Teaching CoP: Teaching Causality in...
Transcript of Teaching CoP: Teaching Causality in...
![Page 1: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/1.jpg)
Bla
nch
ena
y C
oP
201
9
Teaching Causality in ECO372
P. Blanchenay
Teaching and Learning Community of Practice
2019/05/27
![Page 2: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/2.jpg)
Bla
nch
ena
y C
oP
201
9
(Data) science without (data) conscience is but the ruin of the soul.
β (counterfactual) Rabelais.
![Page 3: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/3.jpg)
Bla
nch
ena
y C
oP
201
9
Angrist & Pischke βThrough our classes darklyβ (JEP, 2017)
Undergrad econometrics does not address causality
Focus on Gauss-Markov assumptions & their failures
Implicit or explicit focus on estimator efficiency
Little on identification strategies (e.g. diff-in-diff, RDD)
![Page 4: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/4.jpg)
Bla
nch
ena
y C
oP
201
9
Outline
Today: Approach to causal inference in econometrics
β’ Structure of ECO372
β’ Two frameworks to explain causality (not regressions)
β’ How I test students
β’ Causality beyond econometrics
Not today: settling debates
β’ reduced form vs. structural
β’ potential outcomes vs. causal graphs (DAGs)
![Page 5: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/5.jpg)
Bla
nch
ena
y C
oP
201
9
ECO372 APPLIED REGRESSION ANALYSIS
![Page 6: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/6.jpg)
Bla
nch
ena
y C
oP
201
9
ECO372 Applied Regression Analysis and Empirical Papers
2019H1 renamed : βData Analysis and Applied Econometrics in Practiceβ
Objective: quant. methods β applied empirical research
β’ Causal inference and identification
β’ Empirical strategies
β’ Stata, replication
![Page 7: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/7.jpg)
Bla
nch
ena
y C
oP
201
9
Motivation#1Getting teaching closer to practice
Many questions in economics are causal
Identification central in applied empirical work
Variety of empirical strategies (diff-in-diff, RDDβ¦)
Reliability of findings
Classic econometrics instruction does not focus on this
![Page 8: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/8.jpg)
Bla
nch
ena
y C
oP
201
9
Angrist & Pischke approach
Potential Outcomes (Roy-Rubin) causal model
Start with RCTs as βperfect settingβ
Regression to deal with selection on observables
Quasi-experimental approaches:
Instrumental variables
Diff-in-diff, segue into Panel Data
Regression Discontinuity
![Page 9: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/9.jpg)
Bla
nch
ena
y C
oP
201
9
Gauss-Markov assumptions failure
GM assumptions βClassicβ metrics Angrist Pischke
Linear model with mean zero errors
Get functional form right CEF Linear approximation
Errors are homoskedastic GLS βJust add βrobustββ
Errors are serially uncorrelated
GLS, time series β’ βJust add βrobustβββ’ Clustered SE in
clustered RCTs
(Errors are normally distributed)
Alternative estimators Focus on large sample / asymptotics
Exogeneity β’ Extensive discussion of measurement error
β’ Technical discussion of IV
β’ Small discussion of measurement error
β’ Extensive focus on empirical strategies that yield CIA
![Page 10: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/10.jpg)
Bla
nch
ena
y C
oP
201
9
Motivation #2Economics comparative advantage
Think hard about data!
Many disciplines do stats; not many causal inference
Big data not the solution to all problems
![Page 11: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/11.jpg)
Bla
nch
ena
y C
oP
201
9
Course structure
RCTs as βperfect settingβ
Regression to deal with selection on observables
Instrumental Variables
RCTs with imperfect compliance
Difference-in-differences
Regression Discontinuity
![Page 12: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/12.jpg)
Bla
nch
ena
y C
oP
201
9
Causal frameworks
Two causal frameworks as Ariadneβs threads
β’ Potential Outcomes
β’ Causal graphs (DAGs)
Emphasis on identification assumptions
![Page 13: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/13.jpg)
Bla
nch
ena
y C
oP
201
9
A different take on regressions
π¦π = πΌ + π½π·π + πΎππ + νπ
Start with binary treatment
Not all regressors equal
![Page 14: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/14.jpg)
Bla
nch
ena
y C
oP
201
9
Correlation and causation
Two distinct questions:
1. βIf there were a correlation between π· and π, would this represent the effect of π· on π?β
β’ Tools: assumptions, DAGs, sometimes regressions
2. βIs there a correlation between π· and π?β
β’ Tools: regressions, statistical inference
![Page 15: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/15.jpg)
Bla
nch
ena
y C
oP
201
9
![Page 16: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/16.jpg)
Bla
nch
ena
y C
oP
201
9
TWO CAUSAL FRAMEWORKS
Potential Outcomes, DAGs
![Page 17: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/17.jpg)
Bla
nch
ena
y C
oP
201
9
Combining two approaches
β’ Potential Outcomes (PO)
β’ Used by the textbook
β’ Causal graphs (mostly Pearl, 2000)
![Page 18: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/18.jpg)
Bla
nch
ena
y C
oP
201
9
Potential Outcomes (PO) / Roy-Rubin
β’ Binary treatment π·
β’ Potential Outcome: π0π if untreated ; π1π if treated ;
β’ Treatment effect: π1π β π0π
β’ But only observe either ππ0 or π1π
![Page 19: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/19.jpg)
Bla
nch
ena
y C
oP
201
9
Potential Outcomes (PO) / Roy-Rubin
Assume constant effect π½: ππ = πΌ + π½π·π + νπBaseline potential outcome: π0π = πΌ + νπ
Then:πΈ ππ π·π = 1 β πΈ ππ π·π = 0
observed
= π½ + πΈ νπ π·π = 1 β πΈ νπ π·π = 0πππ₯ππππ’π¨π§ ππ’ππ¬
Conditional Independence Assumption (CIA)πΈ νπ π·π = 0 = πΈ νπ π·π = 1
![Page 20: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/20.jpg)
Bla
nch
ena
y C
oP
201
9
All design assumptions deal with CIA
Diff-in-diff: common trend assumptionAbsent treatment, treated units would have evolved the same way as untreated units
π¦ππ‘ = πΌ + π½ β ππ πΈπ΄ππ + πΎ πππππ‘+πΏ ππ πΈπ΄π Γ ππππ ππ‘ + νππ‘
CIA : πΈ νππ‘ ππ πΈπ΄π, ππππ = πΈ[νππ‘]
![Page 21: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/21.jpg)
Bla
nch
ena
y C
oP
201
9
Directed Acyclical Graphs (DAGs)
β’ X has a (possible) effect on D, and on Y
β’ D has an effect of Y
β’ ν is unobserved (and has an effect on Y)β’ Typically omitted from graph
β’ Directed: causal relationships have a direction (effect of X on Y)
β’ Acyclical: Forbids cycles such asπ·
π
π
π·
π
π
ν
![Page 22: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/22.jpg)
Bla
nch
ena
y C
oP
201
9
π is a confounder(common cause)
π is a collider(common outcome)
Terminology
π·
π
π π·
π
π
![Page 23: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/23.jpg)
Bla
nch
ena
y C
oP
201
9
Causal paths
2 causal paths from D to Y:
Direct path: π· β π
Backdoor path: π· β π β π
π·
π
π
ν
![Page 24: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/24.jpg)
Bla
nch
ena
y C
oP
201
9
Open causal paths
β’ Open if either:
β’ There is no collider on the path
β’ There is a collider π, and we control / hold it constant π
π·
π
π π·
π
π
π
![Page 25: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/25.jpg)
Bla
nch
ena
y C
oP
201
9
Closed causal paths
β’ Closed if either:
β’ There is a collider on the path
β’ We control for a non-collider on the path
π·
π
π
π
π·
π
π
![Page 26: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/26.jpg)
Bla
nch
ena
y C
oP
201
9
A and B correlated because open paths
β’ π΄ β π΅
β’ π΄ β π· β π΅
Correlation between A and B does not represent only the direct effect of A on B
Open paths create correlations
π΄
πΆ
π΅
πΈ
π·
πΉ
![Page 27: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/27.jpg)
Bla
nch
ena
y C
oP
201
9
DAGs and identification
Backdoor criterion (sufficient)
The covariance between π· and π identifies the causal effect of π· on π if all backdoor paths from π· to π are closed.
Identification strategies try to rule out backdoor paths
![Page 28: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/28.jpg)
Bla
nch
ena
y C
oP
201
9
Different benefits
β’ Potential Outcomes
β’ Easy to talk about counterfactuals
β’ Neat interpretable algebra, formula for bias
β’ Causal graphs
β’ Visual
β’ Connects the assumptions of each empirical strategy
β’ Offers immediate reasoning about control variables
![Page 29: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/29.jpg)
Bla
nch
ena
y C
oP
201
9
Collider bias (~ βbad controlsβ)
β’ Controlling on a collider (common outcome) re-opens a causal path
Collider bias, βbad controlsβ, endog. selection bias, Simpsonβs paradox
(Conditional) Correlation between D and Y does not reflect causal effect of π· on π
π·
π
π
![Page 30: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/30.jpg)
Bla
nch
ena
y C
oP
201
9
Collider bias (~ βbad controlsβ)(1) (2)
SAT Maths SAT Maths
SAT Verbal 0.029 -0.251***
(0.0364) (0.0350)
Accepted 0.598***
(19.26)
Observations 800 800
(1)
SAT Maths
SAT Verbal 0.029
(0.0364)
Accepted
Observations 800
![Page 31: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/31.jpg)
Bla
nch
ena
y C
oP
201
9
2 frameworks = 2 Ariadneβs threads
Identification strategies rule out
β’ Violation of CIA
β’ Open backdoor paths
Examples:
β’ RCT
β’ Multivariate regression (control variables)
β’ Individual fixed effects
β’ Instrumental variables
![Page 32: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/32.jpg)
Bla
nch
ena
y C
oP
201
9
TESTING STUDENTS
![Page 33: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/33.jpg)
Bla
nch
ena
y C
oP
201
9
How I test studentsβ understanding
Questions on specific papers/studies
β’ If the researchers estimate Eq(1), would απ½ estimate the causal effect of π on π? Why or why not?
Make students create data and then run estimations
True or False questions (h/t Karen Bernhardt-Walther)
![Page 34: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/34.jpg)
Bla
nch
ena
y C
oP
201
9
![Page 35: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/35.jpg)
Bla
nch
ena
y C
oP
201
9
Effect of Facezon HQ on wages
Q1: Generate wages according to:
π€ππ‘ = 10 + 1.3 π»πππ‘+0.2 π¦ππππ‘ Γ πΆππ‘π¦π΄π + 0.6 π¦ππππ‘ Γ πΆππ‘π¦π΅π + νππ‘
Q2: You receive the data on wages in each city. How would you estimate the effect of Facezon HQ on wages? Estimate diff-in-diff:
π€ππ‘ = πΌ + π½πΆππ‘π¦π΅π + πΎ ππππ2016 π‘
+πΏ πΆππ‘π¦π΅ Γ ππππ2016 + π’ππ‘
Does your estimate απΏ correspond to what you expected from Q1? Why or why not?
![Page 36: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/36.jpg)
Bla
nch
ena
y C
oP
201
9
True/false questions
For an RCT on the effect of receiving food stamps on the decision to work, participants were recruited at Whole Foods and No Frills supermarkets. True or false? In that RCT, one should not control for the recruitment location, as this is a βbad controlβ.
The Ontario government considers offering a subsidy for childcare to families that fall below $40,000 of yearly joint income. True or false? Families are likely to under-report their income in order to qualify, but a researcher could always use an instrumental variable approach to estimate the effect of the childcare subsidy.
![Page 37: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/37.jpg)
Bla
nch
ena
y C
oP
201
9
Trade-offs
Few proofs & little maths (students selection)
Little discussion of heteroskedasticity
β’ Just add vce(robust) or vce(cluster β¦)
No time series
OLS only (IV as 2SLS)
Little discussion of heterogeneous effects
![Page 38: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/38.jpg)
Bla
nch
ena
y C
oP
201
9
CAUSAL INFERENCE BEYOND ECONOMETRICS
![Page 39: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/39.jpg)
Bla
nch
ena
y C
oP
201
9
Causal inference beyond econometrics
β’ An economic theory generates causal statements
β’ Empirics allow to sort between theories
β’ ββ Min wage β β unemploymentβ
β’ How would you test that?
![Page 40: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/40.jpg)
Bla
nch
ena
y C
oP
201
9
Example: Price elasticity
https://www.dropbox.com/s/8nujfq892ut5a37/Lecture%2016%20Estimating%20Elasticity.pptx?dl=
![Page 41: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/41.jpg)
Bla
nch
ena
y C
oP
201
9
Example: Demand elasticity
β’ We only observe equilibrium values of P,Q
β’ How can we find demand elasticity?
https://www.dropbox.com/s/8nujfq892ut5a37/Lecture%2016%20Estimating%20Elasticity.pptx?dl=
P
QQ1 Q2
P1
P2
![Page 42: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/42.jpg)
Bla
nch
ena
y C
oP
201
9
Example: Demand elasticity
Which is it?
https://www.dropbox.com/s/8nujfq892ut5a37/Lecture%2016%20Estimating%20Elasticity.pptx?dl=
P
QQ1 Q2
P1
P2
P
QQ1 Q2
P1
P2
Scenario 1:Less elastic demand, positive supply shock.
Scenario 2:More elastic demand, positive supply shock and negative demand shock.
D
D1
D2
S1
S2
S1
S2
![Page 43: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/43.jpg)
Bla
nch
ena
y C
oP
201
9
Example: Identifying demand elasticity
β’ Suppose you have information on average price and quantity of bread sold per month in Cleveland when there are 30 bakeries. Suppose three new bakeries open on April 1st, increasing supply for April.
β’ νπ· = β
ππ΄ππ βπππ΄π πππ΄π
ππ΄ππ βπππ΄π πππ΄π
P
QQ1 Q2
S1
S2
D
β’ If we are sure that an elasticity is estimated by an exogenous shock only to supply, we say it isidentified.
![Page 44: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/44.jpg)
Bla
nch
ena
y C
oP
201
9
N. Huntington-Klein ECO305Economics, Causality, and Analytics
β’ Focus on causal inference and programming
β’ No regression!
β’ Controlling done through subsamples/matching
![Page 45: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/45.jpg)
Bla
nch
ena
y C
oP
201
9
Some resources
Angrist-Pischke / Potential Outcomes
β’ Textbooks: Mostly Harmless Econometrics , Mastering βMetrics
β’ Angrist & Pischke (2017), Journal of Economic Perspectives, βThrough our classes darklyβ.
Directed Acyclical Graphs
β’ Scott Cunningham (regularly updated) βCausal Inference: The Mixtapeβ, particularly section 3: accessible intro to DAGs
β’ Nick Huntington-Klein ECO305: causal inference without regressions; causal graphs examples of common empirical strategies
β’ Morgan & Winship (2007, 2nd ed 2015) Counterfactuals and Causal Inference: combines Potential Outcomes & DAGs, focus on economics
β’ Judea Pearl, Causality (2000, 2nd ed 2009), particularly chapter 3: full formalism of DAGs & causal inference
![Page 46: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/46.jpg)
Bla
nch
ena
y C
oP
201
9
Thank you!
![Page 47: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/47.jpg)
Bla
nch
ena
y C
oP
201
9
EXAMPLES
![Page 48: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/48.jpg)
Bla
nch
ena
y C
oP
201
9
RCT
β’ Ensures πΈ νπ π·π = 1 = πΈ νπ π·π = 0
β’ CIA satisfied
β’ Ensures no causal path between π· and other covariates
Backdoor criterion
randomized π·
π
π
π1 π2
Back
![Page 49: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/49.jpg)
Bla
nch
ena
y C
oP
201
9
Control variables
β’ DGP: ππ = πΌ + π½π·π + πΎππ + ππβ’ Run: ππ = πΌ + π½π·π + π’πβ’ Omitted Variable Bias if πΈ π’π ππ = 1 β πΈ π’π ππ = 0
β’ Controlling for π closes causal path π· β π β π
π·
π
π
Back
![Page 50: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/50.jpg)
Bla
nch
ena
y C
oP
201
9
Fixed effects (within) in panel data
β’ Controlling for individual closes backdoor path
U
π
Individual
π
Time
π
Individual
π
Time
Back
![Page 51: Teaching CoP: Teaching Causality in ECO372homes.chass.utoronto.ca/~murdockj/econ-cop/Blanche...Teaching Causality in ECO372 P. Blanchenay Teaching and Learning Community of Practice](https://reader033.fdocuments.in/reader033/viewer/2022042303/5ece191ea7d13f3a6036f9e9/html5/thumbnails/51.jpg)
Bla
nch
ena
y C
oP
201
9
Instrumental Variables
Req1 (first stage): instrument π has an effect on π·
Req2 (exogeneity): π is as good as randomly assigned
No unobserved confounder between π and π
Req3 (exclusion restriction)
No other causal path between π and π
π·
π
ππ
Back