«SUMMARY | APRIL 2 WORKING PAPER No. 030 | May 2015 Religion, Land and Politics: Shrines and Literacy in Punjab, Pakistan Adeel Malik and Rinchan Ali ...»
64 See Appendix 2 for a complete list of tehsils.
65 When separately included, in the absence of the interaction term, the coefficient on distance from river is statistically insignificant.
66 The precise size of the impact can be calculated as:
-2.697 + 0.114 x (21.155).
Robust standard errors in parentheses: *** p0.01, ** p0.05, * p0.1 Latitude, longitude and elevation are expressed in natural logs.
The following variables are expressed in natural log: latitude, longitude, elevation, rainfall and distance from GT Road.
Overall, the results strongly support the suggestion that the impact of riverine shrines on literacy, captured by shrine pc, is significantly different from that of shrines situated in tehsils relatively distant from river. It is worth noting that the coefficient on elevation is positive and statistically significant at the 1% level. Apart from capturing specific geographic features, the elevation measure is likely to pick up the impact of other development dimensions, such as rainfall patterns, soil quality and land inequality, with which it is highly correlated. 67 Extended Specifications We next explore extended specifications to ascertain the strength of these empirical patterns. A key challenge in reduced form specifications is to ensure that the results are not contaminated by omitted variables. One crucial dimension in this regard is land inequality. To what extent, we might ask, is the effect of riverine shrines simply acting as a proxy for inequitable land distribution? The fact that most pīrs in riverine regions are also landowners 67 Geographically the average altitude falls as we move from North to South Punjab. Both rainfall and land inequality follow a similar pattern.
We do this in column (2) by adding a proxy for land inequality, the proportion of Muzaara (landless tenants). As expected, the results indicate a very strong negative association between land inequality and literacy.
This is consistent with the argument that landlessness combined with productive agriculture is associated with illiteracy. Measured at the end of the sample period, the negative effect of Muzaara suggests that land inequality and literacy are jointly determined. Importantly, however, the impact of riverine shrines remains robust to the inclusion of this powerful correlate of literacy.
It is important to ensure that the shrines effect is not simply driven by systematic differences in regional development across Punjab. As suggested by Cheema et al. (2008), the south and western parts of Punjab are considerably more underdeveloped than the rest of Punjab. Coincidentally, these regions are also home to some of the politically most influential shrines. To ensure that the shrines effect is not conflated by these regional differences, we include in column 3 three dummy variables for north, central and south Punjab (with west Punjab as the base category). Coefficients on none of these regional dummy variables are statistically significant, whilst the impact of riverine shrines and land inequality remain negative and significant.
A related concern arises if the settlement patterns of sufi saints were determined by ecological endowments and these, in turn, shaped the long-term conditions for literacy. If saints were more likely to settle down in riverine regions and the riverine tracts were more backward, the shrine-river interaction would just pick up these generalized development effects independent of the influence of shrines. To control for this possibility we include a commonly used proxy for income, mean rainfall over the period, 1960-2008. The result is presented in column
4. As expected, rainfall is a strong predictor of literacy: its coefficient is positive and significant at 1% level. Apart from being associated with land inequality68, rainfall captures many unobserved dimensions of income and development. It is reassuring, however, that the negative impact of riverine shrines, captured by the coefficient on shrines pc, continues to be robust.
Some previous work on Punjab has also controlled for distance from Lahore to reflect core-periphery dynamics in development. Other studies have sometimes controlled for distance from the GT (Grand Trunk) road.
69 It is possible that tehsils with characteristics least favourable to literacy were also remote from the historical centre of power and, consequently, suffered from a substantial power vacuum that was filled by shrines. Shrines 68There is a strong negative correlation between rainfall and land inequality. Land is more unequally distributed in low-rainfall regions.
69 GT Road stands for Grand Trunk Road, a major road artery connecting Central Asia with eastern and western parts of the Indian subcontinent. We use the natural log of the variable.
The results are instructive. Inclusion of the politics interaction renders the coefficient on our main interactive term between shrines pc and the River dummy statistically insignificant. Instead, the Politics interaction now has a negative and statistically significant coefficient, supporting our hypothesis that the shrine effect is principally mediated through politics.
The model in column 3 imposes a more stringent test by including all three dimensions together (River, Muzaara and Politics). This allows us to estimate the differential impact of shrines per capita across the three categories defined above. We introduce a dummy variable, Muzaara, that takes a value of one in tehsils where the proportion of Muzaara is greater than or equal to the 75th percentile of the variable’s distribution; zero otherwise. 71 The interaction of the Muzaara dummy with shrines indicator effectively allows the coefficient on shrines pc to be different for tehsils with high and low levels of land inequality. The shrines-politics interaction still has a robust SUMMARY | APR negative and statistically significant impact on literacy. There is no additional impact of Muzaara or its interaction with shrines. Adding average rainfall in column 4 slightly weakens the impact of politics, but it still remains negative and statistically significant at the 10 percent level. The coefficient on the Muzaara dummy itself is now negative and weakly significant. The inclusion of regional dummy variables in column 5 leaves the results unchanged. In column 6 we drop the river and Muzaara interactions and explore a specification with Proportion Muzaara. Despite the inclusion of two powerful correlates of literacy, rainfall and Muzaara, the politics interaction continues to exert a negative and significant impact.
Together, these results are consistent with the suggestion that the impact of riverine shrines is primarily mediated through politics. As the earlier narrative argues, prominent shrines dotted along rivers of Punjab have traditionally acted as important nodes of power, where structures of religious and political authority have been historically co-determined. Through their control of the three critical resources—religion, land and politics—they are a key constitutive element of the local power structure. This is confirmed by eyeballing the data on shrinerelated political families and matching it with selective interviews and qualitative data. Invariably, politically linked shrines have either direct or indirect control over vast tracts of agricultural land. Caretakers of these shrines are therefore not simply pīrs, but pīr-zamindars. Today, nearly all leading pīr families own substantial tracts of land.72 Their landed power is largely hereditary. For example, even in the 1890s the Chishtis of Pakpattan Sharif owned 9 percent of all land in the Pakpattan tehsil.73 Thus, politically influential shrines encapsulate the impact of both landed and political power. However, the effect of political shrines is over and above any direct role of land inequality measured through the Muzaara variable.
Overall, these results provide first-cut evidence on the empirical relationship between shrines, politics and development. Since our results are based only on direct evidence of a shrine family’s entry into politics, they understate the true relationship between shrines and politics. Even if a shrine lacks direct electoral representation, it can play a crucial indirect role by garnering support for election candidates. A relevant example is the shrine of Shaikh Fazil in Burewala that is better known for its indirect political brokerage than direct electoral contest, and has an impressive political footprint on multiple constituencies of Punjab.74 Support of the shrine of Golra Sharif is similarly deemed critical for candidates from many neighbouring constituencies in Islamabad and Rawalpindi.
SHRINES AND POLITICSIn this section, we probe the political dimension in greater detail by considering rival historical explanations and investigating the determinants of political selection. A key challenge is that initial tehsil characteristics might determine the extent to which shrines had an influence over literacy, and that these characteristics may either persist affecting literacy today, or that they might have influenced development outcomes in the past through channels other than the influence of shrines. One crucial dimension that could conflate the impact of political shrines is the historical literacy rate. It is a well-established fact that literacy rates display considerable persistence over time. This has important implications for our results. If regions with political shrines had historically lower literacy rates to begin with then their negative impact on literacy could simply capture these adverse initial conditions. The exclusion of other historically determined tehsil characteristics from our model poses a similar challenge. We try to address these concerns by re-estimating our model on the sample of colonial tehsils and including historically pre-determined characteristics in the model. This requires that we fix tehsil boundaries at 1931, the census year for which historical data is readily available. Data on 115 contemporary tehsils were therefore collapsed into data for 61 colonial tehsils. The underlying mapping strategy is discussed in detail in Appendix 3.
Results for this exercise are presented in Table 5. Our historical sample, consisting of 61 colonial tehsils, is dictated by data availability and is listed in Appendix 4.75 Several historical variables were compiled from district gazetteers and Census Reports. We first assess the robustness of our final explanation, the political influence of shrines, on the historical sample.
71 The 75th percentile of the Proportion Muzaara is equivalent to.0636.
72 The pīrs also consolidate their power through strategic marital alliances with large landowners. Urban shrines have also continuously added to their existing land holdings. For example, the Pīr of Golra Sharif has sizeable landholdings in and around Islamabad. Even shrines with more limited economic fortunes tend to function with the patronage and support of local landlords.
73 Final Report of the Revision of Settlement of the Montgomery District, 1892-99.
74 As a regular election ritual, candidates from neighbouring constituencies queue up in Shaikh Fazil to seek spiritual and political support.
75 The sample size reduced from 69 to 61 tehsils due to data unavailability.
Column 1 includes the Politics dummy variable and its interaction with our main variable of interest, shrines pc. The usual set of geographic controls, including average rainfall measured for 1923, is also included.
Consistent with our argument, greater concentration of shrines in tehsils where shrine families have a direct political presence have systematically lower literacy rates. Results on the historical sample confirm that the presence of shrines is harmful for literacy only in tehsils where shrines have a manifestly political role. As expected, rainfall is a strong and positive predictor of literacy. The statistically significant coefficient on the regional dummy for north Punjab suggests an unexplained literacy differential between north and west Punjab.
We then control for historical land inequality and literacy rate in column 2. To measure land inequality we use detailed information on land tenure contracts from various editions of district gazetteers compiled by the
British colonial administration. Typically, district gazetteers divided land tenure contracts into four categories: