World Development Vol. 84, pp.

282–298, 2016
0305-750X/Ó 2016 Elsevier Ltd. All rights reserved.

www.elsevier.com/locate/worlddev
http://dx.doi.org/10.1016/j.worlddev.2016.02.003

Collective Action in Decentralized Irrigation Systems: Evidence

from Pakistan
AATIKA NAGRAH a, ANITA M. CHAUDHRY b and MARK GIORDANO c,*
a
Independent scholar, Pakistan
b
California State University, Chico, United States
c
Georgetown University, United States
Summary. — Pakistan, home of the world’s most extensive irrigation system, has followed a global trend in irrigation to devolve man-
agement from government to farmers and farmer organizations. We implement the most extensive survey of irrigation management ever
conducted in Pakistan to examine variation in farmer participation in managing local irrigation systems under the new governance re-
gime. Building on and adding to previous work in Pakistan and elsewhere, we examine the effects of privately accessed groundwater,
groundwater quality, surface water, and other factors along 5 different community management responsibilities, including roles that ex-
isted prior to decentralization (maintenance of the watercourse and dispute resolution) and new roles that were created after (voting to
elect representative at the higher level, collection of water charges, and holding internal meetings). Each responsibility represents a col-
lection action problem for the community. We find that while group leader education is important in successful execution of the roles,
voting has in general been taken up enthusiastically while water charge collection remains, not surprisingly, low. We find a previously
unreported inverse-U shaped relationship between groundwater availability and farmer participation, and that availability of high
groundwater quality (less-saline) groundwater significantly reduces participation, as does high variation in canal water supply within
a growing season. We also find, contrary to expectation, that communities in the tail reaches of the system, known to receive worse canal
water service, more actively participate in managing their local irrigation systems. The findings have important implications for the con-
tinued push for decentralized surface irrigation management in South Asia and elsewhere as groundwater irrigation continues to expand
and equitable access to surface water remains a policy concern.
Ó 2016 Elsevier Ltd. All rights reserved.

Key words — Asia, Pakistan, groundwater, water user associations, devolution

1. INTRODUCTION Meinzen-Dick, Raju, and Gulati (2002) found that water

In the late 1990s, the provincial governments of Pakistan
adopted laws and policies supported by international donors
to transform their centralized irrigation departments and
experiment with new forms of local governance. The move
in Pakistan was part of a broader and continuing trend to
decentralize irrigation system management and give greater
responsibility to farmers in that management (Bardhan &
Mookherjee, 2008; Ramesh, 2013; Senanayake, Mukherji, &
Giordano, 2015). Under the reforms, known variously as Irri-
gation Management Transfer (IMT), Participatory Irrigation
Management (PIM) or Farmer Managed Irrigation Systems
(FMIS), responsibility is transferred from government to
farmers or groups of farmers within Water User Associations
to actively manage local irrigation-related resources, collect
water charges to finance system maintenance, and coordinate
water use with higher structures in the irrigation system.
Empirical evidence of the achievements of reform through
increased farmer participation is mixed (Senanayake et al.,
2015; Shyamsunder, 2008). This is also the case in Pakistan
where studies, although limited in scope and methodology,
have found varied outcomes on system financial performance
(cost recovery), quality of operations and maintenance, relia-
bility of water delivery (Haq, 2010; Latif & Tariq, 2009),
and effect on farm productivity (Raza, Ashfaq, Zafar, &
Baig, 2009). It is clear though that reform success in terms
of improved irrigation management, farm productivity and
farmer incomes depends in part on whether farmer participa-
tion is able to fill the management space vacated by the gov-
ernment. A key question is thus what prompts and sustains
effective participation in a decentralized and devolved system?
282
scarcity, access to markets, leadership potential, and size of
the community play an important role in emergence of farmer
organization at the secondary canal level. While this work
focused on why farmer participation and organization emerge,
it leaves open the questions of whether and why farmer partic-
ipation and organization continues after reform implementa-
tion. Other studies that have discerned empirical regularities
in community cooperation in managing irrigation-related pub-
lic goods using statistical methods have focused on the case of
India (Bardhan, 2000; Mishra, Ghosh, Nanda, & Kumar,
2011), Mali (Vandersypen et al., 2008), and Philippines
(Araral, 2009; Fujiie, Hayami, & Kikuchi, 2005). Araral
(2009) makes several significant improvements in methodology,
* The authors wish to thank the participants of the Association of
Environmental and Resource Economists sessions at the 88th Annual
Conference of the Western Economic Association International, held in
June 28–July 3, 2013 at Seattle, Washington, for their constructive
comments. Also, comments and suggestions received from John Eckalbar,
Hasan MacNeil, Ruben Sargsyan, Pete Tsournos, and Kai Wegerich were
useful in improving the manuscript. We also thank the organizers and
participants of the Center of Water and Environment Lecture and the
Anthropology Forum (both as CSU Chico) for providing a platform for
constructive discussion. The data used here were collected by the
International Water Management Institute with financial support of the
Embassy of The Netherlands, Islamabad. The views presented here are
those of the authors and do not necessarily represent the views or policies
of the International Water Management Institute or the Embassy of the
Netherlands. Final revision accepted: February 29, 2016.
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN
most notably a large sample size and rigorous testing of interac-
tion effects.
This study contributes to this literature in three key ways.
First, we study the effect of privately accessed groundwater
on the degree of collective action for public irrigation systems.
Surface irrigation economy of South Asia, including Pakistan,
has been transformed by groundwater use in recent decades.
As of 2011, of the 19.41 million hectares of irrigated land in
Pakistan, canals alone irrigated 34% of area, 43% was irrigated
by ‘‘canal, tubewells and wells” and 22% are irrigated by wells
outside formal canal commands (Pakistan Bureau of
Statistics, 2012). Surface water is managed as part of the for-
mal irrigation system as described above, but groundwater,
accessed by millions of individual farmers using their private
resources to install and operate diesel-powered groundwater
pumps, is essentially unmanaged (Shah, 2008; Shah et al.,
2009). As Shah (2008), p. 137 notes, this extensive groundwa-
ter use effectively informalizes the water economy and takes
most water appropriation and use outside the formal control
of the irrigation bureaucracy. The decentralization of
Pakistan’s irrigation system, and the discussion surrounding
its mechanisms and effects, was entirely focused on the
management of surface irrigation systems, even though the
primary source of irrigation for many farmers is groundwater.
We investigate how dependence on groundwater affects the
efficacy of decentralized institutions.
Second, we improve our measurement of availability of sur-
face water by including measures of its supply variability in
addition to the standard water supply measures. Variation in
canal water supply is an important aspect of performance of
service delivery and is known to have important effects on
agricultural production and farmer incomes. Studies on collec-
tive action in irrigation so far have only examined the effects of
total canal water delivery, whereas variability of water
received may also have a strong affect on productivity and
willingness to participate in maintaining the irrigation system.
Third, we measure collective action of farmers in managing
their irrigation systems along 5 different domains, each of
which relates directly to the responsibilities awarded to farm-
ers since the transfer of irrigation management. Two of these
domains, maintenance of watercourse channel and dispute res-
olution within the watercourse community, are formalization
of traditional watercourse community roles while the other
three, collecting seasonal water charges, holding internal meet-
ings, and electing farmer representatives, are new responsibil-
ities for communities. Comparison of performance across the
new and the old responsibilities measures provides insights
into the challenges of adapting to new roles especially as we
find that certain community characteristics are correlated with
improved performance in one role but not in another.
One of the prominent features of the statistical collective
action literature is the disagreement across case studies on
the effects of various resource or community characteristics
on successful collective action. For example, size of the com-
munity is seen as positively influencing collective action in
some case studies (Araral, 2009), while it may have a negative
effect in other settings (Meinzen-Dick et al., 2002). This may
be taken as a shortcoming or an indication of the ad-hoc nat-
ure of this literature even though the goal should not be to find
generalizable universal results but rather to find micro-
situational variables that work in a particular socio-
ecological context (Ostrom, 2005, 2011). In our statistical case
study we compare the effects of resource and community
characteristics across multiple indicators of collective action,
which suggests an alternative explanation of this apparent
283
disagreement. We propose that it is the underlying nature of
the task or activity that may explain the differences in results
of a community or resource characteristic.
We use a large data set of 1845 watercourses in Pakistan,
which like Araral (2009) allows us the degrees of freedom to
test the effects of a large number of independent variables.
We selected our sample watercourses randomly and collected
data on the performance of each selected farmer group includ-
ing non-performers. As noted by Araral (2009) and Agrawal
(2001), this sampling strategy helps avoid selection bias.
Watercourse-level farmer organizations around the tertiary
irrigation system are the fundamental building block of the
decentralized irrigation management (Uphoff, 1986). Farmers
sharing a watercourse are located in one or more neighboring
villages so we are able to control for watercourse community
characteristics in a more fine-grained analysis. This strategy
has a distinct advantage over Meinzen-Dick et al. (2002) and
Mishra et al. (2011) who analyzed performance at the
secondary-level canal systems, and therefore could not exam-
ine watercourse-specific and associated village-community-
specific factors that may explain the incidence of and variation
in farmer organization performance.
Finally, to the best of our knowledge, this is the most exten-
sive statistical case study of famers’ participation of irrigation
management in Pakistan since the irrigation system was decen-
tralized and several key responsibilities have been devolved to
farmer groups. There is a growing belief that the present form
of irrigation decentralization in Pakistan has not achieved the
intended goals and that alternatives need to be considered
(Government of Pakistan, 2012). While the Government of
Pakistan has continued to reiterate its commitment to the
reforms process, it has identified lack of concrete evidence
on performance of farmer organizations as hindering its abil-
ity to formulate improved approaches (Government of
Pakistan, 2012). A direct consequence of this knowledge gap
is the frequency with which the regulatory framework is being
modified and experimented with at the provincial level in an
attempt to ‘‘fix” the performance issues of the new organiza-
tions. However, without a clear understanding of farmers’ par-
ticipation in irrigation management, it is not clear what the
best paths forward might be.
The rest of the paper is organized as follows: Section 2 gives
an overview of the literature on empirical studies of collective
action for natural resource management, focusing on
irrigation-related studies; Section 3 presents our methods
and results. Section 4 concludes with a discussion of results
in the broader context of irrigation management transfer.
2. COLLECTIVE ACTION FOR LOCAL PUBLIC
GOODS: THEORY AND EVIDENCE
Collective action problems as described by Ostrom (2005)
refer to settings in which individuals select strategies, e.g., level
of effort to maintain the local canal, based on a calculus that
maximizes their private material benefits, but individual strate-
gies are interdependent in that the final outcome, e.g., the state
of the canal, is a result of everyone’s effort and generates
benefits to the entire community. Ostrom (2005) referred to
social dilemmas as situations where individual chose strategies
that yield Nash equilibria that generate lower joint outcomes
than could have been achieved. Research on determinants of
cooperation among self-interested agents on provision of local
public goods or management of common property resources
has suggested a large set of variables that could potentially
284 WORLD DEVELOPMENT
affect successful self-organization and collective action. These
variables can be broadly grouped into three categories: (i)
resource characteristics, (ii) resource user characteristics and,
(iii) characteristics of governance structures (Baland &
Platteau,1996; Ostrom, 2005; Sandler, 1992; Uphoff, 1986;
Uphoff, Wickramasinghe, & Wijayaratna, 1990; Wade, 1988).
The degree of water scarcity in an irrigation system has been
identified as one of the most important factor influencing
farmers’ incentives to organize around water management
(Bardhan, 2000; Bardhan & Dayton-Johnson, 2002;
Meinzen-Dick et al., 2002). An inverse-U relationship between
water scarcity and farmer participation has been proposed in
which there is little incentive to cooperate at very low or very
high scarcity levels. If farmers receive adequate water supplies
the marginal benefit of organizing to further improve water
supply are very low, and similarly if water supplies are highly
scarce farmers may again perceive very low marginal benefit
from organizing to maintain the watercourse.
Size of the resource user community is also important in
determining whether the community is successful in organizing
around a common purpose. Uphoff (1986) found group sizes
vary depending on size and hierarchical nature of the irriga-
tion system. Theoretically, smaller groups have an advantage
in cooperation because strategies are more likely observable;
the share in the loss from not following the rules are larger;
interlinkages among group members are likely to be more
important; and negotiation costs are lower (Olson, 1965). This
was indeed found by empirical evidence (Bardhan, 2000;
Bardhan & Dayton-Johnson, 2002; Meinzen-Dick et al.,
2002; Wade, 1988).
The role of leadership in tapping into the capacity of the
watercourse community to self organize, improve their water-
course, and lobby for better service cannot be underestimated.
This is especially true when collective action needs to serve as a
substitute for the state. However, because it cannot rely on the
coercive power of the state (e.g., the ability to tax and enforce
contracts), collective action is much harder to set in motion.
Two essential ingredients are then required: leadership and
trust. A leader is required who is capable of convincing com-
munity members that they should voluntarily contribute to the
public good. Krishna (2001) has argued that good leaders are
agents that can convert the stock of social capital into a flow
of benefits. Trust is necessary to resolve conflicts among com-
peting interests and to reduce fears of free riding. This was
indeed found by Fujiie et al. (2005).
Another variable that affects successful collective action for
resource management is its salience or relative importance to
the users. Gemma and Tsur (2007) and Provencher and Burt
(1993) have shown that groundwater allows farmers to aug-
ment canal irrigation supply and buffer against its variability.
Even though collective action is organized for management of
public goods for surface irrigation, privately accessed ground-
water increases the payoff from well-functioning surface sys-
tems, increasing the incentive for collective action (Latif,
2007). In other words, better access to groundwater can
improve collective action. But Meinzen-Dick et al. (2002)
found an opposite effect: surface irrigation systems with higher
number of groundwater wells exhibited less collective action.
This may be an indication of what Shah (2008) later found:
in South Asia good quality groundwater is offering a possibil-
ity of exit from canal irrigation. Even earlier, Hirschmann
(1970) had shown that the opportunity to exit from a public
good may lead to underinvestment in that good. If Shah
(2008) is correct, groundwater may be offering an exit
opportunity, at least to farmers who can afford the expense
of a groundwater pump. The option of good quality ground-
water may reduce the need to depend on, and invest in collec-
tive action for the provision of surface water irrigation
services. This may have significant implications for the viabil-
ity of canal-based management organizations. Groundwater’s
role deserves to be examined more carefully.
Women constitute an increasing proportion of the agricul-
tural labor force, at 30%, double the proportion of 20 years
ago (Yu et al., 2013). Part of this increase has been due to
increased outmigration of men to cities or other countries,
leaving women members in charge of farming operation.
Woman-headed households tend to be poorer and may be
excluded from irrigation-related community activities due
to cultural norms dictating that canal irrigation is a man’s
job (Agarwal, 1997). Women are deeply engaged in agricul-
tural labor but evidence suggests that they are not as
engaged in the irrigation-related community meetings or
activities. For example, Meinzen-Dick and Zwarteveen
(1998) found no women members in water user associations
in Pakistan. In addition to the social norms restricting
women’s participation in the public domain, women may
also face high opportunity costs of their time due to high
domestic workloads. As devolution brings irrigation man-
agement closer to the communities, we explore whether
communities with a greater proportion of woman-headed
households tend to perform better or worse in managing
irrigation-related tasks.
None of the governance-related variables identified by the
collective action research are relevant for us here, because
the genesis of all irrigation associations we examine lies in
the irrigation management transfer. For example Shivakoti
and Ostrom (2002) had shown that irrigation associations
(IAs) have a stronger sense of ‘‘ownership” when the commu-
nity is self-organized compared with associations organized by
government agencies. In our study area, formation of all irri-
gation associations were prompted by the PIDA Act of 1997,
by which Provincial Irrigation Departments were transformed
into financially autonomous entities, referred to as Provincial
Irrigation and Drainage Authorities. In Punjab, the provincial
government structured the authority into Area Water Boards,
each of which consisted of several Farmer Organizations
(FOs). 1 An FO took over the management responsibilities
of the secondary canals (distributary) and consisted of directly
elected farmer representatives from each tertiary canal (water-
course). Each of these directly elected members head their
respective watercourse-level Khal Panchayat (KP)
(Government of Sindh, 1997; Government of the Punjab,
1997). See Figure 1 for a comparison of the old and the new
system.
Moreover, age of the irrigation system itself, considered
important by Fujiie et al. (2005), is fairly uniform across dis-
tributaries. So, even though irrigation management transfer
brings the capability of a community to manage its local pub-
lic goods into sharp focus, some of the key governance vari-
ables are not at play in our research.
But despite the inorganic appearance of watercourse-
governance bodies, there are hopes that decentralization
would create a trajectory of change to improve the social orga-
nization of irrigation. By the early 1980s, Pakistan’s irrigation
systems were widely recognized to be performing substantially
below their potential. Problems included irrigation infrastruc-
ture in desperate need of repair, poor revenue collection to
fund that repair, expansion of irrigated area beyond design,
vast inequalities in the distribution of canal water and overall
very low agricultural productivity (Murgai, Ali, & Byerlee,
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN 285

Old system (since ~ 1870) New System (since 1997)

Provincial Irrigation and Drainage Authority
Province (PIDA)

Canal Circle Area Water Board (AWB)

10 farmer and 9 non-farmer members and engineers
from Irrigation Department
Responsible for managing and distributing irrigation
water, through formal volume-based contracts with
Farmer
Organizations, and trade water with other utilities
Irrigation Department (Govt
appointed officials and
Distributary engineers) Farmer Organization (FO)
10 elected farmers from KPs

Responsible for (1) supplying water to irrigators (2)

O&M of distributaries (3) collect water charges (4)
make payments to Area Water Board
Watercourse Khal Panchayat (KP)
5 elected farmers from the watercourse community

Responsible for (1) maintaining the watercourse (2)

dispute resolution (3) collect water charges (4) hold
internal meetings (5) vote in the FO election
Responsible for
(1) maintaining the watercourse
Farmers (2) dispute resolution

Figure 1. Comparison of Centralized Irrigation Department with the PIDA.

2001;World Bank, 1994; Yu et al., 2013). Larger and more
powerful landowners were paying bribes to the irrigation offi-
cials to withdraw more than their share, which exacerbated the
inequities in the system (Mustafa, 2001; Rinaudo, 2002; Van
Halsema & Vincent, 2006). These discouraging trends,
together with fiscal constraints limiting the state’s ability to
improve the infrastructure, helped create the momentum for
a policy shift to devolve Pakistan’s irrigation management
from state agencies to user groups. It was argued that restruc-
turing and decentralizing the government-managed irrigation
department into autonomous units and devolving manage-
ment, maintenance, and fiscal responsibilities to local farm
communities would give greater responsibility to farmers
and increase their participation in management. This, in turn,
would improve the physical state of the infrastructure,
improve on farm water management, and increase revenue
collection in a virtuous circle (Vander Velde & Tirmizi,
2004; World Bank, 1994).
Even before decentralization, a certain level of collective
action and self-governance was needed for effective operation
of the system. Farmers were already maintaining their water-
courses and resolving disputes; a concern is whether the insti-
tutional change will ignore the potential of the community to
self-organize or build on that potential. As Ramesh (2013) has
pointed out that local resources, institutions, political pro-
cesses and social capital play a larger role in shaping outcomes
of decentralization than the decentralization policy itself. Our
explanatory variables may help identify what contributes to
higher social capital. A community’s social capital and the
resulting successful collective action in the community may
therefore be correlated with degree of water scarcity, size of
the community, opportunity to exit from the local public
good, and effective leadership in the community.
3. FRAMEWORK AND HYPOTHESES
(a) Unit of analysis: watercourse community and Khal
Panchayat (KP)
Our unit of analysis is a watercourse community that shares
a watercourse and resides in one or more neighboring villages.
The watercourse carries water from the secondary outlet to the
farmers’ fields. Regardless of the size of the irrigation system,
this field channel level—the unit command area—is the com-
mon denominator in all irrigation systems around the world
(Uphoff, 1986, p. 88).
Each watercourse community is governed by a Khal Pan-
chayat; ‘‘Khal” means watercourse and ‘‘Panchayat” refers
to the five-member governing body. The 1997 PIDA Act lays
out its responsibilities of the Khal Panchayat, referred to as KP
from here onward, as: (1) maintenance of the watercourse, (2)
dispute resolution, (3) collection of water charges (abiana)
from irrigating farmers on the watercourse, (4) hold monthly
meetings within KP and, (5) participation in distributary-
level Farmer Organizations (Government of the Punjab,
2010).
Each KP responsibility is a particular form of collective
action problem. Watercourse maintenance refers to removing
sediment and weeds from the watercourse channel. 2 Periodic
watercourse maintenance, a local public good, is important
for unimpeded water flow to all farmers, especially those
appropriating water in the tail reaches. Resolving disputes,
especially when they arise over sharing of water flowing in
the watercourse, can avoid disruption of the flow of service.
KP meetings can aid exchange of information and developing
trust. As noted by Frohlich and Oppenheimer (2000), individ-
uals do not have complete information in most settings, but we
286 WORLD DEVELOPMENT
can assume that individuals lean more accurate information
about the structure of the situation they are in and behavior
of others involved when they communicate in a repeated situ-
ations. Finally, electing upper tier members can be a direct for-
mal route to voice local concerns at a higher level. We examine
the performance of a KP in fulfilling its responsibilities as a
reflection of the degree of collective action within the water-
course community. This is our key assumption, which enables
us to interpret KP performance as the community’s capacity
of self-organization and collective action.
Each KP consists of five members, all farmers from the
respective watercourse, with four general members and one
chairperson. KPs are formed via elections at the watercourse
level and all irrigating farmers are eligible. 3 Each KP and its
watercourse community represent a clearly defined resource
user group. Although a KP does not have a formal role in
determining the allocated share of water to each member farm,
it is responsible for maintaining its watercourses in good con-
dition in order to receive the agreed canal water delivery. Ver-
ified violations or non-payment of water charges can result in
suspension of water delivery to the watercourse. Theoretically,
KPs can also liaise or bargain with agency officials, or the FO
since the implementation of PIDA Act, for adjustments in
cases when flow timing and thus allocated volume of water
is disturbed, for instance due to a damaged outlet. These roles
of direct service delivery to farmers in the watercourse com-
munity in addition to the KP’s responsibility of electing FO
members, makes the KPs the fundamental institution on
which the decentralized management structure is built.
It is worth noting that some of the KP responsibilities in the
PIDA Act were a mere formalization of existing watercourse
community roles, though others created entirely new roles
for farmers. For example, since the Canal and Drainage Act
of 1873, watercourse communities were responsible for peri-
odic maintenance and upkeep of watercourses, although the
exact mechanism for sharing responsibility was left to the dis-
cretion of the farmers (de Klein & Wahaj, 1998). Similarly,
farmers had been de facto responsible for resolving disputes
arising from water sharing. 4 Assessment and collection of
water charges, on the other hand, had traditionally been the
role of the Revenue Department, but the PIDA Act awarded
this responsibility to the KPs. 5 Similarly, recommendations
on holding monthly community meetings and electing farmer
representatives at the distributary level formalized the func-
tioning of the watercourse community. Anecdotal experience,
and the authors’ field experience, suggests that relatively influ-
ential farmers had always engaged with higher level authorities
for their personal gains. The key difference was that now a
much broader base was eligible to participate in the water-
course management in a more structured and formalized
format.
However, it must be mentioned that the role of a KP in fully
controlling its water supply is somewhat constrained due to
the nature of Pakistan’s run-of-the-river, hierarchical irriga-
tion system in which water flows from branch canals into sec-
ondary canals (distributaries), which in turn flows into tertiary
canals (watercourses). KPs, the farmer body at this last irriga-
tion channel or watercourse, have therefore limited control of
the service delivered to them from upstream except through
their Farmer Organization or Area Water Boards.
(b) Hypotheses
Given the literature reviewed in the previous section, we
propose hypotheses related to characteristics of the resource
(watercourse system), and characteristics of the resource users
(KP community), in order to explain variation in KP perfor-
mance.
(i) Canal water delivery hypotheses
We expect poor KP performance when canal water avail-
ability is very scarce or very abundant. The ‘‘canal water deliv-
ery hypothesis” is the inverse-U hypothesis noted earlier that
marginal increase in canal water availability improves KP per-
formance at low levels of canal water availability, but if canal
water availability is already high, a marginal improvement in
delivery supply worsens KP performance.
(ii) Groundwater hypotheses
Given extensive conjunctive use of surface water and
groundwater in our sample area we explore the role of the
canal water delivery and groundwater availability separately.
We propose a ‘‘groundwater productivity hypothesis”, which
states that better groundwater availability improves KP per-
formance. When groundwater and surface water are joint
inputs or complements in farm production, improved ground-
water availability increases gains from improving canal water
delivery system.
The ‘‘exit hypothesis” states that better groundwater avail-
ability leads to worse KP performance. If farmers view
groundwater as substitute of canal water, then availability of
groundwater may undermine incentives to improving the per-
formance of surface water irrigation systems and therefore
would have worse KP performance. There is theoretical justi-
fication of both groundwater productivity and exit hypothesis
and the net result would be determined by which effect domi-
nates in a particular irrigation system. We allow for both
effects in our estimation.
(iii) Community size hypothesis
Size of the community is known to be an important determi-
nant of collective action and we expect KPs of larger commu-
nities to perform worse. We refer to this as the ‘‘Watercourse
community size hypothesis”, which states that large communi-
ties will have worse KP performance.
(iv) Leadership hypothesis
The literature suggests that a community with better leaders
will have a more effective KP. We refer to this as the ‘‘Leader-
ship hypothesis”.
(v) Gender hypothesis
Does a community with a greater share of women as its
shareholders perform differently than others? There is no for-
mal mechanism to include women members in KPs and the
informal mechanisms may constrain women from participat-
ing in certain KP activities such as meetings or watercourse
maintenance. Therefore we expect a KP with a greater share
of woman-headed households to perform worse.
(c) Study area and data collection
Our study area is Bahawalnagar Canal Circle in eastern
Punjab, Pakistan, where water is scarce, irrigation develop-
ment is critical, and was one of the first locations where pilot
programs for farmer participation in irrigation management
were developed. Bahawalnagar Canal Circle is one of the lar-
gest canal systems in Pakistan with a gross command area of
approximately 690,000 ha. It receives 3 billion cubic meters
(2.5 million acre feet) of surface water annually or nearly 8%
of Punjab’s total irrigation water (4% of the national total).
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN
Wheat is the primary Rabi (winter) crop and cotton is the pri-
mary Kharif (summer) crop.
Of the total 70 distributaries and 4,259 watercourses in the
Bahawalnagar Canal Circle, this study is based on a random
sample of 1,845 watercourses located across 44 distributaries
in head, middle and tail reaches of their respective main canals
and distributaries. 6 The sample watercourses fall across three
different canal divisions, Fordwah, Hakra, and Sadiqia, which
vary in their size, status of infrastructure, percentage of non-
perennial distributaries, and groundwater quality. Yet, this
area is part of a single agro-climatic zone so altitude, rainfall
or temperatures are fairly uniform. In-person interviews of
KP chairpersons and farmers in sampled watercourses were
conducted over the period of 15 July–15 August 2012.
Detailed information was collected regarding activities of the
KP in the two most recent growing seasons: Rabi (winter)
2011–12 and Kharif (summer) 2011. 7 Data were compiled
regarding the activities of the KP, and physical and socioeco-
nomic environment of the watercourse community.
Primary data were supplemented with two secondary
sources. First, daily observation of canal water discharge
recorded at the head of the distributary for each of the 44 dis-
tributaries in our sample. These data were obtained from the
Programme Monitoring and Implementation Unit (PMIU)
of the irrigation department, for a seven-year period (Rabi
2005–06 to Rabi 2012–13). A total of 108,390 daily discharge
observation points were used to construct variability in daily
discharge in canal water. Second, data on socioeconomic
development indicators such as literacy rates, child malnour-
ishment, percentage of population using improved drinking
water sources, were collected from the most recent, 2007–08
Multiple Indicator Cluster Survey. Table 2 shows that our
sample watercourses fall in six tehsils: Bahawalnagar, Chish-
tian, Fort Abbas, Haroonabad, Minchinbad, and Hasilpur. 8
Our study area is one of the poorest regions in the province
of Punjab, Pakistan (Cheema, Khalid, & Patnam, 2008).
(d) Definition and measurement of KP performance: 5 domains
of collective action
We examine KP performance on each of its five legal
responsibilities under the PIDA rules. These include: (1) Main-
tenance of watercourse, (2) Collection of water charges from
irrigating farmers on the watercourse, (3) dispute resolution,
(4) hold monthly meetings and, (5) participation in
distributary-level Farmer Organizations. Descriptive statistics
of KP performance indicators are given in Table 1.
(i) Watercourse maintenance
Watercourse maintenance is a labor-intensive process and
has traditionally been accomplished by farmers volunteering
their time. We calculate the percentage of farmers in the water-
course community who contributed time for maintenance
Table 1. Summary statistics—Khal Panchayat (KP) performance
Khal Panchayat (KP) Performance
Farmers contributing labor for maintenance works (%)
Acceptance of KP’s role in dispute resolution (Yes = 1, No = 0)
KP involved in collecting water charges (Yes = 1, No = 0)
Farmers attending KP meetings (%)
KP voted in Farmer Organization elections (Yes = 1, No = 0)
Source: Authors.
a
Reported only for dummy variables.
287
works organized by the KP in winter (Rabi) season. Table 1
shows that on average 54% of the farmers contributed labor
toward watercourse maintenance.
(ii) Dispute resolution
Degree of involvement of a KP in local dispute resolution
was measured by asking the chairperson of each KP if he
agreed with the statement, ‘‘the farmers on the watercourse
accepted the KP’s role in dispute resolution in the watercourse”.
We constructed a dummy variable with a value of 1 when the
chairman responded with ‘‘Strongly Agree” or ‘‘Agree” and 0
for other responses (‘‘Strongly Disagree”, ‘‘Disagree” and ‘‘To
some extent”). Table 1 shows that the dummy variable for dis-
pute resolution takes the value of 1 in a 91% of cases. Disputes
around irrigation water arise for a whole host of reasons, such
as water sharing schedule or unauthorized use of watercourse
channel, but these disputes are seldom officially reported.
Farmers generally do not admit to having disputes to out-
siders, but our indirect question of whether the KP is seen
as arbiters in case of disputes yields a more honest and easy
response. We interpret the answer to this question as the KP
chairperson’s self-perception as a respected figure and author-
ity in the community.
(iii) Attendance in KP meetings
We calculated a dependent variable indicating the percent-
age of community attending KP meetings. Table 1 shows that
about 68% of the community participates in the KP meetings.
(iv) Collection of water charges
We defined a dummy variable that takes a value of 1 if KP
was involved in water charge collection, and 0 if it was not.
Table 1 shows that only about 33% of KPs reported that they
were successfully taking over collection. Water charge collec-
tion dummy variable indicates success in the KP taking over
this new responsibility. 9
(v) Voting in farmer organization elections
Farmer Organization is the governing body for the sec-
ondary canal (or distributary), which is the parent channel
for the watercourse. We constructed a dummy variable that
takes a value of 1 if the KP voted in the election, and 0 other-
wise. Table 1 shows that an overwhelming majority of the KP
chairpersons, 85%, had voted in the most recent Farmer Orga-
nization election. 10
Each of the public goods provided by the KP has a different
distribution of costs and benefits across its members, which
may partly explain the relative success of KP in each role.
Maintenance of watercourse requires participation of several
community members to perform the physical task of removing
silt although benefits may disproportionally accrue to farmers
in the tail reach of the watercourse. Also, unlike collection of
water charges, it is an older responsibility with very clear and
Mean Std. dev. Min Max Count if 1a Count if 0a
54.19 39.05 0 100
0.91 0.28 0 1 1,683 162
0.33 0.47 0 1 611 1,234
68.27 37.40 0 100
0.85 0.36 0 1 1,569 276
288 WORLD DEVELOPMENT

tangible benefits. Performance of KPs in collecting water
charges is perhaps the least successful of KP roles, with only
33% of the KPs involved, partly because of unclear benefits
and very clear costs. Voting in FO elections, although another
new responsibility has a very average high participation rate of
85%. It is only an annual activity, perhaps less costly, and its
benefits, e.g., voicing the communities’ concerns in the FO or
learning of the latest development in irrigation management
and opportunities, although less tangible may be more evenly
distributed across members. 11
(e) Definition and Measurement of Explanatory Variables
Table 2 provides the summary statistics of all explanatory
variables used to test the hypotheses in Section (b).
(i) Location of the watercourse
In much of South Asia, including Pakistan, discharge of
canal water delivered to the plot or even the head of the water-
course is not measured. Discharge of canal water is measured
only at the head of the distributary, which could be several
hundred meters away from the head of the watercourse and
the plot. We therefore use location of the watercourse in the
irrigation system, infrastructure condition of the distributary
and watercourse, and inter and intra seasonal variability of
the canal water delivery, as proxies of the quality of water ser-
vice delivery to the watercourse. Given the spatial inequities in
canal water delivery, the distance of the watercourse from the
head of the distributary and the main canal can serve as a
strong proxy for canal water availability (Latif, 2007). Chan-
nels closer to the head typically receive higher flows than those

Table 2. Summary statistics—watercourse explanatory variables

Variable Mean Std. Dev. Min Max Count if 1 Count if 0
Canal water delivery
Distributary head (1 = Yes, 0 = No) 0.34 0.47 0 1 621 1,224
Distributary tail (1 = Yes, 0 = No) 0.39 0.49 0 1 714 1,131
Main head (1 = Yes, 0 = No) 0.24 0.43 0 1 440 1,405
Main tail (1 = Yes, 0 = No) 0.57 0.50 0 1 1,045 800
Absolute tail (Distributary Tail  Main Tail) 0.22 0.41 0 1 406 1,439
Absolute head (Distributary Head  Main Head) 0.08 0.28 0 1 154 1,691
Intra-seasonal variation in distributary discharge (average standard 97.28 62.41 2.19 204.20
deviation within season discharge in Kharif seasons for 2006-12)
(cubic meters per second)
Inter-seasonal variation in distributary discharge (standard deviation 22 17.02 0.48 54.48
of average discharge in Kharif 2006–12) (cubic meters per second)
Percentage of watercourse lined 27.73 24.69 0 100
Physical state of the distributary (1 = poor condition, 0.79 0.41 0 1 1,455 390
0 = good condition)
Groundwater availability
Groundwater quality in a watercourse (1 = Poor, 1.69 0.71 1 3
2 = Fair, 3 = Good)
Percentage of gross command area irrigated by groundwater 47.31 33.66 0 100
Size of the watercourse community
Number of shareholders or farmers on watercourse 76.40 66.50 20 1,300
Gross command area of the watercourse (acres) 381.52 178.52 8 2,500
Leadership quality and potential
KP chairperson’s education (years in school) 7.27 4.20 0 14
KP chairperson’s age (years) 48.77 11.24 20 90
KP chairperson’s land located in tail of watercourse (1 = Yes, 0 = No) 0.16 0.36 0 1 291 1,554
College graduates (%) 10.83 13.37 0 100
Influential people (%) 6.93 8.34 0 83.48
Shareholders with a military background (%) 2.56 5.10 0 75
Gender of household heads in watercourse community
Female-headed households (%) 5.60 5.62 0 50
Number of primary schools in the watercourse community 1.22 0.78 0 5
Perennial canal (1 = Yes, 0 = No) 0.67 0.47 0 1 1,229 616
Literacy rate (in population 15 years and oldera 43.95 8.67 30.5 55.3
Child malnourishment, weight for age, % below 2SDa 35.92 4.04 31 42.6
Population using improved sources of drinking water and sanitation (%)a 51.35 11.89 30.8 66.6
Women reporting a Lady Health Worker visit during the past month (%)a 60.54 21.88 23.1 86.2
Unemployment rate (in population 15 years)a 5.03 1.13 3.4 6.9
Fordwah canal division (1 = Yes, 0 = No) 0.33 0.47 0 1 600 1,245
Sadiqia canal division (1 = Yes, 0 = No) 0.41 0.49 0 1 755 1,090
Haroonabad canal division (1 = Yes, 0 = No) 0.27 0.44 0 1 490 1,355
Source: Authors.
a
Source: Multiple Indicator Cluster Survey (MICS), Government of the Punjab, 2007–08.
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN
at the tail of the channel. Following convention, e.g., Meinzen-
Dick et al. (2002), we divide a channel into three parts, head,
middle and tail, such that the head refers to the first 30% of a
channel, the middle refers to the next 30%, and the remaining
40% of a channel is referred to as the tail.
Our sampling scheme at the head-middle-and tail of the dis-
tributaries as well as main canals allows us to improve this
measurement. We develop 4 dummy variables that assign a
head, middle and tail location identifier to each watercourse
in our sample. First, Distributary_Head = 1 if the watercourse
originates from the head of the distributary and 0 otherwise,
and second, Distributary_Tail = 1 if the watercourse origi-
nates from the tail of the distributary, 0 otherwise. Location
of a distributary along the main canal would also affect the
water delivered to all watercourses so two additional dummy
variables are created to identify the location of the distributary
along the main channel. Main_Head, which takes a value of 1
if the watercourse originates from a distributary, located at the
head of the main canal, and 0 otherwise. Main_Tail, which
takes a value of 1 if the watercourse, originates from a dis-
tributary at the tail of the main canal, and 0 otherwise. Addi-
tional dummy variables created from the product of head
dummies (Distributary_Head  Main_Head) and tail dum-
mies (Distribuary_Tail  Main_Tail) are used to explore if
KPs on the absolute head or the absolute tail of the irrigation
system respectively perform differently from others.
Table 2 shows that distribution of the sample watercourses
across the head, middle and tail of the distributaries and main
canals. In line with inverse-U hypothesis, we expect that KPs
located along the head (Main_Head and Distributary_Head
and Distributary_Head  Main_Head) and tail reaches
(Main_Tail, Distribuary_Tail and Distribuary_Tail  Main_Tail)
would display worse KP performance than those in the middle
reaches.
(ii) Variability in canal water supplies
The contribution of canal supplies in agricultural produc-
tion depends not only on cumulative quantity received in a
growing season but also on timing and variability of these sup-
plies. Supply certainty has been found to increase agricultural
productivity (Provencher & Burt, 1993). We use daily dis-
charge data for the Kharif growing season for 7 preceding
years 2006–12 to develop measures of inter-seasonal variability
and intra-seasonal variability of discharge received at the head
of the distributary. Inter-seasonal variability is computed as
the standard deviation in average per acre volume of canal
water received across the 7 growing seasons. Intra or within
season variability, is measured by the standard deviation of
daily discharge of canal water at the head of each distributary.
We compute both measures for summer (Kharif) season,
rather than winter (Rabi) because of presence of non-
perennial canals in our sample, which are shut off during the
Rabi season. While it is not technically possible to have no
variability in canal water supply, high variability within a sea-
son or from one season to the next could be an indication of
poor service delivery and can undermine KP performance. 12
(iii) Condition of canal infrastructure
Upstream infrastructure condition can be a key determinant
of canal water delivery. Distributaries in poor condition will
render the efforts of downstream KPs to desilt and maintain
watercourses meaningless. Conversely, better-maintained
infrastructure may provide greater marginal benefits to collec-
tive action as that action will be more likely to be reflected in
delivery results. We measured infrastructure condition both in
terms of facilities for delivering of the water to the watercourse
289
(the secondary canals) and for delivering of the water from the
watercourse to farmers’ fields (the tertiary canals). Distribu-
tary condition is measured both through the subjective assess-
ment of the KP representative (1 = poor condition, 0 = good
or excellent condition).
(iv) Condition of watercourse infrastructure
The condition of the watercourse was measured as the share
of the watercourse permanently lined with cement or bricks, a
task that is always performed by an outside agency. This is not
to be confused with the watercourse maintenance, which is
about de-silting the watercourse and is the community’s
responsibility. Permanent lining of the watercourse reduces
seepage losses and reduce maintenance efforts, but is also
known to reduce recharge of groundwater.
(v) Groundwater use
We calculated the percentage of watercourse command area
that is irrigated by groundwater and included that as an
explanatory variable that captures the contribution of ground-
water in agricultural production. We add a squared term to
test exit hypothesis for non-linear effects of groundwater use
on KP performance.
(vi) Groundwater quality
Detailed groundwater quality data for each watercourse
command area in the sample were not available for each
watercourse so we elicited farmers’ perceptions regarding the
quality of groundwater (i.e., salinity levels) for agricultural
production. We asked the KP chairmen to rank the quality
of groundwater for agricultural production along poor, fair,
and good rankings. We used two dummy variables: fair, and
good, as explanatory variables to test if groundwater quality
matters for KP performance.
(vii) Watercourse command area
We include gross command area of the watercourse to mea-
sure the total agricultural area under the KP management.
(viii) Number of shareholders
We include total number of irrigating farmers in the water-
course community. In our sample, the number of farmers in a
watercourse varies from a minimum of 20 to 1300 farmers.
The average size of the watercourse is 76 farmers, with a gross
command area of 382 acres.
(ix) Leadership quality
We measure leadership quality by the years of education
and age of the current KP chairperson. We also considered
the influence of the location of KP chairperson’s land in the
watercourse by including a dummy variable, which takes a
value of 1 if the KP chairperson’s land was located in the tail
reach of the watercourse. The general hypothesis is that if
those with power are located at the tail end where water deliv-
ery is lowest, there will be more incentive to encourage collec-
tive action (de Klein & Wahaj, 1998; Latif, 2007).
(x) Leadership potential
We also explored leadership potential by including the per-
centage of ‘‘influential persons” in the community. Meinzen-
Dick et al. (2002) found that the greater number of ‘‘influential
persons” in the community positively and significantly
improved collective action. Ul Hassan (2011) however has
documented that large landowners who were benefiting from
the status quo had vigorously opposed the institutional
reforms. Especially as the new roles of the KP may bring it
290 WORLD DEVELOPMENT
to challenge the older power centers, these groups may try to
undermine its effectiveness. We include ‘‘influential persons”
defined as ‘‘individuals generally recognized as arbitrators, com-
munity leaders” in the respective watercourse to examine their
potential role. We also measure leadership potential with per-
centage of persons with college degree and ex-soldiers in the
community. Military service can offer rare educational oppor-
tunities to disadvantaged groups (Angrist, 1998). Given the
low level of education in this area, literacy rates ranging from
30% to 55%, a few years or a career in military service may not
only be a source of human capital but also a reason for higher
standing and influence in the community. 13
(xi) Percentage of woman-headed households
We collected data on the number of female-headed farming
households in the watercourse community and include their
percentage as a measure of the share of potentially excluded
community members. We expect that greater percentage of
women may undermine the performance of KPs.
(xii) Additional variables
We include a dummy variable for perennial canals (66% of
our sample). We also include tehsil-level development indica-
tors (Table 2) to control for the broader development status
including literacy rate (15 years and above), child malnourish-
ment (weight for age, % below 2SD), population using
improved sources of drinking water and sanitation (%),
women reporting that a lady health worker visited the house
during the past month (%) and unemployment rate (%). We
also include dummies for canal divisions, Fordwah, Hakra,
and Sadiqia. The sample contains 529 watercourses from
Fordwah, 451 from Hakra, and 679 from Sadiqia canal divi-
sions.
4. RESULTS
(a) Estimation models and analytic approach
To test the hypotheses outlined in the earlier section, we
employ two models, Ordinary Least Squares (OLS) and Pro-
bit. We use OLS to estimate our single model with a continu-
ous dependent variable, percentage of community
participating in watercourse maintenance and percentage of
community that attends the KP meetings, and Probit model
for remaining three models with binary dependent variables,
water charge collection, effective dispute resolution and voting
in FO elections. Each KP performance measure is an endoge-
nous variable. We estimate the effect of all explanatory vari-
ables on each KP performance measure separately. Table 3
presents the results of the OLS regressions (with robust stan-
dard errors) and Table 4 reports the results for the Probit
regression. For each Probit regression, we use three
approaches to evaluate the fit of the Probit model: McFad-
den’s pseudo R-squared, Hosmer–Lemeshow statistic
(Woolridge, 2001) which compares group-averaged predicted
probabilities with sample frequencies, and comparison based
on classification. Each of these is shown in the bottom rows
of Table 4. Below, we first present the regression results for
each hypothesis and then compare the results for old versus
new KP roles.
(b) Results by hypothesis
In this section we summarize the results around our key
hypotheses of interest. In the next subsection we explore
results by KP responsibilities in a greater depth, comparing
old versus new KP roles. First we give a broad brushstroke
of results for each hypothesis that are generalizable across
all domains of collective action.
(i) Canal water delivery: location, variability, and infrastructure
Each KP responsibility reveals slightly different patterns
with canal water delivery, but we can generalize one overall
result: KPs in the tail reaches of the system (along a dis-
tributary or a main canal) tend to perform no worse, some-
times significantly better, than KPs in the upstream reaches
of the system. Watercourse communities in the tail reaches
have significantly broader participation in maintaining their
watercourses and voting in the FO elections. If water avail-
ability is worse in the tail reach, then our results indicate
that worse canal water delivery improves collective action.
The KP performance indicator for dispute resolution and
collection of water charges exhibit U-shaped patterns, i.e.,
the middle reaches perform worse than the head or the tail
reaches.
Overall, these results are counter to an inverse-U relation-
ship between community performance and water scarcity
found by Bardhan (2000). Could it be that in the presence of
privately accessed groundwater, canal water delivery is only
a partial measure of water availability? To clarify the relation-
ship between water scarcity and KP performance, we examine
the variation in percentage of watercourse area irrigated by
groundwater by location of the watercourse in the irrigation
system and discover that percentage of watercourse command
area irrigated by groundwater is significantly lower if the
watercourse is in the tail reach of the system. We also examine
if groundwater quality declines from head to the tail reach of
the system and find no such pattern. 14 These results then sup-
port our conclusion that total water supplies may be scarcer in
the tail reaches, and that scarce water supplies do not under-
mine KP performance. Unlike the 48 South Indian communi-
ties surveyed by Bardhan (2000) that were sampled from
across public and self-managed irrigation systems and showed
poorer collective action when water scarcity was higher, our
sample watercourse communities are sampled from more sim-
ilar systems. In the tail reaches communities may be increasing
effort, because they perceive a stronger benefit from collective
action.
Increase in variation in canal discharge and infrastructure
conditions are significant determinants of KP performance
across the entire spectrum of responsibilities. The results are
as expected: higher variability undermines and better condi-
tion of canals strengthens KP performance.
Poor state of infrastructure inhibits collective action. KPs
performed poorly in all dimensions of their responsibilities if
their respective communities had reported that their distribu-
tary was in poor condition. The state of the distributary is
out of the direct control of the watercourse community yet
its effect on all downstream channels is highly significant.
Also, if the watercourse is lined, there is significant improve-
ment in the degree of collective action.
(ii) Groundwater productivity and exit
We find strong evidence that, all else equal, heavy ground-
water use and better groundwater quality, in a watercourse
community is negatively associated with effectiveness of the
KP. We find this result for all but two KP responsibilities.
For watercourse maintenance, internal meetings and voting
in FO elections, we found an inverse-U shaped relationship
between the percentage of watercourse command area irri-
gated by groundwater and KP’s effectiveness. At low levels
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN 291

Table 3. Explaining Khal Panchayat (KP) performance-OLS model

(1) (2)
Percentage of watercourse community Percentage of watercourse
contributed labor for maintenance community attends KP meetings
Main head (1 = Yes, 0 = No) 3.687 (3.826) 6.751 ( 4.306)
Main tail (1 = Yes, 0 = No) 1.557 (3.528) 3.431 ( 3.607)
Distributary head (1 = Yes, 0 = No) 1.712 (2.266) 2.344 ( 2.528)
Distributary tail (1 = Yes, 0 = No) 6.803*** (2.602) 0.58 ( 3.037)
Main Tail  Distributary Tail 10.065*** (3.524) 0.559 ( 4.092)
Main Head  Distributary Head 7.101*(3.864) 9.093** (4.536)
Intra-seasonal variability in canal discharge 0.141*** (0.045) 0.002 (0.046)
Inter-seasonal variability in canal discharge 0.486*** (0.173) 0.09 ( 0.185)
Lined watercourse (%) 0.061* (0.033) 0.151*** (0.037)
Physical state of the distributary (1 = Poor, 0 = good) 0.636 (1.832) 4.987** (2.129)
Watercourse area irrigated by gw (%) 0.242*** (0.077) 0.329*** (0.093)
Watercourse command area irrigated by gw squared 0.003*** (0.001) 0.003*** (0.001)
Fair quality groundwater (1 = Yes, 0 = No) 2.581 (1.647) 4.535** (1.974)
Good quality groundwater (1 = Yes, 0 = No) 8.874*** (2.926) 9.607*** (3)
Total farmers sharing the watercourse 0.120*** (0.023) 0.043*** (0.015)
Area of the watercourse 0.005 (0.004) 0.001 (0.005)
Education of KP chairman 0.101 (0.19) 0.592*** (0.219)
Age of KP chairman 0.015 (0.069) 0.1 (0.079)
KP chairman’s farm located in tail of watercourse 12.016*** (1.998) 9.735*** (2.378)
Influential persons (%) 0.405*** (0.101) 0.297*** (0.114)
College graduates (%) 0.024 (0.06) 0.058 (0.071)
Shareholders with a military background (%) 0.205 (0.17) 0.137 (0.176)
Percentage of female-headed households 0.429*** (0.129) 0.031 (0.154)
Percentage of WC in need for repair 0.066*** (0.025)
Number of primary schools in the watercourse community 3.102*** (0.993) 2.693** (1.144)
Perennial canal (1 = Yes, 0 = No) 33.626*** (3.456) 2.965 (3.399)
Literacy rate (%) 9.623*** (3.515) 4.565 (3.696)
Child Malnourishment (%) 1.309** (0.634) 0.713 (0.635)
Improved sources of drinking water and sanitation (%) 5.591** (2.365) 3.45 (2.486)
Lady Health worker (%) 0.932*** (0.249) 0.039 (0.268)
Unemployment rate (%) 14.938*** (4.204) 7.545* (4.542)
Fordwah canal division 4.942 (4.994) 5.42 (5.304)
Sadiqia canal division 0.177 (4.309) 10.249** (4.818)
Constant 79.200*** (29.227) 39.763 (35.771)
Observations 1,845 1,845
R-squared 0.393 0.082
Notes: Robust standard errors in parentheses.
***
p < 0.01.
**
p < 0.05.
*
p < 0.1.

of groundwater use, an increase in share of groundwater irri-
gated is positively associated with collective action, i.e., the
production-enhancing role of groundwater dominates, but at
high levels of groundwater use the exit effect dominates, i.e.,
farmers tend to substitute away from canal water which weak-
ens their incentives to contribute toward watercourse mainte-
nance. Figure 2 shows that, community contributions toward
watercourse maintenance peak when 45–50% of the water-
course area relies on groundwater for agricultural production,
but decline for higher levels of groundwater dependence.
(iii) Size of the community
We found mixed evidence for the effects of the community
size. For watercourse maintenance and internal meetings, we
cannot reject the hypothesis that larger communities tend to
perform worse in achieving better functioning KPs. But the
size of the community (or area of the watercourse) was not
a deterrent to achieving the other KP responsibilities such as
voting, dispute resolution and collecting water charges.
(iv) Leadership quality and potential
Our evidence suggests that the education level of the elected
KP chairman has a significant and positive association with the
KP performance for three of the five KP responsibilities under
study, namely collecting water charges, voting in the FO elec-
tions, and percentage of community participating in the KP
meetings. Education level of the KP chairman had no signifi-
cant effect for community participation for watercourse main-
tenance or dispute resolution. We also find that if the KP
chairperson’s own farm was located in the tail reaches of the
watercourse, KPs were less effective. This was a universal find-
ing in our research. The magnitude of the negative effect of the
location of KP chairperson was larger than the magnitude of
the positive education effects whenever present.
Next, we turn to the effect of various measures of leadership
potential in the community. We find that a greater percentage
of college graduates or ex-military personnel in the water-
course community significantly improved collective action.
We interpret this finding as human capital contribution to
292 WORLD DEVELOPMENT

Table 4. Explaining Khal Panchayat (KP) Performance –Probit Model

Dependent variables are dummy variables (1 = Yes, 0 = No) (1) (2) (3)
KP collecting Community accepts KP’s dispute KP chairman voted
water charges? resolution role? in the FO election?
Main head (1 = Yes, 0 = No) 0.460*** (0.157) 0.568** (0.245) 0.164 (0.198)
Main tail (1 = Yes, 0 = No) 0.114 (0.132) 0.574*** (0.207) 0.008 (0.170)
Distributary head (1 = Yes, 0 = No) 0.189* (0.098) 0.071 (0.137) 0.070 (0.117)
Distributary tail (1 = Yes, 0 = No) 0.164 (0.110) 0.011 (0.162) 0.204 (0.138)
Main Tail  Distributary Tail 0.462*** (0.153) 0.208 (0.219) 0.401** (0.188)
Main Head  Distributary Head 0.226 (0.166) 0.606** (0.291) 0.028 (0.218)
Intra-seasonal variability in canal discharge 0.009***(0.002) 0.001 (0.003) 0.002 (0.002)
Inter-seasonal variability in canal discharge 0.012*(0.007) 0.017*(0.009) 0.014*(0.008)
Lined watercourse (%) 0.001 (0.001) 0.000 (0.002) 0.007*** (0.002)
Physical state of the distributary 0.263*** (0.079) 0.399*** (0.140) 0.014 (0.100)
Watercourse area irrigated by gw (%) 0.004 (0.004) 0.001 (0.005) 0.019*** (0.004)
Watercourse command area irrigated by gw squared 0.000 (0.000) 0.000 (0.000) 0.000*** (0.000)
Fair quality groundwater (1 = Yes, 0 = No) 0.086 (0.077) 0.075 (0.106) 0.496*** (0.090)
Good quality groundwater (1 = Yes, 0 = No) 0.103 (0.111) 0.051 (0.159) 0.298** (0.142)
Total farmers on the watercourse 0.000 (0.001) 0.000 (0.001) 0.000 (0.001)
Gross command area of watercourse 0.000 (0.000) 0.000 (0.000) 0.000 (0.000)
Education of KP chairman 0.024*** (0.008) 0.001 (0.011) 0.034*** (0.010)
Age of KP chairman 0.004 (0.003) 0.003 (0.004) 0.012*** (0.004)
KP chairman’s farm located in tail of watercourse 0.091 (0.092) 0.230* (0.139) 0.255** (0.106)
Influential persons (%) 0.022*** (0.005) 0.005 (0.007) 0.008 (0.006)
College graduates (%) 0.010*** (0.003) 0.005 (0.004) 0.013*** (0.004)
Shareholders with a military background (%) 0.021*** (0.006) 0.056*** (0.019) 0.028** (0.011)
Percentage of female-headed households 0.010* (0.006) 0.011 (0.009) 0.013* (0.008)
Number of primary schools in the watercourse community 0.005 (0.043) 0.067 (0.063) 0.048 (0.053)
Perennial canal (1 = Yes, 0 = No) 0.162 (0.124) 0.023 (0.190) 0.227 (0.151)
Literacy rate (%) 0.078 (0.139) 0.345 (0.248) 0.318* (0.170)
Child malnourishment (%) 0.005 (0.024) 0.047 (0.045) 0.008 (0.030)
Improved sources of drinking water and sanitation (%) 0.022 (0.094) 0.212 (0.161) 0.220* (0.114)
Lady health worker (%) 0.012 (0.010) 0.027 (0.019) 0.008 (0.012)
Unemployment rate (%) 0.212 (0.171) 0.609* (0.312) 0.243 (0.208)
Fordwah canal division 0.156 (0.195) 0.608 (0.391) 0.046 (0.245)
Sadiqia canal division 0.022 (0.175) 0.796** (0.391) 0.117 (0.236)
Constant 3.061** (1.341) 1.920 (2.698) 2.850* (1.617)
Observations 1,845 1,845 1,845
McFadden’s pseudo R–squared 0.1230 0.1205 0.1497
Hosmer–Lemeshow statistic 1861.41 1821.92 2068.39***
Correctly classified 70.35% 91.22% 87.48%
Notes: Standard errors in parentheses
***
p < 0.01.
**
p < 0.05.
*
p < 0.1.

success of local governance. Also noteworthy was the effect of
influential persons in the community. Like earlier studies, we
find that a community that reported a greater percentage of
influential persons tended to demonstrate a greater degree of
collective action. A greater percentage of influential persons
is also positively associated with improved performance in
all dimensions of KP responsibilities except for collecting
water charges.
Taken together, these results are suggestive of the complex
role of leaders and local elites in communities undergoing
institutional change. Choosing a farmer from the tail reach
was seen as a strategy to circumvent ‘‘capture” of the irriga-
tion authority by the local elite, but we found that this strategy
may come at a cost of KP’s effectiveness in conducting its
responsibilities. Even though the KP chairman who farms a
tail-end agricultural plot has a direct private benefit from
maintaining the watercourse, his relatively weaker position
in the community may constrain his organizing influence.
The converse of this result, that a greater percentage of
community-identified elites is beneficial for fostering collective
action, is not always true either, as we note in the case of col-
lecting water charges. Local influentials may impede effective
execution of this responsibility.
(v) Percentage of woman-headed households
A greater percentage of woman-headed households in the
community significantly lowered the percentage of contribu-
tions in watercourse maintenance, suggesting women may be
facing significant barriers to engage in public aspects of water-
course management activities, even when they are heading
their respective farming households. Yet, communities with
greater percentage of female-headed households tended to be
more involved in collecting water charges and voting in FO
elections. We think that results of this explanatory variable
are reflective of patterns of women’s engagement in public,
especially political, aspects of managing local resources.
Women themselves may not participate in KP or higher level
public fora, unless specific arrangements are made for them. It
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN
60
Proportion of watercourse community contributing
58
56
labor for maintenance works
54
52
50
48
46
44
42
40
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Percentage of Watercourse Command Area Irrigated with Groundwater
Figure 2. Estimated effect of groundwater use on labor contributions for watercourse maintenance.
is no accident that all KP members in our sample were men.
As Agarwal (2006) and Kevany and Huisingh (2013) note,
poor and rural women are disproportionately and adversely
impacted by domestic and public gendered roles in relation
to resource accessibility, system design and management, as
well as by the high costs of water distribution. The situation
is unlikely to have changed with the decentralization of the
irrigation system.
(vi) Other results
We also notice that there is no difference in KP performance
in perennial and non-perennial canals, except that perennial
canals are significantly more likely to maintain their water-
courses in winter season when the perennial canals are shut
down for a 4–5 week period.
(c) Collective action in old versus new roles
Do watercourse communities display systematic differences
in how they organize themselves in new roles compared to
the older roles? As mentioned before, watercourse communi-
ties have traditionally been responsible for watercourse main-
tenance and dispute resolution. Irrigation management
transfer carved out three new responsibilities for them: collect-
ing water charges, holding internal meetings and voting in
electing the distributary-level farmer organization members.
One clear pattern that emerges is regarding the role of educa-
tion: Education of the KP chairperson was a significant deter-
minant of successful collective action for all new KP roles. All
else equal, a more educated KP chairperson was more likely to
vote in the FO elections, be engaged in water charge collec-
tion, and hold better attended KP meetings, but the coefficient
of education of the KP chairman was statistically insignificant
for the older roles. This points to the importance of educated
leaders in the community for successful uptake of new rules. It
is interesting to note that even though the percentage of influ-
ential persons in the community was negatively associated
with collection of water charges the magnitude of the positive
effect of KP chairperson’s education was larger than the mag-
nitude of the negative effect of influential persons in the com-
munity.
293
Among the new roles, collection of water charges seems to
have been taken up with least enthusiasm and the voting in
FO members was taken up with most enthusiasm. A majority
of KPs surveyed, 67%, reported not being involved in collecting
water charges but 85% of KP surveyed reported that they voted
in electing their FO representatives—a rate higher than voter
turnout of 55% in the most recent national elections
(Election Commission of Pakistan, 2013). Clearly, the water-
course community sees voting to elect their distributary-level
representative as a highly relevant activity although participat-
ing in collecting water charges is not. The penalties for not pay-
ing water charges by farmers are not fully enforced, and the
institutional reforms have not yet changed that at the KP level.
Second, this may be an outcome of how the new institutions of
FOs and KPs, were ‘‘marketed” to the farmers: as a benefit-
providing rather than rule-enforcing entity (See Table 5).
Table 6 presents the pairwise correlation coefficients among
the KP responsibilities, and while we note that the correlation
is mostly positive, it is not very high. 15 Voting performance
positively correlated with percentage of participation in
monthly meetings: the correlation coefficient is 0.2. But
monthly meetings had a very low correlation with water
charges collection.
Results show that new roles assigned to farmers are more
likely to have been taken up by communities in the tail reaches
of the irrigation system, in particular the collecting water
charges and voting in the FO elections. This is a promising
result for decentralization and devolution of irrigation man-
agement that shows the vibrancy of the farming communities
in accepting and adopting these new roles for farmers.
5. DISCUSSION
Farmers located on watercourses in the tail end of the
irrigation system are generally regarded as farmers with
poorer canal water service, and a concern at the time of irriga-
tion management transfer was that they might be sidelined and
be further deprived by the devolution. Our results provide
evidence contrary to that claim. Our results suggest a wide-
spread interest in the KP activities and that farmers at the tail
294 WORLD DEVELOPMENT

Table 5. Summary of results

Labor cont. for watercourse Dispute resolution Meetings Voting Collecting
maintenance water charges
Canal water supply (head is better than tail) U U
Canal water variability
Condition of infrastructure + + + + +
Groundwater use Inverse-U Inverse-U Inverse-U
Groundwater quality
Size of the community
Education KP chairman + + +
KP chairman’s own land in tail +
% of influentials + +
% ex-military persons + + +
% woman-headed households + +
+ ( ) means positive (negative) and significant effect of the explanatory variable with a P-value of 10% or lower, blank cell means estimated coefficients
were insignificant with P-value of 0.1.

Table 6. Correlation coefficients among KP performance show correlations

Farmers contributing KP involved in Farmers attending Acceptance of KP voted in
labor for maintenance collecting KP meetings (%) KP’s role in farmer organization
works (%) water charges dispute resolution elections
Farmers contributing labor 1.0000
for maintenance works (%)
KP involved in collecting water charges 0.0789 1.0000
Farmers attending KP meetings (%) 0.1725 0.0653 1.0000
Acceptance of KP’s role in dispute resolution 0.0118 0.1166 0.1237 1.0000
KP voted in Farmer Organization elections 0.0416 0.1305 0.2300 0.1652 1.0000

end are showing a greater dynamism in maintaining their
watercourses, being more involved in all roles, especially the
new KP roles. A significantly greater percentage of water-
course communities located in the tail reaches of the system
contributed labor toward maintenance of their watercourses
than those anywhere else in the system. Given that canal water
flow is known to be worse in the tail reaches of the system, this
may indicate that the watercourse communities were compen-
sating for worse canal water service by stepping up their main-
tenance efforts.
While watercourse communities are partially able to over-
come the location disadvantage in terms of self-organizing to
manage their irrigation, other aspects of service delivery, nota-
bly the canal water variability and state of infrastructure is
persistently detrimental to KP activity. The quality of the ser-
vice received determines the range of possible benefits that can
be generated by collective action. In particular, high variability
of water and poor state of the distributary lowered resource
mobilization in the community. A watercourse community
has little or no control over the quality of service delivered
to it, but a community would have little incentive to undertake
the effort to self-organize if the benefit of that effort is per-
ceived as too little. Studies on social capital and development
have shown that state-society synergy is instrumental to mate-
rializing development potentials in various domains of collec-
tive action (Ostrom, Lam, & Pradhan, 2011).
Education level of the KP chairperson was a strong predic-
tor of KP functionality measures. KPs lead by more educated
chairpersons performed significantly better in conducting their
responsibilities, especially the new responsibilities. As a KP’s
functional powers expand, it may step on the toes of existing
power centers within the community who may or may not con-
cede their power. KPs may allow them to capture this new
institution for their benefit or they may try to jeopardize its
viability by not complying. This should be most obvious in
water charge collection. We found that a greater percentage
of influential persons in the community was negatively associ-
ated with KP’s involvement in collection of water charges,
although the magnitude of the positive effect of educated lead-
ers is greater than the negative effect of influential persons.
While we do not have data on the identity of non-payers, it
is possible that if community influentials are not paying water
charges, it creates a sense of immunity from rules in the com-
munity and further reduces payments. 16
A key objective of this study was to examine the role of
groundwater use on farmer participation in surface irrigation
management. Two competing hypotheses existed: one that
groundwater increased the value of surface water, increasing
farmer incentives to collectively invest in canal performance,
and the other that groundwater gives farmers an option to exit
from surface water management. Our results show evidence of
both effects, i.e., that the relationship between groundwater
and farmer participation in surface irrigation management is
non-linear. At low levels of groundwater availability, marginal
increase in groundwater access is associated with higher levels
of farmer participation in surface system management. But
when groundwater use reaches 40–50% of cultivated area,
marginal increase in groundwater access is associated with
lower levels of farmer participation in surface system. This
relationship is shown in Figure 2. The inflection point for
the three KP performance measures that show this inverse-U
relationship comes at a point when groundwater use reaches
40–50% of cultivated area. Also, coefficients of groundwater
quality also support the exit hypothesis: better quality of
groundwater is significantly associated with poorer KP perfor-
mance for watercourse maintenance, KP meetings and voting.
Is high groundwater use in a community a result or a cause
of poor watercourse maintenance? 17 Although groundwater
use in a community may be endogenous, i.e., determined by
farmers decisions, groundwater quality however can be argued
to be exogenous, i.e., determined outside the control of water-
course community, because it is determined by geographic,
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN
hydrologic, and anthropogenic factors spread over an area lar-
ger area than the watercourse command area. Our results
show that better groundwater quality significantly lowers KP
performance in watercourse maintenance, participation in
KP meetings, and voting in the FO elections. This suggests
that we cannot reject the exit hypothesis i.e., access to ground-
water reduces incentives of collective action.
But an overall caveat in interpretation of all results is that
our results establish the strength of association not the direc-
tion of causality. For example, the choice of leaders by water-
course communities could be endogenous, i.e., a community
chooses good leaders because it intends to perform good col-
lective action.
6. CONCLUSION
The decentralization and devolution of irrigation manage-
ment in Pakistan was marked with a great deal of controversy.
From the farmers’ perspective, the administrative and the legal
structure under which Indus Basin Canal systems had oper-
ated had been in existence long before their lifetimes. Critics
of the decentralization pointed out that farmers may not be
ready or even interested in the task. The concern was that
the decentralization and lack of public oversight, no matter
how imperfect, would increase the incentives of abuse and
might even facilitate elite domination of irrigation service
delivery and systematic exclusion of the disadvantaged (Ul
Hassan, 2013; Vander Velde & Tirmizi, 2004).
We reported on farmer participation and the functioning of
tertiary-level farmer organizations, known in Pakistan as Khal
Panchayat (KP), which were formed as a partial solution to
poor irrigation performance in Pakistan. KPs are expected
to maintain their watercourses, collect seasonal water charges
from within the watercourse community, resolve disputes in
the community, meet at least once per month, and play a vot-
ing role in higher level irrigation management organizations.
Our study relies on data collected from canals after the decen-
tralization and devolution had occurred, therefore our results
cannot be interpreted as a study of the impact of decentraliza-
tion on farmer participation in irrigation management. While
KPs form the backbone of the institutional reforms imple-
mented in Pakistan in the 1990s, the methodology adopted
here cannot be used to assess if KPs can manage irrigation sys-
tems better than the remote bureaucracies that they replaced
or would be better at controlling disputes, collecting fees, or
maintaining the system.
We explained the variation in KP’s fulfillment of their tasks
as a function of variety of variables related to surface water
supply and its variability, state of physical infrastructure,
NOTES
1. The farmer representatives in the Area Water Board are elected by the
chairpersons of all KPs within the jurisdiction of the Area Water Board.
Upon election to the Area Water Board, a representative must relinquish
any existing office within the FO (cannot hold dual office).
2. Sediment drawn along with the water from the distributary and main
canals is deposits on the bed and along the sides of the watercourse,
resulting in distortion of the cross-section, and therewith, a reduction of its
carrying capacity of the watercourse. The embankments of the water-
course, especially of unlined and earthen watercourses, need strengthening
to avoid breaches and excessive seepage losses. Furthermore, vegetation in
295
access to groundwater and its quality, size of the watercourse
community, and role of leaders in the community. KP func-
tional status reflects the effective self-organization and collec-
tive action within the watercourse community, and we relied
on the literature on collective action to identify the explana-
tory variables to analyze the variation in functioning of KPs.
Our overall results suggest that canal water delivery matters
in performance of KPs—intra-seasonal variability is especially
detrimental for KP performance, but tail location of a water-
course, typically taken as a proxy for worse canal water sup-
ply, is not a curse in terms of active participation in
watercourse governance. In fact, absolute tail-enders are more
involved in decentralized management. Leadership quality, as
measured by the education of the KP chairperson, was espe-
cially important for new roles of KPs. Higher percentage of
‘‘influentials” is associated with a broader participation in
maintenance works but not with the new responsibilities, par-
ticularly collecting water charges.
The watercourse communities and their KPs have shown
very active participation in voting to elect their members.
Research on common-pool resource has found that self-
governing institutions that allow farmers to design their own
rules to govern themselves are more likely to bring about pro-
ductive relationships and hence better irrigation performance.
Of all the new roles, voting to elect their representatives has
been most enthusiastically taken up by watercourse communi-
ties, indicating the presence of a latent demand for participation
in irrigation management. Hence, KPs with more educated
leaders were more effective in the new roles assigned to them.
Of particular importance is our finding on groundwater.
Groundwater use for agricultural production in the water-
course community can weaken the incentives for resource
mobilization, especially if groundwater use is high and
groundwater quality is good. In this we depart from the find-
ings of Meinzen-Dick et al. (2002) who only found a negative
relationship between groundwater availability and farmer
organization. We show that watercourses with less than 50%
of the area being irrigated by groundwater tend to have more
effective KP performance but higher dependence on ground-
water weakens the incentives for canal irrigation management.
Also, results on groundwater quality show that better quality
groundwater tends to weaken KP performance. Groundwater
is now a major, and sometimes dominant, source of irrigation
supply in Pakistan and much of Asia. The role of groundwater
in a surface-water-based irrigation system needs to be better
understood and while there may not be an agreement on
whether groundwater management should be considered
within the mandate of watercourse or distributary-level farmer
associations, our results indicate that ignoring groundwater
use can reduce their effectiveness.
the watercourse hampers the flow of water, leading to an inadequate
discharge capacity (de Klein & Wahaj, 1998).
3. Initially only ‘‘smallholder” farmers were included though this
condition was later relaxed.
4. Heretofore, these activities were considered ‘‘interference”. As
Meinzen-Dick (1997) notes, even before the institutional reforms, some
farmers were actively participation in the management of their irrigation
systems by enlarging outlets, taking water out of turn, or unauthorized
adjusting of the levels of gates and even channel beds.
296 WORLD DEVELOPMENT
5. According to the Land Revenue Rules, assessment and collection of
water charges (referred to as abiana) was via the appointment of a
Numberdar (or Lambardar) who was responsible for collection of water
charges from landowners within his village. There were instances when the
local lumberdar was seen as an extractive agent and during the institutional
reforms farmers voiced their preference for eliminating this office and
taking over the responsibility of revenue collection.
6. Overall, 2,044 watercourses were surveyed in this study. About 200 of
these were dropped from this analysis because they were very small or all
landowners belonged to the same extended family.
7. KPs often do not keep written records. The best way to measure their
performance is via interviews of the chairperson and the member famers.
One concern was that KP chairperson could strategically misreport KP
performance. Therefore in a random subsample of watercourses, addi-
tional farmers were also interviewed for the KP performance. The results
of the farmers reporting were not significantly different from the KP
chairperson response and team felt confident there was no systematic
misreporting of KP activities by the KP chairperson.
8. Tehsil is a sub-district. Our sample watercourse fall in 2 districts:
Bahawalnagar and Bahawalpur. Watercourses in Bahawalnagar District
were located in 5 different tehsils: Bahawalnagar, Chishtian, Fort Abbas,
Haroonabad and Minchinbad. Watercourses in Bahawalpur district were
in one tehsil, Hasilpur.
9. For example, if the KP chairperson stated that farmers deposited
water charges directly to the irrigation department, which was the norm
prior to the decentralization, then the variable water charges would be 0.
Only if some farmers submitted dues through the KP is the score 1.
10. Membership of distributary-level Farmer Organization (FO) consists
of directly elected farmer representatives from each watercourse. FO is the
forum where complaints and issues can be voiced, and water charges are
deposited. It is formed at the secondary distributary canal level and will be
composed of farmers elected from member KPs. A minimal KP
participation in Farmer Organization was measured by examining whether
the KP chairperson voted in elections for their Farmer Organization
representatives (a onetime election at the beginning of the FO tenure). See
Figure 1 for the detail of the role and composition of the FO.
11. At the time the survey was conducted, KP voted once every three
years to elect its FO representative, which was the length of the tenure of
FOs but it may have to vote annually for approving operational issues
such as budget, works etc.
REFERENCES
Agarwal, B. (1997). Gender, environment, and poverty interlinks:
Regional variations and temporal shifts in rural India, 1971–199.
World Development, 25(1), 23–52.
Agarwal, B. (2006). Gender inequality, cooperation, and environmental
sustainability. In J. Baland, P. Bardhan, & S. Bowles (Eds.), Inequality,
cooperation, and environmental sustainability (pp. 274–313). New
Jersey: Princeton University Press.
Agrawal, A. (2001). Common property institutions and sustainable
governance of resources. World Development, 9(10), 1649–1672.
Angrist, J. (1998). Estimating the labor market impact of voluntary
military service using social security data on military applicants.
Econometrica, 66(2), 249.
Araral, E. Jr., (2009). What explains collective action in the commons?
Theory and evidence from the Philippines. World Development, 37(3),
687–697.
12. High variability in water flow at the distributary has been accepted as
evidence of high uncertainty of canal water availability for the farmer
although no field study (to the knowledge of the authors) has been
conducted to study this difference carefully.
13. As one reviewer has pointed out, it is possible that for some
communities the ‘influential persons’ and ‘ex-military persons’ are the
same individuals in the community, and in that case the correlation
between these variables of leadership potential would be very high and
statistical significance of each measure of leadership potential will be
lowered due to multicollinearity. Even though the correlation is high
between these variables, we find our overall results do not change by
adding different combinations of variables.
14. We regressed percentage of watercourse area irrigated by ground-
water in a watercourse on location dummy variables and groundwater
quality. Results indicate non-linear relationship between percentage of
area irrigated by groundwater in the watercourse and location of the
watercourse in the irrigation system. For example, watercourses along the
head and the tail reaches of the main canal have a significantly smaller
percentage area irrigated by groundwater than those in the middle reaches
of the main canal. Also, watercourses emanating from the head reaches of
their respective distributaries have a greater percentage of area irrigated by
groundwater. Absolute head or tail location was statistically insignificant.
The coefficients are: percent groundwater area irrigated = 129.4 (30.45)***
– 9.34 (3.24)*** Main_Head – 7.32(2.85)*** Main_Tail + 4.62(2.05)***
Distributary_Head 3.02(2.46) Distributary_Tail .057(3.29) Main
Tail  Distributary Tail 5.45(3.63) Main Head  Distributary Head
+ 10.02(1.59)*** groundwater_quality_fair + 21.86(2.39)*** groundwa-
ter_quality_good 0.023(0.004)*** Watercourse_gross command_area.
R-squared = 0.2416. Replacing the dependent variable by groundwater
quality and keeping all the explanatory variables as before shows none of
the location variables are significantly different from zero i.e., reported
groundwater quality does not vary systematically by location. We also
confirm, as expected, that better groundwater quality is strongly and
positively associated with greater percentage area of the watercourse
irrigated with groundwater.
15. The outcome along each dimension of KP responsibility is an
endogenous variable so we do not regress these on each other. But future
research can utilize more structural modeling approaches to understand
the effect of KP roles on each other.
16. Possibility of immunity from punishment for non-payment of water
charges may be lower if the influentials are the ones not paying. We are
grateful for this insight by our anonymous reviewer.
17. We thank an anonymous referee for pointing this out.
Baland, J., & Platteau, J. (1996). Halting degradation of natural resources:
Is there a role for rural communities? Oxford, U.K.: Clarendon Press.
Bardhan, P. (2000). Irrigation and cooperation: An empirical analysis of
48 irrigation communities in South India. Economic Development and
Cultural Change, 48(4), 847–865.
Bardhan, P., & Dayton-Johnson, J. (2002). Unequal irrigators: Hetero-
geneity and commons management in large-scale multivariate research.
In E. Ostrom (Ed.), The drama of the commons (pp. 87–112).
Washington, DC, USA: National Academies Press.
Bardhan, P., & Mookherjee, D. (2008). Decentralization and local
governance in developing countries. A comparative perspective. Cam-
bridge, MA: MIT Press.
Cheema, A., Khalid, L., & Patnam, M. (2008). The geography of poverty:
Evidence from the Punjab. The Lahore Journal of Economics, 13,
163–188.
 COLLECTIVE ACTION IN DECENTRALIZED IRRIGATION SYSTEMS: EVIDENCEFROM PAKISTAN
de Klein, C. H., & Wahaj, R. (1998). Institutional and physical determi-
nants of water management performance at the tertiary level: The
dynamics of watercourse maintenance in the Pakistan Punjab. Interna-
tional Water Management Institute, Report No. R-56.
Election Commission of Pakistan (2013). Turnout Comparison of Election
Result 2008 and 2013. , Available at <http://ecp.gov.pk/Misc/GE-
2013-Graphs/02_na_turnout_comparison.png>.
Frohlich, Norman., & Oppenheimer, Joe. (2000). How people reason
about ethics. In Arthur. Lupia, Mathew. McCubbins, & Samuel.
Popkin (Eds.), Elements of reason: Cognition, choice, and the bounds of
rationality (pp. 85–107). New York: Cambridge University Press.
Fujiie, Masako., Hayami, Yujiro., & Kikuchi, Masao. (2005). The
conditions of collective action for local commons management: The
case of irrigation in the Philippines. Agricultural Economics, 33,
179–189.
Gemma, M., & Tsur, Y. (2007). The stabilization value of groundwater
and conjunctive water management under uncertainty. Applied Eco-
nomic Perspectives and Policy, 29(3), 540–548.
Government of Pakistan (2012). Canal water pricing for irrigation in
Pakistan: Assessment, issues and options. Islamabad, Pakistan: Plan-
ning Commission, Retrieved from <http://www.waterinfo.net.pk/> on
May 13th 2012.
Government of Sindh (1997). The Sindh Irrigation and Drainage Authority
Act, 1997. Provincial Assembly of Sindh- Notification No. PAS/Legisl-
B-8/97. , Karachi. Pakistan.
Government of the Punjab (1997). The Punjab Irrigation and Drainage
Authority Act, 1997. Punjab Act XI of 1997, Provincial Assembly of
Punjab. Notification No. Legis-2(10)/97/91. , Retrieved from <http://
pida.punjab.gov.pk/> on June 30, 2013.
Government of the Punjab (2010). Notification: The Punjab Irrigation and
Drainage Authority (Farmers Organizations) Rules No. OSD/L&WD/
10/2002. Lahore, Pakistan: Provincial Irrigation and Drainage Author-
ity, Retrieved from <http://pida.punjab.gov.pk/> on June 30, 2013.
Haq, A. (2010). Performance of farmers’ organizations in Punjab,
Pakistan: The challenges and way forward. Irrigation and Drainage,
59(1), 23.
Hirschmann, A. O. (1970). Exit, voice and loyalty. Responses to decline in
firms, organizations, and states. Cambridge, MA: Harvard University
Press.
Kevany, Kathleen., & Huisingh, Donald. (2013). A review of progress in
empowerment of women in rural water management decision-making
processes. Journal of Cleaner Production, 60, 53–64.
Krishna, Anirudh. (2001). Moving from the stock of social capital to the
flow of benefits: the role of agency”. World Development, 29(6),
925–943.
Latif, M. (2007). Spatial productivity along a canal irrigation system in
Pakistan. Irrigation and Drainage, 56(5), 509–521.
Latif, M., & Tariq, J. A. (2009). Performance assessment of irrigation
management transfer from government-managed to farmer-managed
irrigation system: A case study. Irrigation and Drainage, 58(3),
275–286.
Meinzen-Dick, R. (1997). Farmer participation in irrigation–20 years of
experience and lessons for the future. Irrigation and Drainage Systems,
11(2), 103–118.
Meinzen-Dick, R., Raju, K. V., & Gulati, A. (2002). What affects
organization and collective action for managing resources? Evidence
from canal irrigation systems in India. World Development, 30(4),
649–666.
Meinzen-Dick, R., & Zwarteveen, M. (1998). Gendered participation in
water management: Issues and illustrations from water users’ associ-
ations in South Asia. Agriculture and Human Values, 15(4), 337–345.
Mishra, A., Ghosh, S., Nanda, P., & Kumar, A. (2011). Assessing the
impact of rehabilitation and irrigation management transfer in minor
irrigation projects in Orissa, India: A case study. Irrigation and
Drainage, 60(1), 42–56.
Murgai, R., Ali, M., & Byerlee, D. (2001). Productivity growth and
sustainability in post–Green Revolution agriculture: the case of the
Indian and Pakistan Punjabs. The World Bank Research Observer, 16
(2), 199–218.
Mustafa, D. (2001). Colonial law, contemporary water issues in Pakistan.
Political Geography, 20(7), 817–837.
297
Olson, M. (1965). The logic of collective action. Cambridge, MA: Harvard
University Press.
Ostrom, E. (2011). Reflections on ‘‘Some Unsettled Problems of Irriga-
tion”. The American Economic Review, 101, 49–63.
Ostrom, E. (2005). Collective action theory. In C. Boix, & S. Stokes (Eds.),
Oxford handbook of comparative politics. Oxford, U.K.: Oxford
University Press.
Ostrom, E., Lam, W. F., & Pradhan, P. (2011). Improving irrigation in
Asia: Sustainable performance of an innovative intervention in Nepal.
Edward Elgar Publishing.
Pakistan Bureau of Statistics (2012). Agricultural statistics of Pakistan
2011. Islamabad, Pakistan: Pakistan Bureau of Statistics, Retrieved
from <http://www.pbs.gov.pk> on March 18th 2014.
Provencher, B., & Burt, O. (1993). The externalities associated with the
common property exploitation of groundwater. Journal of Environ-
mental Economics and Management, 24, 139–158.
Ramesh, M. (2013). Decentralization in Asia: A survey. Policy and
Society, 32(1), 1–5.
Raza, M. A., Ashfaq, M., Zafar, M. I., & Baig, I. A. (2009). Impact
assessment of institutional reforms in irrigation sector on rice
productivity: A case study of Punjab, Pakistan. Pakistan Journal of
Life and Social Sciences, 7(1), 16–20.
Rinaudo, J. D. (2002). Corruption and allocation of water: The case of
public irrigation in Pakistan. Water Policy, 4(5), 405–422.
Sandler, T. (1992). Collective action: Theory and application. Ann Arbor:
University of Michigan Press.
Senanayake, N., Mukherji, A., & Giordano, M. (2015). Re-visiting
what we know about irrigation management transfer: A
review of the evidence. Agricultural Water Management, 149,
175–186.
Shah, T. (2008). Taming the anarchy: Groundwater governance in South
Asia. Washington D.C.: RFF Press.
Shah, T., Ul-Hassan, M., Khattak, M. Z., Banerji, P. S., Singh, O. P., &
Rehman, S. U. (2009). Is irrigation water free? A reality check in the
Indo-Gangetic Basin. World Development, 37(2), 422–434.
Shivakoti Ganesh, P., & Elinor, Ostrom (Eds.) (2002). Improving irrigation
governance and management in Nepal. Oakland, California, USA: ICS
Press.
Shyamsunder, P. (2008). Decentralization, devolution and collective
action-a review of international experience. In Rucha Ghate, Narpat
Jhoda, & Pranab Mukhopadhay (Eds.), Promise trust and evolution:
Managing the commons of South Asia. USA: Oxford University
Press.
Ul Hassan, M. M. (2011). Analyzing governance reforms in irrigation:
Central, South and West Asian experience. Irrigation and Drainage, 60
(2), 151–162.
Ul Hassan, M. M. (2013). Credit conditionality and strategic sabotage:
The tale of first decade of Pakistan’s irrigation reforms. In C. G.
Goodrich, S. Janakarajan, Anjal Prakash, & Seroshi Singh (Eds.),
Water Resources Policies in South Asia (pp. 170–187). India: Rout-
ledge.
Uphoff, N. T. (1986). Improving international irrigation management with
farmer participation: Getting the process right. Boulder, CO: Westview
Press.
Uphoff, N. T., Wickramasinghe, M., & Wijayaratna, C. (1990). ‘‘Opti-
mum” participation in irrigation management: issues and evidence
from Sri Lanka. Human Organization, 49(1), 26–40.
Van Halsema, G. E., & Vincent, L. (2006). Of flumes, modules and
barrels: the failure of irrigation institutions and technologies to achieve
equitable water control in the Indus Basin. In T. Tvedt, & E.
Jakobsson (Eds.), A history of water. London, New York: I.B. Tauris
& Co. Ltd.
Vander Velde, Edward, & Tirmizi, J. (2004). Irrigation policy reforms in
Pakistan: Who’s getting the process right? In P. P. Mollinga, & A.
Bolding (Eds.), The politics of irrigation reform: Contested policy
formulation and implementation in Asia, Africa and Latin America
(pp. 207–239). Aldershot, UK: Ashgate.
Vandersypen, K., Bastiaens, L., Traore, A., Diakin, B., Raes, D., &
Jamin, J. Y. (2008). Farmers’ motivation for collective action in
Irrigation: A statistical approach applied to the Office du Niger in
Mali. Irrigation and Drainage, 57(2), 139–150.
298 WORLD DEVELOPMENT

Wade, R. (1988). Village republics: Economic conditions for collective
action in South India. Cambridge, UK: Cambridge University Press.
Woolridge, J. (2001). Econometric analysis of cross section and panel data.
The MIT Press.
World Bank (1994). Pakistan irrigation and drainage: Issues and options.
Washington D.C: World Bank.
Yu, Winston., Yang, Yi.-Chen., Savitsky, Andre., Alford, Donald.,
Brown, Casey., Wescoat, James., ... Robinson, Sherman. (2013). The
Indus Basin of Pakistan: The impacts of climate risks on water and
agriculture. Washington, DC: World Bank.

Available online at www.sciencedirect.com

ScienceDirect