ECON742:ECON742:Empirical microeconomicsEmpirical microeconomics
Lecture 1Lecture 1January 2014January 2014
correlation≠causality In economics, we want to measure the causal impact of X on Y.
But crucial distinction between correlation and causality
Correlation: relationship between X and Correlation: relationship between X and Y Y
Causality: X causes YCausality: X causes Y Maybe X->Y or Y->XMaybe X->Y or Y->X Maybe Z->X, Z->Y and so X<->YMaybe Z->X, Z->Y and so X<->Y
Z->X, Z->Y donc X<-Z->X, Z->Y donc X<->Y>Y
Ice cream sales and shark attacksIce cream sales and shark attacks Number of decayed teeth and Number of decayed teeth and
size of vocabularysize of vocabulary Recent article in the British Recent article in the British
Medical Journal: Brush your teeth Medical Journal: Brush your teeth 2 times per day, less 2 times per day, less cardiovascular diseasescardiovascular diseases
X->Y ou Y->X?X->Y ou Y->X? Number of doctors and patientsNumber of doctors and patients
Causal impact of X on Y: Causal impact of X on Y: OLS?OLS? Lecture 1: OLSLecture 1: OLS Lecture 2: LimitationsLecture 2: Limitations Lecture 3: Panel Data Lecture 3: Panel Data
Lecture 4: Program evaluation
Causal impact of X on YCausal impact of X on Y Ideal?Ideal? Typical problem of evaluation: the Typical problem of evaluation: the
counterfactualcounterfactual Bad control group because of:Bad control group because of:– Self-selectionSelf-selection– Non random program placementNon random program placement
Empirical techniques Lecture 5: Randomized experiments Lecture 5: Randomized experiments Lecture 6: Instrumental variablesLecture 6: Instrumental variables Lecture 7: Difference-in-differencesLecture 7: Difference-in-differences Lecture 8: Regression discontinuity Lecture 8: Regression discontinuity
designdesign Lecture 9: Selection modelsLecture 9: Selection models Lecture 10: MatchingLecture 10: Matching Lecture 12: …Lecture 12: …
Conclusion Correlation does not imply causalityCorrelation does not imply causality The economists have tools The economists have tools
(randomized experiments, (randomized experiments, instrumental variables, DID, instrumental variables, DID, discontinuity…) to get the causal discontinuity…) to get the causal impact of programs, reforms, impact of programs, reforms, policies on the behaviour of human policies on the behaviour of human beingsbeings
Evaluation Final exam: 40% (Date TBA)Final exam: 40% (Date TBA) Mid-term exam: 20% (03/11)Mid-term exam: 20% (03/11) Exercises (four): 30%Exercises (four): 30% Presentation: 10% (6 papers)Presentation: 10% (6 papers)
Objectives of the course correlation≠causality Master the main Master the main
microeconometric techniquesmicroeconometric techniques Be able to understand a research Be able to understand a research
paper (presentation)paper (presentation) Be able to maipulate data Be able to maipulate data
(exercises)(exercises)
Correlation/causality
http://xkcd.com/552/http://xkcd.com/552/