Experimental Design
We are conducting the same interventions in 4 distinct experimental samples. Three of these samples are in rural areas, and one is in urban areas.
In rural areas, we combine several data sources. We use the SHRUG database to identify villages (using their polygons). We use DISE data to identify schools, their characteristics (private or public, whether they offer primary schooling, and the proportion of children from Schedule Caste and Schedule Tribe enrolled in the schools), and their location. We also use the SHRUG data to identify villages with high poverty rates and a high proportion of Schedule Caste and Schedule Tribe populations; however, this data comes from the 2011 census.
First, we restrict the sample to rural villages with at least one private school in five study districts: Bemetara, Kanker, Raigarh, Raipur, and Surajpur. Then, we label villages as "unsubscribed" if they have unfilled seats (based on the administrative data from the state RTE program). We label villages as "disadvantaged" if they are in the top tercile of either the fraction of the SC/ST population or the top tercile of the poverty rate.
Then, we divide the sample of all rural villages with at least one private school into three (randomly assigned) groups. Within each of these groups, we randomly pick 80 villages. For the first group, we restrict the selection to villages with under subscription. For the second group, we restrict it to "disadvantaged" villages. For the third group, we do not impose any restrictions. Then, within each group, we randomly assigned 40 villages to the information treatment.
The surveyor goes to the villages selcted. At this stage, he decides whether the site is fit for survey or not. S/he can reject the site mainly on 4 grounds, a. No private school in the locality, b Posh colony, c. Industrial/Market area, d. Not enough houses , e. Site falling in another
district. Once the site is accepted for the sample, the surveyor looks for a randomly given private school within the village. S/he records the location of the pinned school if s/he can locate this school in the vicinity of the chosen village. At this stage, he decides whether the pinned school can be taken as the center point for the village. If yes, he goes ahead and marks the pinned school as
the center point. In case s/he can locate the pinned school but the school is not located such that it can be marked as the center, s/he looks for another center point which preferably should be a private school. In cases where we can’t find any private school which can be marked as the center, they mark an Anganwadi center or a govt. school whichever is centrally located in the community as the center point. In case he cannot locate the pinned school in the village, he looks for another privateschool. If the village has no private school, we dropped the village from the sample. After selecting the center point, he goes around the village and selects 5 points such that the village
is best captured and records the GPS of these five points. Each of the 5 points should be approximately 1 km from the selected
center point. However, in some cases, these points might not be exactly 1km meters from the center point because of some hindrances while recording one or more points (e.g., there is a pond/road/posh colony/ the village is small. In most cases, we mapped the whole village for our survey because villages were small, and we had to reach the village boundary before completing the 1km distance from the center point. For large enough villages, we went up to 1km from the center point on each side.
After we visited each location, we discarded 3 villages from the "disadvantaged" sample and 3 from the "undersuscription" sample because no private schools operating in the area, the area was an upscale colony, or there were no dwellings in the locations.
For urban areas, we followed a different strategy. First, we restricted the sampling to the Raipur/Bhilai/Durg metropolitan area. Specifically, we took the train station of each of these towns and created an 8 km buffer around them. We then identified all the schools within those limits. We removed any school that is more than 1km apart from any other school (we want to keep densely populated urban areas). We then created a buffer of 500 mts around each of the remaining schools. We then made a grid of this area of pixels that are 600x600 mts. We removed any pixel without a private school within 1.5km (walking distance). This is the "sampling" frame. We found the average fee charged by private schools within 1.5km of each pixel. We then classified pixels into deciles of average fees. We sampled 20 pixels from each decile. At this point, we send a surveyor to the pixel. He/she decides whether the site is fit for survey or not. He can reject the site mainly on 4 grounds: a. No Anganwadi or private or govt. school b. Posh colony, c. Industrial/Market area, d. Not enough houses, e. Site falling in another district. Once the site is accepted for the survey, the surveyor looks for all the Anganwadi within the pixel. He selects the Anganwadi that is most centrally located in the community and records its location. For sites that have no Anganwadi falling within the pixel, we look for an Anganwadi within 300 meters outside the box. If they still don’t find any Anganwadi, they look for a govt. school or a private school, which can be marked as the center of the community. After selecting the center point, he goes around the community and selects 5 points to capture the community and record the GPS of these five points. Each of the 5 points should be approximately 300 meters from the selected center point. However, in some cases, these points might not be exactly 300 meters from the center point because of some hindrances while recording one or more points (e.g., there is a pond/road/posh colony on one or more sides). This is the enumeration site.
The final sample has 158 urban "sites", of which 81 are randomly assigned to treatment.
We conducted a baseline survey. During the survey, the information treatment was delivered in treated locations. We only surveyed households with children in the 3-7 age range. We went to a site, to the center (the center that was chosen during the mapping survey), and then outwards. We surveyed the village until we had 30 complete surveys. In the baseline survey, we collected basic household information, demographic details, and child education information. We also gathered information on the household’s socio-economic status. To comprehend the factors influencing the household's decision regarding their child's education, we presented 25% of the surveyed households with hypothetical enrolment choices for their children and asked them to rank these choices in order of preference under various hypothetical scenarios. Additionally, we posed a series of questions to ascertain the household's anticipated benefits and costs of different school options. Finally, in approximately 50% of the urban and rural sites (treated sites), we asked households RTE-related questions and provided them with detailed information on the RTE policy and its application process.
We then randomized individual households (across villages regardless of the "information" treatment status) to receive additional application support. In this round, we offered in-person application support to a random number of households. We visited the selected households and offered them in-person support to complete the application form. Upon document verification, our surveyor filled out the online form, and the application form details were shared with the household.
In rural areas, we provided application support in 160 (out of 240) rural sites. At each site, we’ll have a list of children to provide application support. In total, we provided application support to around 2000 children. In urban areas, we provided application support in all 160 urban sites. We provided support to roughly 1500 children in urban sites.