Religious Education in the Age of Mass Schooling (Tables)

Tables Table 1: Exposure to Islamic Education Source Exposure Definition Cohort IFLS, 1993–2014 at given level all in school (1) (2) Susenas, 2012–18 at final level all in school (3) (4) Admin., 2019 enrolled in school (5) Education Level All Primary Junior Secondary Senior Secondary 20% 25% 7% 10% 21% N=64,141 N=10,573 N=5,240,958 N=1,652,990 N=59,387,784 11% 16% 4% 6% 13% N=55,912 N=10,572 N=3,187,724 N=1,263,12 N=29,309,849 23% 28% 12% 14% 23% N=32,221 N=4,282 N=1,394,572 N=629,061 N=13,708,973 20% 24% 6% 7% 11% N=21,522 N=2,587 N=1,476,917 N=389,880 N=12,412,256 Notes: This table summarizes Islamic education rates across multiple levels of schooling using three different sources. The ‘All’ row includes madrasa enrollment as well as (where possible) pesantren enrollment which cannot be assigned to specific grade levels. Hence Islamic education includes only madrasa in the Primary, Junior Secondary and Senior Secondary rows. The sample sizes reflect the total number of observations over which the percent exposed to Islamic education is computed. Columns 1 and 2 used the Indonesian Family Life Survey (IFLS) longitudinal records from 1993, 1997, 2000, 2007 and 2014. This data is representative of 83% of the Indonesian population and does not cover many districts. This survey records the complete educational history of respondents. Column 1 reports the exposure across all individuals spanning the five survey rounds. Column 2 restricts to the 2014 round and looks only at currently enrolled students. The ‘All’ row includes any pesantren enrollment. Columns 3 and 4 use the nationally-representative annual Susenas data from 2012–2018, which covers all districts and which we deploy in our main empirical analysis. Unlike the IFLS, this data only captures the type of the final year of schooling completed by respondents and only allows respondents to indicate madrasa but not pesantren. Column 3 reports the exposure across all individuals spanning the six Susenas rounds. The Primary, Junior Secondary, and Senior rows are restricted to individuals that completed exactly 6, 9, and 12 years of education, respectively. Column 4 restricts to individuals currently enrolled in school in each round of the survey. These estimates are computed using the sampling weights to obtain national representativeness. Column 5 uses administrative data for the 2019 school year from the Ministry of Education (MEC) and Ministry of Religion (MORA). The former records madrasa attendance while the latter records pesantren attendance. The ‘All’ row includes pesantren enrollment. Table 2: INPRES Exposure and Islamic Schooling (2) (1) (3) (4) (5) (6) (a) Highest Education Level: [. . . ] Islamic Elementary Junior Secondary Senior Secondary INPRES × young Observations Dependent Variable Mean R2 -0.0013** (0.0005) -0.0006 (0.0008) 0.0020*** (0.0005) 0.0047*** (0.0011) 0.0011*** (0.0003) 0.0033*** (0.0007) 839,026 0.014 0.027 3,938,728 0.010 0.019 839,026 0.011 0.011 3,938,728 0.016 0.023 839,026 0.008 0.007 3,938,728 0.012 0.014 (b) Highest Education Level is Islamic, (b) Conditional on Completing [. . . ] 6 Years 9 Years 12 Years INPRES × young Observations Dependent Variable Mean R2 Number of Districts Cohorts aged 2-6 vs. 12-17 in 1974 ———”——– ≤6 vs. ≥12 in 1974 -0.0021*** (0.0006) 0.0011 (0.0013) 0.0053*** (0.0018) 0.0097*** (0.0021) 0.0005 (0.0011) 0.0057*** (0.0015) 457,020 0.025 0.044 2,918,805 0.030 0.049 121,758 0.073 0.076 1,313,827 0.103 0.081 169,914 0.038 0.036 1,349,798 0.056 0.049 275 X 275 275 X 275 275 X 275 X X X Notes: This table reports estimates of equation (1) based on annual Susenas data from 2012 to 2018. INPRES refers to SD INPRES schools constructed from 1973-78 per 1,000 children in 1971. The dependent variables include an indicator equal to one if the individual’s final year of schooling was completed in an Islamic elementary (columns 1–2), junior secondary (columns 3–4), and senior secondary (columns 5–6). Panel (a) includes all individuals regardless of their years of schooling. Panel (b) includes only individuals with the given years of schooling corresponding to each level such that columns 1–2 look at Islamic elementary completion among individuals with 6 years of schooling, columns 3–4 look at Islamic junior secondary completion for those with 9 years, and columns 5–6 look at Islamic senior secondary completion for those with 12 years. All specifications include survey year dummies, district of birth dummies and year of birth dummies interacted with the 1971 children population, the 1971 enrollment rate, and exposure to the water and sanitation program in the district of birth. In odd-numbered columns, the sample is composed of all individuals aged 2–6 (young) or 12–17 in 1974. In even-numbered columns, the sample is composed of all individuals aged less than 6 (young) or more than 12 in 1974. * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by district of birth. Table 3: Effects of INPRES Exposure on Quantity and Type of Schooling Years of Schooling (1) (2) Highest Level Islamic (3) (4) Islamic | Highest Level (5) (6) (a) Reduced Form INPRES × young 0.1392*** (0.0267) 0.2824*** (0.0479) 0.0017** (0.0007) 0.0070*** (0.0020) 0.0005 (0.0007) 0.0054*** (0.0016) (b) Two Stage Least Squares Years of Schooling [weak-instrument-robust p-value] Observations Number of Districts Cohorts aged 2-6 vs. 12-17 in 1974 ———”——– ≤6 vs. ≥12 in 1974 Dependent Variable Mean R2 (panel a) First-stage F-statistic (panel b) Underidentification Test, p-value 839,019 275 X 7.456 0.163 3,938,710 275 X 7.664 0.376 0.0120** (0.0056) [0.022] 0.0247*** (0.0064) [0.000] 0.0065 (0.0102) [0.489] 0.0480*** (0.0177) [0.001] 839,019 275 X 3,938,710 275 717,583 275 X 3,185,314 275 0.031 0.030 27.11 0.000 X 0.037 0.040 34.70 0.000 0.036 0.034 7.45 0.011 X 0.046 0.044 15.70 0.001 Notes: This table reports (in panel a) estimates of equation (1) based on annual Susenas data from 2012 to 2018. INPRES refers to SD INPRES schools constructed from 1973-78 per 1,000 children in 1971. The dependent variables include total years of schooling (columns 1–2), an indicator equal to one if the individual’s final year of education is in an Islamic school (columns 3–4), conditional on the given years of schooling completed (columns 5–6). The dependent variable in columns 3–4 aggregates over all levels in panel (a) of Table 2, and columns 5–6 aggregate over panel (b). The specification in panel (a) is otherwise identical to that in Table 2; see the notes therein for details. Panel (b) reports two-stage least squares (2SLS) estimates of years of schooling on the likelihood to complete an Islamic education overall (columns 3–4) or conditional on the highest level attained (columns 5–6). We instrument for years of schooling using the IN P RESj × youngij interaction in equation (1). The endogenous regressor (years of schooling) is identical to the outcome used in panel (a), columns 1–2. We report the Kleibergen-Paap first-stage F statistic computed using standard errors clustered at the district of birth level. The null of the underidentification test is that the equation is underidentified. The weak-instrument robust p-value is based on the Anderson-Rubin test. The 2SLS specification otherwise includes the same set of controls and fixed effects as those included in equation (1). * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by district of birth. Table 4: INPRES Intensity and Entry of Islamic Schools New schools per 1,000 children Elementary (1) New Madrasa Junior Sec. Senior Sec. (2) (3) Diniyah (4) New Pesantren All (5) (a) Baseline INPRES × post-1972 Observations Number of Districts Avg. new establishments/year R2 0.0043*** (0.0009) 0.0042*** (0.0010) 0.0023*** (0.0006) 0.0095*** (0.0028) 0.0029*** (0.0007) 27,500 275 0.008 0.168 27,500 275 0.006 0.192 27,500 275 0.003 0.233 27,500 275 0.019 0.238 27,500 275 0.005 0.224 (b) Heterogeneity by Islamic Assets (waqf ) and Muslim Share INPRES × post-1972 INPRES × post-1972 × waqf, 1972 INPRES × post-1972 × Muslim share, 1972 Observations Number of Districts Avg. new establishments/year R2 -0.0008 (0.0010) 0.0209** (0.0096) 0.0050*** (0.0017) -0.0022*** (0.0007) 0.0221*** (0.0084) 0.0064*** (0.0017) -0.0009** (0.0004) 0.0108** (0.0051) 0.0032*** (0.0010) -0.0073** (0.0030) -0.0133 (0.0397) 0.0235*** (0.0066) -0.0009 (0.0009) 0.0061 (0.0073) 0.0046** (0.0018) 27,300 273 0.008 0.187 27,300 273 0.006 0.221 27,300 273 0.003 0.266 27,300 273 0.019 0.270 27,300 273 0.005 0.248 Notes: This table reports estimates of equation (2). This specification is estimated on a panel at the district-year level spanning 1920–2019. INPRES refers to SD INPRES schools constructed from 1973-78 per 1,000 children in 1971. The dependent variables are measured as new schools of a given type created per district per year and per 1,000 children in 1971. Madrasa diniyah and pesantren are Islamic afternoon schools and Islamic boarding schools, respectively. Panel (a) is the baseline while panel (b) allows the effects of INPRES to vary with the size of Islamic endowments (waqf ) at the district-level in 1972. The measure captures total waqf land (in square kilometers) held by mosques. All specifications include district fixed effects and year fixed effects interacted with the 1971 children population, the 1971 enrollment rate, and exposure to the water and sanitation program. We lose two districts in panel (b) on account of missing data on waqf endowments. * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by district. Table 5: Islam–State Competition at the Local Level Islamic School Entry Elem.=1 Elem.=0 Elem.=1 Jun. Sec.=0 Jun. Sec.=1 Jun. Sec.=1 (1) (2) (3) (a) Entry 1973–1983 SD INPRES built in village, 1973–78 0.002 (0.002) 0.005*** (0.001) 0.0001 (0.0005) SD INPRES saturation at subdistrict level -0.012 (0.017) 0.009*** (0.002) 0.002 (0.002) 0.049 75,090 0.009 0.004 Number of Villages Share of Villages with Given Profile (b) Entry 1984–1998 SD INPRES built in village, 1973–78 0.006*** (0.002) 0.015*** (0.002) 0.004*** (0.001) SD INPRES saturation at subdistrict level 0.029*** (0.007) 0.022*** (0.007) 0.011*** (0.003) 0.035 75,090 0.039 0.009 Number of Villages Share of Villages with Given Profile Notes: This table presents average marginal effects from a village-level multinomial logit regression relating SD INPRES entry to Islamic school entry at the village level with four categorical outcomes: no madrasa entry (the base, reference), elementary madrasa but not junior secondary madrasa, junior secondary madrasa but not elementary madrasa, and entry of both elementary madrasa and junior secondary madrasa. The reported regressors are an indicator for whether the given village had any SD INPRES schools constructed from 1973 to 1978, and the share of all villages in the subdistrict (a proxy for the local education market) with any SD INPRES construction from 1973 to 1978. The latter excludes the own village from the subdistrict share calculation. Panel (a) reports estimates from a regression with the dependent variable capturing entry from 1973–1983, and panel (b) looks at the remaining years of the Suharto regime from 1984–1998. We end in 1983 as this was the culmination of the five-year planning horizon under the initial SD INPRES plan. The marginal effects are with reference to villages with no madrasa entry over the given time horizon. Each regression also includes controls for the usual district-level controls (1971 children population, the 1971 enrollment rate, and exposure to the water and sanitation program). * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by district. Table 6: Curriculum Differentiation in Islamic Schools All Levels (1) Primary (2) Jun. Sec. (3) Sen. Sec. (4) (a) Islamic Subject Share INPRES × post-1972 Dependent Variable Mean 0.012* (0.007) 0.012* (0.006) 0.021*** (0.007) -0.050** (0.023) 0.262 0.255 0.269 0.268 (b) Pancasila/Civic Share INPRES × post-1972 Dependent Variable Mean -0.001 (0.001) n/a 0.026 -0.004* (0.002) 0.008*** (0.003) 0.060 0.041 (c) Arabic Share INPRES × post-1972 Dependent Variable Mean 0.002 (0.001) 0.003* (0.001) 0.001 (0.002) 0.014*** (0.004) 0.056 0.051 0.064 0.057 (d) Bahasa Indonesia Share INPRES × post-1972 -0.003* (0.002) -0.001 (0.003) -0.006* (0.003) 0.002 (0.002) Dependent Variable Mean 0.055 0.007 0.121 0.081 Number of Observations Number of Districts 16,889 263 8,559 245 5,077 250 3,251 225 Notes: This table presents estimates from a modified version of equation (2). We use an unbalanced panel at the schoolgrade × district × year level, including only years in which the given district had any schools enter. The estimating equation is ysjt = α + β(IN P RESj × P ost1972t ) + (Xj × P ost1972t )0 Θ + ηs + µj + δt + εisjt , where s is a school-grade and other terms are defined as in equation (2). The dependent variable measures the mean share of weekly instruction time devoted to Islamic subject material in panel (a), Pancasila and civic education in panel (b), Arabic instruction in panel (c), and instruction of the national language and literature, Bahasa Indonesia in panel (d). The measures come from the SIAP registry for the 2018–19 school year, and we categorize subject material using a procedure detailed in Appendix C. It is not possible to identify Pancasila and civic subjects for primary schools (see the discussion in the text) and hence the omission of column 2 in panel (b). All specifications include district fixed effects, grade level fixed effects, year-of-entry fixed effects, and a post-1972 dummy interacted with the 1971 children population, the 1971 enrollment rate, and exposure to the water and sanitation program. * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by district. Table 7: INPRES Exposure, Identity, and Religiosity (a) Identity, Proxied by Language National Language Use at Home All Muslims Non-Muslims Which Sub-Sample? INPRES × young Observations Number of Districts Dep. Var. Mean Arabic Literacy Islamic- SecularEducated (4) (5) (6) All (1) (2) (3) -0.0018 (0.0014) -0.0040** (0.0017) -0.0001 (0.0020) 0.0046* (0.0028) 0.0179* (0.0103) 0.0038 (0.0027) 31,678,510 273 0.166 27,811,101 273 0.150 3,867,324 273 0.275 839,026 275 0.343 25,935 275 0.688 813,087 275 0.332 Pay Zakat (7) Index Group (6) (b) Islamic Piety and Practice INPRES × young Observations Number of Districts Dep. Var. Mean Pray 5x daily (1) Fast during Ramadan (2) Reads the Qur’an (3) Friday (4) Prayer: Sunna (5) 0.0615 (0.0400) 0.0029 (0.0217) 0.0666** (0.0318) 0.0614** (0.0280) 0.0886*** (0.0239) 0.0847** (0.0343) -0.0138 (0.0220) 0.0537*** (0.0149) 1,864 150 0.655 1,865 150 0.811 1,860 150 0.267 1,856 150 0.226 1,846 150 0.176 1,859 150 0.246 1,860 150 0.832 1,866 150 0.436 (8) Notes: This table reports estimates of equation (1) using data from multiple sources. The dependent variable in columns 1–3 of panel (a) is an indicator for whether the individual speaks the national language, Bahasa Indonesia, as his/her main language at home. The data come from the complete-count 2010 Population Census. Columns 4–6 in panel (a) look at an indicator for whether an individual reports literacy in Arabic in the annual Susenas data from 2012 to 2018. Panel (a) sample splits across Muslims and non-Muslims in the Population Census (where we do not observe Islamic education) and across Islamiceducated and non-Islamic-educated in Susenas (where we do not observe religion). The specifications in panel (a) are restricted to mothers and fathers (husbands and wives) that fall within the original birth cohorts: aged 2–6 (young) or 12–17 in 1974. The dependent variables in panel (b) include indicators for whether an individual reports partaking in a range of Islamic practices as reported in the Pepinsky et al. (2018) survey data from 2008. The final column is a mean index across all 7 prior outcomes. The sample in panel (b) is restricted to Muslim respondents and compares individuals aged 6 or less (young) in 1974 with individuals aged 12 or more in 1974. The national language regressions in columns 1–3 of panel (a) also include around 1,200 ethnicity fixed effects. The specification is otherwise identical to that in Table 2; see the notes therein for details. * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by district (of birth). Table 8: INPRES Exposure and Ideology (1) (2) (3) (4) (5) (a) Citizens Supports Pancasila Which Sub-Sample? INPRES × young Number of Individuals Number of Districts Dep. Var. Mean Supports sharia Subjective Objective Muslims Muslims All Non-Muslims Muslims 0.0025 (0.0315) 0.2450** (0.1179) 0.0088 (0.0378) -0.0040 (0.0222) -0.0311 (0.0328) 2,034 159 0.840 205 33 0.927 1,798 145 0.829 1,790 145 0.433 1,703 145 0.681 (b) Candidates INPRES × young Number of Candidates Number of Districts Dep. Var. Mean Golkar Party United Development Party (PPP) -0.0106* (0.0059) 0.0073* (0.0043) -0.0112* (0.0059) 0.0019 (0.0021) -0.0111** (0.0055) 17,710 273 0.119 17,710 273 0.046 17,710 273 0.117 17,710 273 0.027 17,710 273 0.111 Platform Appeal Nation Building Islam Nation Building Excl. Islam Notes: This table reports estimates of equation (1) for ideological outcomes. The dependent variable in columns 1–3 of panel (a) is an indicator for whether the individual supports the national, inclusive secular ideology of Pancasila, or thinks some other ideology would be preferable. The data come from the Pepinsky et al. (2018) survey data from 2008, and we examine the outcome separately for Muslims and non-Muslims. Columns 4 and 5 consider measures of support for the sharia law. Column 4 is an indicator for whether the Muslim respondent express strong or very strong support for the implementation of sharia law. Column 5 is a mean index across several specific components of sharia law (e.g., prohibiting interest, mandating hijab for women), each of which is elaborated in Appendix Table A.17. The specification in panel (a) compares individuals aged 6 or less (young) in 1974 with individuals aged 12 or more in 1974. The dependent variables in panel (b) are based on legislative candidates in 2019. Columns 1 and 2 are indicators for whether the candidates are running on the party tickets of Golkar (Suharto’s party) and the Islamic United Development Party (PPP), respectively. Columns 3–5 are indicators for whether the candidate’s campaign platform mentions concepts that appeal to Indonesian nation building and Pancasila (column 3), to Islam and religious themes (column 4), and nation building exclusive of Islam and religious themes. The specifications in panel (b) are restricted to the original birth cohorts: aged 2–6 (young) or 12–17 in 1974. The specification is otherwise identical to that in Table 2; see the notes therein for details. * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by district. Table 9: INPRES Exposure and Religious Cultural Transmission Horizontal Transmission: Child Islamic Education Elementary Islamic Jun. Sec. Islamic Sen. Sec. Islamic (1) (2) (3) (4) (5) (6) INPRES × young (Father) -0.0010** (0.0004) INPRES × young (Mother) Observations Number of Districts Dependent Variable Mean R2 0.0011* (0.0006) -0.0009** (0.0004) 304,048 275 0.005 0.014 0.0007 (0.0006) 246,060 275 0.005 0.014 Marriage Matching 304,048 275 0.026 0.026 INPRES × young (Father) 0.0019** (0.0009) INPRES × young (Mother) Observations Number of Districts Dependent Variable Mean R2 (2) 544,174 275 0.024 0.024 304,048 275 0.034 0.029 246,060 275 0.033 0.028 Arabic in the Home Parents & Children (3) (4) Child’s Arabic No Islamic Schooling (5) (6) 0.0049* (0.0027) 0.0069* (0.0036) 0.0000 (0.0007) 725,803 275 0.039 0.035 246,060 275 0.026 0.028 0.0021** (0.0010) Vertical Transmission Arabic Literacy Islamic-Educated Partner (1) 0.0025** (0.0012) 0.0053** (0.0026) 304,048 275 0.213 0.111 246,060 275 0.268 0.137 0.0053 (0.0046) 95,678 272 0.877 0.047 77,068 272 0.887 0.042 Notes: This table reports estimates of a modified version of equation (1) where young now denotes the INPRES exposure of a parent (father or mother). INPRES refers to SD INPRES schools constructed from 1973–78 per 1,000 children in 1971. In panel (a), the dependent variable is no longer an individual’s own Islamic education exposure but their children’s education. In panel (b), we look at spouses’ Islamic education in columns 1–2. In columns 3 and 4, the dependent variable is an indicator for all 3 members of the household (father, mother, and child) being literate in Arabic. In columns 5 and 6, the dependent variable is an indicator equal to 1 if the child is literate in Arabic, conditional on the parent being literate in Arabic and the child having received no Islamic schooling. All specifications are restricted to children with mothers and fathers (or to husbands and wives) that fall within the original birth cohorts: aged 2–6 (young) or 12–17 in 1974. We restrict to co-resident children that are at least 18 years old and hence likely to have completed their secondary schooling. The regressions additionally control for child birth cohort fixed effects. The specification is otherwise identical to that in Table 2; see the notes therein for details. * p<0.1, ** p<0.05, *** p<0.01. Robust standard errors clustered by the parent’s district of birth.