Policy Research Working Paper 10439 An Assessment of Community-Led Lift Irrigation Systems in Jharkhand, India Avinash Kishore Manavi Gupta Felipe Dizon Priti Kumar Agriculture Global Practice May 2023 Policy Research Working Paper 10439 Abstract The Jharkhand State Livelihoods Promotion Society is roll- money on irrigation, especially compared with nonbene- ing-out solar powered community lift irrigation systems to ficiaries who use their own or who rent diesel pumps to provide access to irrigation to 23,580 farming families in irrigate their fields. Although the solar powered community 13 districts of the state. This paper assesses these irrigation lift irrigation systems offer multiple benefits, they remain systems using data collected from 297 farmers in the com- severely underutilized. The median hours of operation of mand area of 50 randomly selected irrigation systems. The the 50 sampled irrigation systems was less than 100 hours study also interviewed 457 nonbeneficiaries. Farmers in per year, and the average operating time was 192 hours per the command area of the irrigation systems irrigate more year. Solar irrigation systems cannot be economically viable land, have higher cropping intensity, are more likely to grow with low levels of capacity utilization, indicating that incen- high-value crops, and had higher gross value of output in tives are needed for system managers to increase utilization, the Rabi (winter) season. The beneficiaries also spend less gross irrigated area, and irrigation surplus. This paper is a product of the Agriculture Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at a.kishore@cgiar.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team An Assessment of Community-Led Lift Irrigation Systems in Jharkhand, India Avinash Kishore1, Manavi Gupta, Felipe Dizon, and Priti Kumar JEL classifications: O13, Q01, Q15 Keywords: solar, irrigation, agriculture productivity 1 a.kishore@cgiar.org (corresponding author). Shri Bipin Bihari, Sanjay Das, Abhishek Gupta, and Pratyush Singh generously shared with us their deep knowledge of the functioning of the JOHAR irrigation projects and the communities that manage and use them. The primary survey and the fieldwork could not have been completed without their cooperation and support. Hina Sherwani provided helpful research support. Sanjay Prasad of KAABIL led a dedicated team of enumerators to collect the primary data from farmers while his colleague Pratyaya Jagannath conducted semi-structured interviews with the farmers and stakeholders. Their insights from the field have enriched our understanding of the context and the JOHAR intervention. Most of all, we are deeply indebted to the women and men farmers who spared their scarce time to answer our questions and help us carry out this study. This research was funded by the Foreign, Commonwealth and Development Office (FCDO) of the Government of the UK under the project Building Agriculture and Rural Knowledge of Best Practices in Jharkhand. The findings, interpretations and conclusions expressed herein are those of the authors and do not necessarily reflect the view of the World Bank Group, its Board of Directors or the governments they represent, nor of the UK government’s official policies. 1. Background Jharkhand is one of the poorest states in India. A nationwide survey of the situation of agricultural households (SASH) in 2019 found that farmers in Jharkhand have the lowest average farm incomes in the country (SASH, 2021). The average farmer in Jharkhand earned only Rs. 8,000 in 2018-19 from all crops, compared to the national average farm income of Rs. 37,000. The average total income of agricultural households in Jharkhand is the lowest in India (Rs. 3,993/month, compared to the national average of Rs. 8,337/month) and the share of crop income in the total household income is also the lowest (17%) after Jammu & Kashmir and Kerala (ibid). The most recent NSSO Consumption Expenditure Survey, carried out in 2011-12, found that more than 45 percent of farmer households in Jharkhand lived in poverty – the highest in India. Income from agriculture in Jharkhand is also highly vulnerable to droughts, dry spells, and other weather shocks. In 2018- 19 for example, 75 percent of all paddy growers and 96 percent of farmers growing black matpe (urad) in Jharkhand reported losses due to droughts or inadequate rains. Poor access to irrigation is a major reason why agriculture is a high-risk, low-return endeavor in Jharkhand. An analysis carried out in 2016 found that 18 districts in Jharkhand were among the most irrigation deprived in India, and that high energy costs rather than a physical scarcity of water was the main reason for poor access to irrigation in the state (Shah et al., 2016). According to the most recent Minor Irrigation Census, 94 percent of small irrigation structures in Jharkhand were powered by diesel or muscle in 2013-14, and both were significantly more expensive than grid electricity as a source of energy for irrigation. While 65 percent of farmers in the state irrigated at least a part of their land in the Rabi (winter) season, only 19 percent of the cropped area was irrigated, compared to 58 percent at the all-India level (SASH, 2021). The state government’s Jharkhand Vision & Action Plan 2021 sought to increase the gross irrigated area from 0.5 million hectares (ha) in 2016-17 to 0.8 million ha by 2020-21 as an element of its larger objective of doubling farmers’ incomes. Resources from the Pradhan Mantri Krishi Sinchai Yojana (PMKSY) and other government programs were to be utilized for this purpose. The Jharkhand Opportunities for Harnessing Rural Growth (JOHAR) project was launched in 2017 with support from the World Bank and implemented by an agency established by the state Ministry of Rural Development titled the Jharkhand State Livelihoods Promotion Society (JSLPS). JOHAR’s development objective is “to enhance and diversify the incomes of targeted rural 2 households” in 68 blocks in 17 districts in the state. It encourages participating farmers to shift production towards higher value crops which can be sold at more remunerative prices, and improving access to irrigation is an important element of this aim. More specifically, JOHAR seeks to improve access to irrigation through community owned and managed lift irrigation systems. This access is the purpose of the JOHAR Community Led Lift Irrigation Scheme (JCLIS). The JCLIS was assigned to install 2,310 solar pumps by June 2022. Each 5 to 7.5 horsepower pump is connected to a buried pipe distribution network that delivers water to a command area covering between six and eight hectares. All but a few of the systems are powered by solar energy, and 1,310 of the pumps are fixed and 1,000 are mobile. Together, they are used to irrigate 10,600 hectares of land belonging to 23,580 farming families across 39 blocks in 13 districts. Of the fixed systems, 572 were already operational in the state as of September 2021, and 242 more were nearing completion. The project provides farmers with improved access to affordable irrigation by i) creating a new water source or augmenting an existing one; ii) installing a solar pump with very low operating costs to lift the water, and iii) building a network of buried pipelines to deliver water to the fields. In addition to physical infrastructure, the project actively encourages the creation and training of a water users’ group (WUG) with the capacity to operate and manage the irrigation system. The JCLIS schemes are all surface or subsurface micro-irrigation schemes, and are not groundwater schemes, which extract water from aquifers. The source of water for these schemes are dug wells or ponds, which get recharged by rain-fed perennial and seasonal streams. This paper presents the findings of an independent study of the community led irrigation systems created under the JOHAR project. The findings are described in Section 1 and based on qualitative and quantitative data collected from the farmers using these irrigation systems, their neighbors, and those responsible for operating and managing the systems. Section 2 of the report presents the objectives of this study. The authors explain the sampling and data collection methods used, as well as the limitations of overall methodology in Section 3. Section 4 focuses on the impact of the community irrigation systems on the project participants and Section 5 on the sales of high value crops by farmers in the study area. Strategies to increase the benefits of the project’s JCLIS Lift Irrigation Scheme are considered in Section 6. Section 7 concludes with a discussion of opportunities and challenges which affect the prospects of expanding the area of irrigated land in Jharkhand. 3 2. Objectives of the Study The main objective of this study was to assess the impacts of JCLIS on farmers’ access to irrigation, cropping patterns, and the gross value of crop output. These impacts were measured in terms of a) the gross irrigated area across Kharif, Rabi, and Zaid (summer) seasons; b) the hourly cost of irrigation; c) the cropped area in Rabi and Summer seasons; d) the increase in the area under high-value crops, and e) the gross value of crop outputs of the beneficiary farmers in comparison with that of the control group. The second objective of the study was to assess farmers’ satisfaction with the JCLIS and to identify areas of improvement in the management and the performance of the irrigation systems. 3. Methodology 3.1 Sampling The primary survey undertaken during the course of the study began in September of 2021, at a time when a substantial number of working JCLISs had already been installed. Of the 572 working JCLISs, 230 had been installed by a year prior, before September 2020. Ninety-three percent of these systems were concentrated in five districts: Gumla, East Singhbum, Hazaribagh, Khunti, and Lohardaga, which are spread across two of Jharkhand’s three agro-climatic zones. These five districts were selected for the primary survey, and within them, 50 lift irrigation systems were selected based on probability proportional to size (PPS) (Table 1). The sample was selected from irrigation systems that had 10 or more beneficiaries per system. These covered 13 blocks across the five districts.2 The sample was stratified to include an equal proportion of older and newer systems as well as of solar and diesel pumps. However, 68 percent of the systems in the sample had just become operational in 2020, and only 6 percent used diesel pumps. Table 1. Distribution of complete systems by district District Name Population Proportion of Complete Sample Proportion of Complete Systems Systems Gumla 43% 49.2% East Singhbhum 14.4% 20.2% Khunti 12.6% 14.5% Lohardaga 12.6% 10.1% 25 systems were found not to be working when the survey team reached the field. They were replaced by 5 other functioning systems from the neighbouring areas. 4 Hazaribagh 10% 6.1% Six participating farmers and a control group of four nonparticipating farmers were randomly selected for the survey. The control group consisted of farmers with land parcels in close proximity to the local JCLIS command area but not irrigated by it and were presumed to be directly comparable to their participating neighbors by all other parameters affecting agricultural productivity. Five additional nonparticipating farmers from an adjacent village were surveyed as well to account for any spillover effects that may confound results based on comparisons of farmers in the same village. These farmers in nearby villages were counted as part of the control group. This selection process led to a total sample consisting of 757 farmers in 95 villages across 13 blocks in Jharkhand. 297 were participants and members of their local JOHAR Water User Group. 460 nonparticipating farmers comprised the control group. 92 percent of farmers in the sample, including those in the control group, were members of the JOHAR Producer Groups (PG). Farmers in the control group may also have received some training and extension from the larger JOHAR project, but they did not use JCLISs for irrigation. In fact, less than 10 percent of the beneficiary and non-beneficiary farmers in the sample reported having received agricultural extension from JOHAR or any other institutional sources. This makes it safe to assume that the main results based on comparisons between the treatment and control groups are not confounded by some control farmers benefiting indirectly from JOHAR project interventions. (For robustness, the authors also omitted all farmers in the sample who reported having received extension support from the JOHAR project. The results, which are not included in this report, do not differ from the smaller sample analyzed here.) 3.2 Data collection Primary surveys were used to collect data from participating farmers and from JCLIS operators or managers. The farmer surveys collected data on household demography (social group, age, and education of the head of the household, main occupation, etc.); ownership of land and other durable assets; area allocated to different crops and area left fallow during the Kharif, Rabi, and Summer seasons of crop year 2020-21; area irrigated from different sources in each season and the hourly cost of irrigation; sales channels used to sell fruits and vegetables and prices realized; and challenges inhibiting increases in crop production and incomes. Farmers were also questioned 5 regarding their impressions of how well their local JCLIS functioned and how fairly irrigation water was allocated, as well as how reliable and affordable the irrigation service provided were. JCLIS operators or managers representing 50 community-led lift irrigation schemes were interviewed as well. These interviews were used to collect system level data on the actual command area served in each season; fees collected; rules for water allocation; experiences of system breakdown and repair, and opportunities to expand the command area. 3.3 The Econometric Strategy OLS regressions, linear probability models, and Tobit models are used to estimate the impact of irrigation from JCLIS on a range of agricultural outcomes including area irrigated by farmers, the probability of growing a crop in the Rabi or Summer season, the probability of growing at least one high-value crop, and the gross value of output in the Rabi season. In our main specification, we estimate the following equation: = 0 + 1 + 3 + + (1) + 2 Where Yfv is the outcome variable for farmer f in village v. Yfv takes the value of 0 or 1 for binary outcomes (for instance, if the farmer grows a high value crop) and we use the natural logarithm of the Gross Value of Output (GVO) when estimating the impact of JCLIS on the GVO in the Rabi season of 2020-21. Both OLS and Tobit Models are estimated to account for the truncation in data due to zero values for farmers who did not cultivate crops in the given season. Our coefficient of interest is 1 that measures the average impact of being a JCLIS beneficiary (Tf = 1) on the selected outcome. Pv accounts for system-specific characteristics like months since operation and type of pump (solar or diesel). Xf is a vector of household controls and BFE is the block fixed effects to account for unobservable heterogeneity at the block level. In an alternate specification, our sample is limited to compare farmers within a given village – between JOHAR project participants and their neighbors who do not participate in the project. We include village fixed effects instead of block fixed effects in this model to account for all time-invariant village characteristics that may affect the impact of JCLIS on agricultural outcomes. 11 percent of participants and 35 percent of non-participants in our sample did not grow a Rabi crop in 2020-21, and therefore report 0 as the GVO for the season. A Tobit model is also estimated to account for the censoring of data when estimating the impact of JCLIS on the GVO. 6 3.4 Limitations of the study This assessment of the impact of the JCLIS relies on simple comparisons of the average outcomes of interest between participants of the lift irrigation system and their neighbors in the same village or in a neighboring village who do not participate. We use self-reported recall data collected from farmers through in-person interviews with them. JCLIS sites and their beneficiaries were not selected randomly. Site selection was guided by hydrological and topological characteristics. Farmers with land closer to favorable sites may differ from their neighbors in skills, motivation, social capital, and other qualities, any of which can also affect agricultural outcomes. Farmers were moreover required to pay a small one-time Rs. 1,100 fee and join the local water user group to irrigate their land from the JCLIS, and their decision to do so may also differentiate them from others who chose to stay out or wait longer before joining the group. The comparisons made are therefore qualified as suffering from omitted variables bias. Nor did the study benefit from baseline data on variables of interest prior to the intervention. Attempts to collect recall data from farmers with questions about cropping areas, patterns, and yields before the JOHAR irrigation systems were introduced led to unreliable data, particularly for earlier years, and were discontinued. We also considered revisiting farmers who were interviewed in the baseline survey carried out by Oxford Policy Management (OPM), but at the time of our survey (August-September 2021), very few villages in the baseline survey had a functioning JCLIS. We compare beneficiary and non- beneficiary farmers across economic variables that presumably do not change over time or as a result of the JCLIS irrigation. These include average land holding size and asset holdings. We find that the two are not significantly different across the two groups, providing some evidence of baseline comparability. In an alternate specification, we drop the control farmers from the neighboring villages and compare JCLIS beneficiaries only with their geographical neighbors while controlling for village fixed effects and find results similar to our main specification. Results from this alternate specification are reported in Table A.6 in the appendix. We also try to mitigate the selection bias in our study by using a Coarsened Matching and comparing only the matched sample of treated and control farmers. Results with the matched sample (Table A.7 in the appendix) are also qualitatively similar to the results from our main specification. Lastly, as a robustness check, we leverage remotely sensed data and control for two years of pre-treatment measures of Normalized Difference Vegetation Index (NDVI) for both groups of farmers (at a spatial resolution of 10 meters) in our main analysis. We find that our results remain the same, 7 including for the effects of JOHAR irrigation on Area under Irrigation, High-Value Crop Cultivation, and Gross Value of Output in Rabi 2021 (Tables A.8 through A.10 in the appendix). This enables us to conclude with some confidence that (i) beneficiary farmers were not better off before the installation of the JCLIS system, and that (ii) their increase in GVO and area under high value crop cultivation can be attributed to the increase in irrigation delivered by the JOHAR systems. JCLIS operators and managers do not keep written records of system use. Therefore, our assessment of system performance is also based on self-reported recall data from these respondents which may contain recall errors or systematic biases, including the possibility of deliberate misreporting. Data on system performance collected from beneficiary farmers were also used to validate operators’ responses. However, data on important aspects like the total operating hours of the system and the total command area served in each season could only be collected from the operators. Disruptions from the COVID-19 pandemic may have led to lower price realizations for high-value crops grown by farmers and may even have affected their crop choices. It is therefore possible that the impacts of JCLIS may become measurably greater once the pandemic subsides. These limitations of the study should be kept in mind when interpreting the results of this analysis. 4. Data 4.1. System characteristics 50 system operators were interviewed. (Table A.1 in the Appendix reports the summary statistics for systems and their operators.) 48 of them were men. 36 of the 50 operators had received training from the JSLPS team. Most of them had been trained in the operation and management of the irrigation system including basic troubleshooting and care. Only a few of the operators had any training in record keeping or financial management, and none of them maintained written records of hours of irrigation services provided to farmers, fees collected, or area irrigated. The rest of the discussion in this section, therefore, relies on the operators’ recall. Most JCLISs have started irrigating farmers’ fields only recently. Our sample of 50 systems was selected from those which had been installed for more than 12 months when we went to the field. 9 of the 50 systems had started irrigation after September 2020. The newest system in our sample 8 had been serving farmers for 9 months while the oldest had been irrigating for 42 months at the time of the survey (Figure 1). 41 of the 50 systems were less than two years old and only one system was older than three years. This study therefore presents an early assessment of the functioning and the impact of JCLISs. There has not been a drought year since most of these systems started working. Nor have they been used during the Kharif season critical for the household food security of farmers, since it was a good monsoon year. Norms around water allocation, water fees, and fee collection, and resource mobilization for repair and maintenance of system water sources, pumps, panels, and distribution networks are still evolving. Only a few instances of physical or technical breakdown in the sampled irrigation systems have taken place, and the water user groups managing them therefore have not experienced stresses that would test their solidarity, discipline, initiative, and resourcefulness. Figure 1. The frequency distribution of months since the sampled JOHAR systems started irrigating farmers' fields (as of 1 September 2021) 20 12 9 5 4 <1 year 12-18 months 18-24 months 24-30 months 30-42 months Source: Created by authors using primary data collected from system operators 47 of the 50 systems in the sample are powered by solar energy. The remaining three use diesel pumps. Each solar pump was connected to 15-16 solar panels with a total capacity of 4.7 to 5.6 kilowatt (kW). The capacity of the three diesel pumps varied between 5 and 8.5 horsepower (hp). 44 of the 50 system operators knew the design command area of their system. The average design command area of JCLIS in our sample is 13.7 acres (95 percent confidence interval: 11.4 to 16.0 acres) and a typical system benefits between 15 and 18 farmers. According to the operators' report, 9 the actual area irrigated of 30 out of 44 systems was smaller than the designed command area while five systems irrigated more area than originally planned.3 Nearly 80 percent of the JCLISs in our sample draw water from dug wells. Seasonal or perennial streams (with or without check dams) are the other important sources of water for irrigation systems. Only 21 systems had reliable access to water throughout the year. Low availability of water in April, May, and June is a constraint to growing summer crops in the command area. Every second operator mentioned water shortage during the critical periods as a constraint to expanding the command area of their JCLIS. In one village in the Khunti district, farmers in the water user group were planning to invest in a buried pipeline to connect the dug well with a nearby stream to improve water availability in the summer season. We do not know how user groups managed water scarcity. In our informal discussions, farmers reported bearing the yield loss due to soil moisture stress in the summer season after the source of the water dried up. On the other hand, the systems remain idle in the monsoon season (July, August, and September) because of a lack of demand for irrigation. The villages we visited had not experienced droughts or long dry spells during the monsoon season in the last two years. In years of drought or during long dry spells, JCLISs may become useful even during the Kharif season by providing life-saving irrigation to the main paddy crop if water is available in the source. Operators were also asked to recount the total hours of operation of pumps during the three cropping seasons of the 2020-21 crop year. Their recall data suggests low-capacity utilization of most JCLISs. The median system worked for just 93.5 hours throughout 2020-21, while the average operating time was 191.8 hours. Where water sources like dug wells, ponds, or rivers dried up, the average total hours of operation were lower (Table A.5 in Appendix). Looking at the hours of operation across the three seasons, 26 of the 50 JCLISs remained completely idle in Kharif 2020 owing to normal monsoon rains. The other 24 systems were also only barely used. More than 40 percent of the systems faced water scarcity during summer, and this limited both their operating hours and the command area they served. Water scarcity occurred mainly in the summer months from March to June, yet even during the Rabi season, when wells and streams have sufficient 3 These comparisons are based on operators’ recall since they do not maintain written records of the area irrigated or fees collected from the farmers. 10 water, low levels of system use were evident. Moreover, 15 percent of farmers reported non- availability of water from the JCLIS during the Rabi season as well (Table A.5 in Appendix). Table 2. Median and average hours of operation of JICS in the crop year 2020-21 Season Median hours of operation Average hours of operation Kharif 2020 0 16.7 Rabi 2020-21 60 84.4 Summer 2021 27.5 90.8 Total in 2020-21 93.5 191.8 Source: Created by authors using primary data collected from system operators The average irrigation fee for solar pumps is Rs. 34/hour. Operators of three solar-powered systems reported that the irrigation fee had yet to be decided. Nine other operators also said that while the irrigation fee has been decided, fee collection was not yet underway. Because solar pumps have no variable costs, water user group members do not feel the need to collect any fees even if money may be required in the future for repair and maintenance. Irrigation fees for diesel pumps vary from Rs. 120 to Rs. 140 per hour. Though the discharge from solar pumps varies during the day and across the seasons with the intensity of the sunlight, the irrigation fees remain the same across the year and have been set on an hourly basis for all systems. Among those who cultivated in the given season, 71 percent of farmers reported paying an irrigation fee in Rabi 2021, 30 percent in Kharif 2020, and 66 percent in summer 2020. 35 of the 50 systems experienced no breakdowns. Of the 15 systems that did, the most common problems were related to leakage in the distribution system and glitches in the control box, the control panel, or the wiring. In most of these cases the repair costs were negligible (5 cases) or small (less than Rs. 2,000). Managers of the two systems had to spend Rs. 27,000 and Rs. 35,000 on the repair. Water users mobilized the money for repair in both these cases. 45 of the 50 operators reported no significant disputes or discord in managing the irrigation systems. Where disputes did occur, they were triggered by disagreements over irrigation sequencing and water was scarce. 4.2 Farmer characteristics 757 farmers were interviewed in the primary survey. Nearly 90 percent of the respondents were women. 39 percent of the respondents in the beneficiary group and 32 percent in the control group could not read or write. 65 percent of beneficiary households and 59 percent of households in the 11 control group belonged to scheduled tribes, implying that the SIPs are not being captured by Dikus or non-tribal farmers in the villages. The non-tribal respondents were mainly from the OBC group with very little representation of Dalits and General castes. Farming is the main source of income for 80 percent of the beneficiary families and 69 percent of the non-beneficiaries. Most of our respondents were marginal farmers with average land ownership of 2.9 acres for beneficiaries and 2.3 acres for non-beneficiaries. These smallholdings are further sub-divided into more than three or more separate parcels. The average monthly per capita expenditure (MPCE) is only Rs. 705 for beneficiaries and Rs. 855 for non-beneficiaries. Thus, the JOHAR beneficiaries, on average, own more land, but they have lower consumption levels. More details on some of the key characteristics of farmers in both the treated and control groups are presented in Table A.2 of the Appendix. 4.3. Access to irrigation Nearly half of the respondents have access to a source of water, most commonly a private dug well. The beneficiary households rely mainly on JICS for irrigation while dug wells fitted with small pump sets are the main source of irrigation for farmers in the control group. 201 of the 757 farmers in our sample (26.5%) owned a diesel pump and 206 (27%) of them had an electric pump. (Table A.3 in Appendix). We had not expected such high levels of pump ownership in Jharkhand. 180 of the 201 diesel pump owners and all 206 electric pump owners reported not renting their pumps out to neighbors or not providing irrigation services for a fee. This may be the reason behind such high levels of pump ownership. Detailed data were collected on the ownership and the use of diesel pumps. Two-thirds of the diesel pump owners reported using kerosene as the main fuel because it was cheaper by Rs. 20-40 per liter. Most diesel pump owners used their pump for less than 50 hours in Rabi 2021. Pump ownership, though widespread, remains unequal. Farmers who own an electric pump have the largest average landholding size and the highest MPCE followed by those who own diesel pumps. Farmers renting diesel pumps and those who do not use pumps are significantly poorer (Table 3). The thin or absent pump rental markets (commonly called water markets in the Indian literature after Shah, 1992) underline the need for community irrigation systems like JCLIS to ensure equitable access to irrigation in Jharkhand and elsewhere. 12 Table 3. Average landholding size (acre) and MPCE of farmers with different types of irrigation pump Farmer Type Land Holding (Acre) Monthly Per Capita Expenditure (In Rs.) Don’t use a pump 1.9 738 Own a diesel pump 2.8 808 Rent in a diesel pump 2.1 690 Own an electric pump 3.3 889 Source: Created by authors using primary data collected from farmers Access to grid electricity in Jharkhand has improved rapidly over the last few years. The percentage of households with electricity connections increased from 32 percent in 2011 to 87 percent in 2019. The share of rural households using electricity as the primary source of light—as opposed to kerosene—increased from 20 percent in 2015 to 84 percent in 2019. Even as access to electricity increased, the duration and the quality of the power supply also improved. Between 2015 and 2018, about one-fifth of the rural households in Jharkhand moved from consuming electricity for 0-4 hours/day (tier 0) to 4-8 hours/day (tier 1).4 Encouraged by the recent improvements in power supply, many farmers in Jharkhand are investing in electric irrigation pumps. Because the electricity consumption for irrigation is heavily subsidized in the state, cheap irrigation from the growing number of private electric pumps can make the community irrigation systems promoted by JSLPS and other institutions less attractive to farmers. These implications are discussed in more detail in the final section of the paper. Paddy is the dominant crop in the Kharif season and the largest crop across all three seasons in terms of the cropped area. All farmers in our sample grow Kharif paddy. Pulses, millets, groundnut, and maize are other common Kharif crops. Vegetables dominate the cropped area in the Rabi and the summer season. Potato, tomato, peas, onion, and cabbage are the most commonly grown Rabi crops while tomato, long beans, and squashes are common in Summer. Table A.14 in the appendix shows the cropping pattern of the farmers in our sample in all three seasons. 4Data reported in this paragraph have been taken from the ‘Power for All Factsheet’ available at: https://www.powerforall.org/application/files/9615/8470/4205/FS_Jharkhand_sets_to_improve_rural_electricity_service_and_pr ovide_green_jobs_with_renewables7.pdf 13 5. Results The Impacts of JCLIS We estimate the impact of JCLIS on the irrigated area, the average cost of irrigation, the probability of growing rabi and summer crops and the cropped area in the two seasons, the probability of growing high-value crops, the area under HVA, and the gross value of agricultural output. Estimating the impact of JCLIS is a challenge because the beneficiaries were not selected randomly. Sites for installing the systems were selected based on water availability and the distribution systems were designed according to the topography (Singh, Sharma, and Bihari, 2020). However, site selection and the selection of beneficiaries may also have been driven by farmers’ interest, and by their willingness to pay the Rs. 1,100 membership fee to join a water user group. We find statistically significant differences in the mean values of treatment and the control group for several key variables (Table A.2 in the Appendix). In our regressions, we control for the variables for which there is a statistically significant difference in the mean values of the treatment and the control group. Our results may be biased due to the unobservable or unmeasured differences between the two groups. We run a matched sample estimation using Coarsened Exact Matching to account for the same. Matching results can be seen in Table A.7 of the Appendix. 5.1. Irrigated area The JCLIS can help farmers by increasing the irrigated area and/or by lowering the cost of irrigation. Therefore, we first compare the gross irrigated area by the beneficiaries and the farmers in the control group. Since there is a significant difference between the average land owned by the two groups, we compare the gross irrigated area per acre of cultivable land owned by the farmers. Nearly half of all farmers in our sample own a diesel or an electric pump. Farmers more interested in protection against drought or dry spells may be more likely to own or use a private pump. On the other hand, ownership of the pump may also make it easier to expand the irrigated area. The type of pump one owns also matters because irrigation with electric pumps is significantly cheaper than with a rented diesel pump, whether the farmer owns or rents it. Therefore, we control for the use of different types of private pump-sets by farmers. We also control for a range of demographic characteristics of the household and the respondent (household size, social group, literacy, age, sex, years of experience in farming, etc.). Finally, we control for the block fixed effects to control for any unobserved time-invariant differences across blocks. 14 Table 4 shows the regression results. The gross irrigated area across all three seasons in 2020-21 is the dependent variable in column 1. Column 2 shows the results with gross irrigated area per acre of land across all three seasons in 2020-21 owned as the dependent variable. 15 Table 4. OLS Estimation Results for the effect of JOHAR irrigation on Area under Irrigation Area under Irrigation (1) (2) across all three seasons Gross Area Irrigated Area Irrigated/Acre land JOHAR Farmer (vs non-beneficiary) 0.0220 0.0751** (0.0709) (0.0290) Months of Operation of JOHAR System (Base: <=18 months) >18 months 0.260*** 0.0700* (0.0921) (0.0401) Total Land Holdings (Acre) 0.190*** (0.0261) Pump Access (Base: No source) Diesel (own) 0.547*** 0.268*** (0.0791) (0.0390) Diesel (rented) 0.401*** 0.272*** (0.1000) (0.0465) Electric 0.418*** 0.214*** (0.0702) (0.0292) Pond/Canal 0.449*** 0.205*** (0.112) (0.0351) JOHAR farmer used a supplementary source of -0.0339 -0.0291 irrigation (0.119) (0.0441) Household Size 0.00873 -0.00217 (0.0163) (0.00490) At least one member migrates for work -0.0732 -0.0116 (0.0521) (0.0251) Type of JOHAR pump=Diesel (vs Solar) -0.119 0.0940* (0.109) (0.0479) The proportion of land that is upland 0.166** 0.0508 (0.0833) (0.0367) (0.0534) (0.0171) Caste Category (Base = General) Scheduled Caste (SC) -0.00478 -0.0391 (0.225) (0.0673) Scheduled Tribe (ST) -0.0547 -0.0182 (0.120) (0.0406) Other Backward Classes (OBC) -0.118 -0.00546 (0.136) (0.0442) Gender of Respondent = Male 0.0772 0.00693 (0.0920) (0.0326) Age of Respondent (In years) -0.00564* -0.000730 (0.00295) (0.00136) Literacy of Respondent (Can read and write) -0.00245 -0.00278 (0.0659) (0.0233) Experience of farming in family (in years) 0.00433 0.000913 (0.00430) (0.00173) Farming is the main source of family income 0.0940 -0.00996 (0.0590) (0.0211) Block FE Yes Yes Observations 757 757 R-squared 0.484 0.353 Note: Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 The coefficients in the table above are based on an OLS estimation that analyses the difference in area irrigated between JOHAR beneficiary 16 farmers vs control (non-beneficiary) farmers. Column 1 reports results from a specification where the dependent variable is the total area irrigated by a farmer across the three cropping seasons. Column 2 reports results where the dependent variable is the area irrigated per acre of land owned by the farmer. In each regression, we control for the number of months a JOHAR system has been operational (less than or more than 18 months) as well as the access of a farmer to a diesel pump, electric pump, or a pond/canal vs a farmer with no source of irrigation. In Column 1 we control for the total land holding of a farmer in acres. We also control for whether a JOHAR farmer used a supplementary source of irrigation and whether the pump used was a solar or diesel pump. Finally, we control for other demographic characteristics such as household size, migration, upland plot proportion, use of fertilizers, pesticides, weedicides and farmyard manure, use of hybrid seeds, caste category of family, gender, age, and literacy of the respondent, experience of farming in the family and whether farming is the main source of income. We also include block fixed effects to account for unobserved time invariant difference across blocks. The area irrigated per acre of land owned is significantly higher for the JCLIS beneficiaries, and the difference is greater for older systems. If we look at the gross irrigated area (GIA), instead of irrigated area/acre of land owned, then only the beneficiaries of the older systems have a significant edge over non-beneficiaries. Otherwise, there is no significant difference between the GIA of the two groups. For every acre of land owned by the beneficiaries, GIA is higher by 0.08 acres. Farmers who own or rent a pump-set and those with access to gravity irrigation from a pond or canal irrigate significantly larger portions of their land. Farmers with a higher percentage of land in the uplands also have more land area under irrigation. 5.2. Cost of irrigation In addition to expanding the irrigated area, JCLIS also lowers the cost of irrigation for farmers by providing solar powered pumps whose operating cost is negligible. Table 5 shows the average hourly cost of irrigation from various sources in our study area. It also shows the average hours of irrigation/acre of Rabi crop and the total cost of irrigation/acre of Rabi crop. Irrigation from electric pumps is the cheapest, followed by irrigation from JCLIS. A JCLIS beneficiary saves nearly Rs. 1,500 per acre of irrigated land compared to their neighbor who has to use a diesel pump they own or rent themselves. Ideally, one would like to measure the cost of irrigation in terms of Rs/m3 of water applied but we do not have data on the discharge rates of different types of pump-sets. Since the private diesel and electric pumps are significantly smaller than the JCLIS pumps, their hourly discharge rate would be smaller too. The effective cost-saving/m3 of water applied from a JCLIS compared to a diesel pump may be greater still. 17 Table 5. Average Cost of Irrigation by Source in Rabi 2021 Source (1) (2) (3 = 1 * 2) of irrigation Average Cost Average Irrigation use The overall cost of irrigation (Rs/hour) intensity (Hours/acre) (Rs/acre) JCLIS 29 19 551 Own diesel pump 75 27 2025 Rented diesel pump 120 19 2280 Own electric pump 8 31 248 Source: Calculated by authors using primary data collected from farmers. Note: This table has been generated using simple means or average values of hours irrigated/acre and average hourly cost of irrigation. These two values are multiplied to generate the average cost of irrigation/acre for each group of farmers. JCLIS refers to the cost for JOHAR farmers while the diesel/ electric/ rental pump values pertain to control farmers. 5.3. Cropped area Jharkhand has one of the lowest cropping intensities among all states of India. Limited access to irrigation is a major reason for the low cropping intensity in the state. Large areas in the state are under the rice-fallow system where no second crop is grown after the Kharif rice is harvested. The cropped area in Rabi and the summer season can increase if farmers have access to affordable irrigation. Kharif is the main crop growing season, paddy in particular. However, the JCLIS beneficiaries are much more likely to grow a Rabi or a Summer crop (Figure 2a). The JCLIS affects Rabi and Summer cropping on both extensive and intensive margins. The beneficiaries are not only more likely to cultivate in Rabi and summer seasons, but they also bring larger portions of their land under cultivation in both seasons (Figure 2b). We collected data from all respondents on the area left fallow in the Rabi and the summer season and asked them the main reasons for leaving land fallow. 64 percent of the non-beneficiary farmers cited lack of access to irrigation as the reason for not cultivating the land, compared to 45 percent of JCLIS beneficiaries. 18 Figure 2a. The proportion of beneficiary and non-beneficiary farmers growing at least one crop in the three cropping seasons 99% 99% 89% 65% 43% 28% Kharif 2020 Summer 2020 Rabi 2021 Johar Beneficiary Non-Beneficiary Figure 2b. The cropped area as a percentage of cultivable land in Rabi and Summer seasons 32% 22% 7% 6% Summer 2020 Rabi 2021 Johar Beneficiary Non-Beneficiary Source: Created by authors using primary data collected from farmers 5.4. Cultivation of vegetables Engaging in high-value agriculture is essential for smallholders to earn a decent income from agriculture. Therefore, the JOHAR project encourages and supports farmers to grow high-value crops (HVCs). Jharkhand has favorable agro-climatic conditions for horticultural crops. Our data shows that access to irrigation from JCLIS increases the likelihood of a farmer engaging in high value agriculture (HVA) on both extensive and intensive margins. The beneficiaries are more 19 likely to grow a vegetable crop in all three seasons (Figure 3a) and they allocate a larger portion of their cultivable land to HVCs (Figure 3b). Figure 3a. Probability of cultivating a high-value crop across beneficiaries and control groups in 2020-21 89% 81% 67% 60% 39% 31% 27% 24% Kharif 2020 Summer 2020 Rabi 2021 Overall Johar Beneficiary Non-Beneficiary Figure 3b. Percentage of cultivable land allocated to high-value crops in 2020-21 34% 29% 26% 22% 19% 17% 15% 15% Kharif 2020 Summer 2020 Rabi 2021 Overall Johar Beneficiary Non-Beneficiary Source: Created by authors using primary data collected from farmers Column 1 in Table 6 shows regression results from a linear probability model where the dependent variable is whether a farmer cultivated any high-value crops across all three seasons in 2020-21. Column 2 reports results on the effects of JOHAR irrigation on the share of cultivable land under HVCs. Whether the farmer is a beneficiary of a JCLIS is the main variable of interest in both the regressions. We also control for access to irrigation from other sources (own or rented pump or gravity irrigation), distance to markets (because HVCs are grown for the market), whether the 20 farmer is a member of the JOHAR producer group (because the producer group members receive support and encouragement to grow HVCs) and the usual demographic controls. We find that JCLIS beneficiaries were significantly more likely to grow an HVC in 2020-21, and that they allocated a greater share of their land to HVCs. Members of JOHAR producer groups are not more likely to grow an HVC or allocate a larger portion of land to HVCs if they are not beneficiaries of a JCLIS. Larger households with more farming members are more likely to grow vegetables, probably because vegetable cultivation is highly labor-intensive, and this confers an advantage to those with more working hands. Yet larger farming households do not increase the proportion of their land that is cultivated with HVCs. Distance to markets or mandis does not have a significant effect on the extensive or the intensive margins of the uptake of HVA by farmers. Access to irrigation from JCLIS or other sources has the largest impact on the uptake of HVA in our study area. We also find a higher probability of summer cropping among JCLIS beneficiaries versus non-beneficiaries. Farmers mostly grow fruits and vegetables in the summer season. Table 6. Regression Estimates for effects of JCLIS on High-Value Crop Cultivation (1) (2) High value crop cultivation Cultivated HV crops The area under HV across all three seasons across all seasons? cultivation/Acre land (Yes/No) JOHAR Farmer (vs non-beneficiary) 0.160*** 0.111*** (0.0463) (0.0390) Months of Operation of JOHAR System (Base: <=18 months) >18 months 0.0164 0.0184 (0.0516) (0.0492) Total Land Holdings (Acre) 0.00277 (0.00551) Pump Access (Base: No source) Diesel (own) 0.467*** 0.273*** (0.0432) (0.0379) Diesel (rented) 0.327*** 0.0796** (0.0601) (0.0359) Electric 0.474*** 0.196*** (0.0342) (0.0289) Pond/Canal 0.363*** 0.220*** (0.0506) (0.0506) JOHAR farmer used a supplementary source of -0.0818** -0.0676 irrigation (0.0411) (0.0541) Household Size 0.0168*** 0.00283 (0.00601) (0.00636) At least one member migrates for work -0.0145 -0.00882 (0.0179) (0.0137) Type of JOHAR pump=Diesel (vs Solar) 0.0261 -0.0228 21 (0.0624) (0.0788) The proportion of land that is upland -0.0175 0.115** (0.0509) (0.0466) Caste Category (Base = General) Scheduled Caste (SC) -0.106 -0.0624 (0.107) (0.0833) Scheduled Tribe (ST) 0.00369 -0.0339 (0.0695) (0.0392) Other Backward Classes (OBC) -0.0371 0.0203 (0.0659) (0.0445) Gender of Respondent = Male -0.0824 -0.0318 (0.0608) (0.0416) Age of Respondent (In years) 0.00417** 0.000284 (0.00193) (0.00152) Literacy of Respondent (Can read and write) 0.0412 0.0288 (0.0253) (0.0248) Experience of farming in the family (in years) -0.00265 0.000573 (0.00247) (0.00194) Farming is the main source of family income 0.0181 -0.0156 (0.0321) (0.0271) Member of the JOHAR Producer Group (PG) in the village (0.00411) (0.00453) 0.0948 0.0623 Block FE (0.0617) (0.0419) Observations 757 757 R-squared 0.406 0.215 Note: Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 The coefficients in the table above are based on an OLS estimation that analyses the difference in high value crop cultivation (cultivation of fruits and vegetables) between JOHAR beneficiary farmers vs control (non-beneficiary) farmers. Column 1 reports results from a Linear Probability Estimation where the dependent variable is the probability of cultivating an HV crop by a farmer across the three cropping seasons. Column 2 reports OLS results where the dependent variable is the area under HV crop cultivation per acre of land owned by the farmer. In each regression, we control for the number of months a JOHAR system has been operational (less than or more than 18 months) as well as the access of a farmer to a diesel pump, electric pump, or a pond/canal vs a farmer with no source of irrigation. In Column 1 we control for the total land holding of a farmer in acres. We also control for whether a JOHAR farmer used a supplementary source of irrigation and whether the pump used was a solar or diesel pump. Finally, we control for other demographic characteristics such as household size, migration, upland plot proportion, use of fertilizers, pesticides, weedicides and farmyard manure, use of hybrid seeds, caste category of family, gender, age and literacy of the respondent, the experience of farming in the family and whether farming is the main source of income, membership of the JOHAR Producer Group. We also include block fixed effects to account for unobserved time invariant difference across blocks and the distance to the closest fruit/vegetable market in the village . 5.5 The gross value of agricultural output in the Rabi season Better access to irrigation can increase the gross value of agricultural output by expanding the area cropped, by raising crop yields, and by facilitating a change in the cropping pattern towards more water-intensive, higher-value crops. In this sub-section, we compare the gross value of crop output in the Rabi season of JCLIS beneficiaries with non-beneficiaries in the same block. The gross value of agricultural output, therefore, combines all three effects: the effect on the cropped area, crop yields, and choice of what crops to grow. We focus on the Rabi season because our analysis shows that irrigation is redundant in the Kharif season when the rainfall is normal like it was in Jharkhand in 2020, and there was very little area under cultivation in the summer season even in JCLIS command areas (only 6-7% of the total cultivable land) due to limited availability of water and other issues. 22 Table 7 shows results of OLS models with the Y-variable expressed in rupees of the gross value of output (column 1) and also as the natural log of the rupees of the gross value of the output (column 2) in the Rabi season. We estimate the value of Rabi output from plot-level data of the total production of all crops grown by the farmers in the season and the median price realized for the crop. We value the entire production, including the portion saved for home consumption or use as seeds, at the price farmers received for the portion they sold in the market. 200 of the 757 farmers in our sample did not cultivate any Rabi crop in 2020-21. The gross value of crop output in the Rabi season for these farmers is zero. This significant clustering of the values of the Y variable around 0 is the reason we also estimate a Tobit model (column 3). Tobit is our preferred model here because the OLS results are likely to be biased and inconsistent. Table 7. OLS and Tobit Regression Estimates on the effects of JCLIS on Gross Value of Output in Rabi 2021 (1) (2) (3) Gross Value of Output in Rabi 2021 Log GVO GVO (Rs.) Tobit-GVO (Rs.) JOHAR Farmer (vs non-beneficiary) 1.275*** 4,864* 9,627*** (0.430) (2,747) (3,507) Months of Operation of JOHAR System (Base: <=18 months) >18 months 0.320 -647.0 1,463 (0.547) (4,148) (4,757) Total Land Holdings (Acre) 0.0684 990.0** 1,050** (0.0438) (423.8) (490.1) Pump Access (Base: No source) Diesel (own) 4.523*** 12,606*** 26,603*** (0.360) (2,403) (3,210) Diesel (rented) 3.503*** 10,068** 23,468*** (0.423) (4,598) (5,504) Electric 4.607*** 8,022*** 23,088*** (0.315) (1,946) (2,906) Pond/Canal 3.399*** 4,362* 16,327*** (0.595) (2,550) (3,740) JOHAR farmer used a supplementary source of -0.417 -1,004 -3,632 irrigation (0.349) (4,335) (4,402) Household Size 0.187*** 1,076** 1,631*** (0.0482) (437.1) (539.2) At least one member migrates for work -0.160 -2,099*** -2,408** (0.119) (758.6) (978.3) Type of JOHAR pump=Diesel (vs Solar) 0.228 -2,792 -2,682 (0.683) (3,442) (3,808) The proportion of land that is upland 0.152 3,277 2,494 (0.447) (3,232) (4,433) Used urea -0.690** 2,240 -43.39 (0.341) (1,601) (2,273) Used DAP 0.752 -1,083 3,286 23 (0.474) (1,928) (3,137) Used farmyard manure (FYM) 0.230 -4,278 -4,269 (0.400) (3,835) (4,699) Used Potash 0.725*** 4,020** 6,160*** (0.273) (1,797) (2,360) Used pesticide 0.603** 5,234*** 6,351*** (0.257) (1,960) (2,375) Used weedicide -0.283 -4,241 -6,160 (0.650) (3,513) (4,696) Hybrid variety of seeds used (vs Local) 0.184 3,077** 3,946** (0.248) (1,533) (1,994) Caste Category (Base = General) Scheduled Caste (SC) -0.548 300.2 -1,924 (1.007) (6,730) (8,641) Scheduled Tribe (ST) 0.340 -3,728 -4,065 (0.474) (4,521) (5,516) Other Backward Classes (OBC) 0.561 -2,216 -2,470 (0.480) (4,330) (5,153) Gender of Respondent = Male -0.0621 616.8 -928.5 (0.395) (2,119) (3,080) Age of Respondent (In years) 0.0195 14.33 77.15 (0.0157) (74.85) (107.0) Literacy of Respondent (Can read and write) 0.407* 980.0 2,694 (0.233) (1,315) (1,823) Experience of farming in the family (in years) -0.0212 -127.9 -201.7 (0.0175) (118.0) (149.5) Farming is the main source of family income 0.460* 602.3 2,550 (0.272) (1,509) (2,085) OLS Yes Yes No Tobit No No Yes Block FE Yes Yes Yes Observations 757 742 742 R-squared 0.521 0.198 Note: Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 The coefficients in the table above are based on estimations that analyze the difference in gross value of output in Rabi 2021 between JOHAR beneficiary farmers vs control (non-beneficiary) farmers. Column 1 reports results from an OLS specification where the dependent variable is the log of the total gross value of output for a farmer in Rabi 2021. Column 2 reports results where the dependent variable is the total GVO (measured in Rs.) for a farmer. Column 3 reports Tobit regression estimates where the dependent variable is the total GVO (measured in Rs.), taking into account censoring at 0 for farmers who do not cultivate in Rabi. In each regression, we control for the number of months a JOHAR system has been operational (less than or more than 18 months) as well as the access of a farmer to a diesel pump, electric pump, or a pond/canal vs a farmer with no source of irrigation. We also control for whether a JOHAR farmer used a supplementary source of irrigation and whether the pump used was a solar or diesel pump. Finally, we control for other demographic characteristics such as the total land holding of a farmer in acres, household size, migration, upland plot proportion, use of fertilizers, pesticides, weedicides and farmyard manure, use of hybrid seeds, caste category of family, gender, age and literacy of the respondent, the experience of farming in the family and whether farming is the main source of income. We also include block fixed effects to account for unobserved time invariant difference across blocks. The results of the log-linear model show that the gross value of Rabi output of JCLIS beneficiaries is nearly 3.6 times higher than that of the non-beneficiaries within the same block even after we control for the total land owned and access to other sources of irrigation. The linear OLS models show that the beneficiaries' GVO is higher by Rs. 4,825, but the coefficient is statistically significant only at the 10 percent level. The Tobit model shows the GVO of JCLIS beneficiaries is higher by Rs. 9,629 compared to non-beneficiaries. The GVO is significantly higher for farmers 24 with access to an alternate source of irrigation. The GVO is higher for farmers who own more land, are literate, live with larger families, and rely on farming as their main source of income. Higher input usage also appears correlated with higher GVO; farming households where fewer members migrate for work (more labor availability) and those who are more likely to use farming inputs like potash, pesticides, and hybrid seed variety report significantly higher GVO. 37 percent of the current JCLIS beneficiaries in our sample had access to irrigation from their own or rented diesel pumps for irrigation before the system was installed in their village, while 26 percent used electric pumps for irrigation and 10 percent had access to water from a nearby pond or stream. Only 27 percent of the JCLIS beneficiaries reported having had no access to any source of irrigation before JOHAR. Diesel pump owners and renters switched to irrigating from JCLIS because it was much cheaper. Electric pump owners made the switch because the JCLIS was more reliable compared to the power supply in the village, and faster and more convenient than using a domestic power connection for irrigation, which most farmers did. Our regressions show that the impact of a private source of irrigation—own or rented—is much larger than the impact of access to JCLIS across most key indicators like cropped area, the area under irrigation, the area allocated to HVCs, and the gross value of crop output that we collected data on. The cost of irrigation is the only exception. The cost of irrigation from JCLIS is significantly lower compared to both own and rented diesel pumps. However, farmers who have access to their own irrigation source may be different from other farmers in important ways that also affect agricultural outcomes but are not controlled for in our models. For example, pump owners may be more skilled and motivated farmers. The greater impact of pump ownership may be partly driven by these unmeasured qualities of farmers. We are not able to control for the endogeneity in our models, and therefore comparisons between the impact of the two different types of irrigation access (private and community-managed) may be misleading. 5.6 Satisfaction with the JCLIS We asked beneficiary farmers to list the benefits realized from access to this new source of irrigation. 55 percent of the beneficiaries said that it led to an increase in cropping intensity and crop production. 19 percent of the beneficiaries reported that they had started growing new crops they did not grow earlier, and one-in-three (35%) reported an increase in farm incomes. 7 percent 25 of the beneficiaries said that irrigation from JSIC had contributed to greater food security and an increase in consumption expenditure. 5.7 The low-capacity utilization of the irrigation system The JCLISs confer multiple benefits to farmers in their command area. However, they are severely underutilized. The median system in our sample operated for less than 100 hours/year in 2020-21 and the average annual operating time was only 192 hours. For comparison, Jharkhand receives more than 2,000 hours of solar radiation over about 300 clear sunny days in a year. Neither water scarcity during the summer months, nor the low demand during Kharif season in a normal monsoon year can justify such low levels of capacity utilization of an enterprise with very high fixed costs and low variable costs. Both the hours of operation and the command area served by JCLISs need to increase significantly to make them economically viable. Capacity utilization can increase if the system managers are incentivized to serve more land and farmers. At present, no such incentives are in place. To the contrary, managing the system is easier for the current users if there are only a few users. Figuring out ways to increase their capacity utilization of JCLISs should be the top priority of the Jharkhand government. Without this change, investing more money into new solar powered community irrigation systems is not advisable. 6. The sustainability of the community-led irrigation systems in Jharkhand The JOHAR project has created robust irrigation infrastructure with significant capital investments and technical expertise. The project team also made serious efforts to create community institutions and build their capacity to operate the irrigation systems independently. Once the systems were ready, the responsibility to manage them was immediately transferred to the water user group. The question now becomes will the farmer groups continue to manage these irrigation systems successfully in the years to come? And relatedly, how do we measure or assess the success of an irrigation system? Literature on irrigation management transfer (IMT) uses the seven following indicators to assess the success of community managed irrigation systems: i) irrigation service fee collection; ii) financial viability of the system; iii) timely repair and good maintenance of the infrastructure; iv) fair and equitable allocation of water to all users; v) reliable and adequate water supply; vi) active 26 participation of group members in the management of the system; and vii) low frequency of disputes among group members (Mukherji et al., 2009). We assessed JCLIS projects when they were still quite new. 80 percent of the systems in our sample had been operating for less than two years, and one in four were less than a year old. Therefore, it is difficult for us to assess their performance on some of the indicators listed above or to predict their future success. In this section, we present our conjectures based on the insights from the large existing body of work on irrigation management transfer and from the data we gleaned from our survey of farmers and system operators and managers during semi-structured discussions with them. Since solar pumps have very low operating costs, many groups in our sample were being lax in setting the irrigation fees and collecting them regularly from farmers. Three of the 50 groups had yet to set an hourly irrigation charge, and another nine had yet to start collecting fees from the farmers. Altogether, collection of irrigation fees had not started for 12 of the 44 solar powered systems in our sample. None of the water user groups or their operators maintained any written records of hours of irrigation service provided to different farmers and the fees collected from them. Fifteen of the systems in our sample had experienced a breakdown, and in each of these cases water group members promptly arranged the necessary repairs. In 6 of the 15 cases, the machine was still under warranty, and therefore no payments were needed. In other cases, farmers pooled money to pay for the repair or used the irrigation fee they had collected. Members of water user groups that were not collecting any fees told us that they were not keen to start doing so and suggested that they will pool funds when there is a need for repair and maintenance. This ad hoc approach to mobilize resources for repair and maintenance of systems may become hard to sustain with time as systems get older and need more frequent repairs. This system may also be more taxing for the poorer farmers who may not always have the necessary resources with them. Paying regular fees may be easier for them than finding large sums of money for major repairs or replacements. More than 80 percent of beneficiaries in our sample reported that irrigation from JCLIS was reliable, affordable, fair, and adequate in the Rabi season. Very few of them used any other source of irrigation in the command area of the JCLIS. However, limited availability of water in the dug 27 wells and ponds in the summer season may pose a serious challenge for the fair allocation of water among group members. Some water user groups are responding to this problem by pooling resources to augment recharge by connecting the JCLIS well to a nearby river or reservoir. We did not come across any groups planning demand management strategies like investing in micro- irrigation or restricting the cropped area or crop choices to cope with water scarcity. Most JCLISs in our study had seen only one summer season. In the next few years, it will be interesting to see the strategies different water user groups follow to manage water scarcity. Despite seasonal water scarcity, less than 10 percent of farmers and system managers in our sample reported any disputes related to water allocation, the quality of irrigation service, or fee collection. The few disputes that were reported in our survey related to the sequencing of irrigation among group members or to disagreements over whether the irrigation provided was adequate. In water scarce conditions, some farmers wanted more water in their fields than what they had received. Most of these disputes were mild and were amicably resolved without any external interference. All JCLISs in our sample were operating at low capacity. There is scope to increase their command areas, their total hours of operation, and the number of beneficiaries they serve. Managing the group dynamics, water allocation, and sequencing of irrigation may become more challenging, and disputes may become more frequent with increases the size of water user groups and higher demand for irrigation – especially in times of peak demand or water scarcity. In their early years of operation, JCLISs score high on the fairness of water allocation and the affordability, adequacy, and reliability of irrigation. Farmers value the irrigation assets, and they are keen to maintain them. Small group sizes (average 16 members/group), small and well-defined command areas (average size = 13.6 acres), and homogeneity of identities (same tribe or caste group) and interests of group members may also contribute to the successful management of JCLISs (Agrawal, 2001). In the past, poor electricity supply and the high cost of diesel have been the most pressing problems for community tube well irrigation systems in eastern India (Pant, 1984; Kishore, Joshi, and Pandey, 2017). Most JCLISs do not have these problems because they are powered by solar energy, which is both predictable and free. A meta-analysis of 108 studies of water user associations (WUAs) in public irrigation systems from 20 countries in Asia by Mukherji et al. (2009) also found that pump based lift irrigation (LI) schemes, like the JCLISs in Jharkhand are more likely to be successfully managed than gravity 28 flow systems after they are transferred to the farmers, probably because they afford their users a greater degree of control and can respond to irrigation demands much better than storage or surface diversions schemes. Thus, several design features of JCLISs suggest that these systems may continue to operate successfully under farmer management. Their long-term success, however, requires all water user groups to create a system of regular irrigation fee collection. Increasing the capacity utilization of these capital-intensive assets and evolving fair norms to manage water scarcity in the summer season are other key steps needed to improve the performance and the sustainability of JOHAR irrigation systems. 7. The way forward The government of Jharkhand set an ambitious goal of increasing the gross irrigated area in the state from 0.5 million ha to 0.8 million ha by 2021. Solar powered community irrigation systems created by JOHAR are a capital intensive but environmentally sustainable way to expand the area irrigated. Despite a sharp decline in the cost of solar panels over the last few decades (Chawla, Aggarwal, and Dutt, 2020), solar pumps are still quite expensive, especially if they are installed under a government program like JOHAR (Kishore, Shah, and Tiwari, 2014). However, even with high capital outlays, the amortized cost of irrigation from solar pumps is significantly lower than diesel pumps (Kolhe et al 2002; Odeh et al 2006; Meah, et al., 2008; GIZ, 2013). The sharp increase in the price of diesel in recent years tilts the balance further in favor of solar pumps. The ownership and use of grid connected electric pumps is rising in Jharkhand and electric pumps are cheaper than solar pumps even if farmers were to pay the full cost of electricity. Most farmers in our sample paid highly subsidized tariffs or no tariffs for electricity used for irrigation.5 The improvement in grid power supply in rural Jharkhand poses both an opportunity and a challenge for expanding access to affordable irrigation in the state. If the power supply becomes more reliable, then the government can install electric pumps to power the community lift irrigation systems. The capital cost/ha of the command area created will be much lower for electric pumps. Charging a reasonable price for electricity will also create a significant variable cost of irrigation and incentivize more sustainable use of water. On the other 5 Many connections were illegal. Farmers often used their domestic power supply to irrigate their crops. 29 hand, if the uncontrolled, even illegal, ownership of electric pumps keeps growing, it may encourage unsustainable use of both electricity and water and exacerbate inequality in access. Furthermore, cheap irrigation from electric pumps may dampen farmers’ interest in using and maintaining the community irrigation systems created by JOHAR and other government and non- government programs. It may not be feasible for the state government to subsidize the capital cost of private electric pumps (and dug wells) to ensure more equitable ownership. Ninety percent of the farmers in our sample who owned pumps did not rent them out to their neighbors, which is a common practice in many other parts of India. This near-complete absence of pump rental markets in Jharkhand makes promoting private pump ownership a less effective strategy to expand access to irrigation in the state. Because water markets in Jharkhand are thin or completely missing, the government will have to provide a pump subsidy to almost every farmer whose land is to be irrigated. Even subsidized cheap pumps may be unaffordable and unviable for the poorest of smallholders. Furthermore, a large number of poorly utilized pump-sets may also be an inefficient use of scarce public resources. Access to JCLIS is more equitable than the ownership of diesel or electric pumps and a private water source like dug wells. Investing in public irrigation systems should therefore be an essential element of any government program to increase the irrigated area in the state for purposes of both equity and efficiency. The JCLISs are almost fully funded by the state government. Farmers' contribution is small and private vendors work only as contractors who install the solar panels and the pumps. The government of Jharkhand may explore other models of creating irrigation capacity in which a greater share of the capital cost is covered by the community or by enterprising farmers. Farmers may be allowed to repay a part of the capital expenditure on the system in monthly or seasonal installments from the irrigation fees collected from the beneficiaries. If farmers share the capital cost of the system, they will have a greater stake in maintaining the system. The pressure to repay a part of the CAPEX will also incentivize system managers to increase the capacity utilization by expanding the command area and reaching more farmers. The Agha Khan Rural Support Program (AKRSP) and Claro Energy, a private firm, have field tested this model with success in North Bihar (Durga and Rai). Such models may be tested in parts of Jharkhand where water availability 30 is not a problem. Cost sharing by the community can increase the impact of solar powered community lift irrigation systems and free up public money to bring more area under irrigation. Another possible way to increase the capacity utilization of the JCLISs may be to record the hours of operation and set incentives for system manager(s) for higher use. The data on the hours of operation of solar pumps can be collected remotely with a small additional investment. This approach also needs to be pilot tested in the field before it is scaled up across the state. 31 References Agrawal, A. 2001. Common property institutions and sustainable governance of resources, World Development, 29(10): 1649-1672 Beccar, L., Boelens, R. and Hoogendam, P. (2002) Water Rights and Collective Action in Community Irrigation. In Boelens, R. & Hoogendam, P. (Eds.) Water Rights and Empowerment. Assen, the Netherlands, Koninklijke Van Gorcum. Chawla, K., Aggarwal, M., and Dutt, A. 2020. Analyzing the Falling Solar and Wind Tariffs: Evidence from India. ADBI Working Paper 1078. Tokyo: Asian Development Bank Institute. https://www.adb.org/publications/analyzing-falling-solar-wind-tariffs-evidence-india Durga, N. and Rai, G. undated. Catalysing Competitive Irrigation Service Markets in North Bihar: The Case of Chakhaji Solar Irrigation Service Market. https://www.livelihoods- india.org/uploads-livelihoodsasia/subsection_data/catalysing-competitive-irrigation-service- markets-in-bihar-the-case-of-chakhaji-solar-irrigation-service-market-by-neha-durga-and-gyan- rai.pdf Government of Jharkhand (201x). Jharkhand Vision & Action Plan 2021. GIZ, 2013. Solar water pumping for irrigation: Opportunities in Bihar, India. New Delhi, India. https://tuewas-asia.org/wp-content/uploads/2017/05/30-Solar-Water-Pumping-for- Irrigation-Opportunities-in-Bihar-India.pdf [viewed on 31st October 2021]. Kishore, A., Shah, T., & Tewari, N. P. 2014. Solar irrigation pumps: Farmers' experience and state policy in Rajasthan. Economic and Political Weekly, Vol. 49(10), 55-62. Kishore, A., Joshi, P.K. and Pandey, D., 2017. Harnessing the sun for an evergreen revolution: a study of solar-powered irrigation in Bihar, India. Water International, 42(3), pp.291-307. Kolhe, M., Kolhe, S., Joshi, J., 2002. Economic Viability of Stand-alone Solar Photovoltaic System in Comparison with Diesel-Powered System for India. Energy Economics. 24, 155–165. Meah, K., Ula, S., Barrett, S., 2008. Solar photovoltaic water pumping — opportunities. Renewable and Sustainable Energy Reviews. 12, 1162–1175. Mukherji, A., Fuleki, B., Shah, T., Suhardiman, D., Giordano, M. and Weligamage, P., 2009. Irrigation reform in Asia: A review of 108 cases of irrigation management transfer. Background paper, 3. https://www.researchgate.net/profile/Aditi-Mukherji- 2/publication/311066233_Irrigation_Reform_in_Asia_A_Review_of_108_Cases_of_Irrigation_ Management_Transfer/links/5940ca6eaca272371225214f/Irrigation-Reform-in-Asia-A-Review- of-108-Cases-of-Irrigation-Management-Transfer.pdf Odeh I., Yohanis Y.G., Norton, B., 2006. Economic Viability of Photovoltaic Water Pumping Systems. Solar Energy. 80, 850–860. 32 Narayanan, S. (2021). Agricultural Markets and Value Chains in Jharkhand: Reflections on the Way Forward. Unpublished report prepared for the World Bank. Pant, N., 1984. Community Tubewell: An Organisational Alternative to Small Farmers' Irrigation. Economic and Political Weekly, pp.A59-A66. SASH (2021). Situation Assessment of Agricultural Households and Land Holdings of Households in Rural India, 2019. NSS Report No. 587. National Statistical Office (NSO), Government of India. Satpathy, M. K. (2002) Irrigation for Livelihoods Improvement: Small Holder Tribal Irrigation in Jharkhand. IWMI (International Water Management Institute)-TATA Water Policy Research Program: Annual Partners' Meet 2002. Shah, Tushaar., Shilp Verma, Neha Durga, Abhishek Rajan, Alankrita Goswami, and Alka Palrecha (2016). Har Khet of Pani (Water to Every Farm) Rethinking Pradhan Mantri Krishi Sinchai Yojana (PMKSY). IWMI-Tata Policy Paper. http://www.iwmi.cgiar.org/iwmi- tata/PDFs/iwmi_tata_pmksy_policy_paper_june_2016.pdf Singh, P., Sharma, S., and Bihari, B. 2020. Learnings from Community-Based Small Scale Irrigation in Tribal Areas of Jharkhand, India. South Asia Agriculture and Rural Growth Discussion Note Series 11. 33 Appendix Table A.1: Descriptive Statistics for JOHAR systems and their operators System Variable Mean Std. dev. Min Max Operator Characteristics Gender of Operator Male 96% 0.197949 0 1 Operator Received Training Yes 73% 0.446071 0 1 System Characteristics Pump type Solar 94% 0.239898 0 1 Diesel 6% 0.239898 0 1 Year of Installation 2018 2% 2019 30% 2020 68% Source of Water Perennial Stream 2% 0.141421 0 1 Seasonal Stream 12% 0.328261 0 1 Pond 6% 0.239898 0 1 Dug well 78% 0.418452 0 1 Dam 2% 0.141421 0 1 Is the source of water reliable all year round? Yes 43% 0.5 0 1 How is water brought from outlets to the field? Open channel 8% 0.274048 0 1 Plastic pipes 32% 0.471212 0 1 Buried pipes 60% 0.494872 0 1 If diesel pump, horsepower capacity (hp) 5 96% 0.197949 0 1 7.5 2% 0.141421 0 1 8.5 2% 0.141421 0 1 If solar, number of panels 15.87 0.337318 15 16 If solar, total watt rating 5068.83 140.5162 4725 5600 34 Fee amount 38.3 25.26613 0 150 No fee received till date 24% 0.431419 0 1 Payments made on time by beneficiaries 52% 0.504672 0 1 Operators report outstanding payments by farmers 60% .4948717 0 1 Disputes regarding use of system 10% 0.303046 0 1 Reasons for disputes Sequencing 60% Disagreement on Adequacy of irrigation 20% Water Scarcity 20% Months where system remains idle due to low demand July 62% August 73% September 70% Months where system demand exceeds capacity April 79% May 71% June 43% Designated Command Area 13.6 7.40 2 42 Actual area under irrigation 8.74 5.38 1.5 25 Number of Farmers in Command Area 16.36 5.86 6 32 Hours of operation in Rabi 84.38 99.20 0 516 Hours of operation in Kharif 16.68 38.52 0 215 Hours of operation in Summer 90.76 150.64 0 650 Total area irrigated in Rabi 7.20 5.08 0 22 Total area irrigated in Kharif 3.10 4.46 0 15 Total area irrigated in Summer 4.45 5.43 0 28 System Breakdowns System has broken down at least once 30% 46% 0 1 Days taken to repair system 61.87 131.81 0 450 Cost of repair (in Rs.) 4780.07 10798.72 0 35000 Who paid for the repair PG/Farmers/Warranty/TSP/Not repaired 40% Surplus from fee collection 46.67% External JOHAR funds 13.33% N 50 35 Table A.2: Summary Statistics for JOHAR vs Non-Beneficiary Farmers JOHAR Non-Beneficiary Difference Variables Beneficiary (T-test) Time-Invariant Variables Respondent Gender (Female) 90% 86% -0.0391 Respondent Literacy (cannot read or write) 39% 32% 0.0756** Household size 6.09 5.81 0.286* Farming is the main source of income 80% 69% 0.109*** Caste Category General Category 4% 5% Scheduled Caste 4% 2% Scheduled Tribe 65% 59% 0.0641* Other Backward Classes (OBC) 26% 33% Other household characteristics Monthly household per capita expenditure (Rs.) 705.46 855.33 -149.9*** Number of Assets (total number =36) 9.43 9.07 0.361 At least one member of the family migrates for work 32% 40% -0.0768** Total Operational Land Holding (In Acres) 2.87 2.3 0.558*** Number of plots 3.05 3.14 -0.0963 Own source of water for irrigation 47% 52% -0.0559 Source of irrigation in Rabi 2021 JOHAR pump 74.2% 0.4% Dug well with pump set 19.1% 67% Others (river, pond, canal, tube well) 5% 21% Has received any agricultural extension in the last 10% 9% 1 year? 0.0174 Able to get loan easily 79% 83% -0.0447 Pump Ownership Diesel/ Kerosene Pump 28% 18% 0.100*** Electric Pump 26% 28% -0.0255 Do you rent out your diesel pump to other farmers? 11.10% 9.90% 0.012 (Yes) Input Usage in Rabi 2021 Hybrid Seed Variety 62.00% 58.00% 0.0486 Used urea 85.60% 84.30% 0.0303 Used DAP 93.20% 91.70% 0.0506** Used Potash 55.90% 46.30% 0.164*** Used pesticides 57.40% 51.30% 0.131*** Used weedicides 4.60% 3.00% 0.00899 Hired labor 50.20% 41.70% 0.0347 N 297 460 Source: Created by authors using primary data collected from farmers Note: The assets include livestock, agricultural machinery, household appliances, furniture, phone, computer, automobiles. 36 Table A.3. Access to irrigation JOHAR Beneficiary Non Beneficiary Variables (Treated) (Control) Own source of water for irrigation 46.8% 52.4% If yes, source of water used for irrigation (Multiple options allowed) Own Dug well 81% 78% Neighbour/Friend/Relative’s Dug well 4% 9% Community Dug well 1% 4% Own Tube well 1% 2% Farm Pond 15% 10% Own a diesel pump 30.3% 24.1% Fuel used in pump owned Kerosene 41.1% 24.3% Diesel 56.7% 73.9% Both 2.2% 1.8% Rent out pump-set to other farmers (Yes) 11.1% 9.9% Average hourly rate charged (In Rs.) 80 60 Own an electric pump 25.9% 28.3% (Plot-crop level data) At least one crop irrigated in Rabi 2021 48.8% 35.2% Source of irrigation in Rabi 2021 JOHAR Pump 74.2% 0.4% Dug well with pump set 19.1% 67% River/pond 5% 21% Source of energy for irrigation in Rabi 2021 Solar 69.9% 2% Diesel/Kerosene 13.7% 40.9% Electricity 15.6% 50.2% Reason for not irrigating in Rabi 2021 High demand for system 5.9% 0% Water unavailability 11.8% 28.9% Irrigation not required 76.5% 71.2% At least one crop irrigated in Summer 2020 17.2% 12.1% Source of irrigation in Summer 2020 JOHAR Pump 69.2% 0% Dug well with pump set 21.2% 60.25% River/pond 9.6% 25.5% Reason for not irrigating in Summer 2020 Water unavailability 100% 33.3% Irrigation not required 0% 66.7% Source: Created by authors using primary data collected from farmers 37 Table A.4. Supplementary Irrigation details (For JOHAR farmers who cultivated in Rabi 2021) Variables % N JOHAR farmers who used a supplementary source of irrigation 25% 265 What source of suppl. Irrigation was used? Diesel pump (own) 32% 66 Diesel pump (rented) 14% 66 66 Electric pump 55% Reason for using supplementary source (multiple choices allowed) 66 JOHAR pump was not working 14% 66 Discharge from JOHAR pump was too low 12% 66 Other farmers were using water from the JOHAR pump 18% JOHAR pump is far from the plot 51% 66 Among those farmers who used a supplementary source of irrigation Suppl. JOHAR (Hours of irrigation/Acre) Source Diesel pump (own) 3.80 17.10 Diesel pump (rented) 4.39 19.92 Electric pump 6.58 31.00 Source: Created by authors using primary data collected from farmers Table A.5. Average hours of JCLIS operation in 2020-2021 and water availability Hours of operation Constraints to increasing command Mean Median N area Too few outlets/ hilly terrain 325.0 325.0 2.0 Water shortage 98.8 62.0 27.0 Resource constraints 157.7 170.0 3.0 Low demand 313.0 313.0 2.0 No constraints 323.4 168.5 16.0 Water source Mean Median N Perennial Stream 490.0 490 1 Seasonal Stream 238.5 250 6 Pond 129.7 117 3 Dug well 159.1 70 39 Dam 1075.0 1075 1 Total 191.8 93.5 50 As reported by farmers Water is regularly available for the RABI KHARIF SUMMER JCLIS in this season 84.5% 96.4% 35.6% Source: Created by authors using primary data collected from system operators 38 Table A.6. Estimation results for within village comparison of JCLIS and non-beneficiary farmers (1) (2) (3) (4) Within Village Comparison Gross GVO (Rs.) HV HV Irrigated (Tobit) Area/acre Cropping Proby area/acre JOHAR Farmer vs non-Beneficiary Neighbour 0.134*** 13,283*** 0.130*** 0.126** (0.0355) (3,649) (0.0471) (0.0601) Pump Access (Base: No source) Diesel (own) 0.204*** 21,440*** 0.182*** 0.357*** (0.0496) (3,690) (0.0488) (0.0597) Diesel (rented) 0.163*** 24,352*** 0.00990 0.246*** (0.0600) (8,552) (0.0414) (0.0638) Electric 0.201*** 23,095*** 0.170*** 0.401*** (0.0392) (3,368) (0.0363) (0.0398) Pond/Canal 0.179*** 11,586*** 0.193*** 0.305*** (0.0482) (4,217) (0.0638) (0.0575) Village FE Yes Yes Yes Yes Observations 500 493 500 500 R-squared 0.376 0.298 0.390 Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 (Controlling for system and farmer characteristics) Table A.7. Estimation Results from Coarsened Exact Matching (1) (2) (3) (4) (5) CEM Matching Results Area Irrigated/ Log GVO GVO (Rs.) ProbY of HV HV area/acre Acre across Rabi 2021 Rabi 2021 cropping across across seasons seasons seasons JOHAR farmer vs non- 0.123*** 1.637*** 5,978*** 0.185*** 0.105*** Beneficiary Farmer (0.0291) (0.353) (1,756) (0.0369) (0.0327) Operational Land Holdings 0.333*** 1,870*** 0.0248*** (Acre) (0.0830) (491.1) (0.00901) Household Size -0.00638 0.0266 -75.72 0.00230 -0.00462 (0.00500) (0.0682) (249.4) (0.00813) (0.00582) Block FE Yes Yes Yes Yes Yes Observations 631 631 620 631 631 R-squared 0.139 0.173 0.090 0.183 0.060 Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 Note: The coefficients above are from regressions using weights from Coarsened Exact Matching Estimation where treated and control farmers are matched on pump ownership (diesel, electric, other), caste category, respondent gender and literacy, farming is the main source of income, and the percentage of upland land owned by the farmer. We control for family size in all regressions and for total operational land holdings (in acres) in Columns (2) , (3) and (4). 39 Table A.8. OLS Estimation Results for the effect of JOHAR irrigation on Area under Irrigation (1) (2) Area under Irrigation Gross Area Area Irrigated/Acre across all three seasons Irrigated land JOHAR Farmer (vs non-beneficiary) 0.0134 0.0716** (0.0728) (0.0283) Months of Operation of JOHAR System (Base: <=18 months) 0.263*** 0.0742* >18 months (0.0948) (0.0377) Total land holdings (Acre) 0.189*** (0.0260) Pump Access (Base: No source) Diesel (own) 0.558*** 0.266*** (0.0811) (0.0397) Diesel (rented) 0.403*** 0.275*** (0.101) (0.0478) Electric 0.411*** 0.214*** (0.0702) (0.0296) Pond/well/canal 0.423*** 0.194*** (0.118) (0.0361) JOHAR farmer used a supplementary source of -0.0353 -0.0224 irrigation (0.120) (0.0440) Household size 0.00878 -0.00231 (0.0163) (0.00498) At least one member migrates for work -0.0738 -0.00941 (0.0530) (0.0258) Type of JOHAR pump= Diesel (vs Solar) -0.120 0.103** (0.112) (0.0469) The proportion of land that is upland 0.165* 0.0500 (0.0834) (0.0359) Caste category Scheduled Caste (SC) 0.0104 -0.0358 (0.224) (0.0688) Scheduled Tribe (ST) -0.0529 -0.0137 (0.122) (0.0429) Other Backward Classes (OBC) -0.110 0.00160 (0.137) (0.0469) Gender of Respondent = Male 0.0756 0.00976 (0.0960) (0.0348) Age of Respondent -0.00564* -0.000531 (0.00301) (0.00137) Literacy if Respondent (Can read and write) -0.00105 -0.00179 (0.0663) (0.0230) Experience of farming in family (in years) 0.00468 0.00104 40 (0.00441) (0.00174) Farming is the main source of income 0.0940 -0.0149 (0.0593) (0.0213) NDVI Summer 0.391 0.227 (0.659) (0.289) Monsoon 0.124 -0.295** (0.285) (0.131) Autumn -0.0530 0.369 (0.544) (0.271) Winter -0.474 -0.375 (0.799) (0.321) Block FE Yes Yes Observations 756 756 R-squared 0.485 0.376 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 The coefficients in the table above are based on an OLS estimation that analyses the difference in area irrigated between JOHAR beneficiary farmers vs control (non-beneficiary) farmers. Column 1 reports results from a specification where the dependent variable is the total area irrigated by a farmer across the three cropping seasons. Column 2 reports results where the dependent variable is the area irrigated per acre of land owned by the farmer. In each regression, we control for the number of months a JOHAR system has been operational (less than or more than 18 months) as well as the access of a farmer to a diesel pump, electric pump, or a pond/canal vs a farmer with no source of irrigation. In Column 1 we control for the total land holding of a farmer in acres. We also control for whether a JOHAR farmer used a supplementary source of irrigation and whether the pump used was a solar or diesel pump. Finally, we control for other demographic characteristics such as household size, migration, upland plot proportion, use of fertilizers, pesticides, weedicides and farmyard manure, use of hybrid seeds, caste category of family, gender, age, and literacy of the respondent, experience of farming in the family and whether farming is the main source of income. We also include block fixed effects to account for unobserved time invariant difference across blocks. Table A.9. Regression Estimates for effects of JCLIS on High-Value Crop Cultivation High value crop cultivation (1) (2) across all three seasons Cultivated HV crops The area under HV across all seasons? cultivation/Acre land (Yes/No) JOHAR Farmer (vs non-beneficiary) 0.164*** 0.126*** (0.0451) (0.0386) Months of Operation of JOHAR System (Base: <=18 months) 0.164*** 0.126*** >18 months (0.0451) (0.0386) Total Land Holdings (Acre) 0.00244 -0.0282*** (0.00616) (0.00587) Pump Access (Base: No source) Diesel (own) 0.456*** 0.278*** (0.0438) (0.0379) Diesel (rented) 0.328*** 0.0908** (0.0601) (0.0370) Electric 0.466*** 0.217*** (0.0340) (0.0291) Pond/Canal 0.348*** 0.234*** (0.0529) (0.0508) JOHAR farmer used a supplementary source of -0.0711* -0.0569 41 irrigation (0.0408) (0.0529) Household Size 0.0164*** 0.00677 (0.00619) (0.00648) At least one member migrates for work -0.0130 -0.00970 (0.0183) (0.0135) Type of JOHAR pump=Diesel (vs Solar) 0.0205 -0.0392 (0.0587) (0.0805) The proportion of land that is upland -0.0172 0.115** (0.0502) (0.0467) Caste Category (Base = General) Scheduled Caste (SC) -0.107 -0.0639 (0.101) (0.0968) Scheduled Tribe (ST) 0.0118 -0.0199 (0.0679) (0.0418) Other Backward Classes (OBC) -0.0289 0.0275 (0.0646) (0.0465) Gender of Respondent = Male -0.0805 -0.0400 (0.0602) (0.0389) Age of Respondent (In years) 0.00418** 0.000202 (0.00193) (0.00149) Literacy of Respondent (Can read and write) 0.0404 0.0399 (0.0253) (0.0249) Experience of farming in the family (in years) -0.00262 0.00147 (0.00247) (0.00201) Farming is the main source of family income 0.0143 -0.00244 (0.0320) (0.0270) Member of the JOHAR Producer Group (PG) in the village 0.0919 0.0525 (0.0610) (0.0399) NDVI Summer 0.246 0.242 (0.271) (0.272) Monsoon -0.176 0.0136 (0.166) (0.151) Autum -0.458 -0.364 (0.342) (0.375) Winter 0.619* 0.219 (0.328) (0.365) Block FE Yes Yes Observations 756 756 R-squared 0.413 0.253 Note: Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 The coefficients in the table above are based on an OLS estimation that analyses the difference in high value crop cultivation (cultivation of fruits and vegetables) between JOHAR beneficiary farmers vs control (non-beneficiary) farmers. Column 1 reports results from a Linear Probability Estimation where the dependent variable is the probability of cultivating an HV crop by a farmer across the three cropping seasons. Column 2 reports OLS results where the dependent variable is the area under HV crop cultivation per acre of land owned by the farmer. In each regression, we control for the number of months a JOHAR system has been operational (less than or more than 18 months) as well as the access of a farmer to a diesel pump, electric pump, or a pond/canal vs a farmer with no source of irrigation. In Column 1 we control for the total land holding of a farmer in acres. We also control for whether a JOHAR farmer used a supplementary source of irrigation and whether the pump used was a solar or diesel pump. We control for remote sensing measurement using Normalized Difference Vegetation Index (NDVI) for all four seasons. Finally, we control for other demographic characteristics such as household size, migration, upland plot proportion, use of fertilizers, pesticides, weedicides and farmyard manure, use of hybrid seeds, caste category of family, gender, age and literacy of the respondent, the experience of farming in the family and whether farming is the main source of income, membership of the JOHAR Producer Group. We also include block fixed effects to account for unobserved time invariant difference across blocks and the distance to the closest fruit/vegetable market in the village. 42 Table A.10. OLS and Tobit Regression Estimates on the effects of JCLIS on Gross Value of Output in Rabi 2021 Gross Value of Output in Rabi 2021 (1) (2) (3) Log GVO GVO (Rs.) Tobit-GVO (Rs.) JOHAR Farmer (vs non-beneficiary) 1.215*** 5,017* 9,622*** (0.435) (2,804) (3,524) Months of Operation of JOHAR System (Base: <=18 months) >18 months 0.423 -704.7 1,631 (0.524) (4,190) (4,689) Total Land Holdings (Acre) 0.0630 939.4** 990.9** (0.0429) (430.9) (496.9) Pump Access (Base: No source) Diesel (own) 4.516*** 12,500*** 26,303*** (0.377) (2,499) (3,262) Diesel (rented) 3.616*** 10,165** 23,878*** (0.405) (4,618) (5,472) Electric 4.527*** 7,298*** 22,101*** (0.318) (1,938) (2,868) Pond/Canal 3.388*** 3,914 15,779*** (0.581) (2,581) (3,690) JOHAR farmer used a supplementary source of -0.306 -890.3 -3,146 irrigation (0.342) (4,285) (4,391) Household Size 0.177*** 1,045** 1,567*** (0.0483) (452.1) (555.3) At least one member migrates for work -0.137 -2,079*** -2,301** (0.114) (776.8) (995.4) Type of JOHAR pump=Diesel (vs Solar) 0.108 -3,376 -3,428 (0.646) (3,493) (3,785) The proportion of land that is upland 0.143 3,330 2,642 (0.438) (3,234) (4,386) Used urea -0.690* 2,193 -329.8 (0.348) (1,633) (2,333) Used DAP 0.661 -1,236 2,737 (0.492) (1,962) (3,238) Used farmyard manure (FYM) 0.313 -4,287 -3,902 (0.391) (3,894) (4,771) Used Potash 0.728** 4,523** 6,687*** (0.279) (1,774) (2,374) Used pesticide 0.574** 4,930** 6,030** (0.262) (2,000) (2,430) Used weedicide -0.247 -4,130 -5,884 (0.661) (3,533) (4,796) Hybrid variety of seeds used (vs Local) 0.198 2,905* 3,758** (0.246) (1,466) (1,897) Caste Category (Base = General) Scheduled Caste (SC) -0.466 348.3 -1,608 (0.994) (6,643) (8,590) Scheduled Tribe (ST) 0.477 -3,488 -3,375 (0.471) (4,554) (5,549) Other Backward Classes (OBC) 0.680 -2,357 -2,279 43 (0.500) (4,316) (5,244) Gender of Respondent = Male -0.135 614.2 -1,286 (0.419) (2,043) (2,923) Age of Respondent (In years) 0.0230 7.237 74.83 (0.0156) (74.58) (105.0) Literacy of Respondent (Can read and write) 0.417* 857.2 2,600 (0.238) (1,313) (1,844) Experience of farming in the family (in years) -0.0220 -125.1 -198.8 (0.0176) (122.2) (154.6) Farming is the main source of family income 0.459* 686.3 2,586 (0.256) (1,469) (1,974) NDVI Summer (2.790) (24,294) (30,880) -1.336 4,862 -548.0 Monsoon (1.140) (8,583) (10,995) -2.388 -14,136 -20,968 Autum (2.772) (18,422) (23,885) 6.498** 21,903 39,561 Winter (2.980) (20,931) (28,346) 0.189 219.3 1,417 (0.273) (2,142) (2,833) OLS Yes Yes No Tobit No No Yes Block FE Yes Yes Yes Observations 756 740 740 R-squared 0.529 0.199 Note: Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 The coefficients in the table above are based on estimations that analyze the difference in gross value of output in Rabi 2021 between JOHAR beneficiary farmers vs control (non-beneficiary) farmers. Column 1 reports results from an OLS specification where the dependent variable is the log of the total gross value of output for a farmer in Rabi 2021. Column 2 reports results where the dependent variable is the total GVO (measured in Rs.) for a farmer. Column 3 reports Tobit regression estimates where the dependent variable is the total GVO (measured in Rs.), taking into account censoring at 0 for farmers who do not cultivate in Rabi. In each regression, we control for the number of months a JOHAR system has been operational (less than or more than 18 months) as well as the access of a farmer to a diesel pump, electric pump, or a pond/canal vs a farmer with no source of irrigation. We also control for whether a JOHAR farmer used a supplementary source of irrigation and whether the pump used was a solar or diesel pump. We control for remote sensing measurement using Normalized Difference Vegetation Index (NDVI) for all four seasons. Finally, we control for other demographic characteristics such as the total land holding of a farmer in acres, household size, migration, upland plot proportion, use of fertilizers, pesticides, weedicides and farmyard manure, use of hybrid seeds, caste category of family, gender, age and literacy of the respondent, the experience of farming in the family and whether farming is the main source of income. We also include block fixed effects to account for unobserved time invariant difference across blocks. 44 Table A.11. Estimation results segregating the control group (1) (2) (3) Segregating control group Irrigated Tobit-Linear Area under HV area/acre Rabi GVO cultivation/Acre land Base = JOHAR Farmer Neighbour non-beneficiary -0.0912*** -8,117* -0.102** (0.0323) (4,424) (0.0426) Other village non-beneficiary -0.0618** -10,979*** -0.119*** (0.0311) (3,844) (0.0413) Months of Operation of JOHAR System (Base: <=18 months) >18 months 0.0686* 1,464 0.0190 (0.0402) (6,276) (0.0494) Pump Access (Base: No source) Diesel (own) 0.266*** 26,912*** 0.273*** (0.0393) (2,874) (0.0379) Diesel (rented) 0.268*** 23,512*** 0.0808** (0.0458) (6,099) (0.0360) Electric 0.212*** 23,170*** 0.196*** (0.0292) (2,358) (0.0288) Pond/Canal 0.204*** 16,164*** 0.220*** (0.0348) (2,615) (0.0509) Block FE Yes (1,442) Yes Observations 757 Yes 757 R-squared 0.354 742 0.215 Note: Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 45 Table A.12. Estimation results interacting JCLIS access with private pump ownership (1) (2) (3) Irrigated GVO (Rs.) HV area/acre (Tobit) Area/acre JOHAR Farmer vs Non-Beneficiary Farmer 0.225*** 28,536*** 0.250*** (0.0414) (5,795) (0.0520) Pump Access (Base: No source) Diesel (own) 0.368*** 38,984*** 0.342*** (0.0406) (4,507) (0.0395) Diesel (rented) 0.396*** 37,290*** 0.201*** (0.0569) (7,865) (0.0453) Electric 0.292*** 35,090*** 0.259*** (0.0287) (3,993) (0.0242) Pond/Canal 0.246*** 26,680*** 0.268*** (0.0423) (5,383) (0.0598) JOHAR Farmer * Diesel (own) -0.251*** -27,192*** -0.190** (0.0684) (6,677) (0.0730) JOHAR Farmer *Diesel (rented) -0.336*** -32,758*** -0.301*** (0.0835) (9,922) (0.0782) JOHAR Farmer *Electric -0.224*** -28,674*** -0.203*** (0.0525) (5,356) (0.0633) JOHAR Farmer *Pond/Canal -0.125 -22,090*** -0.139 (0.0810) (7,695) (0.151) Block FE Yes Yes Yes Observations 757 742 757 R-squared 0.382 0.228 Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 Table A.13. Estimation results only for Solar pumps (1) (2) (3) Irrigated GVO (Rs.) HV area/acre (Tobit) Area/acre JOHAR Farmer vs Non-Beneficiary Farmer 0.0762*** 9,775*** 0.110*** (0.0290) (3,756) (0.0390) Pump Access (Base: No source) Diesel (own) 0.268*** 27,129*** 0.274*** (0.0397) (2,948) (0.0385) Diesel (rented) 0.272*** 23,392*** 0.0820** (0.0463) (5,931) (0.0359) Electric 0.214*** 23,400*** 0.200*** (0.0293) (2,327) (0.0289) Pond/Canal 0.204*** 15,920*** 0.225*** (0.0359) (2,667) (0.0515) Block FE Yes Yes Yes Observations 745 730 745 R-squared 0.350 0.217 Robust standard errors in parentheses clustered at village level *** p<0.01, ** p<0.05, * p<0.1 46 Table A.14. Crop distribution across seasons for JOHAR and Non-JOHAR farmer 14a. Rabi 2021 Crops RABI Non-JOHAR JOHAR Potatoes 61% 51% Tomato 31% 30% Peas 27% 32% Wheat 17% 26% Onion 11% 9% Cauliflower 11% 6% Cabbage 10% 9% Mustard 9% 8% Eggplant / Brinjal 9% 10% Garlic 8% 5% Chili 8% 8% Long Beans 5% 7% Okra/Lady Finger 5% 6% Chickpeas 3% 2% Sponge gourd 3% 3% Bitter gourd 2% 2% Green Beans 2% 2% Radish 2% 1% Maize 1% 0% Urad Dal 1% 1% Sweet potatoes 1% 0% Cucumber 1% 4% Melon 1% 3% Oil seeds 1% 0% Green leafy vegetables 1% 2% Bottle gourd 1% 5% Finger millet 0% 0% Groundnut 0% 0% Mung Dal 0% 0% Capsicum 0% 1% Pumpkin 0% 3% 47 14b. Summer 2020 Crops SUMMER Non-JOHAR JOHAR Onion 31% 19% Tomato 26% 29% Long Beans 16% 17% Okra/Lady Finger 14% 14% Bitter gourd 13% 13% Chili 9% 7% Garlic 8% 4% Pumpkin 8% 7% Potatoes 7% 4% Cucumber 7% 8% Melon 7% 6% Bottle gourd 7% 10% Rice 5% 3% Sponge gourd 5% 12% Mango 5% 0% Eggplant / Brinjal 3% 4% Green leafy vegetables 3% 3% Cauliflower 3% 0% Wheat 2% 7% Maize 2% 2% Peas 2% 3% Cabbage 2% 2% Green Beans 2% 1% Sweet potatoes 1% 2% Capsicum 1% 0% Ginger 1% 1% Bananas 1% 0% Radish 1% 1% 48 14c. Kharif 2020 Crops KHARIF Non-JOHAR JOHAR Rice 98% 98% Urad Dal 29% 19% Finger millet 21% 20% Groundnut 20% 14% Maize 13% 11% Tomato 8% 9% Long Beans 7% 10% Okra/Lady Finger 3% 4% Chili 3% 4% Bitter gourd 3% 1% Sweet potatoes 3% 3% Cauliflower 3% 1% Toor Dal 2% 1% Eggplant / Brinjal 2% 1% Ginger 2% 2% Sponge gourd 2% 2% Green Beans 2% 0% Green leafy vegetables 1% 1% Cabbage 1% 0% Potatoes 1% 1% Mung Dal 0% 1% Beans 0% 1% Cucumber 0% 0% Bottle gourd 0% 1% Pumpkin 0% 0% Radish 0% 1% Lentil 0% 0% Mango 0% 1% Papaya 0% 0% Mustard 0% 0% Pearl Millet 0% 0% Wheat 0% 0% Sorghum 0% 0% Buckwheat 0% 0% Barley 0% 0% Soybeans 0% 0% Peas 0% 1% Chickpeas 0% 0% Sugarcane 0% 2% 49