Yingying Dong - UC Irvine

Regression Discontinuity Designs with a Continuous Treatment

    Date:  10/19/2017 (Thu)

    Time:  3:30pm- 5:00pm

    Location:  Social Sciences room 113

    Organizer:  Arnaud Maurel

Meeting Schedule: Login or email the organizer to schedule a meeting.

    9:30am - Arnaud (pick up from hotel)

   10:00am - Adam @ 221B

   10:30am - Jia @ 228G

   11:00am - Javier Romero Haaker

   11:30am - Margaux Luflade

   12:00pm - Lunch with Margaux Luflade and Paul Diegert

    1:00pm - Matt @202

    1:30pm - Federico @240

    2:00pm - Rob Garlick

    2:30pm - Jackson Bunting

    3:00pm - Seminar prep

    3:30pm - Seminar Presentation (3:30pm to 5:00pm)

    6:00pm - Dinner with Arnaud, Jia, Fede, Matt

    Additional Comments:  Abstract: This paper provides identification and inference theory for RD designs with a continuous treatment. We show that a control variable can be constructed (a la Imbens and Newey, 2009). Conditional on this control variable, treatment is exogenous to the outcome of interest. Our identification relies on a distributional change in the first-stage, in contrast to a mean change, so as long as there are any changes in the treatment distribution (including the mean change as a special case) at the RD threshold, one can identify causal effects of the treatment. The proposed approach is relevant to a large class of public policies that target not necessarily the average units, but only those in some parts (e.g., top or bottom) of the distribution for treatment. We provide bias-corrected robust inference for our local average treatment effects, either at a given treatment quantile or over the treatment distribution, along with their AMSE optimal bandwidths. The proposed approach is shown to be useful in investigating the impacts of minimum capital requirements on bank stability, where the policy targets small banks, and so the lower tail of the capital distribution shifts up at a certain policy threshold.