\times$| 1,000/tehsil population 0.015 0.024 0 0.110 68 Number of shrines/tehsil area 0.002 0.003 0 0.017 68 High shrine dummy 0.500 0.504 0 1 68 Number of auqaf shrines |$\times$| 1,000/tehsil population 0.003 0.005 0 0.028 68 School size (1982–1999) Average pupils per primary school 77.168 27.746 24.169 187.5 924 Average pupils per middle school 233.160 64.263 62.5 529.851 924 Average pupils per high school 562.668 132.278 181.818 1580.873 924 School access (2007) Proportion of boys public schools less than 2 km 0.928 0.078 0.497 1 123 Proportion of boys public schools from 2 to 5 km 0.051 0.049 0 0.216 123 Proportion of boys public schools greater than 5 km 0.021 0.044 0 0.379 123 Proportion of girls public schools less than 2 km 0.910 0.105 0.421 1 123 Proportion of girls public schools from 2 to 5 km 0.056 0.059 0 0.273 123 Proportion of girls public schools greater than 5 km 0.034 0.065 0 0.486 123 Proportion of boys private schools less than 2 km 0.779 0.238 0.102 1 123 Proportion of boys private schools from 2 to 5 km 0.079 0.089 0 0.394 123 Proportion of boys private schools greater than 5 km 0.143 0.177 0 0.780 123 Proportion of girls private schools less than 2 km 0.778 0.240 0.102 1 123 Proportion of girls private schools from 2 to 5 km 0.078 0.089 0 0.389 123 Proportion of girls private schools greater than 5 km 0.145 0.180 0 0.780 123 Literacy rates Average total literacy rate (1901–2011) 0.214 0.225 0.013 0.814 680 Average total literacy rate (1901–1981) 0.082 0.069 0.013 0.365 476 Average total literacy rate (1998–2011) 0.522 0.149 0.157 0.814 204 Average male literacy rate (1901–2011) 0.180 0.181 0.024 0.841 457 Average male literacy rate (1901–1981) 0.117 0.091 0.024 0.449 389 Average male literacy rate (1998–2011) 0.539 0.143 0.224 0.841 68 Average female literacy rate (1901–2011) 0.069 0.117 0.000 0.638 457 Average female literacy rate (1901–1981) 0.030 0.045 0.000 0.284 389 Average female literacy rate (1998–2011) 0.297 0.142 0.075 0.638 68 Baseline controls High rainfall dummy (1931) 0.426 0.498 0 1.000 68 Proportion nahri (1931) 0.360 0.337 0 0.993 68 Proportion abi (1931) 0.006 0.010 0 0.043 68 Proportion chahi (1931) 0.231 0.226 0 0.804 68 Proportion sailabi (1931) 0.091 0.101 0 0.509 68 Proportion barani (1931) 0.312 0.374 0 0.997 68 Proportion minorities (1931) 0.121 0.047 0.023 0.276 68 Distance to Delhi (km) 414.146 93.963 249.166 629.725 68 Additional controls Distance to river (km) 20.706 18.686 1 77 68 Urbanization (1931) 0.095 0.109 0 0.647 68 Land revenue per capita (1901) 1.859 0.854 0.306 4.801 68 . (1) . (2) . (3) . (4) . (5) . . Mean . Standard deviation . Min . Max . N . Shrine concentration (1931) Number of shrines |$\times$| 1,000/tehsil population 0.015 0.024 0 0.110 68 Number of shrines/tehsil area 0.002 0.003 0 0.017 68 High shrine dummy 0.500 0.504 0 1 68 Number of auqaf shrines |$\times$| 1,000/tehsil population 0.003 0.005 0 0.028 68 School size (1982–1999) Average pupils per primary school 77.168 27.746 24.169 187.5 924 Average pupils per middle school 233.160 64.263 62.5 529.851 924 Average pupils per high school 562.668 132.278 181.818 1580.873 924 School access (2007) Proportion of boys public schools less than 2 km 0.928 0.078 0.497 1 123 Proportion of boys public schools from 2 to 5 km 0.051 0.049 0 0.216 123 Proportion of boys public schools greater than 5 km 0.021 0.044 0 0.379 123 Proportion of girls public schools less than 2 km 0.910 0.105 0.421 1 123 Proportion of girls public schools from 2 to 5 km 0.056 0.059 0 0.273 123 Proportion of girls public schools greater than 5 km 0.034 0.065 0 0.486 123 Proportion of boys private schools less than 2 km 0.779 0.238 0.102 1 123 Proportion of boys private schools from 2 to 5 km 0.079 0.089 0 0.394 123 Proportion of boys private schools greater than 5 km 0.143 0.177 0 0.780 123 Proportion of girls private schools less than 2 km 0.778 0.240 0.102 1 123 Proportion of girls private schools from 2 to 5 km 0.078 0.089 0 0.389 123 Proportion of girls private schools greater than 5 km 0.145 0.180 0 0.780 123 Literacy rates Average total literacy rate (1901–2011) 0.214 0.225 0.013 0.814 680 Average total literacy rate (1901–1981) 0.082 0.069 0.013 0.365 476 Average total literacy rate (1998–2011) 0.522 0.149 0.157 0.814 204 Average male literacy rate (1901–2011) 0.180 0.181 0.024 0.841 457 Average male literacy rate (1901–1981) 0.117 0.091 0.024 0.449 389 Average male literacy rate (1998–2011) 0.539 0.143 0.224 0.841 68 Average female literacy rate (1901–2011) 0.069 0.117 0.000 0.638 457 Average female literacy rate (1901–1981) 0.030 0.045 0.000 0.284 389 Average female literacy rate (1998–2011) 0.297 0.142 0.075 0.638 68 Baseline controls High rainfall dummy (1931) 0.426 0.498 0 1.000 68 Proportion nahri (1931) 0.360 0.337 0 0.993 68 Proportion abi (1931) 0.006 0.010 0 0.043 68 Proportion chahi (1931) 0.231 0.226 0 0.804 68 Proportion sailabi (1931) 0.091 0.101 0 0.509 68 Proportion barani (1931) 0.312 0.374 0 0.997 68 Proportion minorities (1931) 0.121 0.047 0.023 0.276 68 Distance to Delhi (km) 414.146 93.963 249.166 629.725 68 Additional controls Distance to river (km) 20.706 18.686 1 77 68 Urbanization (1931) 0.095 0.109 0 0.647 68 Land revenue per capita (1901) 1.859 0.854 0.306 4.801 68 Open in new tab Additionally, Figure 5 charts the evolution of literacy rates across high- and low-shrine tehsils, defined on the basis of the 95th percentile of the distribution of our historic shrines measure. Revealingly, even without any controls, this crude visual representation shows the emergence of a noticeable gap between literacy rates of the two types of tehsils after the coup-induced shock in 1985. In analysing Figure 5, it is important to recognize that there is a possibility of a trend break in the data between the pre- and post-partition periods. This is because the partition represented a shock along several dimensions (including changes in data quality). Accordingly, if we restrict attention to just the post-partition period, two clear observations can be made. The mean literacy rate in high-shrine regions closely tracks that of low-shrine regions in the pre-shock period. There is also a clear divergence between the high- and low-shrine tehsils in the post-shock period, indicating a more retarded growth of literacy in shrine-dense regions. These findings are corroborated by our formal empirical analysis, which shows that, conditional on baseline controls including the impact of partition, there is no evidence of any pre-trends. Figure 5. Open in new tabDownload slide Evolution of literacy in high- and low-shrine regions. The figure plots mean literacy rates for high- and low-shrine tehsils over time. The tehsil classifications are based on the 95th percentile of the distribution of historic shrines per capita. The dashed lines for 1947 and 1985 signify the partition of Indian subcontinent and the coup-induced policy shock, respectively. 4. Empirical Strategy To empirically examine the impact of shrine elites on literacy after the coup-induced administrative shock, we set out a DID specification: $$\begin{eqnarray} {\rm{{Lit}}}_{it}=\beta ({\rm{Shrines}}_{i}^{\rm{hist}} \times {\rm{Post}}_{t}) + \gamma (x_{i}^{\prime } \times {\rm{Year}}_{t}) + \delta _{i} + \eta _{t} + \epsilon _{it}. \end{eqnarray}$$(2) Here, |${\rm{{Lit}}}_{it}$| is the literacy ratio in tehsil i in year t. |$(\rm{Shrines}_{i}^{\rm{hist}} \times {\rm{Post}}_{t})$| is the interaction between the number of shrines per thousand persons in tehsil i and a post-shock dummy variable that takes a value 1 for years after the coup-induced administrative shock in 1985 and 0 otherwise. |$(x_{i}^{\prime } \cdot {\rm{Year}}_{t})$| is the interaction between a vector of initial tehsil-level characteristics and the year fixed effects. Importantly, these characteristics include amongst others a set of baseline controls to be discussed later.22 To control for any unobservable time-invariant differences across tehsils, we include tehsil fixed effects |$(\delta _{i})$|⁠, and to account for any perturbations that apply equally to all tehsils in a given year, we include year fixed effects |$(\eta _{t})$|⁠. We also cluster standard errors by tehsil. Our coefficient of interest is |$\beta$|⁠, which measures the observed change in literacy in tehsils that had greater concentration of shrines (relative to those that had lower concentration) after the coup-induced administrative shock (relative to before). Our identification strategy rests on the assumption that, conditional on the set of baseline controls, the interaction effect |$(\rm{ Shrines}_{i}^{\rm{hist}} \times \rm{Post}_{t})$| is exogenous with respect to literacy. Several challenges could arise in this regard. First, shrine dense areas could be on a different trajectory in terms of their literacy prior to the coup-induced shock and also be the main catalyst behind the coup itself. Second, even if shrine-dominated areas were not trending differently in terms of their literacy prior to the coup they could still have differed along some other dimension that mattered for literacy. Finally, there could be other tehsil characteristics not included in our baseline controls that could somehow be generally related to shrine location and also shape the relationship between the coup-induced shock and literacy. To address the first challenge, we estimate a fully flexible specification that allows us to investigate trends in literacy in tehsils with greater concentration of shrines relative to tehsils with lower concentration of shrines prior to the coup-induced policy shock: $$\begin{eqnarray} \rm{Lit}_{it}=\Gamma _{t} (\rm{Shrine}_{i}^{\rm{hist}} \times \rm{Year}_{t}) + \Pi _{t} (x_{i}^{\prime } \times \rm{Year}_{t}) + \delta _{i} + \eta _{t} + \epsilon _{it}. \end{eqnarray}$$(3) The main difference from equation (2) corresponds to the term |$(\rm{Shrine}_{i}^{\rm{hist}} \times \rm{Year}_{t})$|⁠, whereby our historic shrine measure is interacted with the full set of year fixed effects. The vector of estimated interaction coefficients, |$\Gamma _{t}$|⁠, shows the relationship between shrine concentration and literacy in each time-period of our panel. If, for instance, areas with a greater shrine concentration were not on a different trend in terms of their literacy prior to the administrative shock, then we would expect the coefficients to be more or less constant over time for the years prior to 1985. Moreover, if shrines were to adversely influence literacy after the policy shock, then we would expect the coefficients to become more negative as we move further into the post-shock period.23 We address the second threat to our identification assumption by testing for levels differences in public infrastructure that matter for literacy between high- and low-shrine tehsils prior to the administrative shock.24 Finally, we address the third challenge to our identification strategy by including interactions between additional tehsil-level characteristics, which we believe are more generally related to shrine location and could have a direct impact on literacy, and the year fixed effects. These characteristics are included alongside the baseline controls in the term, |$(x_{i}^{\prime } \times \rm{Year}_{t})$|⁠, in equations (1) and (2). Given that our estimation is not based on a structural model, we interpret our main result as a “reduced-form” relationship between historically embedded religious power and literacy in the post-shock period. We do, however, provide specific evidence on the pathways through which informal religious power might have shaped the impact of the coup-induced policy shock on literacy. 5. Results In this section, we first provide a detailed motivation of our baseline controls whose inclusion in the main specification forms an important part of our identification strategy. Next, we discuss our main results that show the average effect of historic shrine concentration on post-coup literacy. We then present results from estimating a fully flexible specification that tests for differential trends in literacy between high- and low-shrine tehsils prior to the coup. In addition to testing for pre-trends in literacy, we also test for pre-coup differences in infrastructure related to literacy between the two types of regions. 5.1. Baseline Controls In Section 2.1, we highlighted that historic shrines were more likely to be located in areas that were initially considered marginal to the Delhi Sultanate and had higher agricultural potential. To account for the former, we construct a tehsil-level measure of distance from Delhi, defined as the shortest straight line distance of the centroid of each tehsil in our dataset to Delhi, and include its interaction with year fixed effects in our analysis. Additionally, given the importance of hydrology in determining initial agricultural potential across different regions of Punjab, we construct a comprehensive set of land classifications based on water access regimes and include their interactions with year fixed effects in our specifications. Developed during the Mughal period under its ruler Akbar (1556–1605), these classifications categorize cultivable land based on access to water (Grover 1960). They include cultivable land as being dependent on: rainfall (Barani), flood, and river inundation (Sailabi), water courses from streams and ponds (Abi), wells (Chahi), and canal irrigation (Nahri). Using data from colonial-era District Gazetteers, which retain these land classifications, we construct separate measures for the proportion of total cultivable area classified as falling in one of the above five categories. An advantage of including the measure of canal irrigated land (Nahri) in our baseline controls is that we are able to account for one of the most profound historical legacies of colonial Punjab, commonly known as the canal colonization. The British Empire built the world’s largest canal irrigation scheme in Punjab during the late 19th century. As part of this scheme, dense networks of perennial canals were laid down across the plains of the province, which made cultivation possible in areas that were hitherto considered as agriculturally insecure or pure waste lands. Newly irrigated lands were distributed to upper segments of the agrarian hierarchy who were considered loyal to colonial rulers. Thus, canal colonization not only expanded agricultural possibilities but also influenced the initial distribution of economic and political power (Ali 2014). If this was somehow related to the historical power base of shrines and simultaneously determined the evolution of literacy, then its omission could result in a spurious correlation between our measure of shrine density and literacy. Next, we control for rainfall by including the interaction between a dummy variable for high rainfall regions, which is equal to one for tehsils where average annual rainfall between 1917 and 1923 lied above the sample mean, and year fixed effects. The inclusion of a rainfall dummy as a baseline control is motivated by the strong and well-established association between patterns of rainfall and land inequality in Punjab. In one of the earliest official enquiries into land inequality in the province, Calvert (1925, p. 15) observed that an important “factor determining the size and distribution of holdings in the Punjab” was rainfall. While data on contemporary tehsil-level measures of land inequality are unavailable in Pakistan, it is possible to correlate our rainfall measure with a general measure of landlessness, defined as the proportion of National Identity Card holders who self-classify themselves as landless tenants (muzaara).25 As Figure 6 shows, average rainfall between 1917 and 1923 is strongly and negatively correlated with landlessness in 2011, providing suggestive evidence that high rainfall regions have lower incidence of landlessness. Figure 6. Open in new tabDownload slide Rainfall and landlessness. The figure offers a simple scatter plot of the proportion of landless tenants against average historical rainfall (1917–1923). The former is constructed as the proportion of National Identity Card holders in a tehsil who self-classify themselves as landless peasants. Rainfall data are obtained from colonial-era District Gazetteers of Punjab. The sample consists of 68 historical tehsils. Finally, another important confounding factor in our analysis is partition-induced displacement. Previous research shows that the mass population displacement triggered by the partition of British India in 1947 left an enduring legacy for economic development in the post-independence era (Bharadwaj, Khwaja, and Mian 2008, 2015; Bharadwaj and Mirza 2019; Mirza 2020). If high-shrine regions also witnessed greater displacement at partition, then this could bias our results. To preclude this possibility, we flexibly control for the impact of displacement at partition using the pre-1947 proportion of minorities. Our measure is motivated by prior evidence that population outflows were largely determined by the “relative importance of minorities” whereby “places with greater minorities saw greater outflows”, a pattern that was particularly strong in Pakistan where the “exchange was almost one to one” (Bharadwaj, Khwaja, and Mian 2008, p. 11). 5.2. Baseline Estimates Table 2 presents the results for our main specification in equation (2). In column (1), we estimate a parsimonious model that only includes tehsil and year fixed effects without any additional controls. The estimated coefficient of interaction between our historic shrine measure and the post-shock dummy is negative and statistically significant at 1% level. This provides strong initial evidence that tehsils with higher shrines per thousand persons experienced smaller increases in literacy (relative to tehsils with fewer shrines) after the coup-induced administrative shock (relative to before). Next, in column (2), we add a powerful set of baseline controls to ensure that the shrine effect is not confounded by differences in political geography, hydrology, and partition-induced displacement over the same time-period. Although the coefficient on shrine interaction term is halved in magnitude, it remains strongly negative and statistically significant at 1% level. In columns (3)–(5), we progressively add interactions of a series of additional tehsil-level characteristics with year fixed effects. First, we flexibly control for a measure of the distance of a tehsil from the nearest river (in km) in column (3), which can be correlated with shrine location (see Section 2.1). The inclusion of distance from river further strengthens our set of hydrology-related controls (i.e. agricultural land classifications and rainfall). Table 2. Historical shrines and total literacy. . (1) . (2) . (3) . (4) . (5) . . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** (0.382) (0.324) (0.341) (0.344) (0.370) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorites (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 . (1) . (2) . (3) . (4) . (5) . . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** (0.382) (0.324) (0.341) (0.344) (0.370) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorites (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 Notes: Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The post-coup indicator variable equals 0 for the period 1901–1981 and 1 for the period 1998–2011. All regressions include year fixed effects and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. **Significant at 5%; ***Significant at 1%. Open in new tab Table 2. Historical shrines and total literacy. . (1) . (2) . (3) . (4) . (5) . . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** (0.382) (0.324) (0.341) (0.344) (0.370) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorites (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 . (1) . (2) . (3) . (4) . (5) . . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** (0.382) (0.324) (0.341) (0.344) (0.370) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorites (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 Notes: Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The post-coup indicator variable equals 0 for the period 1901–1981 and 1 for the period 1998–2011. All regressions include year fixed effects and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. **Significant at 5%; ***Significant at 1%. Open in new tab Next, in column (4), we introduce a historic measure of urbanization based on data from colonial District Gazetteers and defined as the proportion of population classified as urban in 1931. This should help to account for the fact that initial differences in urbanization could be correlated with historic shrine location and affect literacy over time. As we will argue later, initial urbanization is also a powerful correlate of partition-induced population flows. Finally, in column (5), we include a measure of land revenue per capita in 1901. Land revenue was a major source of income for the colonial administration and was strongly influenced by underlying land tenurial arrangements. It can thus serve as a proxy for initial tehsil-level differences in state capacity and land inequality. As the results show, despite the inclusion of these additional controls in columns (3)–(5), the estimated coefficients on the shrine interaction remain stable, consistently negative, and statistically significant at least 5% level across all columns. The effect is also economically meaningful. According to the estimate in column (5), a 10 percentage point increase in our shrines measure is associated with a reduction in the literacy rate of 0.0957.26 This is equivalent to 18% of the average post-shock literacy rate.27 To further reinforce the magnitudes of our estimates, we carry out a simple calculation to determine the extent to which the increase in the mean literacy of Punjab during the post-shock period was restrained by our shrine measure. We start by noting that the observed increase in average literacy of Punjab during the post-shock period was 0.15, increasing from 0.420 in 1998 to 0.574 in 2011. We want to compare these numbers to what the counterfactual post-shock increase in mean literacy would have been in the absence of shrines. In order to do so, we first calculate the counterfactual literacy rate for each tehsil in 2011 using the estimate reported in column (5) of Table 2. This is simply the tehsil’s observed literacy rate in 2011 minus the estimated impact of shrines, |$\beta$|⁠, multiplied by the number of shrines per thousand persons in the tehsil |$(\rm{Lit}_{i,2011} - \beta \cdot \rm{ Shrines}_{i}^{\rm{hist}})$|⁠. Next, we compute the average counterfactual literacy of Punjab in 2011 using the counterfactual literacy rate of each tehsil in 2011. Based on our calculations, the counterfactual mean literacy of Punjab in 2011 would have been 0.589 (instead of 0.574) in the absence of shrines. Moreover, the increase in the Punjab’s literacy would have been 0.169 (instead of 0.150) without the restraining influence of shrines. As a consequence, the increase in mean literacy of Punjab would have been 13% higher than the actual observed increase had the impact of shrines not been there to begin with. Overall, our results provide clear evidence that the expansion of literacy that occurred after the coup-induced administrative shock was considerably more restrained in areas with greater concentration of historic shrines. Test for Selection on Unobservables While our baseline specification accounts for a wide range of tehsil-level characteristics, we also conduct sensitivity analysis to any remaining omitted variable bias. Following the tests proposed by Altonji, Elder, and Taber (2005) and Oster (2019), we calculate the bounds that show the robustness of the shrine effect to unobservable characteristics under the assumption that the selection on unobservables is proportional to the selection on observed controls. Results for this exercise are reported in Table 3. Columns (1) and (2) report the main shrine effect from a regression with baseline controls (column (2), Table 2) and additional controls (column (5), Table 2). Column (3) reports the bounded estimates for the shrine effect using the procedure recommended by Oster (2019). The upper bound is based on the assumption that the selection on observables is 0 (⁠|$\delta = 0$|⁠), while the lower bound is estimated under the scenario that the selection on observable characteristics is equal to the selection on unobservables (⁠|$\delta = 1$|⁠). The latter is a particularly reasonable assumption in specifications with high R-squared. The bounds are calculated using the more conservative threshold of |$R_{\rm{ max}}$| of 1 (see Altonji, Elder, and Taber 2005). As column (3) shows, the identified set consists of negative values and excludes zero. Finally, column (4) reports the estimated value for |$\delta$|⁠, which provides an indication of the relative degree of selection that would be needed to eliminate the result. Reassuringly, we find that the selection on unobservables has to be around two and a half times stronger than the selection on observables in order to drive away our estimated shrine effect to zero. Taken together, this sensitivity analysis shows that any omitted variable concern in our analysis is likely to be modest, at best. Table 3. The role of selection on unobservables. . (1) . (2) . (3) . (4) . . Total literacy rate (Rmax |$=$| 1) . Treatment variable . Baseline effect (standard error), [R2] . Controlled effect (standard error), [R2] . Identified set . |$\tilde{\delta }$| for |$\beta =$| 0 Given |$R_{max}$| . Post-coup |$\times$| shrines per thousand persons -0.96511*** (0.324) [0.978] -0.95730*** (0.370) [0.979] [|$-0.95730,-$|0.75865] 2.52225 . (1) . (2) . (3) . (4) . . Total literacy rate (Rmax |$=$| 1) . Treatment variable . Baseline effect (standard error), [R2] . Controlled effect (standard error), [R2] . Identified set . |$\tilde{\delta }$| for |$\beta =$| 0 Given |$R_{max}$| . Post-coup |$\times$| shrines per thousand persons -0.96511*** (0.324) [0.978] -0.95730*** (0.370) [0.979] [|$-0.95730,-$|0.75865] 2.52225 Notes: This table validates our results for the impact of historic shrines on literacy. Baseline effect includes the full set of baseline controls in column (2) of Table 2. Controlled effect includes the full set of controls (baseline |$+$| additional) in column (5) of Table 2. The identified set in column 3 is bounded below by |$\beta ^{\sim}$| and above by |$\beta$|*, calculated based on Rmax given in the top row and |$\delta =$| 1. Column (4) shows the value of |$\delta ^{\sim} $| which would produce |$\beta =$| 0, given the value of Rmax reported in the top row. Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The post-coup indicator variable equals 0 for the period 1901–1981 and one for the period 1998–2011. All regressions include year fixed effcets and tehsil fixed effects. Each control variable is interacted with the full set of year fixed effects. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.***Significant at 1%. Open in new tab Table 3. The role of selection on unobservables. . (1) . (2) . (3) . (4) . . Total literacy rate (Rmax |$=$| 1) . Treatment variable . Baseline effect (standard error), [R2] . Controlled effect (standard error), [R2] . Identified set . |$\tilde{\delta }$| for |$\beta =$| 0 Given |$R_{max}$| . Post-coup |$\times$| shrines per thousand persons -0.96511*** (0.324) [0.978] -0.95730*** (0.370) [0.979] [|$-0.95730,-$|0.75865] 2.52225 . (1) . (2) . (3) . (4) . . Total literacy rate (Rmax |$=$| 1) . Treatment variable . Baseline effect (standard error), [R2] . Controlled effect (standard error), [R2] . Identified set . |$\tilde{\delta }$| for |$\beta =$| 0 Given |$R_{max}$| . Post-coup |$\times$| shrines per thousand persons -0.96511*** (0.324) [0.978] -0.95730*** (0.370) [0.979] [|$-0.95730,-$|0.75865] 2.52225 Notes: This table validates our results for the impact of historic shrines on literacy. Baseline effect includes the full set of baseline controls in column (2) of Table 2. Controlled effect includes the full set of controls (baseline |$+$| additional) in column (5) of Table 2. The identified set in column 3 is bounded below by |$\beta ^{\sim}$| and above by |$\beta$|*, calculated based on Rmax given in the top row and |$\delta =$| 1. Column (4) shows the value of |$\delta ^{\sim} $| which would produce |$\beta =$| 0, given the value of Rmax reported in the top row. Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The post-coup indicator variable equals 0 for the period 1901–1981 and one for the period 1998–2011. All regressions include year fixed effcets and tehsil fixed effects. Each control variable is interacted with the full set of year fixed effects. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.***Significant at 1%. Open in new tab 5.3. Flexible Estimates Going beyond the average effects presented in Table 2, we next provide more fine-grained evidence on how the relationship between historic shrines and literacy evolved over time. To this end, we estimate the flexible specification in equation (3) that interacts the historic shrine measure with year fixed effects and report the results in Table 4. Replicating the empirical set-up in Table 2, we first estimate the flexible specification with only year and tehsil fixed effects (see column (1)). The estimated coefficients on the shrine interaction are negative and increasing in magnitude over time. Subsequently, in column (2), we re-estimate our preferred specification with baseline controls, which include distance from Delhi, five agricultural land classifications, high rainfall dummy, and pre-partition proportion of minorities. As expected, conditional upon these controls, the coefficients on shrine interaction now become statistically insignificant and more or less indistinguishable from zero for the pre-shock period. Moreover, the estimated coefficients for the post-shock period are highly negative and statistically significant. Thus, it is only after the coup-induced administrative restructuring of 1985 that the literacy performance of tehsils with greater informal religious power becomes worse relative to tehsils that lacked such power. Table 4. Historical shrines and total literacy by time period. . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Shrines per thousand persons |$\times$| 1911 -0.0722* -0.0429 -0.0406 -0.0406 -0.0322 (0.0382) (0.0410) (0.0424) (0.0426) (0.0382) Shrines per thousand persons |$\times$| 1921 -0.0789 0.0118 0.0114 0.0128 0.0239 (0.0489) (0.0544) (0.0552) (0.0516) (0.0479) Shrines per thousand persons |$\times$| 1931 -0.262*** -0.0785 -0.0794 -0.0758 -0.0665 (0.0725) (0.0473) (0.0477) (0.0487) (0.0466) Shrines per thousand persons |$\times$| 1961 -0.595*** -0.0805 -0.0900 -0.0832 -0.0598 (0.173) (0.149) (0.141) (0.176) (0.168) Shrines per thousand persons |$\times$| 1972 -0.750*** -0.0500 -0.0605 -0.0554 -0.00502 (0.229) (0.228) (0.216) (0.240) (0.209) Shrines per thousand persons |$\times$| 1981 -0.791*** -0.0443 -0.0582 -0.0515 -0.0285 (0.234) (0.211) (0.206) (0.240) (0.239) Shrines per thousand persons |$\times$| 1998 -2.347*** -1.070*** -1.084*** -1.077** -1.008** (0.505) (0.383) (0.407) (0.407) (0.458) Shrines per thousand persons |$\times$| 2008 -2.318*** -1.018** -1.032** -1.032** -0.999** (0.497) (0.398) (0.409) (0.412) (0.449) Shrines per thousand persons |$\times$| 2011 -2.146*** -0.930** -0.953** -0.952** -0.937** (0.444) (0.360) (0.377) (0.379) (0.402) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.943 0.971 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 P-value for joint significance (1911–1981) 0.003 0.290 0.327 0.412 0.453 . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Shrines per thousand persons |$\times$| 1911 -0.0722* -0.0429 -0.0406 -0.0406 -0.0322 (0.0382) (0.0410) (0.0424) (0.0426) (0.0382) Shrines per thousand persons |$\times$| 1921 -0.0789 0.0118 0.0114 0.0128 0.0239 (0.0489) (0.0544) (0.0552) (0.0516) (0.0479) Shrines per thousand persons |$\times$| 1931 -0.262*** -0.0785 -0.0794 -0.0758 -0.0665 (0.0725) (0.0473) (0.0477) (0.0487) (0.0466) Shrines per thousand persons |$\times$| 1961 -0.595*** -0.0805 -0.0900 -0.0832 -0.0598 (0.173) (0.149) (0.141) (0.176) (0.168) Shrines per thousand persons |$\times$| 1972 -0.750*** -0.0500 -0.0605 -0.0554 -0.00502 (0.229) (0.228) (0.216) (0.240) (0.209) Shrines per thousand persons |$\times$| 1981 -0.791*** -0.0443 -0.0582 -0.0515 -0.0285 (0.234) (0.211) (0.206) (0.240) (0.239) Shrines per thousand persons |$\times$| 1998 -2.347*** -1.070*** -1.084*** -1.077** -1.008** (0.505) (0.383) (0.407) (0.407) (0.458) Shrines per thousand persons |$\times$| 2008 -2.318*** -1.018** -1.032** -1.032** -0.999** (0.497) (0.398) (0.409) (0.412) (0.449) Shrines per thousand persons |$\times$| 2011 -2.146*** -0.930** -0.953** -0.952** -0.937** (0.444) (0.360) (0.377) (0.379) (0.402) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.943 0.971 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 P-value for joint significance (1911–1981) 0.003 0.290 0.327 0.412 0.453 Notes: Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The shrines per thousand persons variable is interacted with each of the year fixed effects. All regressions include year fixed effects and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.*Significant at 10%; **Significant at 5%; ***Significant at 1%. Open in new tab Table 4. Historical shrines and total literacy by time period. . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Shrines per thousand persons |$\times$| 1911 -0.0722* -0.0429 -0.0406 -0.0406 -0.0322 (0.0382) (0.0410) (0.0424) (0.0426) (0.0382) Shrines per thousand persons |$\times$| 1921 -0.0789 0.0118 0.0114 0.0128 0.0239 (0.0489) (0.0544) (0.0552) (0.0516) (0.0479) Shrines per thousand persons |$\times$| 1931 -0.262*** -0.0785 -0.0794 -0.0758 -0.0665 (0.0725) (0.0473) (0.0477) (0.0487) (0.0466) Shrines per thousand persons |$\times$| 1961 -0.595*** -0.0805 -0.0900 -0.0832 -0.0598 (0.173) (0.149) (0.141) (0.176) (0.168) Shrines per thousand persons |$\times$| 1972 -0.750*** -0.0500 -0.0605 -0.0554 -0.00502 (0.229) (0.228) (0.216) (0.240) (0.209) Shrines per thousand persons |$\times$| 1981 -0.791*** -0.0443 -0.0582 -0.0515 -0.0285 (0.234) (0.211) (0.206) (0.240) (0.239) Shrines per thousand persons |$\times$| 1998 -2.347*** -1.070*** -1.084*** -1.077** -1.008** (0.505) (0.383) (0.407) (0.407) (0.458) Shrines per thousand persons |$\times$| 2008 -2.318*** -1.018** -1.032** -1.032** -0.999** (0.497) (0.398) (0.409) (0.412) (0.449) Shrines per thousand persons |$\times$| 2011 -2.146*** -0.930** -0.953** -0.952** -0.937** (0.444) (0.360) (0.377) (0.379) (0.402) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.943 0.971 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 P-value for joint significance (1911–1981) 0.003 0.290 0.327 0.412 0.453 . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Shrines per thousand persons |$\times$| 1911 -0.0722* -0.0429 -0.0406 -0.0406 -0.0322 (0.0382) (0.0410) (0.0424) (0.0426) (0.0382) Shrines per thousand persons |$\times$| 1921 -0.0789 0.0118 0.0114 0.0128 0.0239 (0.0489) (0.0544) (0.0552) (0.0516) (0.0479) Shrines per thousand persons |$\times$| 1931 -0.262*** -0.0785 -0.0794 -0.0758 -0.0665 (0.0725) (0.0473) (0.0477) (0.0487) (0.0466) Shrines per thousand persons |$\times$| 1961 -0.595*** -0.0805 -0.0900 -0.0832 -0.0598 (0.173) (0.149) (0.141) (0.176) (0.168) Shrines per thousand persons |$\times$| 1972 -0.750*** -0.0500 -0.0605 -0.0554 -0.00502 (0.229) (0.228) (0.216) (0.240) (0.209) Shrines per thousand persons |$\times$| 1981 -0.791*** -0.0443 -0.0582 -0.0515 -0.0285 (0.234) (0.211) (0.206) (0.240) (0.239) Shrines per thousand persons |$\times$| 1998 -2.347*** -1.070*** -1.084*** -1.077** -1.008** (0.505) (0.383) (0.407) (0.407) (0.458) Shrines per thousand persons |$\times$| 2008 -2.318*** -1.018** -1.032** -1.032** -0.999** (0.497) (0.398) (0.409) (0.412) (0.449) Shrines per thousand persons |$\times$| 2011 -2.146*** -0.930** -0.953** -0.952** -0.937** (0.444) (0.360) (0.377) (0.379) (0.402) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.943 0.971 0.971 0.971 0.971 F-stat for Joint significance (1911–1981) 3.68 1.26 1.18 1.03 0.97 P-value for joint significance (1911–1981) 0.003 0.290 0.327 0.412 0.453 Notes: Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The shrines per thousand persons variable is interacted with each of the year fixed effects. All regressions include year fixed effects and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.*Significant at 10%; **Significant at 5%; ***Significant at 1%. Open in new tab Next, in columns (3)–(5), we progressively add interactions of three further controls (distance to river, urbanization, and land revenue per capita) with year fixed effects, which we expect to be correlated with shrine location and trends in literacy. As expected, the basic pattern continues to hold in that the coefficients of shrine interaction only become statistically significant after 1985. Scrutinizing the individual coefficients on shrine interaction over time and across columns, we find a stable and consistent pattern after the addition of baseline controls (columns (2)–(5)). The p-value for the F-test, reported in Table 4, fails to reject the null of joint insignificance of shrine interactions in the pre-period (1901–1981) in our preferred specifications in columns (2)–(5). This lends further support to the absence of pre-trends. The top panel in Figure 7 provides a visual representation of these patterns by plotting the coefficients and confidence intervals of the shrine interaction in column (2) of Table 4. Reassuringly, there are no clear observable trends of the estimated interaction effects before the coup-induced policy shock (represented by the vertical dashed line corresponding to the year 1985). In fact, the coefficients are consistently placed on or extremely close to the zero line. The second figure in the bottom panel provides a related visual demonstration of the absence of pre-trends. Taking inspiration from Kahn-Lang and Lang (2020), we estimate an alternative flexible specification that considers the year prior to treatment to be the base year (i.e. 1981) and estimate the difference between the high- and low-shrine regions in each previous and subsequent year relative to 1981. This allows us to test whether average literacy rate prior to the policy-induced shock in 1985 exhibited parallel trends. As the bottom panel of Figure 7 shows, all coefficients in the pre-period are close to zero and statistically significant. Together, both figures provide reassuring evidence that tehsils with a greater concentration of historic shrines did not follow a differential trend in terms of literacy prior to the administrative shock, thereby supporting a key component of our identification strategy. Importantly, it is only after 1985 that the coefficients on shrine interaction become strongly negative and statistically significant. This confirms that the shrine effect remains latent prior to the coup-induced administrative shock in the mid-1980s and only becomes instrumentalized once the policy shock gave politicians direct control over public goods provision. The negative interaction effect is strongly persistent after the coup, which squares well with the fact that administrative arrangements for public goods provision introduced by General Zia continue to persist long after his departure in 1988. Figure 7. Open in new tabDownload slide The shrine effect over time. This figure plots the point estimates and confidence intervals of shrines per thousand persons interacted with year fixed effects. The shaded area corresponds to the post-shock period, 1985–2011. The top panel corresponds to the flexible specification in Table 4 (column (2)), where the base year is 1901. The bottom panel is based on an alternative flexible specification that uses the year prior to the policy shock as the base year (i.e. 1981) and estimates the difference between the high- and low-shrine tehsils in each previous and subsequent year relative to 1981. 5.4. Pre-Shock Differences in Literacy Determinants Despite there being no systematic trends in literacy between high- and low-shrine tehsils prior to the coup-induced administrative shock, it could still be the case that high-shrine tehsils were different along other dimensions that mattered for literacy. Unfortunately, we are limited by the extent to which data are available prior to the shock on characteristics of tehsils that could potentially influence literacy. What we are able to do, however, is to use data from a single pre-shock year (i.e. 1971 or 1973) to demonstrate that, at least in levels, there is no marked difference in the provision of literacy-related infrastructure between high- and low-shrine tehsils. Accordingly, we regress different measures of literacy-related infrastructure provision on each one of the three shrine measures: historic shrines per thousand persons, a dummy variable for high-shrine regions, and historic shrines per area. All regressions include district fixed effects and the full set of baseline controls. The results are reported in Table 5. In columns (1)–(4), we report estimates for models of schools per capita, run separately for total schools per capita as well as disaggregated by different educational levels (i.e. primary, middle, and high schools per capita). Finally, in columns (5)–(6), we present results for canal infrastructure and healthcare provision, measured as proportion of irrigated area fed by canals in 1973 and health units per capita, respectively. The underlying data are compiled from the annual editions of Punjab Development Statistics. As the results show, there are no statistically significant levels differences in the pre-shock period in the supply of public infrastructure (e.g. schools, irrigation, and health units) between high-shrine and low-shrine tehsils. Regardless of the type of infrastructure or shrine measure used, the coefficient on our shrine measure remains statistically insignificant. Table 5. Historical shrines and pre-coup infrastructure. . (1) . (2) . (3) . (4) . (5) . (6) . . Schools per thousand persons in 1971 (Total) . Schools per thousand persons in 1971 (Primary) . Schools per thousand persons in 1971 (Middle) . Schools per thousand persons in 1971 (High) . Proportion of irrigated area fed by canals in 1973 . Health Units per ten thousand persons in 1973 . Panel A Shrines per thousand persons -0.155 0.0676 0.0287 -0.251 -0.621 -0.145 (2.736) (2.161) (0.375) (0.435) (1.180) (0.615) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.448 0.419 0.518 0.484 0.761 0.365 Panel B High shrine dummy 0.0589 0.0454 0.0230 -0.00950 0.0252 -0.0173 (0.126) (0.102) (0.0155) (0.0196) (0.0761) (0.0205) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.450 0.422 0.545 0.483 0.761 0.374 Panel C Shrines per area -9.687 -7.619 -0.899 -1.169 -2.215 0.328 (18.07) (13.91) (2.491) (2.788) (7.178) (3.667) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.451 0.423 0.520 0.482 0.760 0.364 Controls Yes Yes Yes Yes Yes Yes . (1) . (2) . (3) . (4) . (5) . (6) . . Schools per thousand persons in 1971 (Total) . Schools per thousand persons in 1971 (Primary) . Schools per thousand persons in 1971 (Middle) . Schools per thousand persons in 1971 (High) . Proportion of irrigated area fed by canals in 1973 . Health Units per ten thousand persons in 1973 . Panel A Shrines per thousand persons -0.155 0.0676 0.0287 -0.251 -0.621 -0.145 (2.736) (2.161) (0.375) (0.435) (1.180) (0.615) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.448 0.419 0.518 0.484 0.761 0.365 Panel B High shrine dummy 0.0589 0.0454 0.0230 -0.00950 0.0252 -0.0173 (0.126) (0.102) (0.0155) (0.0196) (0.0761) (0.0205) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.450 0.422 0.545 0.483 0.761 0.374 Panel C Shrines per area -9.687 -7.619 -0.899 -1.169 -2.215 0.328 (18.07) (13.91) (2.491) (2.788) (7.178) (3.667) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.451 0.423 0.520 0.482 0.760 0.364 Controls Yes Yes Yes Yes Yes Yes Notes: Each panel uses a different measure for the power of shrine elites. Panel A uses our preferred measure, which is the number of historically important shrines per thousand persons in 1931. Panel B uses a high shrine dummy that takes a value 1 when the number of shrines is above its median value (and 0 otherwise). Panel C uses the number of shrines per spuare kilometre of area. The dependent variables in columns (1)–(4) are the supply of “total”, “primary”, “middle”, and “high” schools normalized by the population in the pre-coup year of 1971. The dependent variable in column (5) is a measure for the supply of canal infrastructure in the pre-coup year of 1973. This measure is contructed by taking the ratio of the total irrigated area to the area that is fed through canals. Finally, the dependent variable in column (6) is the supply of health units normalized by the population in the pre-coup year of 1973. The unit of observation is tehsil. All regressions use a consistent sample of 66 tehsils measured at the 1931 boundaries for which the data on infrastructure variables is available. The baseline controls included in all regressions are high rainfall dummy, agrarian land classifications, proportion minorities, and distance to Delhi. Variables are historically defined using consistently available information from district gazetteers in 1931. As these are cross-sectional regressions the controls have not been interacted with year fixed effects. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. Open in new tab Table 5. Historical shrines and pre-coup infrastructure. . (1) . (2) . (3) . (4) . (5) . (6) . . Schools per thousand persons in 1971 (Total) . Schools per thousand persons in 1971 (Primary) . Schools per thousand persons in 1971 (Middle) . Schools per thousand persons in 1971 (High) . Proportion of irrigated area fed by canals in 1973 . Health Units per ten thousand persons in 1973 . Panel A Shrines per thousand persons -0.155 0.0676 0.0287 -0.251 -0.621 -0.145 (2.736) (2.161) (0.375) (0.435) (1.180) (0.615) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.448 0.419 0.518 0.484 0.761 0.365 Panel B High shrine dummy 0.0589 0.0454 0.0230 -0.00950 0.0252 -0.0173 (0.126) (0.102) (0.0155) (0.0196) (0.0761) (0.0205) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.450 0.422 0.545 0.483 0.761 0.374 Panel C Shrines per area -9.687 -7.619 -0.899 -1.169 -2.215 0.328 (18.07) (13.91) (2.491) (2.788) (7.178) (3.667) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.451 0.423 0.520 0.482 0.760 0.364 Controls Yes Yes Yes Yes Yes Yes . (1) . (2) . (3) . (4) . (5) . (6) . . Schools per thousand persons in 1971 (Total) . Schools per thousand persons in 1971 (Primary) . Schools per thousand persons in 1971 (Middle) . Schools per thousand persons in 1971 (High) . Proportion of irrigated area fed by canals in 1973 . Health Units per ten thousand persons in 1973 . Panel A Shrines per thousand persons -0.155 0.0676 0.0287 -0.251 -0.621 -0.145 (2.736) (2.161) (0.375) (0.435) (1.180) (0.615) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.448 0.419 0.518 0.484 0.761 0.365 Panel B High shrine dummy 0.0589 0.0454 0.0230 -0.00950 0.0252 -0.0173 (0.126) (0.102) (0.0155) (0.0196) (0.0761) (0.0205) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.450 0.422 0.545 0.483 0.761 0.374 Panel C Shrines per area -9.687 -7.619 -0.899 -1.169 -2.215 0.328 (18.07) (13.91) (2.491) (2.788) (7.178) (3.667) Mean outcome 1.168 1.026 0.098 0.044 0.665 0.078 Observations 66 66 66 66 66 66 Adjusted R-squared 0.451 0.423 0.520 0.482 0.760 0.364 Controls Yes Yes Yes Yes Yes Yes Notes: Each panel uses a different measure for the power of shrine elites. Panel A uses our preferred measure, which is the number of historically important shrines per thousand persons in 1931. Panel B uses a high shrine dummy that takes a value 1 when the number of shrines is above its median value (and 0 otherwise). Panel C uses the number of shrines per spuare kilometre of area. The dependent variables in columns (1)–(4) are the supply of “total”, “primary”, “middle”, and “high” schools normalized by the population in the pre-coup year of 1971. The dependent variable in column (5) is a measure for the supply of canal infrastructure in the pre-coup year of 1973. This measure is contructed by taking the ratio of the total irrigated area to the area that is fed through canals. Finally, the dependent variable in column (6) is the supply of health units normalized by the population in the pre-coup year of 1973. The unit of observation is tehsil. All regressions use a consistent sample of 66 tehsils measured at the 1931 boundaries for which the data on infrastructure variables is available. The baseline controls included in all regressions are high rainfall dummy, agrarian land classifications, proportion minorities, and distance to Delhi. Variables are historically defined using consistently available information from district gazetteers in 1931. As these are cross-sectional regressions the controls have not been interacted with year fixed effects. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. Open in new tab 6. Extended Analysis and Robustness Historical Shrines and Post-Coup Enrolment Having shown that regions dominated by shrine elites witnessed a retarded growth in literacy after the coup-induced administrative shock of 1985, we investigate whether these regions also experienced lower enrolment rates. To this end, we compile a tehsil-level panel of annual enrolment data from 1982 to 1998 that encompasses the period during which the state implemented a large-scale programme for public school expansion (see Section 2.2). We then regress these enrolment rates on our historic shrine measure and the full set of baseline controls. Note that we do not include tehsil fixed effects in these regressions since our explanatory variable, shrines per thousand persons, remains fixed for a tehsil over time. However, all regressions do include year fixed effects and control for within tehsil correlation over time by clustering the standard errors at the tehsil level. Results are reported in Table 6 for both total enrolment rates (0–19 years) and disaggregated by primary (5–14 years) and secondary (15–19 years) enrolment. For each regression model, we report the coefficient estimates for shrines per capita with and without the baseline controls. As shown in columns (1)–(2), tehsils with a greater concentration of historic shrines have systematically lower total enrolment rates. Disaggregating by primary and secondary enrolment rates, we show that the result is mainly driven by consistently lower enrolment in primary schools (columns (3)–(4)). Whether the model is estimated with or without baseline controls, the coefficient on shrines per capita remains negative and statistically significant at 1% level. This suggests that shrine dense regions witnessed a relatively more retarded growth in primary enrolment rates than the shrine-poor regions during the post-shock period. Table 6. Historical shrines and post-coup enrolment. . (1) . (2) . (3) . (4) . (5) . (6) . . Total enrolment rate (0–19) . Primary enrolment rate (5–14) . Secondary enrolment rate (15–19) . Shrines per thousand persons -106.7*** -36.95** -91.49*** -46.25*** -218.1*** -24.83 (28.41) (15.53) (25.66) (11.91) (69.33) (51.85) Mean outcome 22.857 22.857 21.150 21.150 60.022 60.022 Observations 924 924 924 924 924 924 Adjusted R-squared 0.196 0.607 0.128 0.535 0.237 0.545 Baseline controls High rainfall dummy (1931) N Y N Y N Y Agrarian land classifications (1931) N Y N Y N Y Proportion minorities (1931) N Y N Y N Y Distance to Delhi (km) N Y N Y N Y . (1) . (2) . (3) . (4) . (5) . (6) . . Total enrolment rate (0–19) . Primary enrolment rate (5–14) . Secondary enrolment rate (15–19) . Shrines per thousand persons -106.7*** -36.95** -91.49*** -46.25*** -218.1*** -24.83 (28.41) (15.53) (25.66) (11.91) (69.33) (51.85) Mean outcome 22.857 22.857 21.150 21.150 60.022 60.022 Observations 924 924 924 924 924 924 Adjusted R-squared 0.196 0.607 0.128 0.535 0.237 0.545 Baseline controls High rainfall dummy (1931) N Y N Y N Y Agrarian land classifications (1931) N Y N Y N Y Proportion minorities (1931) N Y N Y N Y Distance to Delhi (km) N Y N Y N Y Notes: The table shows the impact of historic shrines on post-coup school enrolment during the period 1982–1998. Three separate measures of enrolment are used. Columns (1)–(2) use total enrolment normalized by the population of school going age (i.e. age bracket 0–19). Columns (3)–(4) use primary enrolment normalized by the population of primary school going age (i.e. age bracket 5–14). Finally, columns (5)–(6) use secondary enrolment normalized by the population of secondary school going age (i.e. age bracket 15–19). The unit of observation is tehsil-year. The panel upon which the regressions are based encompass the years 1982, 1984, 1985, 1986, 1988, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, and 1998. We do not include tehsil fixed effects in these regressions since our explanatory variable, shrines per thousand persons, remains fixed for a tehsil over time. However, all regressions do include year fixed effcets. The baseline controls include high rainfall dummy, agrarian land classififcations, proportion minorities, and distance to Delhi. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. **Significant at 5%; ***Significant at 1%. Open in new tab Table 6. Historical shrines and post-coup enrolment. . (1) . (2) . (3) . (4) . (5) . (6) . . Total enrolment rate (0–19) . Primary enrolment rate (5–14) . Secondary enrolment rate (15–19) . Shrines per thousand persons -106.7*** -36.95** -91.49*** -46.25*** -218.1*** -24.83 (28.41) (15.53) (25.66) (11.91) (69.33) (51.85) Mean outcome 22.857 22.857 21.150 21.150 60.022 60.022 Observations 924 924 924 924 924 924 Adjusted R-squared 0.196 0.607 0.128 0.535 0.237 0.545 Baseline controls High rainfall dummy (1931) N Y N Y N Y Agrarian land classifications (1931) N Y N Y N Y Proportion minorities (1931) N Y N Y N Y Distance to Delhi (km) N Y N Y N Y . (1) . (2) . (3) . (4) . (5) . (6) . . Total enrolment rate (0–19) . Primary enrolment rate (5–14) . Secondary enrolment rate (15–19) . Shrines per thousand persons -106.7*** -36.95** -91.49*** -46.25*** -218.1*** -24.83 (28.41) (15.53) (25.66) (11.91) (69.33) (51.85) Mean outcome 22.857 22.857 21.150 21.150 60.022 60.022 Observations 924 924 924 924 924 924 Adjusted R-squared 0.196 0.607 0.128 0.535 0.237 0.545 Baseline controls High rainfall dummy (1931) N Y N Y N Y Agrarian land classifications (1931) N Y N Y N Y Proportion minorities (1931) N Y N Y N Y Distance to Delhi (km) N Y N Y N Y Notes: The table shows the impact of historic shrines on post-coup school enrolment during the period 1982–1998. Three separate measures of enrolment are used. Columns (1)–(2) use total enrolment normalized by the population of school going age (i.e. age bracket 0–19). Columns (3)–(4) use primary enrolment normalized by the population of primary school going age (i.e. age bracket 5–14). Finally, columns (5)–(6) use secondary enrolment normalized by the population of secondary school going age (i.e. age bracket 15–19). The unit of observation is tehsil-year. The panel upon which the regressions are based encompass the years 1982, 1984, 1985, 1986, 1988, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, and 1998. We do not include tehsil fixed effects in these regressions since our explanatory variable, shrines per thousand persons, remains fixed for a tehsil over time. However, all regressions do include year fixed effcets. The baseline controls include high rainfall dummy, agrarian land classififcations, proportion minorities, and distance to Delhi. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. **Significant at 5%; ***Significant at 1%. Open in new tab Placebo Test Our empirical analysis has so far used the measure of historic shrines per capita to proxy for the power of shrine elites. To reinforce our argument that it is religious elites attached to historic shrines and not shrines in and of themselves that matter for literacy, we conduct a placebo exercise that replaces our historical shrine measure with a measure based on the universe of all shrines in Punjab, covering many lesser known shrines that are unattached with elite power. To construct this measure, we leverage our Auqaf database where we compiled information on all shrines in Punjab (see Section 3.1). Results are reported in Table 7. Replicating the empirical set up in Table 2, we start by regressing the total literacy rate on the interaction between the all-shrines measure and the post-shock dummy, while controlling for tehsil and year fixed effects only. Subsequently, in column (2), we flexibly account for baseline controls and the additional controls are successively included in columns (3)–(5). In none of the specifications, whether with or without controls, do we find a statistically significant coefficient on the interaction between the all-shrines measure and the post-shock dummy. Table 7. All shrines and total literacy. . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Post-coup |$\times$| all shrines per thousand persons -2.332 -0.135 -0.108 -0.0128 -0.119 (1.753) (1.433) (1.426) (1.587) (1.622) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.936 0.969 0.969 0.969 0.969 . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Post-coup |$\times$| all shrines per thousand persons -2.332 -0.135 -0.108 -0.0128 -0.119 (1.753) (1.433) (1.426) (1.587) (1.622) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.936 0.969 0.969 0.969 0.969 Notes: Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. All shrines per thousand persons is the total number of shrines per thousand persons in 1931—these include the number of historically important shrines. The post-coup indicator variable equals 0 for the period 1901–1981 and one for the period 1998–2011. All regressions include year fixed effcets and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. Open in new tab Table 7. All shrines and total literacy. . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Post-coup |$\times$| all shrines per thousand persons -2.332 -0.135 -0.108 -0.0128 -0.119 (1.753) (1.433) (1.426) (1.587) (1.622) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.936 0.969 0.969 0.969 0.969 . (1) . (2) . (3) . (4) . (5) . . Total literacy rate . Post-coup |$\times$| all shrines per thousand persons -2.332 -0.135 -0.108 -0.0128 -0.119 (1.753) (1.433) (1.426) (1.587) (1.622) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y Agrarian land classifications (1931) N Y Y Y Y Proportion minorities (1931) N Y Y Y Y Distance to Delhi (km) N Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Urbanization (1931) N N N Y Y Land revenue per capita (1901) N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 Adjusted R-squared 0.936 0.969 0.969 0.969 0.969 Notes: Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. All shrines per thousand persons is the total number of shrines per thousand persons in 1931—these include the number of historically important shrines. The post-coup indicator variable equals 0 for the period 1901–1981 and one for the period 1998–2011. All regressions include year fixed effcets and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. Open in new tab This reaffirms our argument that it is only the subset of shrines that were recognized by the British colonial administration and supported the religious and political authority of shrine elites that shape the impact of the coup-induced administrative shock on literacy. Aside from clarifying the impact of religious elites, this placebo exercise also furnishes crucial evidence against any remaining selection concern on shrine location that our baseline controls might not have fully accounted for. If shrines self-selected into areas that were predisposed to lower literacy expansion, then this selection on location would also apply to all shrines rather than just the historically significant shrines that form the basis of our shrine measure. However, such a concern is assuaged by the absence of any strong predictive power of the all-shrines measure in Table 7. Alternative Baseline Controls To assess the robustness of our results, we include alternative proxies for some of the baseline controls in our empirical analysis. We first replace the binary classification for high and low rainfall regions with a continuous historical measure of rainfall obtained from colonial district gazetteers and averaged over the period 1917–1923. Next, we replace our rainfall dummy with the natural logarithm of elevation in feet.28 The corresponding results are reported in columns (6)–(7), respectively, in Table 8. As the results show, if anything, the coefficient on shrine interaction becomes stronger in magnitude and retains its statistical significance. Finally, we replace the urbanization control with a big city dummy, which uses the classification in Burki (1973) and codes all major cities of Punjab at the time of partition as 1 (and 0 otherwise). As Bharadwaj, Khwaja, and Mian (2008) show, large cities in Pakistan were particularly impacted by the partition-related population flows. The inclusion of big city dummy thus allows us to further strengthen our partition-related controls over and above the proportion of minorities already included in the baseline specification. As column (8) in Table 8 shows, our headline result on historic shrines per capita remains robust to flexibly controlling for the big city dummy. Table 8. Robustness to additional controls. . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . (9) . (10) . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** -1.022*** -1.301*** -0.960** -0.973*** -0.903** (0.382) (0.324) (0.341) (0.344) (0.370) (0.280) (0.414) (0.367) (0.347) (0.358) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y N N Y Y Y Agrarian land classifications (1931) N Y Y Y Y Y Y Y Y Y Proportion minorities (1931) N Y Y Y Y Y Y Y Y Y Distance to Delhi (km) N Y Y Y Y Y Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Y Y Y Y Y Urbanization (1931) N N N Y Y Y Y N Y Y Land revenue per capita (1901) N N N N Y Y Y Y N Y Average Annual Rainfall (mm) N N N N N Y N N N N Elevation (feet) N N N N N N Y N N N Big City Dummy (1931) N N N N N N N Y N N Proportion Landlord (1931) N N N N N N N N Y N Christian Missions per capita (1931) N N N N N N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 0.977 0.966 0.972 0.971 0.973 . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . (9) . (10) . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** -1.022*** -1.301*** -0.960** -0.973*** -0.903** (0.382) (0.324) (0.341) (0.344) (0.370) (0.280) (0.414) (0.367) (0.347) (0.358) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y N N Y Y Y Agrarian land classifications (1931) N Y Y Y Y Y Y Y Y Y Proportion minorities (1931) N Y Y Y Y Y Y Y Y Y Distance to Delhi (km) N Y Y Y Y Y Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Y Y Y Y Y Urbanization (1931) N N N Y Y Y Y N Y Y Land revenue per capita (1901) N N N N Y Y Y Y N Y Average Annual Rainfall (mm) N N N N N Y N N N N Elevation (feet) N N N N N N Y N N N Big City Dummy (1931) N N N N N N N Y N N Proportion Landlord (1931) N N N N N N N N Y N Christian Missions per capita (1931) N N N N N N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 0.977 0.966 0.972 0.971 0.973 Notes: The table shows the robustness of our main results to the inclusion of a whole series of additional controls. Columns (1)–(5) are copied over from the Table 2. From column (6) onwards, we start including, on top of the full set of controls included in Table 2, a series of additonal controls. These include “average annual rainfall in mm”, “elevation of the tehsil headquarter in feet”, “a dummy for a big city in 1931”, “proportion of land tenurial contracts that are classified as landlord (zamindari) in 1931”, and “the number of christian missions per capita in 1931”. Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The post-coup indicator variable equals 0 for the period 1901–1981 and one for the period 1998–2011. All regressions include year fixed effcets and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.**Significant at 5%; ***Significant at 1%. Open in new tab Table 8. Robustness to additional controls. . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . (9) . (10) . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** -1.022*** -1.301*** -0.960** -0.973*** -0.903** (0.382) (0.324) (0.341) (0.344) (0.370) (0.280) (0.414) (0.367) (0.347) (0.358) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y N N Y Y Y Agrarian land classifications (1931) N Y Y Y Y Y Y Y Y Y Proportion minorities (1931) N Y Y Y Y Y Y Y Y Y Distance to Delhi (km) N Y Y Y Y Y Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Y Y Y Y Y Urbanization (1931) N N N Y Y Y Y N Y Y Land revenue per capita (1901) N N N N Y Y Y Y N Y Average Annual Rainfall (mm) N N N N N Y N N N N Elevation (feet) N N N N N N Y N N N Big City Dummy (1931) N N N N N N N Y N N Proportion Landlord (1931) N N N N N N N N Y N Christian Missions per capita (1931) N N N N N N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 0.977 0.966 0.972 0.971 0.973 . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . (9) . (10) . . Total literacy rate . Post-coup |$\times$| shrines per thousand persons -1.906*** -0.965*** -0.977*** -0.979*** -0.957** -1.022*** -1.301*** -0.960** -0.973*** -0.903** (0.382) (0.324) (0.341) (0.344) (0.370) (0.280) (0.414) (0.367) (0.347) (0.358) Baseline controls |$\times$| (year fixed effects): High rainfall dummy (1931) N Y Y Y Y N N Y Y Y Agrarian land classifications (1931) N Y Y Y Y Y Y Y Y Y Proportion minorities (1931) N Y Y Y Y Y Y Y Y Y Distance to Delhi (km) N Y Y Y Y Y Y Y Y Y Additional controls |$\times$| (year fixed effects): Distance to river (km) N N Y Y Y Y Y Y Y Y Urbanization (1931) N N N Y Y Y Y N Y Y Land revenue per capita (1901) N N N N Y Y Y Y N Y Average Annual Rainfall (mm) N N N N N Y N N N N Elevation (feet) N N N N N N Y N N N Big City Dummy (1931) N N N N N N N Y N N Proportion Landlord (1931) N N N N N N N N Y N Christian Missions per capita (1931) N N N N N N N N N Y Mean outcome 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 0.215 Observations 680 680 680 680 680 680 680 680 680 680 Adjusted R-squared 0.945 0.972 0.971 0.971 0.971 0.977 0.966 0.972 0.971 0.973 Notes: The table shows the robustness of our main results to the inclusion of a whole series of additional controls. Columns (1)–(5) are copied over from the Table 2. From column (6) onwards, we start including, on top of the full set of controls included in Table 2, a series of additonal controls. These include “average annual rainfall in mm”, “elevation of the tehsil headquarter in feet”, “a dummy for a big city in 1931”, “proportion of land tenurial contracts that are classified as landlord (zamindari) in 1931”, and “the number of christian missions per capita in 1931”. Observations are at the tehsil-census year level. All regressions use a consistent sample of 68 tehsils measured at the 1931 boundaries. The census years are 1901, 1911, 1921, 1931, 1961, 1972, 1981, 1998, 2007, and 2011. The dependent variable is the proportion of the total population that is literate. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. The post-coup indicator variable equals 0 for the period 1901–1981 and one for the period 1998–2011. All regressions include year fixed effcets and tehsil fixed effects. Each control variable listed is interacted with a full set of year fixed effects. The inclusion of a control variable interacted with the full set of year fixed effects is indicated by a Y; N indicates that the control is not included in the specification. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.**Significant at 5%; ***Significant at 1%. Open in new tab Land Inequality Land inequality is a potentially important confounding factor in our analysis. If tehsils with a greater shrine concentration also had more unequal distribution of land, then this could bias our empirical results. As argued in Section 2.2, shrine guardians are sufficiently differentiated from landed elites both in terms of their incentives and capacity to oppose education. Nevertheless, we try and address this concern in several ways. First, to the extent that patterns of land inequality are highly persistent in Pakistan, especially in the absence of any meaningful land reforms, tehsil-level fixed effects should capture any time-invariant cross-tehsil differences in land inequality. Second, given the well-established association between patterns of rainfall and land inequality in Punjab, we have consistently included a dummy for high rainfall regions as a baseline control in all specifications and assessed the robustness of our results to replacing it with a continuous measure of rainfall. Third, some of our main control variables could serve as indirect proxies for land inequality. These include, in particular, our agricultural land classification indicators and colonial land revenue per capita. As discussed in Section 4.2, British policies for land grants in canal colonies shaped earlier patterns of land inequality that have persisted over time (Ali 2014). Similarly, aside from serving as a measure of state capacity, land revenue per capita is strongly correlated with the size and distribution of landholdings. Our final strategy is to include a more direct proxy for colonial-era land inequality, defined as the proportion of landlord-based land tenurial contracts. As Banerjee and Iyer (2005) argue, regions in British India where proprietary rights were historically given to landlords (zamindars) had higher Gini coefficients of land inequality and suffered from relatively lower investments in health and education in the long-run. Taking cue from this, in column (9) of Table 8, we replace historical land revenue per capita with the proportion of landlord contracts interacted with year fixed effects. As the results show, the coefficient on shrine interaction retains its statistical significance despite the inclusion of Banerjee–Iyer type land tenure variable. Christian Missions Recent research on the relationship between religion and human capital has established that early exposure to Christian missionary activity, especially Protestantism, shaped long-run educational outcomes (e.g. Becker and Woessmann 2009; Gallego and Woodberry 2010). A recent study establishes a similar finding for India (Calvi, Hoehn-Velasco, and Mantovanelli 2020). The presence of historical Christian missions can be a possible omitted variable in our analysis, since it can both have a direct impact on contemporary human capital and be correlated with shrine presence. This is likely to be an important concern as historical accounts suggest that missionaries were less likely to be located in regions with substantial Muslim presence (Richter 1908). If this were the case, then our negative shrine effect on literacy could be spurious. To account for this, we construct a measure of the presence of historic Christian missions, defined as the number of Protestant missions per thousand persons in a tehsil in 1908. As in Calvi, Hoehn-Velasco, and Mantovanelli (2020), the underlying data for this measure are drawn from the Statistical Atlas of Christian Missions, which provides comprehensive information until 1908 on the presence of Protestant Missions across the world, including British India. We include the interaction between our measure of historical Christian missions per capita and year fixed effects in column (10) of Table 8 and demonstrate that the shrine effect is robust to its inclusion. Alternative Shrine Measures In Online Appendix Table A3, we investigate the robustness of our main findings to two alternative measures of religious power: the number of historic shrines per square mile and a high shrine dummy. As our results suggest, the coefficient on shrine interaction remains negative and statistically significant both in specifications with high shrine dummy (panel A) and shrines per area (panel B). Our results thus remain highly robust to the use of alternative shrine measures. Separately, in Online Appendix Table A.4, we present results for a shrine measure that captures the extensive margin through the construction of a simple dummy variable for shrine presence. As Online Appendix Table A4 shows, the interaction of this shrine dummy with the post-coup indicator lacks any predictive power. This further reinforces that our argument is based on the intensive margin of shrines (i.e. how shrine dense a region is relative to others). Disaggregating by Gender We next estimate our baseline specification in Table 2 separately for male and female literacy. Since literacy data disaggregated by gender are not available for the years 1972, 2008, and 2011, our estimations are conducted on a smaller sample. We find a relatively stronger effect of our historic shrines per capita measure on female literacy rate in the post-shock period (see Online Appendix Table A.5). However, given the loss of data, this result should be interpreted with caution. Further Robustness Tests We next consider a battery of additional robustness tests. To conserve space, we report these results in the Online Appendix. In each of our robustness exercises, we preserve the basic set-up in Table 2 where we only include the tehsil and year fixed effects in column (1), add the set of baseline controls in column (2), and progressively include further controls in columns (3)–(5). We first recognize the possibility of a trend break in the data between the pre- and post-partition periods and therefore restrict the sample to the post-partition period (1961–2011). Re-estimating the baseline specification on the post-partition sample delivers a highly consistent set of results whereby the shrine effect remains negative and statistically significant at 1% level in all specifications (see Online Appendix Table A6). In Online Appendix Table A7, we subject our results to a strict test, whereby tehsils at the top and bottom 5th percentiles of literacy (top panel) and shrines per thousand persons (bottom panel) are successively dropped from the sample. As the results show, dropping tehsils at the bottom 5th percentile of literacy or shrine per thousand persons (columns (1)–(5)) preserves our findings. Next, we show robustness to dropping observations in the top 5th percentile of literacy and shrines per thousand persons (columns (6)–(10)). Overall, our findings are fairly robust and consistent to the exclusion of these extreme observations. Next, we replace tehsil fixed effects with more loosely defined district fixed effects, which effectively allows us to compare tehsils within the same district that have more or less historic religious power (see Online Appendix Table A8). Again, results remain unchanged. Our findings are similarly robust to clustering the standard errors by district rather than tehsil (see Online Appendix Table A9) and to alternative functional forms that use the natural logarithm of the dependent and independent variables, respectively.29 Finally, we probe the robustness of our findings to dropping the imputed observations. To do so, we re-estimate our baseline and flexible specifications in Tables 2 and 4 on the unbalanced panel and show that our main results remain robust to such an exercise (see Online Appendix Tables A10–A11). 7. Mechanisms We have argued that shrine elites preside over a hierarchical social structure and are more averse to educational expansion. What remains to be shown, however, are the channels through which shrine elites tend to undermine education. Broadly speaking, shrine elites could affect educational outcomes by influencing the supply or demand for education, or some combination of both. The supply-side story would be relevant if shrine elites were able to influence the construction of schools or provision of school inputs (e.g. teachers). As an alternative supply-side strategy, shrine elites could also influence educational outcomes by making school provision defective. This would be the case, for example, if schools were constructed in more remote and distant locations, if they lacked essential facilities (e.g. electricity and boundary walls), and if schooling resources were spread too thinly. On the other hand, a demand-side explanation would be relevant if shrine elites were able to influence the conditions that shape the demand for education in their regions. Testing both of these explanations, we find stronger evidence in favour of a supply-side story that is based on defective school provision. We begin our discussion by first showing that shrine-dominated regions have a higher presence of shrine elites in electoral politics in the period after the 1977 military coup. We then show that these elites used their political power to restrain education by making school provision defective along three specific dimensions: school siting, school size, and the quality of infrastructure. We conclude our discussion by presenting evidence on demand-side factors. Growing Electoral Presence of Shrine Elites While shrine elites have been a permanent feature of the traditional power structure, they established a stronger electoral foothold during the Zia-era (1979–1988) and became key actors in patronage politics. Using the electoral database discussed in Section 4.1, Figure 8 shows a noticeable increase in the number of shrine contestants per constituency under the Zia regime. There was a three-fold increase in this ratio between the last election before Zia (1977) and the first election after Zia (1988). A growing proportion of shrine contestants also got elected into the parliament (shrine electables) during the same period. As Online Appendix Figures A.4 and A.5 show, the growing electoral participation of shrine families persisted into the post-Zia period. While this provides indicative evidence of the direct participation of shrine elites in politics, it is important to emphasize that shrine elites are also able to impact public goods provision through their indirect political influence on whomsoever is elected in their constituency.30 Such growing politicization of shrines is rooted in the weakening of party politics during the Zia-era when elections were held on a non-party basis. This gave electoral advantage to natural formations, thereby elevating the role of religious prestige, besides wealth and kinship (biraderi), as a crucial determinant of electoral success.31 Figure 8. Open in new tabDownload slide Growing shrine presence in electoral politics. The figure shows the proportion of all shrine-linked candidates in a constituency who either contested or won in National and Provincial Assembly elections in Punjab. The former are described as contestants and the latter as electables. Each year on the horizontal axis refers to an election round. Supply-Side Explanations The 1977 military coup and the associated administrative restructuring thus ushered a period when shrine elites controlled both political power and access to public spending on development. This allowed shrine elites to systematically influence the ecosystem of schooling in ways that undermined mass literacy. An obvious strategy would be for shrine elites to directly oppose educational expansion by reducing public spending on education and obstructing the construction of schools in their regions. While a direct tehsil-level measure of educational spending is unavailable,32 we explore the spending explanation by using a dataset on the dispersal of development funds for all national electoral constituencies of Punjab. The data come from Malik (2020), and covers the electoral cycle, 1993–1996. In Table 9, we empirically examine whether constituencies represented by shrine families systematically under-utilized development funds. We regress the proportion of the maximum possible dispersal of development funds that was actually spent in each constituency on a binary measure that equals 1 if the legislator is a shrine elite (0, otherwise). The unit of analysis is electoral constituency-year. The results show no evidence that shrine-linked legislators systematically under-utilized development funds in their constituencies. Since education forms a major component of development funds in Pakistan, the results in Table 9 provide suggestive evidence against a simple supply-side story based on educational spending. Additionally, we do not find evidence that shrine-dominated regions have systematically lower supply of schools or teachers (see Online Appendix Table A.12). Table 9. Shrines in electoral politics and development funds. . (1) . (2) . (3) . (4) . (5) . (6) . . Utilization rate . Log amount . Shrine MNA in 1993 0.090 0.090 0.090 0.126 0.126 0.126 (0.140) (0.141) (0.142) (0.192) (0.193) (0.195) Controls: Region fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Region-specific linear time trends No Yes No No Yes No Region-by-year fixed effects No No Yes No No Yes Mean outcome 0.572 0.572 0.572 1.005 1.005 1.005 Observations 364 364 364 364 364 364 Adjusted R-squared 0.203 0.205 0.200 0.297 0.299 0.298 . (1) . (2) . (3) . (4) . (5) . (6) . . Utilization rate . Log amount . Shrine MNA in 1993 0.090 0.090 0.090 0.126 0.126 0.126 (0.140) (0.141) (0.142) (0.192) (0.193) (0.195) Controls: Region fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Region-specific linear time trends No Yes No No Yes No Region-by-year fixed effects No No Yes No No Yes Mean outcome 0.572 0.572 0.572 1.005 1.005 1.005 Observations 364 364 364 364 364 364 Adjusted R-squared 0.203 0.205 0.200 0.297 0.299 0.298 Notes: The above models regress the utilization rate (proportion of maximum allowable allocation of development funds actually realized) and the log amount of development funds over the electoral cycle 1993–1996 on a binary variable indicating whether the constituency returned a shrine affiliated Member of National Assembly (MNA) at the start of the cycle in 1993. The unit of observation is constituency-year. All regressions include year fixed effects and region fixed effects. Standard errors clustered by district in parentheses, unless otherwise indicated. Open in new tab Table 9. Shrines in electoral politics and development funds. . (1) . (2) . (3) . (4) . (5) . (6) . . Utilization rate . Log amount . Shrine MNA in 1993 0.090 0.090 0.090 0.126 0.126 0.126 (0.140) (0.141) (0.142) (0.192) (0.193) (0.195) Controls: Region fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Region-specific linear time trends No Yes No No Yes No Region-by-year fixed effects No No Yes No No Yes Mean outcome 0.572 0.572 0.572 1.005 1.005 1.005 Observations 364 364 364 364 364 364 Adjusted R-squared 0.203 0.205 0.200 0.297 0.299 0.298 . (1) . (2) . (3) . (4) . (5) . (6) . . Utilization rate . Log amount . Shrine MNA in 1993 0.090 0.090 0.090 0.126 0.126 0.126 (0.140) (0.141) (0.142) (0.192) (0.193) (0.195) Controls: Region fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Region-specific linear time trends No Yes No No Yes No Region-by-year fixed effects No No Yes No No Yes Mean outcome 0.572 0.572 0.572 1.005 1.005 1.005 Observations 364 364 364 364 364 364 Adjusted R-squared 0.203 0.205 0.200 0.297 0.299 0.298 Notes: The above models regress the utilization rate (proportion of maximum allowable allocation of development funds actually realized) and the log amount of development funds over the electoral cycle 1993–1996 on a binary variable indicating whether the constituency returned a shrine affiliated Member of National Assembly (MNA) at the start of the cycle in 1993. The unit of observation is constituency-year. All regressions include year fixed effects and region fixed effects. Standard errors clustered by district in parentheses, unless otherwise indicated. Open in new tab Besides restraining spending and opposing the construction of new schools, shrine elites could also oppose schooling by rendering existing educational provision defective. Plentiful evidence exists on how political interference has distorted the quality of educational provision through suboptimal siting of schools, poor school infrastructure, and uneconomical school size. The politicization of schooling became a particularly noteworthy concern in the wake of Social Action Programme (SAP), a multi-donor initiative launched in 1992 to rectify Pakistan’s historic deficit in social sector spending. The programme was subjected to such widespread political abuse that independent evaluators termed it a “failure”, and concluded that “politicians used staff recruitment, construction contracts, and site selection for schools and clinics to enrich their kith and kin” (Birdsall, Malik, and Vaishnav 2005, p. 26). A more recent assessment on the quality of school infrastructure shows that around 60% of primary schools do not have electricity, 36% lack any drinking facility, and 42% have no washrooms (GOP 2014). To determine whether shrine regions were more adversely affected by such politicization, we investigate three dimensions of political capture for which tehsil-level data are available. We start by focusing on school location. Elected politicians are known to have influenced the “selection of school sites” (Birdsall, Malik, and Vaishnav 2005, p. 40). A remote location can make schools practically inaccessible to students, especially females, and allows local notables to divert them for private use (e.g. as personal residence, cattle-sheds or stables). To investigate whether schools were more distant in shrine-dominated regions, we use cross-sectional household data from the Multiple Clusters Indicators Survey (MICS 2007) that reports the distance of a household from the nearest government or private school in the year 2007. Analysis is carried out for 123 contemporary tehsils of Punjab. We rely on three categories of physical access to schools provided by the MICS database, all based on distance from the nearest school. Physical access is closest if the nearest school is situated less than 2 km away from the surveyed household. Schools that are located between 2 and 5 km away are relatively more distant. Finally, schools that are more than 5 km away are the farthest. We explore variation in these categories across two metrics: gender (boy versus girl) and provider (public versus private). Besides the main variables, each specification contains latitude, longitude, elevation, the incidence of landlessness, population density, and region-specific fixed effects as controls. As the results in Table 10 show, shrine presence increases the likelihood of a school being distant (the positive and statistically significant coefficients in columns (2) and (5) and decreases the likelihood of a school being situated nearby (the negative and significant coefficients in columns (1) and (4). Importantly, the magnitude of the shrine effect is noticeably higher for girls’ schools, indicating a gender dimension to the problem of school access. Table 10. Historical shrines and physical access to schooling in 2007. . (1) . (2) . (3) . (4) . (5) . (6) . . Proportion of boys public schools . Proportion of girls public schools . . |$<$|2 km . 2–5 km . |$>$|5 km . |$<$|2 km . 2–5 km . |$>$|5 km . Panel A: public schools . Shrines per thousand persons -0.241** 0.214*** 0.0273 -0.432*** 0.397*** 0.0353 (0.0975) (0.0773) (0.0521) (0.159) (0.134) (0.0689) Mean outcome 0.928 0.051 0.021 0.910 0.056 0.034 Observations 123 123 123 123 123 123 Adjusted R-squared 0.451 0.344 0.335 0.419 0.332 0.345 Proportion of boys private schools Proportion of girls private schools |$<$|2 km 2–5 km |$>$|5 km |$<$|2 km 2–5 km |$>$|5 km Panel B: private schools Shrines per thousand persons -0.614*** 0.375** 0.240 -0.613*** 0.388** 0.226 (0.190) (0.158) (0.266) (0.204) (0.150) (0.271) Mean outcome 0.779 0.079 0.143 0.778 0.078 0.145 Observations 123 123 123 123 123 123 Adjusted R-squared 0.470 0.351 0.384 0.456 0.357 0.368 Controls Yes Yes Yes Yes Yes Yes . (1) . (2) . (3) . (4) . (5) . (6) . . Proportion of boys public schools . Proportion of girls public schools . . |$<$|2 km . 2–5 km . |$>$|5 km . |$<$|2 km . 2–5 km . |$>$|5 km . Panel A: public schools . Shrines per thousand persons -0.241** 0.214*** 0.0273 -0.432*** 0.397*** 0.0353 (0.0975) (0.0773) (0.0521) (0.159) (0.134) (0.0689) Mean outcome 0.928 0.051 0.021 0.910 0.056 0.034 Observations 123 123 123 123 123 123 Adjusted R-squared 0.451 0.344 0.335 0.419 0.332 0.345 Proportion of boys private schools Proportion of girls private schools |$<$|2 km 2–5 km |$>$|5 km |$<$|2 km 2–5 km |$>$|5 km Panel B: private schools Shrines per thousand persons -0.614*** 0.375** 0.240 -0.613*** 0.388** 0.226 (0.190) (0.158) (0.266) (0.204) (0.150) (0.271) Mean outcome 0.779 0.079 0.143 0.778 0.078 0.145 Observations 123 123 123 123 123 123 Adjusted R-squared 0.470 0.351 0.384 0.456 0.357 0.368 Controls Yes Yes Yes Yes Yes Yes Notes: The table shows the impact of historic shrines on physical access to primary schools in a contemporary year (2007). The access measures are diaggregated by type of school (public versus private) and by gender (boys versus girls). Columns (1)–(3) of panel A are the proportion of boys public schools within a <2 km, 2 km–5 km and |$>$|5 km radius, respectively. Columns (4)–(6) of panel A are the proportion of girls public schools within a <2 km, 2 km-–5 km and |$>$|5 km radius, respectively. Columns (1)–(3) of panel B are the proportion of boys private schools within a <2 km, 2 km-–5 km and |$>$|5 km radius, respectively. Finally, Columns (4)–(6) of panel B are the proportion of girls private schools within a <2 km, 2 km–5 km and |$>$|5 km radius, respectively. The unit of observation is tehsil. All regressions use a consistent sample of 123 tehsils measured at the 2007 boundaries. The controls are region dummies, latitude, longitude, elevation, population density, and incidence of landlessness. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. **Significant at 5%; ***Significant at 1%. Open in new tab Table 10. Historical shrines and physical access to schooling in 2007. . (1) . (2) . (3) . (4) . (5) . (6) . . Proportion of boys public schools . Proportion of girls public schools . . |$<$|2 km . 2–5 km . |$>$|5 km . |$<$|2 km . 2–5 km . |$>$|5 km . Panel A: public schools . Shrines per thousand persons -0.241** 0.214*** 0.0273 -0.432*** 0.397*** 0.0353 (0.0975) (0.0773) (0.0521) (0.159) (0.134) (0.0689) Mean outcome 0.928 0.051 0.021 0.910 0.056 0.034 Observations 123 123 123 123 123 123 Adjusted R-squared 0.451 0.344 0.335 0.419 0.332 0.345 Proportion of boys private schools Proportion of girls private schools |$<$|2 km 2–5 km |$>$|5 km |$<$|2 km 2–5 km |$>$|5 km Panel B: private schools Shrines per thousand persons -0.614*** 0.375** 0.240 -0.613*** 0.388** 0.226 (0.190) (0.158) (0.266) (0.204) (0.150) (0.271) Mean outcome 0.779 0.079 0.143 0.778 0.078 0.145 Observations 123 123 123 123 123 123 Adjusted R-squared 0.470 0.351 0.384 0.456 0.357 0.368 Controls Yes Yes Yes Yes Yes Yes . (1) . (2) . (3) . (4) . (5) . (6) . . Proportion of boys public schools . Proportion of girls public schools . . |$<$|2 km . 2–5 km . |$>$|5 km . |$<$|2 km . 2–5 km . |$>$|5 km . Panel A: public schools . Shrines per thousand persons -0.241** 0.214*** 0.0273 -0.432*** 0.397*** 0.0353 (0.0975) (0.0773) (0.0521) (0.159) (0.134) (0.0689) Mean outcome 0.928 0.051 0.021 0.910 0.056 0.034 Observations 123 123 123 123 123 123 Adjusted R-squared 0.451 0.344 0.335 0.419 0.332 0.345 Proportion of boys private schools Proportion of girls private schools |$<$|2 km 2–5 km |$>$|5 km |$<$|2 km 2–5 km |$>$|5 km Panel B: private schools Shrines per thousand persons -0.614*** 0.375** 0.240 -0.613*** 0.388** 0.226 (0.190) (0.158) (0.266) (0.204) (0.150) (0.271) Mean outcome 0.779 0.079 0.143 0.778 0.078 0.145 Observations 123 123 123 123 123 123 Adjusted R-squared 0.470 0.351 0.384 0.456 0.357 0.368 Controls Yes Yes Yes Yes Yes Yes Notes: The table shows the impact of historic shrines on physical access to primary schools in a contemporary year (2007). The access measures are diaggregated by type of school (public versus private) and by gender (boys versus girls). Columns (1)–(3) of panel A are the proportion of boys public schools within a <2 km, 2 km–5 km and |$>$|5 km radius, respectively. Columns (4)–(6) of panel A are the proportion of girls public schools within a <2 km, 2 km-–5 km and |$>$|5 km radius, respectively. Columns (1)–(3) of panel B are the proportion of boys private schools within a <2 km, 2 km-–5 km and |$>$|5 km radius, respectively. Finally, Columns (4)–(6) of panel B are the proportion of girls private schools within a <2 km, 2 km–5 km and |$>$|5 km radius, respectively. The unit of observation is tehsil. All regressions use a consistent sample of 123 tehsils measured at the 2007 boundaries. The controls are region dummies, latitude, longitude, elevation, population density, and incidence of landlessness. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. **Significant at 5%; ***Significant at 1%. Open in new tab Next, we test whether the quality of school infrastructure is worse in shrine dense regions. Drawing on a recent database on school facilities, we focus on two infrastructure measures for which tehsil-level data are available: electrification and the presence of boundary walls. In Table 11, we present the results from a regression of proportion of schools with electricity (columns (1)–(2)) and proportion of schools with boundary walls (columns (3)–(4)) on our historic shrine measure, using the same set of controls included in Table 10. The results show that shrine dominated regions have significantly lower proportion of schools with access to electricity and boundary walls. The results are stronger if we control for the area of the tehsil (see columns (2) and (4)). Table 11. Historical shrines and school facilities in 2011. . (1) . (2) . (3) . (4) . . Proportion of schools with electricity . Proportion of schools with boundary walls . Shrines per thousand persons -0.434* -0.434* -0.177* -0.186** (0.238) (0.240) (0.0926) (0.0867) Controls: Yes Yes Yes Yes Mean outcome 0.641 0.641 0.869 0.869 Observations 123 123 123 123 Adjusted R-squared 0.380 0.375 0.428 0.441 . (1) . (2) . (3) . (4) . . Proportion of schools with electricity . Proportion of schools with boundary walls . Shrines per thousand persons -0.434* -0.434* -0.177* -0.186** (0.238) (0.240) (0.0926) (0.0867) Controls: Yes Yes Yes Yes Mean outcome 0.641 0.641 0.869 0.869 Observations 123 123 123 123 Adjusted R-squared 0.380 0.375 0.428 0.441 Notes: The table shows the impact of historic shrines on the quality of school infrastructure in a contemporary year (2011). The unit of observation is tehsil. All regressions use a consistent sample of 123 tehsils measured at the 2007 boundaries. The controls are region dummies, latitude, longitude, elevation and incidence of landlessness. Columns (2) and (4) include land area in (square kilometres) as an additional control. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. *Significant at 10%; **Significant at 5%. Open in new tab Table 11. Historical shrines and school facilities in 2011. . (1) . (2) . (3) . (4) . . Proportion of schools with electricity . Proportion of schools with boundary walls . Shrines per thousand persons -0.434* -0.434* -0.177* -0.186** (0.238) (0.240) (0.0926) (0.0867) Controls: Yes Yes Yes Yes Mean outcome 0.641 0.641 0.869 0.869 Observations 123 123 123 123 Adjusted R-squared 0.380 0.375 0.428 0.441 . (1) . (2) . (3) . (4) . . Proportion of schools with electricity . Proportion of schools with boundary walls . Shrines per thousand persons -0.434* -0.434* -0.177* -0.186** (0.238) (0.240) (0.0926) (0.0867) Controls: Yes Yes Yes Yes Mean outcome 0.641 0.641 0.869 0.869 Observations 123 123 123 123 Adjusted R-squared 0.380 0.375 0.428 0.441 Notes: The table shows the impact of historic shrines on the quality of school infrastructure in a contemporary year (2011). The unit of observation is tehsil. All regressions use a consistent sample of 123 tehsils measured at the 2007 boundaries. The controls are region dummies, latitude, longitude, elevation and incidence of landlessness. Columns (2) and (4) include land area in (square kilometres) as an additional control. Shrines per thousand persons is the number of historically important shrines per thousand persons in 1931. Standard errors clustered by tehsil in parentheses, unless otherwise indicated. *Significant at 10%; **Significant at 5%. Open in new tab Finally, since the mid-1980s political elites have used the construction of new schools as a source of rents, preferring brick and mortar investments that afford opportunities for lucrative construction contracts. Such perverse political incentives result in an inefficient dispersion of school capacity in the form of reduced school size (i.e. fewer pupils per school) rather than better educational provision. As Gazdar (2000, p. 31) argues: the deterioration in standards has been blamed, at least partly, on the large expansion in the school building programme, particularly since the mid-1980s...The growth in the number of schools might have compromised quality by spreading administrative capacity too thinly. In many instances the construction of a new school meant the reduction in the size of another school, as children from formerly school-less villages withdrew from the main school. The reduction in size may have adversely affected quality. Accordingly, we investigate whether schooling resources were spread more thinly in shrine-dominated regions. To do so, we construct a quantitative measure of school size, compiled using annual data on pupils per school for the post-coup period (1982–1998), and regress it on our shrine measure along with our baseline controls. Results are presented in Table 12, columns (1)–(3), for three school types: primary, middle and high. As the results show, the coefficient on our historic shrine measure is negative and statistically significant across all specifications (columns (1)–(3)). Thus, relative to non-shrine areas, tehsils with greater shrine concentration had fewer pupils per school at all levels, including primary, middle and high schools. Table 12. Historical shrines and post-coup school size. . (1) . (2) . (3) . . Primary . Middle . High . Shrines per thousand persons -297.9*** -557.9*** -864.7*** (72.84) (204.4) (264.2) Baseline controls High rainfall dummy (1931) Y Y Y Agrarian land classifications (1931) Y Y Y Proportion minorities (1931) Y Y Y Distance to Delhi (km) Y Y Y Mean outcome 77.169 233.160 562.668 Observations 924 924 924 Adjusted R-squared 0.471 0.438 0.443 . (1) . (2) . (3) . . Primary . Middle . High . Shrines per thousand persons -297.9*** -557.9*** -864.7*** (72.84) (204.4) (264.2) Baseline controls High rainfall dummy (1931) Y Y Y Agrarian land classifications (1931) Y Y Y Proportion minorities (1931) Y Y Y Distance to Delhi (km) Y Y Y Mean outcome 77.169 233.160 562.668 Observations 924 924 924 Adjusted R-squared 0.471 0.438 0.443 Notes: The table shows the impact of historic shrines on post-coup school size (pupils per school) during the period 1982–1998. The underlying data for school size comes from the annual reports of Punjab Development Statistics. Column (1) is the number of primary school pupils per primary school. Column (2) is the number of middle school pupils per middle school. Finally, column (3) is the number of high school pupils per high school. The unit of observation is tehsil-year. We do not include tehsil fixed effects in these regressions since our explanatory variable, shrines per thousand persons, remains fixed for a tehsil over time. However, all regressions do include year fixed effects. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.***Significant at 1%. Open in new tab Table 12. Historical shrines and post-coup school size. . (1) . (2) . (3) . . Primary . Middle . High . Shrines per thousand persons -297.9*** -557.9*** -864.7*** (72.84) (204.4) (264.2) Baseline controls High rainfall dummy (1931) Y Y Y Agrarian land classifications (1931) Y Y Y Proportion minorities (1931) Y Y Y Distance to Delhi (km) Y Y Y Mean outcome 77.169 233.160 562.668 Observations 924 924 924 Adjusted R-squared 0.471 0.438 0.443 . (1) . (2) . (3) . . Primary . Middle . High . Shrines per thousand persons -297.9*** -557.9*** -864.7*** (72.84) (204.4) (264.2) Baseline controls High rainfall dummy (1931) Y Y Y Agrarian land classifications (1931) Y Y Y Proportion minorities (1931) Y Y Y Distance to Delhi (km) Y Y Y Mean outcome 77.169 233.160 562.668 Observations 924 924 924 Adjusted R-squared 0.471 0.438 0.443 Notes: The table shows the impact of historic shrines on post-coup school size (pupils per school) during the period 1982–1998. The underlying data for school size comes from the annual reports of Punjab Development Statistics. Column (1) is the number of primary school pupils per primary school. Column (2) is the number of middle school pupils per middle school. Finally, column (3) is the number of high school pupils per high school. The unit of observation is tehsil-year. We do not include tehsil fixed effects in these regressions since our explanatory variable, shrines per thousand persons, remains fixed for a tehsil over time. However, all regressions do include year fixed effects. Standard errors clustered by tehsil in parentheses, unless otherwise indicated.***Significant at 1%. Open in new tab Demand-Side Explanations Shrine elites could also have restrained literacy in their areas by affecting the conditions that shape the demand for education. We examine whether this was the case by focusing on two leading candidates for a demand-side explanation: urbanisation and economic migration. Both can increase exposure to non-agricultural opportunities, change the returns to schooling, and increase the demand for education. In Online Appendix Table A.13, we include a time-varying measure of urbanization (proportion of the population classified as urban) as an additional control in our baseline specification represented by equation (1). To construct a historic panel on urbanization, we used data from both Colonial and District Census Reports. As expected, urbanization is a positive and statistically significant correlate of literacy. However, its inclusion in the model does not eliminate our shrine effect, even if it somewhat reduces its magnitude. The coefficient on shrine interaction continues to be negative and statistically significant at 5% level. Additionally, when we re-estimate our baseline specification using urbanization as the dependent variable, the coefficient estimate on the shrine interaction is negative but statistically insignificant (see Online Appendix Table A.14). There is thus no evidence to suggest that shrine-dense regions witnessed a differential growth of urbanization in the period after the coup-induced policy shock. Together, these results indicate that the effect of shrines on post-coup literacy is, at best, only partially mediated through urbanization. Next, we test whether our results are due to systematic differences in patterns of migration between high- and low-shrine regions. Accordingly, in Online Appendix Table A.15, we regress two contemporary proxies for economic migration, the proportion of households receiving remittances from abroad and from within the country, on our shrine measure.33 The results show that there is no statistically significant difference between high- and low-shrine tehsils in either models of remittances from abroad (column (1)) or from within the country (column (2)). Overall, the weight of evidence presented above points towards a supply-side explanation. We show that, rather than suppressing education by restraining public spending or blocking the construction of new schools, shrine elites seem to have deployed a more sophisticated strategy that rested on making the educational system defective by influencing the location of schools, quality of school infrastructure, and school size. This is unsurprising in the context of institutional changes brought by General Zia’s military coup, which by assigning elected politicians greater control over public goods provision, made the latter strategy (i.e. defective provision) more incentive-compatible for shrine elites. With more financial resources available for political capture, restraining spending or school construction is now politically suboptimal. 8. Conclusion This paper provides novel evidence on the impact of religious elites on economic development using a rich empirical setting in the second largest Muslim-majority country of the world, Pakistan. Focusing on elites whose religious authority is derived from Sufi shrines, we examine their role in suppressing literacy in the wake of the 1977 military coup that devolved control over public goods provision to elected politicians. Shrine elites—with their increasing presence in politics and their greater aversion to literacy—suppressed its expansion in shrine dense regions in the post-coup period. They were able to do so by making existing school provision defective. Our results shed light on an under-studied dimension of Islam: the impact of Islamic mystical orders (Sufism) on long-run development outcomes. Sufism has long been understood as a central force in the transmission of Islam across vast swathes of Africa, Asia, and the Middle East. The missionary activities of mystical Islamic brotherhoods not only extended the borders of Islam but also left a deep imprint on local social, economic, and political structures. Our paper offers a first systematic empirical enquiry on the longue durée impact of this aspect of Islam. While our findings emanate from a specific historical context, they have broader relevance for other Muslim societies where mystical religious orders are an important part of public life. Morocco offers a case in point where powerful Sufi orders are re-asserting their influence on the political stage and lending legitimacy to authoritarian rule (Werenfels 2014). The impact of such religious elites in different societies will depend on how they are structurally positioned within the prevailing power structure. In contexts where the power of these religious elites is combined with political influence and landed power, and where they are able to influence the provision of public goods, we can expect to see a negative effect on literacy. However, each context is different. As Starrett (1998, p. 235) argues for the Egyptian case, Islam has been “put to work” through the “changing intersections of power, interest, and circumstance”. Future scholarship needs to take into account the considerable heterogeneity in the relationship between Islam, politics and development across Muslim societies (Cammett and Luong 2014). Beyond the world of Islam, our paper also has relevance for long-standing debates on the role of religion in development. In this regard, our analysis highlights the importance of considering the incentives and capacities of religious elites in shaping development outcomes. Furthermore, our empirical framework echoes the important insight of Avner Grief that the impact of informal institutions, including religion, is shaped by historical context and through the interaction between informal and formal institutional structures. Finally, our work has concrete implications for public policy on education. A report prepared by a global task force on education concluded that “fixing the learning crisis will take systemic reform stretching beyond the education sector”.34 This paper has shown that elite incentives and institutional structures that govern public goods provision can together influence the ecosystem of schooling. This is especially pertinent for countries facing a major education emergency. With its 23 million out-of-school children, Pakistan ranks as second (after Nigeria) in the list of top ten countries with the “highest out-of-school populations” (UNESCO 2014). It is also described as “severely lagging” in its goal of achieving universal education. In this milieu, educational expansion is not just about scaling up public spending. It also requires addressing structural inequalities—in this case defined by the powerful configuration of religions and politics. Acknowledgments This research was completed with financial support from an award of the research Competitive Grants Program, Pakistan Strategy Support Program, and International Food Policy Research Institute (IFPRI). Mirza recognizes additional support from the Fund for Scientific Research (FNRS) in Belgium. We are grateful to Cathy Boone, Latika Chaudhry, Ferdinand Eibl, James Fenske, Doug Gollin, Saumitra Jha, Ghazala Mansuri, Muhammad Meki, Simon Quinn, and Jonathan Temple for their detailed comments on an earlier draft version, and to David Orden and Sohail Chaudhry for administrative support. For their constructive feedback, we are also thankful to seminar participants at the World Economic History Congress 2018 Comparative Political Economy Workshops at LSE and Nuffield College, Oxford, and at the University of Oxford, Cambridge, and Namur, Belgium. The usual disclaimers apply. Notes The editor in charge of this paper was Paola Giuliano. Footnotes 1. Role of elites is implicit in most institutional analyses whether it is the differential impact of colonial rule (Acemoglu, Johnson, and Robinson 2001; Huillery 2009), the legacy of historic inequality (Engerman and Sokoloff 1997), the institutional impact of overseas trade engagement (Acemoglu, Johnson, and Robinson 2005; Jha 2013), or the impact of traditional structures of authority built around chieftancy in Africa (Acemoglu, Reed, and Robinson 2014b). 2. Conceptually, this allows us to compare the growth of literacy in tehsils before and after the shock, and to assess how this trajectory of literacy is shaped by shrine-based religious power. 3. Importantly, these literacy differences between high- and low-shrine tehsils persist even after the restoration of electoral democracy in late 1980s. 4. This paper also complements a body of scholarship arguing that educational differences across countries are grounded in history (Galor, Moav, and Vollrath 2009; Gallego 2010; Gallego and Woodberry 2010; Acemoglu, Gallego, and Robinson 2014a). Colonial investments on education are shown to have lingering effects on human capital development in Africa (Huillery 2009). 5. A partial exception is Ticku et al. (2018), who establish that Mughal rulers in medieval India were more likely to desecrate Hindu temples in periods of economic downturns. Relatedly, Barro and McCleary (2003), and Barro, McCleary, and McQuoid (2011) study the socio-economic determinants of sainthood in Christianity. 6. Focusing on the shrine of Baba Farid, Eaton argues that: “Beyond the urban settlements lying along Punjab’s rivers lay vast tracts of sparsely populated land, the so-called barr country, situated between the province’s five rivers. Although scanty rainfall hindered agricultural operations in this region, the barr country was well suited for pastoral nomadism” (p. 210). 7. Census of India 1921, Volume XV, Part I, Punjab and Delhi, 1923, p. 33. Successive Irrigation Commission Reports noted that Punjab’s plains were generally fertile with high levels of soil suitability. 8. The locus of shrine power is in its “informal organization and the extra-associational norms situated in the community” (Malik and Malik 2017, p. 1822). 9. Prior to the British, “jagirs and pensions offered by the state... were non-portable and at the mercy of the government” (Roseberry 1987, p. 81). 10. In the 1920 and 1946 provincial elections, held under British rule, roughly 19% of total rural Muslim constituencies were represented by shrine families. Nearly a century later, their corresponding representation in the national parliament only fell to 16% in 2013. 11. Other shrine notables whose power is similarly spread over multiple generations includes (districts in brackets): the Syeds of Shah Jewna (Jhang), Qureshis of Shah Rukn-e-Alam (Multan), Makhdooms of Jamal Din Wali (Rahimyar Khan), Chishtis of Pakpattan, Kirmani Syeds of Daud Bandagi (Okara), and Khwajas of Taunsa Sharif (D. G. Khan), among others. 12. By facilitating labour mobility, education can offer exit options for people tied in a dependent social and economic structure Galor, Moav, and Vollrath (2009). This can affect both landed and shrine elites. However, aside from its impact on labour mobility, education can threaten the religious authority of shrine elites. 13. As argued in Malik and Malik (2017, p.41): “Compared to the landed class, the pirs seem to possess a stronger sense of class consciousness allowing them to protect their group solidarity more cohesively in and outside the parliament.” 14. The MNAs are also usually required to submit cost estimates and propose the implementation agency. See Afzal (2009) for further details. 15. MNA stands for Member of National Assembly and MPA stands for Member of Provincial Assembly. 16. The administrative arrangements and mandate of development spending remained the same even if it was packaged under different political banners. As Afzal (2009) notes: “Development funds were allocated to MNAs under the Peoples Programme in 1988–90 and 1993–97, under the Tameer-e-Watan Programme in 1991–93 and 1998–2000, and under the Khushal Pakistan Programme from 2002-8.” 17. Such information was typically documented in a separate subsection entitled, “Religious fairs and festivals”. 18. MICS Punjab is a representative household survey designed to produce statistically reliable estimates on key social indicators at the tehsil-level covering both urban and rural areas defined by the 1998 population census. Additional information accessible at http://bos.gop.pk/mics. 19. For the MICS data, we exclude Quranic reading, if this was the only response. 20. The statistics are compiled by the Bureau of Statistics, Government of Punjab. Recent reports are available for download from http://www.bos.gop.pk/developmentstat. 21. The data are compiled by the School Education Department, Government of Punjab, and made available through the following website: https://open.punjab.gov.pk/schools/. 22. Each control is described in detail as and when it is introduced in the analysis. 23. Given that our proxy for the power of religious elites, |$\rm{Shrines}_{i}^{\rm{hist}}$|⁠, in equation (2) is time-invariant and also because equation (2) includes year and tehsil fixed effects, the estimated coefficients in |$\Gamma _{t}$| have to be measured relative to a base time period, which we take to be the first census decade in our panel beginning in 1901. 24. Whilst we would have preferred to have carried out tests for differences in trends, data limitations have meant that we were only able to test for differences in levels. 25. The identity database is maintained by the National Database and Registration Authority (NADRA) and consists of over 96 million records. Applicants for the identity card are required to select a profession from the detailed list of occupational categories. Since researchers do not have access to the database, NADRA was requested to compute the ratio for all tehsils of Punjab. 26. This is calculated by multiplying 0.1 with the coefficient on the shrine interaction in column (5) of Table 2, which is -0.957. 27. This is arrived through the following calculation: 100 |$\times$| (0.0957/0.522). 28. We replace the rainfall measure with elevation since both are highly correlated. 29. Results on functional forms available upon request. 30. In shrine dense regions, election candidates are dependent on the support of influential shrine elites. A prominent example in this regard is the shrine of Sheikh Fazil in Sahiwal whose guardians desist from directly contesting elections, but their support is decisively important for candidates in several neighbouring constituencies. For more illustrations and analysis, see Malik and Malik (2017). 31. Affiliation with a notable shrine becomes a vital political asset in a milieu where, “it is not wealth alone, but wealth plus either kinship or spiritual prestige, or both, that gives political power” (Lieven 2012, p.137). 32. Development spending data are only available at a highly aggregate level (i.e. provincial or district-level). 33. Unfortunately, tehsil-level data on outmigration are unavailable. Data for our dependent variables come from MICS Household Survey 2007-08. See Section 3.2 for a description of MICS. 34. https://www.cgdev.org/sites/default/files/archive/doc/full_text/CGDReports/3120290/schooling-is-not-education.html References Acemoglu Daron , Gallego Francisco A., Robinson James A. ( 2014a ). “Institutions, Human Capital, and Development.” Annual Review of Economics , 6 , 875 – 912 . Google Scholar Crossref Search ADS WorldCat Acemoglu Daron , Johnson Simon, Robinson James ( 2005 ). “The Rise of Europe: Atlantic Trade, Institutional Change, and Economic Growth.” American Economic Review , 95 ( 3 ), 546 – 579 . Google Scholar Crossref Search ADS WorldCat Acemoglu Daron , Johnson Simon, Robinson James A. ( 2001 ). “The Colonial Origins of Comparative Development: An Empirical Investigation.” American Economic Review , 91 ( 5 ), 1369 – 1401 . Google Scholar Crossref Search ADS WorldCat Acemoglu Daron , Reed Tristan, Robinson James A. ( 2014b ). “Chiefs: Economic Development and Elite Control of Civil Society in Sierra Leone.” Journal of Political Economy , 122 , 319 – 368 . Google Scholar Crossref Search ADS WorldCat Acemoglu Daron , Robinson James A. ( 2006 ). Economic Backwardness in Political Perspective . American Political Science Review , 100 , 115 – 131 . Google Scholar Crossref Search ADS WorldCat Albinia Alice ( 2008 ). Empires of the Indus: The Story of a River . John Murray , London . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Ali Imran ( 2014 ). The Punjab under Imperialism , 1885 – 1947 . Princeton University Press , NJ . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Altonji Joseph G. , Elder Todd E., Taber Christopher R. ( 2005 ). “Selection on Observed and Unobserved Variables: Assessing the Effectiveness of Catholic Schools.” Journal of Political Economy , 113 , 151 – 184 . Google Scholar Crossref Search ADS WorldCat Ansari Sarah F.D. ( 1992 ). Sufi Saints and State Power: The Pirs of Sind , 1843 – 1947 , Vol. 50 . Cambridge University Press . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Aziz Khursheed Kamal ( 2001 ). Religion, Land and pPolitics in Pakistan: A Study of Piri-Muridi . Vanguard . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Banerjee Abhijit , Iyer Lakshmi ( 2005 ). “History, Institutions, and Economic Performance: The Legacy of Colonial Land Tenure Systems in India.” American Economic Review , 95 ( 4 ), 1190 – 1213 . Google Scholar Crossref Search ADS WorldCat Barro Robert J. , McCleary Rachel ( 2003 ). “Religion and Economic Growth.” NBER Working Paper No. 9682 . OpenURL Placeholder Text WorldCat Barro Robert J. , McCleary Rachel M., McQuoid Alexander ( 2011 ). “Economics of Sainthood (APreliminary Investigation).” In The Oxford Handbook of the Economics of Religion , edited by McCleary Rachel M.. Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Bazzi Samuel , Koehler-Derrick Gabriel, Marx Benjamin ( 2020 ). “The Institutional Foundations of Religious Politics: Evidence from Indonesia.” The Quarterly Journal of Economics , 135 , 845 – 911 . Google Scholar Crossref Search ADS WorldCat Becker Sascha O. , Woessmann Ludger ( 2009 ). “Was Weber Wrong? A Human Capital Theory of Protestant Economic History.” The Quarterly Journal of Economics , 124 , 531 – 596 . Google Scholar Crossref Search ADS WorldCat Belmonte Alessandro , Rochlitz Michael ( 2019 ). “The Political Economy of Collective Memories: Evidence from Russian Politics.” Journal of Economic Behavior & Organization , 168 , 229 – 250 . Google Scholar Crossref Search ADS WorldCat Bharadwaj Prashant , Khwaja Asim, Mian Atif ( 2008 ). “The Big March: Migratory Flows after the Partition of India.” Economic and Political Weekly , 2008 , 39 – 49 . Google Scholar OpenURL Placeholder Text WorldCat Bharadwaj Prashant , Khwaja Asim I., Mian Atif ( 2015 ). “Population Exchange and its Impact on Literacy, Occupation and Gender–Evidence from the Partition of India.” International Migration , 53 , 90 – 106 . Google Scholar Crossref Search ADS WorldCat Bharadwaj Prashant , Ali Mirza Rinchan ( 2019 ). “Displacement and Development: Long Term Impacts of Population Transfer in India.” Explorations in Economic History , 73 , 101273 . Google Scholar Crossref Search ADS WorldCat Birdsall Nancy , Malik Adeel, Vaishnav Milan ( 2005 ). Poverty and the Social Sectors: The World Bank in Pakistan 1990–2003 . Open Access Publications from University of Oxford . Burki Shahid Javed ( 1973 ). “Rapid Population Growth and Urbanization: the Case of Pakistan.” Pakistan Economic and Social Review , 11 , 239 – 276 . Google Scholar OpenURL Placeholder Text WorldCat Calvert Hubert ( 1925 ). The Size and Distribution of Agricultural Holdings in the Punjab , Vol. 4 . C. & M. Gazette Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Calvi Rossella , Hoehn-Velasco Lauren, Mantovanelli Federico G. ( 2020 ). “The Protestant Legacy: Missions, Gender, and Human Capital in India.” Journal of Human Resources , 18 , 0919 – 10437R2 . Google Scholar OpenURL Placeholder Text WorldCat Cammett Melani , Luong Pauline Jones ( 2014 ). “Is there an Islamist Political Advantage?” Annual Review of Political Science , 17 , 187 – 206 . Google Scholar Crossref Search ADS PubMed WorldCat Campante Filipe , Yanagizawa-Drott David ( 2015 ). “Does Religion Affect Economic Growth and Happiness? Evidence from Ramadan.” The Quarterly Journal of Economics , 130 , 615 – 658 . Google Scholar Crossref Search ADS WorldCat Carvalho Jean-Paul ( 2013 ). “Veiling.” The Quarterly Journal of Economics , 128 , 337 – 370 . Google Scholar Crossref Search ADS WorldCat Carvalho Jean-Paul , Iyer Sriya, Rubin Jared ( 2019 ). Advances in the Economics of Religion . Springer . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Casson Mark C. , Della Giusta Marina, Kambhampati Uma S. ( 2010 ). “Formal and Informal Institutions and Development.” World Development , 38 , 137 – 141 . Google Scholar Crossref Search ADS WorldCat Chaney Eric ( 2013 ). “Revolt on the Nile: Economic Shocks, Religion, and Political Power.” Econometrica , 81 , 2033 – 2053 . Google Scholar OpenURL Placeholder Text WorldCat Chaudhary Latika , Rubin Jared ( 2011 ). “Reading, Writing, and Religion: Institutions and Human Capital Formation.” Journal of Comparative Economics , 39 , 17 – 33 . Google Scholar Crossref Search ADS WorldCat Cinnirella Francesco , Hornung Erik ( 2016 ). “Landownership Concentration and the Expansion of Education.” Journal of Development Economics , 121 , 135 – 152 . Google Scholar Crossref Search ADS WorldCat Clingingsmith David , Ijaz Khwaja Asim, Kremer Michael ( 2009 ). “Estimating the Impact of the Hajj: Religion and Tolerance in Islam’s Global Gathering.” The Quarterly Journal of Economics , 124 , 1133 – 1170 . Google Scholar Crossref Search ADS WorldCat Eaton Richard ( 1984 ). “The Political and Religious Authority of the Shrine of Baba Farid in Pakpattan, Punjab.” In Moral Conduct and Authority: The Place of Adab in South Asian Islam , edited by Metcalf Barbara. University of California Press , pp. 333 – 56 . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Eaton Richard M. ( 2009 ). “Shrines, Cultivators, and Muslim ‘Conversion’ in Punjab and Bengal, 1300–1700.” The Medieval History Journal , 12 , 191 – 220 . Google Scholar Crossref Search ADS WorldCat Engerman Stanley L. , Sokoloff Kenneth L. ( 1997 ). “Factor Endowments, Institutions, and Differential Paths of Growth Among New World Economies.” In How Latin America Fell Behind , edited by Haber Stephen. Stanford University Press . pp. 260 – 304 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Fouka Vasiliki , Voth Hans-Joachim ( 2013 ). “Reprisals Remembered: German-Greek Conflict and Car Sales During the Euro Crisis.” CEPR Discussion Paper No. DP9704 . OpenURL Placeholder Text WorldCat Fouka Vicky ( 2020 ). “Time-Varying Persistence.” Working paper, Broadstreet Blog . OpenURL Placeholder Text WorldCat Gallego Francisco A. ( 2010 ). “Historical Origins of Schooling: The Role of Democracy and Political Decentralization.” The Review of Economics and Statistics , 92 , 228 – 243 . Google Scholar Crossref Search ADS WorldCat Gallego Francisco A. , Woodberry Robert ( 2010 ). “Christian Missionaries and Education in Former African Colonies: How Competition Mattered.” Journal of African Economies , 19 , 294 – 329 . Google Scholar Crossref Search ADS WorldCat Galor Oded , Moav Omer, Vollrath Dietrich ( 2009 ). “Inequality in Landownership, the Emergence of Human-Capital Promoting Institutions, and the Great Divergence.” The Review of Economic Studies , 76 , 143 – 179 . Google Scholar Crossref Search ADS PubMed WorldCat Gazdar Haris ( 2000 ). “State, Community, and Public Schooling—-’A Political Economy of Universal Education in Pakistan.” Working paper , London School of Economics , London . Google Scholar Gilmartin David ( 1979 ). “Religious Leadership and the Pakistan Movement in the Punjab.” Modern Asian Studies , 13 , 485 – 517 . Google Scholar Crossref Search ADS WorldCat Gilmartin David ( 1988 ). Empire and Islam: Punjab and the Making of Pakistan . University of California Press Berkeley . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Grover B. R. ( 1960 ). “Classification of Agrarian Land under Akbar.” Proceedings of the Indian History Congress , 23 , 198 – 209 . Google Scholar OpenURL Placeholder Text WorldCat Guiso Luigi , Sapienza Paola, Zingales Luigi ( 2006 ). “Does Culture Affect Economic Outcomes?” Journal of Economic Perspectives , 20 ( 2 ), 23 – 48 . Google Scholar Crossref Search ADS WorldCat Huillery Elise ( 2009 ). “History Matters: The Long-Term Impact of Colonial Public Investments in French West Africa.” American Economic Journal: Applied Economics , 1 , 176 – 215 . Google Scholar OpenURL Placeholder Text WorldCat Iannaccone Laurence R. ( 1998 ). “Introduction to the Economics of Religion.” Journal of Economic Literature , 36 , 1465 – 1495 . Google Scholar OpenURL Placeholder Text WorldCat Iyer Sriya ( 2016 ). “The New Economics of Religion.” Journal of Economic Literature , 54 , 395 – 441 . Google Scholar Crossref Search ADS WorldCat Jha Saumitra ( 2013 ). “Trade, Institutions, and Ethnic Tolerance: Evidence from South Asia.” American Political Science Review , 107 , 806 – 832 . Google Scholar Crossref Search ADS WorldCat Kahn-Lang Ariella , Lang Kevin ( 2020 ). “The Promise and Pitfalls of Differences-in-Differences: Reflections on 16 and Pregnant and other Applications.” Journal of Business & Economic Statistics , 38 , 613 – 620 . Google Scholar Crossref Search ADS WorldCat Kuran Timur ( 2018 ). “Islam and Economic Performance: Historical and Contemporary Links.” Journal of Economic Literature , 56 , 1292 – 1359 . Google Scholar Crossref Search ADS WorldCat Lieven Anatol ( 2012 ). Pakistan: A Hard Country . PublicAffairs . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Lyall Darling Malcolm ( 1928 ). The Punjab Peasant in Prosperity and Debt . Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Malik Adeel , Malik Tahir ( 2017 ). “Pīrs and Politics in Punjab, 1937–2013.” Modern Asian Studies , 51 , 1818 – 1861 . Google Scholar Crossref Search ADS WorldCat Malik Rabia ( 2020 ). “(A)Political Constituency Development Funds: Evidence from Pakistan.” British Journal of Political Science , 51 ( 3 ), 1 – 18 . Google Scholar OpenURL Placeholder Text WorldCat Martin Nicolas ( 2015 ). Politics, Landlords and Islam in Pakistan . Routledge . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Mills C. Wright , Gerth Hans ( 1953 ). Character and Social Structure . Harcourt, Brace & World Inc. Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Mirza Rinchan Ali ( 2020 ). “Displacement and Development: Evidence from Pakistan through Partition.” D.Phil Thesis, University of Oxford . OpenURL Placeholder Text WorldCat Mufti Masood ( 2020 ). Do-Minar: A Reportoire . Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Nizami Khaleeq Ahmad ( 1953 ). Tareekh-e-Mashaikh-e-Chisht. Delhi: Idara-e-Adbiat Dilli . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC North Douglass C. , Wallis John Joseph, Weingast Barry R. et al. ( 2009 ). Violence and Social Orders: A Conceptual Framework for Interpreting Recorded Human History . Cambridge University Press . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Nunn Nathan ( 2009 ). “The Importance of History for Economic Development.” Annual Review of Economics , 1 , 65 – 92 . Google Scholar Crossref Search ADS WorldCat Nunn Nathan ( 2010 ). “Religious Conversion in Colonial Africa.” American Economic Review , 100 ( 2 ), 147 – 52 . Google Scholar Crossref Search ADS WorldCat Ochsner Christian , Roesel Felix ( 2019 ). “Populist Campaigning and Salient History: The Case of the Turkish Sieges of Vienna.” Working Paper , University of Zurich . Google Scholar Oster Emily ( 2019 ). “Unobservable Selection and Coefficient Stability: Theory and Evidence.” Journal of Business & Economic Statistics , 37 , 187 – 204 . Google Scholar Crossref Search ADS WorldCat Platteau Jean-Philippe ( 2000 ). Institutions, Social Norms, and Economic Development , Vol. 1 . Psychology Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Platteau Jean-Philippe ( 2017 ). Islam Instrumentalized . Cambridge University Press . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Qian Nancy , Xu Yiqing, Yao Yang et al. ( 2015 ). “Making Democracy Work: Culture, Social Capital and Elections in China.” Working Paper No. 21058 . OpenURL Placeholder Text WorldCat Richter Julius ( 1908 ). A History of Missions in India , Vol. 60 . Oliphant . Robinson James A. ( 2012 ). “ Elites and Institutional Persistence.” In The Role of Elites in Economic Development , edited by the late Amsden Alice H., DiCaprio Alisa, Robinson James A.. Oxford University Press , pp. 29 – 52 . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Roseberry J. Royal ( 1987 ). Imperial Rule in Punjab: The Conquest and Administration of Multan , 1818 – 1881 . Manohar Publishers . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Rubin Jared ( 2017 ). Rulers, Religion, and Riches: Why the West got Rich and the Middle East Did Not . Cambridge University Press . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Sokoloff Kenneth L. , Engerman Stanley L. ( 2000 ). “Institutions, Factor Endowments, and Paths of Development in the New World.” Journal of Economic Perspectives , 14 ( 3 ), 217 – 232 . Google Scholar Crossref Search ADS WorldCat Soofi Mushtaq ( 2014 ). “ Forgotten Glory of Punjab.” Working Paper, Dawn Newspaper . OpenURL Placeholder Text WorldCat Starrett Gregory ( 1998 ). Putting Islam to Work: Education, Politics, and Religious Transformation in Egypt , Vol. 25 . University of California Press . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Tabellini Guido ( 2010 ). “Culture and Institutions: Economic Development in the Regions of Europe.” Journal of the European Economic Association , 8 , 677 – 716 . Google Scholar Crossref Search ADS WorldCat Ticku Rohit , Shrivastava Anand, Iyer Sriya ( 2018 ). “ Economic Shocks and Temple Desecrations in Medieval India .” Cambridge Working Papers in Economics 1862, Faculty of Economics , University of Cambridge . Google Scholar UNESCO ( 2014 ). Education for All Global Monitoring Report 2013–14 . Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Valencia Caicedo Felipe ( 2019 ). “The Mission: Human Capital Transmission, Economic Persistence, and Culture in South America.” The Quarterly Journal of Economics , 134 , 507 – 556 . Google Scholar Crossref Search ADS WorldCat Wantchekon Leonard , and Natalija Novta Marko Klašnja ( 2014 ). “Education and Human Capital Externalities: Evidence from Colonial Benin.” The Quarterly Journal of Economics , 130 , 703 – 757 . Google Scholar Crossref Search ADS WorldCat Werenfels Isabelle ( 2014 ). “Beyond Authoritarian Upgrading: the Re-Emergence of Sufi Orders in Maghrebi Politics.” The Journal of North African Studies , 19 , 275 – 295 . Google Scholar Crossref Search ADS WorldCat Wilder Andrew ( 1999 ). The Pakistani Voter: Electoral Politics and Voting Behaviour in the Punjab . Oxford University Press . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Woodberry Robert Dudley ( 2004 ). The Shadow of Empire: Christian Missions, Colonial Policy, and Democracy in Postcolonial Societies . University of North Carolina at Chapel Hill . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC © The Author(s) 2021. Published by Oxford University Press on behalf of European Economic Association. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. © The Author(s) 2021. Published by Oxford University Press on behalf of European Economic Association.
TI - Pre-Colonial Religious Institutions and Development: Evidence through a Military Coup
JF - Journal of the European Economic Association
DO - 10.1093/jeea/jvab050
DA - 2022-04-18
UR - https://www.deepdyve.com/lp/oxford-university-press/pre-colonial-religious-institutions-and-development-evidence-through-a-yeq9vGRmC7
SP - 907
EP - 956
VL - 20
IS - 2
DP - DeepDyve
ER -