Upload
others
View
14
Download
0
Embed Size (px)
Citation preview
Lecture 2: Diff-in-Diff and Instrumental Variables
January 14, 2016
Difference-in-Differences Instrumental Variables
Causal Inference
There are 5 basic empirical methods to obtain causal inference:
1 Controls (includes matching/fixed-effects)2 Randomized Experiments3 Difference-in-Differences4 Instrumental Variables5 Regression Discontinuity
Difference-in-Differences Instrumental Variables
Difference-in-Differences
Difference-in-differences (DD) extends the idea behind individualfixed-effects
Instead of just comparing before and after, it compares before andafter among a ‘treated’ and a ‘control’ group
In fact, fixed-effects is a difference; DD just adds another layerSome people add even another layer, making it adifference-in-difference-in-differences
Difference-in-Differences Instrumental Variables
Example
For a little variety, I will do different examples today than my usualeducation cases
Consider the question of immigration
Difference-in-Differences Instrumental Variables
A View on Immigration
Ted Cruz (Nov 10, 2015 GOP debate): “I will say, the politics of it[immigration] would be very, very different if a bunch of lawyers orbankers were crossing the Rio Grande. Or if a bunch of people withjournalism degrees were coming over and driving down the wages inthe press, then we would see stories about the economic calamity thatis befalling our nation.”
Difference-in-Differences Instrumental Variables
Immigration and Wages
This is one of the major arguments against immigration: thatimmigrants will drive down the wages of locals
Especially in the low-skilled sector (depending on the type ofimmigrant)
We would like to look at this empirically
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does immigration drive down the wages (or theemployment rate) of locals?
1 What is the unit of analysis?
2 What is the treatment?3 What outcome are we interested in?4 What are the counterfactual outcomes?5 What is the causal link?6 How could we mimic this? Can we do a randomized experiment?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does immigration drive down the wages (or theemployment rate) of locals?
1 What is the unit of analysis?2 What is the treatment?
3 What outcome are we interested in?4 What are the counterfactual outcomes?5 What is the causal link?6 How could we mimic this? Can we do a randomized experiment?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does immigration drive down the wages (or theemployment rate) of locals?
1 What is the unit of analysis?2 What is the treatment?3 What outcome are we interested in?
4 What are the counterfactual outcomes?5 What is the causal link?6 How could we mimic this? Can we do a randomized experiment?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does immigration drive down the wages (or theemployment rate) of locals?
1 What is the unit of analysis?2 What is the treatment?3 What outcome are we interested in?4 What are the counterfactual outcomes?
5 What is the causal link?6 How could we mimic this? Can we do a randomized experiment?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does immigration drive down the wages (or theemployment rate) of locals?
1 What is the unit of analysis?2 What is the treatment?3 What outcome are we interested in?4 What are the counterfactual outcomes?5 What is the causal link?
6 How could we mimic this? Can we do a randomized experiment?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does immigration drive down the wages (or theemployment rate) of locals?
1 What is the unit of analysis?2 What is the treatment?3 What outcome are we interested in?4 What are the counterfactual outcomes?5 What is the causal link?6 How could we mimic this? Can we do a randomized experiment?
Difference-in-Differences Instrumental Variables
The Mariel Boatlift
What was the Mariel Boatlift?
Difference-in-Differences Instrumental Variables
DD
The DD term refers to two levels of differenceThe first difference is (almost) always timeThe second difference can be: schools, grades, geography, individuals,etc.
Difference-in-Differences Instrumental Variables
First-Difference
Let’s look at the first-difference of timeNotation: Y1,city ,t is treated city at time tY0,city ,t is control city at time t
What outcomes do we observe?
Write out the treatment effect and the selection bias:
Difference-in-Differences Instrumental Variables
Second-Difference I
What is a reasonable second layer of difference?Two possible:
David Card (1990) v. George Borjas (2015)
Difference-in-Differences Instrumental Variables
Second-Difference II
We now incorporate the second layer of difference
What outcomes do we observe:
What idea can we use to mimic the counterfactual?
Difference-in-Differences Instrumental Variables
Estimating Equations
Y1,miami ,1981 − Y0,atlanta,1981 − [Y0,miami ,1979 − Y0,atlanta,1979]
In regression form:
Open up STATA!
Difference-in-Differences Instrumental Variables
Interpretation of DD
How do we interpret the DD estimate?
We call this a Average Treatment Effect (ATE)
Difference-in-Differences Instrumental Variables
Internal Validity I
What is our key assumption?
Visual Representation:
Difference-in-Differences Instrumental Variables
Internal Validity II
How do we check our identifying assumption? Possible failures?Ashenfelter dipPolitical Endogeneity (i.e. Reverse Causality)
How can we weaken the key assumption?
http://www.motherjones.com/kevin-drum/2015/09/another-shot-fired-great-immigration-vs-wages-war
Difference-in-Differences Instrumental Variables
External Validity
Generalizability?
General Equilibrium Effects?
Difference-in-Differences Instrumental Variables
Pros and Cons of DD
Pros:
Cons:
Difference-in-Differences Instrumental Variables
Causal Inference
There are 5 basic empirical methods to obtain causal inference:
1 Controls (includes matching/fixed-effects)2 Randomized Experiments3 Difference-in-Differences4 Instrumental Variables5 Regression Discontinuity
Difference-in-Differences Instrumental Variables
Basic Intuition
Basic idea: In a randomized experiment, we randomized people to‘treatment’ and ‘control’
What happens if some external ‘thing’ (geography, mass layoff,storms, etc) strikes some people and not others (at random!)
Well the external ‘thing’ has randomized for us! We call this anInstrumental Variable (IV)Randomized experiments can thus be treated as IVs
Difference-in-Differences Instrumental Variables
Example I
Does earlier colonization improve economic outcomes?
Problem: European empires likely colonized the ‘best’ places (mostfertile, etc) first
Solution: Ships had to sail with the wind
Difference-in-Differences Instrumental Variables
Example II
Thought Experiment:
Compare two islands: Guam and Fefan (in Micronesia)
Guam was directly on the East-West route across the pacific (used byMagellan)
Fefan was not on this route (was on the much more difficultWest-East route)
https://www.google.ca/maps/dir/Guam/Fefan,+Federated+States+of+Micronesia/@10.3941378,130.3823886,4z/data=!4m13!4m12!1m5!1m1!1s0x671f76ff930f24ef:0x5571ae91c5b3e5a6!2m2!1d144.793731!2d13.444304!1m5!1m1!1s0x6667a4d6a8ce100d:0xc47882f565ab012a!2m2!1d151.8379961!2d7.3487617
Difference-in-Differences Instrumental Variables
Example III
Therefore, Guam was colonized before Fefan
‘Wind’ randomized the colonization for us!
Difference-in-Differences Instrumental Variables
IV Basics I
The (basic) math:
What outcomes do we observe?
What is our treatment and selection effect?
Difference-in-Differences Instrumental Variables
IVs Basics II
The (more complex) math:
In IVs, randomization is not perfect
Fefan could have been discovered ‘by luck’ earlierFor example, Pitcairn was discovered earlier even though it was not onthe main wind routeThis was because of the mutiny on the ship “HMS Bounty”
We account for this by dividing by the probability randomizationaffected your treatment status:
Difference-in-Differences Instrumental Variables
IV Notation
Notation:
We call the instrument (here wind patterns) Z
We call the ‘endogenous regressor’ (here colonization date) X
Difference-in-Differences Instrumental Variables
IV Assumptions
For an IV to be valid it must be both:
Relevant =⇒ Corr(X ,Z ) 6= 0All this says is that there was randomization (i.e. Islands on favorablewind routes were colonized first)We can check this through the equation:ColonizationDateisland,t = α + βFavorableWindisland,t + εit
Exclusion =⇒ Corr(Z , ε) = 0This says that randomization was ‘proper’ (i.e. Islands on favorablewind routes had no other advantage (more fertile soil, etc.))This assumption cannot be tested as ε is unobserved
Do these assumptions seem likely to hold in this case?
Difference-in-Differences Instrumental Variables
IV as a Regression
We have two terms to estimate: E[D1 − D0] (first-stage) andE[Y1 − Y0|Z = 1] (reduced-form)
First-stage is the effect of Z on X (i.e. how much more likely was it forislands on favorable wind routes to be colonized first)Reduced-form is the effect of Z on Y (i.e. how much better off areislands on favorable wind routes?)
First-stage:ColonizationDateisland ,t = α + β1FavorableWindisland ,t + εit
In general: Xi = α + β1Zi + εi
Reduced-form:EconomicOutcomesisland ,t = α + β2FavorableWindisland ,t + εit
In general: Yi = α + β2Zi + εi
The IV estimate is then: β2β1
Difference-in-Differences Instrumental Variables
For the Keeners
In the last slide, we got the Wald estimator
However, when implementing IVs we use the IV estimatorIntuition is identical; the IV estimator is just more efficient
Only difference is that in the reduced-form regression we plug in a‘predicted’ X rather than Z
First-stage: X̂i = α + β1Zi + εi
Reduced-form: Yi = α + β2X̂i + εi
In STATA you do all this by simply typing: “ ivregress 2sls y (x=z),vce(robust) first”
Difference-in-Differences Instrumental Variables
Another IV
Our question is: Does having another child affect a mother’s laboursupply?
OLS: HasWorkedi = α + β#Kidsi + εi
Possible failures of OLS in this instance?
Idea: Use the fact parents like to have a boy AND a girl in theirfamily (Angrist and Evans, 1998)
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does having another child affect the mother’s laboursupply?
1 What is the unit of analysis?
2 What is the treatment?3 What outcome are we interested in?4 What are the counterfactual outcomes?5 What is the causal link?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does having another child affect the mother’s laboursupply?
1 What is the unit of analysis?2 What is the treatment?
3 What outcome are we interested in?4 What are the counterfactual outcomes?5 What is the causal link?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does having another child affect the mother’s laboursupply?
1 What is the unit of analysis?2 What is the treatment?3 What outcome are we interested in?
4 What are the counterfactual outcomes?5 What is the causal link?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does having another child affect the mother’s laboursupply?
1 What is the unit of analysis?2 What is the treatment?3 What outcome are we interested in?4 What are the counterfactual outcomes?
5 What is the causal link?
Difference-in-Differences Instrumental Variables
Framing the Research Question
Our question is: Does having another child affect the mother’s laboursupply?
1 What is the unit of analysis?2 What is the treatment?3 What outcome are we interested in?4 What are the counterfactual outcomes?5 What is the causal link?
Difference-in-Differences Instrumental Variables
To STATA
How do we mimic this? (think of the IV as having your first 2 kids bethe same gender)
Open up STATA
Difference-in-Differences Instrumental Variables
Interpretation of IVs
How do we interpret the IV estimate? (i.e. who complies withtreatment?)
We call this a Local Average Treatment Effect (LATE)
This is (in my view) the biggest weakness of IVs
Difference-in-Differences Instrumental Variables
Internal Validity
Relevance?
Exclusion?
Difference-in-Differences Instrumental Variables
External Validity
Generalizability?
Mechanisms?
Difference-in-Differences Instrumental Variables
Pros and Cons of IV
Pros:
Cons: