Occupations, Diversity and Public Goods: Evidence from Pakistan through Partition. Rinchan Ali Mirza Candidate for the Dphil in History (HSM & ESH)

Size: px

Start display at page:

Download "Occupations, Diversity and Public Goods: Evidence from Pakistan through Partition. Rinchan Ali Mirza Candidate for the Dphil in History (HSM & ESH)"

Kristopher Houston
5 years ago
Views:

1 Occupations, Diversity and Public Goods: Evidence from Pakistan through Partition Rinchan Ali Mirza Candidate for the Dphil in History (HSM & ESH)

2 MOTIVATION AND CONTRIBUTION

3 Why study the impact of religious diversity? Empirical studies have devoted ample attention to studying the relationship between ethnic diversity and public goods provision. Alesina, Baqir and Easterly (1999); Easterly and Levine (1997); Miguel and Gugerty (25); Alesina and LaFerrara (25); Posner (25) and Jackson (27). This has meant that another important dimension of diversity (i.e. religious) has received relatively scant attention in the literature Mixed success in studies in dealing with endogeneity bias when investigating the relationship between diversity and public goods provision. Need for an effective identification strategy that reduces endogeneity bias Example: Diversity could lower school provision by creating collective action problems. However it could also be the case that poorer minorities face discrimination, and are only able to attend poorly funded schools. The poorly funded schools would then appear more diverse

4 Contribution Project utilizes the almost complete eradication of religious diversity across all districts of West Punjab, Sind and NWFP due to the partition as an identification strategy It constructs a historical panel dataset that studies how school provision has evolved over a period stretching back to the pre-partition era Its empirical analysis shows that those districts that had a higher proportion of minorities pre-partition experienced slower growth in the number of schools per thousand persons post-partition relative to those district that a lower proportion of minorities Mechanisms: The effect of minorities is in part explained by the pre-partition occupational structure of Hindus and Sikhs in Pakistan This does not, however, explain the whole of the influence of these groups. Other mechanisms, such as the social capital sustained by shared religious festivals, also mattered.

5 Map of India before Partition

6 Map of Punjab at Partition

7 IDENTIFICATION STRATEGY

8 Empirical Equation The central variable of interest for the study is: (Pre-partition minorities proportion) (Indicator for Post-partition) More simply, the above variableisdenoted where ( j) is the district by ( H Pr j e Post), Similar to the variable constructed by Hoyt Bleakley in his study on the effects of Hookworm Eradication on years of schooling in Southern US Based on the central variable the following regression is estimated: Y jt 1 ( H Pre j Post) X Where Y jt is the outcome of interest (schools per thousand persons), the t are year dummies, the j are district fixed effects, and X is a vector of time -varying controls jt t j jt jt

9 Identification Strategy Three main components of the identification strategy: First: Prevalence of differences in minorities proportions across districts in the pre-partition period. Such heterogeneity allows for a treatment-control strategy Second: Partition of India was the result of a political struggle that eliminated diversity in districts, irrespective of their economic characteristics or later potential for school expansion Third: Religious diversity in the districts that constituted West Punjab, Sind and NWFP was eradicated within a span of a few years after partition. This is a sudden and sharp drop in diversity on historical time scales

10 Eradication of religious diversity West Punjab Attock Bahawalpur State Dera Ghazi Khan Gujranwala Gujrat Jhang Jhelum Lahore Lyallpur Mianwali Montgomery Multan Muzaffargarh Rawalpindi Shahpur Sialkot

11 Eradication of religious diversity NWFP Bannu Dera Ismail Khan Hazara Kohat Peshawar Year

12 Eradication of religious diversity Sind Hyderabad Karachi Khairpur State Larkana Sukkur Thar Parkar Upper Sind Frontier States

13 DATA AND MAIN RESULTS

14 Data Sources Data on schools, population, the proportion of minorities, and other controls are taken principally from the District census reports for Pakistan, Census reports for British India, and District Gazetteers and District Development Statistics The base year used to construct the Panel Dataset is The number of districts in the base year is 27. The years covered are 1891, 191, 1911, 1931, 1961, 1972, 1981, 1998 and 28. The total number of observations are: 27 x 9 = 243 Used a mapping procedure to ensure that the number of districts are kept fixed throughout the panel period. Starting from the base year as the number of districts increased they were collapsed to their previous census round Dependent variable: total number of secular schools per thousand persons

15 Change in school provision and diversity HAZ BWS USF BNU KHA HYD PHW DIK LAR KOH JHE GUJ MIA JHA MUL ATK DGK MUZ SHP RWP SUKSKT GURMGR KHI LYP LHR.4 Pre-partition proportion of minorities (1931)

16 Main Results Table 1: Proportion of Minorities and Public School Provision: Results (1) (2) (3) (4) Dependent variable: Schools per thousand persons Schools per thousand persons Schools per thousand persons Schools per thousand persons Central Variables Post-partition Dummy X Pre-partition minorities proportion (1931) Post-partition Dummy X Pre-partition minorities proportion (1911) Post-partition Dummy X Pre-partition minorities proportion (191) Post-partition Dummy X Pre-partition minorities proportion (1891) Controls Post-partition Dummy X Literacy of In- Migrants at Partition Post-partition Dummy X Average Rainfall during Pre-partition *** (-3.92) (-1.11).17 (.38) Log (District Population) * (-2.27) *** (-3.32) (-1.16).25 (.54) * (-2.16) ** (-2.92) (-1.18).25 (.51) -.23 * (-2.16) *** (-3.1) (-1.2).26 (.53) * (-2.45) Sample Size Note: t statistics in parentheses. In addition to the controls included in the table above district and year level fixed effects were also controlled for in the regressions. Moreover the results are robust to clustering at the district and year levels. Another control which is not included in the table above was a dummy for a big city that captured the urban fixed effect. * p <.5, ** p <.1, *** p <.1

17 Overview The results show a negative impact of religious diversity eradication at the time of partition on school provision in the post-partition period 1931 Specifically, the coefficient on ( H j Post) implies that a district that had 5% of its population composed of religious minorities in 1931 would experience a reduction of.81 schools per thousand persons, relative to a district that had no religious minorities in 1931 The regressions are run separately using interactions between post-partition dummy and each of the four pre-partition years (i.e. 1891, 191, 1911, 1931) This is in comparison to: a standard deviation of.5 over the whole panel of 243 observations; a standard deviation of.42 over the sample for year 28 and; a standard deviation of.13 over the sample for the year 1931

18 ROBUSTNESS

19 Robustness Several controls included: (literacy of in-migrants at partition X Post), (average rainfall during pre-partition X Post), Log(population of district), dummy for city, district fixed effects and year fixed effects The empirical equation was re-run separately for the central variables: (i) prepartition proportion of ethnic minorities x Post and; (ii) pre-partition linguistic minorities x Post The coefficients of neither of the above central variables were significant This further suggests that it was pre-partition religious diversity that mattered and not ethnic and/or linguistic diversity

20 Which type of school matters? The empirical equation was re-run separately for subgroups: primary, middle and high schools Results show a significantly negative impact of religious diversity eradication on the number of primary schools per thousand persons. However, the results for middle and high schools are not significant The empirical equation was re-run separately for subgroups: male and female schools Results show a significantly negative impact of religious diversity eradication on the number of male schools per thousand persons. However, the results for females is not significant

21 MECHANISM

22 Mechanisms (I) The minorities by virtue of their higher literacy formed a constituency that was crucial to school provision Hindu/Sikh literacy rates were consistently higher than Muslim literacy for the period 191 to 1931 for all the districts in West Punjab and Sind. Chaudhry and Rubin (29) explain literacy gap in terms of the pervasiveness and strength of Muslim religious schools in Muslim majority areas that restrained the progress of literacy However in regions considered in this study it was the differences in the educational wants of Hindus/Sikhs and Muslims. The Muslims were largely agriculturists for whom education was not necessary. On the other hand a majority of Hindus were in mercantile professions that required the ability to read and write

23 Pre-partition differences in literacy (Western Punjab) Attock Bahawalpur State Dera Ghazi Khan Gujranwala Gujrat Jhang Jhelum Lahore Lyallpur Mianwali Montgomery Multan Muzaffargarh Rawalpindi Shahpur Sheikhupura Sialkot Hindu/Sikh Literacy Rate Muslim Literacy Rate

24 Muzaffargarh Lyallpur Bahawalpur Shahpur Rawalpindi Mianwali Jhelum Gujrat Attock Multan Montgomery Jhang D. G. Khan Sialkot Lahore Gujranwala Muzaffargarh Lyallpur Bahawalpur Shahpur Rawalpindi Mianwali Jhelum Gujrat Attock Multan Montgomery Jhang D. G. Khan Sialkot Lahore Gujranwala Minorities and Occupational Selection Occupation Agricultural Religious Mercantile Artisan Menial & Outcast Other Total Muslim Source: Census of Punjab (1931) Occupation Agricultural Religious Mercantile Artisan Menial & Outcast Other Total Hindu Source: Census of Punjab (1931)

25 Mechanisms (II) Table 2: Proportion of Minorities, proportion of Mercantile Castes and Public School Provision: Results Schools per thousand persons Schools per thousand persons Dependent Variable Without With Dependent Variable Without With Post-partition Dummy X Pre-partition minorities proportion (1931) Post-partition Dummy X Pre-partition mercantile caste proportion (1931) *** (-3.92) * (-2.54) * (-2.6) Post-partition Dummy X Pre-partition minorities proportion (1911) Post-partition Dummy X Pre-partition mercantile caste proportion (1931) ** (-3.32) * (-2.21) ** (.9) Sample Size Sample Size Note: t statistics in parentheses. All the controls used in regression Table 1 have also been included in the above regressions. However, one further control was added in the above regressions. It was the interaction between pre-partition proportion of depressed castes and post-partition dummy. It is only after using such a control that the coefficient on the Post X Pre-partition mercantile proportion gains significance and the coefficient on Post X Pre-partition minorities proportion loses significance. * p <.5, ** p <.1, *** p <.1

26 Mechanisms (III) Influence of minorities in terms of school provision is not wholly explained away by their selection into mercantile castes. There are other factors that contributed to importance of minorities The social capital created and sustained by minorities through mixed religious gatherings in the population also mattered for schools Successive census reports show that large numbers of Hindus regularly undertook pilgrimages to Muslim shrines and many Muslims conducted part of their life-cycle rituals as if they were Hindus Charles Ibblestone, the author of the Punjab Census Reports of 1881, mentions the mixed religious gatherings that took place at Muslim shrines

27 Mechanisms (III) For instance Sakhi Sarwar s shrine in the Dera Ghazi Khan district was held up by Ibblestone as exemplifying the extraordinary manner in which religions are intermingled in the Punjab He further described the numerous saints of the Punjab as generally Mahomedan, but adds that they were worshipped by Hindus and Muslims alike with the most absolute impartiality Eradication of minorities through partition meant the destruction of such social capital which had the potential to facilitate better collective action

28 CONCLUSION

29 Conclusion This paper utilizes the eradication of minorities at partition as an identification strategy to determine the impact of religious diversity on school provision Its results show that those areas that had a higher proportion of minorities before partition experienced a slower growth in the number of schools per thousand persons after partition This was in part because religious minorities were more literate by virtue of their occupational selection and thus better at demanding public goods such as schools It was also due to the destruction of the social capital that minorities had created and sustained which facilitated collective action required for school provision The results are not sensitive to the inclusion of several controls such as average rainfall, population, literacy of in-migrants and urban, district and year fixed effects The results are being driven by male schools and primary schools as they are the most abundant

Statistical Analysis of Endorsement Experiments: Measuring Support for Militant Groups in Pakistan

Statistical Analysis of Endorsement Experiments: Measuring Support for Militant Groups in Pakistan Kosuke Imai Department of Politics Princeton University Joint work with Will Bullock and Jacob Shapiro