Orthodox Islamic institutions and individual income: evidence from Pakistan

The study investigates whether religiosity influences individual income in Pakistan, a predominantly Muslim, lower–middle-income country with substantial regional and ethnic diversity. Motivated by mixed evidence in prior research and the importance of religion in shaping social norms, networks, and economic behavior, the authors focus on Pakistan where 96% of the population is Muslim and 4% belong to non-Muslim minorities. The paper addresses rising income inequality in Pakistan and asks whether income differs by ethnic status and whether religiosity interacts with ethnic minority status to shape income. Unlike much of the literature centered on developed countries and conventional measures of religiosity, this study uses affiliation with deeni madrassas (orthodox Islamic institutions) as a proxy for religiosity, aiming to capture how membership in a strong religious ecosystem may affect individual economic outcomes.

The literature identifies both positive and negative channels linking religion and income. Theoretically, religion can affect time allocation and utility (Azzi & Ehrenberg, 1975), operate as a club good that provides social insurance and combats free-riding (Iannaccone, 1992), and enhance social capital and mobility via networks, education, health, and behavior (Chetty et al., 2014; Lim & Putnam, 2010; Hummer et al., 1999; Gruber & Hungerman, 2008). Conversely, higher opportunity costs may reduce religious participation as incomes rise (Lipford & Tollison, 2003; Hungerman, 2014), and cross-country studies often find religiosity inversely related to per capita GDP (Barro & McCleary, 2003). Empirical findings are mixed: positive effects of religiosity or affiliation on earnings in some contexts (Heath et al., 1995; Beck & Gundersen, 2016), null or negative relationships elsewhere (Bettendorf & Dijkgraaf, 2005; 2009; Iannaccone, 1998; Brown, 2000; Meredith, 2012). Religion also provides informal insurance mitigating shocks (Dehejia et al., 2007; Chen, 2010; Shaw et al., 2016). Prior work is limited by endogeneity concerns, focus on developed-country contexts, and measures emphasizing outcomes rather than institutional ecosystems. In Pakistan, orthodox Islamic institutions (madrassas) function as educational-social hubs with dense support networks (Blanchard, 2008; Rahman, 2009), which may shape incomes. The study posits two hypotheses: H1, active membership in orthodox Islamic institutions positively affects individual income; H2, the interaction of such membership with ethnic minority status further enhances income.

Data come from three waves of the Pakistan Social and Living Standards Measurement (PSLM) surveys (2010–2011, 2014–2015, 2019–2020) by the Pakistan Bureau of Statistics, with household samples of 76,546; 78,635; and 195,000, respectively. From the census populations in these surveys, individuals under 14 were excluded; observations with missing income (about 13.5%) were dropped; and 47,250 individuals (about 11% of the remaining population, via convenience quota sampling) were used for analysis. Variables include: LNINCO (log monthly income), LNAGE (log age), GENDER (1=female), REGION (1=rural), MINORITY (1=ethnic minority: Pashtun, Sindhi, Muhajir, Baluchi, Kashmiri, Saraiki; 0=Punjabi), EMPLOY (1=paid employment/business, 0=unpaid family worker), RELEDU (1=affiliated with orthodox Islamic institution/deeni madrassa as last-attended school), time dummies TDI (2019–20) and TDII (2014–15), and province dummies PDI (Khyber Pakhtunkhwa), PDII (Punjab), PDIII (Sindh), with Baluchistan as reference. Religiosity is proxied by last attendance at a deeni madrassa, drawn from PSLM question Sec C Q7 on last institution type. Baseline model: LNINCO = α0 + α1 LNAGE + α2 GENDER + α3 REGION + α4 MINORITY + α5 EMPLOY + α6 RELEDU + α7 TDI + α8 TDII + α9 PDI + α10 PDII + α11 PDIII + ε. Extended model adds interaction RELEDU×MINORITY: LNINCO = α0 + … + α6 RELEDU + α7 (RELEDU×MINORITY) + … + ε. Estimation uses pooled OLS with Huber-White robust standard errors; individual fixed effects are not feasible due to lack of individual time trajectories. Cohort, regional, provincial, and ethnic controls are included to mitigate unobserved heterogeneity. Structural Equation Modeling (SEM) is also employed to address measurement error in the latent construct of religiosity and to assess direct/indirect effects, allowing correlations among variables. Multicollinearity was assessed and not found problematic.

• Affiliation with orthodox Islamic institutions (RELEDU) positively associates with income. In the baseline pooled OLS, RELEDU coefficient = 0.094 (p<0.001), implying approximately 9.4% higher income; in SEM, 0.018 (p<0.001). In the extended OLS, RELEDU = 0.066 (p=0.003). - The interaction RELEDU × MINORITY is positive and significant: 0.069 (p=0.048) in OLS (0.070, p=0.043 in SEM), indicating an additional income premium for ethnic minority affiliates. - MINORITY has a positive association with income: baseline 0.189 (p<0.001); extended 0.177 (p<0.001), suggesting minorities earn more than the Punjabi majority in the sample. - REGION (rural=1) is negatively associated with income: about −0.117 to −0.119 (p<0.001), indicating urban residents earn more. - EMPLOY (paid employment/business=1) shows a negative coefficient: −0.040 to −0.041 (p≈0.006–0.008), reflecting higher reported income among unpaid family workers, potentially due to unmonetized in-kind benefits affecting comparisons. - Time dummies are positive and significant versus 2010–11: TD1 (2019–20) ≈ 0.760; TDII (2014–15) ≈ 0.361 (both p<0.001), consistent with rising incomes over time. - Province effects (vs. Baluchistan reference) are negative and significant: PDI (KPK) ≈ −0.193 to −0.198; PDII (Punjab) ≈ −0.147 to −0.150; PDIII (Sindh) ≈ −0.191 to −0.197 (all p<0.001), implying relatively higher incomes in Baluchistan during the period. - LNAGE and GENDER are statistically insignificant. - Model fit: R^2 ≈ 0.0653 (baseline) and 0.0654 (extended); N=47,250.

Findings support H1 and H2. Affiliates of deeni madrassas earn significantly more, and this premium is amplified among ethnic minorities. The authors interpret the positive effect as reflecting madrassas’ role beyond education: they function as networked social institutions that provide free education, boarding, and robust placement into employment aligned with religious training, supported by stable funding streams (zakat, waqf, sadaqat). The institutional evolution and enlarged financial base—including engagement with merchants and business groups—may facilitate job opportunities and income stability for affiliates. The stronger positive effect among ethnic minorities is attributed to the concentration of organized madrassa networks in economically dynamic minority-majority regions (e.g., Karachi, parts of KPK) where business ecosystems and religious institutions intertwine, enhancing access to opportunities and mitigating social-economic marginalization. These results address the research question by showing that religiosity, operationalized as affiliation with orthodox Islamic institutions, is positively related to individual income in Pakistan and interacts beneficially with minority status. The broader relevance lies in recognizing the socio-institutional mechanisms through which religious ecosystems can influence economic outcomes in developing-country settings.

The study provides evidence from Pakistan that affiliation with orthodox Islamic institutions (deeni madrassas) is positively associated with individual income, with an additional premium for ethnic minority affiliates. Using large-scale PSLM microdata and both pooled OLS with robust errors and SEM, the paper highlights the socio-economic role of madrassas as embedded networks offering support, education, and employment pathways that sustain incomes through shocks. Policy implications include recognizing the economic value of religious institutional ecosystems in workforce development, considering culturally sensitive integration of religious education within broader human capital strategies, and supporting inclusive access that benefits minority groups. Future research could examine heterogeneous effects across Muslim countries with varying institutional environments, delve deeper into mechanisms (networking, placement, credit access), and employ designs that further address causality (e.g., quasi-experimental variation, panel data) and measurement of religiosity beyond educational affiliation.

• Data are pooled cross-sections without individual time trajectories, precluding individual fixed effects and limiting causal inference. - Religiosity is proxied by last attendance at a deeni madrassa; while motivated and addressed via SEM for measurement error, this surrogate may not capture all aspects of religiosity. - The final analytic sample is a convenience quota subsample (47,250) after exclusions, which may affect generalizability. - The models explain a modest share of income variation (R^2 ≈ 0.065), suggesting many unobserved determinants remain. - Despite controls (cohort, regional, provincial, ethnic), residual endogeneity and omitted-variable bias cannot be fully ruled out.