Toxicity and Price-Quality Information and Pesticide Demand: RCT Evidence from Smallholder Tomato Farmers in Zambia The environmental and human health risks from pesticide use are well documented. Small-scale farmers face large health risks as they regularly come into direct contact with pesticides while handling and applying them to their crops. These risks are a function of two factors: exposure and toxicity. Research has shown that small-scale farmers in Africa, Asia, and Central America seldom use the recommended personal protective equipment (PPE) when working with pesticides, thus their exposures are often high. High exposure makes the toxicity of the pesticides used by farmers all the more important. However, regulation of highly toxic pesticides in many developing countries lags behind many developed countries; small-scale farmers often have access to World Health Organization (WHO) class 1 (highly toxic) pesticides with large human health risks (e.g., several pesticides used by small-scale farmers are banned in the US and Europe). The pesticide safety situation is exacerbated by a poor understanding of toxicity, as farmers often misinterpret or ignore pesticide labels. Researchers agree that safety knowledge can be improved with targeted training; however literature has not yet tested the impacts of such training on pesticide demand and toxicity. We address this gap using a randomized control trial. We conducted baseline and endline interviews on a random sample of 512 tomato farmers in rural Zambia. Using our baseline data, we created 32 spatial clusters of farmers and paired the clusters into 16 blocks based on their initial knowledge of pesticide safety. We then randomly selected one cluster from each block to participate in a Lead Farmer training program – a common, low cost, rural extension method. To assess pesticide demand we used a contingent behavior experiment designed to mirror how farmers typically make pesticide purchase decisions. First, we showed farmers a production scenario which described an observed pest pressure on their tomatoes. We then presented farmers with a choice set of actual pesticides from nearby retailers with varied prices. Farmers then stated their choice of pesticides with an option to select none of the products. We designed two of these contingent behavior experiments, one for nematicides and one for insecticides – representing the categories where Zambian farmers use the most WHO class 1 pesticides – and we evaluated farmers’ choices at the baseline (pre-training) and endline (post-training) surveys. First, our baseline survey revealed large knowledge gaps in pesticide toxicity and pest control properties. Second, our choice data revealed a more fundamental issue in how farmers chose pesticides. Our estimations for both the insecticide and the nematicide experiments showed a positive and significant relationship between pesticide price and the probability that a particular pesticide was selected. In other words, as a pesticide’s price increased, farmers chose that pesticide more often. Qualitative interviews with farmers confirmed this result as many farmers openly expressed their perception that the higher priced pesticides were always the higher quality (i.e., higher efficacy) products. Farmers have numerous retail shop options for their pesticide purchases – more than a dozen outlets were accessible by our sample – each with a constantly changing selection of brands and products. We hypothesize that farmers therefore look for a simplifying heuristic – price – to help them choose their pesticides. Yet price is not a reliable signal for pesticide quality in rural Zambia: we observed price differences as large as 100% across outlets for the same products at the same time. The presence of this price-quality heuristic may reduce the impact of toxicity and health risk information on farmer demand. To combat this problem, we devoted a portion of our RCT pesticide training program to teaching farmers the distinction between price and quality. We provided examples of observed prices from nearby retailers to help break the price-quality association and we informed farmers of the pest control properties for a sample of six insecticides and three nematicides, all of which were available from nearby retailers. Our endline (post-training) pesticide choice data showed that our training successfully broke the price-quality association. The group of farmers that were randomly selected to receive training responded to prices in a normal economic way for both our insecticide and nematicide experiments: all else equal, higher prices were significantly associated with a decreased probability of choosing the pesticide. In contrast, the choices for our entire control group continued to show the positive price-quality perception for our nematicide experiment, and, for our insecticide experiment, farmer choices showed a positive relationship between price and selection probability for the portion of the control group that reported a perceived price-quality relationship. Our training significantly changed farmers’ choices for pesticide toxicities. Our control group demonstrated similar toxicity choices across survey rounds, but our training group showed changes. Farmers in our training group were much more likely to select low toxicity pesticides for both the nematicide and insecticide experiments. While our research design does not allow us to directly parse out the impacts of each type of information, we present evidence that providing farmers with pesticide toxicity and health risk information while also providing evidence against a price-quality association can address incomplete and flawed information driving farmers’ pesticide choices. Although a substantial amount of effort is directed at informing farmers about the health risks of pesticides, our results suggest that informing farmers about other attributes of products in the market may be equally important. This has potentially large implications for farmer health in developing countries as farmers were much more likely to select low toxicity pesticides after our training. Our research will stimulate discussion based on its policy relevance, its contemporary use of a randomized controlled trial for the training treatment/information intervention, and because the results suggest an important economic contribution to health and safety information campaigns: correcting market information imperfections about product efficacy and concurrent toxicity information can have a significant health impact through product substitution.