What Impedes Efficient Adoption of Products? Evidence from Randomized Sales Offers for Fuel-Efficient Cookstoves in Uganda

What Impedes Efficient Adoption of Products? Evidence from Randomized Sales Offers for... Abstract Many consumers do not adopt products with health and wellbeing benefits apparently far greater than their costs. A sales offer combining a free trial, time payments, and the option of returning the product can overcome barriers such as liquidity constraints and poor information about benefits and usability. We tested this sales offer (and alternatives) in an experiment with a fuel-efficient charcoal stove in urban Uganda and a fuel-efficient wood stove in rural Uganda. Consistent with the importance of these barriers, this offer dramatically increased uptake—in urban Kampala, from 4% to 46%, and in rural Mbarara, from 5% to 57%. (JEL: I12, O12, O33, Q56) 1. Introduction Despite the potential of new products and technologies to improve health and wellbeing, products ranging from bednets to household water treatments often have surprisingly low adoption.1 One salient example is the approximately 2.8 billion people who cook with solid fuels (wood, charcoal, dung) on traditional cookstoves (Bonjour et al. 2013). The smoke from these stoves kills roughly four million people a year (Lim et al. 2012), and contributes to deforestation (Bailis et al. 2015) and global climate change (Ramanathan and Carmichael 2008). The high cost of buying or gathering fuel means that the use of traditional stoves can also deepen poverty. If markets worked well, even a poor person would pay for a fuel-efficient cookstove if fuel savings quickly exceeded the cost of the stove. To learn about the market imperfections that limit adoption, we offered fuel-efficient cookstoves at local market prices, but experimentally varied the terms of the sales offer. We tested four sales offers: a cash and carry offer (similar to a retail setting with a posted price), a one week free trial (followed by full payment or returning the cookstove) to test the role of imperfect information about fuel savings and usability, time payments (equal installment payments over four weeks) to examine the role of liquidity constraints, and a one week free trial followed by time payments to examine how informational and liquidity constraints interact. We build upon existing studies that examine how the uptake of (or willingness to pay for) beneficial health products depends on (1) very large discounts (e.g., Cohen and Dupas 2010; Dupas 2009; Kremer et al. 2009), (2) moderate discounts (e.g., Meredith et al. 2013; Mobarak et al. 2012), (3) providing additional information on health or economic benefits (e.g., Beltramo et al. 2015b; Madajewicz et al. 2007; Meredith et al. 2013), (4) linkage with microfinance (e.g., Devoto et al. 2012; Fink and Masiye 2012; Tarozzi et al. 2014), and (5) experimental manipulation of social network effects (e.g., Beltramo et al. 2015a; Luoto et al. 2012; Meredith et al. 2013; Miller and Mobarak 2015). Our study is most similar to Mobarak et al. (2012) and Tarozzi et al. (2014). Mobarak et al. (2012) studied the adoption of fuel-efficient cookstoves in rural Bangladesh and found low adoption levels for two different types of fuel-efficient cookstoves at full price (2% and 5%, respectively), and continued low uptake (5% and 12%, respectively) even with a large (50%) price discount. They conclude that there is an inherently low demand for fuel-efficient cookstoves. Tarozzi et al. (2014) studied the adoption of insecticide treated bednets in India and found low uptake when sold requiring an immediate cash payment (11%) but much higher uptake (52%) when linking the product to microfinance. They argue these results indicate liquidity constraints are an important barrier to adoption. Our study is similar to Mobarak et al. (2012) in that we also study fuel-efficient cookstoves; however, instead of offering a discount price, we charged market prices. Our study pushes further in that our design—which includes sales offers that are designed to alleviate informational and liquidity constraints—allows us to test if low rates of stove uptake at full price are due to inherent low demand (as Mobarak et al. (2012) suggest) or something else (such as the informational and liquidity barriers we test). Our study is similar to Tarozzi et al. (2014) in that we examine the role of liquidity constraints; however, instead of linking the sales of our product with an outside microfinance organization we do so by offering time payments directly from the salesperson. We push further than Tarozzi et al. (2014) by also testing microloans combined with a chance to learn about the product (through the free trial plus time payments offer), as well as a sales offer with only the free trial. As Mobarak et al. (2012) found in Bangladesh, in urban Kampala, uptake was low (4%) for fuel-efficient charcoal-burning stoves with the traditional cash-and-carry offers (we gave households a week to raise funds, if needed). Uptake was much higher (46%) with a sales offer that combined a free trial and time payments. In this experiment, we also tested the free trial (29% uptake) and the time payment (26% uptake) offers separately. We repeated two of these sales offers in a rural district (i.e., Mbarara) in southwestern Uganda, with fuel-efficient wood-burning stoves, and found a similarly large uptake with free trial and time payments (57%) as compared with cash-and-carry (5%) offers. Our confidence in the results’ external validity increased with similar findings in different settings. To understand any health or environmental benefits attributable to fuel-efficient cookstoves requires data on stove usage and other outcomes. We analyze these topics in a subsequent paper. Past studies have had mixed results. For example, Bensch and Peters (2015), Smith et al. (2006, 2010), Smith-Sivertsen et al. (2009), and Yu (2011) found a fairly high usage of new stoves and improvements in measures of health or exposure to emissions. Other studies (e.g., Beltramo and Levine 2013; Burwen and Levine 2012; Hanna, Duflo, and Greenstone 2016) failed to find consistent usage of new stoves and did not find sustained health improvements. 2. Research Setting and Design We ran two randomized control trials. We first describe the key differences between the two experiments and then discuss each in detail. The first experiment was in Kampala, the capital of Uganda, and the second in Mbarara, a rural district about 300 km southwest of the capital. The urban study was done at a household's doorstep individually, whereas the rural study gathered participants in large group meetings before making a sales offer. Purchased charcoal was the dominant fuel type in the urban area, whereas gathered firewood was the most popular in the rural setting. The urban sample had four experimental treatments whereas the rural sample had two treatments. See Table 1 for a listing of the key differences between the two experiments. Table 1. Differences between urban and rural studies. Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team View Large Table 1. Differences between urban and rural studies. Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team View Large 2.1. Urban Study Design The first experiment took place in Kampala, Uganda's capital and largest city, from October to December 2010. The majority of Kampala's households cook with a traditional charcoal stove. Most cooking is outdoors, unless it is raining. We marketed the Ugastove charcoal stove, made by a Ugandan-owned company in Kampala (see Figure 1 for images of a traditional charcoal stove and the Ugastove charcoal stove). The main fuel saving innovations of the Ugastove are its cylindrical sheet metal frame and surrounding heat-insulating ceramic insert. These features reduce the amount of charcoal needed as compared to traditional charcoal stoves. The Ugastove was selected based on evidence that this stove reduces fuel use when tested in controlled settings (Wang et al. 2009) and that it met the voluntary carbon market's Gold Standard in kitchen performance tests (Center for Entrepreneurship in International Health and Development 2008). The Ugastove manufacturer receives carbon credits for the stoves and passes these savings on to customers. Because of the carbon credit subsidy, retail prices in our experiment were $6, $8, or $10, depending on the size. These prices matched what Ugastoves sold for in the local Kampala market. Figure 1. View largeDownload slide Comparison of charcoal stoves: traditional versus Ugastove. Figure 1. View largeDownload slide Comparison of charcoal stoves: traditional versus Ugastove. The Centre for Integrated Research and Community Development (CIRCODU), a research organization based in Kampala that specializes in market research on household energy,2 served as the in-country sales team and data collection partner. For the urban study, we employed a door-to-door sales strategy with enumerators acting as salespeople. Enumerators’ salaries were kept independent from cookstove sales levels to remove incentives for enumerators to target wealthier households. To select households, we began with a list of all Kampala parishes provided by the Ugandan Bureau of Statistics (Kampala has 96 urban parishes with an average population of 12,387 each, Uganda Bureau of Statistics 2005). CIRCODU’s staff categorized about 30% of parishes as “low income”, 60% as “middle income”, and 10% as “wealthy”, based on their personal knowledge. We covered virtually all parishes that were categorized as “low income” or “middle income”. We excluded the “wealthy” parishes, as their residents cook mainly with gas stoves and they would not be interested in charcoal cookstoves. We divided parishes into two to three different “neighborhoods”, ending up with 226 different neighborhoods in our sample. We randomized offers at a sub-parish/neighborhood level to avoid the possibility of upsetting customers who heard that their neighbors received a different sales offer.3 Thus, each household within a neighborhood received the same sales offer. To achieve balance between our study arms, we stratified based on CIRDOCU’s socioeconomic categorizations of the parishes (“low/middle income”). In addition, to reduce the chance of enumerator bias, we developed a work schedule that ensured each enumerator presented each offer approximately the same number of times, and their schedules had a similar breakdown in times and weekdays. In each neighborhood, sales teams of two enumerators marketed the stoves to 10 households in which an adult was home. Enumerators gave presentations to approximately every sixth household, identified by a fixed way finding procedure to ensure randomization.4 Thus, a consumer was unlikely to be approached after having seen or heard enumerators visit their neighbors. To reduce socioeconomic similarity within neighborhoods, after five households, enumerators returned to their vehicle, drove approximately 1 km, and then made the sales offer at five more households (following the same procedure) in the same neighborhood. At each home, the sales team gave a marketing presentation about the stoves and presented the sales offer randomly assigned to that neighborhood. After gathering some basic information pertaining to the household's cooking and fuel purchasing behaviors, enumerators recorded the homeowners’ purchase decisions. 2.2. Urban Sales Offers We tested four sales offers in Kampala. Cash-and-Carry Offer. Consumers were given the opportunity to purchase a stove at a take-it-or-leave-it price. Consumers were told that, if they needed to gather funds or check with family members, they could take a week to decide and pay. In such cases, enumerators returned one week later to collect payment and deliver the stove. Free Trial Offer. If consumers were willing to accept a free trial, enumerators left the stove and returned in one week for the full payment. If consumers did not want to keep the stove, they could return the stove at the end of the week, with no obligation. Time Payments Offer. Consumers were given the opportunity to purchase a stove with four equal weekly installments. This offer included the right to return the stove before all scheduled payments were due, in which case, future payments were canceled but no money refunded (similar to the rent-to-own model in the United States). As with the cash-and-carry offer, enumerators offered to return in a week to deliver the stove and collect the first time payment if consumers wanted to discuss their choice with family members or needed one week to gather the first payment. Free Trial and Time Payments Offer. Consumers were offered a one-week free trial followed by the opportunity to purchase a stove through four equal weekly installment payments. They also received the right to return the stoves before all scheduled payments were due, cancelling future payments, as in the time payments offer. 2.3. Urban Pricing and Sales Visits The stove prices for all sales offers were the standard market prices of $6, $8, or $10, depending on the size. The most popular stove was the $6 version, as it could cook for an average-sized family (i.e., 5–7 people). The most expensive model could cook for 10–12 people. To those offered the cash-and-carry sales terms, enumerators offered the posted prices ($6, $8, or $10, based on stove sizes desired by consumers) or an incentive-compatible Becker–DeGroot–Marschak (BDM) procedure (Becker, DeGroot, and Marschak 1964). In the BDM procedure, enumerators showed each participant a sealed envelope and explained that the stove price was set randomly by their manager and hidden within the envelope. Enumerators asked participants what was the highest price they would agree to pay, explaining that they could purchase the stove at the envelope price if the price they stated was at least as large as the unknown price inside the envelope. If respondents stated a willingness to pay below the envelope price, they would not be able to purchase the stove. Because a stated willingness to pay affects whether someone can purchase a product, but not how much she pays, this procedure provides incentives for respondents to report their willingness to pay truthfully (if participants understand and believe all the instructions). That is, it is not in the best interests of respondents to name a higher price than what the product is worth to them because they may end up agreeing to pay more than they are actually willing to pay. Similarly, if participants understate their true willingness to pay, they might lose the opportunity to buy a stove at the price they were willing to pay. Enumerators followed the BDM procedure to recover a demand curve rather than simply the share of households that accepted a cash-and-carry offer at the stoves’ market prices. Enumerators allowed participants to ask questions prior to participating in the BDM procedure, to ensure they understood. Enumerators also explained to participants that they would have up to seven days to gather funds if needed (exactly the same as the participants offered the posted market price). After consumers made their decisions, enumerators thanked them for their time and offered a small gift (i.e., a bar of soap) in exchange for answering a few more questions. Over a space of weeks, enumerators recorded customers’ payments, return rates, and default rates. The process for the consumer to return a stove was designed to be as simple for the consumer as possible. The consumer could return a stove to an enumerator during a follow-up visit and the enumerator would carry the stove away and no longer visit the consumer. 2.4. Rural Study Design The follow-up experiment took place in 26 rural parishes of the Mbarara district in western Uganda, from March to May 2012. We selected the Mbarara region because it is rural and almost all families cook with wood on traditional three-stone fires. In addition, the district is less than a day from Kampala, families spend considerable time gathering wood, and local leaders were supportive of our project. In contrast to the urban setting, most families cook in a separate cooking hut. See Harrell et al. (2016), Simons et al. (2014, 2017) for additional background of the rural study area. CIRCODU, the same organization that collected our data in the urban study also carried out the rural study. We marketed the Envirofit G3300 wood-burning stove, made by Envirofit International Inc. (see Figure 2 for images of a traditional three-stone fire and the Envirofit G3300). This stove achieves relatively efficient fuel combustion by sending airflow into the fire and directing heat upward to the cooking surface. These design innovations allow fuel to burn at a controlled rate and enable more complete combustion than a three-stone fire. Emissions testing of the Envirofit G3300 in a controlled laboratory setting found average reductions in CO by 65%, particulate matter reductions of 51%, and a reduction in fuel wood use by 50% compared to a three stone fire (see Figure 3 for a copy of the emissions and performance report). Figure 2. View largeDownload slide Comparison of wood burning stoves: three stone fire versus Envirofit G-3300. Figure 2. View largeDownload slide Comparison of wood burning stoves: three stone fire versus Envirofit G-3300. Figure 3. View largeDownload slide Certified Emissions and Performance Report for Envirofit G3300. Figure 3. View largeDownload slide Certified Emissions and Performance Report for Envirofit G3300. 2.5. Rural Sales Offers We randomly selected 12 parishes to receive the traditional cash-and-carry sales offer and 14 parishes to receive the free trial with time payments offer.5 The terms and conditions of both offers were the same as in the urban experiment (see Section 2.2). Within each parish, we recruited a local point person with the help of local government officials. We asked each point person to gather roughly 60 people together for a public sales meeting on a specified day. We told each point person that we would demonstrate a fuel-efficient cookstove and offer it for sale. We did not tell the point person which sales offer his or her parish would receive. At the sales meeting, the study team presented the Envirofit G3300, discussed its benefits relative to traditional three-stone fires, did a cooking demonstration, and presented the terms of the randomly selected sales offer. Then participants completed a questionnaire focused on household cooking and basic socioeconomic indicators. 2.6. Rural Pricing and Sales Visits We purchased stoves from UpEnergy,6 a company that distributed Envirofit stoves in several regions of Uganda but at the time, not in our study area. We offered the stoves for sale at $12 each in four parishes and then, because of high sales, increased the price to $16 in the remaining 22 parishes.7 The $16 price was similar to what UpEnergy's retail partners sold the same Envirofit stoves for in other regions of Uganda. As with the Ugastove, this retail price was reduced in part due to a carbon credit agreement. As in the urban study, participants given the cash-and-carry offer had the option of purchasing the stove that day for cash, or, if they desired additional time to gather funds and/or discuss their choice with family, they could return to purchase a stove at a second meeting about 10 days later. Participants offered the free trial with time payments could sign up for a free trial at the end of the group sales meeting. They could then return their stoves to the village focal point person with no obligation at the end of the free trial.8 If they liked the stove, they could purchase it in four equal weekly payments. Due to scheduling and logistical constraints, households that accepted a free trial generally received their stove about three weeks after the sales meeting. 3. Barriers to Adoption We designed our study to examine market failures that could reduce the purchase of fuel-efficient cookstoves in developing countries, including liquidity or credit constraints, and imperfect information about the quality, usability, and fuel savings of the new appliance. Throughout this section we present a brief theoretical framework that describes some concrete hypothesis about the barriers to demand the different treatment arms were designed to address. 3.1. Imperfect Information about Product Effectiveness and Energy Savings The free trial is designed to remove informational barriers about the fuel savings and usability of the cookstove. A free trial gives consumers an opportunity to try out a stove and decide if it saves fuel and fits the household's cooking needs. Our free trial period was similar (prior to the point money was paid) to a money-back guarantee, which is documented to increase consumers’ willingness to try unfamiliar products when they are unsure of benefits, especially for experience goods9 (Davis, Gerstner, and Hagerty 1995; Grossman 1981; Suwelack, Hogreve, and Hoyer 2011). Although our experiment was designed with this learning channel (about fuel savings and usability) in mind, it is also possible that the free trial could be encouraging additional sales through other channels. The free trial could be a credible signal that the stove will actually save fuel and perform adequately (Shieh 1996; Moorthy and Srinivasan 1995), so rather than influencing sales via a learning channel the free trial is actually serving as a credibility signal. Another possibility is that a trial period may activate norms of reciprocity, which could increase uptake and repayment (Cialdini 2007). We examine disaggregated data to examine whether the sales increases are more likely attributed to the learning channel, or if the free trial activated a credibility signal or some norm of reciprocity. We gathered data on whether households had ever seen or used a fuel-efficient stove, and whether they were aware that fuel-efficient stoves used less fuel. We expect the value of the free trial (and associated learning about fuel savings) to be different between participants that already knew that these stoves saved fuel and those that did not. We hypothesize that households with no knowledge of fuel-efficient stoves would purchase at higher rates in the experimental arms that offered a free trial than households that were already aware of the fuel savings of these stoves. 3.2. Liquidity or Credit Constraints Liquidity constraints are another plausible reason for the low adoption of beneficial products in developing countries (Bensch, Grimm, and Peters 2015; Devoto et al. 2012; Dupas 2011; Mobarak et al. 2012; Tarozzi et al. 2014). In a related study in different Ugandan rural parishes, we documented that willingness to pay—in a within-subject comparison—was on average about 40% higher with time payments (four equal payments over four weeks) than when paying for a new cookstove within a week (Beltramo et al. 2015b). To examine liquidity constraints in our urban sample we examine if stove repayment rates are faster for individuals who have higher weekly charcoal expenditures than those with lower weekly charcoal expenditures. If repayment rates are faster for those with more charcoal expenditures this suggests that liquidity is central (although, given the endogeneity of the choice of uptake, not conclusive). Additionally we examine if any of the treatments induce increases in the size of the stove purchased (recall we offered three sizes that cost $6, $8, and $10 in the urban setting). Although not definitive, we could imagine that treatment arms that reduced liquidity constraints could be associated with the purchase of more expensive stoves (either directly via the installments, or indirectly via the savings of fuel costs). We also gathered data on whether rural households purchase or gather fuel. If stove repayment rates were faster for individuals who purchase fuel it would suggest that liquidity is central (though, as pointed out previously, not necessarily conclusive). If the differences in repayment rates are small, other forces are likely to be more prominent. See Table 2 for a summary of the potential mechanisms, the predictions, and how we test for the mechanism. Table 2. Tests for possible mechanisms. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. View Large Table 2. Tests for possible mechanisms. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. View Large 3.3. Potential Weaknesses of the Free Trial and Time Payments Offer The offer of a free trial followed by time payments has several potential weaknesses. Return rates with this offer will be high if stoves break frequently, especially if consumers are careless with them during the free trial. The free trial will not increase demand if new stoves are a poor fit for the region, for example, if they require cooking patterns so far from traditional practices that the difficulty for cooks to adapt to the new technology outweighs its other benefits. If consumers are frequently not at home, collecting payments becomes costly. Furthermore, both the free trial and time payments offers open the stove provider to consumer moral hazards, such as when consumers move frequently or decide to keep stoves and not pay for them or damage them during the free trial. Low payments are particularly likely if adverse selection exists among consumers, in which those consumers least likely to pay for a stove are more likely to accept free trial and time payments offers. 4. Results 4.1. Verifying Randomization Table 3 presents household summary statistics for the urban Kampala study, for each of the four sales offers. Standard errors were adjusted for clustering at the neighborhood level. The households in each group are similar, with none of the differences across a row being statistically different from zero. Table 3. Household characteristics urban (Kampala) study: summary statistics. Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Notes: Household data collected at initial household sales pitch. We adjust standard errors for clustering at the neighborhood level. Values presented are rounded to two decimal places. View Large Table 3. Household characteristics urban (Kampala) study: summary statistics. Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Notes: Household data collected at initial household sales pitch. We adjust standard errors for clustering at the neighborhood level. Values presented are rounded to two decimal places. View Large In the urban sample, about 87% of respondents were women. Households cooked for about five people at their daily largest meal, of which 95% used charcoal with an average expenditure around $2.60 per week. Table 4 presents the household summary statistics for the rural Mbarara study. The groups are similar. Out of the 20 characteristics shown, only one difference is significant at the 5% level, and three differences are significant at the 10% level, about that to be expected by chance. Although this is what is to be expected by chance these differences are on key variables related to household earnings and how household's secure firewood. More households earn income (93% vs. 89%, difference p = 0.06) and have year-round employment (59% vs. 51%, difference p = 0.09) in the cash and carry offer than in the free trial with time payments offer. These slight differences would be more worrisome if households that were better off (i.e., those that earn income and have year-round employment) received more access to the free trial and time payments offer; however, they did not. A larger percentage of the cash and carry households purchased firewood last month (36% vs. 28%, difference at p = 0.11), whereas a smaller percentage gathered firewood last month (73% vs. 83%, difference p = 0.05). Given that these differences move in opposite directions for households that received the free trial and time payments offer, any possible bias that might be associated with differential practices in acquiring firewood relating to a specific sales offer is minimal. Table 4. Household characteristics rural (Mbarara) study: summary statistics. Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Notes: Household data collected at the parish wide sales meeting. We adjust standard errors for clustering at the parish level. To minimize the effect of outliers the value of assets is bottom and top coded at 2% and 98% of the distribution, respectively. The prices used to calculate asset values are taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Survey (LSMS) published by the World Bank (see Table A.4 in the Appendix). Values presented are rounded to two decimal places, the value in the difference column is calculated prior to rounding. View Large Table 4. Household characteristics rural (Mbarara) study: summary statistics. Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Notes: Household data collected at the parish wide sales meeting. We adjust standard errors for clustering at the parish level. To minimize the effect of outliers the value of assets is bottom and top coded at 2% and 98% of the distribution, respectively. The prices used to calculate asset values are taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Survey (LSMS) published by the World Bank (see Table A.4 in the Appendix). Values presented are rounded to two decimal places, the value in the difference column is calculated prior to rounding. View Large In the rural sample, about 60% of those attending the village sales meeting were women. The largest daily meal was cooked for an average of 5.6 people. About 87% of households used firewood as the primary fuel source, of which approximately 32% had purchased firewood in the last month and 78% had gathered wood in the last month. Slightly less than 80% of households belonged to a savings group, and few had previous experience with other new technologies, such as improved seeds (18%), solar lamps (13%), or fertilizer (6%). 4.2. Urban Sales Offer Results Table 5 presents the results of the urban study in term of purchases, returns, and defaults across the four sales offers. Cash-and-Carry Offer. Only 4% of households given the cash-and-carry offer agreed to purchase the stove at prevailing local retail prices (23 of 579). As mentioned earlier, we offered a stated fixed price or followed a Becker–DeGroot–Marschak price elicitation procedure for the cash-and-carry offer in the urban setting. Of the subsample offered the fixed retail price, 6% accepted (7 of 114), whereas 3% of those in the BDM procedure stated they were willing to pay at least the stove's retail price (16 of 465). The difference in take-up rate across the two procedures is not statistically significant (p-value = 0.21). At the same time, the slightly lower willingness to pay reported with the BDM procedure than for the cash-and-carry offer is consistent with the notion that consumers could have shaved their stated willingness to pay relative to their actual willingness to pay (as in Beltramo et al. 2015b; Berry, Fischer, and Guiteras 2015). Free Trial Offer. Among households that received a free trial offer, 33% accepted (178 of 539). Among those that accepted, 12% returned the stove after the trial (21 of 178), 11% defaulted (19 of 178), and 78% paid in full (138 of 178). Of those in default, 42% paid at least something (8 of 19).10 In total, uptake (minus returns) was 29% (157 of 539), and we collected 90% of the revenue that consumers owed. Table 5. Purchase and payment by sales offer in urban (Kampala) study. Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Notes: Sales offers made at randomly selected urban households. We adjust standard errors for clustering at the neighborhood level. Test of row equality reports the p-value for of the chi-squared statistic testing if uptake across the four offers is equal. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. The one case of a free trial recipient returning the stove after payments began was a person who wanted to pay a small portion of the purchase price when the stove was delivered, but then returned the stove after the free trial. Ten recipients of the free trial offer requested to pay at the time the stove was delivered (in essence not utilizing the free trial), these are marked as paid off early in the free trial offer column. Nine recipients of the free trial offer did not pay the full purchase price at the end of the free trial, but did pay in subsequent weeks. Pairwise comparisons of treatment arms: a. Statistically significant difference (p < 0.001) from treatment in column (1); b. Statistically significant difference (p < 0.001) from treatment in column (2); c. Statistically significant difference (p < 0.001) from treatment in column (3); d. Statistically significant difference (p < 0.001) from treatment in column (4); e. Statistically significant difference (p < 0.10) from treatment in column (2); f. Statistically significant difference (p < 0.10) from treatment in column (3). View Large Table 5. Purchase and payment by sales offer in urban (Kampala) study. Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Notes: Sales offers made at randomly selected urban households. We adjust standard errors for clustering at the neighborhood level. Test of row equality reports the p-value for of the chi-squared statistic testing if uptake across the four offers is equal. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. The one case of a free trial recipient returning the stove after payments began was a person who wanted to pay a small portion of the purchase price when the stove was delivered, but then returned the stove after the free trial. Ten recipients of the free trial offer requested to pay at the time the stove was delivered (in essence not utilizing the free trial), these are marked as paid off early in the free trial offer column. Nine recipients of the free trial offer did not pay the full purchase price at the end of the free trial, but did pay in subsequent weeks. Pairwise comparisons of treatment arms: a. Statistically significant difference (p < 0.001) from treatment in column (1); b. Statistically significant difference (p < 0.001) from treatment in column (2); c. Statistically significant difference (p < 0.001) from treatment in column (3); d. Statistically significant difference (p < 0.001) from treatment in column (4); e. Statistically significant difference (p < 0.10) from treatment in column (2); f. Statistically significant difference (p < 0.10) from treatment in column (3). View Large Time Payments Offer. Among households that received the offer of four equal weekly time payments, 26% accepted the offer (102 of 390). Of those that accepted the offer, only 1% (1 of 102) returned the stove after payments began, 6% defaulted (6 of 102) and 93% paid in full (95 of 102). Among those that paid in full, a third paid the stove off early (32 of 95), whereas 11% paid off late (10 of 95). In addition, 83% (5 of 6) defaults paid at least something. In total, uptake (minus returns) was 26% (101 of 390), and we collected 96% of the revenue that consumers owed. Free Trial and Time Payments. Of households that received the free trial and four equal weekly time payments offer, even more accepted this offer, with 48% accepting the initial free trial (171 of 355). Among those that accepted the free trial, 5% returned the stove after the trial (9 of 171), 6% defaulted (11 of 171), and 88% paid in full (151 of 171). Of those that paid in full, 31% paid off early (47 of 151), whereas 10% paid off late (15 of 151). Among those that ended up defaulting, 100% paid at least something (11 of 11). In total, uptake (minus returns) was 46% (162 of 355), with 97% of the revenue that consumers owed being collected. 4.3. Rural Sales Offer Results Table 6 presents the results of the rural study. Table 6. Purchase and payment by sales offer in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Notes: Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table 6. Purchase and payment by sales offer in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Notes: Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Cash-and-Carry Offer. Only 5% of households given the cash-and-carry offer purchased the Envirofit at prevailing local retail prices (25 of 538). Free Trial and Time Payments Offer. Of households that received the free trial and four equal weekly time payments offer, 62% accepted the initial free trial (538 of 866). Among those that accepted the free trial, 8% returned the stove after the trial (44 of 538), 1% defaulted (5 of 538), and 91% paid in full (489 of 538). Of those that paid in full, 22% paid off early (108 of 489), whereas 6% paid off late (30 of 489). Among those that ended in default, 40% paid at least something (2 of 5). In total, uptake (minus returns) was 57% (494 of 538), with 99% of the revenue that consumers owed being collected. As previously noted, in four of the 26 rural parishes (two parishes for each sales offer), the sales price was $12 rather than $16. Removing the results of these four parishes does not materially change results (see Table A.1 in the Appendix). 4.4. Evidence of Imperfect Information as Binding Constraint Our arguments are consistent with market failures—rather than poor stove design—as a major reason for the observed low demand for fuel-efficient cookstoves in developing countries (Bensch et al. 2015; Mobarak et al. 2012). We offered a free trial to combat the potential market failure of consumers having imperfect information. The missing information could concern the stove's general usability, the stove's performance in cooking common meals, the quantity of fuel it saves, and so forth. The quite large increase in uptake, from 4.0% with a cash-and-carry offer to 29.1% with the free trial offer, suggests the free trial overcomes at least some of these informational gaps. We gathered information about household's pre-existing knowledge of fuel-efficient cookstoves. Approximately 8.2% of the urban sample already owned a fuel-efficient stove, whereas 52.5% of those that did not already own a fuel-efficient stove had heard of what one was, and 53.1% of those that had heard of a fuel-efficient stove knew that they used less fuel than a traditional charcoal stove. In sum, this means 33.1% of the entire sample (combining those that knew about fuel-efficient stoves with those that already owned one) knew that fuel-efficient cookstoves saved fuel. For the two arms of the experiment that were not designed to include a potential learning channel (cash and carry and time payments) the differences in eventual purchase rates were small and not statistically significant between participants that knew that fuel-efficient stoves saved fuel versus those who did not. Uptake (less returns and defaults) for cash and carry was 3.7% among those that did not know fuel-efficient stoves saved fuel and 5.0% for those that did know (the difference of 1.3 percentage points is not significant, p = 0.46), whereas the uptake (less returns and defaults) for the time payment offer was 24.2% versus 25.0% for those not aware of the fuel savings of the new stove versus those aware the stoves were designed to save fuel (the difference of 0.8 percentage points is not significant, p = 0.86). However, in the two experimental arms designed to allow learning (free trial and free trial with time payments) eventual purchase was slightly more likely among those that did not already know about the fuel saving properties of the stoves. Uptake (less returns and defaults) for the free trial offer was 28.2% among those that did not know fuel-efficient stoves saved fuel and 22.9% for those that already knew about the fuel savings properties of the introduced stove. Although the rate of eventual purchase was higher for those without pre-existing knowledge of the fuel saving properties of the stoves, the difference of 5.3 percentage points is not statistically significant (p = 0.19). Uptake (less returns and defaults) for the free trial and time payments offer was 48.5% for those not aware of the fuel savings of the introduced stove versus 36.7% for those that already knew the stoves were designed to save fuel. This difference of 11.8 percentage points is statistically significantly different than zero (p = 0.04). The higher rates of eventual purchase for participants in treatment arms that included a free trial but who were unaware of potential fuel-savings is consistent with learning about the fuel-saving properties of the stoves during the free trial increasing sales. However, it is possible the free trial may have removed other barriers in addition to imperfect product information. Consumers’ perception of the salesperson's trustworthiness may have increased, because free trials are not commonplace. Consumers may have purchased due to a norm of reciprocity after receiving the free trial, or the free trial created an endowment effect. Perhaps the free trial was a credible sign of quality, which was more important to the purchase decision than the information learned during the free trial. Although it is plausible that the higher rates of eventual purchase could be attributed to salesperson's trustworthiness, an endowment effect or norms of reciprocity, we present suggestive evidence supporting the importance of learning about fuel savings from an extension of this study in Dhaka, Bangladesh with water filters (Guiteras et al. 2016). In the related Bangladesh study, the free trial did not increase the adoption of water filters, although time payments had the same large effect as in Uganda with cookstoves. For the water filter, unlike the stove, a short trial cannot show effectiveness. That is, fuel savings show up within a few days of using the new stove, whereas it takes a sample of hundreds of weeks to gain confidence in the health benefits of a water filter. Thus, although this evidence is suggestive (at best) given the different context and product, we interpret this as showing the effect of the free trial is probably due more to learning than due to reciprocity, an endowment effect, or a signal of the salesperson's credibility. 4.5. Evidence of Liquidity as Binding Constraint We offered time payments to test the role of liquidity constraints as a potential market failure. Our offer was equivalent to a 0% interest loan, with four equal payments made weekly. The large increase in uptake from 4.0% with cash-and-carry to 25.9% with time payments suggests that this contract overcomes liquidity constraints. We were surprised that about a third of those who accepted a sales offer with time payments in the urban setting paid it off early (i.e., 33.7% of those offered time payments alone and 31.1% of those offered time payments after a free trial) and about a fifth paid off early in the rural setting (22.1%). We use this unexpected outcome to examine the role of liquidity constraints. If stove repayment rates were faster for individuals who had higher weekly charcoal expenditures it would suggest that liquidity is binding. If the differences in repayment rates are small, other forces are likely to be more prominent. We split the distribution of the urban sample in half based on the amount of weekly charcoal expenditures. When examining the entire sample 4.2% of those in the lower half of the charcoal expenditure distribution paid off their stove early, whereas 5.9% in the upper half of the charcoal expenditure distribution paid off their stove early. This difference is large (about 40% higher) and marginally statistically significant at the p = 0.10 level. Narrowing the sample to focus on the offer that relaxes liquidity constraints the most—the free trial plus time payment offer—we find 9.5% of those in the lower half of the charcoal expenditure distribution paid off their stove early, whereas 17.7% in the upper half of the charcoal expenditure distribution paid of their stove early. This difference is even larger (about 86% higher) and statistically significant at p = 0.03. Although not definitive, these results are suggestive that the liquidity constraint was binding in the urban context. We also examined this relationship in the rural sample, but with less conclusive results. In the rural sample, there are four types of households with regards to gathering and purchasing firewood. Approximately 6.6% of households neither purchased nor gathered wood—that is, they used other fuels higher up the energy ladder such as charcoal; 13.8% purchased wood, but did not gather; 62.6% gathered wood, but did not purchase; and 17.0% both gathered and purchased wood in the previous month. In a logistic regression of membership in these four groups on paying off the stove early, we find much higher (and statistically significant) odds ratios for each group compared to the omitted group of the households that neither purchased, nor gathered their wood. However, none of the three groups that either gathered wood or purchased wood are statistically significantly different from each other. Although higher early repayment rates for households that pay for firewood with either time or with money supports the idea of binding liquidity constraints; we would have also expected the group that only purchased wood with money to have a statistically significantly higher rate of early repayment. Although paying off stoves early can be construed as evidence of the importance of the liquidity channel, we also note that this behavior is also consistent with qualitative evidence that suggests many Ugandans consider debt undesirable. Several respondents, for example, said they were prepaying so that the stove salesperson would not come by for collections: they apparently perceived a stigma in owing additional payments. Prepayments may also have been motivated by the irregular nature of many customers’ incomes, coupled with the challenges of saving. By prepaying when they had cash on hand, they reduced the risk of losing the stove if they had no cash when the next payment was due. It is possible that one attraction of our time payments offer is this flexibility to prepay, in contrast, for example, to many typical microfinance loans. Additionally, we examine liquidity constraints and stove size. Recall that in urban areas we sold stoves in three sizes, with prices of $6, $8, or $10. If liquidity constraints bind more often for larger expenditures, then those who purchase for a lump sum will more often purchase the smallest of the stoves. This test is not perfect as fewer bought with a lump sum than time payments, so selection effects could reverse the effects of liquidity constraints (e.g., if the relatively small share who bought for a lump sum were disproportionately the families with easy access to at least $8). Collapsing offers into two groups (those with lump sum payments and those with extended time to pay), we find that among those that purchased a stove 81% of those with lump sum buyers (cash and carry offer and free trial offer combined) and 77% of those with extended time to pay (time payments offer and free trial plus time payments offer combined) purchased the smallest stove type. To create a p-value, we ran an ordered logit predicting stove size with the sales offer. Our sample is buyers who completed payments. The joint test that lump sum buyers (with or without a free trial) purchased smaller stoves than those offered extended time to pay (with or without a free trial) has a p-value of 0.83. The free trial and time payments offer increased uptake even further to 45.6%. This rate is over 11 times that of the cash-and-carry offer (4.0%) and substantially above the uptake of the free trial (29.1%) or time payments alone (25.9%). An interesting question is whether—when combining the free trial with time payments—these offers are additive (i.e., they alleviate different constraints for different people). If we assume that the offers are additive and that anyone who accepted the cash-and-carry offer would also accept either the free trial or time payments offers, we calculate the union of the free trial only and time payments only offers would induce about a 51.0% uptake (4.0% cash-and-carry offer + 25.1% free trial offer (29.1% − 4.0%) + 21.9% time payments offer (25.9% − 4.0%)). In our experiment, the combined offer with a free trial followed by four equal installment payments induced uptake in 45.6% of the sample. We cannot reject equality (p = 0.14) between 45.6% and 51.0% with a sample size equal to that used in our experiment, which suggests that these offers are additive. Some of the success in the time payments and free trial with time payments offer is likely due to a reallocation of fuel expenditures. The average household in our sample spent $2.61 per week on charcoal in Kampala, and the Ugastove's fuel saving estimates range from 35% to 46% (Partnership for Clean Indoor Air 2011; Wang et al. 2009). This corresponds to $0.91–1.20 per week in charcoal savings for the average household in the urban study. Across the four weeks, a household could have used fuel savings to fund the purchase of $3.64–4.80 worth of the stove. These fuel savings would fund approximately 61%–80% of the $6 stove (i.e., the model chosen 80% of the time). 4.6. Costs Borne by the Supplier The increase in uptake for the free trial and time payments offer comes with additional costs and risks borne by the vendor. The vendor bears the risk of default if a consumer takes a free trial and neither returns the stove nor pays for it; the vendor also bears the risk of damages to returned stoves. Additionally, the vendor needs more working capital to prepurchase fuel-efficient cookstoves. The important question to answer is whether the additional sales gained from any of the marketing offers presented in this paper sufficiently compensate the entrepreneurial vendor for the additional default risks and capital costs. Although exact costs would vary in different contexts, we present conservative estimates of the costs of the four offers in the urban study. Our field staff reported that a salesperson could visit approximately 25 households a day to make a sales pitch and a similar number for collection visits. In Table 7 we normalize effort across the four sales offers so that each salesperson makes 25 household visits per day (either an initial sales pitch or a follow-up collection visit). Therefore, in a month with 20 working days, a sales person can sell about 20 stoves a month with the traditional cash-and-carry offer, 113 stoves a month offering a free trial, 73 stoves per month offering time payments, and 81 stoves offering a free trial followed by time payments (see Table 7, first panel). Table 7. Comparing sales performance and costs in urban setting. Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Notes: The first panel calculates the approximate effort in terms of initial sales visits and followup collection visits required to make 100 sales. We estimate that a salesperson can make 25 household visits per day and has 20 working days per month. We also assume that followup collection visits are equally as time consuming as the initial sales visit. The sales per month column uses this rate of sales and prorates the amount of sales made in a representative month. The second panel uses the sales per month metric from the first panel and calculates additional costs for defaults, the cost of capital, and return losses. We use the default rates observed in our experiment (default rates are the share of revenue that is owed that was not collected in the field experiment) and conservatively estimate capital costs using a 6% monthly interest rate (higher than what microfinance banks in Kampala generally charge). In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate that the vendor suffers damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs could vary depending on the loan terms and how the salesperson structures sales, but this is likely a fair approximation of the working capital needed, and is comparable across the four offers). View Large Table 7. Comparing sales performance and costs in urban setting. Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Notes: The first panel calculates the approximate effort in terms of initial sales visits and followup collection visits required to make 100 sales. We estimate that a salesperson can make 25 household visits per day and has 20 working days per month. We also assume that followup collection visits are equally as time consuming as the initial sales visit. The sales per month column uses this rate of sales and prorates the amount of sales made in a representative month. The second panel uses the sales per month metric from the first panel and calculates additional costs for defaults, the cost of capital, and return losses. We use the default rates observed in our experiment (default rates are the share of revenue that is owed that was not collected in the field experiment) and conservatively estimate capital costs using a 6% monthly interest rate (higher than what microfinance banks in Kampala generally charge). In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate that the vendor suffers damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs could vary depending on the loan terms and how the salesperson structures sales, but this is likely a fair approximation of the working capital needed, and is comparable across the four offers). View Large These higher sales must be weighed against defaults, potential damages to returned stoves, and the cost of capital. To estimate these costs, we used the stove sales per month from the first panel, the default rates we observed in our experiment, and a conservative estimate of 6% per month interest rate (i.e., roughly 100% per year). Additionally, we include a damage charge for returned stoves worth 50% of the value of the returned stove. In our experiment we were able to resell returned stoves as if they were new, but to be conservative we include this damage charge. Even including defaults, possible damage to returned stoves and an interest charge; the experimental offers are preferable to the traditional cash-and-carry offer (see Table 7, second panel).11 Under the stated assumptions, the free trial offer is the vendors’ best option. However, this result only holds if collection visits are as time consuming as the initial sales presentation. As the cost of follow-up visits declines, the free trial and time payments offer becomes more favorable. Sales visits may be slower than follow-up visits because the latter will not include a sales pitch highlighting the stove's features. The vendor also can call or text ahead of time to ensure someone is home with the payment, and the use of mobile money could remove the need to meet. The wide and growing availability of mobile money in developing nations, especially sub-Saharan Africa (Aker and Mbiti 2014; Jack and Suri 2014; Luoto and Levine 2014) suggests that mobile payments will eventually reduce, or perhaps even eliminate, the number of in-person visits required to collect installment payments. However, Luoto and Levine (2014) did not find that collecting time payments with mobile money is profitable in their setting of rural western Kenya. Based on the assumptions underlying Table 7, if the time required for follow-up visits is cut in half, the free trial with time payments becomes the preferred sales offer. The logistical set up for rural sales was different from the urban study as households are widely dispersed and we used a local point person who received a commission for collecting payments. Although the calculations differ (see Table A.2 in the Appendix), the free trial with time payments offer is also preferred to the cash-and-carry offer in rural areas, even when accounting for default risk, a damage charge for returned stoves and increased capital costs. 4.7. Robustness and External Validity In the rural study, meeting attendees who heard they would be offered the free trial with time payments may have sent text messages to friends who might be interested. Such texting could have led to nonrepresentative samples at meetings with different sales offers. To test for this possibility, we reran the results removing the final 33%, 25%, and 10% of participants to arrive at each free trial and time payments meeting. The results were similar irrespective of how many late arrivers were removed (see Table A.3 in the Appendix for results removing the last 25% of arrivers from each meeting). Further questions arose about our results’ level of external validity and the performance of these sales offers with other technologies or products in the developing world. We carried out the field experiment in Kampala, in late 2010, and we were surprised by the large increase in uptake compared to the typical cash-and-carry offer (i.e., over 11 times increase for the free trial with time payments offer). Because the magnitude of this increase was so large, we solicited funding to run the same trial in a different setting. In the rural setting, with a different stove, fuel type, market, and socioeconomic context, the free trial with time payments offer increased uptake by 12 times, about the same as in our first experiment. However, how well the sales offer of a free trial followed by time payments would work with other products is still unclear. Offering time payments without a free trial raises purchase levels considerably for many products. These include insecticide-treated bed nets (Tarozzi et al. 2014), piped water connections (Devoto et al. 2012), and water filters (Guiteras et al. 2016), as well as for charcoal-burning stoves (in this paper's urban study) and for wood-burning stoves (Beltramo et al. 2015b). Furthermore, how well the benefits of a free trial generalize is less obvious. A free trial may decrease sales if the product is unpopular, as Luoto et al. (2012) found for chlorine for water treatment, or if the trial anchors consumers’ willingness to pay at a lower number (Fischer et al. 2014). A free trial will not raise demand if consumers cannot learn much about a product's effectiveness, which perhaps accounts for the lack of detectable effects of a free trial for water filters in Dhaka (Guiteras et al. 2016). Finally, free trials do not work well for products that cannot be returned, such as built-in products (e.g., latrines and some stoves) or consumable products. With these cautions in mind, quite plausibly, many products—besides the cookstove models studied here—could increase sales if vendors offered free trials. Lastly, it is worth noting that the overall welfare of this intervention relies on both the purchase of a fuel-efficient stove (as studied in this paper) and the actual use of the fuel-efficient stove and its potential health and environmental benefits (the subject of future work). Although it appears that households understand the value of the time and/or money the fuel-efficient stove saves, we acknowledge that we cannot assess the complete welfare of the households’ purchase decision without additional data on long-term health benefits and data on fuel use. 5. Conclusion We examined the sale of two different fuel-efficient cookstoves in two different settings in Uganda. In contrast to other studies that have altered cookstove prices and found low demand, we found extremely high demand for fuel-efficient cookstoves sold at local market prices but with altered sales contract terms. In urban Kampala, we offered four contracts for charcoal burning stoves: cash-and-carry (4% uptake), a one week free trial followed by full payment (29% uptake), four equal weekly time payments starting immediately (26% uptake), and a combination of a one week free trial followed by four equal weekly time payments (46% uptake). Default levels were low since the combined offer resulted in recovery of 97% of the purchase price of the stoves sold. We repeated two sales offers in a rural area with a different market, fuel type, stove model, socioeconomic setting, and a slightly different experimental design. We sold wood-burning fuel-efficient stoves but also found a quite large uptake (57%) with the free trial plus time payments offer, when compared to the cash-and-carry offer (5%). Defaults were also low in the rural setting, as we recovered 99% of the stove sales price. Similarly high uptake in both settings reinforces the idea that low demand does not inherently hamper the sale of beneficial environmental and health-improving technologies such as fuel-efficient cookstoves. Instead, other important barriers, including information and liquidity constraints, must be eliminated before these purchases will occur. Appendix Table A.1. Robustness check: Purchase and payment by sales offer in rural (Mbarara) study only in parishes where stove was offered for $16.00 USD. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Notes: The four parishes where stoves were sold for $12.00 USD are excluded to calculate the results in this table. Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.1. Robustness check: Purchase and payment by sales offer in rural (Mbarara) study only in parishes where stove was offered for $16.00 USD. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Notes: The four parishes where stoves were sold for $12.00 USD are excluded to calculate the results in this table. Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.2. Comparing sales performance and costs in rural setting. Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Notes: Sales offers made at village wide sales meeting with approximately 50 attendees each meeting. We simulate one sales meeting per week, and four different villages visited in a month. A village focal point person was recruited to gather the community for the initial sales meeting and then to followup with collecting the installment payments. One village focal point person in each village was paid with a free stove ($16.00 value) and $1.60 per installment payment collected (or $6.40 per stove sold once the full payment amount was collected). We use the default rates and return rates observed in our experiment and conservatively estimate capital costs using a 6% monthly interest rate. In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs would vary depending on the loan terms and actual sales, but this figure is at least comparable across the offers). View Large Table A.2. Comparing sales performance and costs in rural setting. Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Notes: Sales offers made at village wide sales meeting with approximately 50 attendees each meeting. We simulate one sales meeting per week, and four different villages visited in a month. A village focal point person was recruited to gather the community for the initial sales meeting and then to followup with collecting the installment payments. One village focal point person in each village was paid with a free stove ($16.00 value) and $1.60 per installment payment collected (or $6.40 per stove sold once the full payment amount was collected). We use the default rates and return rates observed in our experiment and conservatively estimate capital costs using a 6% monthly interest rate. In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs would vary depending on the loan terms and actual sales, but this figure is at least comparable across the offers). View Large Table A.3. Robustness check: Purchase and payment rates after removing late arrivers in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Notes: As a robustness check we remove the last 25% of the attendees that arrived at each sales meeting. This is to ensure the representativeness of each sample as it is possible that early arrivers to the meeting texted their friends (whom they knew to already be interested in a fuel-efficient cookstove). If those that arrived the latest to the sales meeting had a higher propensity to purchase a stove, our results would be biased. Comparing these results to the full sample results shows no such bias. Results are largely the same for different cutoff points (removing the final 33%, 25%, or 10% of attendees). Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.3. Robustness check: Purchase and payment rates after removing late arrivers in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Notes: As a robustness check we remove the last 25% of the attendees that arrived at each sales meeting. This is to ensure the representativeness of each sample as it is possible that early arrivers to the meeting texted their friends (whom they knew to already be interested in a fuel-efficient cookstove). If those that arrived the latest to the sales meeting had a higher propensity to purchase a stove, our results would be biased. Comparing these results to the full sample results shows no such bias. Results are largely the same for different cutoff points (removing the final 33%, 25%, or 10% of attendees). Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.4. Prices used for construction of aggregate asset values. Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Notes: Price data used to construct value of assets are average prices of durable goods and livestock taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Study (LSMS). This data is publicly available at: econ.worldbank.org. View Large Table A.4. Prices used for construction of aggregate asset values. Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Notes: Price data used to construct value of assets are average prices of durable goods and livestock taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Study (LSMS). This data is publicly available at: econ.worldbank.org. View Large Footnotes 1 Examples of studies examining the reasons for the low adoption of health and welfare improving technologies include (but are not limited to): latrines (Gertler et al. 2015; Guiteras, Levinsohn, and Mobarak 2015), point-of-use water treatment technologies (Albert, Luoto, and Levine 2010; Ashraf, Berry, and Shapiro 2010; Kremer et al. 2009), insecticide-treated bed nets (Cohen and Dupas 2010; Dupas 2009; Tarozzi et al. 2014), hand washing with soap (Halder et al. 2010; Luby et al. 2011; Meredith et al. 2013), vaccines (Clemens and Jodar 2005; Cropper et al. 2004), deworming (Kremer and Miguel 2007; Meredith et al. 2013), and micronutrient fortification (Dewey and Adu‐Afarwuah 2008; Meredith et al. 2013). 2 See the “About CIRCODU” webpage for more details: http://www.circodu.org/about-us/. 3 We did this to provide sufficient separation so that consumers from different neighborhoods would be unlikely to communicate regularly (i.e., they would probably attend different churches, schools, shops, etc.). 4 We elected to use a fixed way finding procedure (i.e., enumerators visited every sixth house while moving down a street in the same direction) as opposed to the more common technique of a full neighborhood census and then a follow-up visit to selected households to make the sales pitch. The fixed way finding procedure was selected to save costs as it required fewer field visits and it would also ensure that the first time a household was contacted they received a sales pitch by the door-to-door salespeople (enumerators). In addition to cost savings, we avoided a census with a follow-up visit as it could have potentially created expectations for assistance on the part of the households if they were initially surveyed about household energy usage and at a subsequent visit received an offer for a device like a fuel-efficient cookstove related to household energy usage. 5 The population of most Ugandan rural parishes ranges from 4,000 to 6,000. 6 See http://upenergygroup.com/projects/uganda/. 7 Because we planned to track usage and health benefits of the stove in a follow-on impact evaluation, we needed to make sure enough stoves sold. We thus started with a lower than market sales price. Once our results clearly showed that the stoves were popular, we increased the price to the approximate market price to ensure that we did not run out of stoves. Excluding the four parishes where the stoves sold for $12 does not materially change our results (see Table A.1 in the Appendix). 8 Because homes were geographically spread out in the rural areas, we hired a local focal point person in each village to organize the sales meetings and to collect the follow-up time payments. We paid the village focal point person for each installment payment he or she collected and as part of this structure he or she was in charge of collecting returned stoves if a consumer chose to return the stove. The focal point person would keep the stove at his or her place of residence until our research team was in the area again and could pick up the returned stove (and use it in another village). 9 An experience good is a product or service whose characteristics, such as quality or price, are difficult to observe in advance but are ascertained upon consumption. 10 The study protocol did not permit households offered the free trial to pay a portion of the stove price. However, when the sales staff went to the house either to collect the stove or full payment, some households offered to pay a portion of the purchase price. We had not anticipated this possibility when we trained the sales staff. In practice, the sales staff accepted partial payment in these cases. Therefore, some ultimate defaulters may have paid at least some of the purchase price in the free trial offer. 11 Note that if default is enough of a problem (higher than what we observed), then this policy may not be feasible. Acknowledgements The editor in charge of this paper was M. Daniele Paserman. Acknowledgments: We received funding from the Goggio Family Foundation, the National Institute of Child Health and Human Development (NICHD R21 HD056581), the Silicon Valley Community Foundation, the Institute for Research on Labor and Employment, the International Institute for Impact Evaluation (3IE), the Haas School of Business, the United States Agency for International Development (Translating Research into Action, Cooperative Agreement No. GHS-A-00-09-00015-00), the Atkinson Center for a Sustainable Future, the Institute for the Social Sciences, and the Cornell Population Center. Data collection was carried out by the Center for Integrated Research and Community Development (CIRCODU), and the project's success depended greatly on its managers—Joseph Arinieitwe Ndemere, Juliet Kyaesimira, Vastinah Kemigisha—and field supervisors—Innocent Byaruhanga, Fred Isabirye, Michael Mukembo, Moreen Akankunda, and Noah Kirabo. Stephen Harrell and Dow-Li Kou (of IPA) expertly oversaw field operations. We thank Impact Carbon, Evan Haigler, Caitlyn Hughes, Matt Evans, Jimmy Tran, Johanna Young, Rashad Korah Thomas, Adam Galinsky, Andy Weiss, and the USAID TRAction Technical Advisory Group. The views expressed in this paper solely reflect those of the authors, and these opinions are not necessarily those of the institutions with which the authors are affiliated. All errors are our own. References Aker Jenny C. , Mbiti Isaac M. ( 2014 ). “ Mobile Phones and Economic Development in Africa .” Journal of Economic Perspectives , 24 ( 3 ), 207 – 232 . Google Scholar CrossRef Search ADS Albert Jeff , Luoto Jill , Levine David ( 2010 ). “ End-User Preferences for and Performance of Competing POU Water Treatment Technologies among the Rural Poor of Kenya .” Environmental Science and Technology , 44 , 4426 – 4432 . Google Scholar CrossRef Search ADS PubMed Ashraf Nava , Berry James , Shapiro Jesse M. ( 2010 ). “ Can Higher Prices Stimulate Product Use? Evidence from a Field Experiment in Zambia .” American Economic Review , 100 ( 5 ), 2383 – 2413 . Google Scholar CrossRef Search ADS Bailis Robert , Drigo Rudi , Ghilardi Adrian , Masera Omar ( 2015 ). “ The Carbon Footprint of Traditional Woodfuels .” Nature Climate Change , 5 , 266 – 272 . Google Scholar CrossRef Search ADS Becker Gordon M. , DeGroot Morris H. , Marschak Jacob ( 1964 ). “ Measuring Utility by a Single-Response Sequential Method .” Behavioral Science , 9 , 226 – 232 . Google Scholar CrossRef Search ADS PubMed Beltramo Theresa , Blalock Garrick , Levine David I. , Simons Andrew M ( 2015a ). “ Does Peer Use Influence Adoption of Efficient Cookstoves? Evidence From a Randomized Controlled Trial in Uganda .” Journal of Health Communication , 20 , 55 – 66 . Google Scholar CrossRef Search ADS Beltramo Theresa , Blalock Garrick , Levine David I. , Simons Andrew M. ( 2015b ). “ The Effect of Marketing Messages and Payment over Time on Willingness to Pay for Fuel-Efficient Cookstoves .” Journal of Economic Behavior & Organization , 118 , 333 – 345 . Google Scholar CrossRef Search ADS Beltramo Theresa , Levine David I. ( 2013 ). “ The Effect of Solar Ovens on Fuel Use, Emissions and Health: Results from a Randomised Controlled Trial .” Journal of Development Effectiveness , 5 , 178 – 207 . Google Scholar CrossRef Search ADS Bensch Gunther , Grimm Michael , Peters Jörg ( 2015 ). “ Why Do Households Forego High Returns from Technology Adoption? Evidence from Improved Cooking Stoves in Burkina Faso .” Journal of Economic Behavior & Organization , 116 , 187 – 205 . Google Scholar CrossRef Search ADS Bensch Gunther , Peters Jörg ( 2015 ). “ The Intensive Margin of Technology Adoption—Experimental Evidence on Improved Cooking Stoves in Rural Senegal .” Journal of Health Economics , 42 , 44 – 63 . Google Scholar CrossRef Search ADS PubMed Berry James , Fischer Greg , Guiteras Raymond ( 2015 ). “ Eliciting and Utilizing Willingness to Pay: Evidence from Field Trials in Northern Ghana .” Discussion Paper, Centre for Economic Policy Research , London, UK , www.cepr.org/active/publications/discussion_papers/dp.php?dpno=10703 . Bonjour Sophie , Adair-Rohani Heather , Wolf Jennyfer , Bruce Nigel G. , Sumi Mehta , Prüss-Ustün Annette , Lahiff Maureen , Rehfuess Eva A. , Mishra Vinod , Smith Kirk R. ( 2013 ). “ Solid Fuel Use for Household Cooking: Country and Regional Estimates for 1980–2010 .” Environmental Health Perspectives , 121 , 784 – 790 . Google Scholar CrossRef Search ADS PubMed Burwen Jason , Levine David I. ( 2012 ). “ A Rapid Assessment Randomized-Controlled Trial of Improved Cookstoves in Rural Ghana .” Energy for Sustainable Development , 16 , 328 – 338 . Google Scholar CrossRef Search ADS Center for Entrepreneurship in International Health and Development ( 2008 ). Cook Stove Projects and Carbon Finance , Center for Entrepreneurship in International Health and Development , San Francisco, CA . Cialdini Robert B. ( 2007 ). Influence: The Psychology of Persuasion , Collins , New York . Clemens John , Jodar Luis ( 2005 ). “ Introducing New Vaccines into Developing Countries: Obstacles, Opportunities and Complexities .” Nature Medicine , 11 , S12 – S15 . Google Scholar CrossRef Search ADS PubMed Cohen Jessica , Dupas Pascaline ( 2010 ). “ Free Distribution or Cost-Sharing? Evidence from a Randomized Malaria Prevention Experiment .” Quarterly Journal of Economics , 125 , 1 – 45 . Google Scholar CrossRef Search ADS Cropper Maureen L. , Haile Mitiku , Lampietti Julian , Poulos Christine , Whittington Dale ( 2004 ). “ The Demand for a Malaria Vaccine: Evidence from Ethiopia .” Journal of Development Economics , 75 , 303 – 318 . Google Scholar CrossRef Search ADS Davis Scott , Gerstner Eitan , Hagerty Michael ( 1995 ). “ Money Back Guarantees in Retailing: Matching Products to Consumer Tastes .” Journal of Retailing , 71 , 7 – 22 . Google Scholar CrossRef Search ADS Devoto Florencia , Duflo Esther , Dupas Pascaline , Parienté William , Pons Vincent ( 2012 ). “ Happiness on Tap: Piped Water Adoption in Urban Morocco .” American Economic Journal: Economic Policy , 4 , 68 – 99 . Google Scholar CrossRef Search ADS Dewey Kathryn G , Adu Seth , Afarwuah ( 2008 ). “ Systematic Review of the Efficacy and Effectiveness of Complementary Feeding Interventions in Developing Countries .” Maternal & Child Nutrition , 4 , 24 – 85 . Google Scholar CrossRef Search ADS PubMed Dupas Pascaline ( 2009 ). “ What Matters (and What Does Not) in Households’ Decision to Invest in Malaria Prevention? ” American Economic Review , 99 ( 2 ), 224 – 230 . Google Scholar CrossRef Search ADS Dupas Pascaline ( 2011 ). “ Health Behavior in Developing Countries .” Annual Review of Economics , 3 , 425 – 449 . Google Scholar CrossRef Search ADS Fink Günther , Masiye Felix ( 2012 ). “ Assessing the Impact of Scaling-up Bednet Coverage through Agricultural Loan Programmes: Evidence from a Cluster Randomised Controlled Trial in Katete, Zambia .” Transactions of the Royal Society of Tropical Medicine and Hygiene , 106 , 660 – 667 . Google Scholar CrossRef Search ADS PubMed Fischer Greg , Karlan Dean , McConnell Margaret , Raffler Pia ( 2014 ). To Charge or Not to Charge: Evidence from a Health Products Experiment in Uganda , National Bureau of Economic Research , Cambridge, MA . Gertler Paul , Shah Manisha , Alzua Maria Laura , Cameron Lisa , Martinez Sebastian , Patil Sumeet ( 2015 ). How Does Health Promotion Work? Evidence from the Dirty Business of Eliminating Open Defecation , National Bureau of Economic Research , Cambridge, MA . Grossman Sanford ( 1981 ). “ The Informational Role of Warranties and Private Disclosure about Product Quality .” Journal of Law and Economics , 24 , 461 – 483 . Google Scholar CrossRef Search ADS Guiteras Raymond , Levine David I. , Polley Thomas , Quistorff Brian ( 2016 ). Credit Constraints, Discounting and Investment in Health: Evidence from Micropayments for Clean Water in Dhaka , University of Maryland , College Park, MD . Guiteras Raymond , Levinsohn James , Mobarak Ahmed Mushfiq ( 2015 ). “ Encouraging Sanitation Investment in the Developing World: A Cluster-Randomized Trial .” Science , 348 , 903 – 906 . Google Scholar CrossRef Search ADS PubMed Halder Amal K , Tronchet Carole , Akhter Shamima , Bhuiya Abbas , Johnston Richard , Luby Stephen P ( 2010 ). “ Observed Hand Cleanliness and Other Measures of Handwashing Behavior in Rural Bangladesh .” BMC Public Health , 10 , 545 . doi:10.1186/1471-2458-10-545 . Google Scholar CrossRef Search ADS PubMed Hanna Rema , Duflo Esther , Greenstone Michael ( 2016 ). “ Up in Smoke: The Influence of Household Behavior on the Long-Run Impact of Improved Cooking Stoves .” American Economic Journal: Economic Policy , 8 , 80 – 114 . Google Scholar CrossRef Search ADS Harrell Stephen , Beltramo Theresa , Blalock Garrick , Kyayesimira Juliet , Levine David I. , Simons Andrew M. ( 2016 ). “ What Is a Meal?: Comparing Methods of Auditing Carbon Offset Compliance for Fuel Efficient Cookstoves .” Ecological Economics , 128 , 8 – 16 . Google Scholar CrossRef Search ADS Jack William , Suri Tavneet ( 2014 ). “ Risk Sharing and Transactions Costs: Evidence from Kenya's Mobile Money Revolution .” American Economic Review , 104 ( 1 ), 183 – 223 . Google Scholar CrossRef Search ADS Kremer Michael , Miguel Edward ( 2007 ). “ The Illusion of Sustainability .” Quarterly Journal of Economics , 122 , 1007 – 1065 . Google Scholar CrossRef Search ADS Kremer Michael , Miguel Edward , Mullainathan Sendhil , Null Clair , Zwane Alix Peterson ( 2009 ). “ Making Water Safe: Price, Persuasion, Peers, Promoters, or Product Design? ” Working paper. Unpublished manuscript, Harvard University , Cambridge, MA . Lim Stephen S. , Vos Theo , Flaxman Abraham D. , Danaei Goodarz , Shibuya Kenji , Adair-Rohani Heather , Amann Markus et al. ( 2012 ). “ A Comparative Risk Assessment of Burden of Disease and Injury Attributable to 67 Risk Factors and Risk Factor Clusters in 21 Regions, 1990–2010: A Systematic Analysis for the Global Burden of Disease Study 2010 .” Lancet , 380 , 2224 – 2260 . Google Scholar CrossRef Search ADS PubMed Luby Stephen P , Halder Amal K , Huda Tarique , Unicomb Leanne , Johnston Richard B ( 2011 ). “ The Effect of Handwashing at Recommended Times with Water Alone and With Soap on Child Diarrhea in Rural Bangladesh: An Observational Study .” PLoS Medicine , 8 , e1001052 . doi:10.1371/journal.pmed.1001052 . Google Scholar CrossRef Search ADS PubMed Luoto Jill , Levine David I ( 2014 ). Mobile Phones, Rent-to-Own Payments & Water Filters: Evidence from Kenya , RAND Labor and Population , Santa Monica, CA . http://www.rand.org/content/dam/rand/pubs/working_papers/WR1000/WR1039/RAND_WR1039.pdf . Luoto Jill , Mahmud Minhaj , Albert Jeff , Luby Stephen , Najnin Nusrat , Unicomb Leanne , Levine David I. ( 2012 ). “ Learning to Dislike Safe Water Products: Results from a Randomized Controlled Trial of the Effects of Direct and Peer Experience on Willingness to Pay .” Environmental Science & Technology , 46 , 6244 – 6251 . Google Scholar CrossRef Search ADS PubMed Madajewicz Malgosia , Pfaff Alexander , van Geen Alexander , Graziano Joseph , Hussein Iftikhar , Momotaj Hasina , Sylvi Roksana , Ahsan Habibul ( 2007 ). “ Can Information Alone Change Behavior? Response to Arsenic Contamination of Groundwater in Bangladesh .” Journal of Development Economics , 84 , 731 – 754 . Google Scholar CrossRef Search ADS Meredith Jennifer , Robinson Jonathan , Walker Sarah , Wydick Bruce ( 2013 ). “ Keeping the Doctor Away: Experimental Evidence on Investment in Preventative Health Products .” Journal of Development Economics , 105 , 196 – 210 . Google Scholar CrossRef Search ADS Miller Grant , Mobarak A. Mushfiq ( 2015 ). “ Learning About New Technologies Through Social Networks: Experimental Evidence on Nontraditional Stoves in Bangladesh .” Marketing Science , 34 , 480 – 499 . Google Scholar CrossRef Search ADS Mobarak Ahmed Mushfiq , Dwivedi Puneet , Bailis Robert , Hildemann Lynn , Miller Grant ( 2012 ). “ Low Demand for Nontraditional Cookstove Technologies .” Proceedings of the National Academy of Sciences of the United States of America , 109 , 10815 – 10820 . Google Scholar CrossRef Search ADS PubMed Moorthy Sridhar , Srinivasan Kannan ( 1995 ). “ Signaling Quality with a Money-Back Guarantee: The Role of Transaction Costs .” Marketing Science , 14 , 442 – 466 . Google Scholar CrossRef Search ADS Partnership for Clean Indoor Air ( 2011 ). “Uganda Stove Manufacturers Ltd (Ugastove)—Global Leadership Award.” In 5th Biennial Partnership for Clean Indoor Air Forum , Lima, Peru . http://www.pciaonline.org/2011ForumAwards/uganda-stove-manufacturers-ltd-ugastove . Ramanathan Veerabhadran , Carmichael Gregory ( 2008 ). “ Global and Regional Climate Changes due to Black Carbon .” Nature Geoscience , 1 , 221 – 227 . Google Scholar CrossRef Search ADS Shieh Shiou ( 1996 ). “ Price and Money-Back Guarantees as Signals of Product Quality .” Journal of Economics & Management Strategy , 5 , 361 – 377 . Google Scholar CrossRef Search ADS Simons Andrew M. , Beltramo Theresa , Blalock Garrick , Levine David I. ( 2014 ). “ Comparing Methods for Signal Analysis of Temperature Readings from Stove Use Monitors .” Biomass and Bioenergy , 70 , 476 – 488 . Google Scholar CrossRef Search ADS Simons Andrew M. , Beltramo Theresa , Blalock Garrick , Levine David I. ( 2017 ). “ Using Unobtrusive Sensors to Measure and Minimize Hawthorne Effects: Evidence from Cookstoves .” Journal of Environmental Economics and Management , 86 , 68 – 80 . Google Scholar CrossRef Search ADS Smith Kirk R , Bruce Nigel , Arana Byron , Jenny Alisa , Khalakdina Asheena , McCracken John , Diaz Anaite , Schei Morten , Gove Sandy ( 2006 ). “ Conducting The First Randomized Control Trial of Acute Lower Respiratory Infections and Indoor Air Pollution: Description of Process and Methods .” Epidemiology , 17 , S44 – S44 . Google Scholar CrossRef Search ADS Smith Kirk R. , McCracken John P. , Thompson Lisa , Edwards Rufus , Shields Kyra N. , Canuz Eduardo , Bruce Nigel ( 2010 ). “ Personal Child and Mother Carbon Monoxide Exposures and Kitchen Levels: Methods and Results from a Randomized Trial of Woodfired Chimney Cookstoves in Guatemala (RESPIRE) .” Journal of Exposure Science and Environmental Epidemiology , 20 , 406 – 416 . Google Scholar CrossRef Search ADS PubMed Smith-Sivertsen Tone , Díaz Esperanza , Pope Dan , Lie Rolv T. , Díaz Anaite , McCracken John , Bakke Per , Arana Byron , Smith Kirk R. , Bruce Nigel ( 2009 ). “ Effect of Reducing Indoor Air Pollution on Women's Respiratory Symptoms and Lung Function: The RESPIRE Randomized Trial, Guatemala .” American Journal of Epidemiology , 170 , 211 – 220 . Google Scholar CrossRef Search ADS PubMed Suwelack Thomas , Hogreve Jens , Hoyer Wayne D. ( 2011 ). “ Understanding Money-Back Guarantees: Cognitive, Affective, and Behavioral Outcomes .” Journal of Retailing , 87 , 462 – 478 . Google Scholar CrossRef Search ADS Tarozzi Alessandro , Mahajan Aprajit , Blackburn Brian , Kopf Dan , Krishnan Lakshmi , Yoong Joanne ( 2014 ). “ Micro-Loans, Insecticide-Treated Bednets, and Malaria: Evidence from a Randomized Controlled Trial in Orissa, India .” American Economic Review , 104 ( 7 ), 1909 – 1941 . Google Scholar CrossRef Search ADS Uganda Bureau of Statistics ( 2005 ). “ 2002 Uganda Population and Housing Census: Main Report .” Uganda Bureau of Statistics , Kampala, Uganda . Wang Jin , Tyler Erika , Adkins Edwin , Modi Vijay ( 2009 ). Field Evaluation of Rocket Design Stoves—Fuelwood Use and User Preferences , Columbia University , New York, NY . Yu Fei ( 2011 ). “ Indoor Air Pollution and Children's Health: Net Benefits from Stove and Behavioral Interventions in Rural China .” Environmental and Resource Economics , 50 , 495 – 514 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of European Economic Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the European Economic Association Oxford University Press

What Impedes Efficient Adoption of Products? Evidence from Randomized Sales Offers for Fuel-Efficient Cookstoves in Uganda

Loading next page...
 
/lp/ou_press/what-impedes-efficient-adoption-of-products-evidence-from-randomized-oX72laMZ0D
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of European Economic Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
1542-4766
eISSN
1542-4774
D.O.I.
10.1093/jeea/jvx051
Publisher site
See Article on Publisher Site

Abstract

Abstract Many consumers do not adopt products with health and wellbeing benefits apparently far greater than their costs. A sales offer combining a free trial, time payments, and the option of returning the product can overcome barriers such as liquidity constraints and poor information about benefits and usability. We tested this sales offer (and alternatives) in an experiment with a fuel-efficient charcoal stove in urban Uganda and a fuel-efficient wood stove in rural Uganda. Consistent with the importance of these barriers, this offer dramatically increased uptake—in urban Kampala, from 4% to 46%, and in rural Mbarara, from 5% to 57%. (JEL: I12, O12, O33, Q56) 1. Introduction Despite the potential of new products and technologies to improve health and wellbeing, products ranging from bednets to household water treatments often have surprisingly low adoption.1 One salient example is the approximately 2.8 billion people who cook with solid fuels (wood, charcoal, dung) on traditional cookstoves (Bonjour et al. 2013). The smoke from these stoves kills roughly four million people a year (Lim et al. 2012), and contributes to deforestation (Bailis et al. 2015) and global climate change (Ramanathan and Carmichael 2008). The high cost of buying or gathering fuel means that the use of traditional stoves can also deepen poverty. If markets worked well, even a poor person would pay for a fuel-efficient cookstove if fuel savings quickly exceeded the cost of the stove. To learn about the market imperfections that limit adoption, we offered fuel-efficient cookstoves at local market prices, but experimentally varied the terms of the sales offer. We tested four sales offers: a cash and carry offer (similar to a retail setting with a posted price), a one week free trial (followed by full payment or returning the cookstove) to test the role of imperfect information about fuel savings and usability, time payments (equal installment payments over four weeks) to examine the role of liquidity constraints, and a one week free trial followed by time payments to examine how informational and liquidity constraints interact. We build upon existing studies that examine how the uptake of (or willingness to pay for) beneficial health products depends on (1) very large discounts (e.g., Cohen and Dupas 2010; Dupas 2009; Kremer et al. 2009), (2) moderate discounts (e.g., Meredith et al. 2013; Mobarak et al. 2012), (3) providing additional information on health or economic benefits (e.g., Beltramo et al. 2015b; Madajewicz et al. 2007; Meredith et al. 2013), (4) linkage with microfinance (e.g., Devoto et al. 2012; Fink and Masiye 2012; Tarozzi et al. 2014), and (5) experimental manipulation of social network effects (e.g., Beltramo et al. 2015a; Luoto et al. 2012; Meredith et al. 2013; Miller and Mobarak 2015). Our study is most similar to Mobarak et al. (2012) and Tarozzi et al. (2014). Mobarak et al. (2012) studied the adoption of fuel-efficient cookstoves in rural Bangladesh and found low adoption levels for two different types of fuel-efficient cookstoves at full price (2% and 5%, respectively), and continued low uptake (5% and 12%, respectively) even with a large (50%) price discount. They conclude that there is an inherently low demand for fuel-efficient cookstoves. Tarozzi et al. (2014) studied the adoption of insecticide treated bednets in India and found low uptake when sold requiring an immediate cash payment (11%) but much higher uptake (52%) when linking the product to microfinance. They argue these results indicate liquidity constraints are an important barrier to adoption. Our study is similar to Mobarak et al. (2012) in that we also study fuel-efficient cookstoves; however, instead of offering a discount price, we charged market prices. Our study pushes further in that our design—which includes sales offers that are designed to alleviate informational and liquidity constraints—allows us to test if low rates of stove uptake at full price are due to inherent low demand (as Mobarak et al. (2012) suggest) or something else (such as the informational and liquidity barriers we test). Our study is similar to Tarozzi et al. (2014) in that we examine the role of liquidity constraints; however, instead of linking the sales of our product with an outside microfinance organization we do so by offering time payments directly from the salesperson. We push further than Tarozzi et al. (2014) by also testing microloans combined with a chance to learn about the product (through the free trial plus time payments offer), as well as a sales offer with only the free trial. As Mobarak et al. (2012) found in Bangladesh, in urban Kampala, uptake was low (4%) for fuel-efficient charcoal-burning stoves with the traditional cash-and-carry offers (we gave households a week to raise funds, if needed). Uptake was much higher (46%) with a sales offer that combined a free trial and time payments. In this experiment, we also tested the free trial (29% uptake) and the time payment (26% uptake) offers separately. We repeated two of these sales offers in a rural district (i.e., Mbarara) in southwestern Uganda, with fuel-efficient wood-burning stoves, and found a similarly large uptake with free trial and time payments (57%) as compared with cash-and-carry (5%) offers. Our confidence in the results’ external validity increased with similar findings in different settings. To understand any health or environmental benefits attributable to fuel-efficient cookstoves requires data on stove usage and other outcomes. We analyze these topics in a subsequent paper. Past studies have had mixed results. For example, Bensch and Peters (2015), Smith et al. (2006, 2010), Smith-Sivertsen et al. (2009), and Yu (2011) found a fairly high usage of new stoves and improvements in measures of health or exposure to emissions. Other studies (e.g., Beltramo and Levine 2013; Burwen and Levine 2012; Hanna, Duflo, and Greenstone 2016) failed to find consistent usage of new stoves and did not find sustained health improvements. 2. Research Setting and Design We ran two randomized control trials. We first describe the key differences between the two experiments and then discuss each in detail. The first experiment was in Kampala, the capital of Uganda, and the second in Mbarara, a rural district about 300 km southwest of the capital. The urban study was done at a household's doorstep individually, whereas the rural study gathered participants in large group meetings before making a sales offer. Purchased charcoal was the dominant fuel type in the urban area, whereas gathered firewood was the most popular in the rural setting. The urban sample had four experimental treatments whereas the rural sample had two treatments. See Table 1 for a listing of the key differences between the two experiments. Table 1. Differences between urban and rural studies. Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team View Large Table 1. Differences between urban and rural studies. Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team Urban (Kampala) study Rural (Mbarara) study Fuel used by new stove Charcoal Wood or other biomass Means of acquiring fuel Majority purchase charcoal Majority gather wood Stove brand sold Ugastove Envirofit G-3300 Price of stove $6, $8, or $10 for size 1,2,3 $12 ($16) in 4 (22) parishes Cooks for how many Size 1, cooks for 5–7 people Cooks for 4–7 people Size 2, cooks for 10–12 people Size 3, cooks for 10–12 people Sales method Door-to-door sales by research staff acting as salespeople Parish-wide sales meeting with research staff acting as salespeople Time payment collected by Research staff acting as salespeople Focal point person from parish recruited by research team View Large 2.1. Urban Study Design The first experiment took place in Kampala, Uganda's capital and largest city, from October to December 2010. The majority of Kampala's households cook with a traditional charcoal stove. Most cooking is outdoors, unless it is raining. We marketed the Ugastove charcoal stove, made by a Ugandan-owned company in Kampala (see Figure 1 for images of a traditional charcoal stove and the Ugastove charcoal stove). The main fuel saving innovations of the Ugastove are its cylindrical sheet metal frame and surrounding heat-insulating ceramic insert. These features reduce the amount of charcoal needed as compared to traditional charcoal stoves. The Ugastove was selected based on evidence that this stove reduces fuel use when tested in controlled settings (Wang et al. 2009) and that it met the voluntary carbon market's Gold Standard in kitchen performance tests (Center for Entrepreneurship in International Health and Development 2008). The Ugastove manufacturer receives carbon credits for the stoves and passes these savings on to customers. Because of the carbon credit subsidy, retail prices in our experiment were $6, $8, or $10, depending on the size. These prices matched what Ugastoves sold for in the local Kampala market. Figure 1. View largeDownload slide Comparison of charcoal stoves: traditional versus Ugastove. Figure 1. View largeDownload slide Comparison of charcoal stoves: traditional versus Ugastove. The Centre for Integrated Research and Community Development (CIRCODU), a research organization based in Kampala that specializes in market research on household energy,2 served as the in-country sales team and data collection partner. For the urban study, we employed a door-to-door sales strategy with enumerators acting as salespeople. Enumerators’ salaries were kept independent from cookstove sales levels to remove incentives for enumerators to target wealthier households. To select households, we began with a list of all Kampala parishes provided by the Ugandan Bureau of Statistics (Kampala has 96 urban parishes with an average population of 12,387 each, Uganda Bureau of Statistics 2005). CIRCODU’s staff categorized about 30% of parishes as “low income”, 60% as “middle income”, and 10% as “wealthy”, based on their personal knowledge. We covered virtually all parishes that were categorized as “low income” or “middle income”. We excluded the “wealthy” parishes, as their residents cook mainly with gas stoves and they would not be interested in charcoal cookstoves. We divided parishes into two to three different “neighborhoods”, ending up with 226 different neighborhoods in our sample. We randomized offers at a sub-parish/neighborhood level to avoid the possibility of upsetting customers who heard that their neighbors received a different sales offer.3 Thus, each household within a neighborhood received the same sales offer. To achieve balance between our study arms, we stratified based on CIRDOCU’s socioeconomic categorizations of the parishes (“low/middle income”). In addition, to reduce the chance of enumerator bias, we developed a work schedule that ensured each enumerator presented each offer approximately the same number of times, and their schedules had a similar breakdown in times and weekdays. In each neighborhood, sales teams of two enumerators marketed the stoves to 10 households in which an adult was home. Enumerators gave presentations to approximately every sixth household, identified by a fixed way finding procedure to ensure randomization.4 Thus, a consumer was unlikely to be approached after having seen or heard enumerators visit their neighbors. To reduce socioeconomic similarity within neighborhoods, after five households, enumerators returned to their vehicle, drove approximately 1 km, and then made the sales offer at five more households (following the same procedure) in the same neighborhood. At each home, the sales team gave a marketing presentation about the stoves and presented the sales offer randomly assigned to that neighborhood. After gathering some basic information pertaining to the household's cooking and fuel purchasing behaviors, enumerators recorded the homeowners’ purchase decisions. 2.2. Urban Sales Offers We tested four sales offers in Kampala. Cash-and-Carry Offer. Consumers were given the opportunity to purchase a stove at a take-it-or-leave-it price. Consumers were told that, if they needed to gather funds or check with family members, they could take a week to decide and pay. In such cases, enumerators returned one week later to collect payment and deliver the stove. Free Trial Offer. If consumers were willing to accept a free trial, enumerators left the stove and returned in one week for the full payment. If consumers did not want to keep the stove, they could return the stove at the end of the week, with no obligation. Time Payments Offer. Consumers were given the opportunity to purchase a stove with four equal weekly installments. This offer included the right to return the stove before all scheduled payments were due, in which case, future payments were canceled but no money refunded (similar to the rent-to-own model in the United States). As with the cash-and-carry offer, enumerators offered to return in a week to deliver the stove and collect the first time payment if consumers wanted to discuss their choice with family members or needed one week to gather the first payment. Free Trial and Time Payments Offer. Consumers were offered a one-week free trial followed by the opportunity to purchase a stove through four equal weekly installment payments. They also received the right to return the stoves before all scheduled payments were due, cancelling future payments, as in the time payments offer. 2.3. Urban Pricing and Sales Visits The stove prices for all sales offers were the standard market prices of $6, $8, or $10, depending on the size. The most popular stove was the $6 version, as it could cook for an average-sized family (i.e., 5–7 people). The most expensive model could cook for 10–12 people. To those offered the cash-and-carry sales terms, enumerators offered the posted prices ($6, $8, or $10, based on stove sizes desired by consumers) or an incentive-compatible Becker–DeGroot–Marschak (BDM) procedure (Becker, DeGroot, and Marschak 1964). In the BDM procedure, enumerators showed each participant a sealed envelope and explained that the stove price was set randomly by their manager and hidden within the envelope. Enumerators asked participants what was the highest price they would agree to pay, explaining that they could purchase the stove at the envelope price if the price they stated was at least as large as the unknown price inside the envelope. If respondents stated a willingness to pay below the envelope price, they would not be able to purchase the stove. Because a stated willingness to pay affects whether someone can purchase a product, but not how much she pays, this procedure provides incentives for respondents to report their willingness to pay truthfully (if participants understand and believe all the instructions). That is, it is not in the best interests of respondents to name a higher price than what the product is worth to them because they may end up agreeing to pay more than they are actually willing to pay. Similarly, if participants understate their true willingness to pay, they might lose the opportunity to buy a stove at the price they were willing to pay. Enumerators followed the BDM procedure to recover a demand curve rather than simply the share of households that accepted a cash-and-carry offer at the stoves’ market prices. Enumerators allowed participants to ask questions prior to participating in the BDM procedure, to ensure they understood. Enumerators also explained to participants that they would have up to seven days to gather funds if needed (exactly the same as the participants offered the posted market price). After consumers made their decisions, enumerators thanked them for their time and offered a small gift (i.e., a bar of soap) in exchange for answering a few more questions. Over a space of weeks, enumerators recorded customers’ payments, return rates, and default rates. The process for the consumer to return a stove was designed to be as simple for the consumer as possible. The consumer could return a stove to an enumerator during a follow-up visit and the enumerator would carry the stove away and no longer visit the consumer. 2.4. Rural Study Design The follow-up experiment took place in 26 rural parishes of the Mbarara district in western Uganda, from March to May 2012. We selected the Mbarara region because it is rural and almost all families cook with wood on traditional three-stone fires. In addition, the district is less than a day from Kampala, families spend considerable time gathering wood, and local leaders were supportive of our project. In contrast to the urban setting, most families cook in a separate cooking hut. See Harrell et al. (2016), Simons et al. (2014, 2017) for additional background of the rural study area. CIRCODU, the same organization that collected our data in the urban study also carried out the rural study. We marketed the Envirofit G3300 wood-burning stove, made by Envirofit International Inc. (see Figure 2 for images of a traditional three-stone fire and the Envirofit G3300). This stove achieves relatively efficient fuel combustion by sending airflow into the fire and directing heat upward to the cooking surface. These design innovations allow fuel to burn at a controlled rate and enable more complete combustion than a three-stone fire. Emissions testing of the Envirofit G3300 in a controlled laboratory setting found average reductions in CO by 65%, particulate matter reductions of 51%, and a reduction in fuel wood use by 50% compared to a three stone fire (see Figure 3 for a copy of the emissions and performance report). Figure 2. View largeDownload slide Comparison of wood burning stoves: three stone fire versus Envirofit G-3300. Figure 2. View largeDownload slide Comparison of wood burning stoves: three stone fire versus Envirofit G-3300. Figure 3. View largeDownload slide Certified Emissions and Performance Report for Envirofit G3300. Figure 3. View largeDownload slide Certified Emissions and Performance Report for Envirofit G3300. 2.5. Rural Sales Offers We randomly selected 12 parishes to receive the traditional cash-and-carry sales offer and 14 parishes to receive the free trial with time payments offer.5 The terms and conditions of both offers were the same as in the urban experiment (see Section 2.2). Within each parish, we recruited a local point person with the help of local government officials. We asked each point person to gather roughly 60 people together for a public sales meeting on a specified day. We told each point person that we would demonstrate a fuel-efficient cookstove and offer it for sale. We did not tell the point person which sales offer his or her parish would receive. At the sales meeting, the study team presented the Envirofit G3300, discussed its benefits relative to traditional three-stone fires, did a cooking demonstration, and presented the terms of the randomly selected sales offer. Then participants completed a questionnaire focused on household cooking and basic socioeconomic indicators. 2.6. Rural Pricing and Sales Visits We purchased stoves from UpEnergy,6 a company that distributed Envirofit stoves in several regions of Uganda but at the time, not in our study area. We offered the stoves for sale at $12 each in four parishes and then, because of high sales, increased the price to $16 in the remaining 22 parishes.7 The $16 price was similar to what UpEnergy's retail partners sold the same Envirofit stoves for in other regions of Uganda. As with the Ugastove, this retail price was reduced in part due to a carbon credit agreement. As in the urban study, participants given the cash-and-carry offer had the option of purchasing the stove that day for cash, or, if they desired additional time to gather funds and/or discuss their choice with family, they could return to purchase a stove at a second meeting about 10 days later. Participants offered the free trial with time payments could sign up for a free trial at the end of the group sales meeting. They could then return their stoves to the village focal point person with no obligation at the end of the free trial.8 If they liked the stove, they could purchase it in four equal weekly payments. Due to scheduling and logistical constraints, households that accepted a free trial generally received their stove about three weeks after the sales meeting. 3. Barriers to Adoption We designed our study to examine market failures that could reduce the purchase of fuel-efficient cookstoves in developing countries, including liquidity or credit constraints, and imperfect information about the quality, usability, and fuel savings of the new appliance. Throughout this section we present a brief theoretical framework that describes some concrete hypothesis about the barriers to demand the different treatment arms were designed to address. 3.1. Imperfect Information about Product Effectiveness and Energy Savings The free trial is designed to remove informational barriers about the fuel savings and usability of the cookstove. A free trial gives consumers an opportunity to try out a stove and decide if it saves fuel and fits the household's cooking needs. Our free trial period was similar (prior to the point money was paid) to a money-back guarantee, which is documented to increase consumers’ willingness to try unfamiliar products when they are unsure of benefits, especially for experience goods9 (Davis, Gerstner, and Hagerty 1995; Grossman 1981; Suwelack, Hogreve, and Hoyer 2011). Although our experiment was designed with this learning channel (about fuel savings and usability) in mind, it is also possible that the free trial could be encouraging additional sales through other channels. The free trial could be a credible signal that the stove will actually save fuel and perform adequately (Shieh 1996; Moorthy and Srinivasan 1995), so rather than influencing sales via a learning channel the free trial is actually serving as a credibility signal. Another possibility is that a trial period may activate norms of reciprocity, which could increase uptake and repayment (Cialdini 2007). We examine disaggregated data to examine whether the sales increases are more likely attributed to the learning channel, or if the free trial activated a credibility signal or some norm of reciprocity. We gathered data on whether households had ever seen or used a fuel-efficient stove, and whether they were aware that fuel-efficient stoves used less fuel. We expect the value of the free trial (and associated learning about fuel savings) to be different between participants that already knew that these stoves saved fuel and those that did not. We hypothesize that households with no knowledge of fuel-efficient stoves would purchase at higher rates in the experimental arms that offered a free trial than households that were already aware of the fuel savings of these stoves. 3.2. Liquidity or Credit Constraints Liquidity constraints are another plausible reason for the low adoption of beneficial products in developing countries (Bensch, Grimm, and Peters 2015; Devoto et al. 2012; Dupas 2011; Mobarak et al. 2012; Tarozzi et al. 2014). In a related study in different Ugandan rural parishes, we documented that willingness to pay—in a within-subject comparison—was on average about 40% higher with time payments (four equal payments over four weeks) than when paying for a new cookstove within a week (Beltramo et al. 2015b). To examine liquidity constraints in our urban sample we examine if stove repayment rates are faster for individuals who have higher weekly charcoal expenditures than those with lower weekly charcoal expenditures. If repayment rates are faster for those with more charcoal expenditures this suggests that liquidity is central (although, given the endogeneity of the choice of uptake, not conclusive). Additionally we examine if any of the treatments induce increases in the size of the stove purchased (recall we offered three sizes that cost $6, $8, and $10 in the urban setting). Although not definitive, we could imagine that treatment arms that reduced liquidity constraints could be associated with the purchase of more expensive stoves (either directly via the installments, or indirectly via the savings of fuel costs). We also gathered data on whether rural households purchase or gather fuel. If stove repayment rates were faster for individuals who purchase fuel it would suggest that liquidity is central (though, as pointed out previously, not necessarily conclusive). If the differences in repayment rates are small, other forces are likely to be more prominent. See Table 2 for a summary of the potential mechanisms, the predictions, and how we test for the mechanism. Table 2. Tests for possible mechanisms. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. View Large Table 2. Tests for possible mechanisms. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. Potential mechanism Prediction of mechanism How we test for mechanism Imperfect information (1) Many customers are not confident in vendors' claims of fuel savings, a free trial allows customers a low-cost check on fuel savings and increases uptake. Compare the differences in purchase rates between treatment arms with a free trial versus treatment arms without free trials (i.e., cash-and-carry vs. free trial; time payments vs. free trial and time payments). Imperfect information (2) Learning about fuel savings during free trial results in higher purchase rates from those who did not already know that fuel-efficient stoves were designed to save fuel. Compare purchase rates between households that knew fuel-efficient cookstoves were designed to save fuel versus those that did not know. Examine differences in purchase rates between these two groups in treatment arms with a free trial versus treatment arms without free trials. Liquidity constraints (1) Liquidity constraints impede a household's ability to pay a lump sum for a stove. The time payments offer removes this constraint and increases uptake. Compare the differences in purchase rates between treatment arms with time payments versus treatment arms without time payments (i.e., cash-and-carry vs. time payments; free trial vs. free trial and time payments). Liquidity constraints (2) Households spending the most cash on weekly charcoal purchases have a greater ability to repay time payments early. Compare early repayment rates between households in the upper half versus households in the lower half of the distribution of weekly charcoal expenditures. Liquidity constraints (3) Liquidity constraint binds most tightly for the most costly stove model ($10 for the large model) than for the less costly models ($6 for small, $8 for mid-sized). Time payments have the largest effect on the most costly ($10) model of stove, and the smallest effect on the least costly ($6) model. Liquidity constraints (4) Households already spending cash on fuelwood (instead of gathering it) have a greater ability to repay time payments early. Examine if early repayment rates are faster for households that purchase wood compared to households that gather wood. View Large 3.3. Potential Weaknesses of the Free Trial and Time Payments Offer The offer of a free trial followed by time payments has several potential weaknesses. Return rates with this offer will be high if stoves break frequently, especially if consumers are careless with them during the free trial. The free trial will not increase demand if new stoves are a poor fit for the region, for example, if they require cooking patterns so far from traditional practices that the difficulty for cooks to adapt to the new technology outweighs its other benefits. If consumers are frequently not at home, collecting payments becomes costly. Furthermore, both the free trial and time payments offers open the stove provider to consumer moral hazards, such as when consumers move frequently or decide to keep stoves and not pay for them or damage them during the free trial. Low payments are particularly likely if adverse selection exists among consumers, in which those consumers least likely to pay for a stove are more likely to accept free trial and time payments offers. 4. Results 4.1. Verifying Randomization Table 3 presents household summary statistics for the urban Kampala study, for each of the four sales offers. Standard errors were adjusted for clustering at the neighborhood level. The households in each group are similar, with none of the differences across a row being statistically different from zero. Table 3. Household characteristics urban (Kampala) study: summary statistics. Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Notes: Household data collected at initial household sales pitch. We adjust standard errors for clustering at the neighborhood level. Values presented are rounded to two decimal places. View Large Table 3. Household characteristics urban (Kampala) study: summary statistics. Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Cash and carry Free trial offer Time payments Free trial and time payments Mean SD Mean SD Mean SD Mean SD Row equality p-value N Household characteristics Female respondent (share) 0.86 (0.35) 0.88 (0.33) 0.87 (0.34) 0.88 (0.32) 0.778 1863 Number at largest daily meal 4.98 (2.71) 4.96 (2.56) 5.33 (2.76) 4.86 (2.67) 0.205 1740 Charcoal is primary fuel source (share) 0.94 (0.25) 0.95 (0.21) 0.95 (0.22) 0.94 (0.24) 0.823 1863 Charcoal expenditures (USD per week) 2.52 (1.54) 2.53 (1.82) 2.72 (1.92) 2.76 (1.77) 0.253 1724 Stove usage and knowledge Use more than one stove weekly (share) 0.71 (0.46) 0.64 (0.48) 0.68 (0.47) 0.71 (0.45) 0.343 1815 Use clay stove weekly (share) 0.89 (0.32) 0.85 (0.36) 0.86 (0.35) 0.82 (0.38) 0.135 1815 Use basic metal stove weekly (share) 0.33 (0.47) 0.34 (0.48) 0.35 (0.48) 0.37 (0.48) 0.891 1815 Use three stone fire weekly (share) 0.02 (0.12) 0.01 (0.10) 0.00 (0.07) 0.01 (0.09) 0.392 1863 Owns electric, gas or kerosene stove (share) 0.02 (0.14) 0.04 (0.19) 0.04 (0.19) 0.03 (0.16) 0.547 1815 Already owns fuel-efficient stove (share) 0.07 (0.25) 0.09 (0.29) 0.09 (0.29) 0.08 (0.27) 0.393 1796 Have seen a fuel-efficient stove before (share) [among HHs that do not already own one] 0.51 (0.50) 0.57 (0.50) 0.51 (0.50) 0.50 (0.50) 0.495 1639 Aware fuel-efficient stoves use less fuel (share) [among HHs that do not own, but seen one] 0.55 (0.50) 0.48 (0.50) 0.56 (0.50) 0.55 (0.50) 0.569 848 Number of households receiving offer 579 539 390 355 1863 Notes: Household data collected at initial household sales pitch. We adjust standard errors for clustering at the neighborhood level. Values presented are rounded to two decimal places. View Large In the urban sample, about 87% of respondents were women. Households cooked for about five people at their daily largest meal, of which 95% used charcoal with an average expenditure around $2.60 per week. Table 4 presents the household summary statistics for the rural Mbarara study. The groups are similar. Out of the 20 characteristics shown, only one difference is significant at the 5% level, and three differences are significant at the 10% level, about that to be expected by chance. Although this is what is to be expected by chance these differences are on key variables related to household earnings and how household's secure firewood. More households earn income (93% vs. 89%, difference p = 0.06) and have year-round employment (59% vs. 51%, difference p = 0.09) in the cash and carry offer than in the free trial with time payments offer. These slight differences would be more worrisome if households that were better off (i.e., those that earn income and have year-round employment) received more access to the free trial and time payments offer; however, they did not. A larger percentage of the cash and carry households purchased firewood last month (36% vs. 28%, difference at p = 0.11), whereas a smaller percentage gathered firewood last month (73% vs. 83%, difference p = 0.05). Given that these differences move in opposite directions for households that received the free trial and time payments offer, any possible bias that might be associated with differential practices in acquiring firewood relating to a specific sales offer is minimal. Table 4. Household characteristics rural (Mbarara) study: summary statistics. Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Notes: Household data collected at the parish wide sales meeting. We adjust standard errors for clustering at the parish level. To minimize the effect of outliers the value of assets is bottom and top coded at 2% and 98% of the distribution, respectively. The prices used to calculate asset values are taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Survey (LSMS) published by the World Bank (see Table A.4 in the Appendix). Values presented are rounded to two decimal places, the value in the difference column is calculated prior to rounding. View Large Table 4. Household characteristics rural (Mbarara) study: summary statistics. Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Cash and carry Free trial and time payments Mean SD Mean SD Difference p-value N Household demographics Female respondent (share) 0.55 0.50 0.65 0.48 0.10 0.053 1404 Age of respondent 39.12 13.65 39.62 13.65 0.50 0.643 1397 Married (share) 0.76 0.43 0.78 0.41 0.02 0.386 1404 Wife is primary cook (share) 0.85 0.35 0.88 0.32 0.03 0.245 1404 Spouses make decisions jointly (share) 0.52 0.50 0.52 0.50 −0.00 0.976 1404 Socioeconomic status Earns income (share) 0.93 0.26 0.89 0.31 −0.04 0.057 1398 Self-employed (share) 0.68 0.47 0.66 0.47 −0.02 0.467 1404 Year round employment (share) 0.59 0.49 0.51 0.50 −0.08 0.092 1404 Identify as subsistence farmers (share) 0.85 0.36 0.87 0.33 0.02 0.236 1404 Value of assets (USD) 697.39 1131.58 833.94 1294.64 136.55 0.315 1404 Stove use and fuels Three stone fire is primary stove (share) 0.67 0.47 0.70 0.46 0.03 0.610 1404 Number at largest daily meal 5.56 2.46 5.70 2.43 0.14 0.618 1388 Always boils drinking water (share) 0.66 0.47 0.71 0.45 0.06 0.402 1404 Firewood primary fuel source (share) 0.85 0.36 0.90 0.29 0.06 0.203 1404 Purchased firewood last month (share) 0.36 0.48 0.28 0.45 −0.09 0.105 1396 Gathered firewood last month (share) 0.73 0.44 0.83 0.37 0.10 0.048 1397 Openness to new technologies or products Household in savings group (share) 0.77 0.42 0.80 0.40 0.03 0.415 1398 Household uses improved seeds (share) 0.18 0.38 0.18 0.38 0.00 0.944 1398 Household uses fertilizer (share) 0.06 0.23 0.05 0.22 −0.01 0.597 1398 Household owns solar lamp (share) 0.12 0.32 0.14 0.34 0.02 0.484 1398 Number of households receiving offer 538 866 1404 Notes: Household data collected at the parish wide sales meeting. We adjust standard errors for clustering at the parish level. To minimize the effect of outliers the value of assets is bottom and top coded at 2% and 98% of the distribution, respectively. The prices used to calculate asset values are taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Survey (LSMS) published by the World Bank (see Table A.4 in the Appendix). Values presented are rounded to two decimal places, the value in the difference column is calculated prior to rounding. View Large In the rural sample, about 60% of those attending the village sales meeting were women. The largest daily meal was cooked for an average of 5.6 people. About 87% of households used firewood as the primary fuel source, of which approximately 32% had purchased firewood in the last month and 78% had gathered wood in the last month. Slightly less than 80% of households belonged to a savings group, and few had previous experience with other new technologies, such as improved seeds (18%), solar lamps (13%), or fertilizer (6%). 4.2. Urban Sales Offer Results Table 5 presents the results of the urban study in term of purchases, returns, and defaults across the four sales offers. Cash-and-Carry Offer. Only 4% of households given the cash-and-carry offer agreed to purchase the stove at prevailing local retail prices (23 of 579). As mentioned earlier, we offered a stated fixed price or followed a Becker–DeGroot–Marschak price elicitation procedure for the cash-and-carry offer in the urban setting. Of the subsample offered the fixed retail price, 6% accepted (7 of 114), whereas 3% of those in the BDM procedure stated they were willing to pay at least the stove's retail price (16 of 465). The difference in take-up rate across the two procedures is not statistically significant (p-value = 0.21). At the same time, the slightly lower willingness to pay reported with the BDM procedure than for the cash-and-carry offer is consistent with the notion that consumers could have shaved their stated willingness to pay relative to their actual willingness to pay (as in Beltramo et al. 2015b; Berry, Fischer, and Guiteras 2015). Free Trial Offer. Among households that received a free trial offer, 33% accepted (178 of 539). Among those that accepted, 12% returned the stove after the trial (21 of 178), 11% defaulted (19 of 178), and 78% paid in full (138 of 178). Of those in default, 42% paid at least something (8 of 19).10 In total, uptake (minus returns) was 29% (157 of 539), and we collected 90% of the revenue that consumers owed. Table 5. Purchase and payment by sales offer in urban (Kampala) study. Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Notes: Sales offers made at randomly selected urban households. We adjust standard errors for clustering at the neighborhood level. Test of row equality reports the p-value for of the chi-squared statistic testing if uptake across the four offers is equal. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. The one case of a free trial recipient returning the stove after payments began was a person who wanted to pay a small portion of the purchase price when the stove was delivered, but then returned the stove after the free trial. Ten recipients of the free trial offer requested to pay at the time the stove was delivered (in essence not utilizing the free trial), these are marked as paid off early in the free trial offer column. Nine recipients of the free trial offer did not pay the full purchase price at the end of the free trial, but did pay in subsequent weeks. Pairwise comparisons of treatment arms: a. Statistically significant difference (p < 0.001) from treatment in column (1); b. Statistically significant difference (p < 0.001) from treatment in column (2); c. Statistically significant difference (p < 0.001) from treatment in column (3); d. Statistically significant difference (p < 0.001) from treatment in column (4); e. Statistically significant difference (p < 0.10) from treatment in column (2); f. Statistically significant difference (p < 0.10) from treatment in column (3). View Large Table 5. Purchase and payment by sales offer in urban (Kampala) study. Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Cash and carry (1) Free trial offer (2) Time payments (3) Free trial and time payments (4) Test of row equality N % N % N % N % p-value Number of offers made 579 539 390 355 Purchased or accepted free trial 23 4.0b,c,d 178 33.0a,d,f 102 26.2a,d,e 171 48.2a,b,c 0.000 Among those that accepted free trial  Returned after free trial 20 11.2 – – 7 4.1  Returned after payments began 1 0.6 1 1.0 2 1.2  Paid in full 138 77.5 95 93.1 151 88.3  In default 19 10.7 6 5.9 11 6.4 Among those that paid in full  Paid off early 10 7.2 32 33.7 47 31.1  Paid off late 9 6.5 10 10.5 15 9.9 Among those in default  Paid more than zero 8 42.1 5 83.3 11 100.0  Paid at least half price 6 31.6 3 50.0 9 81.8 Uptake (less returns) 23 4.0b,c,d 157 29.1a,d 101 25.9a,d 162 45.6a,b,c 0.000 Revenue as share of owed 100.0 90.2 96.2 96.9 Uptake (less returns and defaults) 23 4.0b,c,d 138 25.6a,d 95 24.4a,d 151 42.5a,b,c 0.000 Revenue as share of owed 100.0 100.0 100.0 100.0 Notes: Sales offers made at randomly selected urban households. We adjust standard errors for clustering at the neighborhood level. Test of row equality reports the p-value for of the chi-squared statistic testing if uptake across the four offers is equal. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. The one case of a free trial recipient returning the stove after payments began was a person who wanted to pay a small portion of the purchase price when the stove was delivered, but then returned the stove after the free trial. Ten recipients of the free trial offer requested to pay at the time the stove was delivered (in essence not utilizing the free trial), these are marked as paid off early in the free trial offer column. Nine recipients of the free trial offer did not pay the full purchase price at the end of the free trial, but did pay in subsequent weeks. Pairwise comparisons of treatment arms: a. Statistically significant difference (p < 0.001) from treatment in column (1); b. Statistically significant difference (p < 0.001) from treatment in column (2); c. Statistically significant difference (p < 0.001) from treatment in column (3); d. Statistically significant difference (p < 0.001) from treatment in column (4); e. Statistically significant difference (p < 0.10) from treatment in column (2); f. Statistically significant difference (p < 0.10) from treatment in column (3). View Large Time Payments Offer. Among households that received the offer of four equal weekly time payments, 26% accepted the offer (102 of 390). Of those that accepted the offer, only 1% (1 of 102) returned the stove after payments began, 6% defaulted (6 of 102) and 93% paid in full (95 of 102). Among those that paid in full, a third paid the stove off early (32 of 95), whereas 11% paid off late (10 of 95). In addition, 83% (5 of 6) defaults paid at least something. In total, uptake (minus returns) was 26% (101 of 390), and we collected 96% of the revenue that consumers owed. Free Trial and Time Payments. Of households that received the free trial and four equal weekly time payments offer, even more accepted this offer, with 48% accepting the initial free trial (171 of 355). Among those that accepted the free trial, 5% returned the stove after the trial (9 of 171), 6% defaulted (11 of 171), and 88% paid in full (151 of 171). Of those that paid in full, 31% paid off early (47 of 151), whereas 10% paid off late (15 of 151). Among those that ended up defaulting, 100% paid at least something (11 of 11). In total, uptake (minus returns) was 46% (162 of 355), with 97% of the revenue that consumers owed being collected. 4.3. Rural Sales Offer Results Table 6 presents the results of the rural study. Table 6. Purchase and payment by sales offer in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Notes: Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table 6. Purchase and payment by sales offer in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 538 866 Purchased or accepted free trial 25 4.6 538 62.1 0.000 Among those that accepted free trial  Returned after free trial 3 0.6  Returned after payments began 41 7.6  Paid in full 489 90.9  In default 5 0.9 Among those that paid in full  Paid off early 108 221  Paid off late 30 6.1 Among those in default  Paid more than zero 2 40.0  Paid at least half price 1 20.0 Uptake (less returns) 25 4.6 494 57.0 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 25 4.6 489 56.5 0.000 Revenue as share of owed 100.0 100.0 Notes: Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Cash-and-Carry Offer. Only 5% of households given the cash-and-carry offer purchased the Envirofit at prevailing local retail prices (25 of 538). Free Trial and Time Payments Offer. Of households that received the free trial and four equal weekly time payments offer, 62% accepted the initial free trial (538 of 866). Among those that accepted the free trial, 8% returned the stove after the trial (44 of 538), 1% defaulted (5 of 538), and 91% paid in full (489 of 538). Of those that paid in full, 22% paid off early (108 of 489), whereas 6% paid off late (30 of 489). Among those that ended in default, 40% paid at least something (2 of 5). In total, uptake (minus returns) was 57% (494 of 538), with 99% of the revenue that consumers owed being collected. As previously noted, in four of the 26 rural parishes (two parishes for each sales offer), the sales price was $12 rather than $16. Removing the results of these four parishes does not materially change results (see Table A.1 in the Appendix). 4.4. Evidence of Imperfect Information as Binding Constraint Our arguments are consistent with market failures—rather than poor stove design—as a major reason for the observed low demand for fuel-efficient cookstoves in developing countries (Bensch et al. 2015; Mobarak et al. 2012). We offered a free trial to combat the potential market failure of consumers having imperfect information. The missing information could concern the stove's general usability, the stove's performance in cooking common meals, the quantity of fuel it saves, and so forth. The quite large increase in uptake, from 4.0% with a cash-and-carry offer to 29.1% with the free trial offer, suggests the free trial overcomes at least some of these informational gaps. We gathered information about household's pre-existing knowledge of fuel-efficient cookstoves. Approximately 8.2% of the urban sample already owned a fuel-efficient stove, whereas 52.5% of those that did not already own a fuel-efficient stove had heard of what one was, and 53.1% of those that had heard of a fuel-efficient stove knew that they used less fuel than a traditional charcoal stove. In sum, this means 33.1% of the entire sample (combining those that knew about fuel-efficient stoves with those that already owned one) knew that fuel-efficient cookstoves saved fuel. For the two arms of the experiment that were not designed to include a potential learning channel (cash and carry and time payments) the differences in eventual purchase rates were small and not statistically significant between participants that knew that fuel-efficient stoves saved fuel versus those who did not. Uptake (less returns and defaults) for cash and carry was 3.7% among those that did not know fuel-efficient stoves saved fuel and 5.0% for those that did know (the difference of 1.3 percentage points is not significant, p = 0.46), whereas the uptake (less returns and defaults) for the time payment offer was 24.2% versus 25.0% for those not aware of the fuel savings of the new stove versus those aware the stoves were designed to save fuel (the difference of 0.8 percentage points is not significant, p = 0.86). However, in the two experimental arms designed to allow learning (free trial and free trial with time payments) eventual purchase was slightly more likely among those that did not already know about the fuel saving properties of the stoves. Uptake (less returns and defaults) for the free trial offer was 28.2% among those that did not know fuel-efficient stoves saved fuel and 22.9% for those that already knew about the fuel savings properties of the introduced stove. Although the rate of eventual purchase was higher for those without pre-existing knowledge of the fuel saving properties of the stoves, the difference of 5.3 percentage points is not statistically significant (p = 0.19). Uptake (less returns and defaults) for the free trial and time payments offer was 48.5% for those not aware of the fuel savings of the introduced stove versus 36.7% for those that already knew the stoves were designed to save fuel. This difference of 11.8 percentage points is statistically significantly different than zero (p = 0.04). The higher rates of eventual purchase for participants in treatment arms that included a free trial but who were unaware of potential fuel-savings is consistent with learning about the fuel-saving properties of the stoves during the free trial increasing sales. However, it is possible the free trial may have removed other barriers in addition to imperfect product information. Consumers’ perception of the salesperson's trustworthiness may have increased, because free trials are not commonplace. Consumers may have purchased due to a norm of reciprocity after receiving the free trial, or the free trial created an endowment effect. Perhaps the free trial was a credible sign of quality, which was more important to the purchase decision than the information learned during the free trial. Although it is plausible that the higher rates of eventual purchase could be attributed to salesperson's trustworthiness, an endowment effect or norms of reciprocity, we present suggestive evidence supporting the importance of learning about fuel savings from an extension of this study in Dhaka, Bangladesh with water filters (Guiteras et al. 2016). In the related Bangladesh study, the free trial did not increase the adoption of water filters, although time payments had the same large effect as in Uganda with cookstoves. For the water filter, unlike the stove, a short trial cannot show effectiveness. That is, fuel savings show up within a few days of using the new stove, whereas it takes a sample of hundreds of weeks to gain confidence in the health benefits of a water filter. Thus, although this evidence is suggestive (at best) given the different context and product, we interpret this as showing the effect of the free trial is probably due more to learning than due to reciprocity, an endowment effect, or a signal of the salesperson's credibility. 4.5. Evidence of Liquidity as Binding Constraint We offered time payments to test the role of liquidity constraints as a potential market failure. Our offer was equivalent to a 0% interest loan, with four equal payments made weekly. The large increase in uptake from 4.0% with cash-and-carry to 25.9% with time payments suggests that this contract overcomes liquidity constraints. We were surprised that about a third of those who accepted a sales offer with time payments in the urban setting paid it off early (i.e., 33.7% of those offered time payments alone and 31.1% of those offered time payments after a free trial) and about a fifth paid off early in the rural setting (22.1%). We use this unexpected outcome to examine the role of liquidity constraints. If stove repayment rates were faster for individuals who had higher weekly charcoal expenditures it would suggest that liquidity is binding. If the differences in repayment rates are small, other forces are likely to be more prominent. We split the distribution of the urban sample in half based on the amount of weekly charcoal expenditures. When examining the entire sample 4.2% of those in the lower half of the charcoal expenditure distribution paid off their stove early, whereas 5.9% in the upper half of the charcoal expenditure distribution paid off their stove early. This difference is large (about 40% higher) and marginally statistically significant at the p = 0.10 level. Narrowing the sample to focus on the offer that relaxes liquidity constraints the most—the free trial plus time payment offer—we find 9.5% of those in the lower half of the charcoal expenditure distribution paid off their stove early, whereas 17.7% in the upper half of the charcoal expenditure distribution paid of their stove early. This difference is even larger (about 86% higher) and statistically significant at p = 0.03. Although not definitive, these results are suggestive that the liquidity constraint was binding in the urban context. We also examined this relationship in the rural sample, but with less conclusive results. In the rural sample, there are four types of households with regards to gathering and purchasing firewood. Approximately 6.6% of households neither purchased nor gathered wood—that is, they used other fuels higher up the energy ladder such as charcoal; 13.8% purchased wood, but did not gather; 62.6% gathered wood, but did not purchase; and 17.0% both gathered and purchased wood in the previous month. In a logistic regression of membership in these four groups on paying off the stove early, we find much higher (and statistically significant) odds ratios for each group compared to the omitted group of the households that neither purchased, nor gathered their wood. However, none of the three groups that either gathered wood or purchased wood are statistically significantly different from each other. Although higher early repayment rates for households that pay for firewood with either time or with money supports the idea of binding liquidity constraints; we would have also expected the group that only purchased wood with money to have a statistically significantly higher rate of early repayment. Although paying off stoves early can be construed as evidence of the importance of the liquidity channel, we also note that this behavior is also consistent with qualitative evidence that suggests many Ugandans consider debt undesirable. Several respondents, for example, said they were prepaying so that the stove salesperson would not come by for collections: they apparently perceived a stigma in owing additional payments. Prepayments may also have been motivated by the irregular nature of many customers’ incomes, coupled with the challenges of saving. By prepaying when they had cash on hand, they reduced the risk of losing the stove if they had no cash when the next payment was due. It is possible that one attraction of our time payments offer is this flexibility to prepay, in contrast, for example, to many typical microfinance loans. Additionally, we examine liquidity constraints and stove size. Recall that in urban areas we sold stoves in three sizes, with prices of $6, $8, or $10. If liquidity constraints bind more often for larger expenditures, then those who purchase for a lump sum will more often purchase the smallest of the stoves. This test is not perfect as fewer bought with a lump sum than time payments, so selection effects could reverse the effects of liquidity constraints (e.g., if the relatively small share who bought for a lump sum were disproportionately the families with easy access to at least $8). Collapsing offers into two groups (those with lump sum payments and those with extended time to pay), we find that among those that purchased a stove 81% of those with lump sum buyers (cash and carry offer and free trial offer combined) and 77% of those with extended time to pay (time payments offer and free trial plus time payments offer combined) purchased the smallest stove type. To create a p-value, we ran an ordered logit predicting stove size with the sales offer. Our sample is buyers who completed payments. The joint test that lump sum buyers (with or without a free trial) purchased smaller stoves than those offered extended time to pay (with or without a free trial) has a p-value of 0.83. The free trial and time payments offer increased uptake even further to 45.6%. This rate is over 11 times that of the cash-and-carry offer (4.0%) and substantially above the uptake of the free trial (29.1%) or time payments alone (25.9%). An interesting question is whether—when combining the free trial with time payments—these offers are additive (i.e., they alleviate different constraints for different people). If we assume that the offers are additive and that anyone who accepted the cash-and-carry offer would also accept either the free trial or time payments offers, we calculate the union of the free trial only and time payments only offers would induce about a 51.0% uptake (4.0% cash-and-carry offer + 25.1% free trial offer (29.1% − 4.0%) + 21.9% time payments offer (25.9% − 4.0%)). In our experiment, the combined offer with a free trial followed by four equal installment payments induced uptake in 45.6% of the sample. We cannot reject equality (p = 0.14) between 45.6% and 51.0% with a sample size equal to that used in our experiment, which suggests that these offers are additive. Some of the success in the time payments and free trial with time payments offer is likely due to a reallocation of fuel expenditures. The average household in our sample spent $2.61 per week on charcoal in Kampala, and the Ugastove's fuel saving estimates range from 35% to 46% (Partnership for Clean Indoor Air 2011; Wang et al. 2009). This corresponds to $0.91–1.20 per week in charcoal savings for the average household in the urban study. Across the four weeks, a household could have used fuel savings to fund the purchase of $3.64–4.80 worth of the stove. These fuel savings would fund approximately 61%–80% of the $6 stove (i.e., the model chosen 80% of the time). 4.6. Costs Borne by the Supplier The increase in uptake for the free trial and time payments offer comes with additional costs and risks borne by the vendor. The vendor bears the risk of default if a consumer takes a free trial and neither returns the stove nor pays for it; the vendor also bears the risk of damages to returned stoves. Additionally, the vendor needs more working capital to prepurchase fuel-efficient cookstoves. The important question to answer is whether the additional sales gained from any of the marketing offers presented in this paper sufficiently compensate the entrepreneurial vendor for the additional default risks and capital costs. Although exact costs would vary in different contexts, we present conservative estimates of the costs of the four offers in the urban study. Our field staff reported that a salesperson could visit approximately 25 households a day to make a sales pitch and a similar number for collection visits. In Table 7 we normalize effort across the four sales offers so that each salesperson makes 25 household visits per day (either an initial sales pitch or a follow-up collection visit). Therefore, in a month with 20 working days, a sales person can sell about 20 stoves a month with the traditional cash-and-carry offer, 113 stoves a month offering a free trial, 73 stoves per month offering time payments, and 81 stoves offering a free trial followed by time payments (see Table 7, first panel). Table 7. Comparing sales performance and costs in urban setting. Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Notes: The first panel calculates the approximate effort in terms of initial sales visits and followup collection visits required to make 100 sales. We estimate that a salesperson can make 25 household visits per day and has 20 working days per month. We also assume that followup collection visits are equally as time consuming as the initial sales visit. The sales per month column uses this rate of sales and prorates the amount of sales made in a representative month. The second panel uses the sales per month metric from the first panel and calculates additional costs for defaults, the cost of capital, and return losses. We use the default rates observed in our experiment (default rates are the share of revenue that is owed that was not collected in the field experiment) and conservatively estimate capital costs using a 6% monthly interest rate (higher than what microfinance banks in Kampala generally charge). In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate that the vendor suffers damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs could vary depending on the loan terms and how the salesperson structures sales, but this is likely a fair approximation of the working capital needed, and is comparable across the four offers). View Large Table 7. Comparing sales performance and costs in urban setting. Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Effort to make 100 sales Sales rate Initial sales visits Collection visits Total visits Working days Sales per month Sales price Sales per work day  Cash and carry 4.0% 2500 0 2500 100.0 20.0 $6.00 1.0  Free trial 29.1% 344 100 444 17.7 112.7 $6.00 5.6  Time payments 25.9% 386 300 686 27.4 72.9 $6.00 3.6  Free trial and time payments 45.6% 219 400 619 24.8 80.7 $6.00 4.0 Monthly capital, default, Default Return Gross Monthly Default Return Net Revenue and return costs rate rate revenue interest losses losses revenue ratio  Cash and carry 0.0% 0.0% $120.00 $7.20 $0.00 $0.00 $112.80 1.0  Free trial 9.8% 11.8% $676.22 $40.57 $66.27 $39.89 $529.49 4.7  Time payments 3.8% 1.0% $437.25 $26.24 $16.62 $2.14 $392.26 3.5  Free trial and time payments 3.1% 5.3% $484.42 $29.07 $15.02 $12.75 $427.59 3.8 Notes: The first panel calculates the approximate effort in terms of initial sales visits and followup collection visits required to make 100 sales. We estimate that a salesperson can make 25 household visits per day and has 20 working days per month. We also assume that followup collection visits are equally as time consuming as the initial sales visit. The sales per month column uses this rate of sales and prorates the amount of sales made in a representative month. The second panel uses the sales per month metric from the first panel and calculates additional costs for defaults, the cost of capital, and return losses. We use the default rates observed in our experiment (default rates are the share of revenue that is owed that was not collected in the field experiment) and conservatively estimate capital costs using a 6% monthly interest rate (higher than what microfinance banks in Kampala generally charge). In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate that the vendor suffers damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs could vary depending on the loan terms and how the salesperson structures sales, but this is likely a fair approximation of the working capital needed, and is comparable across the four offers). View Large These higher sales must be weighed against defaults, potential damages to returned stoves, and the cost of capital. To estimate these costs, we used the stove sales per month from the first panel, the default rates we observed in our experiment, and a conservative estimate of 6% per month interest rate (i.e., roughly 100% per year). Additionally, we include a damage charge for returned stoves worth 50% of the value of the returned stove. In our experiment we were able to resell returned stoves as if they were new, but to be conservative we include this damage charge. Even including defaults, possible damage to returned stoves and an interest charge; the experimental offers are preferable to the traditional cash-and-carry offer (see Table 7, second panel).11 Under the stated assumptions, the free trial offer is the vendors’ best option. However, this result only holds if collection visits are as time consuming as the initial sales presentation. As the cost of follow-up visits declines, the free trial and time payments offer becomes more favorable. Sales visits may be slower than follow-up visits because the latter will not include a sales pitch highlighting the stove's features. The vendor also can call or text ahead of time to ensure someone is home with the payment, and the use of mobile money could remove the need to meet. The wide and growing availability of mobile money in developing nations, especially sub-Saharan Africa (Aker and Mbiti 2014; Jack and Suri 2014; Luoto and Levine 2014) suggests that mobile payments will eventually reduce, or perhaps even eliminate, the number of in-person visits required to collect installment payments. However, Luoto and Levine (2014) did not find that collecting time payments with mobile money is profitable in their setting of rural western Kenya. Based on the assumptions underlying Table 7, if the time required for follow-up visits is cut in half, the free trial with time payments becomes the preferred sales offer. The logistical set up for rural sales was different from the urban study as households are widely dispersed and we used a local point person who received a commission for collecting payments. Although the calculations differ (see Table A.2 in the Appendix), the free trial with time payments offer is also preferred to the cash-and-carry offer in rural areas, even when accounting for default risk, a damage charge for returned stoves and increased capital costs. 4.7. Robustness and External Validity In the rural study, meeting attendees who heard they would be offered the free trial with time payments may have sent text messages to friends who might be interested. Such texting could have led to nonrepresentative samples at meetings with different sales offers. To test for this possibility, we reran the results removing the final 33%, 25%, and 10% of participants to arrive at each free trial and time payments meeting. The results were similar irrespective of how many late arrivers were removed (see Table A.3 in the Appendix for results removing the last 25% of arrivers from each meeting). Further questions arose about our results’ level of external validity and the performance of these sales offers with other technologies or products in the developing world. We carried out the field experiment in Kampala, in late 2010, and we were surprised by the large increase in uptake compared to the typical cash-and-carry offer (i.e., over 11 times increase for the free trial with time payments offer). Because the magnitude of this increase was so large, we solicited funding to run the same trial in a different setting. In the rural setting, with a different stove, fuel type, market, and socioeconomic context, the free trial with time payments offer increased uptake by 12 times, about the same as in our first experiment. However, how well the sales offer of a free trial followed by time payments would work with other products is still unclear. Offering time payments without a free trial raises purchase levels considerably for many products. These include insecticide-treated bed nets (Tarozzi et al. 2014), piped water connections (Devoto et al. 2012), and water filters (Guiteras et al. 2016), as well as for charcoal-burning stoves (in this paper's urban study) and for wood-burning stoves (Beltramo et al. 2015b). Furthermore, how well the benefits of a free trial generalize is less obvious. A free trial may decrease sales if the product is unpopular, as Luoto et al. (2012) found for chlorine for water treatment, or if the trial anchors consumers’ willingness to pay at a lower number (Fischer et al. 2014). A free trial will not raise demand if consumers cannot learn much about a product's effectiveness, which perhaps accounts for the lack of detectable effects of a free trial for water filters in Dhaka (Guiteras et al. 2016). Finally, free trials do not work well for products that cannot be returned, such as built-in products (e.g., latrines and some stoves) or consumable products. With these cautions in mind, quite plausibly, many products—besides the cookstove models studied here—could increase sales if vendors offered free trials. Lastly, it is worth noting that the overall welfare of this intervention relies on both the purchase of a fuel-efficient stove (as studied in this paper) and the actual use of the fuel-efficient stove and its potential health and environmental benefits (the subject of future work). Although it appears that households understand the value of the time and/or money the fuel-efficient stove saves, we acknowledge that we cannot assess the complete welfare of the households’ purchase decision without additional data on long-term health benefits and data on fuel use. 5. Conclusion We examined the sale of two different fuel-efficient cookstoves in two different settings in Uganda. In contrast to other studies that have altered cookstove prices and found low demand, we found extremely high demand for fuel-efficient cookstoves sold at local market prices but with altered sales contract terms. In urban Kampala, we offered four contracts for charcoal burning stoves: cash-and-carry (4% uptake), a one week free trial followed by full payment (29% uptake), four equal weekly time payments starting immediately (26% uptake), and a combination of a one week free trial followed by four equal weekly time payments (46% uptake). Default levels were low since the combined offer resulted in recovery of 97% of the purchase price of the stoves sold. We repeated two sales offers in a rural area with a different market, fuel type, stove model, socioeconomic setting, and a slightly different experimental design. We sold wood-burning fuel-efficient stoves but also found a quite large uptake (57%) with the free trial plus time payments offer, when compared to the cash-and-carry offer (5%). Defaults were also low in the rural setting, as we recovered 99% of the stove sales price. Similarly high uptake in both settings reinforces the idea that low demand does not inherently hamper the sale of beneficial environmental and health-improving technologies such as fuel-efficient cookstoves. Instead, other important barriers, including information and liquidity constraints, must be eliminated before these purchases will occur. Appendix Table A.1. Robustness check: Purchase and payment by sales offer in rural (Mbarara) study only in parishes where stove was offered for $16.00 USD. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Notes: The four parishes where stoves were sold for $12.00 USD are excluded to calculate the results in this table. Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.1. Robustness check: Purchase and payment by sales offer in rural (Mbarara) study only in parishes where stove was offered for $16.00 USD. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 412 720 Purchased or accepted free trial 21 5.1 432 60.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.2  Returned after payments began 35 8.1  Paid in full 393 91.0  In default 3 0.7 Among those that paid in full  Paid off early 93 23.7  Paid off late 4 1.0 Among those in default  Paid more than zero 2 66.7  Paid at least half price 1 33.3 Uptake (less returns) 21 5.1 396 55.0 0.000 Revenue as share of owed 100.0 99.4 Uptake (less returns and defaults) 21 5.1 393 54.5 0.000 Revenue as share of owed 100.0 100.0 Notes: The four parishes where stoves were sold for $12.00 USD are excluded to calculate the results in this table. Sales offers made at parish wide sales meeting. Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.2. Comparing sales performance and costs in rural setting. Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Notes: Sales offers made at village wide sales meeting with approximately 50 attendees each meeting. We simulate one sales meeting per week, and four different villages visited in a month. A village focal point person was recruited to gather the community for the initial sales meeting and then to followup with collecting the installment payments. One village focal point person in each village was paid with a free stove ($16.00 value) and $1.60 per installment payment collected (or $6.40 per stove sold once the full payment amount was collected). We use the default rates and return rates observed in our experiment and conservatively estimate capital costs using a 6% monthly interest rate. In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs would vary depending on the loan terms and actual sales, but this figure is at least comparable across the offers). View Large Table A.2. Comparing sales performance and costs in rural setting. Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Cash and carry Free trial and time payments Monthly sales scenario (1 sales meeting per week) Sales rate 4.6% 57.0% Monthly sales meetings 4 4 Attendees per sales meeting 50 50 Sales 9.2 114 Stove price $16.00 $16.00 Default rate 0.0% 0.9% Return rate 0.0% 8.2% Total revenue $147.20 $1,824.00 Collection costs per installment payment $1.60 $1.60 Installment payments per stove 0 4 Collection costs per stove $0.00 $6.40 Total variable collection costs $0.00 $729.60 Fixed cost to village focal person ($16 per person) $64.00 $64.00 Capital cost (6% per month) $8.83 $109.44 Default losses $0.00 $16.42 Return losses $0.00 $74.78 Total expenses $72.83 $994.24 Net revenue $74.37 $829.76 Revenue ratio 1.0 11.2 Notes: Sales offers made at village wide sales meeting with approximately 50 attendees each meeting. We simulate one sales meeting per week, and four different villages visited in a month. A village focal point person was recruited to gather the community for the initial sales meeting and then to followup with collecting the installment payments. One village focal point person in each village was paid with a free stove ($16.00 value) and $1.60 per installment payment collected (or $6.40 per stove sold once the full payment amount was collected). We use the default rates and return rates observed in our experiment and conservatively estimate capital costs using a 6% monthly interest rate. In our experiment we were able to resell stoves that had been returned, however to account for the risk of possible damages to returned stoves we conservatively estimate damage equal to 50% of the value of the stove for each returned stove. We calculate the interest rate on the retail value of the entire stock of stoves used in one month (actual capital needs would vary depending on the loan terms and actual sales, but this figure is at least comparable across the offers). View Large Table A.3. Robustness check: Purchase and payment rates after removing late arrivers in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Notes: As a robustness check we remove the last 25% of the attendees that arrived at each sales meeting. This is to ensure the representativeness of each sample as it is possible that early arrivers to the meeting texted their friends (whom they knew to already be interested in a fuel-efficient cookstove). If those that arrived the latest to the sales meeting had a higher propensity to purchase a stove, our results would be biased. Comparing these results to the full sample results shows no such bias. Results are largely the same for different cutoff points (removing the final 33%, 25%, or 10% of attendees). Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.3. Robustness check: Purchase and payment rates after removing late arrivers in rural (Mbarara) study. Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Cash and carry Free trial and time payments Test of equality N % N % p-value Number of offers made 396 640 Purchased or accepted free trial 19 4.8 397 62.0 0.000 Among those that accepted free trial  Returned after free trial 1 0.3  Returned after payments began 28 7.1  Paid in full 364 91.7  In default 4 1.0 Among those that paid in full  Paid off early 83 22.8  Paid off late 24 6.6 Among those in default  Paid more than zero 2 50.0  Paid at least half price 1 25.0 Uptake (less returns) 19 4.8 368 57.5 0.000 Revenue as share of owed 100.0 99.1 Uptake (less returns and defaults) 19 4.8 364 56.9 0.000 Revenue as share of owed 100.0 100.0 Notes: As a robustness check we remove the last 25% of the attendees that arrived at each sales meeting. This is to ensure the representativeness of each sample as it is possible that early arrivers to the meeting texted their friends (whom they knew to already be interested in a fuel-efficient cookstove). If those that arrived the latest to the sales meeting had a higher propensity to purchase a stove, our results would be biased. Comparing these results to the full sample results shows no such bias. Results are largely the same for different cutoff points (removing the final 33%, 25%, or 10% of attendees). Two-sample test of proportions used to test equality between sales offers. Revenue as a share of what consumers owed does not count returned stoves as either revenue or what consumers owed. View Large Table A.4. Prices used for construction of aggregate asset values. Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Notes: Price data used to construct value of assets are average prices of durable goods and livestock taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Study (LSMS). This data is publicly available at: econ.worldbank.org. View Large Table A.4. Prices used for construction of aggregate asset values. Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Price in USD Television 134.27 Bicycle 69.70 Radio 13.83 Vehicle 4509.61 Motorcycle 783.78 Mobile phone 34.26 Indigenous cow 252.23 Indigenous goat 33.18 Indigenous sheep 26.43 Indigenous pig 45.27 Notes: Price data used to construct value of assets are average prices of durable goods and livestock taken from the 2011 to 2012 round of the Uganda Living Standards Measurement Study (LSMS). This data is publicly available at: econ.worldbank.org. View Large Footnotes 1 Examples of studies examining the reasons for the low adoption of health and welfare improving technologies include (but are not limited to): latrines (Gertler et al. 2015; Guiteras, Levinsohn, and Mobarak 2015), point-of-use water treatment technologies (Albert, Luoto, and Levine 2010; Ashraf, Berry, and Shapiro 2010; Kremer et al. 2009), insecticide-treated bed nets (Cohen and Dupas 2010; Dupas 2009; Tarozzi et al. 2014), hand washing with soap (Halder et al. 2010; Luby et al. 2011; Meredith et al. 2013), vaccines (Clemens and Jodar 2005; Cropper et al. 2004), deworming (Kremer and Miguel 2007; Meredith et al. 2013), and micronutrient fortification (Dewey and Adu‐Afarwuah 2008; Meredith et al. 2013). 2 See the “About CIRCODU” webpage for more details: http://www.circodu.org/about-us/. 3 We did this to provide sufficient separation so that consumers from different neighborhoods would be unlikely to communicate regularly (i.e., they would probably attend different churches, schools, shops, etc.). 4 We elected to use a fixed way finding procedure (i.e., enumerators visited every sixth house while moving down a street in the same direction) as opposed to the more common technique of a full neighborhood census and then a follow-up visit to selected households to make the sales pitch. The fixed way finding procedure was selected to save costs as it required fewer field visits and it would also ensure that the first time a household was contacted they received a sales pitch by the door-to-door salespeople (enumerators). In addition to cost savings, we avoided a census with a follow-up visit as it could have potentially created expectations for assistance on the part of the households if they were initially surveyed about household energy usage and at a subsequent visit received an offer for a device like a fuel-efficient cookstove related to household energy usage. 5 The population of most Ugandan rural parishes ranges from 4,000 to 6,000. 6 See http://upenergygroup.com/projects/uganda/. 7 Because we planned to track usage and health benefits of the stove in a follow-on impact evaluation, we needed to make sure enough stoves sold. We thus started with a lower than market sales price. Once our results clearly showed that the stoves were popular, we increased the price to the approximate market price to ensure that we did not run out of stoves. Excluding the four parishes where the stoves sold for $12 does not materially change our results (see Table A.1 in the Appendix). 8 Because homes were geographically spread out in the rural areas, we hired a local focal point person in each village to organize the sales meetings and to collect the follow-up time payments. We paid the village focal point person for each installment payment he or she collected and as part of this structure he or she was in charge of collecting returned stoves if a consumer chose to return the stove. The focal point person would keep the stove at his or her place of residence until our research team was in the area again and could pick up the returned stove (and use it in another village). 9 An experience good is a product or service whose characteristics, such as quality or price, are difficult to observe in advance but are ascertained upon consumption. 10 The study protocol did not permit households offered the free trial to pay a portion of the stove price. However, when the sales staff went to the house either to collect the stove or full payment, some households offered to pay a portion of the purchase price. We had not anticipated this possibility when we trained the sales staff. In practice, the sales staff accepted partial payment in these cases. Therefore, some ultimate defaulters may have paid at least some of the purchase price in the free trial offer. 11 Note that if default is enough of a problem (higher than what we observed), then this policy may not be feasible. Acknowledgements The editor in charge of this paper was M. Daniele Paserman. Acknowledgments: We received funding from the Goggio Family Foundation, the National Institute of Child Health and Human Development (NICHD R21 HD056581), the Silicon Valley Community Foundation, the Institute for Research on Labor and Employment, the International Institute for Impact Evaluation (3IE), the Haas School of Business, the United States Agency for International Development (Translating Research into Action, Cooperative Agreement No. GHS-A-00-09-00015-00), the Atkinson Center for a Sustainable Future, the Institute for the Social Sciences, and the Cornell Population Center. Data collection was carried out by the Center for Integrated Research and Community Development (CIRCODU), and the project's success depended greatly on its managers—Joseph Arinieitwe Ndemere, Juliet Kyaesimira, Vastinah Kemigisha—and field supervisors—Innocent Byaruhanga, Fred Isabirye, Michael Mukembo, Moreen Akankunda, and Noah Kirabo. Stephen Harrell and Dow-Li Kou (of IPA) expertly oversaw field operations. We thank Impact Carbon, Evan Haigler, Caitlyn Hughes, Matt Evans, Jimmy Tran, Johanna Young, Rashad Korah Thomas, Adam Galinsky, Andy Weiss, and the USAID TRAction Technical Advisory Group. The views expressed in this paper solely reflect those of the authors, and these opinions are not necessarily those of the institutions with which the authors are affiliated. All errors are our own. References Aker Jenny C. , Mbiti Isaac M. ( 2014 ). “ Mobile Phones and Economic Development in Africa .” Journal of Economic Perspectives , 24 ( 3 ), 207 – 232 . Google Scholar CrossRef Search ADS Albert Jeff , Luoto Jill , Levine David ( 2010 ). “ End-User Preferences for and Performance of Competing POU Water Treatment Technologies among the Rural Poor of Kenya .” Environmental Science and Technology , 44 , 4426 – 4432 . Google Scholar CrossRef Search ADS PubMed Ashraf Nava , Berry James , Shapiro Jesse M. ( 2010 ). “ Can Higher Prices Stimulate Product Use? Evidence from a Field Experiment in Zambia .” American Economic Review , 100 ( 5 ), 2383 – 2413 . Google Scholar CrossRef Search ADS Bailis Robert , Drigo Rudi , Ghilardi Adrian , Masera Omar ( 2015 ). “ The Carbon Footprint of Traditional Woodfuels .” Nature Climate Change , 5 , 266 – 272 . Google Scholar CrossRef Search ADS Becker Gordon M. , DeGroot Morris H. , Marschak Jacob ( 1964 ). “ Measuring Utility by a Single-Response Sequential Method .” Behavioral Science , 9 , 226 – 232 . Google Scholar CrossRef Search ADS PubMed Beltramo Theresa , Blalock Garrick , Levine David I. , Simons Andrew M ( 2015a ). “ Does Peer Use Influence Adoption of Efficient Cookstoves? Evidence From a Randomized Controlled Trial in Uganda .” Journal of Health Communication , 20 , 55 – 66 . Google Scholar CrossRef Search ADS Beltramo Theresa , Blalock Garrick , Levine David I. , Simons Andrew M. ( 2015b ). “ The Effect of Marketing Messages and Payment over Time on Willingness to Pay for Fuel-Efficient Cookstoves .” Journal of Economic Behavior & Organization , 118 , 333 – 345 . Google Scholar CrossRef Search ADS Beltramo Theresa , Levine David I. ( 2013 ). “ The Effect of Solar Ovens on Fuel Use, Emissions and Health: Results from a Randomised Controlled Trial .” Journal of Development Effectiveness , 5 , 178 – 207 . Google Scholar CrossRef Search ADS Bensch Gunther , Grimm Michael , Peters Jörg ( 2015 ). “ Why Do Households Forego High Returns from Technology Adoption? Evidence from Improved Cooking Stoves in Burkina Faso .” Journal of Economic Behavior & Organization , 116 , 187 – 205 . Google Scholar CrossRef Search ADS Bensch Gunther , Peters Jörg ( 2015 ). “ The Intensive Margin of Technology Adoption—Experimental Evidence on Improved Cooking Stoves in Rural Senegal .” Journal of Health Economics , 42 , 44 – 63 . Google Scholar CrossRef Search ADS PubMed Berry James , Fischer Greg , Guiteras Raymond ( 2015 ). “ Eliciting and Utilizing Willingness to Pay: Evidence from Field Trials in Northern Ghana .” Discussion Paper, Centre for Economic Policy Research , London, UK , www.cepr.org/active/publications/discussion_papers/dp.php?dpno=10703 . Bonjour Sophie , Adair-Rohani Heather , Wolf Jennyfer , Bruce Nigel G. , Sumi Mehta , Prüss-Ustün Annette , Lahiff Maureen , Rehfuess Eva A. , Mishra Vinod , Smith Kirk R. ( 2013 ). “ Solid Fuel Use for Household Cooking: Country and Regional Estimates for 1980–2010 .” Environmental Health Perspectives , 121 , 784 – 790 . Google Scholar CrossRef Search ADS PubMed Burwen Jason , Levine David I. ( 2012 ). “ A Rapid Assessment Randomized-Controlled Trial of Improved Cookstoves in Rural Ghana .” Energy for Sustainable Development , 16 , 328 – 338 . Google Scholar CrossRef Search ADS Center for Entrepreneurship in International Health and Development ( 2008 ). Cook Stove Projects and Carbon Finance , Center for Entrepreneurship in International Health and Development , San Francisco, CA . Cialdini Robert B. ( 2007 ). Influence: The Psychology of Persuasion , Collins , New York . Clemens John , Jodar Luis ( 2005 ). “ Introducing New Vaccines into Developing Countries: Obstacles, Opportunities and Complexities .” Nature Medicine , 11 , S12 – S15 . Google Scholar CrossRef Search ADS PubMed Cohen Jessica , Dupas Pascaline ( 2010 ). “ Free Distribution or Cost-Sharing? Evidence from a Randomized Malaria Prevention Experiment .” Quarterly Journal of Economics , 125 , 1 – 45 . Google Scholar CrossRef Search ADS Cropper Maureen L. , Haile Mitiku , Lampietti Julian , Poulos Christine , Whittington Dale ( 2004 ). “ The Demand for a Malaria Vaccine: Evidence from Ethiopia .” Journal of Development Economics , 75 , 303 – 318 . Google Scholar CrossRef Search ADS Davis Scott , Gerstner Eitan , Hagerty Michael ( 1995 ). “ Money Back Guarantees in Retailing: Matching Products to Consumer Tastes .” Journal of Retailing , 71 , 7 – 22 . Google Scholar CrossRef Search ADS Devoto Florencia , Duflo Esther , Dupas Pascaline , Parienté William , Pons Vincent ( 2012 ). “ Happiness on Tap: Piped Water Adoption in Urban Morocco .” American Economic Journal: Economic Policy , 4 , 68 – 99 . Google Scholar CrossRef Search ADS Dewey Kathryn G , Adu Seth , Afarwuah ( 2008 ). “ Systematic Review of the Efficacy and Effectiveness of Complementary Feeding Interventions in Developing Countries .” Maternal & Child Nutrition , 4 , 24 – 85 . Google Scholar CrossRef Search ADS PubMed Dupas Pascaline ( 2009 ). “ What Matters (and What Does Not) in Households’ Decision to Invest in Malaria Prevention? ” American Economic Review , 99 ( 2 ), 224 – 230 . Google Scholar CrossRef Search ADS Dupas Pascaline ( 2011 ). “ Health Behavior in Developing Countries .” Annual Review of Economics , 3 , 425 – 449 . Google Scholar CrossRef Search ADS Fink Günther , Masiye Felix ( 2012 ). “ Assessing the Impact of Scaling-up Bednet Coverage through Agricultural Loan Programmes: Evidence from a Cluster Randomised Controlled Trial in Katete, Zambia .” Transactions of the Royal Society of Tropical Medicine and Hygiene , 106 , 660 – 667 . Google Scholar CrossRef Search ADS PubMed Fischer Greg , Karlan Dean , McConnell Margaret , Raffler Pia ( 2014 ). To Charge or Not to Charge: Evidence from a Health Products Experiment in Uganda , National Bureau of Economic Research , Cambridge, MA . Gertler Paul , Shah Manisha , Alzua Maria Laura , Cameron Lisa , Martinez Sebastian , Patil Sumeet ( 2015 ). How Does Health Promotion Work? Evidence from the Dirty Business of Eliminating Open Defecation , National Bureau of Economic Research , Cambridge, MA . Grossman Sanford ( 1981 ). “ The Informational Role of Warranties and Private Disclosure about Product Quality .” Journal of Law and Economics , 24 , 461 – 483 . Google Scholar CrossRef Search ADS Guiteras Raymond , Levine David I. , Polley Thomas , Quistorff Brian ( 2016 ). Credit Constraints, Discounting and Investment in Health: Evidence from Micropayments for Clean Water in Dhaka , University of Maryland , College Park, MD . Guiteras Raymond , Levinsohn James , Mobarak Ahmed Mushfiq ( 2015 ). “ Encouraging Sanitation Investment in the Developing World: A Cluster-Randomized Trial .” Science , 348 , 903 – 906 . Google Scholar CrossRef Search ADS PubMed Halder Amal K , Tronchet Carole , Akhter Shamima , Bhuiya Abbas , Johnston Richard , Luby Stephen P ( 2010 ). “ Observed Hand Cleanliness and Other Measures of Handwashing Behavior in Rural Bangladesh .” BMC Public Health , 10 , 545 . doi:10.1186/1471-2458-10-545 . Google Scholar CrossRef Search ADS PubMed Hanna Rema , Duflo Esther , Greenstone Michael ( 2016 ). “ Up in Smoke: The Influence of Household Behavior on the Long-Run Impact of Improved Cooking Stoves .” American Economic Journal: Economic Policy , 8 , 80 – 114 . Google Scholar CrossRef Search ADS Harrell Stephen , Beltramo Theresa , Blalock Garrick , Kyayesimira Juliet , Levine David I. , Simons Andrew M. ( 2016 ). “ What Is a Meal?: Comparing Methods of Auditing Carbon Offset Compliance for Fuel Efficient Cookstoves .” Ecological Economics , 128 , 8 – 16 . Google Scholar CrossRef Search ADS Jack William , Suri Tavneet ( 2014 ). “ Risk Sharing and Transactions Costs: Evidence from Kenya's Mobile Money Revolution .” American Economic Review , 104 ( 1 ), 183 – 223 . Google Scholar CrossRef Search ADS Kremer Michael , Miguel Edward ( 2007 ). “ The Illusion of Sustainability .” Quarterly Journal of Economics , 122 , 1007 – 1065 . Google Scholar CrossRef Search ADS Kremer Michael , Miguel Edward , Mullainathan Sendhil , Null Clair , Zwane Alix Peterson ( 2009 ). “ Making Water Safe: Price, Persuasion, Peers, Promoters, or Product Design? ” Working paper. Unpublished manuscript, Harvard University , Cambridge, MA . Lim Stephen S. , Vos Theo , Flaxman Abraham D. , Danaei Goodarz , Shibuya Kenji , Adair-Rohani Heather , Amann Markus et al. ( 2012 ). “ A Comparative Risk Assessment of Burden of Disease and Injury Attributable to 67 Risk Factors and Risk Factor Clusters in 21 Regions, 1990–2010: A Systematic Analysis for the Global Burden of Disease Study 2010 .” Lancet , 380 , 2224 – 2260 . Google Scholar CrossRef Search ADS PubMed Luby Stephen P , Halder Amal K , Huda Tarique , Unicomb Leanne , Johnston Richard B ( 2011 ). “ The Effect of Handwashing at Recommended Times with Water Alone and With Soap on Child Diarrhea in Rural Bangladesh: An Observational Study .” PLoS Medicine , 8 , e1001052 . doi:10.1371/journal.pmed.1001052 . Google Scholar CrossRef Search ADS PubMed Luoto Jill , Levine David I ( 2014 ). Mobile Phones, Rent-to-Own Payments & Water Filters: Evidence from Kenya , RAND Labor and Population , Santa Monica, CA . http://www.rand.org/content/dam/rand/pubs/working_papers/WR1000/WR1039/RAND_WR1039.pdf . Luoto Jill , Mahmud Minhaj , Albert Jeff , Luby Stephen , Najnin Nusrat , Unicomb Leanne , Levine David I. ( 2012 ). “ Learning to Dislike Safe Water Products: Results from a Randomized Controlled Trial of the Effects of Direct and Peer Experience on Willingness to Pay .” Environmental Science & Technology , 46 , 6244 – 6251 . Google Scholar CrossRef Search ADS PubMed Madajewicz Malgosia , Pfaff Alexander , van Geen Alexander , Graziano Joseph , Hussein Iftikhar , Momotaj Hasina , Sylvi Roksana , Ahsan Habibul ( 2007 ). “ Can Information Alone Change Behavior? Response to Arsenic Contamination of Groundwater in Bangladesh .” Journal of Development Economics , 84 , 731 – 754 . Google Scholar CrossRef Search ADS Meredith Jennifer , Robinson Jonathan , Walker Sarah , Wydick Bruce ( 2013 ). “ Keeping the Doctor Away: Experimental Evidence on Investment in Preventative Health Products .” Journal of Development Economics , 105 , 196 – 210 . Google Scholar CrossRef Search ADS Miller Grant , Mobarak A. Mushfiq ( 2015 ). “ Learning About New Technologies Through Social Networks: Experimental Evidence on Nontraditional Stoves in Bangladesh .” Marketing Science , 34 , 480 – 499 . Google Scholar CrossRef Search ADS Mobarak Ahmed Mushfiq , Dwivedi Puneet , Bailis Robert , Hildemann Lynn , Miller Grant ( 2012 ). “ Low Demand for Nontraditional Cookstove Technologies .” Proceedings of the National Academy of Sciences of the United States of America , 109 , 10815 – 10820 . Google Scholar CrossRef Search ADS PubMed Moorthy Sridhar , Srinivasan Kannan ( 1995 ). “ Signaling Quality with a Money-Back Guarantee: The Role of Transaction Costs .” Marketing Science , 14 , 442 – 466 . Google Scholar CrossRef Search ADS Partnership for Clean Indoor Air ( 2011 ). “Uganda Stove Manufacturers Ltd (Ugastove)—Global Leadership Award.” In 5th Biennial Partnership for Clean Indoor Air Forum , Lima, Peru . http://www.pciaonline.org/2011ForumAwards/uganda-stove-manufacturers-ltd-ugastove . Ramanathan Veerabhadran , Carmichael Gregory ( 2008 ). “ Global and Regional Climate Changes due to Black Carbon .” Nature Geoscience , 1 , 221 – 227 . Google Scholar CrossRef Search ADS Shieh Shiou ( 1996 ). “ Price and Money-Back Guarantees as Signals of Product Quality .” Journal of Economics & Management Strategy , 5 , 361 – 377 . Google Scholar CrossRef Search ADS Simons Andrew M. , Beltramo Theresa , Blalock Garrick , Levine David I. ( 2014 ). “ Comparing Methods for Signal Analysis of Temperature Readings from Stove Use Monitors .” Biomass and Bioenergy , 70 , 476 – 488 . Google Scholar CrossRef Search ADS Simons Andrew M. , Beltramo Theresa , Blalock Garrick , Levine David I. ( 2017 ). “ Using Unobtrusive Sensors to Measure and Minimize Hawthorne Effects: Evidence from Cookstoves .” Journal of Environmental Economics and Management , 86 , 68 – 80 . Google Scholar CrossRef Search ADS Smith Kirk R , Bruce Nigel , Arana Byron , Jenny Alisa , Khalakdina Asheena , McCracken John , Diaz Anaite , Schei Morten , Gove Sandy ( 2006 ). “ Conducting The First Randomized Control Trial of Acute Lower Respiratory Infections and Indoor Air Pollution: Description of Process and Methods .” Epidemiology , 17 , S44 – S44 . Google Scholar CrossRef Search ADS Smith Kirk R. , McCracken John P. , Thompson Lisa , Edwards Rufus , Shields Kyra N. , Canuz Eduardo , Bruce Nigel ( 2010 ). “ Personal Child and Mother Carbon Monoxide Exposures and Kitchen Levels: Methods and Results from a Randomized Trial of Woodfired Chimney Cookstoves in Guatemala (RESPIRE) .” Journal of Exposure Science and Environmental Epidemiology , 20 , 406 – 416 . Google Scholar CrossRef Search ADS PubMed Smith-Sivertsen Tone , Díaz Esperanza , Pope Dan , Lie Rolv T. , Díaz Anaite , McCracken John , Bakke Per , Arana Byron , Smith Kirk R. , Bruce Nigel ( 2009 ). “ Effect of Reducing Indoor Air Pollution on Women's Respiratory Symptoms and Lung Function: The RESPIRE Randomized Trial, Guatemala .” American Journal of Epidemiology , 170 , 211 – 220 . Google Scholar CrossRef Search ADS PubMed Suwelack Thomas , Hogreve Jens , Hoyer Wayne D. ( 2011 ). “ Understanding Money-Back Guarantees: Cognitive, Affective, and Behavioral Outcomes .” Journal of Retailing , 87 , 462 – 478 . Google Scholar CrossRef Search ADS Tarozzi Alessandro , Mahajan Aprajit , Blackburn Brian , Kopf Dan , Krishnan Lakshmi , Yoong Joanne ( 2014 ). “ Micro-Loans, Insecticide-Treated Bednets, and Malaria: Evidence from a Randomized Controlled Trial in Orissa, India .” American Economic Review , 104 ( 7 ), 1909 – 1941 . Google Scholar CrossRef Search ADS Uganda Bureau of Statistics ( 2005 ). “ 2002 Uganda Population and Housing Census: Main Report .” Uganda Bureau of Statistics , Kampala, Uganda . Wang Jin , Tyler Erika , Adkins Edwin , Modi Vijay ( 2009 ). Field Evaluation of Rocket Design Stoves—Fuelwood Use and User Preferences , Columbia University , New York, NY . Yu Fei ( 2011 ). “ Indoor Air Pollution and Children's Health: Net Benefits from Stove and Behavioral Interventions in Rural China .” Environmental and Resource Economics , 50 , 495 – 514 . Google Scholar CrossRef Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of European Economic Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Journal

Journal of the European Economic AssociationOxford University Press

Published: Jan 13, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off