Household level data - some variables were not collected in the end line (due to problems of comprehension, some questins were dropped from the end line and these are not included in the archived data). Some variables were processed when it made sense and reintroduced in the database. This sheet therefore contains all variables that were collected in both the baseline and the end line and additional variables which were added in the end line, and some calculated variables. The number of entries are different in the baseline (2623) and the end line (1751). This is because some households were only surveyed in the baseline and not in the end line, as they could not be contacted again (they may have moved away or were not available during the end line survey despite multiple visits). An additional 57 households not in the baseline are included in the endline. The collection also contains a table which links to the household survey via the unique indentifier ID_HOGAR and presents data on cases of diarrhea among children (Child_Diarrhea). Each entry of this table presents the information for each child under 16 years old belonging to each household. If the household (linked to the household table through ID_HOGAR) has no children one line is still added mentioning “0” under “Nb_of_children”. The sheet gives information about the number of times each child under 16 years old had diarrhoea in the last 12 months before both baseline and end line surveys. The third dataset provides information (from the end line only) on the management of a selected number of conservation contracts per household (Conservation_Contract). It is structured according to the conservation agreements (each entry corresponds to one agreement). Each household in the Household dataset (linked here through the unique identifier ID_HOGAR can have several contracts). This information was collected for all the household’s level 1 and level 2 contracts and was only collected on level 3 contracts if the household had less than 4 level 1 and 2 contracts. Baseline and endline data for control and treatment communities in an RCT evaluation of the Watershared scheme. The Rio Grande catchment in the eastern Bolivian Andes plays an important role by providing irrigation and drinking water services to villages, towns and cities downstream. However, deforestation and extensive cattle ranching due to unsustainable agricultural practices along the rivers adversely affect ecosystem services. In 2011, a Bolivian NGO, Fundación Natura Bolivia (known as Natura), launched a Payment for Watershed Services-like scheme aiming at connecting ecosystem service users (Municipal Government, Water Cooperatives and downstream villages) with services providers (upstream farmers and cattle-ranchers). Known as Watershared internationally and Acuerdos Reciprocos por Agua (Reciprocal Watershed Agreement) within Bolivia, the scheme is based on an in-kind contribution (beehives, fruit seedlings, irrigation tubing, barbed wire and other similar products) paid by downstream water-users (with additional support from Natura’s international donors) to upstream small holder farmers ranging from $1 to 10/year/hectare (with an additional joining incentive of $100). The aim is to support farmers reducing the impact of unsustainable agriculture practices on watershed ecosystems. The scheme emphasizes reciprocal relationships over market-based incentives. Because of Natura’s interest in how this novel scheme would work and its potential to deliver benefits, they took the highly unusual steps of setting the project up as a Randomised Control Trial (with support from the UK-government’s Ecosystem Services for Poverty Alleviation programme and technical input from Harvard University). Baseline studies were carried out before the implementation of the project in 2010 in 2623 households in 129 villages in the area. This included surveys on water quality, biodiversity and (most relevant to this data archive) a socio-economic survey. Data from the socio-economic survey was used to stratify communities into Control or Intervention and only households in the 65 intervention communities were offered the intervention when the scheme was rolled out in 2011. In 2015/2016 an endline survey was carried out, again with support from a grant from the Ecosystem Services for Poverty Alleviation programme. This time Bangor University also played a large role through a project funded by the Leverhulme Trust. The aim was to use the baseline survey, and the randomized trial to evaluate the impacts of the Watershared scheme on farmer livelihoods and environmental sustainability. The endline reached a total of 1694 of the initial baseline (plus 57 additional hosueholds). Research funders: The baseline data (2010) was collected under a grant from espa (Ecosystem Services for Poverty Alleviation) to Fundacion Natura Bolivia (PI Nigel Asquith) “What types of investment can most cost-effectively ensure ecosystem service provision? A randomized program evaluation” (NE/I00436X/1) The endline data (2015/2016) was collected with funding from two projects: an espa funded project to Fundacion Natura Bolivia (PI Nigel Asquith) “Under what conditions can Payments for Environmental Services deliver sustainable improvements in welfare? Learning from a Randomized Control Trial” (NE/L001470/1) and a Leverhulme Trust funded research project to Bangor University (PI Julia Jones) “Can payment for ecosystem services deliver environmental and livelihood benefits” (RPG-2014-056).

A household survey was first implemented in 2010 on 2623 households belonging to 136 communities. 50% of the communities were randomly selected (stratified according to the number of cattle and the human population size in communities). From 2011 to 2014 households in the treatment communities were offered the opportunity to enrol their land in Watershared contracts (a Payment for Ecosystem Service like scheme with in-kind incentives). From September 2015 to April 2016 a follow-up survey was carried out with as many of the initial 2623 households as could be found. We were able to reach 1694 households (including 113 in the treatment and 681 in the control) in the follow up. The drop outs were because some households had moved away, the household’s head had died, or we could not refind the households despite up to five attempts. An additional 57 were also surveyed in the endline.