BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Natural Experiment Policy Evaluation: A Critique
DTSTART:20150327T103000
DTEND:20150327T120000
DTSTAMP:20260413T182939Z
UID:bb84e9914875c20daeb96d52f9706966c92dad8c45029bac78a1fd79
CATEGORIES:Conferences - Seminars
DESCRIPTION:Ilya STREBULAEV (Stanford University)\nWe argue exogenous rand
 om treatment is insufficient for valid inference regarding the sign and ma
 gnitude of causal effects in dynamic environments. In such settings\, trea
 tment responses must be understood as contingent upon the typically unmode
 led policy generating process. With binary assignment\, this results in qu
 antitatively signi…ficant attenuation bias. With more than two policy st
 ates\, treatment responses can be biased downward\, upward\, or have the w
 rong sign. Further\, it is not only generally invalid to extrapolate elast
 icities across policy processes\, as argued by Lucas (1976)\, but also to 
 extrapolate within the same policy process. We derive auxiliary assumption
 s beyond exogeneity for valid inference in dynamic settings. If all possib
 le policy transitions are rare events\, treatment responses approximate ca
 usal effects. However\, reliance on rare events is overly-restrictive as t
 he necessary and sufficient conditions for equality of treatment responses
  and causal effects is that policy variable changes have mean zero. If the
 se conditions are not met\, we show how treatment responses can neverthele
 ss be corrected and mapped back to causal effects or extrapolated to forec
 ast responses to future policy changes.
LOCATION:UNIL\, Extranef\, room 126 https://planete.unil.ch/plan/?local=EX
 T-126
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR