Impact of Farm Electricity Supply Management on Farm Households – Evidence from a Natural Experiment in India

20 May 2023

Namrata Chindarkara, Yvonne Jie Chena, and Shilpa Satheb

aAssistant Professor, Lee Kuan Yew School of Public Policy, National University of Singapore

469C Bukit Timah Road, Singapore 259772

bSenior Research Associate, Singapore Management University

Corresponding author: namrata.chindarkar@nus.edu.sg; Tel: +65-65168360

AbstractManaging agricultural and non-agricultural grid electricity supply through feeder segregation is being highlighted as an innovative reform to address India’s electricity supply issues. While this has a direct implication on the availability of grid electricity for farm operations, there is no rigorous empirical evidence available on the effects of feeder segregation on farm households. This paper exploits exogenous variation in quantity and quality of farm electricity supply resulting from Gujarat’s feeder segregation program, “Jyotigram Yojana” (JGY), and examines its impact on farm households. Using the 2004-05 and 2011-12 India Human Development Survey (IHDS) data and applying a difference-in-differences framework, we analyse the impact of rationed but high-quality farm electricity supply on investments in fixed and variable farm inputs and net farm income per acre. We find that, on average, JGY does not lead to an increase in ownership of electric pumps. However, it decreases ownership of diesel pumps. On average, farmers own more tubewells as exposure to JGY increases. In particular, it is the medium-to-large farmers who significantly increase ownership of tubewells. There is a corresponding statistically significant increase in the likelihood of irrigating with a tubewell for farmers across all land sizes. This is accompanied by an increase in per acre cost of purchased water for irrigation for farmers across land sizes. The consequent effect on net farm income per acre on average is a statistically significant decrease, which is more pronounced for medium-to-large farmers. This seems to be driven mainly by an increase in cost of purchased water for irrigation as we do not observe an increase in labor cost or increase in investments in other farm inputs including fertilizers, pesticides, and tractors. Using supplementary analysis we also rule out lower crop yields as the mechanism underlying decreased net farm income.

Keywords: farm electricity supply, farm income, impact evaluation, difference-in-differences, India

1

Authorship statement

NC conducted preliminary field research, performed the data analysis, and wrote the full paper.

YJC conducted preliminary field research and provided inputs on data analysis and paper

framing. SS collected the administrative data on JGY and performed data analysis.

2

Acknowledgments

We thank Ms. Xiao Yun and Mr. Venu Gopal Mothkoor for their excellent research

assistance. We are grateful for the financial support provided by the Lee Kuan Yew School

of Public Policy. The findings, interpretations, conclusions, and any errors are entirely

those of the authors.

3

1. Introduction

Agriculture continues to occupy a centre-stage in the Indian economy employing about 56% of

its labor force and contributing about 17.5% to the national GDP (Census of India, 2011; World

Bank, 2016). The sector is critical to achieving national goals such as reducing poverty,

providing food and nutritional security, supplying raw materials to major industries, and earning

foreign exchange (Kalamkar, et al., 2015). A key input necessary to sustain agricultural growth is

farm electricity, which powers irrigation pumps and other farm equipment. Approximately 83%

of the total irrigation energy demand is estimated to be met through grid electricity with diesel

accounting for about 17%. Solar electricity makes up less than 1% owing to high initial costs

(NITI Aayog Government of India, 2016). For this reason, grid electricity to agricultural

consumers continues to be heavily subsidized in India (World Bank, 2013). These subsidies

impose two significant costs – financial and environmental.

Subsidies for farm electricity have a long history in India starting from the Green Revolution in

the 1960s (Badiani, et al., 2012; Swain & Mehta, 2014). The high-yielding crop varieties

depended on access to irrigation. However, availability of surface water irrigation was low and

the focus therefore shifted to groundwater irrigation (World Bank, 2008). To enable farmers to

extract groundwater, electricity was provided either free-of-cost or at very low flat tariffs. This

was further justified by the need to promote rural livelihoods, food security, and overall rural

development (Swain & Mehta, 2014). Owing largely to clientelism, there has been no reform in

this subsidy policy over the years (Chindarkar, 2017). While the share of electricity consumed by

agricultural consumers is nearly 23%, revenue realization from agricultural consumers is only

about 9% (Power Finance Corporation, 2016). It is estimated that agricultural electricity

subsidies are equivalent to about 25% of India’s fiscal deficit, twice the annual public spending

on health or rural development, and two and a half times the annual spending on developing

surface water irrigation infrastructure (Monari, 2002). A consequence of this increased financial

pressure on public utility companies has been poor transmission and distribution (T&D) in the

form of interrupted and low-voltage electricity supply to agricultural as well as domestic

consumers (World Bank, 2013).

4

There is evidence supporting the positive correlation between subsidized electricity to farmers

and increased productivity in agriculture and allied sectors measured in terms of crop yields per

acre, farm income and wages, labor productivity, and water productivity (Badiani & Jessoe,

2014; Fan, et al., 2002; Kumar, 2005).1 However, this has come at the environmental cost of

increasing groundwater extraction (Badiani, et al., 2012). More than 60% of irrigation in India

depends on groundwater drawn using electric pumps (World Bank, 2012). Further, as surface

water irrigation is low, farmers depend heavily on groundwater irrigation as a buffer against poor

monsoons. It is also argued that as groundwater irrigation gives farmers greater control over

when to irrigate and how much water to use, productivity of farms irrigated with groundwater is

higher than those irrigated using surface water (World Bank, 2012). Farm electricity subsidies

effectively shift the preference of farmers towards increased use of groundwater irrigation as it

lowers the cost of extraction. This is further reinforced by poor monitoring and regulation of

tubewells and borewells mostly due to clientelism (Birner, et al., 2007). It is estimated that about

29% of districts in India are either semi-critical, critical, or over-exploited in terms of

groundwater extraction, which is largely attributable to farm electricity subsidies (Suhag, 2016).

Balancing the competing policy objectives of agricultural growth, financial viability of public

utility companies, and environmental sustainability is challenging yet critical (Chindarkar, 2017).

Several policy tools are available to policymakers to move towards achieving this balance.

Among these are – tariff changes, direct regulation, technical innovation, and a mix of these tools

(Badiani, et al., 2012; Mukherji, et al., 2010; Shah, et al., 2008; World Bank, 2013). Tariff

changes in accordance with demand and cost of supply is expected to increase revenue

generation for the utility companies and also likely to discourage over-extraction of groundwater

(Badiani, et al., 2012). Direct regulation in the form of metering as well as controlling the

number and depth of tubewells and borewells is also predicted to yield similar results. However,

it is likely to impose significant administrative and monitoring costs (Mukherji, et al., 2010;

Shah, et al., 2008). Technical innovation refers to changing the physical power supply

infrastructure and segregating the non-agricultural and agricultural feeders, or in other words,

separating the paid and non-paid (or nominally paid) loads (World Bank, 2013).

1 Fan, et al. (2002) measure labor productivity as real gross agricultural domestic product produced per unit of labor. Kumar (2005) measures water productivity as gross returns from all irrigated crops grown by the farmer divided by the total volume of water used for irrigating all crops.

5

Implementing tariff increases for agricultural users and direct regulation continue to be plagued

by political opposition, corruption, and non-compliance (Gronwall, 2014). These reforms

therefore have had limited success in striking the sensitive balance between demands of

competing sectors. In contrast, feeder segregation has gained considerable traction among both

policymakers and consumers (World Bank, 2013). While feeder segregation has been

implemented or is in-progress in several states in India and each has followed its own tailored

approach, the state of Gujarat in particular, is seen as an example of successful feeder

segregation (Shah, et al., 2008; World Bank, 2013). This is because Gujarat went beyond just

physical segregation and combined it with intelligent supply management. This included – (i)

enhancing the predictability of farm electricity supply, that is, the duration and timings when it

would be available, (ii) improving quality and reliability, that is, uninterrupted and high voltage

farm electricity supply, and (iii) matching the supply timings with peak periods of moisture

stress (Shah, et al., 2008). However, this has not meant that the reform has been completely

devoid of oppositions and criticisms. These include criticisms against rationing without taking

into consideration land sizes and crop types. Further, feeder segregation was accompanied with

direct regulation which entailed disconnecting unauthorized connections and imposing heavy

fines for non-compliance, and this turned out to be unpopular with political opponents

(Chindarkar, 2017).

Theory and evidence suggests that a change in the availability of a key farm input such as

electricity, which is required to power irrigation pumps and other farm equipment, is likely to

result in reallocation of factors of production and consequently affect welfare of farm households

(Barnes & Binswanger, 1986; Barnum & Squire, 1979). Thus, if the quantity and quality of farm

electricity supply changes then farmers are likely to change investments in other factors in order

to maximize their welfare. An empirical challenge however is to generate exogenous variation in

quantity and quality of farm electricity supply as it may be correlated with unobserved factors

such as individual preferences of farmers and external factors such as resource availability for

power generation, political environment, and administrative capacity.

6

This paper overcomes the empirical challenge by exploiting natural variation in quantity and

quality of farm electricity supply resulting from an exogenous policy intervention. In October

2003, the state government of Gujarat implemented its feeder segregation program known as

“Jyotigram Yojana” (JGY) which translates to “lighted village scheme”. JGY was launched with

an investment of US$290 million to separate agricultural and non-agricultural feeders against the

backdrop of rapidly rising electricity demand, depleting ground water tables, and heavy losses to

the state electricity board (Shah, et al., 2008). As previously mentioned, JGY not only physically

segregated the loads but also enhanced the predictability and quality of electricity supply to rural

agricultural and non-agricultural consumers. Under JGY, farms received 8 hours of high-quality

electricity while non-agricultural users received 24 hours of high-quality electricity supply.

There were no significant changes to the tariffs paid by agricultural consumers (Gujarat Urja

Vikas Nigam Limited, 2010).

Our outcome variables of interest are farm investments in fixed and variable inputs and net farm

income per acre. We use household-level survey data for Gujarat from Wave 1 (2004-2005) and

Wave 2 (2011-2012) of the India Human Development Survey (IHDS) and match it with

administrative data on JGY implementation. JGY ‘implementation’ refers to completion of

feeder segregation in a given village. We compute a ‘exposure’ to JGY treatment variable, which

is the cumulative proportion of villages in a district that implemented JGY in each month starting

in July 2003 till its completion in March 2008. Using a difference-in-differences framework we

find that, on average, JGY does not lead to an increase in ownership of electric pumps. However,

it decreases ownership of diesel pumps. On average, farmers own more tubewells as exposure to

JGY increases. In particular, it is the medium-to-large farmers who significantly increase

ownership of tubewells. There is a corresponding statistically significant increase in the

likelihood of irrigating with a tubewell for farmers across all land sizes. This is accompanied by

an increase in per acre cost of purchased water for irrigation for farmers across land sizes. The

consequent effect on net farm income per acre on average is a statistically significant decrease,

which is more pronounced for medium-to-large farmers. This seems to be driven mainly by an

increase in cost of purchased water for irrigation as we do not observe an increase in labor cost

or increase in investments in other farm inputs including fertilizers, pesticides, and tractors.

Using supplementary analysis we also rule out lower crop yields as the mechanism underlying

7

decreased net farm income. Further examination of heterogeneous effects suggests that for farm

households whose main income source is cultivation, there is a clear substitution away from

diesel pumps and towards electric pumps. We do not find significant differences in the outcomes

when households in the top/bottom 10% of the groundwater distribution are excluded.

2. Background on Gujarat and JGY

Much of Gujarat falls under arid, semi-arid, and dry sub-humid climatic zones. The state receives

average annual rainfall of about 1107 mm, however, the northern and north-western regions are

distinctly semi-arid or arid receiving on average between 500-700 mm rainfall annually

(Department of Agriculture & Cooperation, 2013). Due to low rainfall, the state relies heavily on

groundwater for irrigation. Groundwater development (GWD), which is the ratio of the annual

ground water extraction to the net annual ground water availability, was 30.81% in 1984 which

significantly worsened to 75.57% in 1997 (UNDP, 2004)2. In 2004, it remained at 76%,

however, 50% of the 223 sub-districts assessed were classified either as over-exploited

(GWD>100%), critical (90-100% GWD), or semi-critical (70-90% GWD) (CGWB, 2004).

Since 1988, the state has implemented a flat tariff system based on motor capacity of the electric

pumps for agricultural users. Clientelism made it infeasible to increase farm electricity tariffs.

The only alternative available to the severely financially stressed state electricity board was

therefore to reduce the quantity (number of hours) and quality (voltage) of farm as well as

domestic electricity supply. This negatively affected both rural development and agricultural

growth (Chindarkar, 2017). Against this backdrop, the state government of Gujarat launched

JGY, a rural feeder separation program, which separated power loads of rural non-agricultural

and agricultural consumers by connecting them to separate feeders. The program was first

implemented in 8 districts in early-2003 and gradually covered the entire state by early-2008. In

total, it covers more than 18,000 villages. Agriculture feeders provide 8 hours of high-quality

electricity supply to farms while rural non-agricultural users receive 24 hours of high-quality

supply.

2 Groundwater development rate indicates the quantity of groundwater available for use. A higher ratio (or percentage) implies that extraction as a percentage of availability is higher and therefore reflects an unsustainable situation.

8

JGY involved complete transformation of the electricity infrastructure landscape with installation

of new Specially Designed Transformers (SDTs), high tension lines, low tension lines, electricity

poles, electricity conductors, and PVC cables. Feeders were metered to improve the accuracy of

energy accounting. Apart from the infrastructural changes, concerted efforts were taken to

restructure the work culture within the distribution companies (DISCOMs). Figures (1a) and (1b)

illustrate the physical infrastructural changes under JGY.

<Figures 1a and 1b here>

3. JGY and farm households: Theory and evidence

Theory and evidence on the general impact of access to farm electricity has established that

electricity can affect farm households both through the intensive and extensive margins. Farm

electricity can increase yields (intensive margin) and consequently farm incomes mainly through

adoption of electric pumps for irrigation and other electric farm appliances (Barnes &

Binswanger, 1986; Khandker, et al., 2013). It can also increase the value that the farmers can

extract per drop of water as they can shift towards water-intensive, high-value crops (Mukherji,

et al., 2010). Further, it enables farmers to expand the land under cultivation (extensive margin)

as they are no longer constrained by the capacities of manual and animal labor (Bhargava, 2014).

Theoretical understanding of the impact of JGY on input choices and farm income in particular,

requires accounting for (un)certainty of farm electricity supply in the farm production function.

Suppose farm households are set to maximize their farm income by choosing the optimal level of

labor and machinery (such as pumps) subject to availability of land and other inputs (such as

credit). Prior to implementation of JGY, farm households maximized their expected income

given all factor prices and the probability of receiving electricity during production hours. The

probability essentially introduced uncertainty in the farm households’ decision-making regarding

choice of inputs. After the implementation of JGY however, farm households are not constrained

by uncertainty but are constrained by rationed electricity supply. Because we have limited

knowledge about the probability of receiving electricity prior to JGY implementation, it is

9

difficult to derive a comparison of the maximum farm income under these two scenarios

analytically.

We therefore consider a simple scenario of input choice where farmers have to make a decision

regarding the number of electric pumps required to irrigate their fields. Suppose, n is the number

of pumps owned by a farmer, w is the amount of water each pump is able to draw, and p is the

probability of being able to draw w amount of water. The total amount of water the farmer is

able to draw for irrigation is therefore the product n∗w∗p. Under uncertain farm electricity

supply, farmers may decide to own more units of electric pumps to maximize irrigation and

consequently farm income. Under certain but rationed supply, owning more electric pump units

may not be necessary. This is because while JGY rationed farm electricity to 8 hours per day it

also brought about improvements such better quality of supply, enhanced predictability as

farmers were provided a pre-determined schedule that matched peak periods of moisture stress,

and potentially lower rates of replacement of fixed inputs such as electric pumps as there were no

sudden voltage fluctuations, power outages, fuse blackouts, and motor burns. However, overall

shift in input choices and its consequent effect on farm income is still ambiguous as it is subject

to other factors such as land size, crop choice, availability of substitutes such as diesel pumps,

and so on.

Evidence on impact of farm electricity supply management and quality of supply on farm

households is very scarce and largely qualitative. A recent study looks at the impact of quality of

electricity, measured as hours of supply, on rural non-farm income in India and finds that it

increases non-agricultural income by 28.6% during 1994-2005 (Chakravorty, et al., 2014).

Qualitative evidence on JGY finds that the program induced farmers to shift towards high value

crops and efficient use of groundwater (S. G. Banerjee, et al., 2014; Gronwall, 2014; Mukherji,

et al., 2010; Shah, et al., 2008). Overall, the literature suggests that quantity-quality trade-off due

to JGY has not had negative consequences on farm households.

Drawing upon theory and previous evidence, we posit that impact of JGY on farm households’

input decisions and net income is an empirical question and depends on how factors of

production are simultaneously affected.

10

4. Data and empirical strategy

4.1. Data

This paper uses Waves 1 and 2 of the India Human Development Survey (IHDS) conducted by

the University of Maryland and the National Council of Applied Economic Research (NCAER).

IHDS is a nationally representative, multi-topical survey of 41,554 households spread across

1503 villages and 971 urban neighbourhoods in India. IHDS key characteristics of rich

contextual information, evidentially form the basis for the regression analysis. IHDS-1 is a

nationally representative survey of 41,554 households conducted in 2004-2005 while IHDS-2 re-

interviewed 83% of this original household sample in 2011-2012. Due to inherent demographic

shifts and attrition, IHDS-2 interviewed an additional 2134 households, which form the

replacement sample. The rural sample is drawn using stratified random sampling of villages and

the urban sample is drawn using stratified sampling of towns and cities within states (or groups

of states) selected by probability proportional to population (PPP).

We use only the rural sample and retain land-owning farm households in Gujarat in both Wave 1

and Wave 2, which yields an unbalanced panel of 995 households from 13 districts. We match

this with administrative data on village-level JGY implementation sourced from the four utilities

(distribution companies or DISCOMs) in Gujarat. JGY ‘implementation’ refers to completion of

feeder segregation in a given village. These data suggest that JGY was implemented gradually

between 2003 and 2008. The earliest implementation date is July 2003 and the latest is March

2008. As IHDS does not allow us to identify village names, we use precise month-year of JGY

implementation in each village to calculate the cumulative proportion of villages in a district that

implemented JGY in each month starting in July 2003 till its completion in March 2008. We

match these cumulative proportions with districts in the IHDS (see Figure 2). All IHDS Wave 1

interviews in our sample for Gujarat were fielded between February and September 2005.

Therefore, each district has some cumulative proportion of JGY implementation in Wave 1.

However, as observed in Figure 2, districts began implementing JGY in a staggered manner and

the cumulative proportions differ both within and across districts. We exploit this variation for

our difference-in-differences identification explained further in section 4.2.

11

To control for geographic characteristics, data on annual groundwater depth in meters for

districts in Gujarat for 2004-05 and 2011-12 are obtained from Water Resources Information

System of India (WRIS). District-level annual average rainfall data, measured in millimeters

(mm), are sourced from National Oceanic and Atmospheric Administration (NOAA). Additional

district-level controls on percentage of villages in the district having commercial banks,

cooperative banks, and credit societies, and land area irrigated by tubewells are sourced from the

District Census Handbooks of 2001 and 2011.

Data to examine changes in crop yields as an underlying mechanism linking JGY and net farm

income are sourced from ICRISAT Village Dynamics Studies in South Asia (VDSA) meso-level

dataset at the district-level from 1996-2011. These variables include total area under cultivation

in hectares for different crops, total production in tonnes of different crops, annual average

rainfall in millimeters, and total road length in kilometers indicating other infrastructure

development. Here again, we only use districts from Gujarat.

4.2. Empirical strategy

We examine the impact of JGY implementation on farm households using a generalized

difference-in-differences framework. The first difference comes from within district variation in

exposure over time. As we are able to compute exposure for each month-year for a district, there

is variation among households residing in the same district depending on the time they were

interviewed in Wave 1. The second difference comes from across district variation in exposure

over time. JGY was implemented in a staggered manner with some districts implementing it

earlier and some districts later. Therefore, there is variation in exposure for households

interviewed in the same month-year but residing in different districts in Wave 1.

Our treatment variable, which we hereafter refer to as ‘exposure’ to JGY, is the cumulative

proportion of villages in a district that implemented JGY in each month starting in July 2003 till

its completion in March 2008. Note that the variation in JGY exposure comes only from Wave 1

as all districts are fully exposed by Wave 2. The coefficient on our JGY exposure (treatment)

12

variable can therefore be interpreted as the mean change in outcome if exposure increases by 1

percentage point (from 0% to 100%).

Y idt=α0+Exposureidt+X '¿+W 'dt+φd+γ t +ε idt (1)

where, Y idt is the outcome for household i residing in district d at the year t . Exposureidt is the

treatment variable indicating exposure to JGY. X '¿ is a vector of time-varying household

characteristics. As implementation (timing as well as speed) of JGY might be determined by

district characteristics, we control for a vector of time-varying district characteristics and district-

level geographic characteristics represented by the vector W ' dt. φd are district dummies

controlling for district fixed effects, γt are wave dummies controlling for time fixed effects, and

ε idt is the random error. Standard errors are clustered at the district-level.

Our outcome variables of interest are farm investments in fixed and variable inputs and net farm

income per acre. Fixed inputs include ownership of tubewells, electric pumps, diesel pumps, and

tractors. Variable inputs include log of expenditures per acre on purchased water for irrigation,

fertilizers, pesticides, and hired labor in the last year. In addition, our farm investments analysis

includes a dummy variable indicating whether farmers irrigate with tubewells. Net farm income

per acre is computed as gross crop income minus all expenditure on hired labor, seeds, fertilizer,

pesticides, irrigation, hired equipment and animals, interest on loans, and other miscellaneous

farm expenses, and converted to log form. It is possible for net farm income to be negative if

investments or expenditures exceed revenue. To avoid computational issues, we add a constant

value to each observation before the log transformation so that we only have non-negative

values.

X '¿ controls for observed household characteristics including age of household head, gender of

household head, caste and religion of household head, household size and its quadratic form,

dummy indicating whether household is below the poverty line, log of total outstanding debt, and

log of land size. W ' dt includes time-varying district-level characteristics that might be correlated

with JGY implementation and the outcomes such as percentage of villages having commercial

banks, cooperative banks, and credit societies. It also includes district-level geographic

13

characteristics such as log of average annual groundwater depth, log of average annual rainfall,

and log of area irrigated using tubewells and other wells.

In addition, we include district dummies which control for district fixed effects or unobserved

characteristics that do not vary for a given district over time. These might include cultural

factors, political factors, and local administrative capacity that might be correlated with both

implementation of JGY and the outcomes. Further, we include wave fixed effects that capture

unobserved factors common to all districts in a given wave such as national- or state-level

changes in energy policy, energy prices, or climate happening simultaneously with JGY.

4.3. Parallel trends assumption

Our generalized difference-in-differences empirical strategy hinges on the parallel trends

assumption, which implies that in the absence of JGY, households living in different districts

within Gujarat would expect to follow similar trends of changes in farm outcomes over time. The

difference-in-differences estimator would be biased if JGY exposure varies simultaneously with

other factors at the district-level that could affect farm outcomes such as government subsidy

programs, employment programs, guaranteed prices for agricultural output and so on. Decisions

to implement such programs are made at the state-level and all districts are targeted. However,

there might still be observed and unobserved factors resulting in differences in the way these

other policies were implemented in each district which in turn might violate the parallel trends

assumption.

To provide further support for the parallel trends assumption, we compare pre-JGY trends of

districts with ‘high’ versus ‘low’ exposure as of 2005 or end of Wave 1 interviews. The mean

minimum exposure to JGY in 2005 is 52.28% while the minimum is 11.34% and maximum is

99.65%. We categorize districts with greater than 60% JGY exposure as ‘high exposure’

districts.3 This classifies the 13 districts in our sample into 7 ‘high exposure’ districts and 6 ‘low

exposure’ districts. We test the difference in mean outcomes for these two categories of districts

using information from additional data sources. Results are provided in Table 1. We find that

3 Categorization of districts into ‘high exposure’ and ‘low exposure’ does not change even if we use >50% JGY exposure as cut-off.

14

prior to JGY implementation, ‘high exposure’ and ‘low exposure’ districts are similar to each

other on average across a range of variables that might result in the districts having unparallel

trends of changes in farm outcomes. Importantly, difference in percentage of villages with power

supply is significant only at the 10% level and no difference in percentage of villages with

domestic and agricultural electricity supply prior to JGY implementation.

5. Descriptive statistics

We present descriptive statistics of all variables included in our regression models for Gujarat in

Waves 1 and 2. We observe that the average number of electric pumps, diesel pumps, tubewells,

and tractors owned has increased in Wave 2. Proportion of farmers irrigating with tubewells

increased from 27% in Wave 1 to 32% in Wave 2. Cost of all variable inputs increased between

Waves 1 and 2. Cost of irrigation per acre using purchased water increased by approximately

19% while cost of hired labor per acre increased by approximately 28%. Cost of fertilizers and

pesticides per acre increased by about 34% and 20% respectively. Net farm income per acre

increased by approximately 20% while total outstanding household debt decreased by about 66%

between Waves 1 and 2. Depth to groundwater (in meters) worsened by approximately 6%

between the two waves.

<Table 2 here>

6. Results

6.1. Main results

We estimate the average treatment effect of JGY on farm investments in fixed and variable

inputs using equation (1). We present the results on farm investments in Table 3 for all farm

households and separately for farm households with small and medium-to-large landholdings

separately. The definitions of small, medium, and large farmers are as per the guidelines of

Department of Land Resources, Government of India, where large farmer = landholding >5

hectares, medium farmer = landholding >2 hectares and <=5 hectares, and small farmer =

landholding <=2 hectares as per Wave 1 landholding size.

15

< Table 3 here>

Our findings suggest that, on average, farmers do not increase their ownership of electric pumps

with increased exposure to JGY. However, they decrease their ownership of diesel pumps.

Possible explanation lies in the risk diversification theoretical conjecture, where with rationed

but more certain farm electricity supply, farmers do not need to own as many diesel pump units

as before to substitute electric pumps. On average, farmers own more tubewells as exposure to

JGY increases. In particular, it is the medium-to-large farmers who significantly increase

ownership of tubewells. There is a corresponding statistically significant increase in the odds of

irrigating with a tubewell for farmers across all land sizes. Thinking in terms of the certainty and

rationing of farm electricity introduced by JGY, farmers may not need to spread their risk by

owning more electric pump units. In fact, as demonstrated by the results, with higher probability

of being able to use an electric pump to its optimal extraction capacity, the likelihood of

connecting them to more tubewells to optimize groundwater extraction for irrigation seems to

increase.

We find an average increase in cost of irrigation per acre using purchased water and also for

farmers of varying landholding sizes. For all farmers, cost of purchasing water for irrigation

increases by 0.8% and for small and medium-to-large farmers it increases by 0.6% and 1.2%

respectively. The increase in cost of purchased water for irrigation likely results from rationed

farm electricity. Evidence from previous studies finds that while JGY did not result in a decrease

in groundwater extraction on average possibly because farmers tried to maximize extraction

during the eight hours, it did place constraints on informal groundwater markets (Chindarkar &

Grafton, 2017; Shah, et al., 2008). With rationing, the opportunity to extract excess groundwater

became limited resulting in higher prices of purchased water. The increase in cost of purchased

water may have been further compounded by shifts in cropping patterns towards higher value

and water-intensive crops. During the period 2003-04 to 2009-10, total land under cultivation of

cereal crops and pulses in Gujarat fell by 1.83% and 3.55% while area under cultivation of cotton

increased by 32.81%. Even though there was a net decrease in total land under cereal crops,

among the cereal crops grown, there was an increase in area under wheat cultivation by 76.35%

during this period (Department of Agriculture & Cooperation, 2013). Cotton and wheat entail

16

high irrigation costs of INR 1675.96 and INR 2856.29 per hectare respectively compared to other

cereals and cash crops that are widely grown in Gujarat such as yellow lentils (INR

262.67/hectare), millet (INR 745.62/hectare) and groundnut (INR 434.51/hectare) (Department

of Agriculture & Cooperation, 2013). In sum, farmers require more water for irrigating higher

value and water-intensive crops but are limited to eight hours owing to JGY and therefore have

to source any additional requirement from inflated groundwater markets resulting in increased

irrigation costs. Increased irrigation costs may also explain why ownership of tractors decreases

on average and in particular for small farmers. Increased irrigation costs owing to purchased

water may constrain the benefits that can be gained at the extensive margin especially for small

farmers.

Contrary to theoretical expectation, exposure to JGY does not have any significant negative

effect on labor cost per acre. This could be because labor needs to be allocated to operate and

maintain the additional tubewells being used for irrigation. Thinking again in terms of certainty

and rationing, a full-time tubewell operator may not be needed to keep a watch on electricity

supply all day. However, the presence of a tubewell operator may still be needed to power on the

pump, prevent spills, and also do routine maintenance work. No significant change in fertilizer

cost per acre is observed and only a marginally significant increase in pesticide cost per acre is

observed for small farmers.

From the overall picture of shift in farm investments we see that farmers increase tubewell

ownership and irrigate more using tubewells. On average, farmers reduce their ownership of

diesel pumps plausibly because they do not need to invest as much in substitutes with improved

quality and reliability of farm electricity. At the same time however, they incur higher irrigation

costs due to purchased water. Therefore, it is difficult to have an a priori hypothesis about the

direction of the effect on net farm income. We examine the effect of JGY implementation on net

farm income per acre using equation (1). We find that exposure to certain but rationed farm

electricity supply under JGY significantly decreases net farm income per acre for all farmers in

Gujarat by 0.2% on average. This seems to be driven by medium-to-large farmers for whom net

farm income per acre decreases by 0.1%.

17

It is possible that the decrease in net farm income results from a change in crop yields due to

rationed farm electricity. We examine this potential mechanism linking JGY exposure and

decrease in net farm income per acre in the next sub-section.

6.2. Impact on crop yields

In order to explain the increase in net farm income post-JGY for all farmers, we analyse the

effect of JGY on crop yields using a similar difference-in-differences strategy. As IHDS Wave 2

does not contain data on crop choices, we extract data on districts in Gujarat from the ICRISAT

VDSA from 1996-2011. We estimate the following generalized difference-in-differences

regression:

log (Y dtc)=α 0+β PostJGY dt+ X ' dt+φd+γ t+εdtc(2)

where, log(Y dtc) is the logarithm of crop yield for district d in year t for crop type c. We group

crops into five crop types: cereals, pulses, oilseeds, sugarcane, and cotton. For each crop type,

we retain the longest data series available. All yield is computed in the unit of tonnes per hectare. PostJGY dt equals 1 for the year in which the district implemented JGY (100% of the villages in

the district were covered under JGY) and for each subsequent year thereafter, and 0 otherwise.

The effect of JGY exposure on crop yields is thus identified from the difference over time in

yields between districts that implemented JGY earlier versus those that implemented it later.

Following A. Banerjee, et al. (2002), we also include logarithm of gross cultivated area,

logarithm of gross irrigated area, logarithm of length of road, and logarithm of annual average

rainfall for each district in X ' dt. The vector also includes logarithm of average annual

groundwater level, which we match using WRIS data. Further, we control for district (φd ¿and

year (γt ¿ fixed effects for all regressions and cluster standard errors at the district level.

Results for the yield analyses are presented in Table 4. Coefficient of PostJGY dtestimates the

effect of the JGY implementation on growth of crop yield. Results suggest that exposure to JGY

did not result in a significant increase in crop yields for any crop type. Therefore, the decrease in

net farm income is unlikely to be a result of decreased productivity due to rationed farm

18

electricity supply.

<Table 4 here>

6.3. Additional heterogeneous effects

We examine further heterogeneities based on whether a farm household derives its income

mainly from cultivation and level of groundwater access. It is reasonable to assume that

groundwater extraction for irrigation holds greater significance for cultivators and therefore their

input choices may differ. It is also likely that changes in farm investments following JGY

implementation are highly dependent on access to groundwater. We therefore also present results

excluding households that have on average better access to groundwater (top 10% of distribution

of depth to groundwater) and those who are in groundwater-stressed regions (bottom 10% of

distribution of depth to groundwater). The objective is to reject the possibility that our main

results are driven by households in the extremes according to groundwater distribution. Results

are presented in Table 5.

<Table 5 here>

The effect on cultivators is largely comparable to the average effect of JGY on all farmers. A key

distinction we observe is a clear substitution effect between electric and diesel pumps for

cultivators possibly because groundwater extraction is more salient for these households. This is

accompanied by increased likelihood of irrigating using tubewells, increased cost of purchased

water for irrigation per acre, and a decrease in net farm income per acre. Further, we find that the

effect on our outcomes excluding farm households with extreme groundwater distribution is by

and large similar. Differences in fertilizer and pesticides costs per acre are significant only at the

10% level.

7. Conclusion

This study fills a critical gap in the existing literature on farm electricity supply management on

farm households. In particular, we evaluate the impact of JGY on farm households in the Indian

19

state of Gujarat. The state government launched JGY to balance the competing policy objectives

of agricultural growth, financial viability of public utility companies, and environmental

sustainability (particularly groundwater extraction). Using survey data from IHDS Waves 1 and

2 and administrative data from electricity utilities in Gujarat and applying a difference-in-

differences econometric framework, we examine the impact of JGY on farm investments in fixed

and variable inputs and net farm income.

JGY did away with uncertainty of farm electricity supply and provided high voltage electricity

required to run farm equipment. However, it capped the number of hours of electricity available

to farmers. In response, on average, farmers increased their ownership of tubewells and also the

likelihood of irrigating with a tubewell. Medium-to-large farmers increased their ownership of

tubewells and decreased ownership of diesel pumps presumably because they did not need to

own as many diesel pumps as before to substitute uncertain electricity supply. Cost of irrigation

using purchased water increased for all farmers, which may be a consequence of inflated

groundwater markets due to rationing and also due to shifts in cropping pattern. We also find an

average decrease in ownership of tractors, driven mainly by small farmers.

Further, we find that an increase in JGY exposure leads to a decrease in net farm income per acre

on average, which stems mainly from decrease in net farm income for medium-to-large farmers.

The decrease seems to result from the increased cost of irrigation using purchased water as we do

not observe significant increase in other inputs such as hired labor, fertilizers, pesticides, and

tractors. Supplementary analysis on crop yields suggests that there is no significant change in

productivity for any crop type. Therefore, the decrease in net farm income cannot be attributed to

decrease in productivity due to rationed farm electricity supply.

Additional heterogeneity analysis suggests that for farm households whose main income source

is cultivation, there is clear substitution away from diesel pumps and towards electric pumps.

The effect of JGY on farm households which are in the middle 80% distribution based on

groundwater level, is not significantly different from the average effect on all farm households.

20

To our knowledge, this paper provides the first rigorous evidence of JGY on farm households,

which is crucial to examining the effectiveness of innovative yet politically difficult reforms. Our

findings also have significant policy implications. With farmers shifting towards high value and

water-intensive crops, rationed farm electricity may lead to reduction in welfare of farm

households as they have to spend more on purchased water for irrigation. Previous empirical

evidence also does not suggest that groundwater extraction on average has declined. If anything,

farmers seem to want to maximize extraction during the eight hours. Concomitant costs can

therefore negate the effect of improved quality and reliability of farm electricity not to mention

potential long-term negative effects on environmental sustainability. Thus, alternative policy

measures to balance the objectives of agricultural growth, energy security, and environmental

sustainability such as metering, conservation, and technological innovation in irrigation are still

necessary.

21

8. ReferencesBadiani, R., & Jessoe, K. (2014). The Impact of Electricity Subsidies on Groundwater Extraction and Agricultural

Production. Department of Agricultural Economics, University of California, Davis.Badiani, R., Jessoe, K., & Plant, S. (2012). Development and the Environment: The Implications of Agricultural

Electricity Subsidies in India. Journal of Environment & Development, 21(2) 244–262.Banerjee, A., Gertler, P., & Ghatak, M. (2002). Empowerment and Efficiency: Tenancy Reform in West Bengal.

Journal of Political Economy, 110, 239-280.Banerjee, S. G., Khanna, A., Khurana, M., Mukherjee, M., & Saraswat, K. (2014). Lighting rural India : load

segregation experience in selected states. In Asia Sustainable and Alternative Energy (ASTAE) Program: South Asia energy studies,World Bank Group. .

Barnes, D. F., & Binswanger, H. P. (1986). Impact of Rural Electrification and Infrastructure on Agricultural Changes, 1966-1980. Economic and Political Weekly, Volume 21, 26-34.

Barnum, H. N., & Squire, L. (1979). An econometric application of the theory of the farm-household. Journal of Development Economics, Volume 6, 79-102.

Bhargava, A. K. (2014). The Impact of India's Rural Employment Guarantee on Demand for Agricultural Technology. In IFPRI Discussion Paper 01381: International Food Policy Research Institute (IFPRI).

Birner, R., Gupta, S., Sharma, N., & Palaniswamy, N. (2007). The Political Economy of Agricultural Policy Reform in India: The Case of Fertilizer Supply and Electricity Supply for Groundwater Irrigation. In. New Delhi: International Food Policy Research Institute (IFPRI).

Census of India. (2011). Census 2011. In: Office of the Registrar General & Census Commissioner,Ministry of Home Affairs, Government of India.

CGWB. (2004). Dynamic Ground Water Resources of India. In: Central Ground Water Board, Ministry of Water Resources, Government of India.

Chakravorty, U., Pelli, M., & Marchand, B. U. (2014). Does the quality of electricity matter? Evidence from rural India. Journal of Economic Behavior & Organization, Volume 107, 228-247.

Chindarkar, N. (2017). Beyond Power Politics: Evaluating the Policy Design Process of Rural Electrification in Gujarat, India. Public Administration and Development, Volume 37, 28-39.

Chindarkar, N., & Grafton, Q. (2017). India’s Groundwater Crisis and the Energy Nexus: It’s Worse Than We Thought. In.

Department of Agriculture & Cooperation. (2013): Government of Gujarat.Fan, S., Zhang, L., & Zhang, X. (2002). Growth, Inequality, and Poverty in Rural China:The Role of Public

Investments. In. Washington, D.C.: International Food Policy Research Institute.Gronwall, J. (2014). Power to segregate: Improving electricity access and reduced demand in rural India. In:

Stockholm International Water Institute (SIWI).Gujarat Urja Vikas Nigam Limited. (2010). Tariff for Supply of Electricity at Low Tension, High Tension, and Extra

High Tension. In G. E. R. Commission (Ed.).Kalamkar, S. S., Swain, M., & Bhaiya, S. R. (2015). Impact Evaluation of Rashtriya Krishi Vikas Yojana (RKVY) in

Gujarat. Agro-Economic Research Centre, Sardar Patel University and Agricultural Development & Rural Transformation Centre, Institute for Social and Economic Change (ISEC), Bangalore.

Khandker, S. R., Barnes, D. F., & Samad, H. A. (2013). Welfare Impacts of Rural Electrification: A Panel Data Analysis from Vietnam. Economic Development and Cultural Change, Volume 61, 659-692.

Kumar, D. M. (2005). Impact of electricity prices and volumetric water allocation on energy and groundwater demand management:: analysis from Western India. Energy Policy, Volume 33, 39-51.

Monari, L. (2002). Power subsidies - a reality check on subsidizing power for irrigation in India. Public policy for the private sector. In. Washington, D.C: World Bank.

Mukherji, A., Shah, T., & Verma, S. (2010). Electricity reforms and their impact on ground water use in states of Gujarat, West Bengal and Uttarakhand, India. In J. Lundqvist (Ed.), On the Water Front: Selections from the 2009 World Water Week in Stockholm. Stockholm: Stockholm International Water Institute.

NITI Aayog Government of India. (2016). User Guide for 2047 Energy Calculator. In. New Delhi: NITI Aayog.Power Finance Corporation. (2016). The Performance of State Power Utilities for the years 2012-13 to 2014-15. In:

Power Finance Corporation Ltd, Government of India.

22

Shah, T., Bhatt, S., Shah, R. K., & Talati, J. (2008). Groundwater governance through electricity supply management: Assessing an innovative intervention in Gujarat, western India. Agricultural water management, Volume 95, 1233-1242.

Suhag, R. (2016). Overview of Ground Water in India. In: PRS Legislative Research, Government of India.Swain, A. K., & Mehta, U. S. (2014). Balancing State, Utility and Social Needs in Agricultural Electricity SupplyThe case for a holistic approach to reform. The International Institute for Sustainable Development.UNDP. (2004). Gujarat Human Development Report. In: Mahatma Gandhi Labour Institute, Ahmedabad.World Bank. (2008). World Development Report 2008 : Agriculture for Development. In. Washington, DC: World

Bank.World Bank. (2012). India Groundwater: a Valuable but Diminishing Resource. In. Washington, DC: World Bank.World Bank. (2013). Lighting Rural India : Experience of Rural Load Segregation Schemes in States. In. Washington,

DC: World Bank.World Bank. (2016). World Development Indicators 2016. In. Washington, DC: World Bank.

23

9. Tables and figures

Table 1. Pre-JGY comparison of ‘high exposure’ and ‘low exposure’ districts

High Exposure

Low Exposure

High - Low p-value

Total population 2160974 1983922 177052 0.824Rural population 1137224 1316401 -179177 0.449Literacy 69.654 68.542 1.112 0.788Working population 851888 832010 19878 0.942Number of inhabited villages 547.571 831.666 -284.095 0.110% of villages with access to safe drinking water 0.997 0.995 0.002 0.354% of villages with power supply+ 0.996 0.980 0.016 0.074*% of villages with access to domestic electricity^ 0.042 0.061 -0.019 0.502% of villages with access to agricultural electricity^^ 0.020 0.033 -0.013 0.510% of villages with access to primary school 0.988 0.960 0.028 0.055*% of villages with access to medical facility 0.680 0.666 0.014 0.903% of villages with access to paved road 0.904 0.790 0.114 0.077*% of villages with cooperative bank 0.063 0.036 0.027 0.192% of villages with agricultural credit society 0.594 0.467 0.127 0.111Area irrigated by tubewells (hectares) 113537.7 92568.81 20968.9 0.532

Source: Census of India 2001Notes: District is defined as ‘high exposure’ if >60% of villages in the district have implemented JGY in Wave 1.Number of ‘high exposure’ districts = 7; Number of ‘low exposure’ districts = 6+ denotes power supply is available in a village regardless of whether there are connections to households or farms^ denotes household power connections are provided in a village^^ denotes agricultural power connections are provided in a villagep<0.01***, p<0.05**, p<0.10*

24

Table 2. Descriptive statistics

Wave 1 Wave 2N Mean S.D. Min Max N Mean S.D. Min Max

A. Electrification variableExposure to JGY (cumulative)

553 52.275 31.256

11.343 99.656 441 100.000 0.000 100 100

B. Outcome variablesNumber of tubewells 553 0.105 0.307 0 1 441 0.195 0.424 0 2Number of electric pumps 553 0.065 0.247 0 1 441 0.143 0.375 0 2Number of diesel pump 553 0.083 0.295 0 3 441 0.136 0.350 0 2Number of tractors 553 0.063 0.251 0 2 441 0.109 0.326 0 2Irrigates using tubewells 553 0.269 0.444 0 1 441 0.324 0.469 0 1Log irrigation cost of purchased water per acre

553 0.198 0.364 0 1.833 441 0.385 0.586 0 3.584

Log fertilizer cost per acre 553 0.516 0.450 0 2.918 441 0.851 0.650 0 3.408Log hired labor cost per acre

553 0.366 0.531 0 3.761 441 0.649 0.738 0 4.816

Log pesticide cost per acre 503 0.291 0.353 0 2.150 439 0.482 0.434 0 2.442Log net farm income per acre

553 10.800 0.510 0 12.905 441 10.998 0.464 7.268 13.788

C. Household controlsHousehold size 553 5.685 2.702 1 18 441 5.673 2.605 1 19Below poverty line household

553 0.083 0.276 0 1 441 0.104 0.306 0 1

Years of education of household head

553 4.562 4.192 0 15 441 4.993 4.311 0 16

Male household head 553 0.931 0.253 0 1 441 0.914 0.281 0 1Age of household head 553 48.309 12.99

419 90 441 52.254 12.199 21 85

Log total outstanding debt 553 3.493 4.782 0 13.305 441 2.834 4.783 0 15.202Log land owned 553 1.526 0.796 0.108 4.511 441 1.584 0.761 0.118 3.959CasteBrahmin 553 0.018 0.133 0 1 441 0.034 0.181 0 1OBC 553 0.398 0.490 0 1 441 0.444 0.497 0 1Other 553 0.335 0.472 0 1 441 0.288 0.453 0 1SC 553 0.083 0.276 0 1 441 0.050 0.218 0 1ST 553 0.166 0.373 0 1 441 0.184 0.388 0 1

ReligionHindu 553 0.940 0.237 0 1 441 0.948 0.223 0 1Muslim 553 0.047 0.212 0 1 441 0.036 0.187 0 1Christian 553 0.013 0.112 0 1 441 0.000 0.000 0 0Jain 553 0.000 0.000 0 0 441 0.007 0.082 0 1Others 553 0.000 0.000 0 0 441 0.009 0.095 0 1

Agricultural income typeCultivation 553 0.758 0.429 0 1 441 0.746 0.436 0 1Agricultural labor 553 0.130 0.337 0 1 441 0.093 0.291 0 1Other 553 0.112 0.316 0 1 441 0.161 0.368 0 1

D. Geographic controlsLog rainfall 553 2.803 2.745 0 6.321 441 1.392 1.850 0 4.270Log groundwater level 553 2.541 0.615 1.708 4.165 441 2.597 0.776 1.856 4.461E. District controls% of villages having a commercial bank

553 9.612 5.054 3.260 22.000 441 9.759 5.549 2.330 22.480

% of villages having a cooperative bank

553 4.826 3.775 0.720 13.750 441 4.540 3.422 0.540 12.900

% of villages having an agricultural credit society

553 52.494 9.748 27.430 74.270 441 37.187 17.052 11.470

70.160

Log area irrigated by tubewells (hectares)

553 11.226 0.961 9.146 12.352 441 11.499 0.875 9.794 12.391

Note: All monetary values have been adjusted to 2012 rupees.

25

26

Table 3. Difference-in-differences estimates of impact of JGY on farm inputs and net farm incomeOutcome variables All farmers Small farmers Medium-to-large farmers

Number of electric pumps+ 0.016 0.070 -0.037

(0.014) (0.077) (0.036)

Number of diesel pumps+ -0.036*** -0.024 -0.017

(0.007) (0.022) (0.011)

Number of tubewells+ 0.035** 0.029 0.047***

(0.014) (0.020) (0.012)

Number of tractors+ -0.020** -0.207*** -0.003

(0.009) (0.036) (0.008)

Irrigate using tubewells# 0.027*** 0.024** 0.042***

(0.008) (0.010) (0.012)

Log irrigation cost of purchased water per acre^ 0.008** 0.005** 0.012**

(0.004) (0.002) (0.006)

Log hired labor cost per acre^ -0.002 -0.001 -0.000

(0.004) (0.006) (0.003)

Log fertilizer cost per acre^ 0.003 0.002 0.003

(0.003) (0.003) (0.003)

Log pesticides cost per acre^ 0.002 0.004* -0.001

(0.001) (0.002) (0.002)

Log net farm income per acre^^ -0.002*** -0.003 -0.001***

(0.001) (0.002) (0.000)

District FE Y Y Y

Wave FE Y Y Y

Household controls Y Y Y

Geographic controls Y Y Y

District controls Y Y Y

Observations 994 649 345

27

Notes: Difference-in-difference estimates with full set of controls.+ negative binomial, # logit, ^ tobit, and ^^ OLS regressions.Robust standard errors clustered at district-level in parentheses. p<0.01***, p<0.05**, p<0.10*.

1. Household controls include: Household size; household size squared; below poverty line household; years of education of household head; male household head; age of household head; caste of household head; religion of household head; log total outstanding debt; log land owned; type of agricultural income

2. Geographic controls include: log rainfall, log groundwater level3. District controls include: % of villages having a commercial bank; % of villages having a cooperative bank; % of villages having an agricultural credit society; log area

irrigated by tubewells

28

Table 4. Additional heterogenous effects of JGY on farm inputs and net farm income

Cultivators only Excluding top/bottom 10% of groundwater distribution

Outcome variables (1) (2)

Number of electric pumps+ 0.027** 0.015

(0.012) (0.013)

Number of diesel pumps+ -0.031*** -0.035***

(0.008) (0.007)

Number of tubewells+ 0.035** 0.040

(0.015) (0.000)

Number of tractors+ -0.014 -0.019

(0.011) (0.000)

Irrigate using tubewells# 0.039*** 0.033**

(0.010) (0.015)

Log irrigation cost of purchased water per acre^ 0.012*** 0.009**

(0.004) (0.004)

Log hired labor cost per acre^ -0.003 0.000

(0.004) (0.003)

Log fertilizer cost per acre^ 0.002 0.004*

(0.003) (0.002)

Log pesticides cost per acre^ 0.001 0.003*

(0.002) (0.002)

Log net farm income per acre^^ -0.002*** -0.002***

(0.000) (0.001)

District FE Y Y

Wave FE Y Y

Household controls Y Y

Geographic controls Y Y

District controls Y Y

Observations 748 895

29

Notes: Difference-in-differences estimates with full set of control variables for all farm households. + negative binomial, # logit, ^ tobit, and ^^ OLS regressions.Robust standard errors clustered at district-level in parentheses. p<0.01***, p<0.05**, p<0.10*.

1. Household controls include: Household size; household size squared; below poverty line household; years of education of household head; male household head; age of household head; caste of household head; religion of household head; log total outstanding debt; log land owned; type of agricultural income

2. Geographic controls include: log rainfall, log groundwater level3. District controls include: % of villages having a commercial bank; % of villages having a cooperative bank; % of villages having an agricultural credit society; log area

irrigated by tubewells

30

Table 5. Impact of JGY on crop yields  (1) (2) (3) (4) (5)Log yield (‘000 tonnes yield per ‘000 hectares) Log(Cereal) Log(Pulses) Log(Oilseeds) Log(Sugar) Log(Cotton) Post-JGY -0.058 0.050 -0.158 -0.159 -0.092

(0.064) (0.086) (0.143) (0.301) (0.091)Log of annual rainfall (millimetres) 0.208*** 0.217** 0.193* 0.013 0.120

(0.071) (0.090) (0.112) (0.056) (0.146)Log of total cropped area (‘000 hectares) 0.206 0.257 -0.007 -0.165 0.492

(0.243) (0.179) (0.326) (0.260) (0.407)Log of length of roads (kilometres) -0.644** -0.116 -0.178 1.184** -1.057**

(0.294) (0.378) (0.433) (0.502) (0.448)Log of groundwater depth -0.258 -0.239*** -0.335 -0.343* -0.116

(0.161) (0.081) (0.200) (0.170) (0.217)

District FE Y Y Y Y YYear FE Y Y Y Y YSample Years 1996-2011 1996-2011 1996-2011 1996-2011 1996-2011Observations 266 266 266 192 236R-squared 0.778 0.768 0.531 0.170 0.771

Notes: Difference-in-difference estimates using OLS regressions.Robust standard errors clustered at district-level in parentheses. p<0.01***, p<0.05**, p<0.10*.

31

Figures 1a and 1b. Pre- and post-JGY feeder system

11 KV Feeder - Farm and Non-farm

Village-1

DA C

Village-2

DA C

A--> Agricultural ConsumersD--> Domestic ConsumersC--> Commercial Consumers

11 KV Farm Feeder

Village-1

Village-2

A

A

A--> Agricultural ConsumersD--> Domestic ConsumersC--> Commercial Consumers

11 KV Non-Farm Feeder

D C

D C

Source: Shah et al. (2008)

32

1a: Feeder system before JGY 1b: Feeder system after JGY

Figure 2. Rollout of JGY in districts in the sample

0.25

.5.75

1

0.25

.5.75

1

0.25

.5.75

1

0.25

.5.75

1

2004 2006 2008 2004 2006 2008 2004 2006 2008

2004 2006 2008

Ahmedabad Anand Bharuch Gandhinagar

Jamnagar Junagadh Kachchh Kheda

Mahesana Narmada Patan Surendranagar

Vadodara

Pro

porti

on o

f JG

Y vi

llage

s

Date of ImplementationGraphs by District

Source: Authors’ calculations using administrative data on JGY program rollout collected from DISCOMs.

33