The effect of short-term subsidies on demand for potable water in rural Bihar, India

The intervention includes three main activities (social marketing, random price offers, and repeat household visits), only one of which (the price offer) will be randomly allocated to survey participants.

1. Social marketing intervention:
Social marketing teams will provide a marketing demonstration about the water delivery products to all 516 households surveyed during the baseline. These social marketers are young men between the ages of 18 and 25 who have been selected from the local community to impart information about the clean water product and share their own experiences and education about the importance of clean water use. If feasible, this messaging may include a short video explaining the possible routes of diarrheal disease transmission in local Hindi language. After social marketers deliver their messages, they will offer respondents a free taste test of the potable water product. Then, they will give a demonstration of the high iron content of the local water sources as compared to SHRI’s bottled water. The marketer will fill two cups with water. The first from the same pump the family uses for water consumption. The second, from SHRI’s bottled water. Marketers then collect guava leaves (older leaves work better for the demonstration), crush them in their hands, and place the crushed leaves in each of the two cups of water. Tannins in the guava leaves react with iron content in the pump water to turn the water black (if arsenic is present in the water, it will react to turn water a bright blue color). This reaction takes roughly two minutes, and families can repeat the experiment themselves if they choose. Social marketers then allow ample time for families to ask questions about the water product, the delivery service, or any of the health benefits of purchasing water deliveries.

2. Random price offer:
Next, enumerators in the baseline will invite households to participate in the random price offer for water delivery. Enumerators will each have a bag with ten small manila envelopes. Within each envelope will be a card displaying either 5 or 10 rupees per bottle for delivery price (5 each). Each respondent will have a 50:50 chance of drawing either a 5 or 10 rupee per delivery card. After the respondent draws a card, they will then be asked to choose whether they would like to participate by purchasing a bottle and dispenser. Bottles and dispensers will be sold for an up-front price of around 250 INR (the final price will be determined based on interviews with pilot households prior to the survey launch and the same price will be set across all 516 households).

If a household would like to participate but cannot afford the initial investment for the bottle and dispenser, they will be allowed to place a 50% deposit for the hardware and will be given the following two weeks to pay off the remaining 50% of the purchase price. During the two-week period, participating households will have the chance to return the bottle and stand for a full refund so long as neither piece of hardware has been damaged or misused, and as long as each is returned in a clean condition. To limit abuse of this trial period, all purchasing households will be asked to sign (or provide a finger print) on a document ensuring that they understand the terms of the loan period, and that they will be responsible for the full price of the product with no refund if the hardware is damaged, lost or stolen.
Those choosing to purchase the hardware required for water delivery will receive a punch card. These punch cards will be identical in size and appearance except for the price noted on the card based on whether the household drew a 5 or a 10 rupee per delivery envelope. The punch cards will be laminated with a string or band to attach them to the narrow neck of the water bottle in an effort to keep punch cards from being misplaced. Upon each daily delivery, the driver will punch one ‘bottle’ image on the card until the end of the month-long period or after 30 deliveries, whichever comes first. Punch cards will include the head of household name, and households will be instructed that punch cards are non-transferable to other households and have no cash value.
At the end of the household visit, families who choose to participate by purchasing water delivery will be asked how often they would like to receive deliveries (every day, every other day, etc.) and a delivery schedule will be established. If families opt to purchase less frequently than once per day when asked at baseline, we may revisit these households during the auditing process to ask if they would like to change the frequency of their delivery schedules.

Arm 1 – Social marketing visit with 50% subsidy for water delivery. Households will be offered a subsidized price of 5INR / delivery with punch-card coupons for one month. At the end of one month, the price will increase to 10INR / bottle for delivery.

Arm 2 – Social marketing visit with no subsidy for water delivery. Households will be offered the market rate of 10INR / delivery with punch-card coupons for one month. At the end of one month, the price will remain the same.

3. Repeat household visits
We will use our team of four to six social marketers to conduct weekly audits of all households in the sample regardless of treatment status or purchase decision. Each auditor will be responsible for visiting some portion of these homes over the course of each week until all families are revisited at least once per week. These visits will serve three purposes: auditing reported deliveries to ensure completion, asking families whether they would like to purchase water if they have not or have stopped purchasing, and as a reminder and encouragement to families to participate in the program.

2019-06-28

2019-08-09

1. Weekly uptake: A binary measure [0,1] each week, of whether the family purchased water at all during the week; delivery reports validated by audits of participants and re-offers among non-participants

2. Endline uptake: A binary measure [0,1] of intent to purchase at endline interview, measured by whether the family responds that they intend to purchase water (yes/no) during the endline interview (in week 5)

The main study outcome is uptake of potable water from SHRI. This outcome will be measured in one of two ways. The first is the percentage of households purchasing water in each of the study weeks during the intervention and follow-up periods. For our primary research question, we will report this measure for week 7 of the follow-up period, capturing the net impact of subsidies on uptake of water deliveries over the full program period. This requires that all households are visited in week 7 and offered water delivery sales. We can also capture this measure in weeks 5 and 6 of the follow-up period to see if uptake changes over time after the price shock for our primary research question.

Our second measure of uptake is captured during our endline survey during the 5th week of deliveries (the week following the subsidy period). During the interview, subjects are asked whether they intend to purchase water delivery or not at 10/INR per bottle. This is measured using variable FS1 of the endline survey (see attached endline survey questions). While this is a reliable measure because it is taken directly from survey enumerators, it presupposes that respondents will purchase water without direct observation of purchase. This could present measurement error since some families may report wanting to purchase water yet change their minds after the interview, and vice-a-versa. This occurred with several households during out pilot study, leading us to place a higher preference on the first triangulated measure of uptake.

Consumption
2. Weekly consumption
(no. of bottles delivered) Count measure of the number of deliveries purchased per week Total number of deliveries purchased each week; delivery reports validated by audits among participants and re-offers among non-participants

We can measure water consumption as the number of bottles delivered to each family per week. Although this measure provides more granularity in measuring the impact of subsidies on demand, it may also be subject to greater risk of misreporting and confounding. First, this measure relies on accurate reporting from both delivery drivers and household auditors. It also depends on accurate recall from interviewed family members who may not have been present for all household deliveries. Further, the number of bottles consumed by each family per week will depend substantially on the number of individuals in each household, thus this outcome measure requires controlling for a number of factors like household size. Thus, although this measure offers greater specificity it may also be subject to greater bias.

At baseline all households are offered a random price for water delivery (5INR/delivery for treatment households or 10 INR/delivery for control households) for the next month. After the random price is offered, families must choose whether to purchase the hardware (bottle and dispenser) required for water delivery to take place. Those choosing to purchase the hardware and start taking deliveries are provided with a punchcard good for 30 days of water delivery (1 bottle per delivery). At the end of the 30 day subsidy period, enumerators revisit the households in our endline survey. All families are then offered the same price (10INR per delivery) at the endline survey. Families will then choose whether to purchase water during the endline week, as well as for two weeks following endline (weeks 6 and 7). We will examine the difference in uptake between the treatment and control groups at week 7 to determine whether subsidies increase or decrease uptake after the subsidy period has ended.

Households will be randomized using the previously described random price draw. Enumerators will have respondents pick an envelope from a bag containing 10 envelopes (5 each containing a slip of paper with 5INR/delivery; 5 each containing a slip of paper with 10INR/delivery) at the end of the baseline survey.

Individual (household) level of randomization.

No

Not applicable.

516 households in total will be randomized into either treatment (50% subsidy) or control (0% subsidy) conditions.

Not applicable.

Where, for a two-tailed test, α=0.05 and our desired power is β=0.80, both by convention, thus t_1=2.04 and t_2=0.85. We expect the true population proportion of the outcome in the absence of the intervention (i.e. the uptake at time t=1 for the no-subsidy group) to be P=0.15 and the proportion of the study sample randomly assigned to treatment to be T=0.5. Finally, following previous research in this area we consider that an economically relevant minimum detectable effect (MDE) size of 10 percentage point difference in uptake between the treatment arm and control (δ=0.1). Given these parameters and conservatively assuming a naïve model with no covariates to improve precision of our estimates, the equation above dictates that a 2-arm trial would require a minimum 500 households. In our pilot study (taking place over three weeks) we experienced a 1.2% attrition rate. We expect similar attrition over the full study period. Thus, we increase our sample to a total of 516 households.

Cameron & Dow Pre-Analysis Plan