Zu Chongzhi Research Seminar

Date and Time (China standard time): Wednesday, March 4, 2026, 2:00 – 3:00 pm

Location: IB 2025

Title: Optimal Push, Pull, and Failure Funding for Global Health

Speaker: Peng Sun, Duke University

Abstract: Malaria and tuberculosis each cause more than half a million deaths each year. However, commercial incentives for developing drugs and vaccines for malaria and other tropical diseases remain weak. Governments and nonprofits address these market failures through push (e.g., grants) and pull (e.g., prizes) mechanisms. We propose a third approach: the funder pays only if the firm fails, reimbursing part of its testing costs. This failure insurance is optimal when markets are large enough to reward success but too small to spur investment. The mechanism addresses both adverse selection and moral hazard. Failure insurance, like a forgivable loan, is preferred for tuberculosis if testing costs are below $1 billion. Tuberculosis affects many people in both high- and low-income countries, unlike malaria. For most tropical diseases, including malaria, we recommend pull funding with supplemental push support. These findings challenge current push-heavy practice and have broader relevance for innovation policy.

Bio: Peng Sun is a JB Fuqua Professor in the Decision Sciences area at the Fuqua School of Business, Duke University. He researches mathematical theories and models for resource allocation decisions under uncertainty, and incentive issues in dynamic environments. His work spans a range of applications areas, from operations management, economics, finance, marketing, to health care and sustainability. He serves as a Department Editor at Management Science, and an Associate Editor at Operations Research, two leading academic journals of the profession of Operations Research and Management Science. At the Fuqua School, Professor Sun has taught MBA core course Decision Models and elective course Strategic Modeling and Business Dynamics, and PhD course Dynamic Programming and Optimal Control.