Econometrics II, Fall 2017

Department of Economics, University of Copenhagen

Morten Nyboe Tabor

Solution Guide

#5.1 Interpretation of IV and GMM Estimations

Assume that a central bank sets the monetary policy instrument, the short term
interest rate rt , according to a simple policy rule of the form

rt = α0 + α1 · E[πt+12 − π ∗ | It ] + α2 · E[yt | It ], (5.1)

where E[· | It ] denotes the rational expectation conditional on the information

set available at time t; πt is the current inflation rate from the year before; π ∗
is the constant inflation target of the central bank; and yt is a measure of the
output gap of the economy so that an increase corresponds to an expansion of
the economy.

(1) Give an economic interpretation of the equation in (5.1).

What do you think could be reasonable values for the parameters?

Solution: Equation (5.1) is known as the Taylor rule. In a famous paper from
1993, John B. Taylor1 suggested that monetary policy responses over the period
from 1987 to 1992 can be summarized by a simply rule, which states that the
federal funds rate is set in reaction to the deviation of expected inflation from its
target level and the deviation between economic output and the potential output
(the output gap). Deviations from the benchmark rule are typically interpreted
as evidence of discretionary monetary policy.
To stabilize inflation, the coefficient α1 must be greater than one, i.e. α1 > 1,
which has become known as the Taylor Principle. That implies that an increase
in the inflation rate, πt , is matched by an more than one-to-one increase in the
1
Taylor, John B., “Discretion versus policy rules in practice”, Carnegie-Rochester Conference
Series on Public Policy, Vol. 39, pp. 195-214. http://dx.doi.org/10.1016/0167-2231(93)
90009-L

1
federal funds rate, rt , so that the real interest rate, rt − πt , increases, which cools
the economy and lowers inflation towards its target.
The coefficient α2 is the reaction of the federal funds rate to the output gap.
The coefficient should be positive, α2 > 0, so that when output rises above its
potential, the federal funds rate is lowered to cool the economy.
In his original paper, Taylor suggested the parameter values α1 = 1.5 and
α2 = 0.5 as representative.

(2) Explain why the parameters cannot be estimated by OLS.

Solution: In general, we do not observe the market’s expectations for future

inflation and the current output gap (which is not observed), so we replace them
with observed values for the current inflation and some proxy variable for the
output gap (such as, the deviation of output from a smooth estimated trend).
Pluggin in the observed variables in equation (5.1), we get,

rt = α0∗ + α1 πt+12 + α2 yt + ut , (5.2)

where,



ut = α1 E[πt+12 | It ] − πt+12 +α2 E[yt | It ] − yt , (5.3)

and α0∗ = α0 − α1 π ∗ .
As the explanatory variables, πt+12 and yt , are clearly correlated with the
residual term, ut , the moment conditions E[πt+12 ut ] 6= 0 and E[yt ut ] 6= 0 are
violated and the OLS estimator for the model in equation (5.2) is inconsistent.
We can rewrite the model in equation (5.2) as,

rt = x0t β + ut , (5.4)

where xt = (1, πt+12 , yt )0 and β = (α0∗ , α1 , α2 )0 . Again, the moment condition is

violated, E(xt ut ) 6= 0, so the OLS estimator of β is inconsistent.

(3) Now let zt denote and R−dimensional vector of variables in the information
set at time t, It . Explain how the instruments in zt can be used to estimate the
model in (5.1) using GMM. In particular, replace the expected value with actual
observation and state the moment conditions for estimating the parameters.
What is the requirement for zt to be valid instruments?

Solution: If the R-dimensional vector of instruments, zt , are uncorrelated

with the residual term, ut , then we can use the new moment conditions,

E(zt ut ) = 0, (5.5)

2
to derive an estimator of the parameters in β. Note, that since zt is an R-
dimensional vector, there are R moment conditions, E(zit ut ) = 0 for i = 1, 2, ..., R,
while there are K = 3 parameters of interest in β.
For the instruments, zt , to be valid they must be uncorrelated with the resid-
ual term, ut . For the instruments to be relevant, they should be correlated with
the endogenous variables in xt , i.e. πt+12 and yt .
In general terms, the population moment conditions are given by,

g(β) = E(f (wt , zt , β)) = 0, (5.6)

where wt = (rt , x0t )0 are the model variables in equations (5.2) and (5.4), zt is
an R-dimensional vector of instruments, and β is a K-dimensional vector of true
parameters in the model. Here, we consider the linear case where f (wt , zt , β) =
zt ut = zt (rt − α0∗ − α1 πt+12 − α2 yt ).
We can write out the moment conditions as,

E(zt ut ) = E(zt (rt − x0t β)) = E(zt (rt − α0∗ − α1 πt+12 − α2 yt )) = 0, (5.7)

or, by writing out the vector of instruments,

  
z1t
 z2t 
  
∗

E  .  (rt − α0 − α1 πt+12 − α2 yt )
  
 = 0, (5.8)
 ..  
zRt

which is a system of R equations with K = 3 unknown parameters in β,

E(z1t (rt − α0∗ − α1 πt+12 − α2 yt )) = 0

E(z2t (rt − α0∗ − α1 πt+12 − α2 yt )) = 0
..
.
E(zRt (rt − α0∗ − α1 πt+12 − α2 yt )) = 0. (5.9)

As we do not observe the population expectations, we replace them by the sample

averages to estimate the parameters β. The R sample moment conditions are
given by,
T T
1X 1X
gT (β) = f (wt , zt , β) = zt (rt − α0∗ − α1 πt+12 − α2 yt ). (5.10)
T T
t=1 t=1

(4) Explain the minimal value of R necessary for identification of the parameters
in (5.1). Explain in your own words how identification is related to estimation.

3
 Solution: We need a minimum of R = 3 instruments in zt to estimate the
K = 3 parameters in β = (α0∗ , α1 , α2 ). Identification states the the population
moment conditions,
E(zt ut ) = 0, (5.11)

have a unique solution at the true parameters β. Identification is a necessary

condition for estimation, but it is not sufficient to obtain good estimates.

(5) Now consider a specific choice of instruments:

zt = (1, rt−1 , rt−2 , πt−1 , πt−2 , yt−1 , yt−2 , bt−1 , bt−2 , xt−1 , xt−2 )0 ,

where the additional variables bt and xt denote, respectively, the long-term bond
yield and the unemployment rate.
Explain the difference between the role of the variables included in (5.1) and
the instrumental variables in zt . To explain the difference you can think of two-
stage least-squares (2SLS) as a special case of GMM.
Explain (with reference to 2SLS) what it means that a set of instruments can
be weak.

Solution: The variables in xt , πt+12 and yt , are the model variables that
are related to the optimal economic behavior. The variables π ∗ and yt are the
target variables for the Federal Reserve, and the coefficients α1 and α2 can be
interpreted as the reaction of the Federal Reserve to achieve their target. We are
interested in estimating the coefficients in β, i.e. α0∗ , α1 , and α2 .
The variables in zt are instruments. They are included in the information set
at time t, i.e. they are observed at time t.
To understand the role of the instruments, we can consider the two-stage
least squares (2SLS) estimator. In the first step, the instruments can be used to
forecast E(πt+12 |It ) and E(yt |It ), which can be achieved by regressing πt+12 and
yt , respectively, on the set of instruments. That gives the forecasts, πbt+12 and ybt .
In a second step, we replace the endogenous variables by the forecasted ones
and estimate the model,

rt = α0∗ + α1 π
bt+12 + α2 ybt + vt . (5.12)

A consequence of the moment condition, E(zt ut ) = 0, is that the forecasted

variables are uncorrelated with the residual term, so the moment condition holds
in the second step, E(bxt vt ) = 0, where x bt+12 , ybt )0 .
bt = (1, π
Weak instruments means that the forecasts of the endogenous variables based
on the instruments are poor, which is a consequence of a low correlation between
the instruments and the endogenous variables.

4
 For time-series data it is common to use lagged model variables as instruments
as the variables are typically persistent. For example, here we have included πt−1
and πt−2 as instruments as they are often good predictors of πt .

(6) How would the interpretation change if the unlagged variables bt and xt are
included as additional instruments in zt ?
How would the interpretation change if yt was also included?
Which of the cases do you think is the most reasonable in practice?

Solution: If bt and xt (the long-term bond yield and the unemployment

rate) are included in zt they must have been observed at time t, so that they are
observed at the time when rt is determined.
If yt is included in zt we do not need instruments for that variable since we
observe the output gap in real time.