The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Working Paper Series
Discussion on "Scarcity of Safe Assets,
Inflation, and the Policy Trap" by
Andolfatto and Williamson
WP 15-03
This paper can be downloaded without charge from:
http://www.richmondfed.org/publications/
Huberto M. Ennis
Federal Reserve Bank of Richmond
Discussion on “Scarcity of Safe Assets, Inflation, and the Policy Trap”
by Andolfatto and Williamson 1
Huberto M. Ennis
Research Department
Federal Reserve Bank of Richmond
24 March 2015
Working Paper No. 15-03
Abstract
This discussion was prepared for the 84th Meeting of the Carnegie-Rochester-NYU
Conference Series on Public Policy “Monetary Policy: An Unprecedented
Predicament” held on November 14-15, 2014, at Carnegie Mellon University.
1
I would like to thank Arantxa Jarque, Todd Keister, Jeff Lacker, and Pierre Daniel Sarte for comments. The views
expressed herein are those of the author and do not necessarily reflect the views of the Federal Reserve Bank of
Richmond or the Federal Reserve System. E-mail address: huberto.ennis@rich.frb.org.
1
I. Introduction
In the last few years, the U.S economy has experienced very low nominal interest rates with the
short-maturity end of the yield curve pegged effectively at the zero lower bound. At the same
time, inflation has been positive and most of the time fluctuating in a range of 1 to 2 percent on
an annual basis. The obvious implication of these facts is that the effective real interest rate has
been very low (and even negative, depending on maturities) for a considerable period of time.
Andolfatto and Williamson set up a relatively simple model that is able to accommodate these
facts and then study the implications of such a model for monetary policy.
Simple models that generate, in a plausible way, relatively low real interest rates in the medium
to long run are hard to come by. Interestingly, there is a large body of work in the macrofinance literature dealing with this kind of issue (Kocherlakota, 1996). Standard dynamic
macroeconomic models tend to generate risk-free interest rates that are too high for what was
observed in U.S. data during the 20th century. In the process of adapting models to deal with
this fact, it is now well recognized in the macro-finance literature that one way to lower real
rates in a representative-agent model is to introduce a transaction role for risk-free bonds
(Bansal and Coleman, 1996).
Andolfatto and Williamson seem to get to a similar idea but from a different starting point.
Both Andolfatto and Williamson have been active contributors to the literature that emphasizes
the micro-foundations of the use of money in exchange. It is a natural step in that literature to
study the role of other assets besides money in the mechanism of exchange. Williamson (2012)
incorporates some of the insights from that body of work, the so-called “new monetarist”
perspective, into a macroeconomic model suitable for analyzing the impact of different
monetary policies. 2 In this new paper, Andolfatto and Williamson take a step toward simplicity
by imposing ad-hoc Clower-type constraints for both cash and bonds and then studying the
monetary policy implications of assigning a transaction role to government bonds.
There are good reasons, beyond theoretical plausibility, to consider such a specification. As
Krishnamurthy and Vissing-Jorgensen (2012) extensively discuss, there appears to be a strong
negative relationship in the U.S. data between the yield spread of corporate bonds over
Treasuries and the total outstanding amount of government debt over GDP. This relationship
suggests that when there are more government bonds outstanding, the price of bonds falls and
the yield increases, closing the gap with the yield on corporate debt. Krishnamurthy and
Vissing-Jorgensen use a “bonds-in-the-utility-function” model (a la Sidrauski) to interpret their
empirical findings. Andolfatto and Williamson recast the idea in the form of a “bonds-in2
Lagos (2010) is one paper in the new monetarist tradition that addresses the asset-pricing puzzles reviewed by
Kocherlakota (1996) using transactions liquidity as an added characteristic of certain assets.
2
advance” constraint and concentrate on the monetary policy implications of that observed
relationship between total debt and its yield.
During the recent period of very low interest rates in the United States, it has been common to
hear experts and policymakers argue that the zero lower bound has been binding. If it were
possible, these experts conclude, it would be optimal to lower the short-term nominal rate to a
value below zero. A similar logic has been extensively used to justify non-conventional
monetary policy, such as forward guidance and asset purchase programs (see, for example, the
discussion in Woodford, 2012). In sharp contrast with this view, in the model of Andolfatto and
Williamson, under certain circumstances, the economy may find itself with a nominal and a real
interest rate that are too low and further reducing the nominal interest rate actually
exacerbates the problem, causing output (and welfare) to fall further away from its efficient
level. In my opinion, such a clear challenge to what might be considered conventional wisdom
makes the paper by Andolfatto and Williamson worthy of serious consideration.
II. Assessing the results
The paper by Andolfatto and Williamson is a technical paper. As such, it deserves a technical
discussion. To do that, I will briefly describe a simplified version of their model and then point
out some issues and interpretations.
The model is a general equilibrium dynamic macroeconomic model with a large number of
identical households and two goods, 1 and 2, that are produced by the households allocating
labor 𝑛𝑛 to a one-to-one production technology. Utility from each good is given by a smooth
utility function 𝑢𝑢(𝑐𝑐) with the usual properties, and a linear disutility of labor equal to 𝛾𝛾𝛾𝛾 is
assumed. Each household cannot consume its own output but can consume other households’
output. I am going to assume that the economy is non-stochastic and that agents have perfect
foresight. This is all I need for the purpose of my discussion. While Andolfatto and Williamson
set up the model as if there were shocks, as far as I can tell not much of the stochastic aspect of
the model comes into play in their results.
There are two financial assets in the economy: cash and one-period zero-coupon nominal
government bonds. Let 𝑞𝑞𝑡𝑡 be the price of the bonds in period 𝑡𝑡. Households enter each period
with a stock of money that results from holding one-period nominal bonds and from selling
goods in the previous period. Within the period, they first have an opportunity to trade in an
asset market where the government also sells new bonds to households. Households cannot
short bonds.
3
After the asset market closes, households trade goods. In the goods market, each household
faces two “means-of-payment” constraints. First, good 1 must be paid with cash, so that
𝑝𝑝1𝑡𝑡 𝑐𝑐1𝑡𝑡 ≤ 𝑀𝑀𝑡𝑡 , where 𝑝𝑝1𝑡𝑡 is the dollar price of good 1 at time 𝑡𝑡, 𝑐𝑐1𝑡𝑡 is the quantity of good 1
consumed by the household at time 𝑡𝑡, and 𝑀𝑀𝑡𝑡 is the nominal money holdings of the household
after trading in the asset market of period 𝑡𝑡. Second, a portion of the quantity that the
household purchases of good 2 must be paid with bonds, so that 𝑝𝑝2𝑡𝑡 (𝑐𝑐2𝑡𝑡 − 𝜅𝜅) ≤ 𝐵𝐵𝑡𝑡 , where 𝑝𝑝2𝑡𝑡
is the dollar price of good 2 at time 𝑡𝑡, 𝑐𝑐2𝑡𝑡 is the quantity of good 2 consumed by the household
at time 𝑡𝑡, and 𝐵𝐵𝑡𝑡 is the holdings of nominal bonds that the household acquired in the asset
market of period 𝑡𝑡. Here, 𝜅𝜅 is the quantity of good 2 that the household can produce for itself.3
In a situation where both goods are produced, traded, and consumed, the representative
household should be indifferent between producing and selling good 1 or good 2. Furthermore,
one unit of cash and one unit of bonds are equivalent in the payment of goods from the
perspective of the seller: In both cases, the seller ends up with one unit of cash next period.
Denote by 𝑆𝑆𝑡𝑡+1 the quantity of cash that the representative household will have at the
beginning of period 𝑡𝑡 + 1 from selling goods in period 𝑡𝑡. Using this notation, we have that:
𝑆𝑆𝑡𝑡+1 = 𝑝𝑝1𝑡𝑡 𝑐𝑐1𝑡𝑡 + 𝑝𝑝2𝑡𝑡 (𝑐𝑐2𝑡𝑡 − 𝜅𝜅) = (𝑝𝑝1𝑡𝑡 − 𝑝𝑝2𝑡𝑡 )𝑐𝑐1𝑡𝑡 + 𝑝𝑝2𝑡𝑡 (𝑛𝑛𝑡𝑡 − 𝜅𝜅),
which tells us that unless 𝑝𝑝1𝑡𝑡 = 𝑝𝑝2𝑡𝑡 the seller would choose to produce and sell only one of the
two goods. Therefore, from now on, consider only the case where both goods are traded and
denote by 𝑝𝑝𝑡𝑡 the common price of goods. Before moving on, though, note that from the
perspective of the buyer, whenever 𝑞𝑞𝑡𝑡 < 1, buying good 2 is cheaper than buying good 1. This
is the case because good 2 can be paid with “dollars tomorrow” that were acquired today at a
discount (i.e., at less than a dollar).
As is common in cash-in-advance economies, when 𝑞𝑞𝑡𝑡 < 1 in equilibrium we have that
𝑝𝑝𝑡𝑡 𝑐𝑐1𝑡𝑡 = 𝑀𝑀𝑡𝑡 . Assume, for simplicity, that the cash-in-advance constraint also holds with equality
when 𝑞𝑞𝑡𝑡 = 1. This is compatible with equilibrium and it is not crucial for any of the results I will
discuss.
Now denote with lower-case letters the real value of all nominal quantities. The representative
household’s budget set is then described by the following set of equations:
𝑎𝑎 )
𝑚𝑚𝑡𝑡 + 𝑞𝑞𝑡𝑡 (𝑏𝑏𝑡𝑡 + 𝑏𝑏𝑡𝑡+1
= 𝑠𝑠𝑡𝑡 + 𝜏𝜏𝑡𝑡 ,
𝑐𝑐1𝑡𝑡 = 𝑚𝑚𝑡𝑡 ,
𝑐𝑐2𝑡𝑡 − 𝜅𝜅 = 𝑏𝑏𝑡𝑡 ,
3
Andolfatto and Williamson give a slightly different interpretation of 𝜅𝜅. They call 𝜅𝜅 the portion of good 2 that can
be bought with credit.
4
𝑝𝑝𝑡𝑡
𝑠𝑠𝑡𝑡+1 = 𝑝𝑝
𝑡𝑡+1
𝑎𝑎 ),
(𝑛𝑛𝑡𝑡 − 𝜅𝜅 + 𝑏𝑏𝑡𝑡+1
𝑎𝑎
and the no-short-selling constraint 𝑏𝑏𝑡𝑡+1
≥ 0. Here, again, we denote with the letter 𝑠𝑠 the real
value of cash holdings that the representative household has at the beginning of the period as a
result of selling goods the previous period and holding nominal bonds over and above what was
𝑎𝑎
needed to pay for the purchases of consumption good 2 (that is, 𝑏𝑏𝑡𝑡+1
). The household
𝑡𝑡
maximizes the sum of future discounted utility, that is ∑ 𝛽𝛽 [𝑢𝑢(𝑐𝑐1𝑡𝑡 ) + 𝑢𝑢(𝑐𝑐2𝑡𝑡 ) − 𝛾𝛾𝑛𝑛𝑡𝑡 ], subject to
the constraints that describe its budget set.
A key component of the theory developed by Andolfatto and Williamson is their
characterization of government policy. I will argue that they deal with a very particular set of
policies and will discuss how results would change when alternative policies are followed. Here
is what they do. They first define the real value of consolidated government debt as 𝑉𝑉𝑡𝑡 = 𝑚𝑚
� 𝑡𝑡 +
𝑞𝑞𝑡𝑡 𝑏𝑏�𝑡𝑡 . Then, they assume that the fiscal authority adjusts transfers 𝜏𝜏𝑡𝑡 and the supply of bonds to
target a given sequence of 𝑉𝑉𝑡𝑡 . The monetary authority, via open market operations, targets the
value of 𝑞𝑞𝑡𝑡 , which is effectively a way to target the economy’s nominal interest rate.
Let 𝜋𝜋𝑡𝑡+1 = 𝑝𝑝𝑡𝑡+1 /𝑝𝑝𝑡𝑡 , the gross inflation rate. Given a policy {𝑞𝑞𝑡𝑡 , 𝑉𝑉𝑡𝑡 }, an equilibrium can be
characterized by a sequence {𝑐𝑐1𝑡𝑡 , 𝑐𝑐2𝑡𝑡 , 𝜋𝜋𝑡𝑡+1 } satisfying the following set of equations:
𝑢𝑢′(𝑐𝑐2𝑡𝑡 ) = 𝑞𝑞𝑡𝑡 𝑢𝑢′(𝑐𝑐1𝑡𝑡 ),
𝛾𝛾 = 𝛽𝛽
and
𝑢𝑢′(𝑐𝑐1𝑡𝑡+1 )
𝜋𝜋𝑡𝑡+1
𝑢𝑢′ (𝑐𝑐2𝑡𝑡 ) = 𝛾𝛾
,
if
or
𝑐𝑐1𝑡𝑡 + 𝑞𝑞𝑡𝑡 (𝑐𝑐2𝑡𝑡 − 𝜅𝜅) = 𝑉𝑉𝑡𝑡
𝑐𝑐1𝑡𝑡 + 𝑞𝑞𝑡𝑡 (𝑐𝑐2𝑡𝑡 − 𝜅𝜅) ≤ 𝑉𝑉𝑡𝑡 ,
if
𝑢𝑢′ (𝑐𝑐2𝑡𝑡 ) > 𝛾𝛾.
�𝑡𝑡 , which determines the price level given a policy-induced
Note that in equilibrium 𝑝𝑝1𝑡𝑡 𝑐𝑐1𝑡𝑡 = 𝑀𝑀
�𝑡𝑡 . In equilibrium, the real allocation depends not on the price level but on
supply of currency 𝑀𝑀
(gross) inflation 𝜋𝜋𝑡𝑡 .
As a benchmark example, consider the situation when policymakers target a constant 𝑉𝑉𝑡𝑡 and a
constant 𝑞𝑞𝑡𝑡 . Denote these targets by 𝑉𝑉� and 𝑞𝑞�. If 𝑉𝑉� is large enough, then the equilibrium is
unconstrained, in the sense that the constraint on short-selling is not binding in the
optimization problem of the representative household. The real allocation in the unconstrained
equilibrium is equivalent to the one that obtains in the more familiar case when all of good 2
can be produced and consumed at home. We denote this equilibrium with {𝑐𝑐1∗ , 𝑐𝑐2∗ , 𝜋𝜋 ∗ } and note
5
that 𝑐𝑐1∗ and 𝑐𝑐2∗ solve the equations 𝛾𝛾 = 𝑢𝑢′ (𝑐𝑐2∗ ) = 𝑞𝑞�𝑢𝑢′ (𝑐𝑐1∗ ), which also shows why both
equilibrium consumptions are constant over time. Note also that 𝜋𝜋𝑡𝑡 = 𝜋𝜋 ∗ = 𝛽𝛽/𝑞𝑞�. It is easy to
see from the cash-in-advance constraint that the equilibrium is consistent with a constant
money growth rate equal to 𝛽𝛽/𝑞𝑞�. The gross real rate of interest is equal to 1/𝛽𝛽. In summary,
the monetary authority controls inflation by targeting the price of the nominal bonds and the
real interest rate is invariant to policy (as long as the equilibrium remains unconstrained).
Suppose now that 𝑉𝑉� is relatively small. In particular, suppose that 𝑐𝑐2∗ − 𝜅𝜅 > 𝑏𝑏� . This is the
situation that Andolfatto and Williamson call a “scarcity of safe assets” in the economy. In this
case, the equilibrium is constrained and 𝑐𝑐2∗∗ = 𝑏𝑏� + 𝜅𝜅 < 𝑐𝑐2∗ (where we denote with double-stars
the constrained equilibrium). Consumption of good 1 is given by the solution to 𝑢𝑢′(𝑐𝑐1∗∗ ) =
(1/𝑞𝑞�)𝑢𝑢′�𝑏𝑏� + 𝜅𝜅�, and inflation is given by 𝜋𝜋 ∗∗ = (𝛽𝛽/𝛾𝛾)𝑢𝑢′(𝑐𝑐1∗∗ ). The gross real interest rate now
depends on 𝑏𝑏� as follows:
1
𝑞𝑞�𝜋𝜋 ∗∗
1
𝛾𝛾
1
= 𝛽𝛽 𝑢𝑢′(𝑏𝑏�+𝜅𝜅) < 𝛽𝛽 ,
where the last inequality results from the fact that 𝑐𝑐2∗∗ < 𝑐𝑐2∗ . In contrast with the unconstrained
equilibrium, marginal changes in fiscal policy in the constrained equilibrium affect the real
interest rate and inflation, given a policy followed by the monetary authority.
In light of the experience of the United States, where the nominal rate has been close to zero
and inflation has been in the neighborhood of 2 percent for the last few years, it seems most
relevant to understand equilibrium when 𝑞𝑞� ≈ 1. In the unconstrained equilibrium, just as in
standard cash-in-advance models, the real interest rate is positive and equal to 1/𝛽𝛽 and the
gross inflation rate 𝜋𝜋 ∗ is equal to 𝛽𝛽, which implies that the economy actually experiences
deflation. In the constrained equilibrium, in contrast, the gross real interest rate is lower than
1/𝛽𝛽 and 𝜋𝜋 ∗∗ is greater than 𝛽𝛽. In principle, inflation could be positive and the net real interest
rate could be negative, which seems to align well with the recent U.S. experience.
How does monetary policy influence outcomes in the Andolfatto-Williamson model? In the
unconstrained equilibrium, a lower 𝑞𝑞� implies a higher relative price of good 1 from the
perspective of the buyer. In other words, higher nominal interest rates imply that the cash good
is more “expensive” and households consume less of it. This is a standard result in cash-inadvance economies. More concretely, for a given value of 𝑉𝑉� = 𝑚𝑚
� + 𝑞𝑞�𝑏𝑏�, a lower 𝑞𝑞� implies a
lower 𝑐𝑐1∗ and hence a lower 𝑚𝑚
� . Fiscal policy reacts to this by setting a higher value of 𝑏𝑏�. But, as
we saw before, changes in the value of 𝑏𝑏� do not change the real allocation in the unconstrained
equilibrium.
In the constrained equilibrium, the interaction between fiscal and monetary policy is crucial for
the real allocation. For a given value of 𝑉𝑉� , a lower 𝑞𝑞� implies a lower 𝑐𝑐1∗∗ and hence a lower 𝑚𝑚
�.
6
Fiscal policy reacts to this by setting a higher value of 𝑏𝑏�, which changes the value of 𝑐𝑐2∗∗ and the
value of the real interest rate. Andolfatto and Williamson use this logic to argue that in the
constrained equilibrium the monetary authority can change the real interest rate and the
consumption allocation in the economy by changing the target for the price of the nominal
bonds.
Now, consider an alternative fiscal policy. Suppose that instead of targeting 𝑉𝑉� , the real value of
total consolidated government debt, the fiscal authority targets the real value of outstanding
government bonds, 𝑏𝑏�. As before, monetary policy targets the price of bonds, or equivalently,
the nominal interest rate. This change in the specification of policy is inconsequential in the
unconstrained equilibrium, of course. Suppose, then, that the economy is in the constrained
equilibrium. Now, just as before, a lower 𝑞𝑞� implies a lower 𝑐𝑐1∗∗ and hence a lower 𝑚𝑚
� . However,
since now 𝑏𝑏� is fixed, 𝑐𝑐2∗∗ does not change (it is equal to 𝑏𝑏� + 𝜅𝜅), and in consequence, monetary
policy does not affect the real interest rate, but just inflation, as is commonly the case in cashin-advance models.
In the unconstrained equilibrium, the optimal monetary policy is the Friedman rule (i.e., 𝑞𝑞� = 1).
As it turns out, though, in the constrained equilibrium it matters whether the fiscal authority is
targeting 𝑉𝑉� or 𝑏𝑏� for how the monetary authority should set optimal policy. If fiscal policy fixes
𝑉𝑉𝑡𝑡 = 𝑉𝑉� , then the Friedman rule is not optimal. To see this, suppose that 𝑞𝑞� ≈ 1. A lower 𝑞𝑞�
implies a lower 𝑐𝑐1∗∗ and hence a lower 𝑚𝑚
� , which leads the fiscal authority to set a higher 𝑏𝑏�. This,
in turn, increases 𝑐𝑐2∗∗ . It is easy to show that when 𝑞𝑞� ≈ 1, the change in 𝑐𝑐2∗∗ dominates the
change in 𝑐𝑐1∗∗ and welfare is higher when 𝑞𝑞� is set to be lower than unity. Note that the reaction
of fiscal policy to monetary policy is crucial for this result. In fact, if the fiscal authority targets
𝑏𝑏�, then the Friedman rule is again optimal. Perhaps, then, one of the main lessons from the
Andolfatto-Williamson paper is that, at the time of setting policy, the monetary authority needs
to be mindful of the reaction of the fiscal authority, if the ultimate goal is to maximize the
welfare of society. 4
Optimal monetary policy in the constrained equilibrium is non-trivial. Andolfatto and
Williamson do not spend a lot of energy on this issue. Instead, they move on to study how
Taylor rules perform in their environment. They consider a monetary policymaker who follows
a nominal interest rate rule of the form:
1
𝑞𝑞𝑡𝑡
= max[𝜋𝜋𝑡𝑡𝛼𝛼 (𝜋𝜋 ∗ )1−𝛼𝛼 𝑥𝑥𝑡𝑡 , 1],
4
Of course, there is an extensive literature studying this kind of idea. See, for example, the AEA Presidential
Address by Christopher Sims (2013). The title of section II.A of Sims’ paper is: “Monetary Policy Actions, to be
Effective, Must Induce a Fiscal Policy Response.”
7
where 𝑥𝑥𝑡𝑡 is a real interest rate adjustment factor in the sense that if the monetary authority
hits its target for inflation 𝜋𝜋 ∗ in all periods, then the real gross interest rate 1/𝑞𝑞𝑡𝑡 𝜋𝜋𝑡𝑡+1 equals 𝑥𝑥𝑡𝑡 .
There is no clear justification for following a Taylor rule within the context of the model.
Andolfatto and Williamson argue that “real-world central bankers take an interest” in this kind
of rule and, for that reason, it could be interesting to understand how the rule can go wrong in
their model. But, presumably, real-world central bankers would not want to follow a rule that is
obviously suboptimal, unless they have a misperception with respect to how the economy
actually works. In other words, more generally, Andolfatto and Williamson are proposing us to
consider the perils of having “misguided” policymakers.
In the unconstrained case, for the target inflation 𝜋𝜋 ∗ to be attainable in a stationary equilibrium
situation, we need 𝑥𝑥𝑡𝑡 = 1/𝛽𝛽. In this case, then, it is easy to show that when 𝛼𝛼 > 1 (the Taylor
principle) there are two equilibria with constant inflation and no equilibria with transitional
dynamics (i.e., non-stationary equilibria). This situation is similar to the situation discussed by
Benhabib, Schmitt-Grohe, and Uribe (2001b) in their “simple example,” where the cross-partial
derivative between money and consumption in the utility function are equal to zero.
The constrained case is more interesting. Now, the long-run real interest rate is actually
endogenous and, a priori, it is not as obvious what 𝑥𝑥𝑡𝑡 should be if the monetary authority
intends to hit the target 𝜋𝜋 ∗ in the long run. One possibility is to set 𝑥𝑥𝑡𝑡 equal to the current real
gross interest rate at each time 𝑡𝑡. Andolfatto and Williamson show that this way of conducting
policy creates scores of non-stationary equilibria. An alternative specification, not considered
by Andolfatto and Williamson, is to set 𝑥𝑥𝑡𝑡 = 1/𝑞𝑞𝑡𝑡 𝜋𝜋 ∗ . With this value of 𝑥𝑥𝑡𝑡 there are again only
stationary equilibria. I find this value of 𝑥𝑥𝑡𝑡 somewhat attractive to the extent that it suggests
that the policymaker is calculating the adjustment factor with its long-run inflation target in
mind. But, of course, there are no clear standards for how “misguided” a policymaker can be in
a theoretical model.
III. On misguided policymakers
Perhaps another important lesson from the Andolfatto-Williamson paper is that policymakers
need to be wary of the prescriptions derived from simple macroeconomic models when it
comes to monetary policy and stability. This is really not a new insight, of course. There is a
well-known literature on this subject. Andolfatto and Williamson cite the “Perils” paper of
Benhabib, Schmitt-Grohe, and Uribe (2001a), but those authors wrote several other papers on
the subject as well. In fact, in one of those papers (Benhabib, Schmitt-Grohe, and Uribe, 2001b),
they discuss in detail the properties of economies operating under Taylor-type interest rate
8
rules and how stability depends critically on very specific details of the environment.5 More
recently, John Cochrane (2011) has come back to the issue and postulated that many of the
arguments in favor of the Taylor principle are not technically convincing. He reports that a
review of the existing literature did not reveal to him full answers to his concerns. The current
state of the debate suggests to me that models often used to discuss monetary policy
alternatives are ridden with multiple equilibria and hence not very powerful in predicting
outcomes.
Andolfatto and Williamson present a model that is different, of course, from the one analyzed
by Benhabib, Schmitt-Grohe, and Uribe, describing it as an alternative. It is then not surprising
that monetary policy does not perform well under rules that have been designed to deal with
economies of a different structure. Effectively, Andolfatto and Williamson are considering an
environment where policymakers have bounded rationality and are mistaken about the way
the economy works. My casual impression is that, generally speaking, misguided policymakers
are more prevalent in real life than in the academic literature. Yet, in the particular case of
monetary policy, decisions tend to be made by highly qualified and knowledgeable individuals,
at least in the United States. While, for this reason, misguidedness in monetary policy looked to
me like a relatively weak case to make when discussing the United States, the general idea was
still worth thinking about.
In that process, I went back and read some of the statements that U.S. policymakers were
making in the spring of 2012. That period was one in which labor market indicators had finally
started to show some signs of improvement and the inflation rate was hovering around 2
percent after being somewhat higher in 2011. The Fed was relying mainly on forward-guidance
language to fine tune the stance of monetary policy, with the target interest rate pegged at its
effective lower bound. Evaluating monetary policy at that time was not easy, if it ever is.
Interestingly, then-Vice Chair Janet Yellen seemed to have taken a relatively pragmatic view of
the problem. In her April 11, 2012, speech, she said that “because I see no magic bullet for
determining the ‘right’ stance of policy, I commonly consider a number of different
approaches.” She went on to discuss first the prescriptions from an optimal control procedure,
pointing out that the analysis hinges “on the selection of a specific macroeconomic model as
well as a set of simplifying assumptions that may be quite unrealistic.” To complement her
analysis, she found it helpful to “consult prescriptions from simple policy rules,” arguing (based
on work by Taylor and Williams, 2011) that “research suggests that these rules perform well in
a variety of models and tend to be more robust.” However, Yellen went beyond what some of
the relevant literature seems to support when she said that “any benchmark rule should
conform to the so-called Taylor principle.” At this point, I suppose, the examples in the paper by
5
Adão, Correia, and Teles (2011) study variations in the specification of the interest rate rules that could make
them less conducive to multiplicity.
9
Andolfatto and Williamson (and in Benhabib, Schimitt-Grohe, and Uribe (2001b), for that
matter) may look germane again. 6
IV. Closing remarks
How to set monetary policy appropriately appears to be very sensitive to the structure of the
economy. Sometimes practitioners give very precise advice based on details of the model of the
economy that they have in mind. That seems better to me than advice that is just produced out
of thin air. However, as Andolfatto and Williamson illustrate in their paper, equally plausible
models of the economy often deliver very different recommendations. Perhaps this kind of
work tilts the balance yet another notch toward simpler and potentially more robust policy
rules. Lars Hansen and Thomas Sargent, of course, have been advocates of this way of thinking
for a long time (see Hansen and Sargent, 2011, for a review). Furthermore, Sargent reports to
be in good company when he tells us that Milton Friedman wrote in 1953 about “making policy
when you do not trust your model” and recommended “surrounding a macroeconomic
statistical model with shrouds of uncertainty and being cautious in applying it to construct
quantitative policy advice” (Sargent, 2014). As if things were not complicated enough already, it
seems to me that Andolfatto and Williamson just added a non-trivial point to the support of the
distribution of models that Friedman wanted us to use when addressing the policy problem.
References
Adão, Bernardino, Isabel Correia, and Pedro Teles. (2011) “Unique monetary equilibria with interest rate rules.”
Review of Economic Dynamics 14 (3): 432-442.
Bansal, Ravi, and Wilbur John Coleman (1996). “A monetary explanation of the equity premium, term premium,
and risk-free rate puzzles.” Journal of Political Economy 104 (6): 1135-1171.
Benhabib, Jess, Stephanie Schmitt-Grohé, and Martin Uribe (2001a). “The perils of Taylor rules.” Journal of
Economic Theory 96 (1): 40-69.
Benhabib, Jess, Stephanie Schmitt-Grohé, and Martin Uribe (2001b). “Monetary policy and multiple equilibria.”
American Economic Review 91 (1): 167-186.
Cochrane, John H. (2011). “Determinacy and identification with Taylor rules,” Journal of Political Economy 119 (3):
565-615.
6
Full quote from Yellen (2012): “The commitment to a balanced approach has crucial implications when it comes
to choosing sensible benchmarks from among the many alternative policy rules. In particular, any benchmark rule
should conform to the so-called Taylor principle, which states that, other things being equal, a central bank should
respond to a persistent increase in inflation by raising nominal short-term interest rates by more than the increase
in inflation so that the real rate of interest rises, thereby helping to bring inflation back down.”
10
Hansen, Lars Peter, and Thomas J. Sargent (2011). “Wanting robustness in macroeconomics,” in Benjamin M.
Friedman and Michael Woodford, eds., Handbook of Monetary Economics, vol. 3B, (San Diego: North Holland), pp.
1097-1157.
Kocherlakota, Narayana R. (1996). “The Equity Premium: It's Still a Puzzle.” Journal of Economic Literature 34 (1):
42-71.
Krishnamurthy, Arvind, and Annette Vissing-Jorgensen. (2012). “The aggregate demand for treasury debt.” Journal
of Political Economy 120 (2): 233-267.
Lagos, Ricardo. (2010). “Asset prices and liquidity in an exchange economy.” Journal of Monetary Economics 57 (8):
913-930.
Sargent, Thomas J. (2014) “The evolution of monetary policy rules.” Journal of Economic Dynamics and Control 49:
147-150.
Sims, Christopher A. (2013) “Paper money.” American Economic Review 103 (2): 563-584.
Taylor, John B., and John C. Williams (2011). “Simple and robust rules for monetary policy,” in Benjamin M.
Friedman and Michael Woodford, eds., Handbook of Monetary Economics, vol. 3B, (San Diego: North Holland), pp.
829-60.
Williamson, Stephen D. (2012). “Liquidity, monetary policy, and the financial crisis: A new monetarist approach.”
American Economic Review 102 (6): 2570-2605.
Woodford, Michael. (2012). “Methods of policy accommodation at the interest-rate lower bound.” The Changing
Policy Landscape: 2012 Jackson Hole Symposium. Federal Reserve Bank of Kansas City.
Yellen, Janet L. (2012). “The economic outlook and monetary policy,” speech at the Money Marketeers of New
York University, New York, New York. April 11.
11
@FedFRASER