Chapter Four: Endogenity, IV

Regression and Simultaneous Equation

Models(SEMs)

❑ Endogeneity
❑Simultaneous Equations Models (SEMs)

❑Instrumental Variables (IV) Regression

❑ Two Stage Least Squares (2SLS) Estimation

 Endogenity
❖ One of the assumptions of the classical linear regression models is that the
error term (u) does not correlate with any of the explanatory variables. i.e.,
E (x’u) = 0
❖ If this assumption is violated (E(x’u) ≠ 0), OLS estimates are biased and
inconsistent; hence, there is endogeneity problem
Three important sources/threats to endogeneity(internal validity) are:
1. omitted variable (unobserved heterogeneity)- a variable that is correlated
with X but is unobserved, so cannot be included in the regression and
resulted in omitted variable bias;
2. Measurement error- errors-in-variables bias (X is measured with error).
3. Simultaneity in SEMs/reverse causality- simultaneous causality bias (X
causes Y , Y causes X)
❖ A model can have more than one source of endogeneity.
 Cont’d
Terminologies: Endogeneity and Exogeneity
❑ An endogenous variable is one that is correlated with u.

❑ An exogenous variable is one that is uncorrelated with u.

 Endogenous literally means “determined within the system,” that is, a

variable that is jointly determined with Y , or, a variable subject to
simultaneous causality.

 However, this definition is narrow and IV regression can be used to

address OV bias and errors-in-variable bias, not just to simultaneous
causality bias.
 Cont’d---
 If an explanatory variable (xi) is correlated with the error term for whatever
reason, the variable is traditionally called endogenous explanatory variable.
 In strict sense, the term 'endogenous‘ refers to the case where there is
simultaneity.
❖ Let’s see each sources in bird’s eye view
1. Omitted variables
- If the true model underlying the data is
y = a + b1x1 + b2x2 + b3x3 + n (4.1)
but you estimate the model
y = a + b1x1 + b2x2 + e (4.2)
then variable x1 will be endogenous if it is correlated with x3. Why?
Because e = f (n, x3).
 Cont’d---
2. Measurement error
Suppose the true model underlying the data is
y = a + b1x1 + b2x2 + e (4.3)
but you estimate the model
y = a + b1x1 + b2x2* + e (4.4)
where (x2* = x2 + j).
❖ Variable x2 will be endogenous if j depends on x2 . As x2 grows, so
does j. Thus e is correlated with x2, causing endogeneity.
Rearranging the equation, we have
y = a + b1x1 + b2x2* + e
y = a + b1x1 + b2(x2 + j) + e
y = a + b1x1 + b2x2 + (e + b2 j) (4.5)
If j = f(x ) then error term is correlated with x causing endogeneity.
 Cont’d---
➢ In a special case where an omitted variable is a function of an explanatory
variable in the model, we encounter a functional form misspecication.

3. Simultaneity
❑ A system of simultaneous equations occurs when two or more left hand
side variables are functions of each other:
y1 = a + b1x1 + g2y2 + e
y2 = a + g1x1 + g2y1 + e (4.6)
- With some algebra you can rewrite these two equations in “reduced form” as
a single equation with an endogenous regressor
 Or Simultaneity arises when one or more of the independent variables, Xj’s,
is jointly determined with the dependent variable, Y , typically through an
equilibrium mechanism.
❖ If there is no endogeneity, it is more efficient to use OLS. If there is
endogeneity, OLS is inconsistent and so IV using 2SLS is best
 Dealing with Endogenity
 Now let’s see each in a bit detail
 Unobserved heterogeneity problem can be handled using the ff options:
1. Find additional data so that every relevant variable is included.
2. Ignore it
- Acceptable only if omitted variable is uncorrelated with all
included variables; otherwise the coefficient estimates will be biased
up or down.
3. Try to get an appropriate proxy variable
4. Use instrumental variable(s) and two-stage least squares estimation
5. Use the fixed effects method or the first differenced model (panel data)
 Cont’d---
 These methods can eliminate bias when E(u|X) ≠ 0
 Suppose we want to estimate a model y = f (x; q) where x is the variable of
interest and q is control.
 If we cannot obtain data for or observe q, the variable q will be omitted. In a
sense, q will be part of the error term .
 If the omitted variable q is correlated with x, then it will be correlated with u .
That is cov (x; u) or E(u|X) ≠ 0 and hence x is endogenous.
 Consider the following regression
log(wage) = b 0 + b1educ + ( b 2 abil + e) (4.7)

 Since ability is not observed, we can uonly run the following regression
log(wage) = b 0 + b1educ + u (4.8)
 Since ability is which is correlated with education is ommited, education is
endogenous (i.e, correlated with u). Thus, bˆ will be biased
1
 Ability also can affect wage through education.
 Cont’d---
❖ To eliminate the bias we can employ the following three options:
I. Plug in/find the proxy variable; for example IQ for ability
Suppose y is the outcome, q is the omitted variable and z is the proxy for
q. What properties should the proxy z have?
a. Proxy z should be strongly correlated with q.
b. Proxy z must be redundant (ignorable)
E (y | x, q, z) = E (y | x, q) (4.9)
c. Omitted q must be uncorrelated with other regressors conditional on z:
(corr (q , xj) = 0 | z) for each xj
❑ The last two mean roughly that q and z provide similar information
about the outcome
 Cont’d---
❑ What if proxy variable z is correlated with a regressor x?
➢ OLS is inconsistent, but one can hope and argue that the
inconsistency is less than if z is omitted.
❑ Consider using a lagged dependent variable as a proxy variable.
Example: If you believe that omitted variable qt strongly affects
outcome yt, then a lagged value of y (such as yt-2) is probably
correlated with qt as well.
Problem: yt-2 may be correlated with other x’s as well, leading to
inconsistency
❑ Consider using multiple proxy variables for a single omitted
variable
- Simply put all proxy variables in the equation that meet the
requirements for proxies.
 Cont’d---
❑ What if omitted variable q interacts with a regressor x?
y = a + b1x + b2q + b3qx + e (4.10)
∂y/∂x = b1+ b3q
➢ marginal effect of x on y involves q, which is unobserved
 If we can find a good proxy, we would estimate the model using OLS
 Whenever there is no appropriate proxy variable, consistent estimation of the
parameters β0 and β1 requires an instrumental variable.
 Difference between IV and Proxy?
- With IV we will leave the unobserved variable in the error term but use an estimation
method that recognizes the presence of the omitted variable
- With a proxy we were trying to remove the unobserved variable from the error term e.g.
IQ. IQ would make a poor instrument as it would be correlated with the error in our
model (ability in u)
- Need something correlated with education but uncorrelated with ability (parents
education?)
 Cont’d---
II. Instrumental variable approach
❖ For an instrumental variable (an instrument) Z to be valid, it must satisfy two
conditions:
 Instrument exogeneity or Validity Condition: Cov(Zi , ui )= 0
◼ All the instruments are uncorrelated with the error term:

◼ Instrument z should have no partial effect on y and must be uncorrelated with the
error term (z is exogenous).
◼ Must be uncorrelated with potential unobserved determinants of the y variable
 Instrument relevance or Informativeness Condition: Cov(Zi , Xi )≠0
✓ Instrument (z) must be correlated with the troublesome/endogenous variable (x)
✓ This condition can be tested given the random sample from the population.
❑ The second requirement above is also called the identification requirement.
When it does not hold the parameters are not identified. Which means that we
can not calculate unique values for the parameters
 Cont’d---
❖ Thus, if the above both conditions are satisfied, we call z an instrumental
variable
 There are two ways to intuitively understand these conditions.
1. Instrumental variable is a variable that is not correlated with the
omitted variable, but is correlated with the endogenous explanatory
variable.
✓ Instrument would have to be correlated with education level of a
person, but not their ability level
2. Instrumental variable is a variable that affects y only through x.
 The condition Cov(z,u)=0 involves unobserved u. Therefore, we cannot
test this condition. But, when you have extra instrumental variables, you
can test this. This will be discussed later.
 The condition Cov(z,x)≠0 is easy to test(test for instrumental relevance).
 Testing Over identifying Restrictions: The LM or
J-statistic test (Test for exogenity)-for condition 1
 If there is just one instrument for our endogenous variable (just
identified), we can’t test whether the instrument is uncorrelated with
the error
 If we have multiple instruments, it is possible to test the
over identifying restrictions (partially - for instrument exogeneity)- to see
if some of the instruments are correlated with the error
 Estimate the structural model using IV (2SLS) 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 (m-k df)
 Cont’d---
 Just run the OLS:
x=π0+π1z+v (4.11)
❑ Then test: H0: π1=0.
❑ Thus, we should be able to reject the H0 against HA
 We need an instrument z for education which is:
1. not correlated with ability (and other omitted variables)
2. correlated with education.
 IQ cannot be a good instrument for education because it is correlated
with ability.
 Mother's education could satisfy the second condition but could be
correlated with child's ability.
 Number of siblings while growing up might be an instrument.
Measuring the strength of instruments in practice:
The first-stage F-statistic (Test for relevance)- for
condition 2
 The first stage regression:
Regress X on Z1 , … , Zm , W1 , … , Wr .
 Totally irrelevant instruments iff all the coefficients on Z1 , … , Zm are
zero.
 The first-stage F-statistic tests the hypothesis that Z1 , … , Zm do not
enter the first stage regression.
 If instruments are weak, then the TSLS estimator is biased and the
and statistical inferences (standard errors, hypothesis tests, confidence
intervals) can be misleading and also t-statistic has a non-normal
distribution
 Weak instruments imply a small first stage F-statistic.
 Compute the first-stage F-statistic.
 Cont’d---
 Rule-of-thumb:
- If F > 10, instruments are strong - use TSLS.
- But if the first stage F-statistic is less than 10, then the set of
instruments is weak (take some action).
 What to do if you have weak instruments?
- Get better instruments
- If you have many instruments, some are probably weaker than
others, then it’s a good idea to drop the weaker ones (dropping an
irrelevant instrument will increase the first-stage F).
 The stronger the correlation between the instruments and the explanators
(strong instrument), the more efficient IV is.
 If the correlation between Z and X is too low, then Z is a weak instrument,
and 2SLS is not a helpful procedure
 The IV Estimation: One explanatory variable (X)-
one instrument (Z) case
Two Stage Least Squares (TSLS)
As it sounds, TSLS has two stages - two regressions:
(1)First isolates the part of X that is uncorrelated with u:
Regress X on Z using OLS:
X=π0+π1z+v (4.12)
 Because Zi is uncorrelated with ui , is
uncorrelated with ui . We don’t know or but we have
estimated them.
 Compute the predicted values of Xi , , where
, i = 1, ... , n.
 Cont’d---
(2) Replace Xi by in the regression of interest:
regress Y on using OLS:
(4.13)
 Because is uncorrelated with ui in large samples, so the first least squares
assumption holds.
 If the instruments are not uncorrelated with the error term, the first stage
of TSLS doesn’t successfully isolate a component of X that is uncorrelated
with the error term, so is correlated with u and TSLS is inconsistent.
 Thus can be estimated by OLS using regression (2).
 This argument relies on large samples (so and are well estimated
using regression (1)).
 This resulting estimator is called the Two Stage Least Squares (TSLS)
estimator, .
 Cont’d---
Suppose you have a valid instrument, Zi .
 Stage 1:Regress Xi on Zi , obtain the predicted values .
 Stage 2:Regress Yi on , the coefficient on is the TSLS
estimator, .
Then is a consistent estimator of .
 Now, consider: y = b 0 + b1 x + u
 Then we have: Cov(z,y)=Cov(z, β0+β1x+u)
So we have, Cov(z,y)= β1Cov(z,x)+Cov(z,u)
Since Cov(z,u)=0, we have
Cov( z , y ) (4.14)
b1 =
Cov( z , x )
 β1 is thus identified in the sense that it is expressed in the population
moments (covariances) that can be estimated using sample data.
 Cont’d---
 By replacing the population covariances (Cov(z,y) and Cov(z,x))with their
sample covariances, we have the instrumental variable estimator of β1 which
is given nby
 ( zi − z )( yi − y ) ziYi (4.15)
b̂1 = i =1 bˆ IV =
n zi xi
 ( zi − z )( xi − x )
i =1
 The IV estimator of β0 is obtained as usual using means of Y and X
 You can easily show thatbˆ1 is a consistent estimator of β1.
 Consistency of the TSLS estimator

The sample covariances are consistent:

Thus, (4.16)
 The instrument relevance condition, , ensures that you don’t
divide by zero.
 When Z=X we obtain the OLS estimator of β1, i.e. when x is exogenous, it is its own
‘IV'.
 2SLS/IV Estimation (Vector approach)
▪ The 2SLS model
y = xb + u x = z + v
▪ First-Stage estimate the effect of the IVs on the troublesome variable.
▪ Second-Stage, we use predicted-X in determining y

xˆ = zˆ = z ( z ' z ) −1 z ' x y = xˆb 2 SLS + u

bˆ 2 SLS
−1
= ( xˆ ' xˆ ) xˆ ' y
ˆ Cov( z, y )
b IV = ( z ' x) z ' y =
−1

Cov( z , x)
 Cont’d---
❖ Approach 3: Difference it out
❑ Suppose that the endogeneity is fixed over time, such as
measurement error or an omitted variable. Further, suppose that
observe data in two time periods.
❑ A difference-in-difference (DD) model can be used: subtract values
at time 1 (“before”) from values at time 2 (“after”) and the
endogenous variable will drop out.
Limitations:
- DD models will not eliminate selection bias.
- DD models only eliminate fixed variables; sometimes endogenous
variables change values over time
 Statistical inference with IV:
Homoskedasticity case
 Homoskedasticity assumption in the case of IV regression is stated in
terms of z: E(u2|z)=σ2
 It can be shown that the asymptotic variance of bˆ1is given by:
 2
var( bˆ1 ) = (4.17)
n x  x , z
2 2

where  x2is the variance of x, and  x2,z is the correlation between x and z.
 Now, the estimator of var(bˆ1 ) is obtained by replacing σ2,  2, and  x ,z with
2
x
their sample estimates.
 Sample estimator of σ2 is obtained in the following way. First, obtain the IV
estimates for β0 and β1, then compute
uˆi = yi − bˆ0 − bˆ1 xi (4.18)
The estimator for σ2 is then computed
n
as
1
̂ 2 =
n−2
 i
i =1
uˆ 2
(4.19)
 Cont’d---
 The sample estimator for  x2 is given as:
1 n SSTx
̂ x =  ( xi − x ) 2 =
2 (4.20)
n i =1
 n
for  x , z can be most easily obtained in the following way.
SST 2
 Finally, sample estimatorx

First, regress x on z. Then the R-squared from this regression equals the square of
the sample correlation, R2x,z
 Then, the estimator for the variance of is given by:
ˆ 2
va r̂( bˆ1 ) = (4.21)
SSTx  Rx , z
2

 You can show that this is a consistent estimator of the asymptotic variance of (4.17).
 The R-squared for IV regression is computed as
R2=1-SSR/SST (4.22)
Where SSR is the sum of the squared IV residuals. (The IV residual is given by (4.18)).
Unlike in the case of OLS, SSR can be greater than SST. Thus, R2 can be negative.
In IV regression, R2 does not have a natural interpretation.
 IV Estimation of the multiple regression model
 IV estimation can be extended to the multiple regression case. Call the
model we are interested in estimating the structural model
 Our problem is that one or more of the variables are endogenous.
 To estimate a multiple regression consistently, we need at least one
instrumental variable for each troublesome explanatory/endogenous
variable
 Identification
 In general, a parameter is said to be identified if different values of the
parameter would produce different distributions of the data.
 In IV regression, whether the coefficients are identified depends on the
relation between the number of instruments (m) and the number of
endogenous regressors (k).
 Intuitively, if there are fewer instruments than endogenous regressors, we
can’t estimate . . For example, if k = 1 but m = 0 (no instruments)
 Cont’d---
 Three possible cases for the coefficients, to be identified are:
✓ When we do not have enough instruments, the equation is under identified
(m < k). In this case we do not have a consistent estimator
✓ When we have just enough instruments for consistent estimation, we say the
regression equation is exactly identified (m = k). Then we simply use
Instrumental Variables Least Squares
✓ When we have more than enough instruments, the regression equation is over
identified(m > k). Two-Stage Least Squares (2SLS or TSLS) will be
employed for the reduced form equations
 The necessary condition for identification is called order condition.
- There should be at least as many excluded exogenous explanatory variables
as there are included endogenous explanatory variables in the structural
equation.
❑ The sufficient condition for identification is a rank condition
 Cont’d---
 Write the structural model as:
y1 = b0 + b1y2 + b2z1 + u1 (4.23)
, where y2 is endogenous and z1 is exogenous
 Let z2 be the instrument, so Cov(z2,u1) = 0 and
y2 = 0 + 1z1 + 2z2 + v2 (4.24)
, where 2 ≠ 0
 This reduced form equation regresses the endogenous variable on all
exogenous ones
 Two Stage Least Squares (2SLS)
 It’s possible to have multiple instruments
 Consider our original structural model, and let
y2 = 0 + 1z1 + 2z2 + 3z3 + v2 (4.25)
 Cont’d---

 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
 Could use either z2 or z3 as an instrument
 The best instrument is a linear combination of all of the exogenous
variables, y2* = 0 + 1z1 + 2z2 + 3z3
 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
 If there is a z, such that Corr(z,u) = 0 and Corr(z,x1) ≠ 0, then
 IV will remove the attenuation bias (biased OLS estimate towards
zero due to the measurement error) too.
 IV versus 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
 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
 Prefer IV if Corr(z,u)/Corr(z,x) < Corr(x,u)
 Testing for Endogeneity
 If we do not have endogeneity, both OLS and IV are consistent, use OLS
 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
 Idea of Hausman test is to see if the estimates from OLS and IV are
different
 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
 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. A significant t statistic implies that y2 is
endogenous.
 If multiple endogenous variables, jointly test the residuals from each first
stage
 Cont’d---
 A version of the Hausman’s test that has been proposed by Davidson and
Mackinnon (1993), which carries out the test of endogeneity by running an
auxiliary regression.
 In other way, the underlying idea of Hausman test is to compare two sets of
estimates, one that is consistent under both the null and the alternative
hypothesis and another, which is consistent only under the null hypothesis.
 A large difference between the two sets of estimates is taken as evidence of
the alternative hypothesis.
 First we run the explanatory variable suspected to be endogenous on
all exogenous variables and instruments and retrieve the residuals.
 Second, we re-estimate the original regression including the
retrieved residuals as additional regressors.
 Simultaneity
 Simultaneity is an important form of endogeneity.
 Simultaneity arises when one or more of the explanatory variables are
jointly determined with the dependent variable typically through an
equilibrium mechanism.
 The leading method for the estimation of the simultaneous equation models
(SEMs) is the method of instrumental variables.
 The most important point to remember in using simultaneous equations
models is that each equation in the system should have a ceteris paribus
causal interpretation.
 The classical example for the analysis of SEMs is the supply and demand
equations for a certain commodity or factor input.
 In SEMs, endogeneity means that the observed variable is determined by the
equilibrium. For example, an observed quantity is determined by the
 Cont’d
 Consider the following model describing equilibrium quantity of labor (in
hours) in agricultural sector in a country.
Labor supply : hs=α1w+β1z1+u1 (4.26)
Labor demand: hd=α2w+β2z2+u2 (4.27)
, where, hs is the hours of labor supplied, and hd is the hours of labor demanded.
These quantities depends on the wage rate, w, and other factors, z1 and z2
 z1 would be agricultural land area. The more land available, more demand for
labor. z1 is the observed supply shifter. u1 is the unobserved supply
shifter
 z2would be the wage rate of the manufacturing sector. If the manufacturing
wage increases, people would move to manufacturing sector, reducing hours
worked in agricultural sector. z2 is called the observed demand shifter. u2
is called the unobserved demand shifter.
 Cont’d---
 The observed labor quantity and wage rate are determined by the
equilibirum between these two equations. The equilibrium: hs=hd
 The above two equations constitute a simultaneous equations model (SEM)
and the equations are called the structural equations (that it is derivable
from economic theory and has a causal interpretation). The parameters
α1,β1, α2, β2 are called the structural parameters
 In SEM framework, hi and wi are endogenous variables because they
are determined by the equilibrium between the two equations.
 In the same way, zi1 and zi2 are exogenous variables because they are
determined outside of the model.
 u1 and u2 are called the structural errors.
 One more important point: Without z1 or z2, there is no way to
distinguish whether one equation is demand or supply
 Simultaneous equation bias
 Consider the following simultaneous equation model.
y1=α1y2+β1z1+u1 (4.28)
y2=α2y1+β2z2+u2 (4.29)
In this model, y1 and y2 are endogenous variables since they are determined by
the equilibrium between the two equations. z1 and z2 are exogenous
variables
 Since z1 and z2 are determined outside of the model, we assume that z1 and z2
are uncorrelated with both of the structural errors. Thus, by definition, the
exgoneous variables in SEM are exogenous in ‘econometric sense’ as well.
 In addition, the two structural errors, u1 & u2, are assumed to be uncorrelated
with each other.
 Both equations are behavioral, have ceteris paribus interpretation and each
stands by its own.
 Cont’d---
 Now, solve the equations (4.28) and (4.29) for y1 and y2, then you get the
following reduced form equations.
y1=п11z1+п12z2+v1 (4.30)
y2=п21z1+п22z2+v2 (4.31)
where
п11= β1/(1- α1 α2), п12= α1 β2/(1- α1 α2), v1 =(u1+ α1 u2)/(1- α1 α2)
п21 =α2β1/(1- α2 α1), п22 = β2/(1- α2 α1), v2=(α2u1+u2)/(1- α2 α1)
❑ These parameters are called the reduced form parameters
❑ You can check that v1 and v2 are uncorrelated with z1 and z2. Therefore, you
can estimate these reduced form parameters by OLS (Just apply OLS
separately for each equation)
 Cont’d---
❖ There are at least three reasons for using reduced-form equations:
1. Since the reduced-form equations have no inherent simultaneity, they
do not violate Classical Assumption of exogenity
– Therefore, they can be estimated with OLS without encountering the
problems of endogenity
2. The interpretation of the reduced-form coefficients as impact
multipliers means that they have economic meaning and useful
applications of their own
3. Reduced-form equations play a crucial role in Two-Stage Least Squares, the
estimation technique most frequently used for simultaneous equations
 Simultaneity bias refers to the fact that in a simultaneous system, the
expected values of the OLS-estimated structural coefficients are not equal
to the true βs, that is:
 Cont’d---
 The reason for this is that the two error terms of structural Equations are
correlated with the endogenous variables when they appear as
explanatory variables
 However, you cannot estimate the structural equations with OLS. For
example, consider the first structural equation: y1=α1y2+β1z1+u1
Notice that
Cov(y2, u1) =[α2/(1-α2α1)]E(u12)
=[α2/(1-α2α1)]σ21 ≠0 (4.32)
❑ Thus, y2 is correlated with u1 (assuming that α2 ≠0.) In other words, y2 is
endogenous in ‘econometric sense’.
 Identifying and estimating a structural
equation
 In previous session, we learned how to identify (i.e., eliminate the bias) by
apply IV method.
 In SEM, the term ‘identification’ is used slightly differently.
 Suppose the following model describing the supply and demand:
Supply: q =α1p+β1z1+u1 (4.33)
Demand: q =α2p+u1 (4.34)
Note that supply curve has an observed supply shifter z1, but demand has no
observed shifter.
Given the data on q, p and z1, which equation can be estimated? That is, which
is an identified equation?
 Because there is observed supply shifter z1 which is not contained in
demand equation, we can identify the demand equation.
 Cont’d---
 It is the presence of an exogenous variable in the supply equation that
allows us to estimate the demand equation.
 In SEM, identification is used to mean which equation can be estimated
 Now turn to a more general case.
y1 = b10 + a1 y 2 + b11 z11 +  + b1k z1k + u1 (4.35)
y 2 = b 20 + a 2 y1 + b 21 z 21 +  + b 2 l z 2 l + u 2 (4.36)
(z11~z1k) and (z21~ z 2l ) may contain the same variables, but may contain
different variables as well.
❑ When one equation contains exogenous variables not contained in the other
equation, this means that we have imposed exclusion restrictions
❖ The condition for identification: The first equation is identified iff the
second equation contains at least one exogenous variable (non zero coefficient)
that is excluded from the first equation
 Cont’d---
 The above condition have two components.
- First, at least one exogenous variable should be excluded from the first
equation (order condition).
- Second, the excluded variable should have non zero coefficients in the
second equation (rank condition).
✓ Example
 Labor supply of married working women.
Labor supply equation:
hours = a1lwage + b10 + b11educ + b12 age + b13kids6
+ b14 ( NonWifeIncome) + u1
Wage offer equation:
lwage = a 2 hours + b30 + b21educ + b22 exp+ b 23 exp2 + u2
 Cont’d---
❑ In the model, hours and lwage are endogenous variables. All other
variables are exogenous. (Thus, we are ignoring the endogeneity of
educ arising from omitted ability)
 Suppose that you are interested in estimating the first equation.
 Since exp and exp2 are excluded from the first equation, the order
condition is satisfied for the first equation. The rank condition is that,
at least one of exp and exp2 has a non zero coefficient in the second
equation. Assuming that the rank condition is satisfied, the first
equation is identified.
 In a similar way, you can see that the second equation is also identified
 Estimating SEM using 2SLS
❖ Once we have determined that an equation is identified, we can estimate it by
two stage least square
 Consider the labor supply equation example again. You are interested in
estimating the first equation.
 Suppose that the first equation is identified (both order and rank conditions are
satisfied).
 lwage is correlated with u1. Thus, OLS cannot be used.
 However, exp and exp2 can be used as instruments for lwage in the first
equation
 Because, first, exp and exp2 are uncorrelated with u1 by assumption of the
model (instrument exogeneity satisfied). Second exp and exp2 are correlated
with lwage by the rank condition (instrument relevance satisfied).
 In general, the exogenous variables excluded from the equation were called
the instruments; thus, you can use them as the instruments
 Cont’d---
 Two-Stage Least Squares (2SLS) helps mitigate simultaneity bias in
simultaneous equation systems and consist of the following two steps as
usual
 STAGE ONE:
◼ Run OLS on the reduced-form equations for each of the endogenous variables
that appear as explanatory variables in the structural equations in the system
◼ That is, estimate (using OLS):
(4.37)
(4.38)

 STAGE TWO:
◼ Substitute the Ys from the reduced form for the Ys that appear on the right
side (only) of the structural equations, and then estimate these revised
structural equations with OLS
 Cont’d---
❑ That is, estimate (using OLS):
(4.39)
(4.40)
❑ Hence, the simultaneity causality bias can be eliminated using IV-2SLS estimation
approach
Note that:
✓ 2SLS estimates are still biased in small samples, but consistent in large samples
(get closer to true βs as N increases)
✓ If the fit of the reduced-form equation is poor, then 2SLS will not rid the equation of
bias even in a large sample
✓ 2SLS estimates have increased variances and standard errors relative to OLS
✓ Two-Stage Least Squares cannot be applied to an equation unless that equation is
identified. Thus, Identification is a precondition for the application of 2SLS to
equations in simultaneous systems