Evidence from Special Economic Zones in India Appendix

A A.1 Appendix Supplementary Figures and Tables Figure A.1: Timeline of SEZ Approval Procedure · Business plan re-check · Land deeds check · Financing plan re-check Notification Application Time In-principle/formal approval · Business plan · Land suitability · Financing plan Operation N ote : Figure shows the different stages that SEZs pass through before beginning operation. Refer to. Section 2 for detailed stage-wise explanation 44 Figure A.2: Pictorial representation of the Distance Dimension Specification Example: Analyzing the Effects of an SEZ at a distance of 4 to 5 kilometers For conducting the 4 kilometer distance ring analysis, we select all cells that have their closest SEZ between 4 and 5 kilometers, such as the one depicted above. We then control for all the other SEZs, in rings farther away, and in all stages of their life- before they become notified (period 0), after notification (period 1) and after operation (period 2). Specification 2 then isolates the effect of a notified but not yet operational SEZ in the 4 kilometer distance ring through the coefficient γ4 (the baseline is the effect of a period 0 SEZ in the same ring). Similarly the effect of an operational SEZ in the 4th kilometer ring is captured by the coefficient δ4 , our main coefficient of interest. 45 Table A.1: Benefits to SEZ Developers and SEZ Units Developers Administrative Units Single window clearance for Central and State level approval Exemption from Minimum Alternate Tax Exemption from Central and State Sales Tax Tax Service and Dividend Distribution Tax Duty-free domestic procurement of goods, services 100% Tax exemption for Year 1-5: 100% tax exemption 10 consecutive years Year 6-10: 50% tax exemption since SEZ notification Year 11-15: 50% of reinvested profits Infrastructural support Upper limit extended for managerial Others remuneration, external commercial borrowings allowed, etc. Flexible hiring and firing practices Source: Department of Commerce, Government of India 46 Table A.2: Summary of Key Variables of Workers and Firms before 2005 mean sd 25th pctl 50th pctl 75th pctl min max Demographics Density(/sq. Km) 3,989 2,606 2,279 3,597 4,845 250 19,865 Literacy Primary and below Secondary and below 63.48 0.59 0.92 10.60 0.49 0.28 56.67 0 1 63.37 69.68 37.49 0 0 83.40 1 1 Amenities1) Bank Density Primary School Density Secondary School Density 1.50 3.74 1.30 0.59 1.60 0.59 1.10 2.61 0.93 1.37 3.52 1.15 1.73 4.71 1.58 0.11 1.27 0.48 3.22 10.07 3.48 Wage2) Informal Manufacturing Informal Services Formal Manufacturing 8.5 8.1 11.0 1.7 1.4 0.8 7.4 7.3 10.5 8.1 8.2 11.0 9.9 9.0 11.5 3.8 2.0 4.9 12.5 11.0 14.2 Per capita Consumption2) 7.0 0.6 6.6 6.9 7.4 3.3 11.0 Firm Composition Greater than 10 Workers Any Registration Own Account Enterprise Unincorporated Operating with Power No External Finance Informal Finance—Financed Manufacturing Trade & Services Services Infrastructure 0.04 0.53 0.41 0.93 0.34 0.90 0.41 0.15 0.82 0.33 0.02 0.20 0.50 0.49 0.25 0.47 0.30 0.49 0.36 0.38 0.47 0.14 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 122.7 13.7 596.4 1.5 9.0 12.9 24.0 13.8 114.0 14.7 1.0 0.4 45,481 19.9 5.4 10.3 4.8 2.4 2.1 8.8 3.7 9.7 7.0 12.2 1.0 3.6 35.0 15.6 2.0 11.3 1.1 0.7 1.3 10.8 1.8 11.3 2.2 11.8 1.0 7.6 10.0 13.7 Firm Size Formal Manufacturing Employment3) Average Labour Productivity2) Informal Manufacturing Employment Average Labour Productivity Informal Services Employment Average Labour Productivity N otes : All averages are at the district-level, and when possible, at the village-level. Items under Amenities are values of 2001. 1) per 10,000 population, 2) logged real values, 3) in absolute numbers Source : Census Digital Library of India, NSS and ASI surveys, EC data 47 Table A.3: Treatment & Control Group Formation Level of analysis Pre-treatment Year(s) Post-treatment Year(s) Village/Town 2005 2013 NSS Unorganized Firms District 2000-01, 2004-05 2010-11 NSS Worker Survey District 2000-01, 2004-05 2010-11 ASI Formal Manufacturing District 2001 to 2005 2006 - 2009 Data-set Economic Census Considered Set Treated Set 1 or more notified SEZs by 2012 1 or more notified SEZs by 2010 1 or more notified SEZs by 2010 1 or more notified SEZs by 2012 1 or more operational SEZs by 2012 1 or more operational SEZs by 2010 1 or more operational SEZs by 2010 1 or more SEZs notified or operating each year Table A.4: Effect of Every Additional Notified SEZ on a Firm in Formal Manufacturing Dependent Variable Production Assets Used Employment Labour Productivity Wages New1) District-industry FE Year FE Observations All Firms 0.003 (0.005) 0.018 (0.008) 0.007 (0.004) 0.002 (0.005) 0.005 (0.001) 0.002 (0.001) Y Y 122,624 Firm size Size 2: (10,20] 0.002 (0.005) 0.017 (0.007) -0.000 (0.001) 0.002 (0.004) 0.004 (0.002) 0.002 (0.001) Y Y 21,919 Size 1: <=10 0.017 (0.013) -0.010 (0.011) -0.002 (0.002) 0.010 (0.007) 0.008 (0.002) 0.002 (0.001) Y Y 18,554 Size 3: (20,100] 0.006 (0.004) 0.020 (0.005) 0.000 (0.001) 0.003 (0.004) 0.005 (0.002) 0.002 (0.001) Y Y 39,533 Size 4: >100 0.004 (0.002) 0.001 (0.003) 0.005 (0.002) 0.004 (0.002) 0.020 (0.007) 0.001 (0.001) Y Y 42,618 N otes : This table reports the estimated coefficients to the number of notified, non-operational, SEZs in the district of the firm in a year from Specification 2. Dependant variables listed in Column 1 are at the firm-level, and enters in logged values, and the regressions are run separately for each dependent variable and size category. District-industry and year fixed effects are included. Standard errors are clustered at district level and are reported in the parentheses. 1) New takes the value of 1 if firm formed after first SEZ gets notified in the district, 0 otherwise. 48 A.2 Caveat to NTL Usage NTL cannot be a perfect substitute of traditional socioeconomic data. Each cell has a value between 0 and 63, which means that some lights are bottom- or top- coded. Studies that adopt nighttime lights usually suffer from the fact that non-negligible portion of their data is bottom-coded as most of them focus on underdeveloped countries. In our case, on the other hand, there is a high probability that we suffer from top-coded observations. This is because we focus on area that are more likely to be more developed within India, which is relatively developed among developing countries. Although the right-censoring might affect the empirical results, this would only underestimate the positive effects of SEZs on the neighborhood, if there are any. We, therefore, argue that the estimated spillover effects of SEZs are conservative. Figure A.3: Distribution of Nighttime Lights 49 A.3 Alternate specification for Events Study Analysis We carry out an alternate specification where we consider all cells that have at least one SEZ within 3 kilometers, where the event is the earliest notification/operation of an SEZ in the 3 kilometer-neighborhood. We run Specification 1 using this alternate definition of the event, and find no significant differences from the results of the specification in the main paper: Figure A.4 (a) Event: Operation, SEZ neighborhood (b) Event: Notification, SEZ neighborhood γk0 s Note: Figures plot .01*percentage change in NTL backed out from in Specification 1. The year before the event (operation/notification), year -1, is the base year. Cell and year fixed effects are included. 95% confidence intervals are generated based on spatial HAC errors with 30 kilometer cutoff. 50 A.4 Extending Distance Analysis to 20 kilometers Figure A.5 (a) Long-run effect: NTL (b) Short-run effect: NTL Note: Figures plot .01*percentage change in NTL backed out from δx0 s in Specification 2 where the analysis extends beyond 15 kilometers and up to 20 kilometers away from SEZs. The number of observations (cells) is 95,239 per year. 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. 51 A.5 Alternate specification for Distance Dimension Analysis We carry out an alternate, less restrictive specification by considering all cells with at least one SEZ in the distance ring x and controlling for both nearer and farther away SEZs when studying the effects of SEZs at that particular distance ring around the cell. For a cell i that is situated outside of SEZs in year t and has one or more SEZ in x − 1 to x km (1 ≤ x ≤ 15), log(lightit ) = αi + βt + γx ∗ period1ixt + δx ∗ period2ixt d6=x + 2 X X λpd periodθidt + it , 1≤d≤15 θ=0 The long.run effects due to operating SEZs are displayed on the left, while the short-run effect due to notified SEZs is in the bottom. Cell and year fixed effects are included. 95% confidence intervals are generated based on spatial HAC errors with 30 kilometer cutoff. Figure A.6 (a) Long-run effect: NTL (b) Short-run effect: NTL Note: Figures plot .01*percentage change in NTL backed out from δx0 s in the altered version of Specification 2 where all cells with at least one SEZ in distance ring x are considered in the distance ring x analysis (instead of only those cells with their closest SEZ in distance ring x). 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. 52 A.6 SEZs’ Influence on Village-level Economic Activity We follow Specification 3 at the village level with the Economic Census data. Table A.5 reveals a large and significant increase in the total working population and an economically, if not statistically, significant increase in the number of firms in a treated village as compared to the control village. We also observe increased hiring among firms, with average size expanding by 13.5% (which translates to roughly one additional worker to an average firm) and with the proportion of firms with no hired workers decreasing by 4.3%. This analysis gives us a preliminary view of an expansion in industrial activity which is consistent with an expansion in NTL recorded over areas of similar dimensions.45 Table A.5: Village Level Analysis of Firm and Worker Numbers Dependent Variable Total Firms Total Workers Mean 330,675 2,190,081 Average Workers per Firm 6.2 Firms with 0 Hired Worker1) 0.38 Observations Treatment × After 0.252 (0.412) 0.630 (0.341) 0.135 (0.055) -0.043 (0.020) 2,497,090 N otes : This table reports the estimated coefficients to Treatment × After from Specification 3. All outcome variables are in logged values. District-industry and year fixed effects are included. Standard errors are clustered at district level and are reported in the parentheses. 1) Firms with 0 hired workers are also known as Own Account Enterprises (OAEs). 45 The 5-kilometer neighborhood around an SEZ spans an area of roughly 75 square kilometers, almost twice the area of a typical village in our study. 53 A.7 SEZs’ Influence on Sub-district level Population Movements We use the Gridded Population of the World (GPW) data in conjunction with sub-district level administrative boundaries.46 The 251 SEZs in our sample are situated in 126 sub-districts with the number of SEZs per sub-district varying between 1 and 12 (with a mean of 2.02 and a median of 1). We restrict our attention to those with at least one SEZ notified before 2010 and evaluate whether the ones with at least one operating SEZ show faster population growth. For sub-district i in district d at time t: log(populationidt ) = α0 + α1 Y eart + α2 Districtd + α3 Operatingidt (1) + α4 Operatingidt × Af tert + idt , where Operatingidt = 1 if there is at least one operating SEZ in subdistrict i and Af tert = 1 in year 2010. Standard errors are clustered at district level. The estimation result reported in Table A.6 suggests that there is no differential trend of population density growth between the sub-districts whose SEZs started operating before 2010 and those whose SEZs are only notified by 2010. In other words, it is not likely that there are population movement across subdistricts. This is consistent with the fact that Indian labor market tends to be spatially restricted, meaning that the labor mobility is low. Table A.6: Effect of Operating SEZs on Sub-district Population density Dependent variable Year 2005 Log of population density 0.875 (0.012) 0.164 (0.028) 0.332 (0.374) 0.029 (0.026) Y 321 0.629 Year 2010 Operating Operating × After District FE Number of observations Overall R2 N otes : District fixed effects are included, and year 2000 is omitted. Standard errors are clustered at district level and are reported in the parentheses. 46 Acquired from the Survey of India (http://www.surveyofindia.gov.in/). 54