Jump to ContentJump to Main Navigation
Show Summary Details
More options …

The B.E. Journal of Theoretical Economics

Editor-in-Chief: Schipper, Burkhard

Ed. by Fong, Yuk-fai / Peeters, Ronald / Puzzello , Daniela / Rivas, Javier / Wenzelburger, Jan

2 Issues per year

IMPACT FACTOR 2016: 0.229
5-year IMPACT FACTOR: 0.271

CiteScore 2016: 0.30

SCImago Journal Rank (SJR) 2016: 0.398
Source Normalized Impact per Paper (SNIP) 2016: 0.232

Mathematical Citation Quotient (MCQ) 2016: 0.08

See all formats and pricing
More options …

Optimal Term Structure in a Monetary Economy with Incomplete Markets

Matthew Hoelle
  • Corresponding author
  • Department of Economics, Purdue University, 403 W. State Street, West Lafayette, IN 47907, USA
  • Email
  • Other articles by this author:
  • De Gruyter OnlineGoogle Scholar
Published Online: 2017-07-27 | DOI: https://doi.org/10.1515/bejte-2016-0158


In a stochastic economy, the rebalancing of short and long term government debt positions can have real effects when markets are incomplete. This paper analyzes both stationary and dynamic policy rules for the term structure of interest rates. After proving the existence of a recursive representation of equilibrium, necessary conditions for Pareto efficiency are characterized. The necessary conditions are equivalent for both stationary and dynamic policy rules.

Keywords: unconventional monetary policy; yield curve; asset span; incomplete markets; Pareto efficiency

JEL Classification: D52; E43; E44; E52

1 Introduction

In recent years, central banks have expanded the set of assets that they are willing to hold on their balance sheets. Conventional monetary policies, typically called open market operations, involve the buying and selling of short-term government bonds. This paper analyzes one class of unconventional monetary policies in which the central bank buys and sells government bonds of varying maturities. In a stochastic setting, the additional policy tools can be used by the central bank to increase welfare by expanding the risk-sharing potential for households in the economy.

The paper uses two classical ideas in general equilibrium to analyze a modern policy problem. First, with incomplete markets and a fixed asset structure, Pareto efficiency can be supported if the asset span contains the excess demand vectors for all households (Magill and Quinzii 1996).1 Second, with incomplete markets, monetary policy can have real effects by changing the asset span (Magill and Quinzii 1992).

Though many applications of the policy-induced asset span can be considered, this paper focuses on the role of monetary policy in mitigating financial frictions caused by incomplete markets. Monetary policy is conducted by the central bank and operates through the purchase and sale of debt issued by the fiscal authority. Whereas conventional monetary policy only targets the short-term bond yields, the unconventional monetary policies analyzed in this paper target the entire term structure of interest rates. To study the welfare implications of unconventional monetary policies, this paper adopts the Pareto criterion as the benchmark. Even in the presence of incomplete markets, the broader tools available with term structure targeting, relative to open market operations, may be able to support a Pareto efficient allocation. The paper’s main contribution is to characterize necessary conditions for Pareto efficiency.

In the model presented in the paper, money is valued via a cash-in-advance constraint as in Herrenbrueck (2016). Heterogeneous households are infinite-lived and face idiosyncratic endowment risk. Households choose portfolios of the bonds of varying maturities. The bonds are nominally risk-free. Since household objective functions are specified in terms of real units of commodities, it is the real payouts of the bonds that matter. The real payouts are inversely related to the price levels supported by the monetary policy choice.

Monetary policy is chosen according to a vector of interest rate rules for the entire term structure. Two types of policy rules are considered: stationary and dynamic. A stationary policy rule is such that the interest rate targets only depend upon the realization of uncertainty in the current period. A dynamic policy rule allows for the targets to depend on a longer history of variables.

When the number of assets equals the number of states of uncertainty (complete markets), Hoelle (2015)shows that the equilibrium allocation is Pareto efficient no matter which equilibrium policy rule is implemented. For this reason, I focus exclusively on settings with incomplete markets, where the number of assets is strictly less than the number of states of uncertainty.

Under incomplete markets, the necessary condition for Pareto efficiency imposes a lower bound on the number of assets such that the asset span contains the portfolio payouts for all households required for Pareto efficiency. This necessary condition is equivalent under both stationary and dynamic policy rules. If a dynamic policy rule can be found to support Pareto efficiency, then a stationary policy rule can be found to support the same equilibrium allocation. This is because any non-stationary elements of the dynamic policy rule have no real effects on the equilibrium under Pareto efficiency as the real side of the economy is characterized by a stationary real wealth distribution.

1.1 Literature Review

This paper considers a specific type of unconventional monetary policy in which, in addition to the short-term bonds typically bought and sold, the central bank buys and sells bonds with longer maturities. In reality, unconventional monetary policy can refer not only to the purchase and sale of bonds of all maturities, but additionally to the purchase of mortgage-backed securities and other real assets. Empirical works by Magill and Quinzii (2014a), Herrenbrueck (2016), Lenza, Pill, and Reichlin (2010), Kapetanios et al. (2012), and Baumeister and Benati (2013)have analyzed the effects of government purchases, including long term treasury purchases, lending to financial institutions, liquidity to financial markets, and mortgage-backed securities, on the yield curve.

The general equilibrium framework of this paper connects it most closely to the works of Peiris and Polemarchakis (2015), Adão, Correia, and Teles (2014), and Araujo, Schommer, and Woodford (2013).2 In standard settings with exogenous probabilities, both Adão, Correia, and Teles (2014)and Peiris and Polemarchakis (2015)analyze determinacy in a complete markets setting. The indeterminacy in such a setting is only nominal, as shown in Hoelle (2015). In Adão, Correia, and Teles (2014), policy refers to the targets for the entire term structure. Such policy uniquely determines the equilibrium, notably the inflation rates. In Peiris and Polemarchakis (2015), policy refers to the total value of purchases made by the central bank, but not the composition of such purchases. In such a setting, nominal indeterminacy persists. It is only by targeting the composition of purchases that Peiris and Polemarchakis (2015)arrive at a result equivalent to Magill and Quinzii (2014b).

Peiris and Polemarchakis (2015)focus on how the targets on long-term bond yields suffice to ensure the uniqueness of the inflation rates, and do so in a model with endogenous probabilities. As in Adão, Correia, and Teles (2014), conventional monetary policy that only targets the short-term bond yields does not suffice. Peiris and Polemarchakis (2015)demonstrate how forward guidance, in which the short-term bond yield is fixed for a number of periods, replicates the result obtained when the entire term structure is targeted.

Araujo, Schommer, and Woodford (2013)is probably closest in spirit to the present paper as it focuses on the real effects of unconventional monetary policies via targeted changes in the collateral constraints.

With similar roots in the class of models with real frictions, New Monetarist models offer additional insights into unconventional monetary policy. Recent works by Williamson (2012)and Kiyotaki and Moore (2012)consider the effects of central bank purchases of liquid assets (conventional monetary policy), while Herrenbrueck (2016)considers the effects of central bank purchases of illiquid assets. These models extend the standard New Monetarist framework to allow for portfolio effects.

The New Keynesian class of models represent a complementary view of the effects of monetary policy as the models are characterized not by real frictions, but by nominal frictions.3 Eggertsson and Woodford (2003)recommend that central banks commit to forward guidance, which are rules for how the short-term interest rates will be targeted after the economy departs from the zero lower bound constraint.4 In contrast, McGough (2005)demonstrate how targets on the long-term bond yields can avoid indeterminacy problems that arise near the zero lower bound when only the short-term bond yields are targeted.5 Curdia and Woodford (2011)espouse policies that target the composition of asset purchases in lieu of policies that simply target the supply of reserves. Gertler and Karadi (2011)extend the business cycle models of Christiano, Eichenbaum, and Evans (2005)and Smets and Wouters (2007)by including credit intermediation frictions. Gertler and Karadi (2011)analyze the effects of an increase of central bank credit intermediation to fill the void left by a decrease in private credit intermediation.

Related to the analysis of term structure policies, Andrés, López-Salido, and Nelson (2004)apply the concept of imperfect substitutability between assets to argue that traditional policies of open market operations can have effects on the term structure. Chen, Curdia, and Ferrero (2012)extends this concept of imperfect substitutability in a model similar to Christiano, Eichenbaum, and Evans (2005)and Smets and Wouters (2007)to simulate the effects of unconventional monetary policies, including the purchase of long-term bonds.

The remainder of the paper is organized as follows. Section 2introduces the model. Section 3introduces the recursive formulation of competitive equilibrium and verifies the existence of a Markov equilibrium. Section 4analyzes the conditions for Pareto efficiency, under both stationary and dynamic policy rules. Section 5provide numerical examples to illustrate the effects of policy on the yield curve. Section 6concludes and the proofs are contained in Appendix A.

2 The Model

The model describes a closed economy with a single infinite-lived monetary authority (or central bank).

Time is discrete and infinite with time periods t{0,1,....}. The filtration of uncertainty follows a one-period Markov process with finite state space S={1,...,S}, where S>1. The realized state of uncertainty in any period t, denoted st, is a function only of the realized state in the previous period t1, denoted st1. This random process is characterized by a transition matrix ΓRS,S whose elements are Γ(s,s) for row s and column s.

Define the history of realizations up to and including the realization st in period t as st=(s0,s1,...,st). For convenience, Γ(st|sτ) for any t>τ refers to the probability that history st is realized conditional on the history sτ. Additionally, let st+kst refer to the Sk histories (st,σ1,...,σk)(σ1,...,σk)Sk that are realized k periods from the date-event st.6

2.1 Households

In each date-event, a finite number of household types hH={1,...,H} trade and consume a single physical commodity. Heterogeneity requires H>1.

Households receive the sequence of endowments {eh(st)}. I assume that the endowments are stationary. Define the stationary endowment mapping as eh:SR++\ such that eh(st)=eh(st) for all date-events. Denote the aggregate endowment as E:SR++ such that E(s)=hHeh(s) sS. The model permits aggregate risk, i.e., E(s)E(σ) for some s,σS. Define [e]=[eh(s)](s,h)S×H as the S×H endowment matrix. For technical reasons, the endowment vectors of households are assumed to be linearly independent

Assumption 1

HS and the endowment matrix [e] has full rank and is in general position. Specifically, this means that any H rows of [e] are linearly independent.

The consumption by household h in date-event st is denoted ch(st)R+. The sequence of consumption for household h is denoted {ch(st)}.

The household preferences are assumed to be identical and satisfy constant relative risk aversion:


Assumption 2

The discount factor β(0,1) and u(c)=c1ρ1ρ for ρ>0 and ρ1 and u(c)=ln(c) for ρ=1.

In each date-event st, the money supply is M(st)>0 and the nominal price level is p(st)>0.

Each household type contains a continuum of households, of equal mass. Households trade in perfectly competitive financial markets. The set of financial assets include a finite number J of nominally risk-free government bonds. The bonds are indexed by jJ={1,...,J}.

Asset 1 is a 1-period nominally risk-free bond, which I\ refer to as the short-term bond. The nominal payouts of a 1-period bond purchased in date-event st equal 1 for all date-events st+1st and 0 otherwise.

The set of long-term bonds is the set of jperiod bonds, where j{2,...,J}. A jperiod bond is nominally risk-free, meaning that the nominal payouts of a jperiod bond purchased in date-event st equal 1 for all date-events st+jst and 0 otherwise. The jperiod nominally risk-free bond can be freely traded in all interim periods up to the maturation date.

The nominal asset price for a jperiod bond issued in date-event st is denoted qj(st). The vector of all bond prices in date-event st is q(st)=(qj(st))jJ.

This paper focuses on economies with incomplete markets.

Assumption 3


Each date-event is divided into two subperiods. In the initial subperiod, the money market and bond markets open. Denote m^h(st) as the money holding by household h by the close of the money market in date-event st. Denote bjh(st)R as the jperiod nominal bond holding by household h by the close of the bond markets in date-event st. Each bond can either be held long or short by the household. Denote the entire portfolio as bh(st)=(bjh(st))jJRJ.

Denote ωh(st)R as the nominal wealth held by household h for use in the date-event st. The initial period value ωh(s0) is a parameter of the model. The budget constraint, at the close of the money market and bond markets in date-event st, is given by:


In the second subperiod of each date-event, the commodity market opens. The purchase of the commodity is subject to the cash-in-advance constraint:


At the same time that consumption is being purchased on the commodity market, the households receive income from selling their endowment. Denote mh(st) as the money holding of household h by the close of the commodity market in date-event st:


Given the money definition eq. (4), the cash-in-advance constraint eq. (3) can be rewritten as:


Entering into the date-events st+1st, the nominal wealth available to household h is equal to the money holding plus the portfolio payout:


For simplicity, I define q0(st)=1 for all date-events. The wealth can then be expressed as:


Updating eq. (2) using eqs. (4) and (6), the standard budget constraint is derived:


Households are permitted to short-sell the nominal bonds, so I require the following implicit debt constraint:


The household optimization problem is given by:


2.2 Monetary Authority

The portfolio of debt positions for the monetary authority in date-event st is B(st)=(Bj(st))jJR+J, where Bj(st)0 refers to the amount of debt issued in terms of the jperiod bond. In reality, the monetary authority does not issue debt, but buys or sells the debt issued by the fiscal authority. The fiscal authority’s only role in this model is the debt choice, so for simplicity I allow the monetary authority to make this choice directly. The net debt position of the fiscal and monetary authority cannot be negative. A negative net debt position implies that the monetary authority is holding more debt than was issued by the fiscal authority.

The money supply issued by the monetary authority in date-event st is M(st)>0.

In the initial period s0, the monetary authority has the nominal obligation W(s0).

The monetary authority has the following budget constraints, where the liabilities of the monetary authority are on the left-hand side of the equations and the assets of the monetary authority are on the right-hand side of the equations:



2.3 Sequential Competitive Equilibrium

Definition 1

A sequential competitive equilibrium (SCE) is the vector of household variables {ch(st),mh(st),bh(st)}hH, the monetary authority variables {B(st),M(st)}, and the price variables {p(st),q(st)} such that:

  1. Given {p(st),q(st)} and ωh(s0), each household chooses {ch(st),mh(st),bh(st)} to solve the household problem eq. (9).

  2. Given W(s0), the monetary authority variables{B(st),M(st)} satisfy eq. (10).

  3. Markets clear:

    1. hHch(st)=hHeh(st)t,st.

    2. hHωh(s0)=W(s0).

    3. hHmh(st)=M(st)t,st.

    4. jJ and t,st.

The equilibrium asset price q1(st)1. Otherwise, the market clearing condition on the bond markets is not satisfied as households prefer to save using money holdings and not bond holdings. If q1(st)<1, the cash-in-advance constraints eq. (5) will bind for all households. With binding cash-in-advance constraints eq. (5), the market clearing condition for the money markets implies that the Quantity Theory of Money holds:


The Friedman rule in date-event st is such that q1(st)=1. Under the Friedman rule, money and the 1-period bond are perfect substitutes. Market clearing for both implies that the sum of the two is pinned down for all households and the monetary authority, but not the composition. The cash-in-advance constraints eq. (5) need not bind under the Friedman rule. It is innocuous (i.e., without real effects) under the Friedman rule to set the household money holdings such that the cash-in-advance constraints eq. (5) bind. This would allow the Quantity Theory of Money eq. (11) to hold.

3 Markov Equilibrium

3.1 Constraints in Real Terms

Define the real debt positions for the monetary authority and the real bond positions for the households as B^j(st)=Bj(st)p(st) and b^jh(st)=bjh(st)p(st), respectively. The portfolios are denoted B^(st)=(B^j(st))jJ and b^h(st)=(b^jh(st))jJ, respectively. Market clearing in terms of nominal bond positions occurs if and only if market clearing in the real bond positions occurs. Additionally, define the inflation rate π(st)=p(st)p(st1).

The monetary authority constraints eq. (10) in real terms, after using the Quantity Theory of Money eq. (11), are given by:


The household problem will be recursive in terms of real wealth. Define the real wealth for household h entering date-event st as


In equilibrium, the household budget constraints in equilibrium are given by:


The first order conditions with respect to bonds b^jh(st)\ are given by:


3.2 State Space

For bonds traded in date-event st, I collect the payouts for all bonds jJ in subsequent date-events (st,σ)σS\ in the real payout matrix R(st)R+S,J, defined by


It is convenient to write the real payout matrix as a product of two underlying matrices. Algebraically,


where Π(st)=[π(st,1)000...000π(st,S)] is called the inflation matrix and Q0J1(st)=[1q1(st,1)..qJ1(st,1)::::1q1(st,S)..qJ1(st,S)] is called the bond price matrix. Each element in Q0J1(st) is inversely related to the interest rate for that particular bond and date-event. The matrix Q0J1(st) characterizes the term structure of interest rates. Since the inflation matrix Π(st) has full rank, the real payout matrix R(st) has full rank iff the bond matrix Q0J1(st) has full rank.

I restrict policy such that Q0J1(st) has full rank in all date-events. Hoelle (2014)shows that if the model includes a social welfare function for the monetary authority to maximize, the full rank condition must be satisfied unless a Pareto efficient allocation can be supported with a rank deficient bond price matrix.

Full rank condition Q0J1(st) has full rank t,st.

I utilize the recursive equilibrium concept of Markov equilibrium. The state space includes the aggregate shock realization in the current period, the wealth distribution, the vector of bond holdings for all households, and the bond prices. Define ω^(st)=(ω^h(st))hH{H}RH1 as the real wealth distribution. Aggregate resource constraints uniquely determine the total real wealth hHω^h(st), so only (H1) real wealth variables need be included in the state space. The bond prices must satisfy eq. (14), implying that q(st)Δq for all date-events, where Δq is the compact set defined by:


Definition 2

The policy rule is the function T:S×RH1Δq such that q(st)=T(st,ω^(st)) in all date-events.

Define b^(st)=(b^h(st))hHRHJ as the set of all household bond holdings. Given that B^j(st)=hHb^jh(st)0, the bond holdings must satisfy b^(st)Δb for all date-events, where Δb is the closed set defined by:


The state space is S×RH1×Δb×Δq with typical element (st,ω^(st),b^(st),q(st)). Given the equilibrium requirement of nonnegative consumption, define the equilibrium state space ΩS×RH1×Δb×Δq as the closed subset defined by: Ω={(st,ω^(st),b^(st),q(st))S×RH1×Δb×Δq:ch(st)=ω^h(st)jJqj(st)b^jh(st)0hHH}hH{H}ch(st)E(s)}.

3.3 Expectations Correspondence

Define ZσRH1×Δb×Δq as the set containing the variables (ω^(σ),b^(σ),q(σ)) such that (i) ch(σ)=ω^h(σ)jJqj(σ)b^jh(σ)0 for hH{H} and (ii)hH{H}ch(σ)E(σ). The expectations correspondence g:ΩσSZσ is defined such that for

z=(ω^,b^,q)andz(σ)=(ω^(σ),b^(σ),q(σ))\ σS,

the vector of variables (z(1),...,z(S))g(s,z) if the following conditions hold:

  1. For all σS,q(σ)=T(σ,ω^(σ)).

  2. For all jJ,B^j=hHb^jh and B^j(σ)=hHb^jh(σ)σS.

  3. For all hH{H},ch=ω^hjJqjb^jh and ch(σ)=ω^h(σ)jJqj(σ)b^jh(σ)σS.

  4. For all hH{H} and all σS,the household wealth ω^h(σ)=1π(σ)(eh(s)+jJqj1(σ)b^jh).(15)

  5. For all σS, the monetary authority constraint eq. (12): 1π(σ)(E(s)+jJqj1(σ)B^j)=E(σ)+jJqj(σ)B^j(σ).(16)

  6. For all jJ and all hH, the Euler eq. (14): qj=βσSΓ(s,σ)(ch(σ)ch)ρqj1(σ)π(σ).(17)

3.4 Markov Equilibrium Definition

A Markov equilibrium is defined by a policy correspondence V:S×RH1Δb×Δq and a transition correspondence F:graph(V)σSZσ satisfying the following two properties:

  1. For all (s,ω^,b^,q)graph(V)and all σS,Fσ(s,ω^,b^,q)gσ(s,ω^,b^,q).(18)

  2. For all (s,ω^,b^,q)graph(V)and all σS,(σ,Fσ(s,ω^,b^,q))graph(V).(19)

Theorem 1

Any Markov equilibrium in which the full rank condition is satisfied is equivalent to a sequential competitive equilibrium.


See Section A.1. □

Existence of a Markov equilibrium requires the following assumption on the policy rule.

Assumption 4

The policy rule T:S×RH1Δq is continuous.

Theorem 2

If Assumption 4 is satisfied, then a Markov equilibrium exists.


The proof of the existence of a Markov equilibrium proceeds by construction as in Duffie et al. (1994)and Kubler and Schmedders (2003). See

4 Necessary Conditions for Pareto Efficiency

Under Assumption 2, the Pareto set is characterized by the consumption fractions (θh)hHΔH1 such that ch(st)=θhE(st) for all h and for all date-events.

4.1 Stationary Policy Rules

Monetary authorities choose from the set of policy rules for which (i) a Markov equilibrium exists and (ii) the Markov equilibrium is equivalent to a sequential competitive equilibrium. Since continuity is sufficient for a Markov equilibrium to exist, monetary authorities are restricted to choose T from the set of continuous mappings for which the full rank condition is satisfied.

I denote a dynamic policy rule as any policy rule satisfying Assumption 4 and the full rank condition. A special case of a dynamic policy rule is a stationary policy rule.

Definition 3

A stationary policy rule T:S×RHΔq\ satisfies Assumption 4, the full rank condition, and specifies (qj(s))(j,s)J×S(Δq)S such that (qj(st))jJ=T(st,ω^(st)) for all date-events.

Lemma 1

If the Markov equilibrium allocation is Pareto efficient and the policy rule is stationary, then the real wealth vectors ω^h(st) are stationary, meaning that there exists (ω^h(s))(h,s)H×S such that ω^h(st)=ω^h(st) for all h and all date-events.


See Section A.3. □

4.1.1 Case 1: J<H

Case 1 also requires (H,J)(2,1).

Define the vector of stationary real household wealth as


With a stationary policy rule, the asset price matrix


The space Φ is referred to as the asset span. The asset payout matrix is (Π(st))1(Q0J1) and the asset span is defined as

(Π(st))1(Q0J1)={xRS:x=(Π(st))1(Q0J1)bfor some bRJ}.

The asset span (Π(st))1(Q0J1) is a linear subspace with dimension equal to rank(Q0J1)=J. The household budget constraints are satisfied provided that

ω^h(Π(st))1(Q0J1) \ hH.(20)

Lemma 2

If H>2, the stationary real wealth vectors have full rank, namelyrank((ω^h)hH)=H.


See Section A.4. □

Under Case 1, there does not exist an asset span Φ such that ω^hΦ hH, which implies that a stationary policy rule cannot support a Pareto efficient Markov equilibrium allocation.

4.1.2 Case 2: JH

Case 2 also includes the economy (H,J)=(2,1). Under Case 2, an asset span Φ can be mathematically found such that ω^hΦ hH, but the asset span Φ must be able to be supported by policy:


Employ the following transformation for the bond holding variables:


b~jh(st)=b^jh(st)for j>1.

The debt positions can be similarly transformed:


B~j(st)=B^j(st)for j>1.

Market clearing for the new variables holds if and only if market clearing for the original variables holds.

Lemma 3

Under Case 2, a stationary policy rule must be supported by stationary inflation rates, meaning that there exists (π(s))sS such that π(st,σ)=π(σ) for all date-events.


See Section A.5. □

Since the real wealth vectors ω^h(st) and the inflation rates π(st)\ are stationary, b~jh(st) must be constant to satisfy the wealth eq. (15):


Define the constant vectors (bjh)jS and (Bj)sS such that b~jh(st)=bjh and B~j(st)=Bj for all h, for all j, and for all date-events.

There are two equivalent arguments to derive the generic necessary conditions for Pareto efficiency.

System of Equations Argument: As a function of (q1(s))sS, the Euler equation for the short-term bond uniquely determines (π(s))sS:


The Euler equation for the longer-term bonds uniquely determine (qj(s))sS for all j>1:


From Walras’ Law, the monetary authority constraint eq. (12)\ is equivalent to the sum of the household budget constraints eq. (13). This implies a total of SH independent budget constraints:


The discounted present value budget constraints are used to determine the initial period price level p(s0) and the consumption fractions (θh)hH. Under stationary policy rules and Pareto efficiency, the initial period discounted present value constraints for households are given by: ωh(s0)p(s0)=k=0βkE0[(E(sk)E(s0))ρ(θhE(sk)q1(sk)eh(sk))].(26)7

Define the stochastic discount factor matrix Γ^ as the S×S matrix with elements Γ^(s,σ)=βΓ(s,σ)(E(σ)E(s))ρ. The initial period discounted present value constraints for households can be expressed recursively as: ωh(s0)p(s0)=(ISΓ^)(s0)1(θhE(s)q1(s)eh(s))sS,(27)

where IS is the S×S identity matrix and (ISΓ^)(s0)1 is row s0 of the matrix (ISΓ^)1.

From the recursive expression eq. (27) for all households hH, the short-term bond prices (q1(s))sS uniquely determine the initial price level p(s0) and the consumption fractions (θh)hH as follows:



Theorem 3

From the equilibrium equation for θh in eq. (28), θh[0,1] \ hH.


See Section A.6. □

The following Table 1displays equations and variables:

Table 1:

Variables for stationary policy rules.

The number of variables exceeds the number of equations when SHS+HJ.8

This is equivalent JS(H1)H.

The generic argument is completed in the proof of Lemma Lemma 4in which I show that the equations are linearly independent (generically). In other words, if J<S(H1)H, then over a generic subset of household endowments, a solution to the equations does not exist.

Lemma 4

If HJ<S(H1)H, then over a generic subset of household endowments, a stationary policy rule cannot support the Pareto efficient Markov equilibrium allocation.


See Section A.7. □

Span Argument: The set of all asset spans is the Grassmanian manifold Gr(J,S). The set Gr(J,S) is J(SJ)dimensional. If H=J, there exists a unique asset span ΦGr(J,S) in this set such that ω^hΦ hH.

If H<J, then there exists a continuum of ΦGr(J,S) such that ω^hΦ hH. The dimension of this continuum equals (JH)(SJ). In the J(SJ)dimensional set of asset spans, any asset span in a (JH)(SJ)dimensional subset satisfies ω^hΦ hH. Thus, the number of policy variables must exceed J(SJ)(JH)(SJ)=H(SJ).9

The variables under a stationary policy rule that change the asset span are (q1(s))sS. The vector (q1(s))sS is Sdimensional. The number of policy variables exceeds H(SJ) when JS(H1)H.10

The generic argument is completed when I show that if J<S(H1)H, then over a generic subset of household endowments, there does not exist a vector (q1(s))sS capable of supporting the Pareto efficient equilibrium asset span.

4.1.3 Necessary Conditions for Pareto Efficiency

The following span condition is a generic necessary condition for Pareto efficiency under stationary policy rules.

Condition SC (span condition)(H,J)=(2,1) or Jmax{H,S(H1)H}.

Theorem 4

Under Assumptions 1-4, Condition SC is a generic necessary condition for a stationary policy rule to support a Pareto efficient Markov equilibrium allocation.


The proof is immediate given the results from Lemmas Lemma 2and Lemma 4. □

4.2 Dynamic Policy Rules

The generic necessary condition is the same for the case of dynamic policy rules.

Theorem 5

Under Assumptions 1-4, Condition SC is a generic necessary condition for a dynamic policy rule to support a Pareto efficient Markov equilibrium allocation.


See Section A.8. □

If no stationary policy rule is consistent with Pareto efficiency, then no dynamic policy rule is either. The reason is that Pareto efficiency imposes a stationarity property on the real wealth distribution. While dynamic policy rules seems to offer more policy tools, they are no better at satisfying this stationarity property than stationary policy rules.

4.3 Not Sufficient for Pareto Efficiency

Condition SC is sufficient for the existence of a solution to the equilibrium eqs. (27), (23), (24), and (25). Condition SC is not sufficient for the existence of a Pareto efficient equilibrium as an equilibrium additionally requires the following equilibrium bounds:

  1. B^j(st)0j and st.

  2. q1(st)[0,1] st. 11

General sufficient conditions are not available as it is not possible to impose conditions on the size of the economy to ensure that the equilibrium bounds are satisfied. The only way to determine if a Pareto efficient equilibrium exists for a given economy is to (i) find the equilibrium variable values\ that satisfy eqs. (27), (23), (24), and (25) and (ii) verify that they satisfy the above equilibrium bounds.

The following section provides an example and illustrates the conditions under which each of the equilibrium bounds is violated.

5 Example

5.1 Economy Parameters

The economy contains J=2 assets, H=2 household types, and S=4 states of uncertainty. The necessary condition Jmax{H,S(H1)H} for J=H=2 corresponds to S4.

The specific parameter values are β=0.92, relative risk aversion ρ=4, and Markov probabilities given by Γ=[]. The household endowments are given in the following Table 2:

Table 2:


The endowment distribution provides the economy with two useful properties. First, the economy contains aggregate risk, with aggregate endowments monotonic in σ. Second, the endowment vectors are negatively correlated (in this case, the vectors have perfect negative correlation)\ and this generates the strongest incentives for trade and the largest portfolio effects. The choice of endowment distribution has no effect on equilibrium existence.

The initial period is s0=1 and the initial period wealth parameters are ω1(s0)=ω2(s0)=10.

The final subsection demonstrates that the existence of a Pareto efficient equilibrium is robust to changes in the parameters (β,ρ,Γ,ω(s0)).

5.2 Equilibrium Values

With only H=2 household types, θ1 uniquely characterizes the Pareto efficient allocation. The equilibrium consumption fraction is θ1=0.434. The initial price level is given by p(s0)=0.307. With ω1(s0)=ω2(s0)=10, the initial real wealth ω^h(s0)=ωh(s0)p(s0) (recall s0=1) is given by ω^1(s0)=ω^2(s0)=32.609.

The optimal policy choice supports the welfare-maximizing Pareto efficient allocation in equilibrium. The policy choice is such that the inflation rates are given in Table 3.

Table 3:

Stochastic inflation rates.

The stationary asset prices, which are the policy choices of the monetary authority, are given in Table 4.

Table 4:

Asset prices.

It is straightforward to verify that the Euler equations for assets j{1,2} and states σ{1,2,3,4} are satisfied given the inflation rates and asset prices in Table 3and Table 4:

qj(σ)=βψSΓ(σ,ψ)(E(ψ)E(σ))ρqj1(ψ)π(ψ)\ j{1,2},σ{1,2,3,4}.

Given the initial price level p(s0) and the stochastic inflation rates (π(σ))σS, it is straightforward to determine the time series for prices {p(st)}. Nevertheless, it is easier to express household wealth as real wealth ω^h(st)=ωh(st)p(st)\ and portfolio choices as the real bond holdings b^jh(st)=bjh(st)p(st). By definition, the real wealth distribution is given by:


Per the equilibrium construction, eh(s)+b^1h(st) is constant and b^2h(st) is constant across all date-events. This implies that both real wealth and real bond holdings have a stationary distribution, meaning that there exists (ω^h(σ))σS such that ω^h(st,σ)=ω^h(σ) and there exists (b^jh(σ))σS such that b^jh(st,σ)=b^jh(σ) for all date-events. The real wealth distribution for households is given in Table 5:

Table 5:

Wealth distribution.

Household h=2 has higher endowments (on average), which results in higher consumption than household h=1. The wealth distribution is not directly tied to consumption. Its values are determined to match the initial wealth positions (equal for both households) and support the equilibrium allocation. In other words, although household h=1 receives higher portfolio payouts (as represented by the real wealth values), it also has larger portfolio expenditures (compare Table 6and Table 7below). The real effect is lower consumption for household h=1 relative to h=2.

The equilibrium budget constraints are given by:


In equilibrium, ch(st,σ)=θhE(σ) h{1,2} and σ{1,2,3,4}. Table 6and Table 7illustrate that the budget constraints hold for both households in all states:

Table 6:

Household h=1 budget constraint.

Table 7:

Household h=2 budget constraint.

Walras’ Law indicates that the real monetary authority debt positions consistent with market clearing must satisfy the monetary authority constraints σ{1,2,3,4}. As with household bond holdings, there exists (B^j(σ))σS such that B^j(st,σ)=B^j(σ) for all date-events. Table 8below displays the debt positions, which are verified to be nonnegative.

Table 8:

Monetary authority debt positions.

5.3 Robustness

The example is robust to changes in economic parameters. Recall that the economic parameters are Γ=[], β=0.92, ρ=4, and ω1(s0)=ω2(s0)=10. Define the transition matrix in terms of a single persistence parameter ϕ such that Γ=[ϕ1ϕ31ϕ31ϕ31ϕ3ϕ1ϕ31ϕ31ϕ31ϕ3ϕ1ϕ31ϕ31ϕ31ϕ3ϕ]. Table 9below captures the ceterus paribus robustness analysis concerning the existence of a Pareto efficient equilibrium. For example, the row for ϕ displays the range of values for ϕ over which a Pareto efficient equilibrium exists, while holding fixed all parameters at the example values except ϕ. Table 9also provides the reason for non-existence outside this range (the reason for non-existence for values below the lower bound on the far left and the reason for non-existence for values above the upper bound on the far right). Similar analysis occurs for the remaining 3 rows.

Table 9:

Robustness analysis over parameters (ϕ,β,ρ,ω1(s0)).

The error "exceeds error bounds" means that the computational algorithm was unable to converge to a solution satisfying the error bounds of the algorithm. Due to the ʼnear’ collinearity in the payout matrix required for a solution, the algorithm is unable to find a solution. Such a solution would not be practical anyway as it prescribes bond positions exceeding 104. Despite the limitations of the model at the lower and upper bounds, it should be clear from Table 9that the example considered above is robust to changes in economic parameters.

6 Conclusion

This paper has analyzed the real effects of policy that targets the term structure of interest rates. In a setting of incomplete markets, it is possible for policy to be chosen to support Pareto efficiency and the necessary conditions for Pareto efficiency are characterized. The necessary conditions are equivalent for both stationary and dynamic policy rules and require that the number of financial markets must be sufficient to span the vector of portfolio payouts for all households.

With this theoretical foundation, a violation of the necessary condition leads to Pareto inefficient allocations for all policy rules (generically). Future work will analyze the welfare implications of dynamic policy rules relative to stationary policy rules when the necessary condition is violated.


A Proof of Theorem 1(Equivalence)

To show that Markov equilibria satisfy the SCE definition, the Euler eq. (17) must be necessary and sufficient for household optimality. Necessary is immediate. Sufficiency follows as in Proposition 3.2 from Duffie et al. (1994)since the full rank condition is satisfied.

B Proof of Theorem 2(Existence)

One necessary condition for a SCE is that b^(st) lies in a compact set for all date-events. Denote the compact set for bonds as ΔbΔb. The set Δq is already compact.

Define the initial correspondence V0:S×RH1Δb×Δq such that V0(s,ω^)=Δb×Δq for all (s,ω^)S×RH1. Define the operator G that maps from the correspondence Vn:S×RH1Δb×Δq to a new correspondence Vn+1:S×RH1Δb×Δq as follows:

Vn+1(s,ω^)={(b^,q)Δb×Δq:(ω^(σ),q(σ),b^(σ))σSsuch that(i) (b^(σ),q(σ))Vn(σ,ω^(σ))for all σS(ii) (b^,(ω^(σ))σS)satisfy eq. (15) and (17)}.

In words, given the correspondence Vn:S×RH1Δb×Δq, the solution to two nonlinear systems of equations determines the image of the new correspondence Vn+1:S×RH1Δb×Δq.

The first nonlinear system of equations takes as given (ω^(σ))σS. Therefore, for all σS, (b^(σ),q(σ)) is determined by:


Given (ω^(σ),b^(σ),q(σ)), the household budget constraint determines a unique value for ch(σ) (for all households):


From the policy rule, q=T(s,ω^). There exists a unique vector (b^,(π(σ))σS) such that the following equations are satisfied: (i) the Euler eqs. (17) for all J assets and all H households given the consumption equation ch=ω^hjJqjb^jh, and (ii) the monetary authority constraints eq. (16) for all σS given the market clearing conditions B^j=hHb^jh and B^j (σ)=hHb^jh(σ).

To verify uniqueness, since the asset prices (q(σ))σS and the consumption (ch(σ))σS vectors are fixed, the right-hand side of the Euler eq. (17) is a strictly decreasing and continuous function of b^.12 With HJ Euler equations and HJ variables b^, a unique solution is guaranteed.

The second nonlinear system of equations takes as given (b^,(π(σ),q(σ))σS). There exists a unique real wealth vector ω^h(σ) for all σS and all hH from eq. (15).

Define V:S×RH1Δb×Δq such that

V(s,ω^)=n=0Vn(s,ω^)\ for all (s,ω^)S×RH1.

Lemma 5

If the policy rule satisfies Assumption 4, then a Markov equilibrium exists.


The argument follows as in Theorem 1 of Kubler and Schmedders (2003). Since a SCE exists for all parameters, then Vn is a well-defined correspondence for all n0. Since T is continuous, the graph of g is a closed subset of Ω×σSZσ. Since the graph of g is a closed subset of Ω×σSZσ, then Vn(s,ω^) is closed for all n0 and all (s,ω^)S×RH1. The images are nested, by construction:\ Vn+1(s,ω^)Vn(s,ω^) for all n0 and all (s,ω^)S×RH1. The infinite intersection n=0Vn(s,ω^) of nested, closed, and nonempty sets is itself nonempty, and this holds for all (s,ω^)S×RH1. Thus, the policy correspondence V:S×RH1Δb×Δq is well-defined.

Additionally, the policy correspondence V:S×RH1Δb×Δq is upper hemicontinuous. The initial correspondence is upper hemicontinuous (as it is compact) and since the images of subsequent correspondences are closed and nested subsets of the image of the initial correspondence, then upper hemicontinuity is preserved.

C Proof of Lemma 1

The household budget constraints for date-event st is given by:


The exact same constraint can be written for date-event (st,σ):


Multiply both sides of eq. (30) by β(ch(st,σ)ch(st))ρ, take the conditional expectation, and cite the Euler eq. (14):


The real household wealth vectors are defined as:


Inserting the new expression (31) back into the date-event st budget constraint eq. (29) and iterating forward yields the discounted present value equation


after citing the transversality condition. The discounted present value eq. (32) must hold in all date-events st.

Under Pareto efficiency, there exist fractions (θh)hH such that ch(st+k)=θhE(st+k) h,k. Under a stationary policy rule, q1(st+k)=q1(st+k) for all date-events. Under both Pareto efficiency and a stationary policy rule, the discounted present value eq. (32) is given by:


The right-hand side of the discounted present value eq. (33) in date-event st only depends on the realization st (and this holds for all date-events). Therefore, the real wealth variables ω^h(st) are stationary.

D Proof of Lemma 2

I first consider the special case of (H,J)=(2,1). By eq. (15), the total real wealth (hHω^h(st,σ))σS(E(σ)+jJqj(σ)hHb^jh(st,σ))σS. By eq. (16), (1π(st,σ))σS=(E(σ)+jJqj(σ)hHb^jh(st,σ)E(s)+jJqj1(σ)hHb^jh(st))σS. For the special case of J=1, (hHω^h(st,σ))σS=(1π(st,σ))σS and therefore (hHω^h(st,σ))σS(Π(st))1(Q0J1) .13 For stationary wealth vectors and H=2:


In equilibrium, rank(ω^1,ω^2)=1, so only J=1 asset is required such that


Recall the discounted present value eq. (33):


This can be expressed recursively as


where Γ^ is the S×S matrix with elements Γ^(s,σ)=βΓ(s,σ)(E(σ)E(s))ρ.

Define zh=(θhE(s)q1(s)eh(s))sS. Since (ISΓ^)1 has full rank,


I claim that rank((zh)hH)=H.

D.1 Case 1: (q1(s))sS1

Consider ((zh)hH)v=0 for any vector vRH. This implies that


Subcase 1(a): If hHvhθh=0, Assumption 1 implies v=0 and rank((zh)hH)=H.

Subcase 1(b): Consider hHvhθh0. By definition, E(s)=hHeh(s). Define μh=vhhHvhθh. The system of eqs. (34) is given by:


By Assumption 1, there exists SS with #S=H satisfying both (q1(s))sS1 and [e](h,s)H×S is an invertible matrix. Consider only the states sS. In matrix notation:


Since [e](h,s)H×S is invertible and diag((q1(s))sS) is a diagonal matrix, the unknown coefficients (μh)hH must satisfy:


Since (q1(s))sS1, then μh1 hH, with strict inequality for at least one h. By definition, this means vhhHvhθh hH, with strict inequality for at least one h. Multiplying both sides θh and summing:


This contradicts that hHθh=1.

There does not exist vRH such that eq. (34) is satisfied. Subcase 1(b) is not possible.

D.2 Case 2: (q1(s))sS=1

Suppose, in order to obtain a contradiction, that (q1(s))sS=1. The Euler equations applied iteratively implies (qj(s))sS=1 j>1. All together, rank(Q0J1)=1 under the Friedman rule.

I claim rank(ω^h,ω^h)=2 for any h,h. Set (zh,zh)(vhvh)=0 for any vector (vhvh)R2. The system of eq. (34) for h,h is:


The system eq. (35) is equivalently expressed as


for χh=1μh, χh=1μh, and χk=1 kH{h,h}. Since rank[e]=H, then [χh]hH=0. Since H{h,h} is nonempty, the contradiction χk=1 and χk=0 completes the argument.

This implies that rank((ω^h)hH)2 and rank(Q0J1)2. Case 2 is not possible.14

E Proof of Lemma 3

With stationary asset prices, the Euler eq. (14) imply that the inflation rates π(st,σ) only depend upon (st,σ). The sequence of inflation rates {π(st)} must satisfy


for all date-events.

I claim that the asset span is constant. For the special case of H=2, only J=1 asset suffices for ω^hΦ hH and there exists a unique asset span Φ. When H=J>2, !Φ such that ω^hΦ hH. If H<J, the asset span need not be constant. However, it is innocuous (in terms of the real allocation) to restrict attention to policies with a constant asset span Φ=(Π(st))1(Q0J1).

A constant asset span with stationary asset price vector dictates that (π(st,σ))σS are proportional for all date-events. Since ω^hΦ hH, then (b~h(st))hH such that ω^h=(Π(st))1(Q0J1)b~h(st) hH. If (π(st,σ))σS=κ(π(sτ,σ))σS for some (t,τ), then b~h(st)=κb~h(sτ) hH. Using the market clearing condition, B~(st)=κB~(sτ). The monetary authority constraint eq. (12) for date-events (st,σ) and (sτ,σ) yield:


This implies either that κ=1 or B~(st)=0 for all date-events.

If κ=1, (π(st,σ))σS=(π(sτ,σ))σS, b~h(st)=b~h(sτ), and B~(st)=B~(sτ) for all date-events. The inflation rates are stationary, meaning that there exists (π(s))sS such that π(st,σ)=π(σ) for all date-events.

If B~(st)=0 for all date-events, the monetary authority constraint eq. (12) implies that (q1(s))sS=1. If q1(s)=1 s, then Q0J1 has rank 1, which violates the full rank condition. Additionally, Case 2 in the proof of Lemma Lemma 2rules out q1(s)=1 s. This contradiction implies that κ=1 must hold and the inflation rates are stationary.

F Proof of Theorem 1

Recalling the expression for θh:


The matrix (ISΓ^)(s0)1 contains only positive elements. Since 0q1(s)1 sS, then θh0.

Select any s. Since eh(s)E(s), I claim that


Algebraically, the previous inequality reduces to:


Since eh(s)E(s), then (1ωh(s0)W(s0)eh(s)E(s)ωh(s0)W(s0))1 and the inequality q1(s)1 is always satisfied. The inequality eq. (36) holds sS. Therefore, θh1.

G Proof of Lemma 4

It suffices to show that if J<S(H1)H, then generically over the subset of household endowments (eh(s))(h,s)H×S and initial period wealth (ωh(s0))hH, the H+SJ+SH eqs. (27), (23), (24), and (25) are linearly independent.

The variables ξRH+SJ+S+HJ are


and the parameters θRH(S+1) are


Define the system of equations as


where Φ(ξ,θ)=0 iff the following equations are satisfied: (i) initial period discounted present value eq. (27), (ii) Euler eqs. (23) and (24), and (iii) budget constraints eq. (25).

Define the projection ϕ:RH+SJ+S+HJ+H(S+1)RH(S+1) as the mapping (ξ,θ)θ such that Φ(ξ,θ)=0. The mapping ϕ is proper iff for any compact subset of the range Y, the inverse image ϕ1(Y) is also compact. The payout matrix has full rank since the only variables considered are those with linearly independent payouts. This implies that the projection ϕ is proper.

To complete the argument, it suffices to prove that Dξ,θΦ(ξ,θ) has full row rank H+SJ+SH and apply the parametric transversality theorem.

Consider the columns for derivatives with respect to the price variables p(s0) and the parameters (ωh(s0))hH. The submatrix in Dξ,θΦ(ξ,θ) for the initial period discounted present value eq. (27) and the derivatives with respect to p(s0) and (ωh(s0))hH has full row rank. Moreover, the variables p(s0) and the parameters (ωh(s0))hH do not appear in any other equations. The initial period discounted present value equations are therefore linearly independent of all other equations.

Consider the budget constraints eq. (25):


Recall that Q0J1=[1q1(1)..qJ1(1)::::1q1(S)..qJ1(S)]. Define the stationary inflation rate matrix Π=[π(1)000...000π(S)]. From the Euler equation, the expression (qj(s)qj1(s)π(s))sS is given by (Γ^IS)Π1Q0J1. The updated household budget constraints are given by:


where bh is a Jdimensional column vector. To this system of equations, I add the total endowment expressions


to account for the fact that individual endowments and total endowment are related.

I claim that the submatrix in Dξ,θΦ(ξ,θ) for the budget constraints eq. (25) and the derivatives with respect to (π(s))sS and (eh(s))(h,s)H×S has full row rank. A sufficient condition for full row rank is that for any


the product


implies αT=0, where Dπ,eΦ(ξ,θ) is the submatrix in Dξ,θΦ(ξ,θ) for the budget constraints eq. (25) and the derivatives with respect to (π(s))sS and (eh(s))(h,s)H×S.

Consider the columns for the derivatives with respect to (eh(s))(h,s)H×S. The equations αTDπ,eΦ(ξ,θ)=0 for these columns are:


Since this holds h, then ΔahT=ΔaT h.

Consider the columns for the derivatives with respect to (1π(s)). The equations αTDπ,eΦ(ξ,θ)=0 for these columns are:


where (Γ^IS)(s) refers to row s of the matrix Γ^IS. Since ΔahT=ΔaT h, then we obtain Δa(s)(Γ^IS)(s)Q0J1hHbh=0.

I claim hHbh0. If, in order to obtain a contradiction, hHbh=0, then B=0. This violates the inequality B^1(st)0 as B1=B~1(st)=E(st)+B^1(st) and E(st)>0.

I claim (Γ^IS)(s)Q0J1hHbh0. If (Γ^IS)(s)Q0J1hHbh=0, then


The budget constraints require that hH(θhE(s)q1(s)eh(s))=E(s)(1q1(s))=0. If q1(s)=1 s, then Q0J1 has rank 1, which violates the full rank condition. Additionally, Case 2 in the proof of Lemma Lemma 2rules out q1(s)=1 s.

For the special case of q1(s)=1, replace one of the household budget constraints (suppose, without loss of generality, the budget constraint for household H) in state s with the Euler equation for q1(s):


Consider the columns for the derivatives with respect to (1π(s)). The equations αTDπ,eΦ(ξ,θ)=0 for these columns are:


Recall that Δah(s)=Δa(s) h. From the budget constraint with q1(s)=1,


The sum (Γ^IS)(s)Q0J1hH{H}bh=(Γ^IS)Q0J1bH. Equation (40) is updated to:


Without loss of generality, the household H is chosen such that θHE(s)eH(s)<0. From Assumption 1, such a household exists. Since


and {π(s)(θHE(s)eH(s))+βΓ(s,s)}>0, then Δa(s)=0.

For the special case with q1(s)=1 and q1(s)=1, replace the (s,H) budget constraint with 1=βσSΓ(s,σ)(E(σ)E(s))ρ1π(σ) and the (s,H) budget constraint with 1=βσSΓ(s,σ)(E(σ)E(s))ρ1π(σ). Without loss of generality, household H is chosen such that θHE(s)eH(s)<0 and household H is chosen such that θHE(s)eH(s)>0. It may be the case that H=H. Consider the columns for the derivatives with respect to (1π(s)) and (1π(s)). Using eq. (41), the equations αTDπ,eΦ(ξ,θ)=0 for these columns are:


If either Γ(s,s)=0 or Γ(s,s)=0, then Δa(s)=Δa(s)=0 as before. If Γ(s,s)>0 and Γ(s,s)>0, the first equation implies Δa(s)Δa(s)0 and the second equation implies Δa(s)Δa(s)0. This is only satisfied if either Δa(s)=0 or Δa(s)=0. Once one of them equals 0, then Δa(s)=Δa(s)=0 as before.

The same approach works for any number of states s such that q1(s)=1, provided that (q1(s))sS1.

Therefore Δa(s)=0 and the same argument holds s, implying ΔahT=0 hH. From eq. (38), ΔEs=0 sS. This implies αT=0, so Dπ,eΦ(ξ,θ) has full row rank.

The submatrix in Dξ,θΦ(ξ,θ) for the Euler eqs. (23) and (24) and the derivatives with respect to (qj(s))(j,s)J×S has full row rank. Given the above findings, the Euler equations are linearly independent of all other equations.

This completes the argument.

H Proof of Theorem 5

Pareto efficiency requires that for all state variables (s,ω^), the corresponding consumption fractions (θh)hH are constant. Consider the recursive derivation of the policy correspondence. Take as given the policy correspondence Vn:S×RH1Δb×Δq. For a given vector of state variables (s,ω^), the following algorithm determines the image Vn+1(s,ω^) for a dynamic policy rule consistent with Pareto efficiency. Guess the Pareto efficient consumption fractions (θh)hH. For each σS, the state variables ω^(σ) are determined such that: (i) (b^(σ),q(σ))Vn(σ,ω^(σ)) and (ii) θhE(σ)=ω^h(σ)jJqj(σ)bjh(σ) for hH{H}.15 Recall that ω^(σ)=(ω^h(σ))hH{H} does not include the real wealth for household h=H. For household h=H, ω^H(σ)=θHE(σ)+jJqj(σ)bjH(σ). Notice that the equilibrium values (ω^(σ),b^(σ),q(σ))σS only depend on (θh)hH and not (s,ω^) (at this point in the iteration).

The equilibrium variables (b^h)hH are determined from the simplified version of the wealth eqs. (15):


The wealth equation for household h=H is redundant. The number of unknowns equals HJ and the number of equations equals S(H1). Using the same arguments as in the proof of Lemma Lemma 4, if J<S(H1)H, then over a generic subset of household endowments, there does not exist a solution to the wealth equations. Using contraposition, then JS(H1)H is a generic necessary condition for a solution.

Given (b^h)hH, the inflation rates (π(σ))σS are uniquely determined from the monetary authority constraints eq. (16):


Given (π(σ))σS, the asset prices q are determined from the Euler equations under Pareto efficiency:


Notice that the equilibrium values ((b^h)hH,(π(σ))σS,q) only depend on (θh)hH and not s (at this point in the iteration).

To complete the determination of the image Vn+1(s,ω^), the initial guess (θh)hH must be updated. The consumption fractions are updated such that:


Iteration continues until convergence to (θh)hH for each element (s,ω^) in the domain of Vn+1:S×RH1Δb×Δq.

Pareto efficiency requires that (θh)hH is identical for state variables (s,ω^A) and (s,ω^B). Using the algorithm above, identical consumption fractions (θh)hH imply identical values for (ω^(σ),b^(σ),q(σ))σS and ((b^h)hH,(π(σ))σS,q). Returning to the iterative step, eq. (42) is satisfied iff ω^A=ω^B. Pareto efficiency implies that the the real wealth vectors ω^h(st) are stationary, meaning that there exists (ω^h(s))(h,s)H×S such that ω^h(st)=ω^h(st) for all h and all date-events.

With stationary real wealth vectors, the recursive representation beginning in date-event st is given by:


Matrix algebra requires that q1(st,σ) are stationary, namely that there exists (q1(s))sS such that q1(st,σ)=q1(σ) for all date-events. Lemma 2can now be applied.

From Lemma 2, if H>2, then rank((ω^h)hH)=H. The necessary condition for ω^h(Π(st))1(Q0J1) hH is JH or (H,J)=(2,1).

In conclusion, Condition SC is a generic necessary condition for Pareto efficiency under dynamic policy rules.


The author wishes to thank Zhigang Feng, Anson Soderbery, and the anonymous referees for helpful feedback. Comments are welcome at mhoelle@purdue.edu or (+1) 765-496-2737. The most recent version of the paper can be found at www.matthew-hoelle.com/1/75/resources/document\_690\_1.pdf.


  • Adão, Bernardino, Isabel Correia, and Pedro Teles. 2014. “Short and Long Interest Rate Targets.” Journal of Monetary Economics 66: 95–107. Web of ScienceCrossrefGoogle Scholar

  • Andrés, Javier, J., David López-Salido, and Edward Nelson. 2004. “Tobin’s Imperfect Asset Substitution in Optimizing General Equilibrium.” Journal of Money Credit and Banking 36 (5): 665–690. CrossrefGoogle Scholar

  • Araujo, Aloiso, Susan Schommer, and Michael Woodford. Conventional and Unconventional Monetary Policy with Endogenous Collateral Constraints. National Bureau of Economic Research Working Paper Series 2013 19711. 

  • Barro, Robert J., and David B. Gordon. 1983. “Rules, Discretion and Reputation in a Model of Monetary Policy.” Journal of Monetary Economics 12: 101–121. CrossrefGoogle Scholar

  • Baumeister, Christiane, and Luca Benati. 2013. “Unconventional Monetary Policy and the Great Recession: Estimating the Macroeconomic Effects of a Spread Compression at the Zero Lower Bound.” International Journal of Central Banking 9 (2): 165–212. Google Scholar

  • Bernanke, Ben S., and Ilian Mihov. 1998. “Measuring Monetary Policy.” Quarterly Journal of Economics 113 (3): 869–902.CrossrefGoogle Scholar

  • Calvo, Guillermo A., and Pablo E. Guidotti. 1993. “On the Flexibility of Monetary Policy: The Case of the Optimal Inflation Tax.” Review of Economic Studies 60 (3): 667–687. CrossrefGoogle Scholar

  • Chari, V.V., Lawrence J. Christiano, and Patrick Kehoe. 1991. “"Optimal Fiscal and Monetary Policy: Some Recent Results.” Journal of Money, Credit and Banking 23 (3): 519–539. CrossrefGoogle Scholar

  • Chen, Han, Vasco Curdia, and Andrea Ferrero. 2012. “The Macroeconomic Effects of Large-Scale Asset Purchase Programmes.” Economic Journal 122 (564): 289–315. CrossrefWeb of ScienceGoogle Scholar

  • Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 1999. Taylor, J.B, and M. Woodford (Eds.), Handbook of Macroeconomics 1A. 65–148. Amsterdam: Elsevier Science Monetary Policy Shocks: What Have we Learned and to What End?.Google Scholar

  • Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 2005. “Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy.” Journal of Political Economy 113 (1): 1–45. CrossrefGoogle Scholar

  • Clarida, Richard, Jordi Gali, and Mark Gertler. 2000. “Monetary Policy Rules and Macroeconomic Stability: Evidence and Some Theory.” Quarterly Journal of Economics 115 (1): 147–180.CrossrefGoogle Scholar

  • Curdia, Vasco, and Michael Woodford. 2011. “The Central-Bank Balance Sheet as an Instrument of Monetary Policy.” Journal of Monetary Economics 58 (1): 54–79. Web of ScienceCrossrefGoogle Scholar

  • Dong, Bingbing. 2014. Forward guidance and credible monetary policy. University of Virgina Working paper.., https://sites.google.com/site/uvabd3h/Research. Google Scholar

  • Duffie, Darrell, John Geanakoplos, Andreu Mas-Colell, and McLennan. Andrew. 1994. “Stationary Markov Equilibria.” Econometrica 62 (4): 745–781. CrossrefGoogle Scholar

  • Eggertsson, Gauti B., and Michael Woodford. 2003. “The Zero Bound on Interest Rates and Optimal Monetary Policy.” Brookings Papers on Economic Activity 34 (1): 139–233. Google Scholar

  • Gagnon, Joseph., Matthew Raskin, Julie Remache, and Brian Sack. 2011. Large-Scale Asset Purchases by the Federal Reserve: Did they work?. Federal Reserve Bank of New York: 41-59 Economic Policy Review.Google Scholar

  • Jordi, Galí 1992. How Well Does the IS-LM Model Fit Postwar US Data?" Quarterly Journal of Economics. 709–738. 2.Google Scholar

  • Gertler, Mark, and Peter Karadi. 2011. “A Model of Unconventional Monetary Policy.” Journal of Monetary Economics 58: 17–34. Web of ScienceCrossrefGoogle Scholar

  • Herrenbrueck, Lucas. 2016. Quantitative easing and the liquidity channel of monetary policy". Working paperhttp://herrenbrueck.weebly.com/uploads/1/3/3/9/13395560/herrenbrueck-2016-qe-liquidity.pdf. Google Scholar

  • Hoelle, Matthew. Quantitative Easing Under Incomplete Markets: Optimality Conditions for Stationary Policy. Krannert Working Paper Series1277, 2014. 

  • Hoelle, Matthew. 2015. “On the Pareto Efficiency of Term Structure Targeting Policies.” Mathematical Social Sciences 77: 62–69. Web of ScienceCrossrefGoogle Scholar

  • Kapetanios, George, Haroon Mumtaz, Ibrahim Stevens, and Konstantinos Theodoridis. 2012. “Assessing the Economy-Wide Effects of Quantitative Easing.” Economic Journal 122 (564):. F316-F347. Web of ScienceGoogle Scholar

  • Kiyotaki, Nobuhiro, and John Moore. Liquidity, Business Cycles, and Monetary Policy. 2012:17934 NBER Working Paper No. 

  • Krishnamurthy, Arvind, and Annette Vissing-Jorgensen. The Effects of Quantitative Easing on Interest Rates: Channels and Implications for Policy. 2011:17555 NBER Working Paper No.

  • Kubler, Felix, and Karl Schmedders. 2003. “Stationary Equilibria in Asset-Pricing Models with Incomplete Markets and Collateral.” Econometrica 71 (6): 1767–1793. CrossrefGoogle Scholar

  • Kydland, Finn E., and Edward C. Prescott. 1977. “Rules Rather than Discretion: The inconsistency of Optimal Plans.” Journal of Political Economy 85 (3): 473–492. CrossrefGoogle Scholar

  • Lenza, Michele, Huw Pill, and Lucrezia Reichlin. 2010. “Monetary Policy in Exceptional Times.” Economic Policy 25 (62): 295–339. CrossrefGoogle Scholar

  • Robert Lucas, E.Jr., and Nancy L. Stokey. 1983. “Optimal Fiscal and Monetary Policy in an Economy with Capital.” Journal of Monetary Economics 12: 55–93.CrossrefGoogle Scholar

  • Magill, Michael, and Martine Quinzii. 1992. “Real Effects of Money in General Equilibrium.” Journal of Mathematical Economics 21: 301–342. CrossrefGoogle Scholar

  • Magill, Michael, and Martine Quinzii. 1996. 1. Cambridge: MIT Press Theory of Incomplete Markets, Vol. Google Scholar

  • Magill, Michael, and Martine Quinzii. 2014a. “Anchoring Expectations of Inflation.” Journal of Mathematical Economics 50: 86–105. CrossrefWeb of ScienceGoogle Scholar

  • Magill, Michael, and Martine Quinzii. 2014b. “Term Structure and Forward Guidance as Instruments of Monetary Policy.” Economic Theory 56: 1–32. CrossrefWeb of ScienceGoogle Scholar

  • McGough, 2005. “Bruce, Glenn D. Rudebusch, and John C. Williams.” Journal of Monetary Economics 52 (5): 855–879. Using a Long-Term Interest rate as the Monetary Policy Instrument. Google Scholar

  • Peiris, Udara, and Herakles Polemarchakis. 2015. “Monetary policy and quantitative easing in an open economy: Prices, exchange rates and risk premia". CRETA Discussion Series. University of Warwick, No. 9. Google Scholar

  • Sargent, Thomas J., and Neil Wallace. 1975. “’Rational’ Expectations, the Optimal Monetary Instrument, and the Optimal Money Supply Rule.” Journal of Political Economy 83 (2): 241–254. CrossrefGoogle Scholar

  • Schmitt-Grohé, Stephanie, and Martín Uribe. 2004. “Optimal Fiscal and Monetary Policy Under Sticky Prices.” Journal of Economic Theory 114: 198–230. CrossrefGoogle Scholar

  • Sims, Christopher. 1992. “Interpreting the Macroeconomic Time Series Facts: The Effects of Monetary Policy.” European Economic Review 36: 975–1011. CrossrefGoogle Scholar

  • Smets, Frank, and Rafael Wouters. 2007. “Shocks and Frictions in US Business Cycles: A Bayesian DSGE Approach.” American Economic Review 97 (3): 586–606. CrossrefGoogle Scholar

  • Taylor, John B. 1999. “The Robustness and Efficiency of Monetary Policy Rules as Guidelines for Interest Rate Setting by the European Central Bank.” Journal of Monetary Economics 43: 655–679. CrossrefGoogle Scholar

  • Uhlig, Harald. 2005. “What are the Effects of Monetary Policy on Output? Results from an Anostic Identification Procedure.” Journal of Monetary Economics 52 (2): 381–419. CrossrefGoogle Scholar

  • Williamson, Stephen D. 2012. “Liquidity, Monetary Policy, and the Financial Crisis: A New Monetarist Approach.” American Economic Review 102 (6): 2570–2605. Web of ScienceCrossrefGoogle Scholar

  • Woodford, Michael. 2003. Interest and Prices: Foundations of a Theory of Monetary Policy. Princeton, NJ: Princeton University Press. Google Scholar

  • Woodford, Michael. 2005. “Comment on: ’Using a Long-Term Interest Rate as the Monetary Policy Instrument’.” Journal of Monetary Economics 52 (5): 881–887. CrossrefGoogle Scholar


  • 1

    In a finite-horizon model with a single commodity, this asset span condition is satisfied under identical utility of the CARA or CRRA form when the endowment vector of each household is contained in the asset span. 

  • 2

    Classic papers in the real business cycle tradition that focus on monetary policy include Barro and Gordon (1983), Calvo and Guidotti (1993), Chari, Christiano, and Kehoe (1991), Kydland and Prescott (1977), and Sargent and Wallace (1975). In terms of general equilibrium models with incomplete markets and monetary policy, the initial papers were Kydland and Prescott (1977). 

  • 3

    A representative sample of the important papers using this class of models in a closed economy include Schmitt-Grohé and Uribe (2004), Sims (1992), Taylor (1999), Uhlig (2005), Williamson (2012), and Woodford (2003, 2005). 

  • 4

    Recent work by Dong (2014)shows that policies of forward guidance are not credible, and the time consistent policies look quite different than the proposed rules of Eggertsson and Woodford (2003). 

  • 5

    Woodford (2005)offers a critique that summarizes the main points from Eggertsson and Woodford (2003). 

  • 6

    When k=0, the set st+kst refers to the date-event st. 

  • 7

    The derivation is contained in the proof of Lemma 1. 

  • 8

    For the special case of (H,J)=(2,1), the number of variables equals S+1 as the bond holdings b2 are redundant given b1 and the fact that ω^1Φ and ω^2Φ, where Φ is 1-dimensional. The number of equations equals S as the budget constraints for household h=2 are redundant given the budget constraints of h=1. The number of variables exceeds the number of equations as S+1 ≥ S is satisfied for any value of S. 

  • 9

    For the special case of (H,J) = (2,1), a policy of dimension J(S-J)= (S-1) is required. 

  • 10

    For the special case of (H,J) = (2,1), the number of policy variables is S and this must exceed S-1 from the previous footnote, which is trivially satisfied. 

  • 11

    If q1 (st) ϵ [0, 1], Euler equations imply that q q(st)ϵ[0,qq(st)] for all jϵ{2,...,J} 

  • 12

    An increase in b^jh means an increase in B^j from the market clearing condition. An increase in B^j means an increase in π(σ) from the monetary authority constraint eq. (16) (for all states σS). An increase in b^jh means a decrease in ch (as qj is positive) and a decrease in (1ch)ρ. Thus, the right-hand side of the Euler eq. (17) is a strictly decreasing function. 

  • 13

    For the general case with J>1, (hHω^h(st,σ))σS(Π(st))1(Q0J1) only holds if (ω^h(st,σ))σS(Π(st))1(Q0J1) for all hH, which requires that (Π(st))1(Q0J1) has at least H dimensions (from Lemma 2). 

  • 14

    Additionally, this argument reveals that the Friedman rule (when H>2) is never consistent with a Pareto efficient allocation. 

  • 15

    The policy rule is recursively updated. At this stage in the iteration, q(σ)=Tn(σ,ω^(σ)). As the policy correspondence converges, so too does the policy rule. 

About the article

Published Online: 2017-07-27

Citation Information: The B.E. Journal of Theoretical Economics, Volume 18, Issue 1, 20160158, ISSN (Online) 1935-1704, DOI: https://doi.org/10.1515/bejte-2016-0158.

Export Citation

© 2017 Walter de Gruyter GmbH, Berlin/Boston. Copyright Clearance Center

Citing Articles

Here you can find all Crossref-listed publications in which this article is cited. If you would like to receive automatic email messages as soon as this article is cited in other publications, simply activate the “Citation Alert” on the top of this page.

Comments (0)

Please log in or register to comment.
Log in