Industry Dynamics with Social Learning: Evidence from Hydraulic Fracturing

I model the interaction between dynamic decision making and social learning about new technologies in driving industry takeoff and productivity growth. Learning about the use of new technologies is an important factor in economic growth, but I demonstrate that anticipated social learning can lead to a free-riding dynamic in scenarios with high uncertainty. I consider the empirical setting of hydraulic fracturing in North Dakota, where firms learn about the optimal use of fracturing technology, in part due to detailed data published by regulators. The cumulative value of this learning process is a ceteris paribus 40% increase in expected profitability. I model the impact of learning externalities on agents’ decisions to drill shale oil wells, an optimal stopping problem. My estimates suggest that the social learning externality is too small to affect investment as the industry develops and uncertainty is reduced. Conversely, I demonstrate that under higher uncertainty, anticipated social learning can lead to significantly lower industry investment and learning rates. Under this scenario, I also demonstrate the potential for public tests of the technology to enhance welfare by leading to more investment and a higher learning rate. ∗Department of Economics, Duke University; [email protected]. Special thanks is owed to my dissertation committee: Allan Collard-Wexler, Chris Timmins, Jimmy Roberts, Steven Sexton, and Daniel Yi Xu. I am also grateful for conversations with Steve Slawson of Slawson Exploration and Rob Jacobs of Caird Energy on the industry. I have also benefited from discussions with T. Robert Fetter, Brian Prest, members of the Duke Economics Public / IO lunch seminar, and of the Duke University Energy Initiative. In addition, I thank DrillingInfo for providing data. Any errors are my own.