Firms and Farms: The Impact of Agricultural Productivity on the Local Indian Economy

Size: px
Start display at page:

Download "Firms and Farms: The Impact of Agricultural Productivity on the Local Indian Economy"

Transcription

1 Firms and Farms: The mpact of Agricultural Productivity on the Local ndian Economy Gabriella Santangelo University of Cambridge May 4, 2017

2 Outline ntroduction Model Empirical Analysis: Agricultural Productivity Empirical Analysis: NREGA Summary and Conclusion

3 Motivation: Firms in Rural Economies Firms in the developing world operate in contexts primarily rural Agriculture still plays a central role in the economy n 2005, rural ndia accounted for: 70% of manufacturing establishments 60% of manufacturing employment How do changes in other sectors of the rural economy affect manufacturing firms?

4 Agricultural Productivity and Firms Consider an increase in agricultural productivity This shifts the demand for labor and raises the wage of agricultural workers But how does it affect firms? Does it affect the cost of labor for manufacturing firms? Does it reduce firms production and employment?

5 This Paper Two effects: 1. Higher wage 2. Higher income! Higher demand? f product markets are imperfectly integrated: Demand channel may dominate Firms production and employment may in fact rise This paper: Estimate the size of the demand channel and show that it is active and important

6 What Do 1. Simple multi-sector general equilibrium model of the local rural economy Highlight the cross-sectoral linkages in factor and product markets 2. Estimate the elasticity of firms wage, output and employment with respect to agricultural productivity Exploit weather-induced variation over time and across districts in ndia Examine mechanisms through unique combination of firm and household microdata

7 Preview of Findings 1. An increase in agricultural productivity: Raises the wage that firms pay Notwithstanding, raises firms production and employment 2. Effect is sizable 10% increase in rainfall raises manufacturing output by 1.2% 3. Effect driven by firms that produce locally-traded goods Negative (though insignificant) effect on firms that produce tradeables Key takeaway: Local rural incomes are an important determinant of the demand that firms face

8 The mpact of Policies What does this imply for policy? Extend the model and empirical analysis to examine the effect of a large-scale public employment program (NREGA) As agricultural productivity, NREGA shifts the demand for labor Existing work shows that it raises the local wage (mbert and Papp, 2015; Muralidharan et al., 2017) But how does it affect firms? Despite raising wages, it may benefit manufacturing if it raises income and demand gnoring demand channel may lead to a partial, and possibly misleading, assessment of the impact of NREGA

9 The mpact of NREGA on Firms NREGA introduced across ndia in three phases between 2006 and 2008 Test whether NREGA affects the response of firms production and employment to agricultural productivity shocks Can NREGA support local demand during bad times? dentification does not rely on common trend assumption across implementation phases Estimate the interaction of weather shocks and NREGA

10 Preview of NREGA Findings 1. NREGA acts as a wage floor t moderates the impact of adverse agricultural productivity shocks on wages and consumption By supporting consumption, it generates increased demand for locally-traded manufacturing goods 2. Large effect The demand effect is not there for firms that produce tradeables NREGA reduces elasticity of local wage, consumption and manufacturing outcomes by more than 70%

11 Related Literature Role of agricultural productivity in the development of the non-farm sector (Foster and Rosenzweig, 2004, 2007; Bustos et al., 2016) Role of market size for industrialization (Murphy et al., 1989; Magruder, 2013) Labor markets in developing countries (Rosenzweig, 1978; Jayachandran, 2006; Kaur, 2015) NREGA (Muralidharan et al., 2017; mbert and Papp, 2015; Fetzer, 2014) Local markets and economic growth (Alcott and Keniston, 2016; Autor et al., 2013; Moretti, 2011) Role of sector-specific shocks in aggregate fluctuations (Acemoglu et al., 2012; Caliendo et al., 2015)

12 Road Map 1. Simple multi-sector model of the local rural economy 2. Empirical analysis of the effects of agricultural productivity 3. Empirical analysis of the impact of NREGA

13 Outline ntroduction Model Empirical Analysis: Agricultural Productivity Empirical Analysis: NREGA Summary and Conclusion

14 Local Rural Economy To guide the empirical analysis, set up a simple multi-sector general equilibrium model of the local economy Empirically, district is the unit of analysis Treat district d as a small open economy Main purpose is to illustrate: 1. Effect of a change in d s agricultural productivity on d s farm and non-farm sectors Heterogenous effect on firms that sell goods within vs. outside district d 2. mpact of a public-works program such as NREGA on firms in district d

15 Multi-Sector Economy Economy (district) has three sectors Agriculture (A), non-farm tradable (T) and non-farm non-tradable (N) Representative firm in sector j produces output Y j using labor n j Yj = q j n a j,whereq j is a sector-specific productivity parameter q A captures agricultural productivity and is the driving force in the model

16 Set-up: Labor and Products Markets Product markets Agricultural and non-farm tradable sectors sell to world markets Prices pa and p T exogenously given Non-farm non-tradable sector sells to local market Labor market Price p N endogenous and determined by local demand conditions Labor receives wage w Labor is: mmobile across districts Mobile across sectors Given local mobility of labor, in equilibrium wage will be equated across sectors

17 Set-up: Consumption Representative agent has Cobb-Douglas preferences over agricultural goods (c A ), traded goods (c T )andnon-traded goods (c N ) Agent receives profits from agricultural and non-tradable sector Profits from tradable sector do not accrue locally Agent endowed with one unit of time, which is supplied inelastically to the labor market

18 Comparative Statics: The Local Effects of a Change in Agricultural Productivity An increase in agricultural productivity q A : 1. ncreases employment in the local agricultural sector: n A q A > 0 2. ncreases the equilibrium wage: w q A > 0 3. ncreases local income and consumption: q A > 0and c j q A > 0 for j 2{A,T,N} 4. Has opposite effects on the local tradable and non-tradable sectors: Employment in tradable firms decreases: n T q A < 0 Employment in non-tradable firms increases: n N q A > 0

19 Outline ntroduction Model Empirical Analysis: Agricultural Productivity Empirical Analysis: NREGA Summary and Conclusion

20 Rainfall and Agricultural Productivity Estimate the response of equilibrium outcomes to changes in agricultural productivity Source of exogenous variation: Yearly fluctuations in agricultural productivity induced by monsoon rainfall Trace the effect of rainfall on: Agricultural and non-agricultural wage Household consumption Firms production and employment in tradable/non-tradable sectors

21 Data Firm data: Annual Survey of ndustries: Household data: National Sample Survey Employment Schedule: National Sample Survey Consumption Schedule: Other data: Rainfall: Tropical Rainfall Measuring Mission (TRMM), Agricultural production: Crop Production and Land Use Statistics nformation System,

22 Empirical Specification log(y dt )=blog(r dt )+g d + t t + e dt Outcome y for district d at time t R dt : Total monsoon rainfall (June to September) g d :DistrictFE t t : Year FE Standard errors clustered at the district level Additional controls and/or FE in regressions at the individual-, household- or firm-level

23 Wage and Consumption Elasticities TABLE - WAGE AND CONSUMPTON ELASTCTY Log(Wage) Log(Consumption Expenditure) Log(Crop Yield) Non- Manufactured All Agricultural All Goods Agricultural Goods (1) (2) (3) (4) (5) (6) Log(Rainfall) 0.178*** 0.080*** 0.050*** 0.111*** 0.069*** 0.079*** (0.020) (0.020) (0.019) (0.030) (0.021) (0.022) Observations 6,763 89,429 44,955 44,474 83,212 83,205 District FE Yes Yes Yes Yes Yes Yes Phase-Year FE Yes Yes Yes Yes Yes Yes Notes: Unit of observation is a district in Column 1, an individual in Columns 2-4, and a household in Columns 5-6. n Columns 2-4, sample is restricted to individuals aged 18 to 60 and the regressions include controls for age, gender, years of schooling, landholdings and crop season.the regressions in Columns 5-6 include controls for landholdings and crop season. Rainfall refers to total rainfall in June-September. Standard errors are clustered at the district level. All estimates are obtained using sampling weights. *** p<0.01, ** p<0.05, * p<0.1

24 Firm Production, Employment and Wage Elasticities TABLE - FRM PRODUCTON AND EMPLOYMENT ELASTCTES Log(Wage) Log(Value of Output) Log(Man-days) Log(Workers) (1) (2) (3) (4) Log(Rainfall) 0.056*** 0.122** 0.066*** 0.058*** (0.017) (0.049) (0.025) (0.022) Observations 17,270 17,296 17,270 17,284 District FE Yes Yes Yes Yes Phase-Year FE Yes Yes Yes Yes Phase-ndustry FE Yes Yes Yes Yes ndustry-year FE Yes Yes Yes Yes Notes: Unit of observation is a district-industry-year. ndustry is defined at the NC 4-digit level. District-industryyear measures are obtained using sampling weights and are weighted in the regression by the corresponding weight. Rainfall refers to total rainfall in June-September. Value of output is in nominal terms. Number of workers and man- days include all types of firm employees. Wage refers to average real daily wage across firms -- CP for Agricultural Labourers is used. Daily wage at the firm level is obtained as total compensation during the year divided by total number of man-days. Standard errors are clustered at the district level. *** p<0.01 ** p<0.05 * p<0.1 AS Summary Stats

25 Tradable vs. Non-Tradable ndustries Classify industries in tradable vs. non-tradable Follow Holmes and Stevens (2014): Higher average product shipment distance = More tradable Examples Least tradable: Bricks, cardboard boxes, ice-creams Most tradable: Watches, X-ray equipment, aircraft parts Robustness: Additional classifications based on geographical concentration and international trade (Mian and Sufi, 2015) Tradable ndustries AS Summary Stats by Tradable Tradable by District

26 Production and Employment Elasticities by ndustry Type TABLE - FRM ELASTCTES BY NDUSTRY TYPE Log(Wage) Log(Value of Output) Log(Man-days) Non-tradable Tradable Non-tradable Tradable Non-tradable Tradable (1) (2) (3) (4) (5) (6) Log(Rainfall) 0.063*** 0.036* 0.164*** *** (0.018) (0.021) (0.059) (0.065) (0.033) (0.064) Observations 13,497 3,773 13,514 3,782 13,497 3,773 District FE Yes Yes Yes Yes Yes Yes Phase-Year FE Yes Yes Yes Yes Yes Yes Phase-ndustry FE Yes Yes Yes Yes Yes Yes ndustry-year FE Yes Yes Yes Yes Yes Yes Notes: Unit of observation is a district-industry-year. ndustry is defined at the NC 4-digit level. District-industryyear measures are obtained using sampling weights and are weighted in the regression by the corresponding weight. Rainfall refers to total rainfall in June-September. Value of output is in nominal terms. Number of workers and mandays include all types of firm employees. Wage refers to average real daily wage across firms -- CP for Agricultural Labourers is used. Daily wage at the firm level is obtained as total compensation during the year divided by total number of man-days. Standard errors are clustered at the district level. *** p<0.01 ** p<0.05 * p<0.1

27 Robustness 1. Rainfall measure 2. nput-output linkages 3. Tradability definition 4. Spatial correlation 5. Placebo tests Highly irrigated districts Non-monsoon rainfall

28 Outline ntroduction Model Empirical Analysis: Agricultural Productivity Empirical Analysis: NREGA Summary and Conclusion

29 The mpact of Policies Local rural demand is important for a large share of manufacturers This highlights the importance of accounting for demand effects in program evaluation Policies that increase wages may benefit manufacturing firms if they boost local demand Examine the effect of a large-scale public employment program (NREGA) NREGA entitles every household in rural ndia to 100 days of minimum-wage public employment per year ntroduced in the model as an additional sector that offers a fixed wage w G

30 The mpact of NREGA on the Local Labor Market How does NREGA impact the local labor market? 1. The availability of NREGA jobs induces a wage floor: w w G for any q A Threshold q A such that w = w G for any q A apple q A 2. NREGA acts as a counter-cyclical stimulus policy Labor market clears with a positive share of agents working for NREGA, n G Share of agents working for NREGA is decreasing in qa : n G q A < 0

31 NREGA as a Local Stabilizer NREGA acts as a stabilization policy. Specifically, NREGA reduces: 1. Wage elasticity: e G w,q A < e w,qa 2. ncome and consumption elasticity: e G,q A < e,qa and ec G j,q A < e cj,q A for j 2{A,T,N} 3. Employment elasticity of local non-farm tradable sector: e G n T,q A < e nt,q A 4. Employment elasticity of local non-farm non-tradable sector: e G n N,q A < e nn,q A

32 mpact on NREGA on Local Economic Activity 1. NREGA increases local wage Reduces local profits of agriculture and non-farm tradable sector Reduces production and employment of tradable firms 2. NREGA also increases local income May increase production and employment of non-tradable firms

33 The ntroduction of NREGA Test whether NREGA affects the response of firms production and employment to agricultural productivity shocks NREGA was gradually implemented throughout ndia Phase 1 (2006): 200 districts Phase 2 (2007): 130 districts Phase 3 (2008): All remaining districts Exploit the timing of implementation across districts to study impact on local non-farm sector

34 Empirical Strategy log(y idpt )=blog(r dt )+qlog(r dt ) NREGA pt + X idpt +g d +t pt +e idpt Outcome y for individual i in district d of NREGA-phase p at time t R dt : Rainfall NREGA pt = 1ifphasep has NREGA at time t; 0otherwise g d :DistrictFE t pt : Phase-year FE X idt :ndividualdemographiccontrols dentifying assumption: NREGA is not correlated with omitted variables that affect the rainfall-dependence of the local economy

35 The Effect of NREGA on the Rainfall-Dependence of Agricultural Productivity and Public Employment TABLE - THE EFFECT OF NREGA ON THE RANFALL-DEPENDENCE OF AGRCULTURAL PRODUCTVTY AND PUBLC EMPLOYMENT Log(Crop Yield) Days in Public Employment (1) (2) Log(Rainfall) 0.178*** (0.020) (0.013) Log(Rainfall) x NREGA *** (0.003) (0.009) Observations 6, ,601 District FE Yes Yes Phase-Year FE Yes Yes Notes: Unit of observation is a district-year in Column 1 and an individual in Column 2. n Column 2, sample is restricted to individuals aged 18 to 60. The regression in include controls for age, gender, years of schooling, landholdings and crop season. Rainfall refers to total rainfall in June- September. Standard errors are clustered at the district level. All estimates are obtained using sampling weights. *** p<0.01, ** p<0.05, * p<0.1

36 The Effect of NREGA on Wage and Consumption Elasticities TABLE - THE EFFECT OF NREGA ON WAGE AND CONSUMPTON ELASTCTY Log(Wage) Log(Consumption Expenditure) Non- Manufactured All Agricultural All Goods Agricultural Goods (1) (2) (3) (4) (5) Log(Rainfall) 0.062*** 0.057*** 0.075*** 0.056*** 0.063*** (0.018) (0.018) (0.021) (0.013) (0.014) Log(Rainfall) x NREGA *** *** *** *** *** (0.013) (0.018) (0.018) (0.011) (0.012) Observations 193,602 92, , , ,313 District FE Yes Yes Yes Yes Yes Phase-Year FE Yes Yes Yes Yes Yes Notes: Unit of observation is an individual in Columns 1-3 and a household in Columns 4-5. n Columns 1-3, sample is restricted to individuals aged 18 to 60 and the regressions include controls for age, gender, years of schooling, landholdings and crop season.the regressions in Columns 4-5 include controls for landholdings and crop season. Rainfall refers to total rainfall in June-September. Standard errors are clustered at the district level. All estimates are obtained using sampling weights. *** p<0.01, ** p<0.05, * p<0.1

37 The Effect of NREGA on Production and Employment Elasticities TABLE - THE EFFECT OF NREGA ON PRODUCTON AND EMPLOYMENT ELASTCTES Log(Wage) Log(Value of Output) Log(Man-days) Non-tradable Tradable Non-tradable Tradable (1) (2) (3) (4) (5) Log(Rainfall) 0.043*** 0.126*** *** (0.012) (0.041) (0.078) (0.025) (0.061) Log(Rainfall) x NREGA ** * (0.014) (0.044) (0.091) (0.028) (0.060) Observations 31,911 25,097 6,887 25,046 6,865 District FE Yes Yes Yes Yes Yes Phase-Year FE Yes Yes Yes Yes Yes Phase-ndustry FE Yes Yes Yes Yes Yes ndustry-year FE Yes Yes Yes Yes Yes Notes: Unit of observation is a district-industry-year. ndustry is defined at the NC 4-digit level. District-industryyear measures are obtained using sampling weights and are weighted in the regression by the corresponding weight. Rainfall refers to total rainfall in June-September. Value of output is in nominal terms. Number of workers and man-days include all types of firm employees. Wage refers to average real daily wage across firms -- CP for Agricultural Labourers is used. Daily wage at the firm level is obtained as total compensation during the year divided by total number of man-days. Standard errors are clustered at the district level. *** p<0.01 ** p<0.05 * p<0.1 Elasticities by Phase

38 Robustness Potential concern: NREGA treatment may be capturing 1. General over-time decline in rainfall-dependence of the local economy 2. mpact of other policies that affect rainfall-dependence Can be addressed by estimating: log(y jdpt )= Â s2{ 7,...,3} where z dt measures years-to-nrega b s log(r dt ) {z dt = s}+d d +t pt +J pj +r jt +e jdst

39 Time Series of Output Elasticity Figure : Output Elasticity Pre- and Post-NREGA Output Elasticity Years to NREGA

40 Time Series of Employment Elasticity Figure : Employment Elasticity Pre- and Post-NREGA Man-days Elasticity Years to NREGA

41 Time Series of Wage Elasticity Figure : Wage Elasticity Pre- and Post-NREGA Wage Elasticity Years to NREGA

42 The Level Effect of NREGA on Firms TABLE - THE EFFECT OF NREGA ON WAGE, PRODUCTON AND EMPLOYMENT Log(Wage) Log(Value of Output) Log(Man-days) Non-tradable Tradable Non-tradable Tradable (1) (2) (3) (4) (5) NREGA 0.043** 0.138** * (0.018) (0.067) (0.117) (0.043) (0.071) Observations 25,479 19,942 5,602 19,893 5,586 District FE Yes Yes Yes Yes Yes Phase-ndustry FE Yes Yes Yes Yes Yes ndustry-year FE Yes Yes Yes Yes Yes District Controls x Time Trend Yes Yes Yes Yes Yes Notes: Unit of observation is a district-industry-year. ndustry is defined at the NC 4-digit level. Districtindustry-year measures are obtained using sampling weights and are weighted in the regression by the corresponding weight. Rainfall refers to total rainfall in June-September. Value of output is in nominal terms. Number of workers and man- days include all types of firm employees. Wage refers to average real daily wage across firms -- CP for Agricultural Labourers is used. Daily wage at the firm level is obtained as total compensation during the year divided by total number of man-days. Standard errors are clustered at the district level. *** p<0.01 ** p<0.05 * p<0.1 Placebo Test

43 Outline ntroduction Model Empirical Analysis: Agricultural Productivity Empirical Analysis: NREGA Summary and Conclusion

44 Summary and Conclusion Exploit shifts in local labor demand to estimate the size of the demand channel Higher agricultural productivity and NREGA raise wages but also raise manufacturing activity Effect driven by locally-traded sector that benefits from increased demand The results highlight that: Demand channel is active and important Local rural demand matters for a large share of manufacturers Rural development policies can strongly affect the industrial sector because of GE effects

45 Manufacturing Firm Data Annual Survey of ndustries data are representative of ndian formal manufacturing sector TABLE - AS SUMMARY STATSTCS Characteristics Mean Median p10 p90 S.D. Number of Workers Value of Output (Millions Rupees) Fixed Capital (Millions Rupees) Daily Wage (Rupees) Output per Worker (Millions Rupees) Capital to Labor Ratio Return

46 Tradable vs. Non-Tradable ndustries ndustry Code TABLE - TRADABLE NDUSTRES ndustry Description Tradable ndustry Output ndustry Output / Total Output / Total ndustry Across Output ndustries 30 Office, Accounting and Computing Machinery Radio, Television and Communication Equipment and Apparatus Medical, Precision and Optical nstruments,watches and Clocks Tobacco Products Textiles Electrical Machinery and Apparatus N.E.C Other Transport Equipment Motor Vehicles, Trailers and Semi-Trailers Machinery and Equipment N.E.C Tanning and Dressing of Leather; Luggage, Handbags and Footwear Chemicals and Chemical Products Furniture; Manufacturing N.E.C Publishing, Printing and Reproduction of Recorded Media Wearing Apparel; Dressing and Dyeing of Fur Food Products and Beverages Wood and of Products of Wood and Cork, Except Furniture Coke, Refined Petroleum Products and Nuclear Fuel Other Non-Metallic Mineral Products Fabricated Metal Products, Except Machinery and Equipments Basic Metals Rubber and Plastic Products Paper and Paper Product Return

47 AS Summary Statistics by ndustry Type TABLE - AS SUMMARY STATSTCS BY NDUSTRY TYPE All ndustries Non-tradable ndustries Tradable ndustries Number of Workers (155.9) (138.2) (205.2) Value of Output (Millions Rupees) (445.2) (426.0) (507.4) Fixed Capital (Millions Rupees) (178.0) (171.3) (199.8) Daily Wage (Rupees) (120.3) (113.5) (139.4) Output per Worker (Millions Rupees) (3.158) (3.354) (2.301) Capital to Labor Ratio (44.69) (49.24) (20.51) Return

48 Share of Tradable Output Across Districts Percent Share of Tradable District Output Return

49 Production and Employment Elasticities by NREGA Phase TABLE - THE EFFECT OF NREGA BY PHASE Log(Value of Output) Log(Man-days) Non-tradable Tradable Non-tradable Tradable (1) (2) (3) (4) Log(Rainfall) x Phase *** *** (0.074) (0.174) (0.050) (0.116) Log(Rainfall) x Phase 1-2 x NREGA ** ** (0.075) (0.143) (0.046) (0.095) Log(Rainfall) x Phase (0.044) (0.082) (0.026) (0.064) Log(Rainfall) x Phase 3 x NREGA (0.051) (0.119) (0.032) (0.084) Observations 25,097 6,887 25,046 6,865 District FE Yes Yes Yes Yes Phase-Year FE Yes Yes Yes Yes Phase-ndustry FE Yes Yes Yes Yes ndustry-year FE Yes Yes Yes Yes Notes: Unit of observation is a district-industry-year. ndustry is defined at the NC 4-digit level. District-industryyear measures are obtained using sampling weights and are weighted in the regression by the corresponding weight. Rainfall refers to total rainfall in June-September. Value of output is in nominal terms. Number of workers and mandays include all types of firm employees. Wage refers to average real daily wage across firms -- CP for Agricultural Labourers is used. Daily wage at the firm level is obtained as total compensation during the year divided by total number of man-days. Standard errors are clustered at the district level. *** p<0.01 ** p<0.05 * p<0.1 AS Summary Stats by Phase Tradable by Phase Return

50 AS Summary Statistics by NREGA Phase TABLE - AS SUMMARY STATSTCS BY NREGA PHASE All Phase 1 Phase 2 Phase 3 Panel A: All ndustries Number of Workers (150.8) (155.4) (141.9) (151.8) Value of Output (Millions Rupees) (445.1) (370.0) (438.3) (468.6) Fixed Capital (Millions Rupees) (178.9) (178.0) (194.1) (174.3) Daily Wage (Rupees) (120.3) (91.82) (110.1) (129.7) Output per Worker (Millions Rupees) (3.162) (2.091) (2.714) (3.555) Capital to Labor Ratio (45.19) (22.00) (91.99) (23.74) Return

51 Share of Tradable Output across Phases Percent Share of Tradable District Output Phase 1 Phase 2 Phase 3 Return

52 Placebo Test Log(Wage) Log(Value of Output) Log(Man-Days) Non-Tradable Tradable Non-Tradable Tradable (1) (2) (3) (4) (5) NREGA (0.016) (0.061) (0.083) (0.038) (0.067) Observations 17,342 13,525 3,854 13,498 3,844 District FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Phase-ndustry FE Yes Yes Yes Yes Yes ndustry-year FE Yes Yes Yes Yes Yes nitial Conditions x Time Trend Yes Yes Yes Yes Yes Notes: Unit of observation is a district-industry-year. ndustry is defined at the NC 4-digit level. Districtindustry-year measures are obtained using sampling weights and are weighted in the regression by the corresponding weight. Value of output is in nominal terms. Number of workers and man-days include all types of firm employees. Wage refers to average real daily wage across firms -- CP for Agricultural Labourers is used. Daily wage at the firm level is obtained as total compensation during the year divided by total number of man-days. Analysis is restricted to a balanced panel of districts with at least 20 percent of population in agriculture. Standard errors are clustered at the district level. *** p<0.01 ** p<0.05 * p<0.1 Return