Does Money Matter? The Effects of Block Grants on Education - PowerPoint PPT Presentation

Does Money Matter? The Effects of Block Grants on Education Attainment in Rural China: Evidence from Intercensal Population Survey 2015 Wei Ha, Peking University Fang Yan, UNICEF China 2017/3/17

Education attendance by age in China

Education profile of labor force in China Education profiles of the labor force in China, 1982-2010 Year <Primary Junior High Senior High Some College + 1982 63 26 11 1 1990 59 29 11 2 2000 41 42 13 5 2005 37 44 12 7 2010 27 49 14 10 Education profiles of the labor force in Urban China, 1982-2010 Year <Primary school Junior High Senior High Some College+ 1982 52 31 15 2 1990 52 32 13 3 2000 39 44 13 5 2005 16 41 24 19 2010 9 40 25 26

Rural Education Finance Reform in China (1980-2005)

Why study rural education finance reform (REFR)? • The largest intergovernmental transfer program in education in China announced in December 2005. • Four arms of the program: – School fees exemption, free textbooks and stipend for boarding students: (“Liangmian Yibu”) – Grants on recurrent expenditure – Major maintenance and repair costs – Teacher salary

The rise in the level of operation expenditures Rural Primary School Rural Middle School 2009 2010 2011 2014 2009 2010 2011 2014 Western and Central Provinces 300 400 500 600 500 600 700 800 Eastern Provinces 300 450 550 650 500 650 750 850 Sources: Authors' own compilation based on published government documents. Li Chelan (2015)

Key institutional features of the block grant • Geographic phase-in: West. China in 2006 and national rollout in 2007 • Differential subsidy rates across and within provinces – Western provinces: 80% – Central provinces: 60% but 243 counties treated as • Unprecedented publicity western counties; campaign – Eastern provinces: case by • Treasury single account case except BJ SH TJ – Provinces share costs with • Special office and prefectures and counties monthly monitoring report

Literature review on REFR • Existing research relies on household surveys – Hu and Lu (2010) no effects on enrollment 69 counties in 2005&2006 and a value-added model – Wang (2009) lowered the middle school dropout rate by 6-13 percentage points 4 counties 2005&2007 in DID – Ha et al (2016) use cross-county variation in the reform and cross- cohort comparison in DID strategy – Xiao et al (2017) exploit the cross-province variation in the roll-out of the reform and apply a DID strategy • Limitations – Cost sharing at and below provincial level largely neglected – Sample of counties very small and external validity questionable

Key treatment variable: differential block grants from higher level to county governments Table X Birth Cohorts and their exposure to the block grant Years Birth time affected by 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 policy primary school junior middle school senior middle school 1989.9—1990.8 0 1 2 3 4 5 6 1 2 3 1 2 3 primary school junior middle school senior middle school 1990.9—1991.8 0 1 2 3 4 5 6 1 2 3 1 2 3 primary school junior middle school senior middle school 1991.9—1992.8 1 1 2 3 4 5 6 1 2 3 1 2 3 primary school junior middle school senior middle school 1992.9—1993.8 2 1 2 3 4 5 6 1 2 3 1 2 3 primary school junior middle school senior middle school 1993.9—1994.8 3 1 2 3 4 5 6 1 2 3 1 2 3 primary school junior middle school senior middle school 1994.9—1995.8 4 1 2 3 4 5 1 2 3 6 1 2 3

Identification strategy: Cross-cohort and cross- county variation Education Attainment Diff-in-diff Counties with higher grants Counties with lower grants “ Endline” Baseline difference difference Older cohorts (1988-1991) Younger cohorts (1992-1995) Figure 2. Difference-in-difference strategy and rural education finance reform

1% population sample survey China 2015 • Two-stage cluster sampling method:14 million people from 60,000 sampling units. • Questions about gender, age, ethnicity, education attainment, migration, employment, social security, marriage, fertility and housing conditions. • Access the 10% sample of the mini census 2015. Variable Obs Mean Std. Dev. Min Max eduyrs 80,305 10.62336 2.997485 0 18 ifcompulsory 80,465 0.9036227 0.2951098 0 1 highsch 80,465 0.4545952 0.4979372 0 1 evermar 80,303 0.4111677 0.4920486 0 1 subsidy 77,063 0.9771414 0.0525239 0.5 1 subsidy1 77,063 0.9331037 0.0865397 0.5 1 trtgrp 80,465 0.4209284 0.4937112 0 1 female 80,465 0.4934195 0.4999598 0 1 ifminority 80,465 0.154241 0.3611818 0 1

DID results (1): Effects on completing compulsory education 1988-1995 （1） （2） （3） （4） VARIABLES -0.325*** -0.321*** -0.200*** 0.037 subsidy [0.051] [0.046] [0.036] [0.066] 0.018 -0.013 -0.113*** trtgrp [0.035] [0.032] [0.030] -0.015 0.027 0.111*** 1.trtgrp#c.subsidy [0.036] [0.033] [0.030] -0.010*** -0.008*** female [0.002] [0.002] -0.184*** -0.099*** ifminority [0.012] [0.009] -0.014*** poverty county=1 [0.004] -0.034*** urban_area=1 [0.003] prefecture FE NO NO NO YES Observations 77,063 77,063 77,063 77,063 R-squared 0.004 0.004 0.054 0.142 Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1

DID results (2): Effects on years of education （1） （2） （3） （4） 1988-1995 VARIABLES -2.545*** -2.927*** -1.937*** -1.969** subsidy [0.666] [0.642] [0.538] [0.840] trtgrp -0.084 -0.346 -1.004** trtgrp [0.400] [0.389] [0.423] 0.498 0.845** 1.034** 1.trtgrp#c.subsidy [0.410] [0.398] [0.430] 0.008 -0.004 female==1 [0.023] [0.021] -1.516*** -0.733*** Ethnic minority==1 [0.091] [0.073] -0.095** poverty county=1 [0.044] -0.946*** urban_area=1 [0.030] prefecture FE NO NO NO YES Observations 76,905 76,905 76,905 76,905 R-squared 0.002 0.007 0.039 0.133 Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1

Placebo test 1: High school completion 1988-1995 （1） （2） （3） （4） VARIABLES -0.182** -0.227** -0.211** -0.058 subsidy [0.073] [0.097] [0.095] [0.509] -0.379*** -0.371*** -0.546*** trtgrp [0.123] [0.122] [0.117] 0.062 0.055 0.071 1.trtgrp#c.subsidy [0.127] [0.126] [0.120] 0.008 0.012 female==1 [0.008] [0.008] -0.070*** -0.03 Ethnic minority==1 [0.021] [0.019] -0.054* poverty county=1 [0.032] prefecture FE NO NO NO YES Observations 15,983 15,983 15,983 15,983 R-squared 0.001 0.097 0.099 0.154 Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1

Placebo test 2: years of schooling 1980-1986 （1） （2） （3） （4） VARIABLES -4.460*** -4.931*** -3.521*** -1.923** subsidy [0.770] [0.829] [0.606] [0.805] -0.604 -0.522 -0.096 trtgrp [0.374] [0.362] [0.335] 0.914** 0.827** 0.641* 1.trtgrp#c.subsidy [0.386] [0.374] [0.344] -0.277*** -0.311*** female==1 [0.023] [0.021] -1.597*** -0.699*** Ethnic minority==1 [0.106] [0.076] -0.155*** poverty county=1 [0.042] -1.340*** urban_area=1 [0.035] prefecture FE NO NO NO YES Observations 60,818 60,818 60,818 60,818 R-squared 0.007 0.009 0.051 0.182 Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1

Placebo test 3: years of schooling 1983-1989 （1） （2） （3） （4） VARIABLES -3.621*** -4.017*** -2.730*** -2.700*** subsidy [0.682] [0.759] [0.546] [0.797] trtgrp -0.268 -0.176 -0.003 trtgrp [0.443] [0.377] [0.331] 0.666 0.544 0.631* 1.trtgrp#c.subsidy [0.454] [0.387] [0.339] -0.191*** -0.231*** female==1 [0.022] [0.021] -1.554*** -0.779*** Ethnic minority==1 [0.103] [0.072] -0.127*** poverty county=1 [0.044] -1.194*** urban_area=1 [0.032] prefecture FE NO NO NO YES Observations 69,681 69,681 69,681 69,681 R-squared 0.004 0.009 0.044 0.154 Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1

Placebo test 4: years of schooling 1974-1980 （1） （2） （3） （4） VARIABLES -4.984*** -4.920*** -3.244*** 1.094 subsidy [0.644] [0.637] [0.488] [0.874] 0.405 0.66 1.132*** trtgrp [0.483] [0.464] [0.418] -0.074 -0.322 -0.595 1.trtgrp#c.subsidy [0.494] [0.474] [0.427] -0.446*** -0.483*** female==1 [0.022] [0.021] -1.627*** -0.745*** Ethnic minority==1 [0.110] [0.076] -0.181*** poverty county=1 [0.041] -1.454*** urban_area=1 [0.034] prefecture FE NO NO NO YES Observations 60,526 60,526 60,526 60,526 R-squared 0.007 0.011 0.059 0.204 Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1