Evidence from Special Economic Zones in India (Figures)

Figure 1: Construction of the Cell-level Panel Note: Figure shows how the cell-level panel is constructed. A grid of 1 square-kilometer cells is overlaid on the map of India and cells that are strictly outside geo-coded SEZ locations, which are depicted by the circles, are considered for the analysis of spillover effects. 29 Figure 2: Event Study of SEZs and SEZ Neighborhoods (a) Event: Operation, Inside SEZs (b) Event: Notification, Inside SEZs (c) Event: Operation, SEZ Neighborhood (d) Event: Notification, SEZ Neighborhood Note: Figures plot 0.01*percentage changes in NTL backed out from γk0 s in Specification 1. Figures (a) and (b) are obtained from cells within SEZs and (c) and (d) are results from cells within 3 kilometers away from SEZs. The base year is the year before operation for Figures (a) and (c) and the year before notification for Figures (b) and (d). Cell and year fixed effects are included. 95% confidence intervals are generated based on spatial HAC errors with 30 kilometer cutoff. 30 Figure 3: Effect of SEZs across Distances (a) Long Run Effect: NTL (b) Short Run Effect: NTL (c) Long Run Effect: NTL per population (d) Short Run Effect: NTL per population Note: Figures (a) and (c) plot 0.01*percentage changes in NTL and NTL per population, respectively, backed out from δk ’s in Specification 1, and figures (b) and (d) present 0.01*percentage change in NTL and NTL per population, respectively, obtained from γk ’s in Specification 1. The base period is the pre-notification period of an SEZ in distance ring x. Cell and year fixed effects included. 95% confidence intervals are generated based on spatial HAC errors with 30 kilometer cutoff. 31 Figure 4: Testing the Strength of Identification: Using NTL (a) Notified SEZs that became operational eventually (b) Notified SEZS that did not reach operation Note: Figures (a) and (b) compare 5 kilometer neighborhoods around SEZs that have begun operation before 2013, and those that have been notified but not yet operational by 2013. The figures plot 0.01*percentage changes in NTL with the the year before notification being the base year. Cell and year fixed effects are included. 95% confidence intervals are generated based on spatial HAC errors with 30 kilometer cutoff. 32 Figure 5: Testing for Pre-trends in District-level Manufacturing Outcomes (a) Actual Trend: Production (b) Linear Fit: Production (c) Actual Trend:Total assets (d) Linear Fit: Total assets (e) Actual Trend: Employment (f) Linear Fit: Employment Note: Treated districts are those with at least one operational SEZ before 2011 while control districts are those with at least one notified SEZ, none of which are operational before 2011. Production, assets and employment are means across district totals, in logged values. Standard errors for differences in trend are reported in parentheses. 33 Table 1: Summary of SEZ & NTL Data Mean Median Std. Dev. Panel A: All SEZs Year of notification 2008.1 2008.0 1.89 Area (sq. km) 1.47 0.27 4.83 Public 0.29 0.00 0.21 Manufacturing 0.24 0.00 0.43 IT/electronics/engineering 0.69 1.00 0.47 Neighborhood NTL (2000) 20.0 11.6 18.6 Neighborhood NTL (2005) 19.8 11.1 19.2 Neighborhood NTL per population (2005) 0.030 0.017 0.048 Neighborhood NTL per population (2010) 0.054 0.034 0.064 Number of observations 251 Panel B: Operational SEZs Year of notification 2007.3 2007.0 1.30 Year of operation 2009.7 2010.0 2.23 Number of observations 133 N ote : An SEZ is considered public if any district or state agency was involved in the development process. Neighborhood of an SEZ is defined as the area within 3km away from the boundary of the SEZ. Source : Ministry of Commerce & Industry, National Centers for Environment Information, NASA 34 Table 2: Falsification Test: Impact on Districts before SEZ Notification Eventual Number of SEZs: Operating (γ) Notified (δ) 0.009 0.023 (0.03) (0.02) 0.004 0.032 (0.03) (0.02) 0.015 0.006 (0.01) (0.01) -0.002 0.016 (0.02) (0.01) 0.015 0.005 (0.02) (0.01) Production Assets used Employment Labour Productivity Wage Industry FE Year FE Y Y N ote : This table contains results from Specification 1. All outcome variables are average logged firm-level variables in the formal sector of a district-industry group. Industry and year fixed effects included, and standard errors are clustered at district level and are reported in parentheses. Source : ASI data on formal manufacturing firms (20002005) 35 Table 3: Comparison of Pre-Treatment Averages Across Districts Variable Control Treated Other Districts Density(Sq. Km) 4,059 3,988 431.8 Primary and Below (%) 66.8 61.7 74.8 Higher Secondary and Below (%) 95.8 92.4 97.4 Demographics Worker Composition Formal Employment1) 0.21 0.27 0.16 HH Employment 0.47 0.46 0.57 Manufacturing 0.67 0.59 0.28 Trade 0.11 0.13 0.03 Services 0.15 0.22 0.05 Agriculture 0.07 0.06 0.64 Formal Firm1 0.07 0.03 0.01 Firms with No Hired Workers 0.38 0.40 0.64 With Power 0.40 0.34 0.24 Average Firm Size 6.20 3.60 2.41 Manufacturing 0.14 0.16 0.2 Services 0.36 0.32 0.24 Trade 0.47 0.49 0.39 1,263.1 1,665.5 934.5 28 40 167 Year of Earliest Formal Approval Year of Earliest Notification 2007 2008.4 2006.1 2006.7 - Year of Earliest Operation 2012.3 2008.3 - Firm Composition Income Monthly Earnings (Rs.) Districts N ote : All values are from 2005 (or 2001 in the case of Census variables). 1) Formal Employment takes the value of 1 if firm employs over 10 workers, 0 otherwise. Source : ASI, NSS Employment Unemployment Surveys, NSS Unorganized Manufacturing, Economic Census, Census Digital Library, Ministry of Commerce & Industry 36 Table 4: Effect of Additional Operating SEZ on Formal Manufacturing Firms All Firms (1) Dependent Variable Production Size 1: <=10 (2) No. Mean Operational 28,630,983 0.022 (3) (4) No. Mean Operational 2,649,814 0.015 (0.011) Assets Used 22,976,901 0.015 43.8 0.009 169,397 579,546 0.018 0.035 6.4 37 59,928 0.012 0.001 545,796 .024 0.005 Operational 9,626,209 0.019 0.014 39,340 0.011 627,814 .019 0.009 (8) No. Mean Operational 33,598,769 0.028 -0.002 14.6 0.001 2,446,087 0.015 0.015 41.7 0.007 0.008 1,045,494 .025 0.009 Operational 303,229,348 0.006 (0.003) 27,508,346 0.004 (0.007) 284.3 0.005 0.021 (0.002) 1,289,803 0.002 (0.003) 54,721 0.010 (0.003) (0.001) No. Mean (0.001) (0.004) 47,099 (10) (0.007) (0.002) 873,270 Size 4: >100 (9) (0.004) (0.007) (0.003) (0.001) Size 3: (20,100] (7) (0.003) (0.005) (0.003) New Firm1) Mean (0.002) (0.011) Wages No. (0.015) (0.004) Labour Productivity (6) (0.007) (0.008) Employment Size 2: (10,20] (5) (0.002) 74,608 0.016 (0.003) 0.030 0.011 (0.001) (0.007) 0.020 0.007 (0.001) (0.001) District FE Y Y Y Y Y Industry FE Y Y Y Y Y Year FE Observations Y Y Y Y Y 122,624 18,554 21,919 39,533 42,618 N ote : This table reports the results from running Specification 2 separately for each dependent variable and size category. All means are in Rupees (except for Employment and New) and are reported in numbers and proportions respectively. Dependant variables listed in Column 1 are at the firm-level in logged values. District, industry and year fixed effects are included, and standard errors are clustered at district level and are reported in parentheses. See Table A.4 in the Appendix for coefficient on number of notified and not yet operational SEZs. 1) New Firm takes the value of 1 if firm formed after first SEZ gets notified in the district, 0 otherwise. Source : ASI data (2000-2009) Table 5: Effect of Additional Operating SEZ on New vs. Old Formal Firms Production Assets Used Employment Labour Productivity Wage Full Sample Old New 0.022 0.013 0.185 (0.011) (0.004) (0.044) 0.015 0.013 0.015 (0.008) (0.006) (0.008) 0.009 0.010 -0.005 (0.004) (0.003) (0.028) 0.018 0.004 0.171 (0.011) (0.002) (0.041) 0.012 0.004 0.051 (0.003) (0.001) (0.028) District FE Y Y Y Industry FE Y Y Y Year FE Observations Y 122,624 Y 105,454 Y 17,170 N ote : This table reports the results from running Specification 2 for the logged firm-level values of the dependent variable separately for two groups of firms- old and new. Firms are categorized as old if they have been operating for 4 years or longer. Industry, district, and year fixed effects are included. Standard errors are clustered at district level and are reported in parentheses. Source : ASI Data (2000-2009) 38 Table 6: Impact on the Informal Sector Manufacturing Services Dependent Variable Mean (1) (2) Mean (3) (4) Production 73,967 -0.765 -0.826 71,264 0.243 0.177 (0.26) (0.38) (0.151) (0.25) -0.646 0.228 0.213 (0.108) (0.18) 0.423 0.444 (0.194) (0.30) 0.182 0.192 (0.060) (0.10) Gross Value Added 18,657 -0.578 (0.18) (0.24) Assets used 200,938 -0.401 -0.419 (0.14) (0.19) -0.192 -0.157 (0.08) (0.11) -0.372 -0.452 (0.12) (0.17) -0.545 -0.618 (0.20) (0.30) -0.574 -0.626 (0.17) (0.25) Employment Gross Value Added per worker Labor Productivity Wage 3.9 4,836 18,506 3,064 38,715 106,424 1.8 21,216 39,054 1,586 0.044 0.022 (0.084) (0.14) 0.058 -0.017 (0.122) (0.20) 0.412 0.435 (0.168) (0.24) District FE Y Y Industy FE Y Y District-Industry FE Year FE Observations Y Y 59,233 Y Y Y Y 42,056 N ote : This table reports the estimated coefficients to the indicator for district treated with at least one operational SEZ before 2011 (α2 ) from Specification 3. Dependant variables listed in Column 1 are at the firm-level and in logged values. All means in Rupees, except for employment reported in numbers. Columns (1) and (3) include district, industry, and year fixed effects. Columns (2) and (4) include district-industry and year fixed effects. Standard errors are clustered at district level and are reported in parentheses. Source : NSS Unorganized Manufacturing & Services 39 Table 7: Impact on Total Informal and Formal Economic Activity Panel A. Formal Sector1) Dependant Variable Production Manufacturing Employment Investment Wage Labor Productivity 0.385 0.166 0.316 0.130 0.214 (0.12) (0.08) (0.14) (0.04) (0.08) Panel B. Informal Sector1) Dependant Variable Manufacturing Services Labor Production Employment Investment Wage -0.694 -0.278 -0.416 -0.574 -0.545 (0.237) (0.149) (0.163) (0.17) (0.20) Productivity 0.605 0.555 0.732 0.412 0.058 (0.356) (0.288) (0.382) (0.168) (0.122) Panel C. Overall Dependant Variable Household Employed in Number of 2) Firms Number of Informal Firms2), 3) Employment Firms>10 Manufacturing -0.082 (0.036) 0.079 (0.036) 0.078 (0.463) -0.238 (0.137) Services -0.039 0.056 0.117 -0.200 (0.034) (0.022) (0.414) (0.103) N otes : This table reports the estimated coefficients to the indicator if district is treated with an operational SEZ before post-treatment year, from Specification 3. Regressions are carried out separately for manufacturing and service sectors, and dependent variables are district-industry totals in logged values. District and year fixed effects are included. Standard errors are clustered at district level and are reported in the parentheses. 1) Post-treatment period for formal sector analysis at district level includes until 2009; For informal sector and worker level analysis, the analysis covers until 2011. 2) Analysis uses the Economic Census data with post treatment year 2012. 3) Informality is defined according to the 10-worker rule. Source : Source: ASI, NSS (Unorganized Manufacturing & Services, Employment survevy), Economic Census 40 Table 8: Wage Effects across the Income Distribution: Quantile Regression Results Quantile Regression: OLS 90% 75% Median 25% 10% 0.351 0.195 0.139 0.052 -0.042 0.165 (0.101) (0.100) (0.108) (0.118) (0.106) (0.082) District FE Y Y Y Y Y Y Industry FE Year FE Y Y Y Y Y Y Y Y Y Y Y Y Observations 63,782 63,782 63,782 63,782 63,782 63,782 Treatment × After N otes : This table reports the average and quantile treatment effects from regressions run on logged monthly worker wages in the 90th, 75th, 50th, 25th and 10th percentiles of the distribution. District, 2-digit industry and year fixed effects are included. Standard errors are clustered at the district level and reported in parentheses. Source : NSS Employment and Unemployment Surveys 41 Table 9: Drivers of Wage Effect Firm Characteristics (Size/Formality)1) Worker Characteristics (Education) (., Primary] (Primary, Higher Secondary] (Higher Secondary,.] Informal Formal Panel A. Manufacturing Treatment × After -0.203 0.153 0.512 0.111 0.407 (0.264) (0.194) (0.21) (0.28) (0.13) 906.9 1,772.2 5,486.2 1,510.2 2,697.3 District FE Y Y Y Y Y Industry FE Y Y Y Y Y Year FE Y Y Y Y Y 15,388 6,686 9,727 5,746 5,314 Mean Wage (Rs.) 42 Observations Panel B. Service Treatment × After 0.340 -.041 0.394 0.022 0.329 (0.296) (0.137) (0.19) (0.13) (0.17) 1,465.6 2,565.7 6,310.7 1,737.1 4,722.1 District FE Y Y Y Y Y Industry FE Y Y Y Y Y Year FE Y Y Y Y Y 5,167 6,962 8,531 8,909 8,189 Mean Wage (Rs.) Observations N otes : This table presents the estimated coefficient to the Treatment × After term from Specification 3. Dependent variable is logged worker wages. Separate regressions are run for sectors formed from every combination of row and column headings. District, industry and year fixed effects are included. Standard errors are clustered at district level and are reported in the parentheses. 1) Firm considered formal if it employs greater than 10 workers, otherwise informal Source : NSS Employment Unemployment Surveys Table 10: Effect of SEZs on Composition of Educated Workforce Logged Worker Wages Treatment × After Mean Proportion Proportion of workers with Below Primary Below Secondary -0.054 -0.034 (0.03) (0.04) 0.59 0.92 Y Y District-industry FE Year FE Observations Y Y 303,055 303,055 N otes : This table reports the estimated coefficients to Treatment × After from Specification 3. District-industry and year fixed effects are included. Standard errors are clustered at district level and are reported in the parentheses. Source : NSS Employment Unemployment Survey 43