Policy Research Working Paper 11016 Conserving Wildlife through Demand Reduction and Supply Alternatives Two Experiments in Restaurants in Kinshasa Abdoulaye Cisse Gabriel Englander Daniel J. Ingram Development Economics A verified reproducibility package for this paper is Development Research Group available at http://reproducibility.worldbank.org, January 2025 click here for direct access. Policy Research Working Paper 11016 Abstract High aggregate levels of wildlife consumption in Central potential alternative to wild meat, affects restaurants’ total African cities may be depleting wildlife populations. This wild meat sales. The elasticity estimate indicates that a 1% study explores the impacts of demand and supply-side reduction in the price of Moambe Chicken reduces total interventions on wild meat consumption through two wild meat sales by 0.91%. Although this relationship is randomized control trials in restaurants in Kinshasa, Dem- not statistically significant, it suggests that interventions ocratic Republic of the Congo: a demand-side experiment increasing the availability and affordability of alternatives and a supply-side experiment. In the demand-side experi- to wild meat may reduce wild meat consumption. The ment, 544 subjects were given a coupon to their restaurant experiments advance previous wild meat research by uti- of choice and randomly assigned to view either a treatment lizing actual consumption data rather than self-reported video discouraging wild meat consumption or a control data, assessing social desirability bias, and pre-registering all video unrelated to wild meat. Treatment group subjects statistical specifications to enhance research integrity. This are 31% less likely to order wild meat than control group paper provides preliminary evidence that both wild meat subjects, though this difference is not statistically signifi- demand reduction through social marketing campaigns and cant and may be affected by social desirability bias. In the supply expansion via affordable alternatives could contrib- supply-side experiment, the research question is whether ute to effective wildlife conservation in Central Africa. randomly reducing the price of Moambe Chicken, a 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 authors may be contacted at aenglander@worldbank.org. A verified reproducibility package for this paper is available at http:// reproducibility.worldbank.org, click here for direct access. RESEA CY LI R CH PO TRANSPARENT ANALYSIS S W R R E O KI P NG PA 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 Conserving Wildlife through Demand Reduction and Supply Alternatives: Two Experiments in Restaurants in Kinshasa † * Gabriel Englander, Abdoulaye Cisse, Daniel J. Ingram‡ Keywords: biodiversity conservation, wild meat consumption, Central Africa, randomized con- trol trials JEL Classification: Q57, O13, Q21, Q56, D12, C93 * Agricultural and Resource Economics, University of California, Berkeley; California, USA † Development Research Group, The World Bank; Washington, USA. Correspondence: aenglander@worldbank.org ‡ Durrell Institute of Conservation and Ecology, University of Kent; Canterbury, CT2 7NR, UK 1 Introduction Millions of people in tropical regions of the world depend on wildlife as a source of food and means of acquiring income (Coad et al., 2019; Wells et al., 2024). Simultaneously, overexploita- tion, including from hunting that exceeds population growth rates, has emerged as the most sig- nificant threat to many species (IPBES, 2019), impacting wildlife populations and threatening ecosystem functioning. The consumption of the meat of wild animals, “wild meat”, is prevalent in rural areas, and there is also substantial demand for wild meat among people living in some urban areas (Carignano Torres et al., 2022; Edderai & Dame, 2006; Ingram et al., 2021; Simo et al., 2024). Consumption of wild meat in towns and cities is driven by a variety of factors, including culture, taste preference, the perception that it is the most natural meat, and the rela- tive availability and price compared to domesticated meats (Chausson et al., 2019; van Vliet & Mbazza, 2011; Wilkie et al., 2016). Where wild meat is more expensive than domesticated meat alternatives it can also be considered a luxury product that signifies status (Sandalj et al., 2016). When demand for wild meat in urban areas drives hunting to unsustainable levels, it poses a unique challenge and opportunity for conservation efforts. Addressing this challenge requires innovative and just solutions that respect local livelihoods and traditions, protect wildlife, and cultivate the relational values between people and nature (Chan et al., 2018; Ingram, 2020). There are two primary approaches to tackle overexploitation of wild meat in urban areas: reducing the demand for wildlife products or limiting the supply of wildlife to the market. ıssimo Recent years have seen a surge in wildlife product demand reduction campaigns (Ver´ & Wan, 2019; Willis et al., 2022). For instance, WildAid, a non-governmental organization (NGO), invests heavily in social marketing campaigns to reduce demand for wildlife products (WildAid, 2020). The World Bank’s Global Wildlife Program has advocated for increased investment in such interventions (Sobrevila, 2016). Alongside efforts to reduce demand, supply- 2 side interventions seek to limit the quantity of wildlife supplied to the market. For example, the Sustainable Wildlife Management Programme aims to regulate hunting and expand poultry production in some of the 15 countries in which it operates (Food and Agriculture Organization et al., 2019). Both approaches have strengths and limitations. Demand-side interventions can create awareness and shift societal norms, but attempts to change preferences in domains other than wild meat consumption often fail (MacFarlane et al., 2022). Careful design and years-long campaigns may be required to change deeply ingrained preferences. But if successful, demand- side interventions can provide a lasting benefit to wildlife populations: even if wild meat were available, fewer people would choose to consume or purchase it, reducing the incentive to hunt wildlife. Supply-side interventions, on the other hand, can have more immediate impacts. By more directly reducing the quantity of wildlife killed for food, they can quickly benefit wildlife populations. However, supply-side interventions alone may not be sufficient to reduce wild meat consumption to sustainable levels, because persistent demand leaves the economic incen- tive for hunting unaddressed. Despite the numerous and extensive initiatives undertaken by governments and NGOs, the effectiveness of both demand-side and supply-side interventions in changing actual wildlife consumption habits, especially in urban settings, remains underex- ıssimo & Wan, plored (Ingram et al., 2021; MacFarlane et al., 2022; Travers et al., 2021; Ver´ 2019). In November 2023, we implemented two complementary randomized control trials in Kin- shasa. Our research questions are: (1) Does a demand-side intervention reduce wild meat con- sumption in restaurants? (2) Does a supply-side intervention reduce wild meat consumption in restaurants? The study focused on restaurants in Kinshasa, the capital city of the Demo- cratic Republic of the Congo (DRC) and one of the fastest growing megacities in the world with a population of ∼17 million people. The DRC is the largest country in Central Africa, 3 comprises 61% of the region’s forests, and is highly biodiverse (Grantham et al., 2020). Kin- shasa is located on the Congo River, south of the world’s second largest rainforest, and it has approximately 3,000 wild meat (“bushmeat”) restaurants (Fa et al., 2019). Wild meat enters the city by road, river, and plane, and it is sold in markets and restaurants across the city (Lucas et al., 2022). A recent study estimated that 8,592 wild meat dishes were sold in restaurants in Kinshasa each day, equating to 1,254 tonnes (by live weight) of wild meat annually (Wright et al., 2022). Primates and ungulates were the most popular types of wild meat, whereas gen- erally the most frequently sold dishes across the city contained either fish, beans or chicken. Whilst there appears to be no single driver of wild meat consumption in Kinshasa, tradition and culture, social status, taste, freshness, cost, and accessibility all play a role (Trefon, 2023; Yocum et al., 2022). Furthermore, wild meat can be more expensive than domesticated meat in restaurants, suggesting it could be consumed more by wealthier residents (LaCerva, 2016; Wright et al., 2022). Restauranteurs in Kinshasa state that they sell wild meat because of high demand, its profitability, to maintain Congolese culture, and menu diversification (Wright et al., 2022). Indeed, some traditional Congolese dishes are made predominantly with wild meat. In restaurants with menus, wild meat is often openly listed for sale. Laws pertaining to wild meat in the DRC refer to the act of hunting, requiring a permit, and to the species that are protected. No specific laws regulate the consumption in urban settings or trade of meat obtained by hunting (Sustainable Wildlife Management Programme, 2024). 1.1 Conceptual Framework and Contributions Relational values refer to the preferences, principles, and virtues associated with relationships, both interpersonal and as articulated by policies and social norms, that contribute to a mean- ingful life (Chan et al., 2016). Unlike instrumental values (nature for human use) or intrinsic values (nature for its own sake), relational values emphasize the connections and responsibili- 4 ties between people and nature (Klain et al., 2017). Our study is situated within the relational values framework, focusing on how human behaviors, cultural ties, and economic incentives impact conservation outcomes. Relational values guide both the design and analysis of our experiments. The demand-side experiment employs a culturally resonant video that connects pride in Congolese wildlife to traditional culinary heritage without wild meat. This video aims to strengthen people’s cultural bonds to wildlife, in order to shift consumer behavior through a renewed sense of stewardship and identity (West et al., 2018). Likewise, in the supply-side experiment, we tested how offering a culturally relevant alternative—Moambe Chicken—at a lower price could reduce wild meat consumption. This intervention respects existing food- related cultural practices while encouraging sustainable consumption, aligning with relational values that prioritize harmonious human-environment interactions (Schulz et al., 2017). Our study offers practical insights for policymakers and conservation practitioners on de- signing culturally sensitive interventions that resonate with local values and traditions. By us- ing actual consumption data and robust methodologies—including randomized control trials, pre-registration, and response bias measurement—we adhere to best practices in conservation research (Cisse et al., 2023). Internationally, the demand and supply-side interventions we evaluate could be tailored to urban areas beyond Kinshasa where wildlife consumption poses conservation challenges. Our study’s theoretical contribution to the relational values framework is demonstrating how economic incentives can be integrated with cultural values to effect be- havioral change. By addressing the socio-cultural drivers of wild meat consumption, our work highlights the critical role of integrating relational values into conservation efforts to foster sustainable behaviors that are both ecologically beneficial and culturally acceptable. 5 2 Methods 2.1 Design of Demand-Side Experiment A fundamental challenge in evaluating the effectiveness of demand-reduction interventions is the difficulty in controlling exposure to specific messages. Mediums like billboards or televi- sion commercials offer limited insight into who views the message and, more critically, who constitutes the control group (e.g., those not exposed to the message). Establishing a credible control group is necessary as it acts as a comparison point or counterfactual, illustrating what the wild meat consumption level might have been in the absence of the intervention. Thus, our demand-side experiment was designed with a dual focus: first, to manage and document each participant’s exposure to the intervention message; and second, to link this exposure to their subsequent consumption choice. Restaurants in Kinshasa present an ideal setting for our experiment due to the real-world decision-making environment they provide. When patrons dine at a restaurant, they make real choices based on their preferences and budget, spending their own money on the dishes they choose. This setting facilitates the measurement of choices representative of actual, rather than hypothetical, consumption behavior. We recruited four restaurants in Kinshasa to participate in our study (Figure 1), based on two criteria: they needed to offer both wild meat and non- wild meat options daily, ensuring a variety of choices for customers; and the restaurant owners agreed, via contract, to report each participant’s order to us, allowing us to link their choices to the intervention message they received. In the experiment, enumerators set up four tables, each representing one of the participat- ing restaurants. These tables were positioned two or three blocks away from their respective restaurants, ensuring that the study and the restaurants were out of sight from each other. This configuration served a dual purpose: it prevented participants from being influenced by the di- 6 4.31°S Latitude 4.33°S 4.35°S Restaurant A Restaurant B Restaurant C Restaurant D 4.37°S 15.26°E 15.28°E 15.30°E 15.32°E 15.34°E 15.36°E Longitude Figure 1: Approximate Locations of Participating Restaurants. To ensure anonymity, the actual locations of the restaurants were randomly adjusted by up to approximately 3 km. The dashed circles, centered on these modified coordinates, indicate that the true locations of the restaurants are situated somewhere within the corresponding circles. The inset map on the left shows the Democratic Republic of the Congo (green polygon) within Africa (grey), while the inset on the right highlights Kinshasa (green triangle) within the Democratic Republic of the Congo (grey). 7 rect visibility of the restaurants during the experiment and ensured that participants were not visible to enumerators when visiting the restaurants and ordering a dish. The placement also guaranteed that each subject had convenient access to at least one restaurant. Over eight days in November 2023, enumerators at each table encouraged passersby to par- ticipate in a survey, offering a coupon to their restaurant of choice as an incentive. Participation was limited to adults (18 years or older) who are active wild meat consumers, defined as sub- jects who reported eating bushmeat at least once in the past month. Subjects were required to use their coupon within two weeks. Enumerators followed a strict protocol, attending to each participant individually to prevent interaction between participants that could bias their choices. Subjects were randomly assigned to watch one of two videos on a tablet with headphones. The treatment group watched a publicly available 90-second clip from the 2019 multimedia campaign “Let’s eat less bushmeat in the city” (“Mangeons moins de viande de brousse en ville”), led by the Government of neighboring Republic of the Congo with the aim of reducing wild meat consumption in the city of Pointe Noire. In the video, two childhood friends re- unite, and when one suggests cooking wild meat, they discuss the environmental consequences ere de l’Economie of urban wild meat consumption on the depletion of forest wildlife (Minist` ere - Congo, 2019). They decide to avoid such consumption in the future and instead Foresti` embrace Congolese dishes that do not contain wild meat (Appendix B). Rather than address all drivers of wild meat consumption simultaneously, the video we used focused on messages that target pride in Congolese forests and wildlife, whilst also supporting traditional Congolese cuisine. These could be considered as targeting drivers primarily linked to culture and tradition. The control group viewed a 90-second clip from a Congolese soap opera unrelated to wild meat or the environment. After watching the video, subjects chose one of four restaurants for a $5 coupon (2,500 Congolese Francs). We provided subjects with the name, location, and a sample menu from 8 each restaurant. We compensated restaurant owners for each redeemed coupon. Subjects were responsible for paying the difference between the price of their meal and the coupon value. For example, a subject would have to pay $4 for a $9 meal ($9 − $5 = $4). Each coupon had a unique identifier, linking subjects’ orders to their treatment assignment status. Subjects were informed during the survey that they should order whatever dish they preferred, regardless of whether it was wild meat or not. Immediately following the video viewing and restaurant selection, subjects completed a survey on their attitudes toward bushmeat and the frequency of consumption by people in their social network (Appendix C). The attitude questions asked subjects to express their level of agreement with statements that wild meat is sustainable, fresh, tasty, cool, legal, healthy, and connects them to their place of origin. These attitudes were selected based on other studies (Chausson et al., 2019; Wilkie et al., 2016), with the exception of legal and sustainable which we added out of interest. Additionally, enumerators asked subjects about their pride in the environment of the DRC. The purpose of the attitudinal questions was to provide insight into why any potential treatment effects occurred. Additionally, subjects answered 13 questions related to social desirability, enabling us to assess response bias (Dhar et al., 2022). 554 subjects participated in the demand-side experiment, of whom 59% used their coupon within two weeks to order a meal at their restaurant of choice. 2.2 Analysis of Demand-Side Experiment 2.2.1 Analysis of Wild Meat Consumption and Coupon Usage To understand the effects of the video aimed at discouraging wild meat consumption (treat- ment), our pre-specified analyses employ ordinary least squares regression to estimate equations 9 of the following form (Cisse et al., 2023): Yijt = βTi + αXi + κj + ζt + ϵijt . (1) In this equation, Yijt is an outcome variable, such as whether a subject ordered wild meat. The subscript i denotes the subject, j represents the table where the subject participated in the experiment, and t denotes the date of participation (not the date at which they ordered a dish in their restaurant of choice). Ti indicates whether the subject was in the treatment group that watched the wild meat demand reduction video (Ti = 1) or in the control group (Ti = 0). Xi is a matrix of eight individual subject characteristics, such as age, gender, education level, salary and business ownership status. κj and ζt are the “fixed effects”, or individual dummy control variables for each table (κj ) and individual dummy control variables for each date (ζt ). Including these fixed effects obviates the need for an intercept term, as it would be collinear with either κj or ζt . The variable ϵijt is the error term, which represents factors that affect the outcome variable but which are not otherwise accounted for in the equation. The coefficient of interest is β , which measures the effect of treatment assignment on the outcome variable. The standard errors of all regression coefficients are clustered at the subject level to match the level at which treatment is assigned (Abadie et al., 2023). This method accounts for potential differences in the variability of responses across subjects. Our clustered standard errors are also “heteroskedasticity-robust”. We do not solely rely on p-values in interpreting our results; we also assess the magnitudes of the treatment effects, and we compare their consistency and inter- pretive validity across the different parts of our experiments (McShane et al., 2019; Wasserstein & Lazar, 2016). When we do assess statistical significance, we use the conventional p-value threshold of 0.05. The inclusion of individual-level controls and fixed effects is aimed at increasing the pre- 10 cision of the estimated impact of the treatment, β . They are not necessary for estimating the magnitude of the treatment effect because random assignment ensures that any differences in outcomes between the treatment and control group can be attributed to the treatment video (Ru- bin, 1974). In Equation 1, Xi includes an indicator for whether the subject reports usually eating wild meat at formal or informal restaurants, the number of days the subject has eaten wild meat in the last 30 days, the age of the subject in years, an indicator variable for whether the subject is male, an indicator for whether the subject has attained university (post-secondary) education, the total number of years of education the subject has, an indicator variable for whether the subject was employed for pay in the previous 7 days, and an indicator variable for whether the subject worked for pay as the owner of a business in the previous 7 days. For any missing values in these control variables (except for indicator variables), we impute the value with the mean among all non-missing values. For example, if age is missing for 44 subjects, we assume their age to be the mean age among the 500 subjects with recorded ages. We chose these eight control variables to capture key behavioral, demographic, and economic factors that may influence wild meat consumption, while maintaining survey brevity (Chausson et al., 2019; WildAid, 2021). Other variables, such as ethnicity or duration of residency in Kinshasa, may also predict wild meat consumption. The eight explanatory variables we chose can be grouped conceptually into three categories of wild meat consumption predictors. We expected the first two, whether subjects usually eat wild meat at restaurants and the number of days subjects have eaten wild meat in the previous 30 days, to positively correlate with subjects using their coupon to order wild meat. We chose the next four—age, sex, whether the subject attained university education, and years of education— as capturing the most essential demographic information, including characteristics that predict wild meat consumption. For example, in Pointe Noire, Republic of the Congo, younger people eat less wild meat, and males eat more wild meat (Chausson et al., 2019). Finally, we expected 11 the last two variables—in paid employment and business owner—to be positively correlated with subjects using their coupon, because we thought richer subjects would be more able to pay the portion of the restaurant meal not covered by the coupon. 2.2.2 Analysis of Selection of Wild Meat-Intensive Restaurants The decision of what dish to order at a restaurant focuses on the quantity of wild meat consumed, ignoring the preceding choice of which restaurant to visit. Our treatment video, and demand- side interventions more broadly, might influence not just the selections made within an eating environment but also the choice of the eating environment itself. With the four restaurants in our study offering varying numbers of wild meat dishes and these dishes constituting different proportions of their total menus, we have the opportunity to assess whether subjects exposed to the treatment are more likely to choose coupons to restaurants that feature less wild meat. For this analysis, we categorize restaurants based on the proportion of wild meat dishes featured on their sample menus, as shown to subjects during the demand-side experiment. Our pre-specified definition identifies “wild meat-intensive” restaurants as those with the highest proportion of wild meat dishes (23% and 21%, compared to 7% and 9%). Alongside the origi- nal control variables in Equation 1, we include as controls the distance from the subject’s table to the chosen restaurant and the average price of dishes on the restaurant’s sample menu. We included these variables to enhance the precision of our estimate of the treatment effect by con- trolling for factors like convenience and cost, which might affect a subject’s choice of restaurant independently of treatment assignment. 2.2.3 Analysis of Attitudes Towards Wild Meat To understand the mechanisms behind the potential changes in wild meat consumption, we in- vestigate how the treatment video affects subjects’ views on various attributes of wild meat. 12 We pre-specified this analysis to discern whether changes in wild meat consumption are due to shifts in perception caused by the treatment video. After watching the video and choosing their restaurant, subjects answered eight questions measuring different aspects of their views of bushmeat (Appendix C). To avoid priming the respondents or the enumerators into choosing the first response options, we randomly varied the order with which we displayed the response options (from completely agree to completely disagree vs. from completely disagree to com- pletely agree). We aggregate responses to these questions into a single index, coding responses to indicate a negative view of wild meat. This coding schema allows us to test whether the treatment video worsened attitudes toward wild meat. For the first seven attributes, we code a response showing disagreement as 1 (indicating a negative view), while coding agreement or neutrality as 0. For the statement about environmental pride, we code agreement as 1. We sum the coded responses for each subject, creating a score ranging from 0 to 8. Then we standardize this score across all subjects by subtracting the mean and dividing by the standard deviation, resulting in a standardized response score. Finally, we re-estimate Equation 1 with this “Unfavorable” standardized score as the dependent variable. 2.2.4 Analysis of Social Desirability Bias During our demand-side experiment, enumerators emphasized to subjects that they should order the dish they most desired at the restaurant, regardless of whether it was wild meat or not. Despite these instructions, there remains a possibility that treated subjects might order less wild meat due to social desirability bias—subjects doing what they believe enumerators want them to do—rather than subjects being genuinely persuaded by the treatment video. To evaluate this potential bias, also known as experimenter demand effects, we replicate an established method to assess its impacts on our results (Dhar et al., 2022). 13 This method involves a 13-question module designed to measure social desirability (Crowne & Marlowe, 1960; Reynolds, 1982). Given each statement, subjects answered whether they completely agree, partially agree, neither agree nor disagree, partially disagree, or completely disagree. To avoid priming the respondents or the enumerators into choosing the first response options, we randomly varied the order with which we displayed the response options (Ap- pendix C). We code the response to each statement as 1 if the subject gives a socially desirable answer. For example, if a subject completely disagrees or partially disagrees with the statement “I sometimes feel resentful when I don’t get my way”, we code their response as 1 (and code the response as 0 if they neither disagree nor agree, partially agree, or completely agree). We sum the coded responses over statements, so that subjects have a social desirability score of between 0 and 13. We standardize the score by subtracting the mean score across all subjects, and then dividing by the standard deviation of the score across all subjects. To investigate social desirability bias’s influence on our wild meat consumption results, we replace the eight individual characteristics controls in Equation 1 with two variables: the standardized social desirability score (StdSoci ), and its interaction with the treatment indicator (Ti × StdSoci ). The dependent variable is an indicator of whether a subject ordered wild meat: 1{OrderedW ildM eat}ijt = βTi + σ1 StdSoci + σ2 Ti × StdSoci + κj + ζt + ϵijt . (2) In this equation σ2 is the coefficient of interest, indicating the potential interaction between treatment and social desirability. A negative σ2 would imply that subjects in the treatment group less likely to order wild meat also exhibited higher social desirability. 14 2.3 Design of Supply-Side Experiment In November 2023, alongside our demand-side experiment, we conducted a supply-side experi- ment at the same four restaurants to explore how changes in the price of Moambe Chicken affect wild meat consumption. Moambe Chicken, a dish specific to Congolese cuisine, is popular in the region and may fill a similar cultural role as wild meat. Our field team, primarily composed of Kinshasa residents, advised us that Moambe Chicken was the most likely substitute dish for wild meat at restaurants in Kinshasa. On randomly selected days, the price of Moambe Chicken was reduced by $2 (5,000 Congolese Francs). On these days, we compensated restaurants $2 for every plate of Moambe Chicken they sold (to all customers, not only those participating in our demand-side experiment). The goal of this experiment was to determine whether making a non-wild meat option more affordable would reduce customers’ decisions to order wild meat. This relationship between wild meat consumption and the price of alternatives is the rationale for supply-side alternative protein programs, such as promoting poultry production, where the aim is to reduce wild meat consumption by increasing the availability and affordability of sub- stitutes (Foerster et al., 2012; Moro et al., 2015). The restaurants provided daily sales data, which included all customers, not only those participating in our demand-side experiment. These data encompassed the number of plates sold in three categories: wild meat, Moambe Chicken, and all other dishes, along with the corresponding revenue for each category. For the purpose of our analysis, “wild meat dishes” referred to all types of wild meat combined into a single category. For instance, if a restaurant sold 2 dishes of antelope and 2 dishes of monkey on a day, it was recorded as 4 dishes of wild meat sold that day. During the supply-side experiment, we gathered data across 68 restaurant-day observations, comprising 17 days for each of the four participating restaurants. Among these, 11 restaurant- days were randomly chosen to lower the price of Moambe Chicken, creating the treatment 15 group, while the other 57 served as the control group with regular pricing. This randomiza- tion enabled a comparison of dish sales between days with reduced prices (treatment group) and days when prices remained unchanged (control group). The smaller number of treatment observations compared to control observations reflects our budget constraints, as we only paid restaurants on days they were treated. Importantly, the treatment assignment in the demand-side and supply-side experiments was independently randomized, ensuring that neither experiment would influence the results of the other on average. This independent randomization means that while subjects in the demand-side experiment may visit a restaurant during the supply-side experiment, these interactions would be randomly distributed and thus not bias the results of either experiment. 2.4 Analysis of Supply-Side Experiment The key assumption in our analysis is that the only way the experiment affected wild meat and Moambe Chicken sales was through the experimentally induced change in Moambe Chicken price. This assumption is likely valid given the randomization of the experiment and its singular focus on subsidizing treated restaurants to reduce Moambe Chicken price by $2 on randomly selected days. Our analysis begins with estimating the effect of the treatment on Moambe Chicken’s price by ordinary least squares regression: log (Pit ) = γ1 Tit + δ1 Qi0 + δ2 P ricei0 + ϵit . (3) Here, log (Pit ) is the log price of Moambe Chicken in restaurant i on day t, with Tit indicating treatment status (Tit equals 1 if treated and equals 0 otherwise). The control variable Qi0 is the number of wild meat dishes sold in the week preceding the supply-side experiment, and 16 P ricei0 is the average price over all dishes on the restaurant’s sample menu, as used in Section 2.2.2. These control variables were pre-specified to improve the precision of our treatment effect estimates (Cisse et al., 2023; McKenzie, 2012). Second, we estimate the treatment’s effect on sales volumes by ordinary least squares re- gression: log (Qit ) = γ2 Tit + δ3 Qi0 + δ4 P ricei0 + ϵit (4) where log (Qit ) is the log number of Moambe Chicken dishes sold in restaurant i on day t or the log number of wild meat dishes sold. We cluster standard errors of all coefficients at the restaurant-day level to match the level of treatment assignment (Abadie et al., 2023). Since the dependent variables in both equations are in the logarithmic form, the treatment coefficients γ1 and γ2 can be interpreted in percentage terms by applying the transformation ecoefficient − 1. Given the logarithmic dependent variables, restaurant-day observations with zero sales of a particular dish were necessarily omitted from that dish’s regression. The ratio of γ2 to γ1 represents the “elasticity” of Moambe Chicken or wild meat sales with respect to Moambe Chicken price. We estimate these elasticities using the feols() function e, 2018). In the elasticity regres- in R, which provides accurate standard error estimates (Berg´ sions, both the dependent and independent variables are in logarithmic form. Consequently, the coefficient on log Moambe Chicken price can be interpreted as the percentage change in the dependent variable (Moambe Chicken sales or wild meat sales) for a 1% change in the indepen- dent variable (price). Elasticities measure the sensitivity of Moambe Chicken sales or wild meat sales to the price of Moambe Chicken (Perloff, 2023). The interpretation of elasticities differs from that of γ1 or γ2 because the independent variable Tit is linear, rather than logarithmic. 17 2.5 Ethics Statement Our study was reviewed and approved by the DRC’s National Institute of Statistics (#0340/INS/DG/fau/2023) and the University of California, Berkeley’s Institutional Review Board (Protocol ID: 2023-05- 16343). The National Institute of Statistics validated the methodology and tools used to carry out our study and granted us the necessary statistical approval for our survey. Prior to initiating any part of the research, informed consent was obtained from all subjects as well as from the participating restaurants. We anonymized all subject data. We implemented distinct randomization procedures for our demand-side and supply-side experiments. In the demand-side experiment, we utilized SurveyCTO’s programming features to assign subjects to treatment or control groups based on a randomly generated number. This process occurred automatically upon survey initiation, with the assignment concealed from enu- merators, subjects, and restaurants. For the supply-side experiment, we used R version 4.1.2 to randomly select one or two restaurants for treatment on specific days from a vector containing all four participating restaurants (R Core Team, 2024). Each evening, our field team informed restaurants of their next day’s treatment status. 3 Results 3.1 Demographic Characteristics of Subjects in Demand-Side Experiment We begin by comparing the characteristics of treatment and control subjects in the demand-side experiment. Each row of Table 1 considers one of the eight individual subject characteristics that are included in the matrix Xi in Equation 1. The first three columns respectively display the mean value of a characteristic among subjects in the control group, the mean difference be- tween the treatment and control groups, and the standard error of this difference. All observed 18 differences are small and not statistically significant, indicating that our randomization proce- dure successfully created comparable groups. This balance is crucial as it ensures that observed differences in outcomes, such as whether a subject orders wild meat, can be attributed to the treatment rather than to pre-existing disparities between groups. In our sample of 544 subjects, 42% typically consume wild meat at formal or informal restaurants, the average number of days wild meat was consumed in the past 30 days is 2.5, the average age is 33, 76% are male, 30% are university graduates, the average number of years of education is 11.7, 43% earned a salary in the past week, and 24% earned income as a business owner in the same period. These mean values differ slightly from the control means (Column 1 of Table 1) because they are average values among all subjects (across both the treatment and control groups). Dependent Variable Control Mean Treatment Difference Standard Error P-Value N Restaurant Habit 0.441 -0.034 (0.042) 0.429 544 Wild Meat Days 2.465 -0.008 (0.223) 0.973 544 Age 33.210 -1.000 (0.931) 0.283 544 Male 0.752 0.020 (0.037) 0.593 544 University Graduate 0.325 -0.050 (0.039) 0.204 544 Years of Education 11.598 0.135 (0.271) 0.620 544 Salary Earner 0.441 -0.018 (0.043) 0.671 544 Business Owner 0.241 -0.001 (0.037) 0.979 544 Table 1: Individual Subject Characteristics in the Demand-Side Experiment by Treatment Status. Control Mean is the mean value of each variable for the control group. Treatment Difference is the mean value in the treatment group minus the Control Mean. Standard Error is the standard error of the Treatment Difference. The Standard Error, clustered at the subject level, is derived from an ordinary least squares regression of each variable on an intercept and a treatment indicator. The P-Value gives the Type 1 error rate for the Treatment Difference. Definitions of all variables are provided in Section 2.2. 19 3.2 Results of Demand-Side Experiment 3.2.1 Effects on Wild Meat Consumption and Coupon Usage We now assess the results of the demand-side experiment. The first question we address is the impact of the treatment, the video aimed at discouraging wild meat consumption, on partic- ipants’ dining choices. Specifically, we examine whether exposure to the video affected the likelihood of subjects ordering wild meat. The results from estimating Equation 1, with the dependent variable being whether or not the subject ordered wild meat (1 for yes, 0 for no), are presented in Column 2 of Table 2. The main result of our demand-side experiment is a 31% decrease in the probability of subjects in the treatment group ordering wild meat compared to those in the control group (3.1% vs. 4.5%). We calculate this 31% reduction by dividing the Column 2 treatment coefficient (- 0.014) by the mean of the dependent variable in the control group (0.045, as shown by the Intercept coefficient in Column 1). This result is not statistically significant according to the conventional p-value threshold of 0.05 (Figure S1a). In addition to examining wild meat ordering behavior, we also investigate whether the treat- ment influenced the overall likelihood of subjects using their coupon to order a dish (Column 4 of Table 2). Differential coupon usage between the treatment and control groups could im- ply a fundamental difference in the composition of the two groups, potentially challenging the validity of our comparison. However, this concern is alleviated by the finding that the differ- ence in coupon usage is minimal, at only 0.6%, and not statistically significant. The similarity in coupon usage, with approximately 59% of subjects in both groups redeeming their coupons for a meal, supports our assumption that the only factor differentiating wild meat consumption between the treatment and control groups is the specific video each group watched. Finally, including individual controls and fixed effects does not substantially change the 20 Dependent Variable: Ordered Wild Meat Used Coupon (1) (2) (3) (4) Treatment -0.011 -0.014 0.009 -0.006 (0.017) (0.017) (0.042) (0.036) Restaurant Habit 0.023 0.059 (0.017) (0.037) Wild Meat Days -0.003 -0.002 (0.002) (0.006) Age 0.000 0.004 (0.001) (0.002) Male -0.015 0.020 (0.020) (0.042) University Graduate -0.036 -0.061 (0.027) (0.058) Years of Education 0.003 0.008 (0.006) (0.009) Salary Earner -0.008 0.061 (0.018) (0.037) Business Owner -0.005 0.056 (0.029) (0.051) Intercept 0.045 0.584 (0.012) (0.029) Date and Table Fixed Effects No Yes No Yes N 544 544 544 544 Percent Change -23.26 -31.29 1.56 -1.07 Table 2: Effect of Treatment on Probability of Ordering Wild Meat (Columns 1 and 2) and on Probability of Using Coupon (Columns 3 and 4). The Intercept equals 1 for all observations (Columns 1 and 3). Due to the inclusion of the Treatment variable, which equals 1 only for subjects in the treatment group, the coefficient on the Intercept is mathematically equivalent to the mean of the dependent variable in the control group. Standard errors are clustered at the subject level. estimated treatment effects (comparing results in Column 1 to Column 2, and in Column 3 to Column 4). The results are similar whether we use a simple regression model with only an in- tercept and a treatment indicator (Columns 1 and 3) or a more complex model that controls for 21 individual characteristics and fixed effects (Columns 2 and 4). This consistency across different model specifications provides further reassurance about the success of our randomization pro- cedure. In our pre-analysis plan, we selected individual characteristics and fixed effects that we thought would increase the precision of our estimates. However, this approach did not yield the anticipated improvement; for example, the standard error of the treatment effect on wild meat consumption is 0.017 in both wild meat consumption specifications (Columns 1 and 2). While individual characteristics also generally do not emerge as significant predictors of wild meat consumption or coupon usage, some variables predict these behaviors in ways that align with our initial expectations. For instance, subjects who typically eat wild meat at formal or informal restaurants are more likely to order wild meat (Column 2) and to use their coupon (Column 4). Likewise, subjects who earn a salary or business income are more likely to use their coupon, as they are likely to have greater financial capacity to pay the difference between the price of their meal and the coupon value (Column 4). 3.2.2 Treatment Effect on Selection of Wild Meat-Intensive Restaurants We find little evidence that treatment reduced subjects’ selection of wild meat-intensive restau- rants. The treatment coefficient is small in magnitude and not statistically different from zero in our pre-specified estimating equation (Column 2 of Table S1). The most likely explanation for this null result is that 74% of subjects chose the restaurant geographically nearest to them, lim- iting the scope for variation in restaurant selection along the dimension of wild meat-intensity. 3.2.3 Exploring Mechanisms: Impact of Treatment on Attitudes Towards Wild Meat Most subjects strongly agree or somewhat agree that wild meat is tasty, sustainable, healthy, fresh, cool, legal, and connects them to their place of origin. (Figures 2(a)-(g)). About three- quarters of subjects also report being very proud or somewhat proud of the DRC’s environment 22 (Figure 2(h)). Visually, the treatment video appears to slightly reduce perceived sustainability of wild meat, the extent to which wild meat connects subjects to their place of origin, and pride in the DRC’s environment. However, we do not find evidence supporting a shift in attitudes in our pre-specified sta- tistical analysis. The intervention did not significantly alter subjects’ overall perceptions of wild meat (first row of Table S2). We also explore treatment effects on each of the eight sep- arate attitudinal questions (second through ninth rows of Table S2). The intervention did not significantly shift perceptions on any specific dimension related to wild meat. 23 (a) Taste (e) Coolness Strongly Agree Strongly Agree Somewhat Agree Somewhat Agree Neither Neither Somewhat Disagree C Somewhat Disagree C T T Strongly Disagree Strongly Disagree 0% 15% 30% 45% 0% 15% 30% 45% (b) Sustainability (f) Legality Strongly Agree Strongly Agree Somewhat Agree Somewhat Agree Neither Neither Somewhat Disagree C Somewhat Disagree C T T Strongly Disagree Strongly Disagree 0% 15% 30% 45% 0% 15% 30% 45% (c) Healthiness (g) Origin Strongly Agree Strongly Agree Somewhat Agree Somewhat Agree Neither Neither Somewhat Disagree C Somewhat Disagree C T T Strongly Disagree Strongly Disagree 0% 15% 30% 45% 0% 15% 30% 45% (d) Freshness (h) Pride Strongly Agree Very Proud Somewhat Agree Somewhat Proud Neither Neither Somewhat Disagree C Somewhat Ashamed C T T Strongly Disagree Very Ashamed 0% 15% 30% 45% 0% 15% 30% 45% Figure 2: Subjects’ Attitudes Towards Bushmeat. Bars represent the percentage of control group (orange) and treatment group (purple) subjects who chose a response option. See Ap- pendix C for question wording. 24 3.2.4 Assessing Social Desirability Bias Subjects in the treatment group less likely to order wild meat also exhibit higher social desir- ability (third row of Table S3). However, this relationship is not statistically significant (p-value is 0.193). 3.3 Results of Supply-Side Experiment Our supply-side experiment focused on the impact of food price on consumer choices. Specifi- cally, we examined how reducing the price of Moambe Chicken, a common alternative to wild meat, affected wild meat and non-wild meat sales. We begin our analysis of the results by validating the supply-side experiment. While the price of Moambe Chicken may or may not directly affect wild meat demand, it should cer- tainly affect Moambe Chicken demand. Indeed, the experiment significantly reduced Moambe Chicken’s price by approximately 30% (Column 1 of Table 3) and correspondingly increased Moambe Chicken sales by about 35% (Column 2). Dividing the treatment effect on Moambe Chicken sales (Column 2) by the treatment effect on its price (Column 1) reveals the “elastic- ity” of Moambe Chicken sales with respect to its price (Section 2.4). The estimated elasticity of -0.837 indicates that a 1% decrease in price leads to approximately a 0.837% increase in Moambe Chicken sales (Column 3). The central finding of our supply-side experiment is the result that reducing the price of Moambe Chicken reduces wild meat consumption. Column 4 shows the effect of the Moambe Chicken price reduction treatment on wild meat sales. The coefficient indicates that the treat- ment reduced restaurants’ total wild meat sales by approximately 26%. In other words, our experiment likely caused some patrons who would have ordered wild meat to order Moambe Chicken instead, though the effect is not statistically significant. 25 Moambe Chicken Wild Meat Dependent Variable: Dependent Variable: Log Price Log Dishes Log Dishes Log Dishes Log Dishes (1) (2) (3) (4) (5) Price Reduction Treatment -0.362 0.303 -0.302 (0.041) (0.119) (0.201) Log Chicken Price -0.837 0.910 (0.341) (0.573) Baseline Wild Meat 0.011 0.028 0.037 0.019 0.008 (0.000) (0.004) (0.006) (0.005) (0.009) Baseline Average Price -0.002 -0.158 -0.159 -0.107 -0.104 (0.002) (0.017) (0.017) (0.021) (0.021) Intercept 9.460 4.415 12.340 3.379 -5.264 (0.028) (0.207) (3.297) (0.268) (5.436) N 67 65 65 54 53 Table 3: Elasticities of Moambe Chicken and Wild Meat Consumption with Respect to Moambe Chicken Price. The unit of observation is a restaurant-day. Effect of price reduction treatment on log(Moambe Chicken price) (Column 1), on log(Moambe Chicken dishes sold) (Column 2), and on log(wild meat dishes sold) (Column 4). Elasticity of Moambe Chicken dishes sold (Column 3) and wild meat dishes sold (Column 5) with respect to Moambe Chicken price. Due to the logarithmic nature of the dependent variables, observations with zero sales of a particular dish were omitted, resulting in slight variations in the numbers of observations across columns. Baseline Wild Meat is the number of plates of wild meat sold by the restaurant the week before the supply-side experiment began. Baseline Average Price is the mean price, in thousand Congolese Francs, of all dishes on the sample menu shown to participants in the demand-side experiment. Standard errors are clustered at the restuarant-day level. We also obtain the elasticity of wild meat sales with respect to Moambe Chicken price (Column 5). This positive elasticity value, approximately 0.91, implies that a 1% decrease in Moambe Chicken price causes a 0.91% reduction in wild meat consumption. However, our estimate does not reach the conventional level of statistical significance (Figure S1b). 26 4 Discussion By conducting two randomized controlled trials in Kinshasa, DRC, our study provides the first experimental evidence regarding the effectiveness of interventions designed to reduce the con- sumption of wild meat in urban restaurants. One experiment targeted demand for wild meat by presenting a subset of participants with a specifically designed treatment video, while the other experiment targeted the supply-side by reducing the price of an alternative domesticated animal protein dish (Moambe Chicken). Our findings represent encouraging results for the effective- ness of both demand- and supply-side interventions, such as social marketing and reducing the price of alternative meats, respectively. 4.1 Wild Meat Consumption in Kinshasa We found that study participants reported consuming wild meat on average 2.5 days in the past month, and across the treatment and control groups, 4% of subjects ordered wild meat in the experiment. Although per capita consumption of wild meat is relatively infrequent, the large and growing human population means urban demand for wild meat may be reducing wildlife populations in the DRC (Batumike et al., 2021; van Vliet et al., 2017). The relatively infrequent per capita consumption is likely driven by a number of factors including: a) availability of wild and domesticated sources of animal protein; b) fewer offerings of wild meat on menus in comparison to other meats and fish — in the four participating restaurants wild meat based dishes were 7%, 9%, 21%, and 23% of all dishes; c) the price of wild meat in comparison to other options — in two of our participating restaurants, wild meat dishes were 38% and 39% more expensive on average than other options. Together these factors highlight the complex drivers of wild meat consumption in urban areas, which could be investigated simultaneously in future studies and intervention evaluations. 27 Separate from our experiments, the citywide Yoka Pimbo campaign between March 2021 ere de l’Environnement and March 2024 aimed to reduce bushmeat demand in Kinshasa (Minist` eveloppement Durable, 2024). This campaign does not bias our treatment effects because et D´ we randomized treatment assignment in our experiments, which ensures that both treatment and control groups had similar average exposure to the Yoka Pimbo campaign. 4.2 Demand-Side Experiment Whilst not statistically significant, our analysis shows that compared to the control group, mem- bers of the treatment group had a 31% lower probability of ordering wild meat. This is a promis- ing result for the potential of demand-reduction interventions that involve video message dis- semination. However, further evaluation is needed to have certainty in their effectiveness. Other wild meat demand reduction interventions had mixed success. A social marketing intervention involving an information campaign and community engagement in Brazil was found to reduce wild meat consumption by 62% (Chaves et al., 2018), while a radio entertainment-education ıssimo et al., 2018). intervention in Tanzania found no significant demand reduction (Ver´ Despite the challenges with conducting an experiment on wild meat consumption in the DRC, we successfully recruited 544 wild meat consumers as participants. A larger sample size, which we recommend for future evaluations, would have yielded more statistical power to bet- ter interpret our results. For example, we found no statistically significant interaction between treatment and social desirability, although the direction and magnitude of the estimate sug- gests that participants in the treatment group who were less likely to order wild meat exhibited higher scores on our social desirability metric. These findings suggest that future evaluations of demand-reduction programmes should measure and account for such potential biases. Wild meat was positively viewed among study participants in terms of its taste, sustain- ability, healthiness, freshness, “coolness”, legality, and to a slightly lesser extent, the ability to 28 connect them to their place of origin. Most participants also expressed pride in the DRC’s natu- ral environment. While our survey measured attitudes rather than values, our results suggest that wild meat consumers may hold diverse values regarding wild meat, including instrumental (e.g., source of food) and relational values (e.g., cultural identity, sense of place) (Chan et al., 2018; Pascual et al., 2017). Values-centred approaches to achieving sustainability may be more ethi- cal and effective (Pascual et al., 2023). The message used in the treatment video, highlighting the decline of wildlife in the forests of Congo and the desirability of cooking traditional recipes without wild meat, is likely to be appropriate in Kinshasa. The video speaks to relational values of wild meat consumption (culture, identity), and the strong levels of pride we observed in the participants towards the DRC’s natural environment. These elements relate most closely to our questions regarding bushmeat’s sustainability, the extent to which it connects subjects to their places of origin, and their pride in the DRC’s environment. While the treatment video slightly shifts responses to these questions compared to responses in the control group, these changes are not statistically significant in our pre-specified analyses. Repeated exposure to messages connecting existing norms (e.g. pride in wildlife) with reduced wild meat consumption could be effective in changing behavior (MacFarlane et al., 2022; Wakefield et al., 2010). 4.3 Supply-Side Experiment Our supply-side intervention tested the theory that prices of alternative meats influence the consumption of wild meat. The results of our experiment showed that reducing the price of Moambe Chicken dishes by $2 (5,000 Congolese francs) reduced the sales of wild meat by 26%. This suggests that Moambe Chicken is a substitute for wild meat in line with theoretical expectations. However, our estimates do not reach the conventional level of statistical signif- icance, likely due to our small sample size. Few such interventions have been experimentally tested for effectiveness (Ingram et al., 2021; Willis et al., 2022), highlighting the significant 29 contribution of our study to the literature. One exception is an experiment that found that pro- viding coupons for chicken in Brazil increased chicken consumption, but did not decrease wild meat consumption (Chaves et al., 2018). Our results therefore cautiously support the role of supply-side interventions in efforts to reduce wild meat consumption, particularly through interventions that increase the affordability of wild meat alternatives. Government agencies (such as environment and agriculture min- istries), international aid organizations, and conservation NGOs could implement and evaluate supply-side interventions. However, reducing urban demand for wild meat through supply-side interventions may have implications for the incomes of rural hunters and traders reliant on wild meat for their livelihoods, which should be considered during intervention implementation. An additional trade-off might also exist between the increased production of domestic livestock and subsequent land needs, and intact habitats. Whether wild meat is elastic to its own price or to the price of other potential substitutes is complex, and depends on the site context (Rentsch & Damon, 2013; Walelign et al., 2019; Wilkie et al., 2005). For example, Walelign et al. (2019) found that wild meat demand in rural Tanzania was more elastic to its own price when the substitute option was beef, but less so when the substitute option was fish or goat — although the latter depended on whether socioeconomic covariates were controlled for in the analyses. Socio-economic household determinants (e.g., household income) and cultural factors (e.g., ethnic group) have been shown to be important factors in mediating wild meat demand and price responsiveness (Walelign et al., 2019). Impor- tantly, none of the previous studies experimentally changed the price of alternatives in a restau- rant setting to investigate the effect on wild meat consumption or sales. The setting is likely important because it influences the types of people attending so may include a different subset of society based on food consumption norms. For example, wild meat may hypothetically be consumed more frequently at home yet consumed outside the home on certain occasions or with 30 certain groups of people. In our study in Kinshasa, the participating restaurants were all formal restaurants where patrons sit at tables inside permanent buildings and are served by waiters; results could differ in informal restaurants. Foods and eating are often connected with identity, rituals, symbols, and belief systems (Mintz & Du Bois, 2002), with wild meat in particular holding diverse roles and values in var- ıguez-R´ ious cultures (Rodr´ ıa-P´ ıos & Garc´ aez, 2018; Rose, 2001; van Vliet & Mbazza, 2011). While Moambe Chicken, a national dish of the DRC, shares some attributes with wild meat, it may not fill the identical social, cultural, or relational function. Therefore, attempting to change potentially culturally important food consumption habits poses ethical questions. Fur- ther research is needed to better understand the prevalence and importance of social and cultural functions that wild meat may have, to inform the design of ethical strategies and interventions, and to ensure sustainable levels of wild meat consumption. 4.4 Limitations As is the case for all studies, ours has some limitations. First, the statistical insignificance of our results means we cannot definitively ascertain whether our interventions reduced wild meat consumption. Second, our demand-side intervention considered the effect of a single exposure of the treatment video on a consumption decision temporally close to the exposure (within two weeks). Our results that a single exposure could result in behaviour change are promising, but it is not clear how long this change would last. Studies suggest that multiple message exposures might increase intervention effectiveness (Montoya et al., 2017), so we suggest future evaluations assess the influence of message frequency, and effectiveness over time. In SMS- based behaviour change interventions, frequency of message exposure matched the behaviour frequency (Fjeldsoe et al., 2009). For TV-based interventions, we suggest demand reduction messages be timed to coincide with typical wild meat consumption patterns. Third, the video 31 was developed for a campaign in Pointe Noire, Republic of the Congo, yet we experimentally tested the video in restaurants in Kinshasa, DRC. Whilst there may be socio-cultural differences in how wild meat is consumed in these cities, the content of the video focused on the generic topics of wild meat consumption leading to empty forests in the Congo, and cooking traditional Congolese dishes without wild meat. The video was also mostly in French, the official language of both countries. Therefore, we suspect the video to be relevant in Kinshasa. 4.5 Conclusions Our results suggest that both demand-side and supply-side interventions can reduce the con- sumption of wild meat dishes in formal restaurants in Kinshasa. While not statistically signifi- cant, the magnitude and direction of our estimates suggest that these interventions can succeed in reducing consumption. Approaches that include both intervention types may be most suc- cessful where wild meat consumption is unsustainable. We recommend further intervention testing with long-term monitoring and evaluation to definitively ascertain effectiveness. We encourage those involved in such interventions to a) publish their results openly, whether suc- cessful or otherwise, so others can learn, b) fully engage with any ethical concerns around intervention design and implementation, and c) randomly assign the intervention to some sub- jects (or units of study) but not to others. Random assignment allows estimation of the causal effects of the intervention, which are most useful for understanding intervention effectiveness and making recommendations for policy and action. References Abadie, A., Athey, S., Imbens, G. W., & Wooldridge, J. M. (2023). When should you adjust standard errors for clustering? The Quarterly Journal of Economics, 138(1), 1–35. 32 Batumike, R., Imani, G., Urom, C., & Cuni-Sanchez, A. (2021). Bushmeat hunting around Lomami National Park, Democratic Republic of the Congo. Oryx, 55(3), 421–431. e, L. (2018). Efficient estimation of maximum likelihood models with multiple fixed- Berg´ effects: The R package FENmlm. CREA Discussion Papers, (13). Carignano Torres, P., Morsello, C., & Parry, L. (2022). Rural–urban mobility influences wild- meat access and consumption in the Brazilian Amazon. Oryx, 56, 864–876. ıaz, S., G´ Chan, K. M., Balvanera, P., Benessaiah, K., Chapman, M., D´ omez-Baggethun, E., Gould, R., Hannahs, N., Jax, K., Klain, S., et al. (2016). Why protect nature? Rethinking values and the environment. Proceedings of the National Academy of Sciences, 113(6), 1462–1465. Chan, K. M., Gould, R. K., & Pascual, U. (2018). Editorial overview: Relational values: What are they, and what’s the fuss about? Current Opinion in Environmental Sustainability, 35, A1–A7. Chausson, A. M., Rowcliffe, J. M., Escouflaire, L., Wieland, M., & Wright, J. H. (2019). Un- derstanding the Sociocultural Drivers of Urban Bushmeat Consumption for Behavior Change Interventions in Pointe Noire, Republic of Congo. Human Ecology, 47 (2), 179– 191. Chaves, W. A., Valle, D. R., Monroe, M. C., Wilkie, D. S., Sieving, K. E., & Sadowsky, B. (2018). Changing Wild Meat Consumption: An Experiment in the Central Amazon, Brazil. Conservation Letters, 11(2), 1–10. Cisse, A., Englander, G., & Ingram, D. (2023). Conserving wildlife by reducing demand: A restaurant experiment in Kinshasa. AEA RCT Registry. https : / / doi . org / 10 . 1257 / rct . 12275-1.0 Cisse, A., Englander, G., & Ingram, D. J. (2024). Reproducibility package for Conserving Wildlife through Demand Reduction and Supply Alternatives: Two Experiments in Restau- 33 rants in Kinshasa [The World Bank Reproducible Research Repository]. https://doi.org/ 10.60572/g9ks-a898 Coad, L., Fa, J. E., Abernethy, K., Van Vliet, N., Santamaria, C., Wilkie, D., El Bizri, H. R., Ingram, D. J., Cawthorn, D.-M., & Nasi, R. (2019). Towards a sustainable, participatory and inclusive wild meat sector. CIFOR. Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psy- chopathology. Journal of Consulting Psychology, 24(4), 349. Dhar, D., Jain, T., & Jayachandran, S. (2022). Reshaping adolescents’ gender attitudes: Evi- dence from a school-based experiment in India. American Economic Review, 112(3), 899–927. e, Edderai, D., & Dame, M. (2006). A census of the commercial bushmeat market in Yaound´ Cameroon. Oryx, 40(4), 472–475. Fa, J. E., Wright, J. H., Funk, S. M., M´ arquez, A. L., Olivero, J., Farf´ ´ Guio, F., Mayet, an, M. A., L., Malekani, D., Louzolo, C. H., et al. (2019). Mapping the availability of bushmeat for consumption in Central African cities. Environmental Research Letters, 14(9), 094002. Fjeldsoe, B. S., Marshall, A. L., & Miller, Y. D. (2009). Behavior change interventions delivered by mobile telephone short-message service. American Journal of Preventive Medicine, 36(2), 165–173. Foerster, S., Wilkie, D. S., Morelli, G. A., Demmer, J., Starkey, M., Telfer, P., Steil, M., & Lewbel, A. (2012). Correlates of bushmeat hunting among remote rural households in Gabon, Central Africa. Conservation Biology, 26(2), 335–344. Food and Agriculture Organization, Centre for International Forestry Research, French Agricul- tural Research Centre for International Development, & Wildlife Conservation Society. (2019). Sustainable Wildlife Management Programme [Accessed: April 4, 2024]. http: //www.fao.org/3/ca5100en/ca5100en.pdf 34 Grantham, H. S., Duncan, A., Evans, T. D., Jones, K. R., Beyer, H. L., Schuster, R., Walston, J., Ray, J. C., Robinson, J. G., Callow, M., Clements, T., Costa, H. M., DeGemmis, A., Elsen, P. R., Ervin, J., Franco, P., Goldman, E., Goetz, S., Hansen, A., . . . Watson, J. E. M. (2020). Anthropogenic modification of forests means only 40% of remaining forests have high ecosystem integrity. Nature Communications, 11, 5978. Ingram, D. J. (2020). Wild Meat in Changing Times. Journal of Ethnobiology, 40(2), 117–130. ıtez-L´ Ingram, D. J., Coad, L., Milner-Gulland, E., Parry, L., Wilkie, D., Bakarr, M. I., Ben´ opez, A., Bennett, E. L., Bodmer, R., Cowlishaw, G., et al. (2021). Wild meat is still on the menu: Progress in wild meat research, policy, and practice from 2002 to 2020. Annual Review of Environment and Resources, 46, 221–254. IPBES. (2019). Summary for policymakers of the global assessment report on biodiversity and ecosystem services of the intergovernmental science-policy platform on biodiversity and ecosystem services (tech. rep.). IPBES Secretariat. Bonn, Germany. Klain, S. C., Olmsted, P., Chan, K. M., & Satterfield, T. (2017). Relational values resonate broadly and differently than intrinsic or instrumental values, or the New Ecological Paradigm. PLoS ONE, 12(8), e0183962. LaCerva, G. (2016). Untamed and rare: access and power in DRC’s emerging luxury bushmeat trade. In C. J. P. Colfer, B. B. Basnett, & M. Elias (Eds.), Gender and Forests: Cli- mate change, tenure, value chains and emerging issues (pp. 243–260). Earthscan from Routledge. Lucas, A., Kumakamba, C., Saylors, K., Obel, E., Kamenga, R., Makuwa, M., Clary, C., Miningue, G., McIver, D. J., Lange, C. E., Mbala Kingebeni, P., & Muyembe-Tamfun, J. J. (2022). Risk perceptions and behaviors of actors in the wild animal value chain in Kinshasa, Democratic Republic of Congo. PLoS ONE, 17 (2), e0261601. 35 MacFarlane, D., Hurlstone, M. J., Ecker, U. K., Ferraro, P. J., van Der Linden, S., Wan, A. K., ıssimo, D., Burgess, G., Chen, F., Hall, W., et al. (2022). Reducing demand for over- Ver´ exploited wildlife products: Lessons from systematic reviews from outside conservation science. Conservation Science and Practice, 4(3), e627. McKenzie, D. (2012). Beyond baseline and follow-up: The case for more T in experiments. Journal of Development Economics, 99(2), 210–221. McShane, B. B., Gal, D., Gelman, A., Robert, C., & Tackett, J. L. (2019). Abandon statistical significance. The American Statistician, 73(sup1), 235–245. ere de l’Economie Foresti` Minist` esentation du premier Spot TV de la ere - Congo. (2019). Pr´ eduction de la consommation de la viande de brousse. campagne sur la r´ ere de l’Environnement et D´ Minist` eveloppement Durable. (2024). Yoka Pimbo [Accessed: 2024-11-08]. https://yokapimbo.org/ Mintz, S. W., & Du Bois, C. M. (2002). The anthropology of food and eating. Annual Review of Anthropology, 31, 99–119. Montoya, R. M., Horton, R. S., Vevea, J. L., Citkowicz, M., & Lauber, E. A. (2017). A re- examination of the mere exposure effect: the influence of repeated exposure on recogni- tion, familiarity, and liking. Psychological Bulletin, 143(5), 459–498. Moro, M., Fischer, A., Milner-Gulland, E., Lowassa, A., Naiman, L. C., & Hanley, N. (2015). A stated preference investigation of household demand for illegally hunted bushmeat in the Serengeti, Tanzania. Animal Conservation, 18(4), 377–386. Pascual, U., Balvanera, P., Anderson, C. B., Chaplin-Kramer, R., Christie, M., Gonzalez-Jimenez, D., Martin, A., Raymond, C. M., Termansen, M., Vatn, A., Athayde, S., Baptiste, B., Barton, D. N., Jacobs, S., Kelemen, E., Kumar, R., Lazos, E., Mwampamba, T. H., Nakangu, B., . . . Zent, E. D. (2023). Diverse values of nature for sustainability. Nature, 620, 813–824. 36 Pascual, U., Balvanera, P., Diaz, S., Pataki, G., Roth, E., Stenseke, M., Watson, R. T., Dessane, E. B., Islar, M., Keleman, E., Maris, V., Quaas, M., Subramanian, S. M., Wittmer, G., Adlan, A., Ahn, S., Al-Hafedh, Y. S., Amankwah, E., Asah, S. T., . . . Yagi, N. (2017). Valuing nature’s contributions to people: the IPBES approach. Current Opinion in En- vironmental Sustainability, 26-27, 7–16. Perloff, J. M. (2023). Microeconomics (9th ed.). Pearson. R Core Team. (2024). R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. https://www.R-project.org/ Rentsch, S., & Damon, A. (2013). Prices, poaching, and protein alternatives: an analysis of bushmeat consumption around Serengeti National Park, Tanzania. Ecological Economics, 91, 1–9. Reynolds, W. M. (1982). Development of reliable and valid short forms of the Marlowe-Crowne Social Desirability Scale. Journal of Clinical Psychology, 38(1), 119–125. ıguez-R´ Rodr´ ıa-P´ ıos, E., & Garc´ aez, B. (2018). Valuing bushmeat for people living at a subsis- ı, Ecuador. Human tence level: the case of Cuniculus paca meat in Flavio Alfaro, Manab´ Dimensions of Wildlife, 23(4), 313–328. Rose, A. L. (2001). Social change and social values in mitigating bushmeat commerce. In M. I. Bakarr, G. A. B. da Fonseca, R. Mittermeier, A. B. Rylands, & K. W. Paine- milla (Eds.), Hunting and bushmeat utilization in the African rain forest. Perspectives toward a blueprint for conservation action. (pp. 59–74). Center for Applied Biodiversity Science at Conservation International, Washington, D. C., USA. Rubin, D. B. (1974). Estimating causal effects of treatments in randomized and nonrandomized studies. Journal of Educational Psychology, 66(5), 688. Sandalj, M., Treydte, A. C., & Ziegler, S. (2016). Is wild meat luxury? Quantifying wild meat demand and availability in Hue, Vietnam. Biological Conservation, 196, 105–112. 37 Schulz, C., Martin-Ortega, J., Glenk, K., & Ioris, A. A. (2017). The Value Base of Water Gov- ernance: A Multi-Disciplinary Perspective. Ecological Economics, 131, 241–249. Simo, F. T., Difouo, G. F., Tchana, C. N., Wandji, A. C., Djomnang-Nkwala, A. L., Mbad- e, M., Kekeunou, S., Swiack´ joun Nzik´ a, M., & Ingram, D. J. (2024). Urban wild meat and pangolin consumption across southern forested Cameroon: the limited influence of COVID-19. People and Nature, 1–15. Sobrevila, C. (2016). Reducing demand must be a core component of combatting wildlife crime [World Bank Voices]. https://blogs.worldbank.org/voices/reducing-demand-must-be- core-component-cutting-wildlife-crime Stantcheva, S. (2023). How to run surveys: A guide to creating your own identifying variation and revealing the invisible. Annual Review of Economics, 15(1), 205–234. ´ Sustainable Wildlife Management Programme. (2024). CONSOMMATION. REPUBLIQUE ´ DEMOCRATIQUE DU CONGO [Accessed: April 4, 2024]. https://www.swm-programme. info/legal-hub/rdcongo/consumption-use Travers, H., Walsh, J., Vogt, S., Clements, T., & Milner-Gulland, E. J. (2021). Delivering be- havioural change at scale: What conservation can learn from other fields. Biological Conservation, 257, 109092. Trefon, T. (2023). Bushmeat: Culture, Economy and Conservation in Central Africa. Hurst Publishers. van Vliet, N., & Mbazza, P. (2011). Recognizing the multiple reasons for bushmeat consumption in urban areas: a necessary step toward the sustainable use of wildlife for food in Central Africa. Human Dimensions of Wildlife, 16(1), 45–54. uggen, B., Muhindo, J., Nebesse, C., Gambalemoke, S., & Nasi, van Vliet, N., Schulte-Herbr¨ R. (2017). Trends in bushmeat trade in a postconflict forest town: implications for food security. Ecology and Society, 22(4). 38 ıssimo, D., Schmid, C., Kimario, F., & Eves, H. (2018). Measuring the impact of an entertainment- Ver´ education intervention to reduce demand for bushmeat. Animal Conservation, 21(4), 324–331. ıssimo, D., & Wan, A. K. Y. (2019). Characterizing efforts to reduce consumer demand for Ver´ wildlife products. Conservation Biology, 33(3), 623–633. Wakefield, M. A., Loken, B., & Hornik, R. C. (2010). Use of mass media campaigns to change health behaviour. Lancet, 376(9748), 1261–1271. Walelign, S. Z., Nielsen, M. R., & Jacobsen, J. B. (2019). Roads and livelihood activity choices in the Greater Serengeti Ecosystem, Tanzania. PLoS ONE, 14(3), e0213089. Wasserstein, R. L., & Lazar, N. A. (2016). The ASA statement on p-values: context, process, and purpose. The American Statistician, 70(2), 129–133. Wells, G. J., Ryan, C. M., Das, A., Attiwilli, S., Poudyal, M., Lele, S., Schreckenberg, K., Robinson, B. E., Keane, A., Homewood, K. M., Jones, J. P. G., Torres-Vitolas, C. A., Fisher, J. A., Ahmad, S., Mulligan, M., Dawson, T. P., Adams, H., Siddappa Setty, R., & Daw, T. M. (2024). Hundreds of millions of people in the tropics need both wild harvests and other forms of economic development for their well-being. One Earth, 7, 311–324. o, M. (2018). Steward- West, S., Haider, L. J., Masterson, V., Enqvist, J. P., Svedin, U., & Teng¨ ship, care and relational values. Current Opinion in Environmental Sustainability, 35, 30–38. WildAid. (2020). 2019 Annual Report. https : / / wildaid . org / wp - content / uploads / 2020 / 06 / WildAid Annual-Report-2019.pdf WildAid. (2021). Understanding Urban Consumption of Bushmeat in Nigeria. https://wildaid. org/wp-content/uploads/2021/01/Nigeria-Bushmeat-Consumption-Survey-Report.pdf 39 Wilkie, D. S., Starkey, M., Abernethy, K., Nstame Effa, E., Telfer, P., & Godoy, R. (2005). Role of prices and wealth in consumer demand for bushmeat in Gabon, Central Africa. Conservation Biology, 19(1), 268–274. Wilkie, D. S., Wieland, M., Boulet, H., Le Bel, S., van Vliet, N., Cornelis, D., BriacWarnon, V., Nasi, R., & Fa, J. E. (2016). Eating and conserving bushmeat in Africa. African Journal of Ecology, 54, 402–414. Willis, J., Ingram, D. J., Abernethy, K., Kemalasari, D., Muchlish, U., Sampurna, Y., Midoko Iponga, D., & Coad, L. (2022). WILDMEAT interventions database: a new database of interventions addressing unsustainable wild meat hunting, consumption and trade. African Journal of Ecology, 60, 205–211. Wright, J. H., Malekani, D., Funk, S. M., Ntshila, J., Mayet, L., Mwinyihali, R., Fa, J. E., & Wieland, M. (2022). Profiling the types of restaurants that sell wild meat in Central African cities. African Journal of Ecology, 60(2), 197–204. Yocum, L. F., Vanegas, L., & Day, B. A. (2022). From the forest to the fork: Why we need to “reframe conservation”’ for conservation behavior change campaigns. Applied Environ- mental Education & Communication, 21(1), 3–6. Data Availability: All anonymized data are publicly available (Cisse et al., 2024). Code Availability: All replication codes are publicly available (Cisse et al., 2024). Acknowledgments: We thank Alexandra Jarotschkin, Megan Lang, John Loeser, James Sallee, and Juliet Wright for helpful comments, and Maximilian Auffhammer for Institutional Review Board assistance. Funding for the experiments provided by the Development Research Group of the World Bank. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the World Bank 40 and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Daniel J. Ingram acknowledges support from UK Research and Innovation (Future Leaders Fellowship awarded to DJI, Grant ref: MR/W006316/1). Conflict of Interest: AC, GE, and DJI declare that they have no conflicts of interest. Author contributions: AC, GE, and DJI conceived the ideas and designed methodology; GE collected the data; AC and GE analyzed the data; GE led the writing of the manuscript with DJI drafting the discussion. All authors contributed critically to the drafts and gave final approval for publication. Declaration of Generative AI and AI-Assisted Technologies in the Writing Process: During the preparation of this work (Introduction through Results) GE used GPT-4 and Claude 3.5 Sonnet in order to draft and revise text. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication. AI and AI-Assisted Technologies were not used for the Discussion drafted by DJI. 41 Supporting Information for Conserving Wildlife through Demand Reduction and Supply Alternatives: Two Experiments in Restaurants in Kinshasa Abdoulaye Cisse, Gabriel Englander, Daniel J. Ingram Correspondence to: aenglander@worldbank.org This PDF file includes: A. Supplementary Tables and Figures B. English Transcript of Treatment Video C. Survey Instruments 1 A Supplementary Tables and Figures (1) (2) Treatment -0.029 -0.0007 (0.043) (0.0014) Distance to Restaurant 0.0043 (0.0018) Baseline Average Price -0.0449 (0.0071) Restaurant Habit -0.0001 (0.0015) Bushmeat Days 0.0000 (0.0004) Age 0.0000 (0.0001) Male -0.0012 (0.0016) University Graduate -0.0088 (0.0030) Years of Education 0.0015 (0.0004) Salary Earner 0.0025 (0.0016) Business Owner -0.0008 (0.0018) Intercept 0.517 (0.030) Date and Table Fixed Effects No Yes N 544 544 Percent Change -5.63 -0.13 Table S1: Effect of Treatment on Probability of Choosing Wild Meat-Intensive Restaurant in Demand-Side Experiment. The dependent variable in both regressions is an indicator that equals 1 for the 2 restaurants whose sample menus contain the highest proportion of wild meat. Distance to Restaurant is the distance in kilometers between where the subject participated in the experiment and the location of the restaurant they chose. Baseline Average Price is the mean price, in thousand Congolese Francs, of all dishes on the sample menu shown to participants in the demand-side experiment. Standard errors are clustered at the subject level. 2 Dependent Variable Treatment Coefficient Treatment Standard Error Control Mean N Unfavorable -0.086 (0.082) 0.041 544 Tasty -0.023 (0.014) 0.038 544 Sustainable 0.023 (0.022) 0.059 544 Healthy -0.038 (0.025) 0.115 544 Fresh -0.004 (0.024) 0.084 544 Cool -0.027 (0.014) 0.042 544 Legal -0.014 (0.018) 0.056 544 Origin 0.028 (0.031) 0.147 544 Proud -0.052 (0.031) 0.783 544 Table S2: Effect of Treatment on Perceptions of Bushmeat in Demand-Side Experiment. Each row corresponds to a separate regression. “Unfavorable” indicates the standardized index variable, which is calculated from the eight other variables (Section 2.2.3). The dependent variables in the second to ninth rows equal 1 if the subject views bushmeat negatively along that dimension, and they equal 0 otherwise. All regressions include the eight pre-specified control variables, date fixed effects, and table fixed effects (Equation 1). Standard errors are clustered at the subject level. Dependent Variable: Ordered Wild Meat (1) (2) Treatment -0.009 -0.012 (0.017) (0.017) Social Desirability 0.043 0.035 (0.018) (0.020) Treatment×Social Desirability -0.028 -0.028 (0.021) (0.021) Intercept 0.044 (0.012) Date and Table Fixed Effects No Yes N 544 544 Table S3: Testing Whether Treatment Subjects Who Provided More Socially Desirable Answers Ordered Less Wild Meat in the Demand-Side Experiment. The dependent variable is an indicator that equals 1 if the subject ordered wild meat and equals 0 otherwise. Standard errors are clustered at the subject level. 3 (a) Demand-Side Experiment (b) Supply-Side Experiment −0.04 0.03 −0.02 −0.01 0 0.01 0 0.5 1 1.5 2 Figure S1: 95% Confidence Intervals Corresponding to Demand-Side Experiment and Supply-Side Experiment. (a) The point is the treatment coefficient from Column 2 of Table 2 and the horizontal line on either side of the point is the coefficient’s 95% confidence interval. (b) The point is the estimated elasticity of wild meat sales with respect to Moambe Chicken price (Column 5 of Table 3) and the horizontal line on either side of the point is the elasticity’s 95% confidence interval. 4 B English Transcript of Treatment Video [Person 1 and 2 meet on the street.] • Person 1: Hello X, it’s been a long time! • Person 2: As you know, my brother, I carry out my activities in the village. That’s what keeps me busy there. [Persons 1 and 2 enter Person 1’s house.] • Person 1: Honey, honey, • Person 3 [Person 1’s Wife]: Yes honey! • Person 1: You won’t guess who’s here. • Person 3: Hey [Person 2], where have you been? • Person 2: Here I am! • Person 3: It’s been a while. [I am] very happy [to see you]. [Person 1 and 2 sit down at a table.] • Person 1: Make yourself at home. [Body greetings between the different persons.] • Person 1: You know, I met him here in front of the gate and I invited him for a beer. • Person 3: [To her daughter], can you go get me some cold beers please? [Person 4 brings the beers to the table.] • Person 1: Tomorrow, in fact darling, when your brother comes back, you can make us a good bushmeat dish. 5 • Person 2: Bush meat? No, I no longer consume it myself. Our forests are empty, due to the consumption of bush meat in the city. • Person 3: Oh yes? • Person 2: I know something about that. I have my activities in the village. • Person 1: Oh so? I didn’t know this. • Person 2: Yes! It’s no longer like during our childhood when there was still bushmeat. Today, it is very rare. That’s why I decided to no longer eat bush meat. • Person 3: Absolutely! • Person 1: [To Person 3] I don’t think we’re going to eat bush meat in this house anymore. • Person 3: You’re right! If our forests are empty, we must eat less bush meat in the city. OK, I’ll make you some good Congolese cuisine then, without the bush meat. • Person 2: [Approval in local language]. [End of the video with an audio and written message: “The large consumption of bush meat in the city is emptying the Congolese forests. Let’s eat less bush meat in the city.”] C Survey Instruments At the start of the demand-side experiment, enumerators asked participants to specify their preferred language among French, Lingala, or Swahili. The experiment was then conducted in the participant’s preferred language; the survey questions and answer choices listed in this appendix have been translated into English. Participants responded to these survey questions after watching the video and choosing the restaurant to which they wanted to receive a coupon. 6 First, enumerators asked subjects six questions about different aspects of bushmeat: 1. Do you agree or disagree that bushmeat is tasty? 2. Do you agree or disagree that bushmeat is sustainable? 3. Do you agree or disagree that bushmeat is healthy? 4. Do you agree or disagree that bushmeat is fresh? 5. Do you agree or disagree that bushmeat is cool? 6. Do you agree or disagree that bushmeat is legal? We designed our survey to conform with the following best practices (Stantcheva, 2023). Stating both sides in the question stem (“agree or disagree”) improves question clarity and reduces response bias. We randomized the order of questions to reduce the influence of question order on responses. Subjects chose from five answer options: strongly agree, somewhat agree, neither agree nor disagree, somewhat disagree, and strongly disagree. The middle, indifferent option allows us to capture subjects’ uncertainty or lack of opinion. We also randomized the order of the question stems (“agree or disagree” or “disagree or agree”) and the order of the response options (strongly agree to strongly disagree, or strongly disagree to strongly agree) to avoid biasing subjects towards more agreement or more disagreement (Stantcheva, 2023). Next, enumerators asked subjects four questions regarding the frequency of wild meat con- sumption in their social network: 1. Do you disagree or agree that bushmeat connects you to your place of origin? 2. Is it uncommon or common for your friends to eat bushmeat? 3. Is it uncommon or common for your family to eat bushmeat? 7 4. Is it uncommon or common for your colleagues to eat bushmeat? As with the first set of questions, subjects chose from one of five answer options, and we randomized the order of questions, question stems, and response options. The answer options for the commonality questions are very common, somewhat common, neither common nor uncommon, somewhat uncommon, and very uncommon. Finally, we asked subjects whether they agreed or disagreed with the following statements. We allowed subjects to choose one of five answer options and we randomized the order of response options. The first question is the eighth and final one we pre-specified to measure attitudes towards wild meat. The final 13 questions follow the social desirability measurement of Dhar et al. (2022), who use a 13-question module in line with Reynolds (1982) and Crowne and Marlowe (1960). 1. Are you proud of the environment of the Democratic Republic of Congo? 2. It is sometimes hard for me to go on with my work if I am not encouraged. 3. I sometimes feel resentful when I don’t get my way. 4. On a few occasions, I have given up doing something because I thought too little of my ability. 5. There have been times when I felt like rebelling against people in authority even though I knew they were right. 6. No matter who I’m talking to, I’m always a good listener. 7. There have been occasions when I took advantage of someone. 8. I am always willing to admit it when I make a mistake. 8 9. I sometimes try to get even rather than forgive and forget. 10. I am always courteous, even to people who are disagreeable. 11. I have never been irked when people expressed ideas very different from my own. 12. There have times when I was quite jealous of the good fortune of others. 13. I am sometimes irritated by people who ask favors of me. 14. I have deliberately said something that hurt someone’s feelings. The answer choices for the first question are very proud, somewhat proud, neither proud nor ashamed, somewhat ashamed/embarrassed, and very ashamed/embarrassed. The answer options for the 13 social desirability questions are strongly agree, somewhat agree, neither agree nor disagree, somewhat disagree, and strongly disagree. 9