Can Energy Pricing Save India’s Groundwater?
Evidence from a Field Pilot in Gujarat
Ram Fishman
1 George
1
Upmanu Lall
2
Vijay Modi
2
Washington University 2 Columbia Water Center, Columbia University
September 2, 2014
IWREC 2014
Thanks to: Principal Secretaries of Energy for the Government of Gujarat MR. D. J. Pandian (current) and Mr. S
Jagdishan (former); Directors of the North Gujarat Electricity Utility (UGVCL), Mr. N. Srivastava (Current) and Mr. A.
K. Verma (former); Columbia Water Center’s Kapil Narula, Shama Praveen, Nikunj Parekh, Dishant Patel, Ankur Patel,
Anil Kumar, Sandeep Mahajan and Sahil Gulati.
Funding: International Growth Center, LSE; Harvard Sustainability Science program, Columbia Water Center, GWU UFF
1 / 23
Overview
1
Introduction
2 / 23
The Groundwater-Energy Crisis in India
Groundwater depletion - a global
problem (Konikow and Kendy
2005, Wada et al 2010)
India - world’s largest consumer
and the country probably most
vulnerable to this threat (Fishman
et al, 2011; Sekhri, 2011; Shah,
2010; World Bank, 1998; World
Bank, 2010).
Introduction
3 / 23
The Groundwater-Energy Crisis in India
Groundwater depletion - a global
problem (Konikow and Kendy
2005, Wada et al 2010)
India - world’s largest consumer
and the country probably most
vulnerable to this threat (Fishman
et al, 2011; Sekhri, 2011; Shah,
2010; World Bank, 1998; World
Bank, 2010).
Introduction
4 / 23
The Groundwater-Energy Crisis in India
0
10000000 15000000
5000000
Pumpsets
Agricultural Electricity Consumption, India
0
20000 40000 60000 80000 100000
30
25
20
Ag Consumption (GWh)
15
4
3.8
3.6
Share of Total (Percent)
Pumpset Power (KW)
4.2
4000000
3.8
3.6
4.2
Pumpset
30
25
20
15
Share
2
100000
80000
60000
40000
20000
Ag
5
0
Pumpsets
15000000
10000000
5000000
1980
1990
2000
2010
Number
Ag
Share
Pumpset
Agricultural
.2Consumption
0000
Consumption
ofof
Total
Power
of
Total
Power
Pumpsets
(Percent)
(KW)
Electricity
(GWh)
(KW)
(GWh) Consumption, India
1980
1990
Number of Pumpsets
Share of Total
2000
2010
Ag Consumption (GWh)
Pumpset Power (KW)
Number of electrical pumpsets (red), Agricultural electricity consumption (GWh, black), its share of total electricity
consumption (percent, grey) and the average pumpset power (KW, grey dashed) from 1978 to 2008. Source: All India
Electricity Statistics, annual volumes.
Introduction
5 / 23
Low Water Use Efficiency
Maintaining food production
while reducing water/energy use
is a key challenge.
Despite increasing scarcity,
adoption of water saving
technologies remains low
(MOWR 2001)
Introduction
6 / 23
Distorted Incentives
Groundwater use/access is not directly regulated in India.
Since the 1980s, electricity for pumping groundwater is mostly
un-metered and billed at flat, subsidized rates, if at all (Shah 2007).
Flat rates widely blamed for groundwater depletion and inefficient
use, deterioration of power utilities finances and infrastructure
(Dubash 2007, Planning Commission 2011).
Electricity Act of 2003: Mandatory Metering
Attempts to restore pricing by financially stressed state govts. have
all but failed (Dubash and Rajan 2001).
Introduction
7 / 23
Research Questions
Can metering and/or accurate pricing be re-introduced to agricultural
electricity use?
Will a marginal price change water usage patterns (reverse GW
depletion)?
Introduction
8 / 23
A New Approach
Provide consumers with Electricity Entitlements reflecting current usage, and
compensate them per unit voluntary reductions in electricity use (Morris 2006,
Gulati 2011).
Farmers must agree to have meters installed and read regularly, but are
otherwise under no commitment to reduce power use.
A politically feasible way to re-introduce a marginal cost structure and an
incentive to reduce water/energy use.
A theoretical win-win: revenue neutral for state, farmers make no
commitment...
The idea was piloted In April 2011 through a partnership between the Govt. of
Gujarat (GoG) the North Gujarat Utility (UGVCL) and Columbia Water Center.
Introduction
9 / 23
Main Results
Prior: Farmers will refuse to install a meter
In Fact: 65% of candidate consumers consented to participate.
Prior: Farmers will tamper with meters
In Fact: Only 3% of meters malfunctioned (unlikely due to tampering)
Prior: Farmers can easily save water if they only have an incentive
In Fact: No evidence of response to the price signal.
Introduction
10 / 23
Main Results
Prior: Farmers will refuse to install a meter
In Fact: 65% of candidate consumers consented to participate.
Prior: Farmers will tamper with meters
In Fact: Only 3% of meters malfunctioned (unlikely due to tampering)
Prior: Farmers can easily save water if they only have an incentive
In Fact: No evidence of response to the price signal.
Introduction
10 / 23
Main Results
Prior: Farmers will refuse to install a meter
In Fact: 65% of candidate consumers consented to participate.
Prior: Farmers will tamper with meters
In Fact: Only 3% of meters malfunctioned (unlikely due to tampering)
Prior: Farmers can easily save water if they only have an incentive
In Fact: No evidence of response to the price signal.
Introduction
10 / 23
Study Area
Rapidly declining water tables (last three decades)
Deep wells (≈ 1000 ft), Powerful pumps (≈ 60 HP)
Introduction
11 / 23
Study Area
Pumping consumes more than half of all electricity supply
A precursor for the rest of groundwater intensive parts of India?
(Shah 2007)
Introduction
12 / 23
Pilot Design
In Gujarat, ag. consumers pay a bi-monthly flat rate amounting to about
15% of total value of electricity used.
GoG decided to deliver the compensation in the form of reductions to flat
bill - compensation thus capped at 15%.
Price was set at 2.5 Rs. / KwH.
Introduction
13 / 23
Pilot Design - Baseline
Determination of baseline is complicated by the lack of individual metering
of past consumption.
All consumers receive the same duration of poser supply, but differ in their
pump capacity.
Baseline consumption was determined as: Pump HP × Average hours of use
Issue: similar to grandfathering plus (lack of) rewards to (real) false
reductions in use.
Introduction
14 / 23
Pilot Implementation
UGVCL implemented the program at 4 feeders of the Kukarwada sub-station.
Out of 113 eligible consumers, 85 consent.
Introduction
15 / 23
Pilot Implementation - Timeline
Meter$
readings$
begin$
Load$
verified.$
Program$
announced.$$
Rainy$season$
First$
compensaEons$
given$
Winter$season$
Feb$ Mar$ Apr$ May$ Jun$
Jul$ Aug$ Sep$ Oct$ Nov$ Dec$ Jan$ Feb$ Mar$
2011$ 2011$ 2011$ 2011$ 2011$ 2011$ 2011$ 2011$ 2011$ 2011$ 2011$ 2012$ 2012$ 2012$
Introduction
Summer$season$
Apr$ May$
2012$ 2012$
16 / 23
Pilot Implementation - Timeline
2012 Monsoon was very abundant.
Consumers had high degree of awareness and confidence about the
program (at least from November onwards).
30#
100%#
25#
80%#
1,000#Rs#
20#
60%#
15#
40%#
10#
5#
4.32%
1.09%
0.08%
0#
Apr.Jun#
Jul.Aug#
Sep.Oct#
0.29%
1.43%
1.22%
0%#
Nov.Dec#
Jan.Feb#
Mar#
Basline#ConsumpHon#(Monthly,#per#Consumer)#
Actual#ConsumpHon#(Monthly,#per#Consumer)#
Amount#Rebated#(Monthly,#per#Consumer)#
FracHon#of#Consumers#Eligble#for#Rebates#
Introduction
20%#
17 / 23
Evaluation I - Feeder Level Data
The pilot was implemented in 4 feeders (not randomly chosen) out of
32 starting April 2011. Feeder level data available since 2009.
Difference in Difference estimation: no effect found. With 95%
confidence: less than 7% saving.
logCf ,y ,m = Tf ,y ,m + af + by ,m + f ,y ,m
Treatment
2012
2012+2013
2012
2012+2013
0.0238
[-0.137,0.184]
0.0609
[-0.0694,0.191]
0.0258
[-0.0742,0.126]
0.0627
[-0.0235,0.149]
927
Yes
Yes
Yes
No
1251
Yes
Yes
Yes
No
927
Yes
No
No
Yes
1251
Yes
No
No
Yes
N
Feeder FE
Year FE
Month FE
Year,Month FE
95% confidence intervals in brackets
∗
p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
Introduction
18 / 23
Evaluation II: Hour Meters
The evaluation is challenging because the control group is not
metered (and suspicious)...
Solution: we provided consumers with hour meters: devices that
measure duration of pump operation rather than electricity
consumption.
Virtually all consumers approached were willing to install.
Treatment effect on the consenting consumers (upward biased) is
bound by 0.6 hours (95% confidence).
Introduction
19 / 23
Self Reported Impacts
Consumers were asked What are you doing to save energy?
Treated consumers reported significantly more (unobservable) effort,
and most claimed this was a response to the incentive.
Saving water
Mean
0.20
Efficient irrigation
0.11
Reducing hours of use
0.11
Introduction
Full Control
0.16*
(0.08)
-0.00
(0.07)
0.22***
(0.06)
Village FE
0.33**
(0.15)
0.06
(0.12)
0.20*
(0.12)
In same villages
0.19*
(0.10)
0.03
(0.08)
0.22**
(0.09)
Village FE
0.33**
(0.15)
0.06
(0.12)
0.20
(0.13)
20 / 23
Discussion
Why did farmers not respond to the incentive?
Possible reasons:
Compensation was too low, in level and in cap
Lack of familiarity with water saving technologies
Specific local factors, in particular the shareholding pattern.
Introduction
21 / 23
Discussion
40
Why did consumers not reduce water sales in favor of the compensation?
Strong social norms control water markets
Contrast to results from West Bengal (Mukerjee 2012)
30
Compensation
Frequency
20
Generation Cost
0
10
Cost to Non-Ag
.5
Introduction
1
1.5
2
2.5
3
3.5
4
4.5
Price of Electricity Sold by Consumer (Rs. per KwH)
5
5.5
22 / 23
Conclusion
The pilot has shown that voluntary shift to metering and billing is
possible: first time in decades in a groundwater intensive part of India.
Social norms around water sales and lack of availability of water
saving technologies may limit the response to the incentive.
Next Steps
A better designed, larger experiment is needed, with higher
compensation.
A survey of the surrounding areas found more favorable conditions for
an expansion (Fishman et al, 2013). Other states?
Private sector involvement?
Introduction
23 / 23