![Page 1: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/1.jpg)
Instrumental Variables Estimation
and Two Stage Least Squares
![Page 2: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/2.jpg)
2
Outline
1. Motivation: Omitted Variables In A Simple Regression Model
2. IV Estimation Of The Multiple Regression Model
3. Two Stage Least Squares
4. IV Solutions To Errors-In-Variables Problems
5. Testing For Endogeneity And Testing Overidentifying Restrictions
6. 2SLS With Heteroskedasticity
![Page 3: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/3.jpg)
3
Motivation
Why use instrumental variables?
Instrumental Variables (IV) estimation is used
when your model has endogenous 𝑥’s.
That is, whenever 𝐶𝑜𝑣(𝑥, 𝑢) ≠ 0.
Thus, IV can be used to address the problem of
omitted variable bias.
Additionally, IV can be used to solve the classic
errors-in-variables problem.
![Page 4: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/4.jpg)
4
Motivation
What is an instrumental variable?
In order for a variable, 𝑧, to serve as a valid
instrument for 𝑥, the following must be true:
The instrument must be exogenous.
That is, 𝐶𝑜𝑣 𝑧, 𝑢 = 0.
The instrument must be correlated with the
endogenous variable 𝑥.
That is, 𝐶𝑜𝑣(𝑧, 𝑥) ≠ 0.
![Page 5: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/5.jpg)
5
Motivation
How about validation of instrumental variable?
We have to use common sense and economic
theory to decide if it makes sense to assume
𝐶𝑜𝑣 𝑧, 𝑢 = 0.
![Page 6: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/6.jpg)
6
Motivation
How about validation of instrumental variable?
We can test if 𝐶𝑜𝑣(𝑧, 𝑥) ≠ 0.
Just testing 𝐻0: 𝜋1 = 0 𝑖𝑛 𝑥 = 𝜋0 + 𝜋1𝑧+ 𝑣.
Sometimes refer to this regression as the first-
stage regression.
![Page 7: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/7.jpg)
7
For 𝑦 = 𝛽0 + 𝛽1𝑥 + 𝑢, and given our assumptions, What is 𝛽1 in term of covariances?
𝐶𝑜𝑣(𝑧, 𝑦) = 𝛽1𝐶𝑜𝑣(𝑧, 𝑥) + 𝐶𝑜𝑣(𝑧, 𝑢), so
𝛽1 =𝐶𝑜𝑣(𝑧,𝑦)
𝐶𝑜𝑣(𝑧,𝑥).
IV Estimation in the Simple
Regression Case
![Page 8: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/8.jpg)
8
Then the IV estimator of 𝛽1 is?
𝛽 1 = (𝑧𝑖 − 𝑧 )(𝑦𝑖 − 𝑦 )𝑛
𝑖=1
(𝑧𝑖 − 𝑧 )(𝑥𝑖 − 𝑥 )𝑛𝑖=1
Is 𝛽 1(IV estimator) consistent?
A simple application of the law of large numbers shows that the IV estimator is consistent:
𝑝𝑙𝑖𝑚 𝛽 1 = 𝛽1.
When 𝑥 and 𝑢 are in fact correlated, IV estimator is biased. But it is consistent.
IV Estimation in the Simple
Regression Case
![Page 9: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/9.jpg)
9
The homoskedasticity assumption in this case?
𝐸 𝑢2 𝑧 = 𝜎2 = 𝑉𝑎𝑟(𝑢)
The asymptotic variance of 𝛽 1 is?
𝜎2
𝑛𝜎𝑥2 𝜌𝑥,𝑧
2
𝜎𝑥2 : population variance of 𝑥
𝜎2 : population variance of 𝑢
𝜌𝑥,𝑧2 : square of the population correlation between
𝑥 and 𝑧.
Inference with the IV Estimator
![Page 10: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/10.jpg)
10
Prove?
𝛽 1 = (𝑧𝑖−𝑧 )(𝑦𝑖−𝑦 )𝑛
𝑖=1
(𝑧𝑖−𝑧 )(𝑥𝑖−𝑥 )𝑛𝑖=1
= 𝑧𝑖−𝑧 𝑦𝑖
𝑛𝑖=1
(𝑧𝑖−𝑧 )(𝑥𝑖−𝑥 )𝑛𝑖=1
= 𝑧𝑖−𝑧 (𝛽0+𝛽1𝑥𝑖+𝑢𝑖)
𝑛𝑖=1
(𝑧𝑖−𝑧 )(𝑥𝑖−𝑥 )𝑛𝑖=1
= 𝛽1 + 𝑧𝑖−𝑧 𝑢𝑖
𝑛𝑖=1
(𝑧𝑖−𝑧 )(𝑥𝑖−𝑥 )𝑛𝑖=1
→ 𝑉𝑎𝑟 𝛽 1 = 𝑧𝑖−𝑧 2𝑛
𝑖=1 𝑉𝑎𝑟(𝑢𝑖)
[ (𝑧𝑖−𝑧 )(𝑥𝑖−𝑥 )] 𝑛𝑖=1
2
=𝜎2
(𝑧𝑖−𝑧 )(𝑥𝑖−𝑥 ) 𝑛𝑖=1
𝑧𝑖−𝑧 2
𝑥𝑖−𝑥 2
2
.𝑛 𝑥𝑖−𝑥
2
𝑛
= 𝜎2
𝜌 𝑥,𝑧2 𝑛𝜎 𝑥
2
→ 𝑝𝑙𝑖𝑚 𝑉𝑎𝑟[𝛽 1] =𝜎2
𝑛𝜎𝑥2 𝜌𝑥,𝑧
2
Inference with the IV Estimator
![Page 11: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/11.jpg)
11
Asymptotic standard error of 𝛽 1?
𝜎 2
𝑆𝑆𝑇𝑥 . 𝑅𝑥,𝑧2
𝜎 2 = 1
𝑛−2 𝑢 𝑖
2𝑛𝑖=1
𝑆𝑆𝑇𝑥 : total sum of squares of the 𝑥𝑖.
𝑅𝑥,𝑧2 : R-squared from regression of 𝑥𝑖 on 𝑧𝑖.
Inference with the IV Estimator
![Page 12: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/12.jpg)
12
Differences between IV and OLS estimation?
Standard error in IV case differs from OLS only in
the 𝑅2 from regressing 𝑥 on 𝑧.
Since 𝑅2 < 1, IV standard errors are larger.
However, IV is consistent, while OLS is
inconsistent, when 𝐶𝑜𝑣(𝑥, 𝑢) ≠ 0.
The stronger the correlation between 𝑧 and 𝑥, the
smaller the IV standard errors.
IV versus OLS estimation
![Page 13: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/13.jpg)
13
What if our assumption that 𝐶𝑜𝑣(𝑧, 𝑢) = 0 is false?
The IV estimator will be inconsistent, too.
Can compare asymptotic bias in OLS and IV.
The Effect of Poor Instruments
![Page 14: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/14.jpg)
14
𝐼𝑉: 𝑝𝑙𝑖𝑚 𝛽 1 =?
𝑝𝑙𝑖𝑚 𝛽 1 = 𝛽1 +𝐶𝑜𝑟𝑟(𝑧,𝑢)
𝐶𝑜𝑟𝑟(𝑧,𝑥).𝜎𝑢
𝜎𝑥
𝑂𝐿𝑆: 𝑝𝑙𝑖𝑚 𝛽 1 =?
𝑝𝑙𝑖𝑚 𝛽 1 = 𝛽1 + 𝐶𝑜𝑟𝑟(𝑥, 𝑢).𝜎𝑢
𝜎𝑥
Prefer IV if 𝐶𝑜𝑟𝑟 𝑧,𝑢
𝐶𝑜𝑟𝑟 𝑧,𝑥 < 𝐶𝑜𝑟𝑟(𝑥, 𝑢).
The Effect of Poor Instruments
![Page 15: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/15.jpg)
15
Outline
1. Motivation: Omitted Variables In A Simple Regression Model
2. IV Estimation Of The Multiple Regression Model
3. Two Stage Least Squares
4. IV Solutions To Errors-In-Variables Problems
5. Testing For Endogeneity And Testing Overidentifying Restrictions
6. 2SLS With Heteroskedasticity
![Page 16: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/16.jpg)
16
IV estimation can be extended to the multiple regression case.
the structural model: the model we are interested in estimating.
Our problem is that one or more of the variables are endogenous.
We need an instrument for each endogenous variable.
IV Estimation in the Multiple
Regression Case
![Page 17: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/17.jpg)
17
Write the structural model as
𝑦1 = 𝛽0 + 𝛽1𝑦2 + 𝛽2𝑧1 + 𝑢1,
𝑦2 : endogenous
𝑧1 : is exogenous
So, what to do?
Let 𝑧2 be the instrument, so 𝐶𝑜𝑣(𝑧2, 𝑢1) = 0 and
𝑦2 = 𝜋0 + 𝜋1𝑧1 + 𝜋2𝑧2 + 𝑣2, where 𝜋2 ≠ 0.
This reduced form equation regresses the endogenous variable on all exogenous ones.
IV Estimation in the Multiple
Regression Case
![Page 18: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/18.jpg)
18
Outline
1. Motivation: Omitted Variables In A Simple Regression Model
2. IV Estimation Of The Multiple Regression Model
3. Two Stage Least Squares
4. IV Solutions To Errors-In-Variables Problems
5. Testing For Endogeneity And Testing Overidentifying Restrictions
6. 2SLS With Heteroskedasticity
![Page 19: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/19.jpg)
19
It’s possible to have multiple instruments.
Consider our original structural model
𝑦1 = 𝛽0 + 𝛽1𝑦2 + 𝛽2𝑧1 + 𝑢1,
Let
𝑦2 = 𝜋0 + 𝜋1𝑧1 + 𝜋2𝑧2 + 𝜋3𝑧3 + 𝑣2
Here we’re assuming that both 𝑧2 and 𝑧3 are valid
instruments – they do not appear in the structural
model and are uncorrelated with the structural
error term, 𝑢1.
Two Stage Least Squares (2SLS)
![Page 20: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/20.jpg)
20
Could use either z2 or z3 as an instrument.
Best instrumental variable?
The best instrument is a linear combination of all of the exogenous variables,
y2* = p0 + p1z1 + p2z2 + p3z3 .
We can estimate y2* by regressing y2 on z1, z2 and z3 – can call this the first stage.
If then substitute ŷ2 for y2 in the structural model, get same coefficient as IV.
Best Instrument
![Page 21: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/21.jpg)
21
While the coefficients are the same, the standard
errors from doing 2SLS by hand are incorrect, so
let Stata do it for you.
Method extends to multiple endogenous variables
– need to be sure that we have at least as many
excluded exogenous variables (instruments) as
there are endogenous variables in the structural
equation.
More on 2SLS
![Page 22: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/22.jpg)
22
Outline
1. Motivation: Omitted Variables In A Simple Regression Model
2. IV Estimation Of The Multiple Regression Model
3. Two Stage Least Squares
4. IV Solutions To Errors-In-Variables Problems
5. Testing For Endogeneity And Testing Overidentifying Restrictions
6. 2SLS With Heteroskedasticity
![Page 23: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/23.jpg)
23
Remember the classical errors-in-variables
problem where we observe 𝑥1 instead of 𝑥1∗.
Where 𝑥1 = 𝑥1∗ + 𝑒1, and 𝑒1 is uncorrelated with
𝑥1∗ and 𝑥2.
If there is a 𝑧, such that 𝐶𝑜𝑟𝑟(𝑧, 𝑢) = 0 and
𝐶𝑜𝑟𝑟(𝑧, 𝑥1) ≠ 0, then
IV will remove the attenuation bias.
Addressing Errors-in-Variables with
IV Estimation
![Page 24: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/24.jpg)
24
Outline
1. Motivation: Omitted Variables In A Simple Regression Model
2. IV Estimation Of The Multiple Regression Model
3. Two Stage Least Squares
4. IV Solutions To Errors-In-Variables Problems
5. Testing For Endogeneity And Testing Overidentifying Restrictions
6. 2SLS With Heteroskedasticity
![Page 25: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/25.jpg)
25
Since OLS is preferred to IV if we do not have an
endogeneity problem, then we’d like to be able to
test for endogeneity.
If we do not have endogeneity, both OLS and IV
are consistent.
Idea of Hausman test is to see if the estimates
from OLS and IV are different.
Testing for Endogeneity
![Page 26: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/26.jpg)
26
While it’s a good idea to see if IV and OLS have
different implications, it’s easier to use a
regression test for endogeneity.
If 𝑦2 is endogenous, then 𝑣2 (from the reduced
form equation) and 𝑢1 from the structural model
will be correlated.
The test is based on this observation.
Testing for Endogeneity
![Page 27: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/27.jpg)
27
How to do the test?
Save the residuals from the first stage.
Include the residual in the structural equation (which of course has y2 in it).
If the coefficient on the residual is statistically different from zero, reject the null of exogeneity.
If multiple endogenous variables, jointly test the residuals from each first stage.
Testing for Endogeneity
![Page 28: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/28.jpg)
28
If there is just one instrument for our endogenous variable, we can’t test whether the instrument is uncorrelated with the error.
We say the model is just identified.
If we have multiple instruments, it is possible to test the overidentifying restrictions – to see if some of the instruments are correlated with the error.
Testing Overidentification
Restrictions
![Page 29: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/29.jpg)
29
How to test overidentification?
Estimate the structural model using IV and obtain
the residuals.
Regress the residuals on all the exogenous
variables and obtain the R2 to form nR2
Under the null that all instruments are uncorrelated
with the error, LM ~ cq2 where q is the number of
extra instruments.
The OverID Test
![Page 30: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/30.jpg)
30
Outline
1. Motivation: Omitted Variables In A Simple Regression Model
2. IV Estimation Of The Multiple Regression Model
3. Two Stage Least Squares
4. IV Solutions To Errors-In-Variables Problems
5. Testing For Endogeneity And Testing Overidentifying Restrictions
6. 2SLS With Heteroskedasticity
![Page 31: Instrumental Variables Estimation and Two Stage Least Squaresgsme.sharif.edu/~barakchian/courses/EconometricsI... · Best instrumental variable? The best instrument is a linear combination](https://reader034.vdocument.in/reader034/viewer/2022050403/5f80b60a2fe67b060672d6af/html5/thumbnails/31.jpg)
31
When using 2SLS, we need a slight adjustment to
the Breusch-Pagan test.
How to test for heteroskedasticity?
Get the residuals from the IV estimation.
Regress these residuals squared on all of the
exogenous variables in the model (including the
instruments).
Test for the joint significance.
Testing for Heteroskedasticity