EPTD DISCUSSION PAPER NO. 68
OF DRYLAND WATERSHED DEVELOPMENT
PROJECTS IN INDIA
John Kerr, with Ganesh Pangare,
Vasudha Lokur Pangare, and P.J. George
Environment and Production Technology Division
International Food Policy Research Institute
2033 K Street, N.W.
Washington, D.C. 20006 U.S.A.
EPTD Discussion Papers contain preliminary material and research results, and are circulated
prior to a full peer review in order to stimulate discussion and critical comment. It is expected that most
Discussion Papers will eventually be published in some other form, and that their content may also be
The author expresses thanks to Peter Hazell, Derek Byerlee, John Pender, Anna
Hazare, Shashi Kolavalli, G.B. Singh, Dayanatha Jha, and numerous investigators and
respondents who contributed to the study. An earlier, abridged version of this study was
published in Kerr et al. (2000).
India's semi-arid tropical (SAT) region is characterized by seasonally
concentrated rainfall, low agricultural productivity, degraded natural resources, and
substantial human poverty. The green revolution that transformed agriculture elsewhere
in India had little impact on rainfed agriculture in the SAT. In the 1980s and 1990s,
agricultural scientists and planners aimed to promote rainfed agricultural development
through watershed development. A watershed is an area from which all water drains to a
common point, making it an attractive unit for technical efforts to manage water and soil
resources for production and conservation.
Watershed projects are complicated, however, by the fact that watershed
boundaries rarely correspond to human-defined boundaries. Also, watershed projects often
distribute costs and benefits unevenly, with costs incurred disproportionately upstream,
typically among poorer residents, and benefits realized disproportionately downstream,
where irrigation is concentrated and the wealthiest farmers own most of the land.
Watershed projects take a wide variety of strategies, ranging from those that are
more technocratic to those that pay more attention to the social organization of
watersheds. By the mid-1990s annual expenditure on watershed development in India
approached $500 million, but there was relatively little information available on the
success of different project approaches.
This study addresses three main research questions: 1) What projects are most
successful in promoting the objectives of raising agricultural productivity, improving
natural resource management and reducing poverty? 2) What approaches enable them to
succeed? 3) What nonproject factors also contribute to achieving these objectives? The
major hypotheses are that participatory approaches that devote more attention to social
organization yield superior project impact, and that favorable economic conditions and
good infrastructure also support better natural resource management and higher
A detailed survey of Maharashtra and Andhra Pradesh states covered 86 villages
under several watershed projects as well as nonproject villages with no project. The
projects covered operated under the Ministry of Agriculture, the Ministry of Rural
Development, various nongovernment organizations (NGOs), and in collaboration between
NGOs and the Government of Maharashtra. The government projects were more
technocratic in focus, whereas the NGO projects focused more on social organization, and
the government-nongovernment collaborative projects tried to draw on the strengths of
The analysis of the Maharashtra and Andhra Pradesh villages compared pre- and
post-project conditions in the study villages. Quantitative analysis at the village level
addressed performance indicators such as changes in access to water for irrigation and
drinking, change in employment opportunities, soil erosion and conservation on
uncultivated lands and drainage lines, and change in availability of various products from
the common (government revenue) lands. At the plot level, performance indicators
included changes in cropping intensity, change in yields, soil erosion on cultivated lands,
farmers' land improvement investments, and annual net returns to cultivation. This
analysis was supplemented by qualitative information about the effects of the projects on
different interest groups in the villages such as farmers with irrigation, farmers without
irrigation, landless people, shepherds, and women.
Findings of the empirical study in Maharashtra and Andhra Pradesh lend support
to the hypothesis that more participatory projects perform better than their more
technocratic, top-down counterparts, and that a combination of participation and sound
technical input may perform the best of all. Evidence about the role of economic
conditions and infrastructure is more limited.
Despite rhetoric to the contrary, successful participatory projects remain few in
number so their impact is limited. In the study area in rainfed areas of Maharashtra's
Pune and Ahmednagar districts, for example, the innovative projects operated in only 40
out of over 1000 villages, even though they are particularly highly concentrated in this
area compared to the rest of India. Also, the most successful projects enjoyed special
treatment that will be difficult to replicate on a large scale. Spreading participatory
watershed development throughout the country will not be easy.
One continuing challenge for almost all projects is in designing interventions and
organizing communities so that benefits are distributed more evenly to landless people,
shepherds and women. These are the least influential community members and their
needs and interests require special attention. Otherwise watershed projects can actually
make them worse off than before by restricting their access to resources that contribute to
their livelihoods. Unstructured interviews with these groups suggested that all of the
Maharashtra projects have room for improvement in serving their needs. Some NGOs in
Andhra Pradesh have developed innovative ways to build everyone's interests into the
projects in advance, and other projects would gain by learning from them.
1. Introduction........................................................................................................ 1
O outline of the Paper ........................................................................................... 2
Conceptual Framework ............................................................................. 3
Watershed Management as a Social Organization Problem.............................. 3
How Economic Forces Can Determine Project Outcomes................................ 5
A analytical Approach .......................................................................................... 7
Selection Criteria under each Project ....... ........................................................ 8
Project Categories Covered in the Analysis ...................................... ............ 8
2. Approaches to Watershed Development in Maharashtra and Andhra Pradesh......... 10
Agroclimatic Conditions in Maharashtra and Andhra Pradesh............................ 10
How Different Projects Approach Raising Agricultural Productivity .................. 12
Projects that Focus on Increasing Irrigation.............................................. 12
Projects that Focus on Rainfed Agriculture................................ .......... .... 13
Projects Covered Under this Study ............................................................... 14
Government Projects that Focus Primarily on Water Harvesting.................. 14
Government Projects that Focus Primarily on Rainfed Agriculture................. 17
Nongovernment Organizations: A Focus on Social Organization................. 20
Government-NGO Collaborative Programs in Maharashtra........................... 23
Investment Costs per Hectare Under Each Project ................................................ 26
3. D ata .............................................................................................................. . 27
Performance Indicators.................................................. ................................. 28
Determinants of Project Performance ......................................... ........... .. 30
Sam pling................................................................................................... 31
Characteristics of the Sampled Villages and Plots ............................................. 33
Village Characteristics .................................................. .......................... 33
Plot Characteristics................................................. ................................. 38
Characteristics of Household and Interest Group Respondents.............................. 42
4. M methods ........................................................................................................... 43
Assessing Endogeneity in Program Placement.................................. ........... ... 44
Qualitative A nalysis..................................................................................... 48
5. How Projects Choose Where to Operate........................................... ........... ... 48
Project Site-Selection Guidelines .................................................................. 48
Comments on the Site Selection Criteria.................................... ........... 50
Analysis of Determinants of Program Placement............................................ 51
Econometric Analysis of the Determinants of Project Placement................... 55
6. Natural Resource Management and Productivity on Uncultivated Lands ................ 58
W hat the projects do......................................................................................... 59
Social Fencing Institutions ............................................................................. 60
Erosion and Conservation Status of the Main Drainage Line .............................. 62
A ll villages....................................................................................................... 64
Erosion of Uncultivated Lands ....................................................................... 67
Change in Availability of Fuel and Fodder from the Common (Government
Revenue) Lands....................................................................................... 71
7. Promoting Irrigation Development............................................................... 80
Changes in Irrigation........................................................................................ 81
Changes in Irrigated Area at the Village Level ............................................ 81
Changes in Cropping Intensity through Increased Irrigation ........................ 83
Respondents' Perceptions of Projects' Effects on Irrigation Development.............. 84
8. Natural Resource Management and Productivity of Rainfed Agricultural Land ...... 87
Project Subsidies to Participants Under Each Project........................... ........... 88
Subsidy Policy and Practice under each Project........................................... 89
Interaction Between Project Staff and Survey Respondents.................................. 91
Adoption of Soil and Water Conservation Practices........................... .......... .. 93
Agronomic Practices ............................................................................... 93
Investment and Maintenance of Soil and Water Conservation Structures......... 96
Use of Credit for Land Improvement Investments ......................................... 99
Maintenance of Soil and Water Conservation Assets ................................... 100
Net Returns to Cultivation.................................................................................... 102
9. C conclusion ........................................................................................................... 106
Evidence of Project Performance.......................................................................... 107
Participatory Projects Perform the Best ....................................................... 107
Factors that Enable Participatory Projects to Perform Better........................ 108
The Role of Infrastructure ............................................................................. 113
Additional Issues for the Future........... ................................................................ 115
Monitoring and Evaluation............................................................................ 115
A Call for Caution in Watershed Investments.............................................. 117
R references ............................................................................................................ 120
AN EVALUATION OF DRYLAND WATERSHED DEVELOPMENT
PROJECTS IN INDIA
John Kerr, with Ganesh Pangare, Vasudha Lokur Pangare, and P.J. George
Rainfed agriculture in India's semi-arid tropics (SAT) is characterized by low
productivity, degraded natural resources, and widespread poverty. Most of the hundreds
of millions of people living in the Indian SAT depend on agriculture and natural resource
management for their livelihoods, so development planners are eager to implement
productive, environmentally sustainable land and water management systems.
Watershed development projects are designed to harmonize the use of water, soil,
forest and pasture resources in a way that conserves these resources while raising
agricultural productivity, both through in situ moisture conservation and increased
irrigation through tank- and aquifer-based water harvesting. Watershed projects have
become widespread in rainfed areas in recent years, with a current annual budget from all
sources that exceeds US $500 million (Farrington et al. 1999). This study examines the
experience of watershed projects in Andhra Pradesh and Maharashtra.'
The literature on watershed development in India is growing rapidly, but most of
it is confined to qualitative descriptions of success stories. Some of these contain
excellent insights into the social processes that contribute to successful watershed
development, but there is little frank discussion of less successful projects. The few
quantitative studies available tend to be based on a small number of heavily supervised
projects, with no information about long-term impacts. Benefits after the first year or two
This study was originally conducted under the Indian Rainfed Agricultural
Research and Development Project, jointly sponsored by the World Bank and the Indian
Council of Agricultural Research. That project also included a companion study on
watershed development that reviewed the literature on watershed projects and drew upon
the findings of a rapid rural appraisal of new projects in Karnataka, Rajasthan, and Orissa
were typically assumed and, not surprisingly, cost-benefit findings were almost always
favorable. At the same time, the vast majority of projects were never subject to
evaluation and there were good reasons to suspect that most of them had little impact
(Kerr and Sanghi 1992).
With this background, the current research was commissioned to analyze the
determinants of agricultural productivity, natural resource management and poverty
alleviation under a wide range of watershed projects. The study is mainly quantitative
but also incorporates qualitative data, explicitly examining the effects of non-project
factors such as infrastructure, access to markets, social institutions in the villages,
agroecological conditions, etc. This broad framework not only controls for the effects of
these factors but also enables identification of other policy-relevant determinants of
improved natural resource management and economic development. It also discusses the
approaches taken by different projects in order to understand the essential elements of
successful projects and make recommendations for the future. To summarize, the study
addresses three related questions: 1) which projects perform the best, 2) what approaches
enable them to succeed, and 3) what additional characteristics of particular villages
contribute to achieving the objectives of improved natural resource management, higher
agricultural productivity and reduced poverty.
OUTLINE OF THE PAPER
After introducing the problem and presenting a conceptual framework in Section
1, Section 2 describes the broad approaches to watershed development in Maharashtra
and Andhra Pradesh and introduces the specific projects operating there. Section 3
describes the data on which this study is based, and Section 4 presents the analytical
model. Characteristics of villages in which each project operates are analyzed in Section
5, while Sections 6-8 analyze project performance in terms of achieving various
objectives related to agricultural productivity, natural resource management and poverty
alleviation. These include the work conducted by watershed projects on protecting and
developing nonarable land, recharging groundwater, improving the management of
agricultural land, and raising agricultural production. These sections also analyze the role
of nonproject factors, such as infrastructure development, on outcomes of interest, and
they examine watershed project activities in relation to villagers' development priorities.
Section 9 concludes with policy implications and recommendations.
Two main hypotheses guided this research. One is that watershed projects cannot
succeed without full participation of project beneficiaries and careful attention to social
organization. This is because the costs and benefits of watershed interventions are
location-specific and unevenly distributed among the people affected. The second
hypothesis is that a variety of factors determine the incentives for people to manage and
protect natural resources and invest in increased agricultural productivity. These factors
may have as great an impact as a watershed project in determining the outcomes that
projects seek to achieve. The issues underlying these two hypotheses are explained next.
Watershed Management as a Social Organization Problem
A watershed (or catchment) is a geographic area that drains to a common point,
which makes it an attractive unit for technical efforts to conserve soil and maximize the
utilization of surface and subsurface water for crop production. A watershed is also an
area that contains administrative and property boundaries, lands that fall under different
property regimes, and farmers whose actions may affect each others' interests. Human-
defined boundaries, however, normally do not match biophysical ones. In watershed
management projects, mechanical or vegetative structures are installed across gullies and
rills and along contour lines, and areas are earmarked for particular land use based on
their land capability classification. Cultivable areas are put under crops according to
strict principles of contour-based cultivation. Erosion-prone, less favorable lands are put
under perennial vegetation. This approach aims to optimize moisture retention and
reduce soil erosion, thus maximizing productivity and minimizing land degradation.
Improved moisture management increases the productivity of improved seeds and
fertilizer, so conservation and productivity-enhancing measures are complementary.
Excess surface runoff water is harvested in irrigation or percolation tanks while
subsurface drainage recharges groundwater aquifers, so conservation measures in the
upper watershed have a positive impact on productivity in the lower watershed.
Reducing erosion in the upper reaches also helps to reduce sedimentation of irrigation
tanks (ponds) in the lower reaches. The watershed approach enables planners to
internalize such externalities and other linkages among agricultural and related activities
by accounting for all types of land uses in all locations and seasons. This systems-based
approach is what distinguishes watershed management from earlier plot-based
approaches to soil and water management.
Socioeconomic relationships among people in a watershed can complicate efforts
to introduce seemingly straightforward technical improvements. This is because, as
mentioned above, a watershed contains multiple decision-makers whom watershed
development affects unequally. When a watershed project is introduced, often the bulk
of the work is done in the upper reaches while the benefits accrue primarily in the lower
reaches. For example, revegetating the upper reaches involves banning grazing and
felling trees so that plants can establish. As a result, the people who utilize the upper
watershed-typically relatively poor people with little or no land-bear the brunt of the
costs of watershed development, which mainly benefits wealthier farmers in the lower
watershed. Those who are made worse off by a watershed project can undermine its
efforts if they refuse to go along with it. Herders, for example, might refuse to abide by
grazing bans and trespass on the common lands if they are able to. In general, watershed
technologies are likely to fail if they divide benefits unevenly but require near-universal
cooperation to make them work. In this case, equity becomes a prerequisite to efficiency
(Kerr and Sanghi 1992).
While early watershed projects failed to recognize the socioeconomic dimensions
of watershed development, this has changed significantly in the last decade. In recent
years there has been a growing appreciation of the need to organize communities to work
collectively, make sure that beneficiaries have an interest in the work that is done, and
ensure that everyone benefits from the project. In the 1990s, every project was designed
to include the "participation" of local people; however, they all defined "participation"
differently. For government programs, typically it meant making the effort to convince
people of the soundness of an approach that was essentially pre-designed without any
input from those who would be affected. Taking people's involvement a step further, in
such projects local committees were established to mobilize laborers for moving earth
and planting vegetation, and to facilitate communication within the village to improve the
management of common lands. On the other extreme, many new projects operate under
the assumption that local people know best how to care for their land and simply need
outside assistance to help them organize and gain access to resources, including funds
and social services.
Approaches to participation are discussed in detail in Section 2, and implications
of alternate approaches for project outcomes are revealed by the analytical findings
presented in Sections 6-8. Based on these findings and various observations from the
field, recommendations for how projects should pursue participation in the future are
presented in Section 9.
How Economic Forces Can Determine Project Outcomes
As mentioned above, performance in improving agricultural production, natural
resource management, and human welfare depends on economic factors beyond the
control of a watershed project. Throughout the world, both today and historically, it is
easy to find areas with a broad range of performance in agricultural growth, natural
resource management and poverty alleviation. For example, evidence abounds of areas
in India with stagnant agricultural production, low real incomes, and environmental
degradation. On the other hand, both the literature and folk wisdom are full of examples
of places in India where villagers manage their natural resources particularly well and the
local economy is unusually vibrant. What determines why some areas are more
productive than others?
Induced innovation theory helps explain the conditions under which agricultural
development will take place along paths that degrade or conserve natural resources.
Induced innovation theory holds that, over time, technological innovations and
institutional changes take place to economize on scarce resources and utilize abundant
ones (Hayami and Ruttan 1984). The theory helps explain why traditional farming
systems have evolved differently in different places. For example, in sparsely populated
areas traditional farming systems were bush-fallow, with forest land being cleared and
farmed for a few years before being left for 20 to 30 years of nutrient-restoring fallow.
On the other hand, in land-scarce areas such as the intensive rice growing areas of
Southeast Asia, elaborate terraces, irrigation systems and nutrient management systems
enabled continuous cultivation without degradation. In the widely cited case of
Machakos, Kenya (Tiffen et al. 1994), rising population density, good access to markets
and off-farm income created incentives and provided resources to raise productivity and
conserve natural resources.
In India, farmer-led agricultural intensification is also widespread. In semi-arid
areas the most obvious example is that of private irrigation investments, which are
typically accompanied by land leveling and application of substantial organic matter and
commercial inputs. On rainfed lands the successes are less dramatic, but evidence shows
that private tree planting has grown steadily in recent years (Chambers et al. 1989), and
that many farmers invest in indigenous soil and water conservation measures
independently of special project efforts (Kerr and Sanghi 1992). Likewise, some villages
have designed social institutions for managing common property resources in ways that
raise their productivity and protect against long-term resource degradation (Wade 1988).
Several exceptional case studies of successful watershed development have been
well-publicized in India, but the common perception is that they remain just that:
exceptional. Success is often attributed to the efforts of a charismatic leader or some other
set of social conditions that would be difficult or impossible to replicate on a wide scale.
There is undoubtedly a great deal of truth in this perception, but to date there has been little
systematic effort to examine the extent to which policy-relevant factors have played a role
in causing some areas to be characterized by better resource management and higher
agricultural production than others. Leaving aside unusual success stories like Ralegan
Siddhi (Hazare et al. 1996) and Sukhomajri (Chopra et al. 1990; Patel-Weynand 1997), are
there village-level or regional differences in natural resource conditions, agricultural
productivity and household incomes that can be explained by induced innovation theory?
From the induced innovation perspective, assessing the performance of watershed
development projects requires examining the effects of such factors as market access,
population density and the economic policy environment. Induced innovation theory
suggests that if market access is favorable and population density is high, people will be
more receptive to projects seeking to conserve soil resources and intensify agricultural
production. In fact, even in the absence of a special project, the economic environment
may be sufficient to induce farmers to adopt resource-conserving, productivity-enhancing
technologies. On the other hand, even a well-designed watershed development project
might be unable to achieve long-term success if enabling conditions are lacking. In such
a case, farmers would have insufficient motivation to adopt and maintain practices
needed to promote sustainable agricultural intensification.
This study examines performance in improving agricultural productivity, natural
resource management, and human welfare. Data on performance indicators, which are
described in subsequent sections, come from a survey of 86 villages reflecting a variety of
project approaches, including villages with no project. Quantitative data collected at the
village, plot and household level provide the basis for econometric analysis of the
determinants of changes in pre- and post-project conditions. Open-ended discussions
provide further qualitative information on the impact of projects on people from various
interest groups, such as farmers with and without irrigation, livestock herders, etc.
This research was originally designed to examine only completed projects where
the staff had withdrawn. However, despite the large literature on watershed development
in India, the number of projects in which work has actually been completed is quite
small, so the intended approach was not feasible. Instead, the study covers mainly well-
established projects, with a few that have been completed.
Selection Criteria under each Project
The criterion by which each project selects participating villages is of critical
importance to the present analysis. If, as argued above, numerous factors can determine a
village's performance in agricultural production and natural resource management, then it
is important to know how these factors are distributed across villages in different project
categories. Otherwise, if villages in different project categories vary in their endowment of
factors that can affect performance, then it is difficult to know whether to attribute
differences in performance to project activities or to the effects of pre-existing village
characteristics. For example, Pitt et al. (1993) describe a case in Indonesia that showed that
villages covered for several years under a major family planning program actually had
higher fertility rates than those outside of the program. One could jump to the conclusion
that the family planning program had failed miserably, but Pitt et al. explain that the
difference was not surprising given that the program consciously worked in villages where
fertility had been higher to begin with. In the absence of the family planning program, the
difference in fertility between the two sets of villages might have been even greater.
An analogous situation could apply in the present study since programs may
choose to operate in particularly favorable or unfavorable villages, either intentionally or
Section 2 describes each project in detail, including its rules for selecting villages.
In Section 5 the data are analyzed to assess the extent to which different projects adhere
to their published guidelines and to identify any other factors that may characterize
villages under each category.
Project Categories Covered in the Analysis
All categories of projects operating in Andhra Pradesh and Maharashtra are
covered by this research. They include the following:
Ministry ofAgriculture (MOA): projects that focus primarily on technical
aspects of developing rainfed agriculture. These include the National
Watershed Development Project for Rainfed Areas (NWDPRA), the Indian
Council of Agricultural Research's Model Watershed Projects, and the World
Bank-assisted Pilot Project for Watershed Development in Rainfed Areas.2
Ministry of Rural Development (MORD): Engineering-oriented projects that
focus on water harvesting through construction of percolation tanks, contour
bunds, and other structures. These fall under the Maharashtra Department of
Soil and Water Conservation projects (Jal Sandharan) and the Drought Prone
Area Project (DPAP).3
Non-government organizations (NGOs): projects that typically place greater
emphasis on social organization and less on technology relative to the
NGO-Government collaboration: projects operated jointly by government and
non-government organizations (Indo-German Watershed Development
Programme (IGWDP), Adarsh Gaon Yojana (AGY)) that seek to combine the
technical approach of government projects with the NGOs' orientation toward
social organization. These projects are found in Maharashtra but not Andhra
Control: villages with no watershed project.
All of these project categories are discussed in detail in the next section.
2 The more recent World Bank-assisted Integrated Watershed Development
Project (IWDP) did not operate in either Maharashtra or Andhra Pradesh, so it is not
covered in the quantitative analysis. This paper draws on other analysis of that project,
including a companion to this study by Kolavalli (1998), to discuss this later generation
of World Bank watershed projects.
3 In 1995, the DPAP guidelines were restructured under radical new, participatory
guidelines. However, only pre-reform DPAP projects are included in the quantitative
research since little progress had been made in implementing the new guidelines at the
time of the fieldwork for this research. Other studies including Kolavalli (1998) and
Farrington et al. (1999) help provide information about this more recent set of projects.
2. APPROACHES TO WATERSHED DEVELOPMENT IN MAHARASHTRA
AND ANDHRA PRADESH
This section begins by describing the agroclimatic characteristics of the study
region and characterizing two alternative technical approaches that have been used in
watershed development. It then describes each of the projects covered in this study,
focusing on their guiding principles and the relative emphasis on social organization
compared to technical assistance. There is also a discussion of how each project selects
the sites where it works, the amount of money they invest, and their policies regarding
cost-sharing with intended beneficiaries.
AGROCLIMATIC CONDITIONS IN MAHARASHTRA AND ANDHRA PRADESH
Maharashtra and Andhra Pradesh both have highly diverse agroclimates. In
Maharashtra, a narrow coastal plain separates the Arabian Sea from the Western Ghat
Mountains. On the eastern side of the mountains, the majority of the state is spanned by
the large Deccan Plateau, which covers much of south-central India. Rainfall is very high
in the coastal mountains, but the western part of the Deccan Plateau (in the rain shadow
of the Ghats) is very dry. The wettest district of the Western Ghats receives an annual
average rainfall of over 4000 mm, while the driest areas of the rain shadow zone (only
about 150 km to the east) receive about 500 mm. The topography of this transitional
zone from wet to dry is a series of tablelands, or flat plateaus that drop sharply to plains
below. Conditions for rainfed agriculture in the driest zones are difficult, and this is
where watershed projects are most concentrated. Moving toward eastern Maharashtra,
average annual rainfall rises gradually to over 1000 mm, making conditions for rainfed
agriculture quite favorable.
Figure 1: Map of Maharashtra and Andhra Pradesh States, India
I I --
* State capital
) Study area
Andhra Pradesh contains similar diversity. The long coastal plain along the Bay
of Bengal receives over 1000 mm average annual rainfall, and much of it is irrigated by
the major canal systems of the Krishna and Godavari rivers. Moving west from the coast
and over the Eastern Ghats (which are much smaller than the western Ghats), inland areas
on the Deccan Plateau are divided into the Rayalseema and Telengana regions.
Rayalseema is the southernmost part of the state; it is highly drought prone with average
annual rainfall as low as 500 mm in some areas. South Telengana (which is south of
Hyderabad) is also drought prone, though not to the same extent, with average annual
rainfall in the 600-700 mm range. Both Rayalseema and south Telengana vary in their
topography, with small hills and valleys that are suitable for traditional irrigation tanks
that capture runoff from rainfall for lowland irrigation. The predominantly red soils of
these regions also favor tank irrigation. North Telengana (which is north of Hyderabad),
on the other hand, is flatter, has black soils, and receives around 800-1000 mm average
annual rainfall. Conditions are much better for rainfed agriculture, comparable to the
conditions across the state border in eastern Maharashtra.
Thus, rainfed agriculture in both states varies between areas of high and low
potential and this heterogeneity has important implications for the approaches to
HOW DIFFERENT PROJECTS APPROACH RAISING AGRICULTURAL
There are fundamental differences between watershed projects that focus on
developing rainfed agriculture and those that focus on increasing access to irrigation.
Projects that Focus on Increasing Irrigation
In western Maharashtra, the scarcity of water and favorable topography make
water harvesting a high priority and the focus of most projects. Where plateaus slope
down to the plains, there are many opportunities to capture water behind small dams for
irrigation in the flat lands below. Soils in these areas are relatively porous and favor
percolation of harvested water into groundwater aquifers; it must be pumped for use as
irrigation. By contrast, in Telengana and Rayalseema regions of interior Andhra Pradesh,
irrigation tanks store water on the surface for irrigation by gravity.
Agricultural engineering to build and protect water-harvesting structures is the
key feature of most watershed projects in western Maharashtra. The structures include
mainly check dams in drainage lines and continuous contour trenches in the uncultivated
catchment areas. Since almost all the structures are built on nonarable lands with
common access by all village inhabitants, the projects also promote collective action to
protect vegetation in the catchment area. This reduces erosion and limits the silting that
would reduce the storage capacity of water harvesting structures.
In these projects there is relatively little focus on plot-level management. This is
because once irrigation is in place, farmers have sufficient knowledge and incentive to
manage a plot and improve its productivity. Rainfed agriculture is a low priority where
projects are successful in increasing irrigated area. For example, Shri Anna Hazare,
known as the "father" of watershed development in the well known success story of
Ralegan Siddhi, explained that watershed efforts there focus exclusively on increasing
irrigation and protecting nonarable lands. Virtually no attention is paid to developing
rainfed agriculture (personal communication 1996). This approach has proven highly
successful in Ralegan Siddhi, where irrigated area went from virtually zero to about 70%
of the cultivated land over the last 25 years. Average annual rainfall is barely 500 mm,
so conditions are not favorable for rainfed agriculture. Project designers clearly perceive
that the real payoffs in such areas lie in irrigation development.
Projects that Focus on Rainfed Agriculture
In areas with limited opportunity for water harvesting, watershed projects
typically devote more attention to developing rainfed agriculture. This is the situation in
eastern Maharashtra and northern Andhra Pradesh, where the terrain is flatter and the
climate less arid. Watershed projects in these areas promote on-site soil and water
conservation measures that improve the resource base for rainfed agricultural production.
This is intended to pave the way for adoption of crop varieties that are responsive to
increased moisture. These projects often build water harvesting structures such as check
dams and percolation tanks, but they cannot offer the spectacular increases in irrigation
achieved in places like Ralegan Siddhi, because the terrain does not provide the same
opportunities for harvesting water.
In southern Andhra Pradesh, the most obvious opportunities for water harvesting
have long since been exploited in the form of traditional irrigation tanks. Some
opportunities remain, but often they lie in the catchment of an existing tank, thus
interfering with the traditional system. This helps explain why most projects in Andhra
Pradesh focus more on rainfed agriculture than irrigation. (The DPAP is the exception.)
Table 1 lists the projects in the sample area according to their primary orientation
toward water harvesting vis-a-vis rainfed agricultural development.
Table 1: Primary orientation of projects in the study toward either water harvesting
or rainfed agriculture
Primary orientation of the technical work Projects and locations
Primarily water harvesting DPAP (with pre-1995 guidelines) and Jal
Sandharan, Maharashtra and Andhra Pradesh
NGOs in Maharashtra
Adarsh Gaon Yojana and Indo-German
Programme (govemment-NGO collaboration)
Primarily rainfed agriculture NWDPRA, both Maharashtra and Andhra
Pradesh study sites
World Bank Pilot Project
ICAR Model Watershed projects
Both rainfed agriculture and water NGOs in Andhra Pradesh
PROJECTS COVERED UNDER THIS STUDY
Government Projects that Focus Primarily on Water Harvesting
This discussion of the different watershed projects operating in the study area begins
with the project sponsored by the Government of Maharashtra, because the Maharashtra
projects represent the roots of watershed development in India. The Jal Sandharan program
is the result of several decades of experience with watersheds in the state.
Watershed projects in Maharashtra:4 The elements of watershed development
date back to the 1942 Bombay Land Improvement Schemes Act. This initiative
resembled modem watershed projects in its focus on soil and water conservation,
improved rainfed farming methods, and controlled grazing. Watershed management
gathered momentum in Maharashtra following the severe 1972 drought. The
Government of Maharashtra launched the Employment Guarantee Scheme (EGS), which
aimed to provide work to anyone who needed it while also creating permanent assets
such as infrastructure. One important objective was to "drought-proof" the land by
building water harvesting structures that would provide drinking water and irrigation
throughout the year.
In 1982, the Government of Maharashtra initiated the Comprehensive Watershed
Development Program (COWDEP). This program was intended to combine the
budgetary resources of the EGS and the technical provisions of the 1942 Bombay Land
Improvement Schemes Act for a large-scale watershed development effort. One notable
problem was that work undertaken by COWDEP was administered by several
government departments, and coordination among them proved to be difficult.
Following COWDEP and other experiments in watershed development, the GOM
launched the Jal Sandharan Program in 1992. It represents an effort to take a more
comprehensive approach to watershed development, with the key innovation being that the
four government departments involved in the work were brought under one umbrella. The
Jal Sandharan, which became a department in itself, would also handle the funds from the
centrally-sponsored Drought Prone Area Program (DPAP), Jawahar Rojgar Yojana (JRY),
and National Watershed Development Project for Rainfed Areas (NWDPRA).
The Jal Sandharan program treats the village as the unit of planning,
implementing the work in microwatersheds that lie within village boundaries. Emphasis
is given to raising the water table to protect and enhance drinking water sources and
provide protective irrigation for at least one crop. The program is implemented by a
4 This discussion of state government programs in Maharashtra draws on Pangare
and Gondhalekar (1998).
committee at the district level representing all the government agencies involved in the
project. The work in each selected village proceeds with the consent of the village
sarpanch (elected leader) after a meeting of villagers is held to discuss the project.
The Jal Sandharan shows the signs of lessons learned from several decades of
state government experience in watershed development, but it also shows clearly the
difficulties of coordinating large-scale activities across government departments. In
particular, coordination in the upper levels of bureaucracy does not always translate into
coordination at the village level, where all the departments involved have separate
budgets and targets (Pangare and Gondhalekar 1998).
Drought-Prone Areas Programme: The Drought Prone Areas Programme
(DPAP) is sponsored by the Ministry of Rural Development in the central government.
The DPAP can be traced back to the Rural Works Programme initiated in 1971-72. It has
evolved gradually over time, initially covering a wide range of labor-intensive activities
such as soil and water conservation, afforestation, and development of irrigation and
infrastructure. Over time the program gradually focused more sharply on area
development for drought-proofing. By the late 1980s, the DPAP became exclusively a
watershed development program focusing on soil conservation, water harvesting, pasture
development and afforestation. A small amount of funds were earmarked for associated
activities such as livestock development, sericulture and horticulture.
As with other government-funded watershed programs, the DPAP was strictly a
technical program in which local people played little or no role. Many NGOs,
meanwhile, had moved toward a more fundamentally participatory approach in which
villagers shared in developing and implementing watershed plans. In 1995, the Ministry
of Rural Development adopted this approach on the basis of the well-known Hanumantha
Rao Committee Report (GOI 1994a). Under the guidelines subsequently drafted, (GOI
1994b), plans were to be developed by the villagers, with an emphasis on the use of local
technologies. Funds would go directly to the village, with villagers working hand in hand
with an independent project-implementing agency that could come from the government,
nongovernment or even corporate sector. A strong effort was made to move away from
the physical target orientation that characterizes most government programs. This radical
restructuring of the program has taken time to operationalize, and by 1997 almost no
work had been undertaken in Maharashtra. Progress was better in Andhra Pradesh, but
insufficient work had been done to warrant analysis of any post-1995 DPAP villages. As
a result, this study covers villages covered by the DPAP in its pre-1995 guidelines. In
Maharashtra the pre-1995 DPAP is synonymous with the COWDEP and Jal Sandharan,
and in Andhra Pradesh the approach is very similar. In fact, before the implementation of
the new DPAP guidelines the Jal Sandharan drew most of its budget from the DPAP.
Government Projects that Focus Primarily on Rainfed Agriculture
While the water harvesting and afforestation approach to watershed management
was gathering momentum in Maharashtra and in the DPAP, alternate approaches were
being introduced that focused more on developing rainfed agriculture through on-site soil
and water conservation practices. These approaches were led in India by on-station
research undertaken by the Indian Council of Agricultural Research (ICAR) institutes
such as the Central Research Institute for Dryland Agriculture (CRIDA), and also by the
International Crops Research Institute for the Semi-Arid Tropics (ICRISAT).
Indian Council of Agricultural Research Model Watersheds: In the mid-1980s,
ICAR decided to implement the findings of its dryland agricultural research in 47 model
watersheds around the country. Many of these pilot sites were treated as research
watersheds, where the work undertaken was closely monitored and changes in land and
water conditions were analyzed. Physical costs of the watershed works were relatively
low, but supervision was intensive, with persistent efforts to introduce new varieties and
other improved technologies and management practices.
CRIDA took up three such watersheds in Andhra Pradesh. One of them,
Chevella, is included in the present study. The other two, both of which are close to
Hyderabad, could not be included because they have since been converted to housing
developments. At least one model watershed was launched in Maharashtra, but it was in
the eastern portion of the state not covered by the present study.
World Bank Pilot Project for Watershed Development in Rainfed Areas: The
World Bank Pilot Project for Watershed Development was initiated in 1984 in Karnataka,
Madhya Pradesh, Maharashtra and Andhra Pradesh. Like the ICAR model watersheds,
this project sought to introduce improved rainfed agricultural technology. The project's
guiding philosophy was that low-cost soil and water conservation measures, including
improved agronomic practices like contour cultivation and vegetative rather than
mechanical bunds, could make a strong contribution to rainfed agricultural development
at a relatively low cost (World Bank 1988). While the work was undertaken on a
watershed basis, additional emphasis was given to proper treatment within each plot as
the project's design team felt this was missing from the watershed approach pioneered in
western Maharashtra. A major thrust of this program would be to promote contour-based
cultivation, which would conserve soil and concentrate moisture at very little monetary
cost. The improved soil moisture regime in turn would make improved seeds, fertilizers
and other inputs more productive.
The project also applied lessons learned from earlier projects regarding
institutional approaches. Efforts were made to streamline state government operations to
support the project; special offices were established at both the central and state level in
order to coordinate the administrative needs of the project.
Like other early government projects, the Pilot Project aimed for universal
implementation of a single, centrally-developed plan, with efforts made to convince local
people of its merits. The project document stressed the need to adapt proven technologies
to local conditions, but in practice there was little flexibility. Techniques not pre-
approved under the project were not supported. Concerning pasture development and
afforestation, it was recognized that the work would have no lasting impact unless people
supported it; accordingly, no pasture development work would be undertaken without
local people's consent. But that was about the extent of participation.
When the Pilot Project ended in 1991, a second phase of the World Bank project
was introduced. This project was called the Integrated Watershed Development Project
(IWDP), with separate components in the hills and plains. The plains portion of the
project was undertaken in Rajasthan, Orissa and Gujarat. The IWDP, representing the
next generation after the World Bank Pilot Project, took essentially the same approach to
developing rainfed crop production as the earlier project. Its main difference was that it
focused greater attention to developing and strengthening local organizations as the
means of garnering people's participation and collective action for protecting common
pasture areas. The project's administrative approach was also restructured. Project
evaluations suggest, however, that the IWDP suffered from the same problems of poor
participation and inflexible technology choice as the Pilot Project (ICRISAT 1996; RAU
1999; personal communication with Director of Watersheds, Rajasthan). Since the
present study is confined to Andhra Pradesh and western Maharashtra, the analysis
addresses only the Pilot Project.
National Watershed Development Project for Rainfed Areas: The National
Watershed Development Project for Rainfed Areas (NWDPRA) is the Ministry of
Agriculture's counterpart to the World Bank-funded Pilot Project and IWDP. Similar in
approach to the World Bank projects, the NWDPRA promoted the same low-cost
vegetative bunding techniques and contour-based cultivation. Vegetative and other low
cost measures were also used in the nonarable lands (Government of India 199 la). The
NWDPRA is centrally funded and operates through state-level Departments of
Agriculture or Watershed Development. In 2000, the NWDPRA adopted the MORD's
more participatory guidelines, and in the coming years the two projects are to be
implemented with common guidelines. This study covers villages developed under the
NWDPRA's earlier approach.
Another similarity to the World Bank projects is that the NWDPRA works on a
watershed basis, where watersheds do not necessarily correspond to village boundaries.
The NWDPRA watersheds are only about 500-5000 ha, or around 5-20% of the area of
the World Bank watersheds. As a result, the NWDPRA watersheds typically cover one
village entirely or nearly so, plus parts of one or two neighboring villages. This approach
is considered to make the most sense from a land and water management perspective, but
it raises administrative complications because project staff have to deal with multiple
village administrations in one relatively small area. Also, organizing local institutional
arrangements for managing nonarable common lands is complicated when working
across village boundaries. The World Bank Pilot Project and IWDP shared this same
problem and had difficulty in making it work (ICRISAT 1996).
The NWDPRA project guidelines mentioned the issues of people's participation
and institution-building, but they presented no clear strategy and only a small budget for
addressing them (GOI 1991a). It appeared that the project's intentions were in line with
modem views about the benefits of participation but that the mechanisms for ensuring
them were not fully developed. This is discussed further in Kolavalli (1998).
In western Maharashtra, implementation of the NWDPRA was strongly influenced
by the fact that the project was implemented by the same agency that plans and implements
the engineering-based approaches of the COWDEP, Jal Sandharan and DPAP. In particular,
the primary focus remained on treating drainage lines and catchment areas to promote
infiltration of water. One of the most notable differences was simply that the technologies in
use were much less expensive. For example, drainage line structures under the NWDPRA
contained no cement and were limited to a maximum cost of Rs 25,000 per structure,
whereas under other projects individual water harvesting structures might cost seven or eight
times as much. As a result, water harvesting was not the NWDPRA's strength.
In Andhra Pradesh, the NWDPRA was operated by the Department of Agriculture
and more clearly matched the approach envisioned in the project guidelines.
Nongovernment Organizations: A Focus on Social Organization
NGO programs are by no means uniform, but they share the common feature of a
strong emphasis on social organization. Their guiding principle is that without proper
social organization, efforts to introduce watershed technology will be fruitless.
The two features that most distinguish NGO watershed programs from government
programs are their scale of operations and their staffing structure. While government
programs have huge budgets and work in hundreds of villages, most NGOs work in only a
handful of villages. They devote more staff time per village, and they often work on a
variety of activities in addition to watershed management. Second, while government
employees concerned with watershed management are almost exclusively trained in
agricultural sciences and engineering, NGO staff members include many more
nontechnical staff trained in community organization. They believe that social organization
contributes as much to successful watershed development as technical input. Some NGOs
collaborate with government agencies that provide technical expertise, but others do not.
It is important to note that NGOs vary a great deal. Some are large and well
established, with access to substantial funding, whereas others are smaller, less
experienced, and underfunded.
NGOs in Maharashtra:5 Watershed management in Maharashtra has roots in the
nongovernment sector that go back nearly as far as those in the government programs. In
the early 1980s two villages became well known for their watershed management
programs: Ralegan Siddhi in Ahmednagar district and Adgaon in Aurangabad district.
Many current government and NGO initiatives draw inspiration from them.
In the 1970s, Ralegan Siddhi was a poorly developed village almost devoid of trees
and grass and a haven for liquor dens. Anna Hazare emerged as a local leader and brought
about various social changes in the village, particularly family planning, a ban on alcohol,
protection of nonarable lands against open grazing and felling of trees, and shramdan, or
voluntary labor for community welfare. Around the same time he also learned about the
benefits of soil conservation and water harvesting. The social changes brought order and a
sense of community to the village, while soil and water conservation work (implemented
by COWDEP) and protection of the common lands helped restore the natural resource
base. This was the beginning of people's participation in watershed development.
Among the many NGOs working in watershed development in Maharashtra, one of
the best established is Social Centre, founded in Ahmednagar in 1969 by Jesuit priests.
Between the period 1969-1988, it was engaged in various activities such as small loans,
community lift irrigation schemes, community health programs, etc. In 1988 it shifted its
5 This discussion of NGO projects in Maharashtra draws on Pangare and
focus towards motivating and organizing entire villages to undertake ecological regeneration
of their own watersheds. The Social Centre played a key role in launching and designing the
statewide Indo-German Watershed Development Programme, discussed below.
NGOs in Andhra Pradesh: NGOs in rural Andhra Pradesh have traditionally
focused on the problems of lower caste communities. Caste structure is more
dichotomized than in Maharashtra, with more villages in which one or two large land
owning families control large tracts of land while many others remain landless. As a
result, NGOs typically focused on non-land based activities such as developing and
strengthening local credit institutions. With the rise of watershed development as a focal
point for rural development, some NGOs gradually adopted it into their project portfolio
(Sanghi personal communication).
In recent decades Andhra Pradesh has had successive waves of large-scale
privatization of common lands in which landless and near landless people were given
legal but nontransferable title to formerly common lands (Pender and Kerr 1999). Many
NGOs expanded their work from credit and other income generation activities to support
agriculture on the privatized land, much of which is of low quality. As they expanded to
a watershed approach they also began to work with other farmers with higher quality
land. But their primary orientation toward helping poor, landless people means that these
watershed agencies tend to be more committed to making landowners pay for work done
on their own property. For example, while most projects in Maharashtra and the centrally
funded government programs typically ask for no more than a 10% contribution from
farmers for work done on their private lands, some NGOs in Andhra Pradesh require a
more substantial contribution for work done on private lands. Some of the implications
of this policy are discussed below, in the discussion on land improvement investments.
One interesting difference between the works conducted by MYRADA, an NGO
operating in Andhra Pradesh, and that by the Maharashtra NGOs is MYRADA's greater
focus on trying to build consensus among different interest groups in a watershed. As
discussed above, the costs and benefits of watershed development can be spread
unevenly. This raises difficulties in project implementation, especially where
socioeconomic diversity is relatively high. MYRADA addresses this problem by trying
to organize communities to develop mechanisms to compensate those who lose so that
they will go along for the greater good (Mascarenhas et al. 1991; Fernandez 1993, 1994).
In all of its rural development projects, MYRADA organizes people in small,
homogeneous groups to work toward one common purpose. In the context of
watersheds, the first step is to work in "miniwatersheds" of no more than a few hundred
hectares and a hundred farmers. Second, MYRADA helps form small subgroups of
farmers based on homogeneity of location, socioeconomic conditions or interests. These
groups all belong to a larger miniwatershed group. This preserves the participatory and
socially functional character of the smaller, homogeneous subgroups while also retaining
advantages of scale in planning watershed works and interacting with government
agencies, banks, and input suppliers. The larger group provides a vehicle for airing
complaints and settling disputes among people from different subgroups.
Government-NGO Collaborative Programs in Maharashtra
The most intriguing aspect of watershed development in Maharasthra in recent
years is the rise of collaborative programs between government and non-government
agencies. The two main examples are the Adarsh Gaon Yojana (Ideal Village Scheme, or
AGY), and the Indo-German Watershed Development Programme.
Adarsh Gaon Yojana:6 The AGY is a major initiative that seeks to replicate the
Ralegan Siddhi model in 300 villages by combining the technical staff of the Jal
Sandharan program with the social orientation of NGOs.
The key elements of the AGY are government-NGO collaboration and strict
guidelines for social organization. Villages participating in the AGY must undertake to
follow the five social principles of Ralegan Siddhi: family planning, a ban on alcohol, a
ban on open grazing, a ban on cutting trees, and shramdan. The idea is that adherence to
these five principles can lead the village towards self-sufficiency by helping them meet
their needs for water, food, fuel and fodder within their own village. The philosophy also
6 This discussion of the AGY draws on Pangare and Ghondhalekar (1998).
promotes a set of values that encourages self-discipline and a willingness to overcome
social barriers and political factionalism to work for the common good.7
Shramdan is intended to foster a spirit of self-sufficiency and self-dependence.
The idea is that when villagers observe the benefits of the physical works carried out for
watershed development, it gives them a sense of satisfaction and achievement. They also
feel responsible for the maintenance of the structures for which they have invested their
own labor. Shramdan is also seen as a good way of getting people together to work for
the welfare of the entire community.
Nongovernment organizations play an important role in the AGY. People in each
village select a local NGO to help them implement the different development activities
and adhere to the social principles. The NGO also maintains records and accounts, and
monitors the project activities. In addition, the NGO coordinates with the government
departments at the state level to access funding and technical guidance. The Jal
Sandharan Department, meanwhile, implements the technical work.
Funds under the project are to be used for two main types of activities, namely,
watershed development (the core activity) and other development activities (non-core
activities). The latter are carried out by the government agencies in question, as listed
above, in consultation with the people of the village. Government departments are
supposed to give AGY villages preference in providing services. Steps are undertaken to
reduce corruption and peripheral expenses.
Indo-German Watershed Development Programme: The Indo-German
Watershed Development Programme (IGWDP) is another example of collaboration
between government and nongovernment organizations that seeks to scale up the success
of small NGO programs (Farrington and Lobo 1997; WOTR 2000; NABARD 1995).
Initiated in 1993, the IGWDP develops microwatersheds in a comprehensive manner
7 It is important to note that shramdan has a long history in Maharashtra and is
considered culturally appropriate. In other areas, other means of promoting cooperation
and social discipline may be preferred. In southern Rajasthan, for example, Seva Mandir
insists that villagers reverse all illegal encroachment on common lands before they will
undertake work there (Seva Mandir 1999; Ahluwalia 1997).
through the initiative taken by village groups. Its guiding philosophy is the need for
collaboration among village level organizations, NGOs skilled in social organization, and
government organizations skilled in technical work. Further, it accepts that although
indigenous knowledge and practices are important, they need to be augmented by modem
techniques and management practices. The IGWDP has developed elaborate procedures
to cut through bureaucratic turf wars and red tape, ensuring that funds move quickly
(Farrington and Lobo 1997). As of July 2000, the IGWDP has developed 123 villages
covering about 130,000 ha, with the involvement of 74 NGOs (WOTR 2000). Plans are
being considered to spread this program to other states.
Investment in physical capital under the IGWDP begins only after evidence of
social organization suggests that people will work together to maintain the investments
on both private and community land. As with the AGY, there is as strong an emphasis on
developing the village's social capital as its natural and physical capital, and the villagers
must submit to similarly strict social conditions.
The work begins with 12 to 18 months of social organization work. This is
almost 12 to 18 months longer than the social organization phase of a typical government
watershed program, but it is shorter than that of many NGOs, which conduct work on
several other areas of village development before venturing into watershed development.
One important early project activity under the IGWDP is to plant trees and grasses in the
catchment area. This is done prior to building water harvesting structures in order to
force the inhabitants of the village to show that they can enforce social fencing to protect
natural vegetation. Only after people demonstrate such social discipline does the project
invest larger amounts of funds in new watershed structures.
The NGO helps organize and develop a Village Watershed Committee (VWC),
which is essentially a village-based NGO. The idea is that the VWC will eventually
outgrow the need for support from the original NGO.
INVESTMENT COSTS PER HECTARE UNDER EACH PROJECT
Information about the cost per hectare under different projects is helpful in
assessing their cost effectiveness. It also helps in interpreting the findings of quantitative
analysis presented later in this paper; it should not be surprising if one project that spends
twice as much as another also has more measurable impact.
Unfortunately, measuring project costs is difficult. Some projects, including the
NWDPRA and the World Bank, have expenditure guidelines that can be taken as a broad
indication of the level of investment per hectare. Others require calculating estimates
based on the total expenditure and the total area covered, but records are difficult to
obtain. For example, officials of the Jal Sandharan project say that their budgets are
constructed on the basis of structures to be built and vegetation to be planted, not the area
of the watershed. Accordingly, calculations of cost per hectare are only approximate.
NGOs tend to keep poor records about costs per ha, and calculating them is very
difficult because NGOs tend to undertake a variety of activities in addition to watershed
development. So even when costs can be calculated, it is not always clear what to
attribute them to. One certainty is that NGOs have much higher administrative costs than
government projects, since they devote much more time to social organization for which
expenditures are not directly tied to treated area.
In Maharashtra, since nearly all projects operate in ex-COWDEP villages,
calculating costs requires taking the sum of expenditures under both the old and new
programs. Records from the old projects are poor, so the cost figures are only approximate.
Rough estimates of project costs per hectare by project category are presented in
Table 2. For the NWDPRA and World Bank Pilot Project the upper range is the cost
listed in the project guidelines, while for the Jal Sandharan and COWDEP it is based on
the total number of structures built divided by the area covered. AGY and IGWDP costs
are calculated similarly but there are higher staff costs. The NGO figures are based on
estimates by the officials interviewed.
Table 2: Estimated cost per ha of watershed development under different programs
Project category Approximate cost per ha in 1998 Rs.
excluding COWDEP Including COWDEP
NWDPRA 2500-3500 4000-6000
Jal Sandharan 2500-4000 4000-6500
NGOs 4000-6000 5500-8500
AGY/IGWDP 3500-5500 5000-8000
World Bank 5500-6500 5500-6500
Evaluating watershed projects requires baseline and monitoring data for
comparison of pre- and post-project conditions, but unfortunately no such information
was available for this study. As a result, the quantitative analysis is based on some
secondary data available for both the pre-project period (19878) and the present (1997),
primary data of current conditions based on interviews and visual assessments, and
primary data of past conditions based on recall by local inhabitants. Inevitably there are
weaknesses in the data that limit the study's analytical power.
A major component of the research was the development and collection of data on
various indicators of performance in natural resource conservation, agricultural
productivity, and equitable distribution of project benefits. These data were collected
through direct observation, group discussions, and published records. Quantitative data
were also collected on the background characteristics of the projects, villages, households
and plots covered under the study. Some of the village-level information came from
public sources, but most of it was collected from group and individual interviews in each
village. In addition, qualitative data were collected regarding the natural resources
people use to earn their livelihoods, the social institutions that govern access to those
resources, and any changes in access to them resulting either from changes in their
8 Work in the World Bank and ICAR project villages began in 1986; in villages
under these projects the baseline period was the year before the project began.
quantity or changes in social institutions. This information was collected in open-ended
discussions with members of specific interest groups in each village, such as farmers with
irrigated land, farmers with rainfed land, landless people, herders, and women.
The village, rather than the watershed, was selected for analysis of community
level indicators of natural resource management and economic performance. This is
because most projects in the sample worked at the village- or sub-village level, people are
organized around villages, and secondary data are recorded at the level of the village. In
some cases, particularly in Andhra Pradesh, villages are disaggregated into hamlets, in
which case primary data were collected at the hamlet level.
There is no single indicator of successful watershed development, so the most
feasible approach is to compare the performance of a variety of indicators. The various
performance indicators also reflect the diversity of project objectives. These include,
among other things, raising rainfed agricultural productivity, recharging groundwater for
drinking and irrigation, raising productivity of nonarable lands, reducing soil erosion,
skewing benefits toward poorer members of society, creating employment (directly and
indirectly), promoting collective action, and building or strengthening social institutions.
All the projects surveyed shared most of these objectives but, as described in Section 2,
they differed in their relative emphasis.
As mentioned above, the indicators vary in their level of rigor and reliability,
which is inevitable given the lack of baseline or monitoring data in the study villages.
Table 3 presents an overview of performance criteria, ideal indicators, and the indicators
actually used in the current study, and Table 4 shows the level at which they operate.
Table 3: Performance indicators used to compare project performance
* measurement of erosion and
associated yield loss
Measures taken to
Proxy indicators used in this study
* visual assessment of rill and gully erosion (current
* inventory, adoption and
effectiveness of soil and water
conservation (SWC) practices
* measurement of groundwater
levels, controlling for aquifer
variation and pumping volume
* time series, intrayear and
interyear variations in soil
moisture, controlling for
* net returns at the plot level
* change in production from
revenue and forest lands (actual
* change in household income
* nutritional status
Note: aAll ideal indicators would be collected both before and after the project.
* visual assessment of SWC investments and apparent
effectiveness (current only)
* adoption of conservation-oriented agronomic
* expenditure on SWC investments
* approximate change in number of wells
* approximate number of wells recharged or defunct
* change in irrigated area
* change in number of seasons irrigated for a sample of
* change in village-level drinking water adequacy
* change in cropping patterns
* change in cropping intensity on rainfed plots
* relative change in yields (higher, same or lower)
* net returns at the plot level, current year only
* relative change in production from revenue and forest
lands (more, same or less than pre-project)
* extent of erosion and SWC on nonarable lands
* perceived effects of the project on the household
* perceived change in living standard (better, same,
* change in housing quality
* change in percentage of families migrating
* perceived changes in real wage and availability of
casual employment opportunities (higher, same,
Table 4: Performance domains and the units of analysis at which they operate
Type of performance Level of measurement
Village Plot Household
Social organization institutions to protect membership in
common lands village
use of voluntary organizations
Natural resource irrigated area and irrigated area drinking water
conservation drinking water supply population of supply
soil erosion and improved
conservation dairy cattle
habitat for wild
animals and migratory
Agricultural productivity soil erosion and irrigation
Equity and poverty assets (wealth)
alleviation access to
standard of living
Determinants ofProject Performance
Village level: Data collected at the village level are based on a survey covering
background information such as access to markets, land use patterns, natural resource
management practices, and description of social institutions operating in the village.
Most background information is available for both 1987 and 1997. A village-level
survey was conducted to obtain most of this information, and additional background
variables were obtained from the 1991 census. Performance indicators at the village level
include some variables from the village survey, but also visual observations of natural
resource conditions from village level transects covering a cross-section of broadly
representative land types and uses.
Plot level: A plot-level survey was conducted to collect data on agricultural
productivity and adoption of improved technologies and practices. This provides
information about changes resulting from the watershed projects and other determining
factors. The sample includes both irrigated and rainfed plots, and both plots covered and
not covered by watershed projects. Village-level information related to each plot is
available from the village survey. Some household-level information for each plot's
operator was also collected as a part of the plot survey.
Household level: A household-level survey supplied detailed information about
household characteristics and changes in household welfare. This provides indications of
how watershed projects and changes in a variety of village- and household-level conditions
have affected household welfare. This survey recorded concrete changes in living standards,
such as ownership of durable goods and quality of housing, as well as respondents'
subjective assessments of changes in their well-being. The household survey was conducted
using a different set of respondents from the plot survey, but in the same villages.
Interest group level: A fourth set of interviews focused on different interest
groups within the village, such as farmers with irrigation, farmers without irrigation,
landless people, and women. The information provided by these interviews offers a
qualitative assessment of project performance from the viewpoint of the intended
beneficiaries, and it provides further insights about how project benefits and costs are
distributed across different groups of people within the village.
Sampling villages for data collection was a major undertaking in itself. The
situation in Pune and Ahmednagar districts of western Maharashtra provides a good
example of the difficulties. Despite widespread publicity about the success of the
watershed approach to agricultural development, hard data were quite limited. A few
widely known success stories were easy to locate, but others were not. This was the case
for two reasons. First, the 1991 Census lists over three thousand villages in the two
districts, but the famous success stories accounted for no more than a handful. Second, a
complete list of villages where projects have operated did not exist. The most active
watershed agency in the area, the Maharashtra Department of Soil and Water
Conservation, kept good records of the villages where work was currently underway, but
lists of villages where work had been completed were archived and could be accessed
only with difficulty. Some government programs, like the NWDPRA, maintained lists of
project locations only at the taluka level. NGOs maintained their own lists, which could
be obtained by visiting the head office. As a result, simply identifying project villages
required a great deal of legwork. The resulting list of project villages was then checked
against the complete list of all villages from the national census so that nonproject
villages could be selected as a control against which to compare project performance. All
the sampled villages were visited to confirm their project status.
The Maharashtra study villages were all located in Pune and Ahmednagar districts
in the western part of the state, where there was a relatively high concentration of
watershed project sites. The eastern side of this study area is drought-prone, while the
area closer to the Western Ghat mountain range has higher rainfall. In Andhra Pradesh
the projects were less concentrated, so the sample villages covered 4 districts: Anantapur
in the far south of the state and Medak, Mahbubnagar and Ranga Reddy, all of which are
in the north, near Hyderabad.
Based on available knowledge about project status of villages, the sample was
selected at random, stratified by the project categories listed above and, in Maharashtra,
by geographic location. With five project categories and two geographical zones, there
are ten strata in Maharashtra. A small amount of resampling was done to replace villages
incorrectly classified as "control" after visits to the villages revealed that watershed
projects had operated there in the 1980s. In Andhra Pradesh, where only sixteen villages
were sampled, geographic stratification was omitted to ensure that each stratum has at
least two observations. Thus Andhra Pradesh has only five strata.
As shown in Table 5, data were collected in 70 villages in Maharashtra and 16
villages in Andhra Pradesh. A full set of quantitative and qualitative data at the village,
household and plot level were collected in a random, stratified subsample of 13 of the
Maharashtra villages and all 16 Andhra Pradesh villages, for a total of 29. In the
remaining 57 Maharashtra villages only village level data were collected. The village-
level analysis is confined to the 70 Maharashtra villages, while the plot-level analysis
covers the 29 villages from both states where more detailed data were collected. The
qualitative data cover primarily these same 29 villages.
Table 5: Location of the study villages
Type of analysis Maharashtra Andhra Pradesh Total
Total 70 16 86
Village-level (quantitative) 57 0 57
Village, plot, household (quantitative) 13 16 29
and interest group (qualitative)
Teams of five to seven village investigators spent four days and nights in each of
the 29 villages where they collected the full set of quantitative and qualitative data. In
the remaining 57 villages in Maharashtra where only village-level data were collected,
teams of three to four investigators spent two to three days.
CHARACTERISTICS OF THE SAMPLED VILLAGES AND PLOTS
Village-level analysis is presented for Maharashtra. Conditions in the two states
are sufficiently different to make it useful to analyze them separately, and the
Maharashtra sample of 70 villages facilitates the analysis.
Villages are characterized by a variety of factors that might affect performance in
agricultural productivity, natural resource management and living standards. Table 6
defines the variables representing these characteristics.
OT CO aa
*? l *?
44 0 "-
's rc A 'o *0
(4 a cc
0 "l > 0 0
-- -- -^-
t0l 0 0o
4) 0~ ~ ~S
00 00 00 00
'-4-- ll -
ce $e c -e r:
C U (g g n -11 -
Co Cd Cd i
o o \OSS5 5 5>
s -t--r- r-r--
oo o 0
cV d o aCA
^0 |0 o a
Table 7 shows the number of sampled villages in each project category and also
the total number of villages in the study area. The relatively small number of sampled
villages under the NWDPRA, NGO and AGY/IGWDP categories reflects the fact that
these projects were not very widespread at the time of the study; the sample includes
nearly the entire population for the projects.9 The Jal Sandharan and nonproject villages
in the sample, on the other hand, represent only a small fraction of their total populations.
Table 7: Number of villages in the Maharashtra village-level survey, by project
Project category Number of Total population of
villages villages in the study
NWDPRA 10 11
Jal Sandharan/DPAP 17 201
NGO 12 13
Government/Non-government collaboration 14 27
No project 17 361
Total 70 613
Project Characteristics: As discussed in Section 2, projects vary in various
characteristics such as the percentage of village area that they cover, the number of years
they operate, the amount of funds they spend, and the training of their staff. Table 8
presents the approximate mean values of such information for the sampled villages in
Maharashtra. The figures in the table show that the percentage of each village covered
and total expenditure per hectare in each village10 are slightly less under the NWDPRA
than other projects, but these differences are not significant. As discussed in Section 2,
an important difference is the fact that the NWPDRA projects cover multiple villages
while the other projects work in only one village at a time.
9 Subsequent expansion of the Indo-German project and the Drought Prone Area
Project raised the number of villages in these categories. Under its new guidelines, the
DPAP works mainly through NGOs, so the number of NGO-led projects in the study area
is slated to rise quickly.
10 Total expenditure per hectare is calculated as the cost per hectare treated
multiplied by the percentage of the village covered by the project.
Villages are also differentiated by the amount of time in which they have
undergone watershed work. This is so for both the current or most recent watershed project
and earlier work (if any) in the 1980s done by the Maharashtra Department of Soil and
Water Conservation under COWDEP. Table 8 shows that, by either measure, the average
duration of projects is not significantly different across watershed project categories. With
the exception of the nonproject villages, all but two villages in the sample also had work
done under COWDEP in the 1980s. The amount of work actually performed by COWDEP
varied by village, but the figures in Table 8 are still striking: in one form or another,
watershed development has taken place in these villages for a long time.
Finally, perhaps the most noticeable point in Table 8 is the difference across
projects in the percentage of staff members trained in social organization skills. In the
two government programs (NWDPRA and Jal Sandharan) no staff members had any
training in social organization, while in those with an NGO component nearly half of
them did. This reflects the differences in relative orientation toward social organization
vs. transfer of technology described in Section 2.
Table 8: Project operations in villages under each project category, Maharashtraa
NWDPRA DPAP/Jal NGO NGO/Govt No project
Mean area of the village (ha) 2102 1422 1209 1144 905
Number of villages in which 9/10 16/17 11/12 14/14 0/17
COWDEP previously worked (out of
Mean % area covered by the old 38 49 43 46
Mean % area covered by new projects 36 39 42 38
Mean % area covered by both old and 74 88 85 84
Mean % of staff members with 0 0 42 45
training in social organization
Mean number of years since the most 0 0 0.25 0
recent project ended
Mean number of years of work under 7.2 8.4 7 8.7
Mean number of years under the new 5.9 4.9 6 5.1
Mean number of years under both 13.1 13.3 13 13.8
Mean cost/ha actually treated under 4500 4783 4989 4963
new project (Rs)
Mean expenditure/ha for entire village 1880 2355 2148 2310
under COWDEP (Rs)
Mean expenditure/ha for entire village 1622 2006 2207 1874
under new project (Rs)
Mean expenditure/ha for entire village 3501 4361 4355 4185
under both projects (Rs)
Notes: a Most figures are approximate, based on calculations from official records.
b Analysis of variance (ANOVA) test shows that group means differ significantly across
project category (excluding nonproject) only for the percentage of staff members with
training in social skills (F=45, 3 df, p<.001). No other variables listed in this table show
significant difference across project types.
Approximately 12 plots were sampled in each of 13 villages in Maharashtra and
16 in Andhra Pradesh. The villages were stratified by watershed project category; the
resulting sample of villages selected is shown in Table 9. Within each village, plots and
households were selected at random, stratified by land capability classification and
irrigation status. Every third plot was selected from the transect line that crossed
representative areas of the village, until four plots were selected in each of three land
capability classes, with three rainfed and one irrigated plot within each category.1
Table 9: Number of villages covered in the plot survey, by state and project category
Project category Number of projects
Maharashtra Andhra Pradesh
World Bank/ICAR 0 4
NWDPRA 2 3
Jal Sandharan/DPAP 2 3
NGO 3 3
AGY/IGWDP 3 0
No project 3 3
Total 13 16
As shown in Table 10, about 30 percent of the plots are irrigated, with the
proportion of irrigated land higher on better quality land. (About 21 percent of plots were
irrigated at the start of the study period in 1987.) 62 percent of class II lands are
irrigated, while the corresponding figures for class III and class IV lands are 37 percent
and 7 percent, respectively. This distribution arises for two reasons. First, given the
choice farmers will irrigate their better lands first, since water will give higher returns
when applied to better soil. Second, land class is somewhat endogenous with respect to
irrigation, since irrigated plots tend to be leveled and receive higher organic matter
inputs. As a result, it is possible for a given plot to change to a higher land classification
after it becomes irrigated.
1' In some villages, the sampling approach was altered somewhat because there
were not enough plots with the desired irrigation status or land capability classification.
For example, in some villages land quality could only be divided into two categories, and
in others there were no irrigated plots on land of below average quality.
Table 10: Number of plots in the plot survey, by irrigation and land capability
Irrigation status Land capability classification Totals
II III IV
Rainfed 22 99 125 246
Irrigated 37 59 9 105
Total 59 158 134 351
Sampled plots were not stratified by the farmer's total land holding size, but as
shown in Table 11 the sample is distributed fairly evenly across respondents of different
land holding size. It is important to note that the sampling approach oversampled large
plots relative to small ones, because the transect line was more likely to cross a large plot
than a small one. However, plots are not large overall; the mean size is 0.72 ha and the
median is 0.42 ha.
Table 11: Sample for the plot survey, categorized by land holding size
Category Hectares operated Number
Small 0-2 118
Medium 2-4 114
Large >4 119
The sampling approach led to a reasonably even distribution of plots across
categories when classified by soil, irrigation and land size holding. Other factors were
not controlled, however; for example, village and household characteristics that may
affect productivity and natural resource conditions at the plot level may vary across
project category. Data were collected to incorporate these factors in the plot-level
analysis. Table 12 presents the various characteristics to be examined in the plot-level
analysis and the source of data used for this study.
Table 12: Plot, household, village and project characteristics that potentially
determine performance at the plot levela
Description of variable Years data are Source of data
Biophysical and management characteristics of the plot
Plot area (hectares)
Average annual rainfall measured at the nearest taluka headquarters
Land capability classification: sample includes plots of class II, III, and IV
1987 and 1997
Irrigation status in 1987 and 1997: 1 if it is irrigated at least one season, 0 if it is rainfed
Slope: 1 if slope is 0-2%, 2 if slope is 2-4%, 3 if slope is 4-8% 1997
Characteristics of the farm household
Total hectares of land owned by the household
Total number of adult workers in the household (men, women and
long term hired workers)
Highest number years schooling of any male household member
Highest number years schooling of any female household member
% income from off-farm sources
Change in percentage of off-farm income between 1987 and 1997
Tenure status: 1 if the farmer owns the plot, 0 if the operator is a tenant
Land title status: 1 if the farmer has a transferable title, 0 if not
Plot rank (relative to farmer's other plots): 1 if it is the farmer's only
plot or a better than the average quality plot in his holding, 2 if it is an
average quality plot, 3 if it is below average.
Farmer interacted with project staff: 1 if the farmer has interacted with
the project staff, 0 if not.
Project staff made technical recommendations to farmer: 1 if the
project staff made technical suggestions, 0 if not.
Farmer adopted technologies or practices recommended by project
staff: 1 if the farmer adopted the agency's suggestion, 0 if not.
1987 and 1997
1987 and 1997
1987 and 1997
1987 and 1997
1987 and 1997
1987 and 1997
Village, taluka, district, state
Altitude range between highest and lowest point in the village (meters)
Position in macrowatershed (lower, middle or upper reaches)
Distance to taluka headquarters (km)
Distance to district headquarters (km)
Distance to a large city (Pune in Maharashtra, Hyderabad or Bangalore in Andhra Pradesh),
Type of road connecting the village (highway, paved road, good 1987 and 1997
unpaved road, bad unpaved road, bullock cart path)
Distance in km to nearest bus stop (km) 1987 and 1997
Number of visits to the village by an extension agent per month 1987 and 1997
Distance to the nearest bank (km) 1987 and 1997
Table 12 continued
Description of variable
Strong leader in the village promotes social and economic
development (1 if yes, 0 if no)
Population density (inhabitants per square km.)
Distance to nearest industrial unit (km)
Distance to nearest regulated market (km)
Nominal daily wage (Rs)
Percentage of houses in the village with an electrical connection
Type of watershed project operating in the village, if any
Number of years the current or most recent watershed project
operated in the village
Combined number of years under the most recent watershed project
and another previous project
Approximate percentage of the village's area covered under the
Notes: a Primary data for 1987 are based on respondents' recall in 1997.
Years data are
1987 and 1997
1987 and 1997
1987 and 1997
1987 and 1997
Source of data
CHARACTERISTICS OF HOUSEHOLD AND INTEREST GROUP RESPONDENTS
In addition to demographic and socioeconomic data collected for the household
operating each sampled plot, an additional detailed household survey was conducted to
learn about the welfare impacts of watershed projects and other factors potentially
influencing development. 347 respondents were sampled randomly, stratified by land
holding size. These respondents differed from those in the plot survey in that about 20%
of them were entirely landless; the remaining respondents also had smaller holdings on
average than those in the plot survey. This report does not draw much on the household
survey, so it is not described in detail here.
Group interviews to collect qualitative data were collected in the same 29 villages
as the plot data. The respondents for these interviews included the village's elected
leader, or sarpanch, representatives of the watershed agency, and specific interest groups
in the village such as farmers with irrigated land, farmers without irrigation, landless
people (often herders), people from low castes, etc. Men and women were interviewed in
separate groups. Facilitators of these discussions had a list of unstructured questions to
ask, but they also encouraged participants to address other issues of importance to them.
This section discusses the approach taken to identify the contribution of
watershed projects to agricultural productivity, natural resource management and poverty
alleviation, taking into account the contributions of other factors such as infrastructure
development, agroclimatic conditions and village-level social capital. The analysis is
structured around the basic program evaluation question: what would have been the state
of agricultural productivity, natural resource management and poverty in the absence of
the project interventions? Answering this question is complicated, because one never
observes the same villages (or households or plots) both participating and not
participating in the program at the same time. As a result, care is needed to identify other
factors that may have contributed to observed outcomes. These include contemporaneous
events, such as changes in infrastructure and market access, and systematic biases in
where the projects choose to operate or which villages (or people) choose to participate.
Since projects are not placed randomly, differences in project outcomes may depend on
pre-existing village characteristics in addition to project activities.
Historically, the evaluation profession has been characterized by a split between
quantitative and qualitative researchers. Recent years, however, have seen a growing
appreciation of the benefits of combining the approaches of both (Greene and Caracelli
1997; Patton 1997). Quantitative evaluation uses statistical analysis to disentangle
project effects from intervening factors, relying mainly on theory to explain how the
project activities lead to impact. It follows the logical positivist belief that a single,
objective truth exists independently of the observer. Qualitative evaluation, on the other
hand, tends to make fewer assumptions about how a project affects individual behavior.
It focuses on the mechanisms of change while also yielding qualitative measures of
impact. Combining the two types of information can yield a particularly thorough
understanding of project impact.
This study uses mainly quantitative analysis, but it also draws on qualitative
information to better understand the relevant research questions, identify projects'
unintended consequences and the mechanisms though which they operate. This sets the
stage for a better-informed quantitative investigation, with greater confidence that
statistical analysis addresses the most important questions and incorporates the most
relevant variables. Subsequent qualitative investigation then helps interpret the findings
of the statistical analysis and rule out competing explanations for observed differences
across projects. This is particularly important given limitations in the data.
ASSESSING ENDOGENEITY IN PROGRAM PLACEMENT
The critical problem in quantitative evaluation is endogeneity, which arises if
some factors affect the project placement and the outcome simultaneously. This makes it
difficult to distinguish the effect of the project from the underlying factors that
determined where the project operated. The only way to solve this problem completely
would be to observe the same individual at the same point in time, both with and without
the project. Of course this is not possible, so the evaluator must try to set up one group of
observations affected by the project treatment, and a control group not affected by the
project but identical in every other way. This follows the standard experimental design
of the natural scientist. In practice, in many social science settings it may be possible to
identify similar but not identical treatment and control groups, so the social scientist's
experimental design can never be perfect (Manski 1995). This is discussed below with
respect to watershed projects.
Evaluators follow three main approaches to establishing control and treatment
groups: randomization, or pure experimental design; quasi-experimental design, and non-
experimental design (Ezemenari et al. 1999). Randomization refers to randomly placing
individuals into two groups-one that receives the project treatment and one that does
not. This solves the endogeneity problem by ensuring that the two groups are statistically
equivalent, so that any difference in average outcomes after the project can be attributed
to the project. In the present context, five separate randomly placed groups would be
needed: four for the different watershed projects and one control. Obviously the
randomization must take place before the projects begin. In the case of Indian
watersheds, the different projects operated independently and little or no advance thought
was given to evaluation, so the randomization approach is not possible.
Quasi-experimental design involves matching program participants with a
comparable group of individuals who did not participate in the program. This simulates
randomization but need not take place prior to the intervention. For example, Pitt and
Khandker (1996) used such an approach to estimate the effects of microcredit programs
in Bangladesh. They matched program and nonprogram villages and then devised an
elaborate set of matched groups in both sets of villages based on eligibility requirements
for participating in the programs. In another example, Jalan and Ravallion (1998) took
advantage of a large existing data set to estimate the probability that households
participated in a public works program in Argentina. They then constructed treatment
and control groups by matching participating and nonparticipating households that had
the same predicted probability of participation (Ezemenari et al. 1999).
In the present study villages in each project category were matched
geographically, but there was insufficient data to match them in a more rigorous manner.
A nonexperimental design was used instead.
Several nonexperimental approaches are possible. One way comes from the Pitt
et al. (1993) study of Indonesian poverty programs referred to in Section 1. Because
those programs were intentionally located in the poorest areas, purely cross-sectional
analysis would have suggested mistakenly that the programs actually increased poverty.
More formally, the simple approach that yielded the incorrect finding was as follows:
Y=a+bW +cX+e (1)
where the outcome (Y) is a function of the watershed project treatment effect (W) and
other determining factors (X), a is the intercept and e is the error term. This approach is
valid if the other factors (X) include all possible determinants of treatment effect and if
program placement is independent of treatment effect. Since this is unlikely to be the
case, the estimated effect of the treatment is likely to be biased. In the case analyzed by
Pitt et al. (1993), the project coefficient b in equation 1 had a negative coefficient; i.e.,
the analysis suggested incorrectly that the project contributed to poverty.
Pitt et al. approached this problem by estimating the change between pre-project
and post-project poverty conditions as a function of the change in hypothesized
determining factors, such as demographic conditions and access to services and markets. In
this model, the nonproject determining factors are grouped into socioeconomic variables
(X) that vary both spatially and temporally, and environmental variables (E) that vary
across villages but are fixed over time. The relationship can be expressed as follows'2:
AY = aAW + bAX + cE +Ae (2)
This approach isolates the changes associated with the project, eliminating the
bias associated with the influence of pre-existing conditions on both the program
placement and the outcome. However, it presumes availability of panel data (containing
conditions both before and after the program), and it presumes that sufficient change
occurred in the socioeconomic variables X to estimate them. In the present study, over
the ten-year study period many socioeconomic variables in question, such as
infrastructure conditions and the distance to markets and services, did not change in most
villages. As a result, many variables contained mainly values of zero, with insufficient
variation within the sample to perform econometric analysis. Wherever possible,
variables are expressed in terms of the change during the project period, but for most
explanatory variables the value at the start of the project period is used.
Another way to control for endogeneity of program placement in estimating
program effects is through instrumental variables. A variety of two-stage models for
estimating treatment effects or sample selection bias provide models for this approach
(Maddala 1993; Greene 1990). One equation yields the predicted probability that any
given case is selected (or self-selects) for treatment under a given program. Then, in a
two-stage model, another regression estimates the outcome in question, replacing the
endogenous treatment variable W with its predicted value eliminating the
endogeneity. In this case the model is as follows:
W = a + bX + cZ + e (3)
12 In this specification W may be considered as program expenditure in a given
location, so AWis the change in program expenditure between two points of time. W
could also be specified as project dummy variables, in which case Wwould replace AW
in equation (2).
Y=f+g +hX +e (4)
where X is a set of variables correlated to both the outcome Y and the placement of
project treatment W, and Z is a set of variables that affect Wbut not Y.
In the present context, equations (3) and (4) can be written more specifically as:
W= a + bV+ cZ+ e (5)
Y=f+g +hV+iH+jP+e (6)
where: Wis a categorical variable indicating one of five watershed project categories; is
the predicted probability that the project falls in each watershed project category; Yrepresents
outcomes defined in terms of the performance indicators introduced in section 3; Vis a set of
village-level explanatory variables; H is a set of household-level variables; and P is a set of
plot-level explanatory variables. Both H and P are omitted from village-level analysis.
Equation 5 is a multinomial logit model because W is categorical. Equation 6
takes different forms depending on the nature of the performance indicator in question;
these variables may be continuous, binary or ordinal. In most of the models equation 6 is
an ordinal logit model, in some it is a binary probit, and in a few it is a tobit or an
ordinary least squares regression. In all of these cases, the models are adjusted for the
use of complex survey data (Stata 1999).
A remaining shortcoming of the model is that, for technical reasons, the standard
errors could not be corrected for the fact that predicted values were used in the
regressions. The author is not aware of formulas to correct the standard errors for the
complex two-stage regressions used in the analysis. Bootstrapping was not justifiable
due to the small number of observations per stratum. Pender and Scherr (1998) faced this
same problem; they examined the robustness of their findings by comparing their
regression results using actual vs. predicted values. This study follows the same
approach, and the findings are discussed below.
Qualitative investigation took the form of detailed, open-ended discussions,
mostly at the group-level with people from different interest groups. The findings from
this work helped identify some of the questions posed in the quantitative analysis, and it
also helped interpret the findings. This study was primarily quantitative, so the
qualitative data played mainly a supporting role. In a few cases data limitations
prevented the quantitative analysis from yielding any useful information, so the
qualitative analysis became the sole source of insight from the fieldwork. However, time
constraints limited the scope of the qualitative investigation to less than would be ideal.
In particular, it would have been desirable to engage in a more thorough qualitative
investigation after having analyzed the quantitative data.
Qualitative data were recorded in written notes and yielded a variety of forms of
data. Some findings from these interviews could be translated into numeric data, while
others helped to explain the specific problems that people faced or the ways that projects
affected them. Findings from these interviews are presented in this report alongside those
from quantitative analysis as a means of providing additional insight.
5. HOW PROJECTS CHOOSE WHERE TO OPERATE
As explained in Section 4, how a project selects villages in which to work can
have a major impact on what it can achieve since, as introduced above, a variety of
conditioning factors can have a strong influence on people's incentives to invest in land
improvement. This section reviews each project's published site selection criteria and
then examines the data to characterize villages under each project.
PROJECT SITE-SELECTION GUIDELINES
DPAP: The pre-1995 DPAP focused on small microwatersheds located within
predominantly rainfed villages with relatively little irrigated area. The irrigation
threshold varied depending on the village's average annual rainfall; in villages with over
1125 mm average annual rainfall the DPAP could operate if irrigated area was less than
10%, whereas if rainfall was less than 750 mm irrigated area could be up to 20%.
Jal Sandharan: The Jal Sandharan project selected villages according to four
conditions.13 First, they would be selected if COWDEP had operated there and
completed more than 50% of the watershed development work. This would enable
treatment of the entire watershed to be completed through the Jal Sandharan program
within five years. Second, they should have a scarcity of drinking water. Third, they
should be located in a taluka (a subdistrict administrative unit containing up to 200
villages) with scarce groundwater (as designated by the state Groundwater Survey and
Development Agency). And fourth, they must lie outside of the outside of the command
of a canal irrigation project.
World Bank Pilot Project: Unlike the other projects in the study that work in one
or two villages at a time, the Pilot Project worked in very large, contiguous areas of about
25,000 ha with at least 750 mm average annual rainfall and little irrigation. Villages in
the core watershed area were treated in their entirety, whereas those in the periphery
typically lie partly in the watershed and partly outside. In that case only the part of the
village lying inside the watershed was treated.
NWDPRA: The NWDPRA operates in areas with less than 30% irrigation, with
no criteria concerning average annual rainfall. Preferably sites should be located in the
upper reaches of the local macrowatershed. Project sites should be close to the taluka or
block headquarters in order to facilitate supervision by officials based at headquarters,
and close to markets so that "farmers from nearby areas can assemble and see the process
and feel the impact of the project interventions (GOI 1991a)." Project sites should be
located on the main road, easily accessible to government officials and other visitors. To
quote the guidelines, "Just a pause on the road would give an opportunity to have a bird's
eye view of the project area. This will ensure visual impact on intentional and
unintentional visitors (GOI 1991 a)."
NGOs: NGOs all have their own guidelines, but virtually all of them stress
working in poverty-stricken areas, often inhabited by tribal groups. MYRADA's
13 Where DPAP funds are used its criteria are also followed.
guidelines are instructive. It operates in remote, unfavorable areas, usually in the border
areas of a state, far from the state capital, that are relatively neglected by state-level
development programs. They have the worst land, the worst infrastructure, and the least
well-off inhabitants. The typical MYRADA village has practically opposite
characteristics of those in NWDPRA project sites described above.
AGY: Any village is eligible for participation in the AGY if it is located in a
drought-prone area of Maharashtra, with no more than 30% cultivated area under irrigation.
The villagers must meet in the Gram Sabha (assembly of all adults in the village) to pledge
to accept and abide by the five social principles listed above. 70 percent of the Gram Sabha
must agree to participate before the application can go through.
IGWDP: Villages are selected for the project should be in a drought prone area
with less than 20 percent area under irrigation and overall water scarcity; they should lie in
the upper part of a macrowatershed and have noticeable erosion, land degradation and
resource depletion problems. Village boundaries should coincide to the greatest extent
possible with watershed boundaries, and the topography should offer good opportunities
for water harvesting. Villages should be predominantly poor with a high proportion of
scheduled castes and scheduled tribes in the population, and land holdings within selected
villages should be relatively equally distributed. Beyond this, the villagers should commit
to the social conditions outlined above. The key factor is that villagers should demonstrate
their capacity for collective action and their concern for resource conservation. Finally, the
village is represented by a village watershed committee that is selected on the basis of
consensus in the Gram Sabha (NABARD 1995; Farrington and Lobo 1997).
Comments on the Site Selection Criteria
The site selection criteria reveal a great deal about each project's orientation. All
of the projects share a bias towards working in areas that are less favored
agroclimatically, although it is stronger in some than others. Agroclimatic conditions are
the most lenient in the NWDPRA due to its interest in developing rainfed agriculture.
The NWDPRA is also the only project that does not seek to work in villages that
are least well off socioeconomically. Selecting the most easily accessible villages reveals
the NWDPRA's orientation toward planning and supervision by people from outside the
village, as well as an optimistic view about the process of dissemination of project
benefits. More subtly, the approach also leads to an apparently unintentional bias in
selection of project sites towards more densely populated areas with better access to
transport and markets. In accordance with the conceptual framework outlined in the
introductory section, these conditions may well be especially favorable for the promotion
ofrainfed areas. In this sense the project's technical interventions may complement other
features of the project sites.
The AGY and IGWDP reveal the greatest focus on social discipline. In many
respects the villages under this project are self-selected for collective action. Many of
them practiced the required social restrictions prior to the onset of the program. Villages
that are not prepared to ban grazing and tree-cutting or to practice shramdan shy away
from these programs. The IGWDP takes the additional step of selecting only villages
with topography favorable for constructing water harvesting structures.
Another point about selecting villages for success is that in the start-up phase of the
IGWDP, which is covered under this study, only well-established NGOs were selected that
were already familiar with the community in which they initiated the project. That means
that they were already working there prior to the start of the IGWDP, and many ongoing
activities were simply brought under the flag of the IGWDP. A similar situation holds for
the AGY, where many project villages are led by disciples of Anna Hazare and tried to
follow his development philosophy even before the project began.
ANALYSIS OF DETERMINANTS OF PROGRAM PLACEMENT
This section investigates how published site-selection guidelines translate into
actual program placement in the study villages. As mentioned above, it focuses on
Maharashtra due to data limitations in Andhra Pradesh.
Table 13 presents mean values of variables that describe the villages in
Maharashtra covered by the village survey. The data suggest that the projects do in fact
follow the principles laid out in their guidelines. NGO and NGO-government collaborative
projects tended to have the lowest levels of infrastructure in the pre-project period, while
NWDPRA villages had the highest. Conditions in nonproject villages were similar to those
under the NWDPRA. All of the NGO-government collaborative project villages practiced
shramdan, as did 75% of the NGO villages. Less than half of the remaining villages
practiced shramdan. None of the projects had a significantly lower percentage of area
irrigated than the nonproject villages, but this is because virtually all villages in the study
area have relatively little irrigation, and also because the government projects target their
work based on the level of irrigation at the taluka level, not the village level.
As mentioned above, nearly all projects take advantage of work done by the
COWDEP project in the 1980s, although only Jal Sandharan advertises this fact in its
published guidelines. Only three project villages in Maharashtra-one each under the
NWDPRA, DPAP and NGO categories-were not previously treated under COWDEP.
For NGOs, selecting COWDEP villages is sensible because technical work that was
already undertaken can be made more productive by developing complementary social
institutions. At the same time, it makes it somewhat difficult to determine how much of
the success of the program stems from the project's current work and how much depends
on earlier watershed structures built under COWDEP. It may have implications for the
time frame required for watershed development interventions when the project expands
to areas beyond those covered by COWDEP.
Table 13: Mean values of selected Maharashtra village characteristics in 1987, by
project category ab
Average annual rainfall
% Irrigated area
% of villages located in the upper reaches
of the macrowatershed
% of villages located in the middle reaches
of the macrowatershed
% of villages located in the lower reaches
of the macrowatershed
% of villages located on a highway
% of villages located on a paved road
% of villages located on a good unpaved
% of villages located on a bad unpaved
% of villages located on a bullock cart path
Distance to nearest regulated market'
Distance to taluka headquarters
Distance to district headquarters
Distance to large cityd
Distance to nearest bus stop
Number of extension agent visits per
Distance to industrial unit
Distance to bank
Population density (from 1991 census)
% of villages with a public health service
Distance to nearest public health service
% of houses with an electrical connection
% of villages with electricity to power
% of villages that had informal credit
% of villages that had adequate drinking
water (i.e. did not need tankers to deliver
% of villages containing some common
(government revenue) land
% of villages that had some land coming
under the Forest Department
Overall adult literacy rate, % (1991 census)
Male literacy rate, % (1991 census)
Female literacy rate, % (1991 census)
20 41 17
10 12 8
60 12 25 0.0
70 35 50 43 71
60 59 33 57 71
80 71 83 86 65
20 35 25 14 41
Table 13 continued
Village characteristic All NWDPRA JS NGO AGY-IG No project
Distance to nearest veterinarian 5.4 3.2 5.5 5.9 6.9 4.9
% practicing shramdan (voluntary 54 20 35 75 100 41
% in which fuel was available from 40 33 40 50 63 25
government revenue land'
% in which grass fodder was available 83 83 90 50 88 83
from government revenue landg
% in which tree fodder was available from 40 50 40 50 50 25
government revenue landg
Notes:a Analysis of variance statistics are reported for continuous variables with significance
level p <.10.
bKruskal-Wallis statistics for ordinal categorical variables are reported for variables with
significance level p <.10 (Agresti 1997).
CDistance to nearest regulated market: F = 2.80, 4 df, p < .03.
dDistance to nearest large city: F = 2.84, 4 df, p < .05.
eDistance to nearest bank: F = 2.07, 4 df, p < .1.
fWhether shramdan (community voluntary labor) is used: chi-square = 21.9, 4 df, p < .001.
gThese measurements cover only those villages that had government revenue land in both 1987
Another interesting observation in Table 13, not revealed by published guidelines,
is that in all the project categories, a smaller percentage of villages contain common land
(owned by the government) than in nonproject villages. The difference is particularly
large in NGO villages; only 33% contain government revenue land compared to 71% for
nonproject villages and 57% overall. Such land usually lies in the upper watershed and is
used as grazing commons; as discussed in Section 1, watershed projects aim to restrict
access by grazing animals to this land. Throughout India, government revenue is in a
notoriously degraded, open access state. Accordingly, the organizational requirements of
watershed development may be significantly less complicated in villages without
government revenue land. Another kind of common land is owned by the Forest
Department, which uses its own resources to restrict access. More project villages than
nonproject villages contain Forest Department land, most likely because they are located
in more hilly terrain with more forested area.
Econometric Analysis of the Determinants of Project Placement
A multinomial logit model is used to examine in more detail the determinants of
which project category a particular village falls into. The dependent variable is the
categorical project variable covering the five categories found in Maharashtra: National
Watershed Development Project for Rainfed Areas (NWDPRA), Jal Sandharan, NGOs,
and Adarsh Gaon Yojana and Indo-German Programme (AGY/IGWDP-combined into
one category), and nonproject.
Explanatory variables: Conditions prevailing in 1987, before the projects began,
represent the potential determinants of a village's selection by a given project. Altitude
range (the difference between the highest and lowest points, in meters) is important since
many projects seek to work in areas with high potential for water harvesting.
Infrastructure variables include the distance to taluka headquarters, the population density
in 1990,14 percent area irrigated, adequacy of drinking water availability, distance to
market, distance to the nearest bus stop, and distance to the nearest public health center.
Other infrastructure variables are omitted due to high correlation with those included.15
The existence of an old COWDEP project in the village is omitted from the analysis
because it perfectly predicts the existence of a current project, making the multinomial
logit analysis infeasible. Variables representing social conditions and social institutions
are whether the village practiced shramdan and the number of communal groups. The
literacy rate was considered but excluded because it is highly correlated with some of the
Results: The findings of the multinomial logit analysis are presented in Table 14.
The analysis supports most of the descriptive findings about project selection and raises
some additional points. With nonproject villages as the base category, the analysis shows
14 Population data come from the 1990 census (GOI 1991b); they are not available
15 An effort was made to build an index of infrastructure quality, but it had limited
Table 14: Determinants of project category in Maharashtra"
Multinomial logit regressions (standard errors in parentheses)
Variable Project category
Min of Rural NGO/govemment
Min of Agriculture Development NGO collaboration
Distance to nearest bus stop in 0.83 -0.16 0.16 -0.34
1987 (km) (.34)** (0.27) (0.32) (0.29)
Paved road in 1987 (dummy) 0.29 -1.58 0.41 -2.49
(1.27) (1.63) (1.11) (1.53)
Whether the village contained -0.32 -2.10 -4.96 -1.16
government revenue land, 1987 (1.16) (1.22)* (1.17)*** (0.88)
Number of communal groups in 1.18 0.76 0.85 0.13
the village (0.25)*** (0.29)** (0.30)*** (0.35)
Altitude range ('00 meters) 3.34 1.93 2.44 2.16
(1.02)*** (1.00)* (1.06)** (1.34)
Distance to taluka headquarters 0.21 0.01 0.35 -0.03
(km) (0.05)*** (0.05) (0.43) (0.04)
Population density in 1990 ('00 3.71 0.88 -1.81 -0.59
persons/sq km) (0.82)*** (1.76) (-1.43) (0.88)
Percent area irrigated in 1987 2.90 -2.39 8.29 1.94
(3.28) (5.76) (3.55)** (4.52)
Whether the village had sufficient 3.31 -1.35 0.26 0.93
drinking water in 1987 (dummy) (1.38)** (1.27) (1.54) (1.49)
Distance to nearest public health -0.38 0.17 0.18 0.33
center, 1987 (km) (0.15)** (0.15) (0.15) (0.14)**
Distance to market for -0.15 0.23 0.34 0.10
agricultural inputs in 1987 (km) (0.11) (0.15) (0.16)** (0.13)
Village practiced community -2.01 -1.31 1.57 8.42
voluntary labor in 1987 (dummy) (1.10)* (1.51) (1.57) (2.35)***
Area of the village ('00 ha) 0.17 1.29 0.09 0.30
(0.13) (1.34) (0.13) (0.13)**
Approx. % of households with at -0.10 -0.06 -0.10 0.09
least one seasonal migrant, 1987 (0.03)*** (0.04) (0.06) (0.03)***
% of inhabitants of low caste 0.047 0.08 0.12 -0.03
(0.025)* (0.03)*** (0.03)*** (0.06)
Note: a Reference category is control (no project); variables reflect values in the pre-
project period. 70 observations. Model is not corrected for choice-based sampling, i.e.
that the sample is stratified on the dependent variable. Coefficients and standard errors
are adjusted to account for sampling weights, stratification and finite population size. *,
**, and *** indicate statistical significance at the 10%, 5% and 1% level, respectively.
All projects have a greater range in altitude between the highest and lowest point
in the village compared to nonproject villages, and this difference is significant for all
except the NGO/government collaborative projects. This is to be expected since hilly
areas are most suited for water harvesting.
The AGY/IGWDP villages were significantly more likely to practice shramdan in
1987. NWDPRA villages actually practiced significantly less shramdan than nonproject
villages; the reasons for this finding are not known. NWDPRA, Jal Sandharan and NGO
villages all had more communal diversity and more people of low caste than nonproject
villages, and the latter is consistent with published guidelines. The AGY/IGWDP, on the
other hand, had no significant differences from nonproject villages in communal diversity
and low caste people. If the analysis is conducted using the AGY and IGWDP as the base
category (not shown) the communal diversity and population of low caste people are
significantly lower than other project categories. The IGWDP requires consensus-based
decision making, which may be easier with communal homogeneity, and the two projects
require a ban on open grazing and tree-cutting, which may be more difficult for poor, low
caste people to accept because they rely on products from the commons for their
livelihoods. NGO and Jal Sandharan villages were significantly less likely to contain
government revenue land, possibly suggesting that these projects sought to operate in such
villages to make their work easier. NWDPRA and AGY/IGWDP villages also were less
likely to contain government revenue land, but the difference is not statistically significant.
NWDPRA villages were likely to have significantly fewer seasonal migrant
workers, while AGY/IGWDP villages were likely to have significantly more. A higher
number of migrant workers can be an indication of poor economic conditions locally.
AGY/IGWDP villages are also significantly larger in area; the reason for this difference
is not clear.
NWDPRA villages were likely to be more densely populated while others were
likely to be less densely populated than nonproject villages, but this difference is
significant only for the NWDPRA villages. This is consistent with the NWDPRA's
published guidelines, which call for working in more accessible, visible villages. Again,
this probably reflects a nonrandom selection process. NWDPRA villages are also closer
to public health clinics and markets, though only the former is significant. NGO villages,
on the other hand, were significantly likely to be located further from markets and the
taluka headquarters, and AGY/IGWDP villages were significantly farther from the
nearest public health office.
Finally, only Jal Sandharan/DPAP villages were more likely to have a drinking
water shortage, consistent with the project's mandate, but the difference was not
6. NATURAL RESOURCE MANAGEMENT AND PRODUCTIVITY ON
Most nonarable lands in the study region are managed either as government revenue
land or Forest Department land. While the latter is managed by the Forest Department and
access is heavily restricted (at least in principle), government revenue land is typically in a
state of open access to all users. Protecting it requires village-level management institutions
based on widespread commitment to improvement of this resource.
Watershed projects seek to develop nonarable lands for a variety of reasons. In
projects oriented toward water harvesting, the ultimate reason is that nonarable lands are
typically in the upper reaches of the watershed, which act as the catchment area for water
harvesting structures downstream. If the upper reaches are poorly maintained, erosion will
silt the water harvesting structures, rendering them useless. So developing and protecting
nonarable lands is a prerequisite to the primary objective of raising the water table.
Developing nonarable lands also has direct benefits, particularly increasing the
long-term availability of products such as fuel and fodder that historically were supplied
by these lands. Soil and water conservation trenches are dug to concentrate water and
soil, with trees and grasses planted in the trenches. In the early years after planting, the
common lands must be strictly protected against grazing so that plants can establish.
After that, they can supply a steady stream of fodder and fuel as long as grazing and
harvesting are restricted.
The typical scenario on the common lands in rural India has been one of gradually
declining productivity due to overexploitation, which in turn resulted from institutional
arrangements that were inadequate to encourage people to protect and develop these
lands (Singh 1997). Historically, management of common lands followed at least three
different patterns (Gadgil and Guha 1992). In some places they were accessible to all, with
insufficient pressure on resources to lead to severe degradation until the last several
decades. In others, management was enforced by powerful landowners such as zamindars,
who acted as "gatekeepers" to make sure that the common lands were not overexploited
(Gadgil and Guha 1992; Bentley 1984). While this system was good for the condition of
the land, it was inequitable, with benefits dominated by the landlords. In a third kind of
situation, democratic village-level institutions resulted in sophisticated, equitable ways of
sharing both rights and responsibilities for managing common lands (Agarwal and Narain
1989). Although this latter situation is sometimes presented as the historical norm in rural
India, there is little evidence that it prevailed beyond a minority of villages. It is often said
anecdotally, however, that such systems are still common in tribal areas.
WHAT THE PROJECTS DO
The idea behind most current watershed project efforts on common lands is to use
a combination of technical and institutional means to move the supply of products such as
fuel and fodder from a low-level equilibrium to a high-level one. In addition to installing
soil and water conservation works and planting vegetation, most projects today seek to
develop institutions for managing government lands based on principles of common
property resource management. They typically encourage villagers to establish users'
committees that are expected to develop and enforce management plans in a way that
satisfies the needs of every interest group. In short, they try to create the kind of ideal,
democratic arrangement mentioned above.
As described in Section 2, the AGY, IGWDP and NGO projects all devote
relatively large efforts to social organization, and particularly to mechanisms to reduce
pressure on common lands. The IGWDP and AGY, for example, only work in villages
that promise to ban grazing and cutting trees. All of these projects also promote "social
fencing," or social mechanisms to achieve protection of the common lands. Their efforts
may include encouraging everyone to comply with restrictions on the commons, devising
arrangements to guard them if necessary, etc. The NWDPRA and Jal Sandharan
guidelines cover the same issues but in a more cursory way, and unlike the other agencies
they do not make it a central component of their work. The fact that none of their staff
members have any training related to social organization helps ensure this. The
NWDPRA and Jal Sandharan may contract this part of the work out to NGOs, but only
for a few weeks, after which social organization is expected to take care of itself.
An important feature of project investment on common lands is that it is entirely
subsidized. Some projects require a 10% in kind contribution (in terms of donated labor),
but this is more than offset by the fact that the projects pay above the market wage. At
the very least, donating 10 percent of one's labor to gain a day of employment is a break-
even proposition. The AGY and IGWDP, along with some NGOs, obtain a local
contribution to developing the common lands through the practice ofshramdan.
Approximately 16% of the costs should be contributed through shramdan, with half the
value returned in cash to a village development fund for maintaining watershed
The indicators presented in this section help identify of the extent to which
various types of watershed projects have succeeded in developing and protecting
common lands. The section on common lands is divided into discussions of four sets of
indicators. One type of indicator is the introduction of social fencing institutions to
encourage protection of the commons, and the other three are rough measurements of
natural resource conditions, including erosion and conservation status of the main
drainage line, erosion status of nonarable lands, and changes in availability of fodder and
fuel. The analysis of the condition of the drainage line covers the 64 Maharashtra
villages that have a main drainage line, while the remaining analysis covers the 40
villages that contain government revenue land, since this is the common land over which
villagers have the authority to manage as they please.
SOCIAL FENCING INSTITUTIONS
The most common social fencing institutions are bans on grazing and cutting
trees. Many villages impose such bans in name, but whether or not they are adhered to
may be another question. Investigators who stayed in each village for 2-4 days sought to
distinguish whether grazing bans and tree cutting bans were active or just in name only.16
Table 15 lists the number and percentage of villages in each project category that
had banned grazing, as well as the number that actually imposed penalties on offenders of
rules against grazing and tree cutting. A traditional penalty against illicit grazing, for
example, is to impound the grazing animals in the panchayat (village government) office
and release them only upon payment of a fee. Panchayats or watershed user
organizations keep records of such payments, so it is not difficult to identify whether or
not such punishment systems were enforced in 1997.
The table shows the presence of grazing and tree cutting bans for both the pre-
project period and the present. It shows that in both 1987 and 1997, banning grazing on
the commons was the exception, not the rule. Only 5 out of 40 villages (12.5%) had
banned grazing before the projects, rising to 35% after introduction of the projects. The
respective numbers for imposing punishments for illicit grazing were even less, with 5%
in 1987 and 22% in 1997. The numbers for punishing illicit cutting of trees are similar at
5% in 1987 and 20% in 1997. Two details in the table are particularly interesting. First,
even some of the nonproject villages have imposed grazing bans; clearly this is not
something that necessarily requires a watershed project. The second is that while none of
the AGY and IGWDP villages had imposed bans or penalties in 1987, by 1997 50 percent
of them had done so. They are the only category of villages with a statistically
significantly higher percentage than the nonproject villages. At the same time, the 50
percent figure is low in comparison with these projects' target of universal compliance
with bans on grazing and cutting trees.
No regression analysis is performed on the determinants of banning grazing and
tree cutting, because so few villages actually imposed these restrictions.
16 For example, if a pasture is protected against grazing there should be no traces
of cow dung on the ground.
Table 15: Number and (percentage) of villages with restrictions on access to
common (government revenue) lands, by project category, 1987 (pre-project) and
Type of restriction NWDPRA DPAP/Jal NGO NGO/Govt No project Total
Open grazing restricted, 1987 1 (17) 2 (20) 0 (0) 0 (0) 2 (17) 5 (13)
Open grazing restricted, 1997 2 (33) 4 (40) 1(25) 4 (50) 3 (25) 14 (35)
Open grazing restriction 1 (17) 2 (20) 1(25) 4 (50) 1(8) 9 (23)
introduced after project began
Punishment for open grazing, 0 (0) 1(10) 0 (0) 0 (0) 1(8) 2 (5)
Punishment for open grazing, 1(17) 2 (20) 0 (0) 4 (50) 2 (17) 9 (23)
Punishment for open grazing 1(17) 1(10) 0 (0) 4 (50) 1(8) 7 (18)
introduced after project began
Punishment for cutting trees, 1(17) 1(10) 0 (0) 0 (0) 0 (0) 2 (5)
Tree cutting restricted, 1997 1(17) 2 (20) 1(25) 4 (50) 0 (0) 8 (20)
Tree cutting restriction 0(0) 1(10) 1(25) 4(50) 0(0) 6 (15)
introduced after project
Notes: a This table is based on the 40 Maharashtra villages that had government revenue
land both in 1987 and 1997.
b Figure in parentheses is percentage of category total. Category totals are NWDPRA: 6,
Jal Sandharan: 10, NGO: 4, AGY/IGWDP: 8, No project: 12.
SKruskall-Wallis test shows that differences across categories are significant for
introduction of punishment for cutting trees (Chi-square = 11.1, 4 df, p < 0.027). Other
differences are not statistically significant.
EROSION AND CONSERVATION STATUS OF THE MAIN DRAINAGE LINE
By definition, the main drainage line is where runoff water concentrates, so it is
highly vulnerable to soil erosion. The drainage line is also usually on government land.
As mentioned above, government land tends to be managed poorly compared to privately
operated land, so the drainage line faces management challenges borne of both
biophysical and social causes. (Out of the 64 villages, 40 have government revenue land
and most of the remainder have Forest Department land.)
Field investigators trained in soil survey methods conducted a transect of the main
drainage line in each village, making several visual observations of its condition and the
extent of erosion on its banks. The transect was divided into segments of equal length
(100 meters), and investigators made an assessment of its overall condition. This is
determined by whether it appears to be under control and not expanding into adjoining
fields; the extent and condition of bunds on the sides of the drainage line, and the extent
of breaches in the sides of the drainage line. Each segment was assigned one of three
possible scores for each of these characteristics: 3 for "good or high," 2 for
"intermediate," and 1 for "poor or low." Overall scores for each village were then
calculated by taking the simple average of all the segment scores. This visually-based
scoring system was the best that could be achieved given the resources available to the
project and the lack of existing data. While it is obviously subjective, the scores should
be consistent across villages within each state because only two teams of people
conducted the transects and the members rotated regularly in an effort to make sure they
all used the same standards. Also, the 1 to 3 scale, while reducing the ability to make
fine distinctions across observations, reduces the likely variation in scoring standards
across data collection teams.
Drainage line transect scores are analyzed in both tabular and econometric form.
First, average values of the drainage line scores are shown in Table 16.17 This table
shows that the kind of project operating in the village has a small but statistically
significant effect on the drainage line, with AGY/IGWDP villages having the best
average score and nonproject villages having the worst. (Note that even the best average
score is only 2.00, indicating intermediate condition.) The same table shows stronger
evidence that the duration of the watershed project and the percentage of the village
covered by a project have a positive effect on the condition of the drainage line,
regardless of the project category. This appears to suggest that some kind of watershed
activity is better than none in determining the condition of the drainage line.
Drainage line scores are also analyzed through multivariate econometric analysis,
which is used to identify village- and project-level factors that determine the drainage
line scores. Each village's score represents the mean value of scores for all the 100-meter
17 Strictly speaking, the scores are ordinal, not cardinal, but average scores are
shown here for ease of presentation and interpretation. An average score of 2.00 means
that average condition is intermediate; less than 2.00 means the average score is low, and
greater than 2.00 means the average score is high.
segments of the drainage line, so the scores take continuous values ranging from 1 to 3.
A Tobit model is appropriate in this case.
Table 16: Drainage line transect scores at the village level by project category and
Village characteristic Average score for condition of the drainage line
All villages 1.70
Jal Sandharan 1.60b
No project 1.40b
Total number ofyears under old and new watershed projects
0 (no project) 1.40c
13 or less 1.60c
14 or 15 1.82c
Percentage of village covered by the project
0 (no project) 1.40d
Notes:a Possible drainage line scores are 1, 2 and 3. Strictly speaking, they are ordinal
categorical variables, not cardinal, but average scores are shown here for ease of
presentation. An average score of 2.00 means that average condition is intermediate;
less than 2.00 means the average score is low, and greater than 2.00 means the average
score is high. The Kruskal-Wallis test for ordinal categorical variables shows that the
following variables are significant at 10%:
b Overall condition of the drainage line varies significantly among project categories
(chi-square = 8.7, 4 df).
c Overall condition of the drainage line varies significantly by number of years under
old and new projects (chi-square = 8.0)
d Overall condition of the drainage line varies significantly by percentage of village
covered by project (chi-square = 6.0, df =2)
Explanatory variables: 1987 variables are used in the model of determinants of
the condition of the drainage line, since its stabilization through soil and water
conservation measures is a long-term process. 1997 values would not be the correct
explanatory variables to explain the effectiveness of conservation measures that took
place prior to 1997. Agroclimatic variables include the village's altitude range, which is
reflected in the course of the drainage line and determines susceptibility to erosion.
Average annual rainfall also determines susceptibility to erosion, but it is highly
correlated to altitude range (>.6), with more hilly areas having higher rainfall, so rainfall
is omitted from the model.
Social institutions and characteristics include the number of different communal
(caste and religious) groups in the village, the percentage of households in the village that
derive their income primarily from herding sheep, and a dummy variable indicating
whether the village contains government revenue land. It is hypothesized that a higher
proportion of shepherds will bring increased resistance to protecting the commons and
thus poorer condition of the drainage line. The practice of shramdan is excluded because
it is highly correlated with the predicted probability that the AGY or IGWDP operates in
the village. The presence of government land almost certainly indicates that the drainage
line runs through common land, which is more difficult to manage.
Economic factors include infrastructure, such as the presence or absence of a
paved road, distance in km to the nearest bus stop, distance in km to the taluka
headquarters, population density (inhabitants per sq km), and the percentage of people in
the village who earn most of their income from a source other than cultivation, livestock
or agricultural labor. Population density, infrastructure and access to markets can
increase the pressure on natural resources, but they can also raise the returns to better
land management. Off-farm income also has an ambiguous effect; it can help finance
land improvement or it can lead people to focus their interests elsewhere, making them
less willing to participate in social action to develop the village's natural resources
(Gebremedhin et al. 2000; Pender and Kerr 1998). Finally, as discussed above, the
project inputs are represented by predicted values of the dummy variables for each
project category and the average project expenditure per hectare in village as a whole.
Results: Table 17 presents the results for two cases, once in which the predicted
project category is interacted with (multiplied by) the project expenditure per hectare, and
Table 17: Determinants of drainage line erosion status
Variable Coefficients in model 1b Coefficients in model 2c
Whether the village contained government 0.42 0.42
revenue land in 1987 (dummy) (0.12)*** (0.11)***
Altitude range ('000 meters) -4.32 -5.85
Distance to nearest bus stop in 1987 (km) 0.04 0.04
Paved road in 1987 (dummy) 0.15 0.17
Population density in 1990 ('00 persons/sq km) 0.06 0.07
Distance to taluka headquarters ('0 km) -0.06 -0.06
% inhabitants working primarily in -0.01 -0.01
nonagricultural sector (0.01) (0.01)
% inhabitants working primarily as shepherds -0.07 -0.06
Mean project expenditure per hectare ('000 Rs) 0.06
DPAP/Jal Sandharan 0.007
Mean expenditure per ha in NWDPRA village 0.10
('000 Rs) (0.05)**
Mean expenditure per ha in DPAP/Jal Sandharan village 0.07
Mean expenditure per ha in NGO village 0.17
Mean expenditure per ha in AGY/IGWDP village 0.27
Notes: a64 observations. Possible transect scores range from 1 to 3. Coefficients and standard errors are adjusted to
account for sampling weights, stratification and finite population size. *, **, and *** indicate statistical significance at
the 1%, 5% and 10% level, respectively. Predicted values based on the multinomial logit regression in Table 14 are
used for the four project category variables. Standard errors are not adjusted for use of predicted values.
b Model 1: mean expenditure per hectare and project category are expressed as separate variables. F(13,42)=6.64
(p>.0000); R2 = 0.38.
SModel 2: mean expenditure per hectare is expressed separately for each project category. F(12,43)=6.20 (p>.0000); R2
once in which these variables are specified separately. There are only 64 observations
because 6 villages have no main drainage line. The models have highly significant F-
statistics, but both R2 values are about 0.38, so the extent of variation explained by the
model is not high. When expenditure and project category are specified separately,
expenditure per hectare is highly significant, but only the AGY-IGWDP project category
variable is.18 The dummy variable indicating the presence of government revenue land is
positive and statistically significant, which was unexpected. The percentage of households
earning their livelihoods as shepherds is negative and statistically significant, as expected.
Most other variables have the expected sign but are insignificant. These include the
percentage of people working outside of agriculture or livestock (-), distance to the taluka
headquarters (-), population density (+), and existence of a paved road in 1987 (+). The
latter three signs are consistent with the induced innovation hypothesis that better market
access may raise the incentives to manage land better.
When the model is respecified so that project expenditure and the project category
variables are interacted, all the project expenditure variables are positive, and all are
statistically significant except the Jal Sandharan/DPAP, which is nearly significant. This
lends support to the notion, expressed above, that all the projects are successful in
improving the condition of the drainage line. The AGY-IGWDP category has a much
higher coefficient than the other categories as well as a higher level of statistical
significance, so these projects appear to perform the best. For every rupee spent by the
AGY or IGWDP, the drainage line score rises by 0.27 on a scale from 1 to 3; for NGOs it
is 0.17; and for government projects the increase is less than 0.10. Other significant
variables remain the same as in the previous specification of the model. A very similar
result is obtained with the use of actual project dummy variables, except with a smaller
effect of each rupee spent (not presented).
EROSION OF UNCULTIVATED LANDS
Field investigators conducted a separate transect covering a route perpendicular to
the main drainage line, designed to cover a representative tract of the village's area, with
variations in soil types, slopes, and land use.19 The route was selected based on discussions
with the sarpanch, groups of farmers, and a soil map of the village where it was available.
18 Note that the standard errors have not been corrected for the use of predicted
values of the project category.
19 The straight line design of the field transect oversamples plots close to the
center of the village relative to those at the periphery, which are more likely to be hilly
The investigators delineated the transect route into separate segments whose boundaries
were defined by changes in either land use (nonarable, rainfed, irrigated), land capability
classification, or the extent of soil erosion and soil conservation measures.
This section examines the findings regarding the extent of soil erosion on
uncultivated lands in the transect, which refers to visible signs of rills and gullies. This is a
rough measure that cannot identify imperceptible sheet erosion processes, but it is
sufficient to identify any form of rill or gully erosion. Inhabitants of the study villages
accompanied the field investigators on the transect to tell them the tenure status of the land
(private, Forest Department or government revenue). The transect scoring system is the
same as in the drainage line transect, with l=low erosion, 2=medium, and 3=high erosion.
The score was recorded for each segment along with the length of the segment. Weighted
averages of the erosion score for each segment can then be summed to give aggregate
village-level scores, which can be expressed either as an overall score for the entire village,
or as separate scores for different land uses and different land capability classifications.
As shown in Table 18, the mean transect scores for uncultivated land show no
significant differences across project categories; nor are any other village characteristics
significantly associated with the erosion score in this bivariate tabular analysis. In fact
the scores are marginally better in nonproject villages than they are for each of the other
categories. One possible explanation for this is that nonproject villages are flatter than
project villages and so less susceptible to erosion. A related interpretation would be if the
project villages were intentionally selected because they had the most problems, so that
nonproject villages are in the best condition today because they were in the best condition
when the projects began.
Table 18: Erosion scores for uncultivated land from the village transect, by project
Project category Erosion score
All villages 2.29
Jal Sandharan 2.32
No project 2.15
Notes: a Possible transect scores are 1, 2 and 3. Strictly speaking, they are ordinal
variables, not cardinal, but the average scores are shown here for ease of presentation.
An average score of 2.00 means average condition is intermediate; less than 2.00 means
the average score is low, and greater than 2.00 means the average score is high.
b Kruskal-Wallis ordinal variables test shows no significant difference between project
categories or any other village characteristics
Econometric analysis is required to gain a more detailed understanding of the
determinants of soil erosion on uncultivated land. The unit of observation for the
econometric analysis is the transect segment rather than the village level average; this
allows controlling for land tenure status. The model specification accounts for the fact
that observations "clustered" within villages are not independent of each other (Stata
1999). An ordinal probit model is used because the transect scores are ordinal values of
1, 2 or 3. The observations are not weighted by segment length because that would
prohibit weighting them for sampling weights and clustering.20
The variables used in the analysis are nearly the same as in that for the
determinants of the drainage line scores. One additional variable that was not applicable
in the analysis of drainage lines is the ownership status of the segment of land in
question; a dummy variable indicates whether land is private or common (government
revenue or Forest Department land).
Table 19 displays the regression results. As in the drainage line analysis, the
analysis is conducted twice, once in which the project category variables are specified
20 When the analysis was conducted without accounting for the survey data, there
was practically no difference in the analysis when weighted or unweighted by segment
separately from the funds invested per acre, and once in which they are interacted. When
they are specified separately, all the project category variables have a negative coefficient,
indicating that the extent of erosion is reduced. The coefficient is significant only for the
NGO category, but it is nearly significant for the AGY/IGWDP category.21 This is
consistent with the finding that these projects are more successful than others in protecting
the common lands, and also that many of the villages where NGOs operate have no
common land. Expenditure per hectare is statistically significant, suggesting that the
specific project category may be less important than the fact that at least some kind of
investment takes place. The only other significant variable is population density; higher
density indicates lower erosion, which is consistent with the induced innovation hypothesis
that land will be better managed when it is more scarce. Infrastructure variables are
insignificant. Neither the property rights status of the plot nor the number of shepherds in
the village is significant, but both are nearly so and both have the expected sign.
The results are similar when the predicted project category and expenditure per
hectare are interacted. In this case expenditure under any project reduces erosion, but it
is only statistically significant for the NGOs, the AGY-IGWDP, and the DPAP. It is
nearly significant for the NWDPRA. The NGO and AGY-IGWDP coefficients have a
much greater magnitude than those of the other projects.
The use of actual project dummy variables (not shown) yields very similar results.
All project categories have statistically significant coefficients, but the degree of
significance and the magnitude of the coefficient is higher for NGOs and the AGY-
21 The standard errors are not corrected for the use of predicted project categories.
Table 19: Transect scores: erosion status of uncultivated lands'
Ordered probit regression
Variable Coefficients in model 1C Coefficients in model 2c
Altitude range ('000 meters) 0.23 0.33
Distance to nearest bus stop in 1987 (km) 0.00 -0.02
Paved road in 1987 (dummy) 0.36 0.31
Population density in 1990 ('00 persons/sq km) -0.65 -0.66
Distance to taluka headquarters ('0 km) 0.05 0.04
% inhabitants working primarily in nonagricultural sector 0.005 0.008
% inhabitants working primarily as shepherds 0.06 0.04
Whether the land is operated privately (dummy) -0.59 -0.57
Mean project expenditure per hectare ('000 Rs) -0.17
DPAP/Jal Sandharan -0.34
Mean expenditure per ha in NWDPRA village ('000 Rs) -0.20
Mean expenditure per ha in DPAP/Jal Sandharan village -0.20
Mean expenditure per ha in NGO village -0.35
Mean expenditure per ha in AGY / IGWDP village -0.45
Notes: a 174 observations from 70 villages. Possible transect scores are 1, 2 and 3. Coefficients and standard errors are
adjusted to account for sampling weights and stratification. *, **, and *** indicate statistical significance at the 1%,
5% and 10% level, respectively. Predicted values based on the multinomial logit regression in Table 14 are used for
the four project category variables. Standard errors are not adjusted for use of predicted values.
b Model 1: mean expenditure per hectare and project category are expressed as separate variables. F(13,42)=3.56,
cModel 2: mean expenditure per hectare is expressed separately for each project category. F(13,42)=3.45, p >.002.
Change in Availability of Fuel and Fodder from the Common (Government Revenue)
Information on products collected from the common lands was obtained as part of
the village-level survey. Respondents were asked in groups about what kinds of products
were available today, what kinds were available in 1987, and whether and in which direction
the quantity had changed between 1987 and 1997. The questions covered grass fodder, tree
fodder, fuel, timber, and building materials, and respondents mentioned several other
products. However, only grass fodder, tree fodder and fuel were found in more than a few
villages, so the analysis presented here is restricted to those commodities. The mean values
of responses by project category are presented in Table 20; it covers only the subset of 40
Maharashtra villages that had government revenue land in both 1987 and 1997.
The table shows no significant differences across project categories or any other
village characteristics. For grass fodder, most villages reported that there was less
available in 1997 than 1987, and this was the case for all project categories except NGOs,
which had closer to the same amount in both years. For tree fodder and fuel, most
villages reported having the same amount in both years, but more villages reported a
decline than an increase. It appears that the watershed projects have had difficulty in
raising availability of these products on the government revenue lands.
Econometric analysis is needed for more thorough examination of the determinants
of changes in access to products of the commons. Ordered probit models are used to
analyze the determinants of whether a village has more, less or the same amount of grass
fodder available from the government revenue lands. The explanatory variables are the
same as in the analysis of the condition of the drainage line, with the addition of a dummy
variable indicating whether or not grass fodder was available in 1987. As in the earlier
analyses, the model is run both with the predicted project category variables expressed
separately from the expenditure per hectare, and with them interacted.
Table 21 shows that the projects have led to reduced access to grass fodder
compared to nonproject villages. The variables for expenditure per hectare and the
AGY/IGWDP and Jal Sandharan/DPAP project categories have negative, statistically
significant signs. Where the expenditure and project category variables are interacted,
the AGY/IGWDP and DPAP variables remain significantly negative, while the other
project categories are insignificant.22 The AGY-IGWDP coefficient also has a much
22 The standard errors are not corrected for the use of predicted project categories.
tn o o C) W) W
fim >/i r) eq en
0 1 0 0) W) r0
o o' vir~'
r"0 C)1 tn n 0M
ID cn " o
en cn C' n r- N
g 0'0^g t
-t o v5.
0 o *
'dO 04 q
o c D
Z 00. m
>j Cd) *" 0
00- ^- ^
Table 21: Econometric analysis of determinants of change in availability of grass
fodder and fuel on government revenue lands
Ordered probit regression
Variable Grass fodder Fuel
Model lb Model 2c Model 1b Model 2c
Availability of grass fodder (fuel) in 1987 2.13 2.09 1.92 1.46
(0.64)*** (0.61)*** (0.55)*** (0.83)*
Altitude range ('00 meters) 3.76 3.71 0.04 0.02
(1.26)*** (0.96)*** (0.01)*** (0.11)**
Distance to nearest bus stop in 1987 (km) 0.34 0.53 -0.10 -0.08
(0.16) (0.19)*** (0.17) (0.15)
Paved road in 1987 (dummy) 0.75 0.92 -0.40 -0.36
(0.56) (0.66) (0.51) (0.42)
Population density in 1990 ('00 persons/sq km) -0.75 -1.06 -0.64 -0.36
(0.23)*** (0.51)** (0.41) (0.21)*
Distance to taluka headquarters ('0 km) 0.05 0.33 -0.06 -0.06
(0.04) (0.33) (0.04)* (0.02)*
% inhabitants working primarily in nonagricultural sector 0.11 0.10 0.04 0.04
(0.03)*** (0.04)** (0.03) (0.02)**
% inhabitants working primarily as shepherds 0.69 0.62 0.13 0.12
(0.19)*** (0.17)*** (0.18) (0.13)
Mean project expenditure per hectare ('000 Rs) -0.43 -0.71
NWDPRA 0.90 7.56
DPAP/Jal Sandharan -2.40 5.15
NGO 7.95 7.75
AGY/IGWDP -5.23 2.26
Mean expenditure per ha in NWDPRA village ('000 Rs) 0.06 0.60
Mean expenditure per ha in DPAP/Jal Sandharan village -0.89 0.32
Mean expenditure per ha in NGO village 1.35 -0.71
Mean expenditure per ha in AGY/IGWDP village -2.04 -0.33
Notes: '40 observations. Possible transect scores range from I to 3. Coefficients and standard errors are adjusted to account for
sampling weights, stratification, and finite population size. *, **, and *** indicate statistical significance at the 1%, 5% and 10%
level, respectively. Predicted values based on the multinomial logit regression in Table 14 are used for the four project category
variables. Standard errors are not adjusted for use of predicted values.
b Model 1: mean expenditure per hectare and project category are expressed as separate variables.
' Model 2: mean expenditure per hectare is expressed separately for each project category. F-statistics: Fodder: F(13,19)=6.27 (Model
1); F(13,19) = 6.88 (Model 2); Fuel: F(13,19)=2.94, p>0.02 (Model 1); F(13,19) = 2.06, p > 0.08.
This finding is consistent with those presented above showing that the AGY and
IGWDP are particularly successful in restricting access to common lands and reducing
erosion in the drainage line and pasture lands. Improving the condition of these lands
requires restricting access to them, and Table 21 suggests that access is in fact still
restricted. Several other variables are significant as well. Population density has a
negative sign, while the variables with positive signs include availability of grass fodder
in 1987, altitude range, distance to the nearest bus stop in 1987, percentage of households
working primarily outside of agriculture, and percentage of households working
primarily as shepherds. The highly significant, strongly positive coefficient for
shepherds is again consistent with the finding that it was more difficult to reduce erosion
in the villages with the most shepherds, presumably because access restrictions were
difficult to enforce. The positive sign for altitude range may reflect high rainfall, which
is omitted because it is highly correlated with altitude range. High rainfall stimulates
rapid growth of natural vegetation, so it may be that access restrictions can be less strict
in these villages. The negative sign for population density either means that availability
of fodder has declined due to population pressure, or that more densely populated villages
were more likely to impose access restrictions. The positive sign for the percentage of
households working outside of agriculture means either that this caused less competition
for fodder, or that there was less pressure to impose restrictions in these villages.
When actual project dummy variables are used (not shown), the result is very
similar when the project dummies and expenditure per hectare are interacted. When they
are not interacted, all of the project variables have positive signs but none are significant
(while expenditure per hectare is negative and statistically significant).
Table 21 also shows the determinants of changes in availability of fuel from
government revenue lands. Most of variables have the same signs as in the model for
changes in grass fodder, and most of the same variables are significant. One notable
difference is that the NDWPRA and DPAP project categories have significantly greater
availability of fuel than the nonproject villages, while the AGY/IGWDP villages have
less. The finding for the AGY and IGWDP is consistent with that for grass fodder, while
that for the government projects could signify that they succeeded in planting trees and
getting them established, but then did not enforce their protection. It is important to note
that these results are not duplicated when actual project dummies are used; in that case
the NWDPRA coefficient is insignificant and all the others are negative.
The results for tree fodder are similar to those for fuel, which is not surprising
since trees are the main source of fuel. These results are not shown.
The strong finding of reduced availability of fodder from the common lands
deserves more detailed investigation, as does that for reduced fuel in the AGY/IGWDP
project category. Findings from qualitative investigations provide further insight into this
issue. In particular, women and livestock herders in many project villages complained in
group interviews that they had suffered from loss of access to common lands sealed off to
Herders: Livestock herders in many project villages complained in group
interviews that they had suffered from loss of access to their traditional grazing lands,
which were sealed off to promote regeneration. All of the projects had provided
employment opportunities to the herders, but they said it was not enough to compensate
their loss. This problem commonly arose in Maharashtra, where landless, low caste people
are a small minority in most villages and the decision to close the common lands was
usually based on a majority-rule vote. In the IGWDP villages the decision to begin the
project is based on consensus, but some landless people stressed in the group interviews
that it was not feasible for them to stand up to the will of a more powerful majority.
In some villages herders said that they had been promised that access restrictions
would be temporary while vegetation was allowed to regenerate. However, they
complained that regeneration had already taken place yet the common lands remained
off-limits to them. As mentioned above, such inequity is more likely to be a problem
where projects succeed in productivity and environmental objectives. In other places,
herders were able to ignore grazing restrictions, protecting their immediate livelihoods
but undermining project objectives. These findings from qualitative discussions are
consistent with the result in the quantitative analysis that a high population of shepherds
raised the extent of erosion but also raised access to grass fodder, compared to other
villages. To reiterate, this does not necessarily mean that these villages are more
productive, just that grass fodder from the commons was more readily available at the
time of the survey, possibly due to lack of restrictions.
Additional data from open-ended questions at the household level support these
findings. Table 22 shows that in 13 Maharashtra villages, respondents' perception that they
had benefitted from the projects rose with land holding size. Table 23 shows that landless
people were much more likely to indicate that the project had harmed their interests; among
landless people the unanimous complaint was lost access to common lands.
As revealed in Section 5, NGOs and the Jal Sandharan project in Maharashtra
appear to have dealt with this problem by selecting many villages that have no
government revenue land, thus avoiding the issue. Obviously this approach provides no
lessons about how to address the problem in the majority of villages that do have
government revenue land, but it may be an intelligent approach for agencies with limited
budgets that can only operate in a limited area.
Table 22: Percentage of respondents in Maharashtra who say they benefited from
the watershed project, by project category and landholding size
Landholding size category
Project category All respondents Landless 0-1 ha 1-2 ha > 2 ha
All projects 26 12 19 26 45
NWDPRA 8 0 17 0 17
JS/DPAP 17 0 0 33 20
NGO 39 29 44 25 63
AGY/IGWDP 31 14 0 33 60
Notes: a Findings based on household survey; 120 respondents in Maharashtra.
Table 23: Percentage of respondents in Maharashtra who say they were harmed by
the watershed project, by project category and landholding size
Landholding size category
Project category All respondents Landless 0-1 ha 1-2 ha > 2 ha
All projects 11 19 8 10 7
NWDPRA 4 0 17 0 0
JS 13 33 0 11 0
NGOs 8 14 0 8 13
AGY/IGWDP 17 29 14 17 10
Notes: a Findings based on household survey; 120 respondents in Maharashtra.
On the other hand, a few NGOs, such as Chaitanya and MYRADA in Andhra
Pradesh, have explicitly aimed to develop innovative solutions to the problem of
managing common lands. They try to try to build the interests of different groups into
the project design at the outset. For example, in some projects landless people are
granted fishing rights in the water bodies protected by soil conservation and revegetation
of the common lands. Unlike in Maharashtra, landless and near-landless respondents in
Andhra Pradesh unanimously reported having benefitted from NGO projects.
Social Centre, a Maharashtra NGO, grants fishing rights to landless people in
some villages including Mendhwan, covered under the current study (WOTR 1999).
Some projects encourage farmers without irrigation to dig group-owned wells so that they
have an interest in promoting groundwater recharge. Outside of the study area in the
famous Sukhomajri and Pani Panchayat projects, landless people even own rights to
water for tank or lift irrigation, which they utilize by leasing in farmland or, in the case of
Sukhomajri, sell to other farmers (Chopra et al. 1990; Patel-Weynand 1997). And in
several Andhra Pradesh villages not covered by any kind of project, shepherds lease
cultivated land and manage it as pasture. Such an arrangement could be made in a
watershed project: if shepherds had exclusive rights to grazing lands they would have an
incentive to invest in raising their productivity, and this would likely include reduced
grazing pressure and thus reduced erosion. A wide assortment of such arrangements can
be devised to spread the benefits of watershed development and, as a consequence,
increase its chances of success.
Finally, watershed agencies argue that if their work is successful, landless people
will benefit in the long term. In the famous Adgaon watershed, annual employment rose
from 75 days to 200 days, and laborers' incomes rose above those of small farmers
according to an NGO involved in the project (WOTR 1999). Social Centre found that
after 4 years of watershed management, laborers in Mendhwan village could find eight
months of employment whereas previously they could only find three months. In
Sherikoldara, landowners began to lease land to laborers rather than pay the high wage
costs (WOTR 1999).
Respondents in this study were asked whether they obtain more, less or the same
number of employment days than before the project period. No distinction was made
between short-term work generated as part of the project and long-term changes in
demand for labor. Table 24 shows that respondents in the AGY-IGWDP and NGO
project villages indicated with much greater frequency that employment opportunities
had risen, whereas those under the NWDPRA, the DPAP-Jal Sandharan, and in
nonproject villages indicated that employment had declined.
Table 24: Reported changes in number of days of employment between 1987 and
1997, by project category"'b
% of respondents indicating more, less or
same access to employment
Project category More Same Less
All villages 33 61 6
NWDPRA 9 91 0
Jalsandharan 29 65 6
NGO 43 47 10
AGY/IGWDP 72 17 11
No project 18 78 4
Notes: a Findings from household-level interviews; n = 85. 35 respondents who do not
engage in wage labor did not respond.
b Kruskal-Wallis test for ordinal variables shows that change in number of employment
days varies significantly across project categories (Chi-square = 13.6, 4 df, p < .01)
Women:23 Project officials rarely understood that watershed projects can increase
women's workloads. This happens for two reasons. First, if a project succeeds in raising
agricultural production, women will have to devote more labor to various cultivation
operations. Second, restrictions on collecting fodder and fuelwood from common lands
forces women to collect these resources elsewhere, increasing the time they must allocate
to these tasks.
23 This section on women and watershed projects draws on Pangare (1998).
Women had little opportunity to voice their concerns about watershed projects.
Guidelines for all projects contain language about promoting women's welfare, but in
practice virtually no projects created a role for women or addressed their interests. For
example, in about half the villages surveyed the actual watershed committees had no
women members, and most of the remainder had only one or, occasionally, two. In every
case a lone woman committee member proved to be a token to fill a bureaucratic
requirement. This is not surprising, as an individual woman on a male-dominated
committee in rural India will always find it difficult to make her voice heard. Moreover,
women are a heterogeneous group whose diverse interests cannot normally be
represented by just one or two women.
Just as some projects have taken innovative steps to incorporate the interests of
landless people and herders and give them a role in project management, all projects can
do the same for women. A few simple steps that can be easily adopted are to ensure that
women attend all project meetings (in part by scheduling meetings at times when women
are available to attend), give them 50% representation in project committees, listen to
them to find out their interests and concerns, identify the contributions they can make,
and train them in various watershed activities, among other things. The findings
regarding project impacts on women and recommendations for improvement are
presented in more detail in Pangare (1998).
7. PROMOTING IRRIGATION DEVELOPMENT
Raising the water table to promote irrigation development is a primary objective
of most watershed projects operating in Maharashtra, and a secondary objective of those
in Andhra Pradesh. The projects achieve this through soil and water conservation (SWC)
and revegetation measures that encourage rainwater to infiltrate into the soil, gradually
augmenting groundwater. Primary among the SWC investments are large structures
placed in the main drainage lines that impound water; they range from small "gully
checks" to major structures such as percolation tanks or check dams. Outside of the
drainage line, projects dig contour trenches along uncultivated hillsides.
This section examines the projects' impacts in promoting irrigation development.
Data limitations concerning the irrigation potential of each village make it impossible to
estimate precisely the contribution of the watershed projects in promoting irrigation, so the
analysis relies on several sources of information. It begins with a presentation of changes
over time in the mean irrigated area in the villages under each project category, and the
change in the number of seasons irrigated on the plots sampled in the study. Econometric
modeling to identify the determinants of these changes did not yield any insight, so
findings from qualitative investigations are presented to gain additional information.
CHANGES IN IRRIGATION
It is important to acknowledge weaknesses in the data used in this analysis: not all
of the numerous factors that determine irrigated area could be incorporated into this
analysis. In particular, the hard rock aquifers of the Deccan Plateau are known for high
spatial variation in irrigation potential. In some villages-or in some areas within some
villages-the potential for raising irrigated area is quite favorable, but in others it is
minimal. Unfortunately no data are available on the nature of aquifers in each village.
Changes in Irrigated Area at the Village Level
At the village level, the relevant measure of increased irrigation is the change
between 1987 and 1997 in the percentage of cultivated area that is irrigated.24 This
information is recorded each year for every village and stored at the taluka headquarters.
Table 25 shows that in 1987, prior to the introduction of the current projects, NWDPRA
villages had by far the highest area irrigated. This is consistent with their higher level of
infrastructure, such as electricity to power irrigation pumps and access to markets to sell
irrigated produce. The table also shows that during the period under study, nonproject
24 Virtually all arable land was under cultivation by 1987, the total area under
cultivation was roughly constant between 1987 and 1997.
villages actually enjoyed the highest average increase in percent area irrigated, more than
doubling their 1987 level by 1997. This may be due to the fact that the nonproject villages
had relatively good infrastructure but relatively undeveloped irrigation in 1987. They are
relatively flat and half of them are in the lower and middle part of the macrowatershed;
these characteristics are more likely to be associated with a higher water table.
Table 25: Average change in village-level percent irrigated area, by project category
Project category % area irrigated, % area brought % increase in irrigated
1987 under irrigation, area, 1987-1997
All villages 12.9 7.1 55
NWDPRA 23.7 5.2 22
Jal Sandharan 12.9 4.8 37
NGO 8.5 4.0 47
AGY/IGWDP 16.0 8.6 54
No project 10.5 10.9 104
Note: Analysis of variance shows no significant difference between project categories or
any other village characteristics.
AGY/IGWDP villages are the only other category with above-average increase in
irrigated area; with 50% more irrigated area in 1997 than 1987. This probably reflects
strong improvements in infrastructure in these villages during the period, including
electricity to power irrigation pumps. NWDPRA villages began the study period with the
highest level of irrigation, so they may have had less room for further expansion. Also,
the NWDPRA places less emphasis on irrigation development than the other projects as
is made clear by the exclusion of large water harvesting structures from its portfolio of
project activities. Villages under Jal Sandharan, which focuses especially on water
harvesting, had a particularly low increase in irrigated area. However, this might indicate
that the Jal Sandharan works in the most water-scarce areas.
Changes in Cropping Intensity through Increased Irrigation
Plot-level data provide more disaggregated information about irrigation
development. In particular, data on changes in cropping intensity give additional detail
regarding more subtle changes in irrigated area. For example, a plot that was irrigated for
one season in 1987 may be irrigated for two seasons in 1997, but the village level data on
gross irrigated area would not show the change.
The indicator for increased cropping intensity measures the change in the number
of seasons irrigated for each plot in the sample. For example, if a plot was rainfed in
1987 but irrigated two seasons in 1997, its score is +2. This information is collected
through the recall of the plot's owner.
Table 26 shows the average change in cropping intensity by state and project
category. Irrigation intensity increased much more in Maharashtra than Andhra Pradesh,
with a mean increase of 0.35 in Maharashtra compared to 0.20 in Andhra Pradesh. The
difference across project categories is significant only in Andhra Pradesh, where plots
under the World Bank project had the highest increase in cropping intensity. In
Maharashtra, plots under the AGY/IGWDP had the highest irrigation increase, but this
difference was not statistically significant. As in the Maharashtra village level data,
nonproject villages had a higher mean increase in cropping intensity than most projects.
Table 26: Mean increase in number of seasons irrigated 1987-1997, by project
Plot characteristic Mean increase in seasons irrigated
All plots Maharashtra Andhra Pradesh
NWDPRA .37 .10
DPAP or JS .33 .00
NGO .25 .11
AGY or I-G .44 n.a.
World Bank or ICAR n.a. .49
No project .25 .21
Note: Kruskal-Wallis test shows that differences across project categories are statistically
Multivariate analysis is needed to gain more detailed information about the
determinants of increases in irrigation development. Unfortunately, analysis at both the
village and plot levels failed to reveal any additional information; this is due almost
certainly to the lack of data on such important confounding variables as the nature of the
aquifer. Regression findings are not presented here due to their inability to provide
RESPONDENTS' PERCEPTIONS OF PROJECTS' EFFECTS ON IRRIGATION
Qualitative discussions revealed that respondents are keenly aware that water-
harvesting structures in the drainage line can raise the groundwater level, thus promoting
irrigation development. In several villages they indicated that water levels in open wells had
risen visibly following the construction of water harvesting structures. In several of the
Maharashtra villages, respondents indicated that low rainfall in recent years made it difficult
to discern the effectiveness of water harvesting. And in some villages, respondents reported
that certain water harvesting structures leaked water, making them ineffective.
As mentioned above, all of the Maharashtra projects focused on water harvesting,
whereas in Andhra Pradesh the World Bank, ICAR and NWDPRA devoted only minimal
attention to it. Only the DPAP focused primarily on water harvesting, and two of the
three NGO projects also included water harvesting as a major project activity. In Andhra
Pradesh, group discussions with owners of irrigated land revealed a good impression of
the DPAP's efforts in this regard.
Discussions in both states revealed a keen sense among farmers of the types of
structures that could promote water harvesting. For example, the DPAP and Jal Sandharan,
for which water harvesting was the main project objective, had large budgets for gully
structures and they built the largest and most solid, impermeable structures. The NWDPRA
and World Bank projects, on the other hand, were not designed with water harvesting in
mind and so they budgeted much smaller amounts for mainly vegetative or loose stone
structures. Respondents were keenly aware of these differences, especially in Maharashtra
where they could compare the NWDPRA gully structures with those built under COWDEP
in the 1980s. They did not perceive that the NWDPRA's work had much impact.
A similar issue arose among NGOs. As discussed in Section 5, projects vary in the
number of technically trained people on their staff. Some NGOs, like Chaitanya, employ
no technically trained staff and focus exclusively on social organization, relying on
indigenous technical knowledge in the design of their watershed interventions. Some other
NGOs, like MYRADA, employ engineers to oversee the technical work. Similar
differences are found in Maharashtra. Not surprisingly, respondents reported better water-
harvesting impact where projects employed technical experts. In the Chaitanya village, for
example, the water harvesting structure was not effective because it leaked. Such a finding
underscores the philosophy behind the AGY and IGWDP, which sought to combine the
technical expertise of government agencies with the social organization skills of NGOs.
In a semi-structured interview as part of the household survey, respondents were
asked to list the kinds of benefits they perceived from the project operating in their
village. Table 27 shows the number of respondents who mentioned irrigation benefits
and displays this as a percentage of 1) the total number of respondents, 2) the number of
respondents who are farmers, and 3) the number who are farmers with irrigation. In fact
all of those who reported benefitting had irrigation. They are best suited to explain
whether they thought project activities had helped raise the water table. Figures in the
table show that a much higher percentage of respondents perceived benefits in
Maharashtra than Andhra Pradesh, and this is consistent with project objectives in the
two states. In Maharashtra only the NWDPRA had low reported benefits among irrigated
farmers, and this is consistent with that project's lack of focus on water harvesting; they
are highest for the AGY and IGWDP. In Andhra Pradesh, perceived irrigation benefits
are very low for all projects.
Table 27: Number and percentage of respondents reporting that water-harvesting
investments improved their access to irrigation
State All respondents Farmers Farmers with irrigation
Maharashtra 21 18 23 46
NWDPRA 2 8 11 13
DPAP 3 13 17 50
NGO 6 17 21 60
AGY-IGWDP 10 28 37 71
Andhra Pradesh 9 6 8 13
NWDPRA 2 6 8 22
DPAP 3 8 11 18
NGO 2 6 7 9
World 2 4 5 11
Bank / ICAR
One obvious point in the table is that perceived benefits from irrigation are highly
concentrated among farmers with access to irrigated land. There are also indirect
benefits, such as higher employment demand, that respondents did not refer to. In any
case, the skewed distribution of the most valuable project benefits to those who already
have the most prized asset (irrigated land) is a source of concern to many project officials
and other commentators. There has been much discussion of what can be done to distribute
project benefits more evenly. For example, as mentioned in Section 6, in some projects
outside the current study area all village inhabitants share equally in water resources
generated by the project. No project in this study undertook such ambitious steps, but some
of them did try to help spread the benefits of irrigation. In particular, the IGWDP agreed to
take up work only in villages that agreed not to drill any borewells, which draw more water
than traditional open wells and would appropriate harvested water disproportionately. For
similar reasons, the IGWDP also insists that no farmers may take up water-intensive crops
such as sugarcane in response to higher water supplies. Sugarcane farmers would draw
more water from their wells, reducing the water level in other wells. Also, farmers with
excess water might choose to sell it to their neighbors if they cannot grow water-intensive
crops. A few other NGOs mentioned similar restrictions, but most did not. None of the
government projects tried to impose any such restrictions.
Another approach to sharing the benefits of water harvesting is to help resource-
poor farmers invest in their own wells. India has quite a bit of experience in this regard; a
centrally sponsored program, for example, digs individual private wells for landless, low
caste farmers. Some projects have invested in group owned wells, but the most common
experience was that groups had difficulty in working together to manage and maintain their
wells cooperatively. This matches the experience of state-owned cooperative tubewells
(Shah 1993). In the villages in this study, there were numerous cases of group-owned
wells, but in nearly every case they were jointly owned by brothers who inherited the well
from their father. There was one recorded case of some neighbors (not relatives) who
jointly invested their own funds in a well, but within a few years a dispute emerged and the
case ended up in court. Against this backdrop, most projects are hesitant to invest in group
wells. In the current study, only one NGO, Gramayan, invested funds in a group-owned
well. According to respondents it is managed effectively.
8. NATURAL RESOURCE MANAGEMENT AND PRODUCTIVITY OF
RAINFED AGRICULTURAL LAND
As introduced in Section 2, raising the productivity of rainfed agriculture is the most
important objective for some watershed projects, particularly the NWDPRA, the ICAR
model watersheds, and the World Bank's Pilot Project and IWDP (Plains). It is particularly
important where opportunities for water harvesting are limited, as with many Andhra Pradesh
project locations. The watershed approach to raising the productivity ofrainfed agriculture
begins with conserving soil on rainfed plots, which also implies retaining soil nutrients and
concentrating moisture. This in turn creates opportunities for planting high yielding varieties
that require more water and nutrients, or, in some areas with black soils and high rainfall,
may enable farmers to harvest an additional crop each year.
This section examines the experience of the projects in promoting rainfed
agricultural development. Following a review of projects' subsidy policies for
developing private land, it investigates the nature and extent of interaction between
project staff and farmers, since technical assistance for rainfed agriculture presumably
involves working closely with farmers. It then focuses on efforts to conserve soil and
moisture, both through improved agronomic methods and investment in soil conservation
structures. Analysis of rainfed farmers' adoption of new varieties and their net returns to
cultivation follows. Because of the focus on rainfed agriculture, the quantitative analysis
focuses on plots that were unirrigated in 1987 and 1997.
PROJECT SUBSIDIES TO PARTICIPANTS UNDER EACH PROJECT
Subsidies are a contentious and increasingly complex issue in watershed projects.
Approaches have evolved over time, with significant trial and error. In the early days of
the Bombay Land Improvement Scheme, bunds were installed on some farmers' fields
without their consent, yet they were expected to repay the bank for the cost of the work
undertaken. They were listed as defaulters if they refused (World Bank 1988).
In some watershed projects, the opposite approach is taken today; with watershed
works being heavily subsidized and no thought given to cost recovery. The rationale for
this approach is that farmers who benefit from canal irrigation did not have to pay for the
canal, so why should rainfed farmers who benefit (much less) from watershed projects
have to pay for the works undertaken? This argument is really a matter of opinion, but it
matters because rainfed farmers often are not interested in the measures introduced under
watershed projects and have no intention to maintain them once the project ends (Kerr et
al. 1996; Sanders et al. 1999). Under these circumstances, it is important to require some
kind of payment or other sacrifice by "beneficiaries" simply to make sure that they really
want the work and are likely to maintain the assets created. Otherwise the project will
simply be a waste of money. (This problem does not arise in irrigation projects, because
there was never a farmer in India who did not want irrigation!)
There are two main reasons why farmers would allow measures to be taken on
their land that they do not really want. The first is that some projects install structures on
farmers' plots without their consent, although this practice is diminishing. Watershed
officials increasingly appreciate the fact that a structure built in one location can generate
on-site costs but only downstream benefits, so in current projects measures are rarely
undertaken in farmers' fields without their consent.
A new problem that may lead farmers to "accept" measures they do not want
results from the fact that in most projects, watershed works are labor-intensive and very
highly subsidized. All projects covered in this study subsidized work on common lands
at the rate of 100%, generating ample employment for workers to plant vegetation, dig
trenches and build structures. Even on private land, the typical subsidy rate was 90%,
and the remaining 10% was not paid in cash but in kind (in the form of labor). Moreover,
much of the project work was undertaken in the dry season when labor demand is scarce,
and in many projects wages exceeded the slack season market wage. So even if a project
paid only 90% of the subsidized wage it may still represent more than the farmer could
earn under other available opportunities. Under these circumstances it may well make
sense for a farmer to accept an unwanted structure on his field, provided of course that
the costs of dismantling it are low.
Subsidy Policy and Practice under each Project
NWDPRA: Project guidelines call for a contribution by farmers only for work
undertaken on private lands, except that no single family may receive subsidized
assistance worth more than Rs 5000 (GOI 1991a). Specific terms are not mentioned.
From the experience of the present study it is not clear how the farmer's contribution
works in practice. In the Maharasthra villages, no work was done on private lands, so the
issue did not arise. In Andhra Pradesh, work was undertaken on private lands using labor
paid for by the project, but respondents did not indicate that they had contributed
anything. Kolavalli (1998) found that the NWDPRA collected a small farmer's
contribution in only one of the four project sites he visited.
DPAP and Jal Sandharan: These projects had no beneficiary contribution. Most
work was conducted on nonarable land, but even the minority of work done on private
land was entirely subsidized. Usually this work is done using contracted labor. This
increased the possibility that the farmer would not be aware that the work is taking place,
but it sharply reduced the incentive for the farmer to "accept" unwanted work.
NGOs: Several NGOs called for a 10% farmer's contribution for work on private
land, paid in kind (in the form of labor). As mentioned above, however, the wage scale
was inflated so that employment benefits remained substantial to the farmer. Many
NGOs liked to contract farmers to do the work on their own field on the principle that
this would raise the quality of the work.
Two NGOs in Andhra Pradesh, on the other hand, required a much more
substantial farmer's contribution on private land. Chaitanya required a 50% contribution
while MYRADA recently introduced a requirement of 33%. In some villages not
included in this study, MYRADA is experimenting with zero subsidies for work on
private land (Fernandez 1998). Chaitanya and MYRADA offered lower subsidies in
recognition that the farmer would be the primary beneficiary of the work and that farmers
would certainly pay attention to its quality if they helped pay for it. There was no
contribution for work done on common land.
AGYandIGWDP: In these projects a private landowner's contribution was about
8 percent, but this figure was inflated because dry season wages under the project often
exceeded existing market wages.
ICAR: Under the ICAR model watersheds, all costs were paid by the project and
farmers were provided improved seeds and other inputs free of charge. Little or no
employment was generated as part of project implementation.
World Bank: The World Bank Pilot Project and IWDP both called for a farmer's
contribution of 10% on cultivated lands. The contribution was in kind in the form of the
farmer's labor. Farmers also received various free inputs such as improved seeds and
fertilizers that more than made up for the value of any contribution. There was no cost-
sharing for work on common lands.
INTERACTION BETWEEN PROJECT STAFF AND SURVEY RESPONDENTS
Table 28 shows the percentage of respondent farmers from each project category
that interacted with project staff. To distinguish among different types of interaction, it
also displays the percentage of respondents who received technical recommendations
related to rainfed agriculture, and the percentage who actually adopted some practice
recommended by the watershed project staff.
Table 28: Percentage of farmers who interacted with watershed project staff, by
Project category Percentage who Percentage who Percentage who
interacted received technical adopted a technical
Overall 30 17 9
NWDPRA 0 0 0
DPAP/JS 0 0 4b
NGO 44 25 8
AGY/IGWDP 56 28 19
Overall 67 53 50
NWDPRA 56 51 49
DPAP/JS 58 22 22
NGO 70 57 54
World Bank/ICAR 78 72 67
Notes: a This table excludes respondents from nonproject villages.
b Some projects installed watershed measures in farmers' fields without consulting them.
Three main points are worth mentioning from the table. First, overall interaction
rates were not very high. This reflects the fact that watershed projects rarely cover every
farmer's field in every project site.
Second, interaction was much higher in Andhra Pradesh than Maharashtra, and it
was much more likely to include technical recommendations. This reflects the way in
which projects operated in the two states. There was less scope for interaction in western
Maharashtra because most projects there focused on soil and water conservation on
nonarable lands rather than technical interventions on farmers' fields.
The picture in Andhra Pradesh is very different. Here, the level of interaction
between staff and respondents was much higher, and most of that interaction came in the
form of technical recommendations for rainfed agriculture. Only the DPAP, whose
primary mission was to develop water resources through groundwater recharge, had low
levels of technical interaction. The World Bank and ICAR projects had high levels of
interaction, almost all of it in the form of technical recommendations.
A third noticeable finding is that in Maharashtra, no farmers in the NWDPRA and
Jal Sandharan project villages ever interacted with project staff. This is quite surprising,
particularly for the NWDPRA, whose mandate is to promote rainfed agricultural
development. Most likely it reflects the focus on water harvesting in the taluka-level line
departments that implemented the project. Even so, at first glance it is surprising that
there was no interaction based on employment of labor. But this is explained by the fact
that the project officials worked through an intermediary in the village who in turn hired
workers, so that there was no explicit interaction between project officials and laborers.
NGOs and the AGY and IGWDP projects in Maharashtra had more interaction
with farmers, but the figures are still lower than in Andhra Pradesh. They reflect mainly
project efforts to facilitate social organization and to mobilize laborers rather than
technical assistance for rainfed agriculture.
A fourth point of interest concerns the percentage of respondents who actually
adopted a practice or technology recommended by the watershed project on the plot in
question. These figures quite closely reflect the figures for technical recommendations.
In Andhra Pradesh, almost all farmers who received technical recommendations also
adopted them. In Maharashtra there was very little adoption of specific technologies.
These findings suggest that the analysis of project impacts on rainfed agriculture
should focus primarily on the Andhra Pradesh villages. Most of the rest of this chapter
ADOPTION OF SOIL AND WATER CONSERVATION PRACTICES
Despite the historic focus of most Indian soil and water conservation programs on
mechanical measures, soil scientists and agronomists often stress that there is much more
to soil conservation than trapping runoff water behind mechanical or vegetative barriers.
Conservation begins with sound agronomic practices such as maintaining soil cover and
cultivating across the slope to encourage infiltration and reduce runoff. Accordingly, this
section examines farmers' adoption of both approaches to conserving soil.
Respondents were asked about a variety of conservation-oriented agronomic
practices, including strict contour cultivation, cultivation across the slope, retaining
stubble in the plot, and applying mulches to cover bare soil. Of all of these practices,
cultivation across the slope was the only one practiced by more than a handful of farmers.
Farmers indicated that they recognized the value of applying mulches and retaining
stubble in the fields throughout the dry season, but they rarely carried out these practices
due to the high opportunity cost of biomass for use as fuel and feed.
Respondents uniformly said that strict contour farming is impractical except on
irrigated land and plots steeper than those covered in this survey. Not a single respondent
practiced contour cultivation. This finding echoes the points about contour cultivation
made by Kerr and Sanghi (1992). In short, numerous basic features of indigenous rainfed
farming systems are integrally linked to quadrilateral plot boundaries, and contour bunds
and contour cultivation directly interfere with them. As a result, adopting contour
farming carries high opportunity costs. In the 1990s, many watershed projects still
officially recommended contour cultivation on rainfed plots, but in practice this was
ignored. Efforts were limited to promoting "modified contour cultivation," which simply
means cultivating across the slope. Project staff ignored the official instructions to
promote contour cultivation because farmers simply would not adopt it, except on
irrigated and very steep rainfed plots.