| Call Number | 17182 |
|---|---|
| Day & Time Location |
T 9:00am-10:50am 801 International Affairs Building |
| Points | 3 |
| Grading Mode | Standard |
| Approvals Required | None |
| Instructor | Douglas V Almond |
| Type | LECTURE |
| Method of Instruction | In-Person |
| Course Description | Prerequisites: SIPA U6501 The goal of this course is to enable students to evaluate the policy relevance of academic research. While academic research frequently considers treatments that approximate a potential public policy, such prima facie relevance alone does not inform policy. In particular, public policy is predicated on the credible estimation of causal treatment effects. For example, although researchers frequently document the strong correlation between years of schooling and better health, this tells us surprisingly little (and arguably nothing) about the health effects of public tuition assistance, compulsory school laws, or any other program that raises educational attainment. Policies guided by statistical correlations - even the regression-adjusted estimates that dominate the academic literature - will frequently have unintended and even perverse real-world effects. Policymakers must distinguish between causal estimates that should inform policy design and statistical correlations that should not. The catch is that distinguishing correlation from causation in empirical studies is surprisingly difficult. Econometric technique alone does not provide a reliable path to causal inference. Applications of instrumental variables (IV) techniques, while wildly popular, arguably obscure sources of identification more often than isolating exogenous variation. Similar concerns apply to popular panel data and fixed effects (FE) models, which can eliminate certain unobservable sources of bias. Furthermore, causal claims by a study's author should be regarded with skepticism - frequently this is merely the marketing of a non-transparent statistical correlation. Put differently, when has a researcher portrayed his empirical result as a mere correlation when in fact he/she had identified a credible causal impact? A basic theme of the course is that identification strategy - the manner in which a researcher uses observational real-world data to approximate a controlled/randomized trial (Angrist & Pischke, 2009) - is the bedrock of causal inference. Econometric technique cannot rescue a fundamentally flawed identification strategy. In other words, econometrics and identifications strategies are complements in the production of causal estimates, not substitutes. Examples of appropriate econometric technique applied to compelling identification strategies will be described to illustrate this approach (most often from health economics), along with their implications for public policy. |
| Web Site | Vergil |
| Department | International and Public Affairs |
| Enrollment | 17 students (20 max) as of 1:06PM Saturday, May 11, 2024 |
| Subject | International Affairs |
| Number | U6604 |
| Section | 001 |
| Division | School of International and Public Affairs |
| Campus | Morningside |
| Note | Pre-req: Quant II |
| Section key | 20233INAF6604U001 |