Policy Research Working Paper 8928 Stability and Evolution of Preferences for Improved Cookstoves A Difference-in-Difference Analysis of a Choice Experiment from Ethiopia Sahan T. M. Dissanayake George Voigt Abbie Cooper Abebe Damte Beyene Randall Bluffstone Zenebe Gebreegziabher Daniel LaFave Peter Martinsson Alemu Mekonnen Michael Toman Development Economics Development Research Group June 2019 Policy Research Working Paper 8928 Abstract There is a growing effort in the non-market valuation lit- 504 households in 36 communities in 2013, and 486 of erature toward better understanding of the stability and the same households participated in 2016 (a 96 percent evolution of preferences over time. The study uses a novel retention rate). The results show that preferences of the approach combining a repeated choice experiment with respondents from the control group are stable over the study a randomized controlled trial on stove adoption in Ethi- period, while preferences of the respondents from the treat- opia to analyze the stability and evolution of preferences. ment group evolve. Moreover, households in the treatment The treatment group in the randomized controlled trial group still using the stoves have significantly higher willing- received an improved fuelwood stove with less fuelwood ness to pay for all the stove’s attributes in 2016 compared use, whereas the control group continued to use traditional with 2013, indicating how longer experience can increase cooking methods. Respondents were given the exact same the willingness to pay for technology with environmentally choice questions in 2013 and 2016. The study began with preferable attributes. This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The lead author and task team lead leader may be contacted at sdissan2@gmail.com and mtoman@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Stability and Evolution of Preferences for Improved Cookstoves - A Difference-in- Difference Analysis of a Choice Experiment from Ethiopia Sahan T. M. Dissanayake*, George Voigt, Abbie Cooper, Abebe Damte Beyene, Randall Bluffstone, Zenebe Gebreegziabher, Daniel LaFave, Peter Martinsson, Alemu Mekonnen, Michael Toman Keywords: temporal stability, preferences, improved cookstoves, choice experiment, Ethiopia JEL Codes: Q51, Q56, D00                                                         * Corresponding author: Dissanayake, Portland State University, Portland, OR, USA sdissan2@gmail.com Dissanayake: Portland State University, Portland, OR, USA; Beyene: Ethiopian Development Research Institute, Addis Ababa, Ethiopia; Cooper: Colby College, Waterville, ME, USA; Gebreegziabher, Ethiopian Development Research Institute and Mekelle University, Addis Ababa and Mekelle, Ethiopia; LaFave: Colby College, Waterville, ME, USA; Martinsson, P: Gothenburg University, Gothenburg, Sweden; Mekonnen, Ethiopian Development Research Institute and Addis Ababa University, Addis Ababa, Ethiopia; Toman, The World Bank, Washington DC, USA; Voigt: Colby College, Waterville, ME, USA. Financial support for this research from the World Bank and from the Knowledge for Change Trust Fund is gratefully acknowledged. Stability and Evolution of Preferences for Improved Cookstoves - A Difference- in-Difference Analysis of a Choice Experiment from Ethiopia 1. Introduction Non-market valuation methods are used to determine the value, preferences and potential consumer demand for non-market environmental goods and policies. Valuation studies, in particular stated- preference methods that involve surveys, are often costly, and therefore conducted infrequently (Bliem et al. 2012, Brouwer and Bateman 2005, Mørkbak and Olsen 2014, Price et al. 2016). If the values obtained from these studies are stable over time, meaning there is no significant change in preferences among the given survey sample over time, they can be used in future analysis (Bliem et al. 2012, Brouwer and Bateman 2005, Morten and Olsen 2014, Price et al. 2016). Therefore, rather than acquiring funding for data collection, historical data that exhibit stability can be used for benefit transfer and cost-benefit analysis in the future. However, due to the hypothetical nature of these valuation techniques, it is often difficult to determine the credibility or validity of the results as time progresses. Credibility is a measure of temporal validity, or as Mørkbak and Olsen (2014) state, “Does an individual respondent provide identical answers to identical questions over time, assuming that preferences are stable?” If preferences are not found to be stable, the use of results from stated preference studies conducted in the past for benefit transfer, cost-benefit analysis and calculation of welfare over time becomes problematic. Further, the few studies on measuring stability of preferences elicited from choice experiment surveys have been conducted in developed country settings. At the same time there is a growing use of non-market valuation surveys, especially choice experiments, in developing country contexts to guide environmental policy and regulations. Understanding if preferences elicited in developing country settings, especially in rural communities, are stable is important with regard to using results from non-market valuation studies over time. Therefore, the first focus of this work is to determine if preferences elicited using choice experiment surveys in developing country settings are stable over time. We test the temporal stability of preferences for identical respondent groups over a three-year period, using identical choice experiment surveys, in a test-retest experimental setting. The background of the study is a valuation study on preferences for clean cookstoves and a randomized controlled trial (RCT) of improved cookstoves conducted in Ethiopia. The main objective of the valuation study was to use a choice experiment survey to examine Ethiopian households’ 2    valuation of attributes of cookstoves, including durability, fuel reduction, smoke reduction, and amount of time saved by using new clean, cooking technology. The overall goal of the survey was to determine features of clean cookstoves that were well-liked by communities in order to help implement sustainable cooking practices in rural Ethiopia, with the hopes of reducing the harmful health and environmental impacts of traditional cooking methods. We use the choice experiment survey on preferences for attributes of clean cookstoves conducted in Ethiopia in 2013 and repeated in 2016 to analyze the temporal stability of preferences. The unique setting of having a repeated choice experiment and an RCT allows us to also study technology adoption and how preferences and willingness-to-pay for improved cookstoves evolve with use and experience. There is a large economics literature on technology adoption (see Marra et al. 2003 and Conley and Udry 2010) and a smaller but growing literature on adoption and learning with regard to cookstoves (see Ruiz-Mercado 2011 and Miller and Mobarak 2014). At the same time, it has been found that the Mirt stoves save fuelwood (Gebreegziabher et al., 2018) and that it may reduce household air pollution (Bluffstone et al. 2018). Therefore, the second focus of this work is to evaluate the degree to which such experience with such outcomes causes users to shift their preferences and change their demand for attributes. Therefore, our research questions become Q1: Are the preferences for cookstoves elicited from choice experiments stable over time? Q2: How do users’ preferences and willingness to pay for cookstove attributes evolve after extensive experience with the Mirt stoves? The combined RCT and repeated choice experiment setting allows us to use a difference-in- difference framework to analyze the results from the choice experiment, the first instance of a difference-in-difference choice experiment analysis that we are aware of. We use the control group to evaluate stability of preferences and the treatment group to evaluate the degree to which experience with the Mirt stove shifted user preferences and attribute values. Results show that the willingness to pay for stove attributes of the control group are stable over the three-year period from 2013 to 2016 and that the preferences of the treatment group have evolved over this time period. To better understand how long-term stove use impacts preferences, we then refine the treatment group to only include those currently using the stoves and we find that, compared to the control group, long-term stove use increases the WTP for most of the stove attributes. 3    2. Literature on Stability of Preferences There are a small number of studies that examine the temporal stability of contingent valuation and choice experiment studies but the results on the stability of preferences are mixed. Brouwer and Bateman (2005) studied the temporal stability of WTP responses from two contingent valuation studies (on flood control and wetland conservation) administered five years apart. They find that when examining WTP values over an extended time period, there is a statistically significant decrease in how people value flood control and wetland conservation attributes over time. They attribute these results to the five-year gap in the survey administration, which is much longer than the test-retest period that most researchers use, and also to the utilization of expanded models to include more, ad hoc transitory factors. Price et al. (2016) and Bliem et al. (2012) find differing results from Brouwer and Bateman (2005) in their analysis of the stability of WTP values obtained from stated preference surveys regarding the health effects of tap water consumption. Price et al. (2016) conducted their survey in 2004 and again in 2012 using multiple survey formats. Their findings show that there is no significant difference in WTP between the two survey formats, and they support the transfer of non-market benefits over extended time periods. Bliem et al. (2012) test for the temporal stability of individual preferences by using two identical surveys for a restoration project of a stretch of the Danube River in Austria repeated over a one-year period. The researchers find that the preferences are stable across the one-year time frame. Mørkbak and Olsen (2014) test the reliability and stability using an incentivized test-retest choice experiment. In their study, they use a choice experiment for a market good with real economic incentives, and sample respondents using the exact same questionnaire twice within a two-week lag time. Analysis of the data shows that error variances between the two identical surveys do not differ significantly, but the larger the utility difference in a choice task, the larger the probability that the respondent will choose the same alternative in the retest. Morten and Olsen also determine that the longer the time a respondent takes to answer the choice sets in the retest, the lower the probability that they will choose the same alternatives in the retest as they did in the original test. Our study assesses how households in Ethiopia value different attributes of improved cookstove technology, and if these values are stable over a period of three years. This work contributes to the existing body of work on the stability of WTP values elicited from CE surveys using a test- 4    retest structure and differs from previous studies in that a randomly selected portion of the sample, the RCT treatment group, received an actual improved Mirt stove after completing the survey in 2013. Therefore, when the same sample was given an identical survey in 2016, a portion of them had already experienced cooking with the new, efficient stove technology. Results from this study not only contribute to the growing body of research about temporal stability of preferences, but also contribute to the existing literature about global sustainable development through the adoption of clean technologies. 3. Empirical Approach Given the improved cookstove RCT and the repeated choice experiments that were conducted in 2013 and in 2016, our data allow us to study temporal stability and the evolution of preferences elicited using a choice experiment survey within a difference-in-difference framework. The choice experiment in 2013 was conducted immediately prior to the improved stove RCT. Choice Experiment Design Choice experiments are a stated preference valuation method used to determine individuals’ willingness to pay (WTP) for goods that do not have attainable market values. A choice experiment survey was chosen for this study over other stated preference valuation methods because they provide information about individual attributes of the good being valued. We follow standard practice in the choice modeling literature for this research (Adamowicz et al. 1997, Adamowicz et al. 1998, Carlsson and Martinsson 2003, Louviere et al. 2000). We initially conducted informal focus groups, engaged in discussions with researchers and then finally conducted 15 formal focus groups and a field trial of the survey. Through this process we identified five attributes that we use for the final choice experiment; durability of the stove, the reduction of fuel use, the reduction of smoke, decrease in cooking time, and the cost of the stove. The attribute levels are presented in Table 1. 5    Attributes Description Levels Durability of stove How long the stove would be - No improved stove functional for. - 1-5 years - 6-10 years - 11-15 years Reduction in amount of The reduction in fuel wood use in - No reduction fuel wood used comparison to using the - 25% reduction traditional cookstove. - 50% reduction Reduction of smoke The reduction of smoke as a - No reduction result of using an improved stove. - 25% reduction - 50% reduction Amount of cooking time The amount of cooking time - No reduction reduction reduced by adopting an improved - 25% reduction stove. - 50% reduction Cost of improved stove The amount paid to acquire a new - No payment improved stove. - 100 Ethiopian Birr - 200 Ethiopian Birr - 300 Ethiopian Birr - 400 Ethiopian Birr Table 1: Attribute Levels The survey presents a number of choice questions where each choice question consists of two stove alternatives that contain varied levels of the five stove attributes that a respondent can choose from, and an additional status quo option of no improved cookstove at no cost.1 To account for learning, the first and seventh questions of the survey were identical, and the first question was dropped for data analysis (Carlsson, Mørkbak, and Olsen 2012). Figure 1 shows an example of one choice experiment question.                                                         1 It is important to note that for the RCT treatment group, the status-quo option reflects going back to a traditional stove. We discuss how this might impact the interpretation of the results in the conclusion. 6    Figure 1: Choice Experiment Question Example We conduct a national survey in Ethiopia starting with 504 households in 36 communities in 2013 and are able to conduct the study with 486 of the same households in 2016, giving an extremely low attrition rate of just 4%. The 2016 sample includes 342 treatment households and 139 control households. For more details about the choice experiment and the data collection please refer to Dissanayake et al. 2016 which analyzes the preferences for stove attributes from the data collected in 2013. 4. Methods of Analysis Econometric model specification For all our analysis we use a difference-in-difference specification that compares the 2013 responses with the 2016 responses across the treatment and control groups. In our first specification we look at how preferences have evolved between the treatment and control groups. 7    , ∑ , , , , ∑ , , , , ∑ , , , 16 (1) Where ASC is the alternative specific constant, Xk is each of the four stove attributes, Cost is the cost of the stove, and 13 denotes 2013 base data, 16t denotes the treatment group in 2016 and 16c denotes the control group in 2016. The coefficients of interest for evaluating the evolution of preferences, the causal impact of the treatment, are , , , , and , . The coefficients of interest for evaluating the stability of preferences, the change in the untreated, are, , , , , and , . The coefficient estimates from the conditional logit model and the mixed multinomial logit model cannot be interpreted directly. Therefore, the average marginal WTP is calculated for a change in each attribute i by dividing the coefficient estimate for each attribute by the coefficient estimate for the payment term, as given in (3). . (3) We conduct the difference-in-difference choice experiment estimation using the MIXLOGIT command in STATA and analyze the data using both a conditional logit model (CL) and a mixed multinomial model (MMNL). 5. Results and Discussion The results from the main effects estimation for the mixlogit analysis are provided in Table 2 (mean coefficient estimates in column 1 and 2 and the standard deviation estimates in columns 3 and 4). Columns (1) and (3) present results for the sample with all treatment households and columns (2) and (4) present the results for the sample where the treatment households are limited to the long-term stove use households. Table , and Figures 2 and 3 provide the WTP estimates for both the full set of treatment households and the long-term stove use treatment households. In Figures 2 and 3, the first panel includes the base year, the 2016 treatment and the 2016 control, and the second panel drops the 2016 control (to be able to clearly compare the base 2013 results with the 2016 treatment). The WTP results (Table 3 and Figures 2 and 3) show that between 2013 and 2016 the WTP for all attributes have evolved for the treatment group for both the full sample of treatment households and long-term use treatment households. The WTP for stove attributes for the control 8    group remain stable for both categories of treatment households (except for stove durability for the second sample). The ASC estimates (Table 2) for the control households are not significant, indicating that the overall preferences for the stoves have not changed for control households. This conclusion is robust across both definitions of treatment households. The standard deviation estimates from the MMNL indicate that there is significant heterogeneity for most attributes in both the treatment group and the control group for both definitions of treatment households. Given the complexity of the model that we are estimating, we are not able to include interaction terms to better analyze heterogeneity in the estimates. The average WTP results for the base/2013 and the 2016 treatment households are provided in Table 4. We find that households in the treatment group are WTP on average 54 birr (about 2 USD) for a stove that lasts a year longer (17 birr in 2013), 86 birr (about 3 USD) for a stove that reduces cooking time by 10% (47 birr in 2013), 142 birr (about 5 USD) for a stove that reduces fuelwood use by 10% (62 birr in 2013), and 112 birr (about 4 USD) for a stove that reduces smoke by 10% (45 birr in 2013). 9    Table 2: Diff-in-Diff Results – Mean Estimates Mean Estimates SD Estimates (1) (2) (3) (4) Treatment Defined As All Households in Households Still All Households in Households Still Treatment Group Using Mirt Treatment Group Using Mirt 2013 Baseline Values ASC 25.79*** 17.53*** -12.83*** -7.475*** (4.874) (3.364) (2.430) (1.398) StoveDurability 0.0597*** 0.0608*** -0.0237 -0.0301* (0.00876) (0.00915) (0.0194) (0.0170) TimeReduction 0.0155*** 0.0166*** 0.00283 0.00869** (0.00209) (0.00221) (0.00593) (0.00414) FuelReduction 0.0207*** 0.0221*** 0.0242*** 0.0290*** (0.00226) (0.00247) (0.00271) (0.00301) SmokeReduction 0.0143*** 0.0163*** 0.0122* -0.0154** (0.00356) (0.00377) (0.00697) (0.00607) Cost -0.00333*** -0.00348*** 0.00507*** 0.00612*** (0.000663) (0.000707) (0.000656) (0.000713) 2016 Treatment Group relative to 2013 Baseline ASCxtreatment16 -2.324* 36.03 -1.530 4.435 (1.305) (1259604.0) (1.084) (652309.5) StoveDurabilityxtreatment16 0.0994*** 0.141*** 0.192*** 0.247*** (0.0226) (0.0317) (0.0262) (0.0351) TimeReductionxtreatment16 0.0108** 0.0145** 0.0510*** 0.0496*** (0.00527) (0.00660) (0.00672) (0.00959) FuelReductionxtreatment16 0.0284*** 0.0191*** 0.0569*** 0.0557*** (0.00572) (0.00707) (0.00615) (0.00824) SmokeReductionxtreatment16 0.0236*** 0.0180 -0.0466*** 0.0561*** (0.00823) (0.0112) (0.0124) (0.0137) Costxtreatment16 0.000780 -0.000938 -0.00569*** 0.00469 (0.00137) (0.00169) (0.00202) (0.00311) 2016 Control Group relative to 2013 Baseline ASCxcontrol16 15.73 12.91 0.286 0.974 (1260.4) (465.9) (979.0) (517.3) StoveDurabilityxcontrol16 0.0549* 0.0794** 0.124*** 0.167*** (0.0293) (0.0335) (0.0328) (0.0421) TimeReductionxcontrol16 0.00653 0.000729 -0.0453*** 0.0353*** (0.00714) (0.00651) (0.0104) (0.00993) FuelReductionxcontrol16 0.0311*** 0.0272*** 0.0510*** 0.0506*** (0.00846) (0.00825) (0.00885) (0.00949) SmokeReductionxcontrol16 0.0538*** 0.0373*** -0.0912*** -0.0863*** (0.0159) (0.0142) (0.0181) (0.0180) Costxcontrol16 0.00186 0.000916 0.00837*** 0.00113 (0.00206) (0.00190) (0.00236) (0.00469) Observations 17712 15318 Log lik. -3424.0 -2971.0 Chi-squared 423.2 305.8 Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01 10    Table 3: Diff-in-Diff Results – WTP Estimates (1) (2) MMNL MMNL Treatment-All Treatment-Use 2013 Baseline Values ASC 7742.7*** 5034.6*** (2027.4) (1355.7) StoveDurability 17.93*** 17.48*** (2.717) (2.745) TimeReduction 4.651*** 4.776*** (1.198) (1.232) FuelReduction 6.205*** 6.348*** (1.595) (1.648) SmokeReduction 4.284*** 4.673*** (1.582) (1.659) 2016 Treatment Group (total) ASCxtreatment16 9198.2* 12117.2 (4738.2) (284992118.2) StoveDurabilityxtreatment16 62.36** 45.66*** (27.32) (13.81) TimeReductionxtreatment16 10.31* 7.040** (5.605) (2.994) FuelReductionxtreatment16 19.22* 9.331** (10.26) (3.993) SmokeReductionxtreatment16 14.83* 7.759* (8.863) (4.273) 2016 Control Group (total) ASCxcontrol 28254.2 11866.4 (858440.3) (181816.1) StoveDurabilityxcontrol16 78.01 54.67* (95.03) (32.83) TimeReductionxcontrol16 14.98 6.765 (21.49) (5.585) FuelReductionxcontrol16 35.19 19.24 (49.78) (14.62) SmokeReductionxcontrol16 46.31 20.90 (66.35) (17.07) Observations 17712 15318 Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01 Note: WTP values are presented as totals for each group (after accounting for interactions) 11    0 20 40 60 80 100 -100 0 100 200 300 StoveDurability StoveDurability TimeReduction TimeReduction FuelReduction SmokeReduction FuelReduction StoveDurabilityxtreatment16 SmokeReduction TimeReductionxtreatment16   12  FuelReductionxtreatment16 StoveDurabilityxtreatment16 SmokeReductionxtreatment16 WTP - Full Treatment WTP - Full Treatment TimeReductionxtreatment16 Figure 2: Marginal WTP Estimates StoveDurabilityxcontrol16 TimeReductionxcontrol16 Marginal WTP estimates for all treatment households FuelReductionxtreatment16 FuelReductionxcontrol16 SmokeReductionxtreatment16 SmokeReductionxcontrol16 Marginal WTP estimates for all treatment households – displayed without control households 0 20 40 60 80 0 20 40 60 80 100 StoveDurability StoveDurability TimeReduction TimeReduction FuelReduction SmokeReduction FuelReduction StoveDurabilityxuse16 SmokeReduction TimeReductionxuse16   13  FuelReductionxuse16 StoveDurabilityxuse16 SmokeReductionxuse16 TimeReductionxuse16 StoveDurabilityxcontrol16 WTP - Long Term Use Treatment WTP - Long Term Use Treatment TimeReductionxcontrol16 FuelReductionxuse16 FuelReductionxcontrol16 Marginal WTP estimates for treatment households with long-term use SmokeReductionxuse16 SmokeReductionxcontrol16 Figure 3: Marginal WTP estimates for treatment households with long-term use Marginal WTP estimates for long-term use households – displayed without control households Table 4: Average WTP Average WTP 2013 2016 Stove Durability (1 year) 17 birr 54 birr ($2) Time Reduction (10%) 47 birr 86 birr ($3) Fuel Reduction (10%) 62 birr 142 birr ($5) Smoke Reduction (10%) 45 birr 112 birr ($4) Note: The average is calculated between all treatment households and long-term use treatment households. The conversation to $ (USD) is approximate (1 USD = 27.52 birr May 2018) The results support the view that the WTP estimates obtained from choice experiment surveys (for the control group) are stable, at least over a three-year period, and can therefore be used for cost- benefit analysis and welfare calculations over time. The results also show that respondents in treatment groups, especially the respondents who continue to use the stoves long-term, are willing to pay significantly more for the individual benefits of the stoves than in 2013. Next, we discuss the policy implications of this in the conclusion. 6. Conclusions and Policy Implications There is a growing effort in the non-market valuation literature to better understand the stability and evolution of preferences over time. A small number of studies (Bliem et al. 2012, Brouwer and Bateman 2005, Mørkbak and Olsen 2014, Price et al. 2016) have used the test-retest method to analyze stability of preferences but have found mixed results. A combined repeated choice experiment and an RCT on stove adoption provide a unique opportunity to analyze both the stability and the evolution of preferences. Respondents were given the exact same choice question in 2013 and 2016 and the treatment group in the RCT received improved firewood stoves, whereas the control group continued to use traditional stoves. This setting allows us to use a difference-in-difference framework to analyze the results from the choice experiment, the first instance of a difference-in-difference choice experiment analysis that we are aware of. Further, we exploit the heterogeneity in use in the treatment group to analyze the evolution of preferences in response to the effectiveness of the stoves. We conduct a national survey in Ethiopia starting with 504 households in 36 communities in 2013, and we are able to conduct the study with 486 of the same households in 2016, giving an extremely low attrition rate of just 4%. 14    Results show that WTP for stove characteristics for the control groups are stable over the study period. As noted in the introduction, this was the first test-retest choice experiment study conducted in a developing country setting. This result supports the use of values obtained from non- market studies over multiple years. The results also show that the WTP for stove attributes for the treatment group increases over time and this is consistent for all stove attributes. We analyze the robustness of this result by focusing on respondents in the treatment group that are long-term stove users. It is important to note that about 1/3 of the treatment group has stopped using the stoves. We are unable to analyze this subgroup given the complexity of the estimation process and model convergence issues. Understanding the preferences of the households that have stopped using the stoves could be very important. The results show that fuelwood reduction is the most highly valued attribute of the stoves and WTP for fuelwood reduction more than doubles between 2013 and 2016. Overall, the evidence from this study highlights that the Mirt stove, a technology developed by the Federal Government of Ethiopia in conjunction with donors, is well liked by long-term users and their overall valuation of the benefits of the improved stoves increases with long-term use of the Mirt stove. References Adamowicz, W., Boxall, P., Williams, M. and Louviere, J. (1998). Stated Preference Approaches for Measuring Passive Use Values: Choice Experiments and Contingent Valuation. American Journal of Agricultural Economics 80, 64-75. Adamowicz, W., Swait, J., Boxall, P., Louviere, J. and Williams, M. (1997).Perceptions Versus Objective Measures of Environmental Quality in Combined Revealed and Stated Preference Models of Environmental Valuation. Journal of Environmental Economics and Management 32, 65-84. Bliem, M., Getzner, M., Rodiga-LaBing, P. (2012) “Temporal Stability of Individual Preferences for River restoration in Austria Using a Choice Experiment. Journal of Environmental Management 103, 65-73. Brouwer, R., Bateman, I. (2005). Temporal Stability and Transferability of Models of Willingness to Pay for Flood Control and Wetland Conservation. Water Resources Research 41, W03017. Carlsson, F. and Martinsson, P. (2003) Design Techniques for Choice Experiments in Health Economics, Health Economics 12, 281-94. Carlsson, F., Mørkbak, R. and Olsen, B. (2012). The First Time is the Hardest: A Test of Ordering Effects in Choice Experiments. Journal of Choice Modelling 5(2), 19-37. Conley, T. and Udry C. (2010). “Learning about a New Technology: Pineapple in Ghana.” American Economic Review 100, 35-69. Dissanayake, S. T. M, Beyene, A. D., Bluffstone, R, Kiggundu, G., Kooser, S, Gebreegziabher, Z, Mekonnen, A, Martinsson, P. and Toman, M. (2018). Improved Cook Stoves for Climate 15    Change Mitigation? Evidence of Values, Preferences and Carbon Savings from a Choice Experiment in Ethiopia. World Bank Policy Research Working Paper 8499 (June). Gebreegziabher, Z. Beyene, A, Bluffstone, R, Martinsson, P, Mekonnen, A, Toman, M. (2018). Fuel Savings, Cooking Time and User Satisfaction with Improved Biomass Cookstoves: Evidence from Controlled Cooking Tests in Ethiopia. Resource and Energy Economics 52: 173-185. Marra, M., Pannell, D. J., & Ghadim, A. A. (2003). The Economics of Risk, Uncertainty and Learning in the Adoption of New Agricultural Technologies: Where are We on the Learning Curve?. Agricultural Systems 75(2-3), 215-234. Miller, G., & Mobarak, A. M. (2014). Learning about New Technologies through Social Networks: Experimental Evidence on Nontraditional Stoves in Bangladesh. Marketing Science 34(4), 480-499. Mørkbak, M.R., and Olsen, S.B. (2014) A Within-Sample Investigation of Test-Retest Reliability in Choice Experiment Surveys with Real Economic Incentives. Agricultural and Resource Economics 59, 375-392. Price, J., Dupont, D., and Adamowicz, W. (2016) As Time Goes By: Examination of Temporal Stability Across Stated Preference Question Formats. Environmental and Resource Economics 68(3), 643-662. Ruiz-Mercado, I., Masera, O., Zamora, H., & Smith, K. R. (2011). Adoption and Sustained Use of Improved Cookstoves. Energy Policy 39(12), 7557-7566. 16