Quasi-Experiments in Empirical Economics: An Application to Extreme Levels of Lake Michigan Michael R. Moore School of Natural Resources and Environment University of Michigan
Outline Quasi-experiments in empirical economics Application to extremely low levels of Lake Michigan Research question: did the extremely low levels of Lake Michigan in 2000-07 cause economic impacts in Lake Michigan coastal counties?
Quasi-experiment Experiment Scientific method of control group, treatment group, and experimental treatment What is the effect of the treatment on the treatment group? Quasi-experiment Research design of groups and a treatment not arranged in advance A by-product of public policy, public program, or nature Desirable feature: establishes causal evidence, not simply association evidence
Average Treatment Effect (ATE): what is the effect of the treatment on an outcome variable in the treated counties? ATE = [(Average 4 – Average 3) – (Average 2 – Average 1)] (approximated ATE)
Policy dialogues on lake levels (IJC) Lake Ontario – St. Lawrence River Study (2006) Lake Ontario is a regulated lake “Lake levels are too even; not enough fluctuation” Upper Great Lakes Study (2012) Lake Superior is regulated; Lakes Michigan-Huron and Erie are unregulated Joint management of these lake levels “levels of Lake Michigan-Huron are too extreme”
Great Lakes lake levels, 1960-2007 Michigan-Huron Erie Superior Ontario
Extremely low lake levels as a quasi experiment Control group: Counties adjacent to Lakes Ontario and Erie Treatment group: Counties adjacent to Lake Michigan-Huron Treatment: 2000-07 episode of extremely low levels
Data: NAICS Sector 72, Accommodation and Food Services County-level data for 1990-2007 Outcome variables: establishments, employment, and wage rate in Sector 72 We expect the Average Treatment Effects to be negative