Transcript Using MoPoS in the Classroom

Using MoPoS in the
Classroom
Yvan Lengwiler
WWZ, Economic Theory
University of Basel, Switzerland
2½ learning aims
1) MoPoS aims at providing a complement to the
usual curve shifting exercises.
Students should understand the key macroeconomic relationships as correlations, not as
abstract curves.
And they should recognize the most
important stylized facts in the simulation.
2) MoPoS tries to convey a realistic experience of
the job of a central banker.
Lots of information is missing, but decisions
have to be taken, and they have strong
effects, sometimes with long lags.
2½) Students should have fun doing this.
Curve shifting…
When teaching introductory macro we
usually use a more or less involved
combination of coordinate systems in which
we shift several more or less abstract
curves around.
A third semester student once told me:
"Essentially, economics is curve
shifting, right?"
Of course, nothing could be more wrong,
but this is the impression we give.
…vs. stochastic simulation
MoPoS is a stochastic simulation of a small
reduced-form mainstream macro model, which
also allows the user to interact with it.
Unlike in comparative statics (i.e. curve shifting)
exercises, the user of MoPoS is confronted with a
constant flow of shocks. He experiences the
dynamics of his policy decisions through time.
Interaction with a stochastic simulation provides
a different view of the most important
macroeconomic relationships, such as the IS or
the Phillips curve.
It focuses on correlations and lags rather than
comparative statics and abstract curves.
Monetary policy making
MoPoS allows the user to play the role of a central
bank governor.
It provides a realistic simulation of the decision
problem of the monetary authority, because…

The economy is subject to a constant stream of
unobservable shocks. The player constantly experiences
surprises.

Decisions have to be made based on very limited and
not perfectly reliable information.

Decisions interact with the (virtual) economy in complex
ways. Effects of these decisions are often visible only
with a considerable lag.

Strategy of the player affects expectations, and these
influence the available trade-offs.

Neither a very cautious nor a stop-go policy are
successful.
The model

Standard IS-LM-PC model with stochastic
potential growth. Calibrated to quarterly data.

All shocks are AR(1). Innovations are normally
distributed (or in some instances, can be chosen
to be leptokurtic).

Variance and autoregressive coefficient of
shocks can be set for each equation.

Inflation expectations are convex combination of
static expectation and OLS-expectation.

OLS-expectation is an inflation forecast
[regression of current inflation on lagged
inflation, real growth, and money growth, with
1 to 4 lags].
Core of the model
y *  g   y ,
4
* 
p  E {p}   t (y t  y t )   p ,
 t 1

4
y  y *    (y 1  y *1 )   t (i t  E {p t }  r * )  d ,
t 1
E {p}  pols  (1  )p 1.
Stochastic potential growth, Phillips-Curve, IS
curve, quasi-rational ("OLS-learning") expectation.
pols is forecast of regression of current inflation on
lagged inflation, real growth, money growth, with 1
to 4 lags.  is inflation stickiness parameter.
Monetary block and
control block

There is also a monetary block (an LMcurve with money demand shocks).

But money is passive in the model (except
that it enters inflation forecast).

Control is either given to the user or to
the Taylor rule ("autopilot").

Output gap is estimated as residual of
log-linear regression of real output on
time trend.
Observation errors
Macroeconomic statistics are typically subject to a
great deal of imprecision. They are often revised,
replacing the "initial estimate" with the "final
estimate."
Yet, this is what the monetary authority has, and it
has to decide upon the policy stance based on such
imprecise information.
p  p
observed p  
p
y   y
observed y  
y
before revision,
after revision,
before revision,
after revision.
The model
nominal
interest rate
(indirectly
through
money
demand)
inflation
expectation
real interest
rate
IS
business
cycle
PC
inflation
Credibility

The literature on the role of credibility in many
kinds of strategic situations — but especially in
monetary policy — is large.

In the simulation, a player who follows some
pattern (rule?), will earn a significant parameter
of his decision variable on inflation forecast.

Through this simple mechanism, he has to build
up this reputation by demonstrating that he does
not tolerate inflation.

Such a reputation makes the player's job easier
because his acts affect not only the interest rate,
but inflation expectations as well.
Liquidity trap
It can happen that the virtual economy falls into
a deflationary spiral and ends up in a liquidity
trap, because the nominal interest rate is not
allowed to become negative.
In reality, a true liquidity trap is a very rare
phenomenon. It requires the nominal return rates
on all assets to vanish.
In the simulation, however, there is only one
interest rate, and when this goes to zero and
deflation is strong enough — bang! — you cannot
escape the trap anymore.
So, the simulation is much more likely to
experience the trap than a real economy.
Liquidity trap
Possible exercises

Several scenarios come with the download,
such as "stability," "recession," or "new
economy," but students can also generate
new, random scenarios.

Students should make log of their decisions
and observations. Leads to more reflection.

Usually I ask the students first to try to make
"smiley" happy as long as possible.

…won't last, due to long run vertical Phillips
curve, but takes out the obvious temptation to
see a happy face.
Possible exercises


Next I ask the students to turn on the
autopilot (the Taylor rule).

The students can try to recognize the "stylized
facts" they know from the lecture.

They can also observe how the Taylor rule
behaves. Why does it do what it does? Does it
make miskates?
Finally, the students are asked to keep the
economy stable (starting from any of the
standard scenarios).

This is very difficult. Hopefully, students learn
what lags really mean, and why sometimes a
"preemptive strike" constitutes good policy.
Possible exercises


In order to share it, or for discussion in
class, students can…

save a simulation (with Simulator > Save
Simulation)

or print it (with File > Print).
So far, I have used the game with an
intermediate macro class, with a course
for continuing education, and with an
Executive MBA class.

But I think that one could also use it with
some benefit in some more advanced classes.
More advanced classes

"Advanced mode" allows much more diverse
exercises.

Compare true output gap with estimated gap (maybe
using different methods of estimation).

Make additional graphs (for
instance, graph the empirical
Phillips curve is "real" time).


Explore the effect of the parameters of the Taylor rule on the
performance of this particular
feedback rule.
One can also implement a whole new feedback rule.

Experiment with the parameters of the simultation.

One can even make the parameters themselves
stochastic, giving rise to a "robust control problem."
Download
Check out
www.wwz.unibas.ch/lengwiler
and open the «software» link to download
software and for additional information.
Email me at
[email protected]
for comments or questions.