"What do I do about the chickens?"

When assistant professor of medicine Eran Bendavid began a study on livestock in African households to determine impact on childhood health, he'd already anticipated common field problems like poorly captured or intentionally misreported data, difficulty getting to work sites, or problems with training local volunteers.

But he'd never gotten that particular question from a fieldworker before. It didn't occur to him that participating families, in reporting their livestock holdings, would completely omit the chickens running around at their feet, thereby skewing the data.

"They didn't consider chickens to be livestock," recalled Bendavid. Along with Scott Rozelle, the Helen F. Farnsworth Senior Fellow at FSI, and associate professor of political science and FSI senior fellow Beatriz Magaloni, Bendavid spoke to a full house last week on lessons learned from fieldwork gone awry. The return engagement of FSI's popular seminar, "Everything that can go wrong in a field experiment” was introduced by Jesper Sørensen, executive director of Stanford Seed, and moderated by Katherine Casey, assistant professor of political economy at the GSB. The seminar is a product of FSI and Seed’s joint Global Development and Poverty (GDP) Initiative, which to date has awarded nearly $7 million in faculty research funding to promote research on poverty alleviation and economic development worldwide.

Rozelle, co-director of the Rural Education Action Program, spoke of the obstacles to accurate data gathering, especially in rural areas where record-keeping is inaccurate and participants' trust is low. Arriving in a Chinese village to carry out child nutrition studies, said Rozelle, "we found Grandma running out the back door with the baby." The researchers had worked with the local family planning council to find the names of children to study, but the families thought the authorities were coming to penalize them for violation of the one-child policy.

Cultural differences make for entertaining and illuminating (if frustrating) lessons, but Beatriz Magaloni, director of FSI's Program on Poverty and Governance at the Center on Democracy, Development and the Rule of Law had a different story to tell. Over the course of three years, her GDP-funded work to investigate and reduce police violence in Brazil - a phenomenon resulting in more than 22,000 deaths since 2005 - has encountered obstacle after obstacle. Her work to pilot body-worn cameras on police in Rio has faced a change in police leadership, setting back cooperation; a yearlong struggle to decouple a study of TASER International’s body worn cameras from its electrical weapons in the same population; a work site initially lacking electricity to charge the cameras or Internet to view the feeds; and noncompliance among the officers. "It's discouraging at times," admitted Magaloni, who has finally gotten the cameras onto the officers' uniforms and must now experiment with ways to incentivize their use. "We are learning a lot about how institutional behavior becomes so entrenched and why it's so hard to change."

Experimentation is a powerful tool to understand cause and effect, said Casey, but a tool only works if it's implemented properly. Learning from failure makes for an interesting panel discussion. The speakers' hope is that it also makes for better research in the future.

The Global Development and Poverty Initiative is a University-wide initiative of the Stanford Institute for Innovation in Developing Economies (Seed) in partnership with the Freeman Spogli Institute (FSI). GDP was established in 2013 to stimulate transformative research ideas and new approaches to economic development and poverty alleviation worldwide. GDP supports groundbreaking research at the intersection of traditional academic disciplines and practical application. GDP uses a venture-funding model to pursue compelling interdisciplinary research on the causes and consequences of global poverty. Initial funding allows GDP awardees to conduct high-quality research in developing countries where there is a lack of data and infrastructure.