This is not the document you are looking for? Use the search form below to find more!

0.00 (0 votes)

This paper analyzes a small open economy Ramsey growth model with convex investment costs and a collateral constraint on borrowing. Optimal control methods are used to characterize the dynamics of investment, consumption, and debt. The analysis demonstrates that the economy's adjustment speed depends on the fraction of the capital stock that can be used as collateral. In the presence of non-convexities, a higher loan-to-value of the capital stock may produce a bifurcation in the dynamics by increasing the economy's adjustment speed. In contrast to the canonical small open economy model with convex investment costs, domestic and foreign savings are growth-rate complements due to the interaction between domestic savings, the price of capital, and the borrowing constraint. The standard closed economy Ramsey model, the Cohen-Sachs debt repudiation model, and the canonical small open economy model with adjustment costs are shown to be special cases of the analysis.

- Added:
**May, 17th 2010** - Reads:
**242** - Downloads:
**0** - File size:
**264.67kb** - Pages:
**35** - content preview

- Username: samanta
- Name:
**samanta** - Documents:
**1258**

Related Documents

Motivated by Japan's economic experiences and policy debates over the past two decades, this paper uses an open economy dynamic stochastic general equilibrium model to examine the volatility and ...

This paper provides a simple dynamic neo-Keynesian model that can be used to analyze the impact of monetary policy that considers inflation targeting in a small open economy. This economy is ...

The authors assess the stabilization properties of simple monetary policy rules within the context of a small open-economy model constructed around the limited-participation assumption and calibrated ...

In this paper, we develop adynamic stochastic general equilibrium (DSGE) model for an open economy, and estimate it on Euro area data using Bayesian estimation techniques. The model incorporates ...

This paper computes welfare maximizing Taylor-style interest rate rules, in a business cycle model of a small open economy. The model assumes staggered price setting and shocks to domestic ...

We study how determinacy and learnability of global rational expectations equilibrium maybe affected by monetary policy in a simple, two country, New Keynesian framework. The two blocks ...

This paper proposes an open-economy Phillips Curve that features a real exchange rate channel. The resulting target rule under optimal policy from a timeless perspective (TP) involves additional ...

This paper considers implications of the neoliberal shift for forms of national integration that were achieved in sub-Saharan Africa during the developmentalist era. National integration strategies ...

Creating Spatio-Temporal Contiguities Between Real And Virtual Rooms In An Assistive Living Environment

Irrespective of whether discretion or commitment to a binding rule guides the conduct of monetary policy, the existence of a direct exchange rate channel in the Phillips Curve causes the behavior of ...

Content Preview

January 2009

COLLATERAL CONSTRAINTS AND MACROECONOMIC ADJUSTMENT

IN AN OPEN ECONOMY

Philip L. Brock

Department of Economics

University of Washington

Seattle, Washington 98195

plbrock@u.washington.edu

Abstract

This paper analyzes a small open economy Ramsey growth model with convex investment

costs and a collateral constraint on borrowing. Optimal control methods are used to

characterize the dynamics of investment, consumption, and debt. The analysis demonstrates

that the economy’s adjustment speed depends on the fraction of the capital stock that can be

used as collateral. In the presence of non-convexities, a higher loan-to-value of the capital

stock may produce a bifurcation in the dynamics by increasing the economy’s adjustment

speed. In contrast to the canonical small open economy model with convex investment costs,

domestic and foreign savings are growth-rate complements due to the interaction between

domestic savings, the price of capital, and the borrowing constraint. The standard closed

economy Ramsey model, the Cohen-Sachs debt repudiation model, and the canonical small

open economy model with adjustment costs are shown to be special cases of the analysis.

1

I. Introduction

During the late 1970s and early 1980s economists began a concerted effort to develop

optimizing intertemporal models of open economies1. As part of this push various attempts

were made to open the Ramsey growth model to capital flows. Simply opening the standard

Ramsey model produces instantaneous adjustment of the capital stock, since domestic saving is

no longer a constraint on capital formation. In order to slow down the adjustment of the capital

stock two principal approaches were taken. The first opened the Uzawa (1961, 1964) two-

sector growth model and made the investment goods sector nontradable.2 The second

approach incorporated convex investment costs into the basic Ramsey model so that the cost of

investment increased with the rate of investment. Both approaches slowed the speed of capital

stock adjustment and created a relative price of investment goods to consumption goods that

could be interpreted as a real exchange rate.

The adoption of the second approach was initially slowed by the difficulty of

integrating the microeconomics literature on firm adjustment costs, such as Lucas and Prescott

(1973), into a Ramsey model with a representative price-taking agent. Hayashi (1982) showed

that linearly homogenous convex adjustment costs were required for the specification of a

Ramsey model with price-taking firms. 3 Hayashi’s paper proved to be instrumental in the

development of the open economy Ramsey model. In short order papers by Blanchard (1983),

Lipton and Sachs (1983), and Giavazzi and Wyplosz (1984) incorporated linearly

1 See, e.g., Sachs 1981 and Svensson and Razin 1983 for early models.

2 Papers laying the groundwork for this approach include Fischer and Frenkel (1972) and Bruno (1976). See

Murphy (1986), Brock (1988), and Obstfeld (1989) for early examples of explicitly intertemporal optimizing open

economy models with nontraded investment

3Without a linearly homogeneous installation cost function, firms are price makers that generate monopoly rents.

Hayashi (1982) characterized the problem as one of determining the assumptions necessary to make the marginal

“q” valuation of the capital stock the same as the Tobin’s “q” average value of the capital stock.

2

homogeneous investment costs into open economy versions of the Ramsey model, and many

other papers soon followed.4

By the 1990s the open economy Ramsey model with convex investment costs had been

incorporated into graduate textbooks and became one of the canonical models of international

macroeconomics.5 Nevertheless, there has remained the question of how to close this model

with a borrowing constraint. Any infinite-horizon Ramsey model must impose a no-Ponzi-

game condition on the representative agent, which in the presence of complete markets is

equivalent to a solvency condition requiring that the value of debt cannot exceed the present

value of the economy’s net capital income plus wage income.6 The solvency condition

generates an upper bound on debt and a lower bound of zero on consumption. For standard

utility functions that imply positive consumption (e.g., ones that satisfy the Inada conditions),

assumptions in addition to the solvency condition must be incorporated to solve for the

model’s steady state equilibrium.

As is well known, if both the rate of time preference ( ρ ) and the world interest rate (*r*)

are taken as constants, then the existence of a steady state for the small open economy model

requires, in addition to the solvency condition, that ρ =*r *. This in turn gives rise to a zero root

in the dynamics, so that the steady state depends on initial debt and consumption equals

4 See, e.g., Matsuyama (1987), Brock (1988), and Turnovsky and Sen (1991).

5 See, e.g., Blanchard and Fischer (1989), Obstfeld and Rogoff (1995), Turnovsky (1997), and Barro and Saleh-i-

Martin (2004). Convex investment costs also have an increasingly important role in dynamic stochastic general

equilibrium models such as Christiano, Eichenbaum and Evans (2005) and Adolfson et. al (2007).

6 See Levine and Zame (1996). This condition can be expressed algebraically as

∞

− ( − )

*b *≤ *q k *+

(

*w k *) *r s t*

*e*

*ds*

∫

*t*

*t*

*t*

*s*

*t*

where*b* is the stock of debt, *qk *is the value of the capital stock , (

*w k*) is the real wage, and *r* is the world interest

rate. With convex, linearly homogenous investment costs, the value of the capital stock equals the present

discounted value of future rentals on capital net of adjustment costs.

3

permanent income.7 For many purposes modelers may wish to have steady states that are

invariant to initial levels of debt. One alternative way to close the model is to pin down long-

run consumption (*c *) by assuming that the rate of time preference is increasing in consumption

(e.g. Obstfeld 1981). A second alternative is to specify the borrowing rate as an increasing

function of the amount borrowed where the steady-state level of debt (*b *) is assumed to satisfy

the solvency constraint (see, e.g., Schmidt-Grohé and Uribe 2003). The assumption I make in

this paper is that debt must be collateralized by a fraction*a* of the value of the capital stock.

Current and future labor income cannot be pledged against repayment of a loan.8

Evans and Jovanovic (1989) were among the first to use a collateral requirement on

capital in an optimizing model.9 Cohen and Sachs (1986) were the first to introduce a

collateral constraint in an open economy Ramsey model with convex investment costs, but

their model proved to be analytically tractable only for linear production functions. Mendoza

(2008) has recently used a collateral constraint, but focuses on whether or not the constraint is

7 Three papers in the early 1980s that developed techniques which permitted the analytical characterization of

open economy models with zero roots are Blanchard and Kahn (1980), Buiter (1984), and Giavazzi and Wyplosz

(1985).

8 The four alternative closure assumptions generate the following alternative steady-state borrowing constraints,

where ρ is the rate of time preference,*r *is the world interest rate, and *b *, *k *, and *c *are steady-state values of

debt, capital, and consumption:

(

*w k *) − *c *(*b *)

0

.

*i *ρ = *r*

⇒

*b *= *k *+

*r*

(

*w k *) − *c *(*r*)

.

*ii *ρ (*c*) = *r*

⇒

*b *= *k *+

*r*

(

*w k *) − *c*

*ii *.

*i *ρ = *r*(*b*%)

⇒

*b *(ρ ) = *k *+

*r*

.

*iv b *≤ *aqk*

⇒

*b *≤ *ak*

0 ≤*a *≤ 1

Given the steady-state capital stock (*k *), the first closure assumption determines the steady-state level of debt as a

function of the initial stock of debt (*b *). The second assumption determines steady-state debt by pinning down

0

steady-state consumption via the exogenous world interest rate, while the third assumption pins down steady-state

debt by equating the exogenous rate of time preference with the endogenous debt-determined interest rate.

If*a *= 1 and the collateral constraint binds, consumption will equal the real wage in equilibrium.

9 The paper by Aiyagari and Gertler (1999) is also an early closed-economy model that employs a collateral

constraint of the type used in this paper.

4

binding rather than on the effect of fractional changes in the collateral requirement* *when the

constraint is binding. Barro, Mankiw and Sala-i-Martin (1995) and Lane (2001) modify the

Cohen-Sachs model by replacing convex investment costs with an additional factor of

production that requires nontraded investment. Caballero and Krishnamurthy (2001),

Devereux and Poon (2004), Chari, Kehoe, and McGrattan (2005), and Braggion, Christiano,

and Roldós (2007) also employ collateral constraints, but like Mendoza (2008) restrict their

analyses to the change in model dynamics when there is a switch from a binding to a non-

binding constraint.

Several papers motivate the collateral constraint by the ability of foreign debtors to

seize the capital stock in the case of default (see, e.g., Barro et. al. 1995, Lane 2001, and Chari

et. al. 2005). In this paper, like that of Cohen and Sachs, international lenders cannot seize the

capital stock due to adjustment costs that fix capital in the short run, so the borrowing

constraint refers to the amount of present and future capital income that can be pledged against

the debt. This limit on pledgeable income could reflect the ability of foreign creditors to

impose a penalty equal to a fraction of the economy’s output in the case of default, as in Cohen

and Sachs, or it could reflect the inability of agents to pledge wage income because of the

inalienability of human capital as in Kiyotaki and Moore (1997).10 Following Matsuyama

(2008) and Mendoza (2008), I assume that the collateral requirement captures the outcome of

unmodeled agency problems or legal restrictions that prevent the representative agent from

pledging more than a fraction of capital income when incurring debt.

10 There are, of course, other reasons related to adverse selection and moral hazard. Similarly, Holmstrom and

Tirole (1998) posit a private benefit that cannot be pledge by an entrepreneur to creditors.

5

2. The Model

The open economy model with convex investment costs is well-known and is exposited by a

number of graduate-level textbooks, such as Blanchard and Fischer (1989, 58-69) and

Turnovsky (1997, 57-77). The canonical form of the model is the following, where*f *(*k*) is a

neoclassical production function, ψ (*i*, *k*) represents convex investment costs, *b* is debt, *r* is the

world interest rate, ρ is the rate of time preference, and*u*(*c*) is a concave utility function:

∞

−ρ

max*u*(*c*)

*t*

*e*

*dt*

*subject to k *, *b and*

∫

0

0

*c*,*i*

0

*b*& = *c *+ψ (*i *, *k *) + *rb *− *f *(*k *)

(1)

*t*

*t*

*t*

*t*

*t*

*t*

*k*& = *i *− δ *k*

*t*

*t*

*t*

With the rate of time preference set equal to the interest rate, the model is separable in

production and consumption decisions. With a linearly homogenous investment costs, the

value of the capital stock (*qk*) is equal to the sum of the discounted value of rental income from

capital net of investment costs:

∞

−

−

*q k *=

*f *′ *k k *−ψ *i k*

*e*

*ds*

∫

*t*

*t*

[

*s*

*s*

*s*

*s *]

(

)

( )

( ,

)

*r s t*

(2)

*t*

Consumption is constant and is equal to permanent income:

∞

∞

−

−

−

−

*c *= −*rb *+ *r*

*f *(*k*) −ψ (*i*, *k*)

*r s t*

*e*

*ds *= *r*(*q k *− *b *) + *r*

(

*w k *) *r s t*

*e*

*d*

∫

∫

*s*

0

[

] ( )

(

)

(3)

*o *0

0

*t*

0

0

Consumption is constant because the agent can use the world capital market to create an

annuity from current debt and future income.

6

In this paper the agent has more limited access to the world capital market. This

limited access takes the form of a borrowing constraint:

*b *≤ *aq k * (4)

*t*

*t*

*t*

The agent can pledge only a fraction*a* of the capital stock as collateral and no part of wage

income. The capital stock cannot literally serve as collateral since adjustment costs prevent its

direct pledging to creditors. Since the value of the capital stock equals the present value of

capital income net of investment costs, the better interpretation of the borrowing constraint is

that at most a fraction*a* of present and future net capital income can be pledged to lenders.

This section looks at the effect of the constraint when all net capital income can be

pledged (*a *= 1 ). Sections 5 and 6 will generalize the model to allow *a * to take on fractional

values. The assumption*a *= 1 is the assumption made by Kiyotaki and Moore (1997). It

signifies that capital income can be pledged to lenders, but that borrowers cannot pledge their

wage income. The borrowing constraint*b *≤ *q k * is then added to the maximization problem

*t*

*t*

*t*

given by (1). The present value Hamiltonian associated with the maximization problem) is:

ρ*t*

*e H *= *u*(*c*) + λ [ *f *(*k*) − *rb *− *c *−ψ (*i*, *k*)] + *q **(*i *− δ *k*) + γ *(*qk *− *b*) (5)

where −λ is the shadow value of debt,*q ** is the shadow value of installed capital, and γ * is

the shadow value attached to the borrowing constraint. The first-order necessary conditions

are:

*u *(

′*c*) = λ (6)

ψ (*i*,*k*) = *q * (7)

*i*

where*q *= *q ** λ is the market value of capital. The Euler equation for consumption is:

7

*c*

*c*& =

(*r *+ γ − ρ)

σ

(8)

where σ = −*cu *(

′′*c*) *u *(′*c*) and γ = γ * λ .

The evolution of the price of capital is given by

*q*& = − *f *(′*k*) +ψ (*i*, *k*) + (*r *+ δ )*q * (9)

*k*

The state equation for capital is derived from the first order condition ψ (*i*, *k *) = *q *(equation 7):

*i*

*k*& = *i*(*q*, *k*) − δ *k * (10)

These latter two equations are identical to the canonical model. The consumption equation

differs from the canonical open economy Ramsey model by the term γ in the consumption

Euler equation. Thus, when*a *= 1 the borrowing constraint of this model alters the Euler

equation while leaving the dynamics of capital stock adjustment unchanged from the canonical

model. Capital accumulation with the borrowing constraint*a *= 1 is governed by the same

speed of adjustment as the canonical model:

2

µ

*r*

⎛*r *⎞

*f *′ (

′*k*)

1

*k *= *k *+ (*k *− *k *)

*t*

*e * where µ

=

−

*t*

0

m

(11)

1,2

⎜ ⎟

2

⎝ 2 ⎠

ψ*ii*

are the characteristic roots of equations (9) and (10). The adjustment dynamics are the same

for the canonical and borrowing-constrained economy because with*a *= 1 borrowing for

investment is completely collateralized by future capital income.

With a binding constraint (*b=qk*) the current account equation (1) implies that

consumption in the borrowing-constrained economy is equal to the real wage:

*b*& = *c *+ψ (*i*, *k*) + *rqk *− *f *(*k *) = *q*&*k *+ *qk*&

⇓

(12)

*c *= *f *(*k*) − *f *'(*k*)*k *=

(

*w k *)

8

The borrowing constraint will bind in equilibrium if γ = ρ −*r *> 0 where γ is the

shadow value of the borrowing constraint in steady-state equilibrium. As distinct from the

canonical model, where ρ =*r * is required for a stationary steady state, the borrowing

constrained model will have a stationary steady state provided that*r *≤ ρ . In essence, when the

autarchic interest rate ( ρ ) exceeds the world interest rate (*r*), imposing a borrowing constraint

is an alternative assumption to imposing equality between the world interest rate and the rate of

time preference in the steady state, including the assumptions that the interest rate is an

increasing function of the stock of debt or that the rate of time preference is a decreasing

function of consumption.

In summary, with a binding collateral constraint*b *= *q k *, the capital stock is financed

*t*

*t*

*t*

by borrowing from foreigners. Agents cannot borrow against future wage income so that

consumption is equal to the real wage rather than to a constant value as in the canonical model.

3. Nonbinding borrowing constraint

The borrowing constraint may be non-binding (*b *< *q k *) under several conditions. First, if the

*t*

*t*

*t*

initial capital stock (*k *) is sufficiently large, the agent can borrow against it to perfectly

0

smooth consumption, assuming that ρ =*r *. Second, if the initial capital stock is greater than

the long-run capital stock, labor income will decline over time. In this case the agent will

accumulate foreign assets to smooth consumption so that the borrowing constraint will not

bind. Once again, a stationary steady state will exist provided that ρ =*r *. In these two cases,

consumption will equal permanent income and the model’s dynamics will be the same as those

of the canonical model.

9

In addition, there is a third possibility. The initial capital stock may not be large

enough to allow the agent to borrow against it to perfectly smooth consumption, but it may be

large enough to allow the agent to temporarily smooth consumption by borrowing against it.

In this case, the dynamics are described by a temporary period in which consumption exceeds

wage income before the borrowing constraint becomes binding and consumption equals wage

income.11

The canonical model is described by four dynamic equations in*k*, *q*, *b*, and *c*.12 The

first two eigenvalues, which correspond to capital and the price of capital, are given by

equation (11). Debt accumulation and consumption dynamics are governed by the third and

fourth eigenvalues,

µ = 0, µ =*r *

3

4

During the temporary period in which the borrowing constraint is not binding, the

trajectory of consumption will be governed by the third and fourth eigenvalues as follows:

µ

µ

3*t*

4*t*

*c *= *c *+ *A e*

+*A e*

*t*

1

2

*rt*

=*c *+ *A *+ *A e*

1

2

where*A * and *A * are constants that are determined by the model. If the borrowing constraint

1

2

is never binding, adjustment is saddlepath so that

µ3*t*

*c *= *c *+ *A e*

*t*

1

11 See Turnovsky (1997), pp. 94-98 for a discussion and several mathematical examples of this kind of temporary

dynamics.

12 The linearized system (with ρ =*r *) can be written as:

⎛*q*&⎞ ⎛ *r*

−*f *(

′′*k*) 0 0⎞⎛ *q *− *q *⎞

⎜ ⎟ ⎜

⎟⎜

⎟

*k*&

1 ψ

0

0

0

*k *−

⎜ ⎟

*k*

*ii*

⎜

⎟⎜

⎟

=

⎜ ⎟ ⎜

&

− ′

⎟⎜

⎟

*b*

1 ψ

*f *(*k*)

*r*

1

−

*ii*

*b b*

⎜ ⎟ ⎜

⎟⎜

⎟

⎜ ⎟

⎜

⎟

⎝*c*& ⎠ ⎝ 0

0

0

0 ⎠⎝*c *− *c *⎠

COLLATERAL CONSTRAINTS AND MACROECONOMIC ADJUSTMENT

IN AN OPEN ECONOMY

Philip L. Brock

Department of Economics

University of Washington

Seattle, Washington 98195

plbrock@u.washington.edu

Abstract

This paper analyzes a small open economy Ramsey growth model with convex investment

costs and a collateral constraint on borrowing. Optimal control methods are used to

characterize the dynamics of investment, consumption, and debt. The analysis demonstrates

that the economy’s adjustment speed depends on the fraction of the capital stock that can be

used as collateral. In the presence of non-convexities, a higher loan-to-value of the capital

stock may produce a bifurcation in the dynamics by increasing the economy’s adjustment

speed. In contrast to the canonical small open economy model with convex investment costs,

domestic and foreign savings are growth-rate complements due to the interaction between

domestic savings, the price of capital, and the borrowing constraint. The standard closed

economy Ramsey model, the Cohen-Sachs debt repudiation model, and the canonical small

open economy model with adjustment costs are shown to be special cases of the analysis.

1

I. Introduction

During the late 1970s and early 1980s economists began a concerted effort to develop

optimizing intertemporal models of open economies1. As part of this push various attempts

were made to open the Ramsey growth model to capital flows. Simply opening the standard

Ramsey model produces instantaneous adjustment of the capital stock, since domestic saving is

no longer a constraint on capital formation. In order to slow down the adjustment of the capital

stock two principal approaches were taken. The first opened the Uzawa (1961, 1964) two-

sector growth model and made the investment goods sector nontradable.2 The second

approach incorporated convex investment costs into the basic Ramsey model so that the cost of

investment increased with the rate of investment. Both approaches slowed the speed of capital

stock adjustment and created a relative price of investment goods to consumption goods that

could be interpreted as a real exchange rate.

The adoption of the second approach was initially slowed by the difficulty of

integrating the microeconomics literature on firm adjustment costs, such as Lucas and Prescott

(1973), into a Ramsey model with a representative price-taking agent. Hayashi (1982) showed

that linearly homogenous convex adjustment costs were required for the specification of a

Ramsey model with price-taking firms. 3 Hayashi’s paper proved to be instrumental in the

development of the open economy Ramsey model. In short order papers by Blanchard (1983),

Lipton and Sachs (1983), and Giavazzi and Wyplosz (1984) incorporated linearly

1 See, e.g., Sachs 1981 and Svensson and Razin 1983 for early models.

2 Papers laying the groundwork for this approach include Fischer and Frenkel (1972) and Bruno (1976). See

Murphy (1986), Brock (1988), and Obstfeld (1989) for early examples of explicitly intertemporal optimizing open

economy models with nontraded investment

3Without a linearly homogeneous installation cost function, firms are price makers that generate monopoly rents.

Hayashi (1982) characterized the problem as one of determining the assumptions necessary to make the marginal

“q” valuation of the capital stock the same as the Tobin’s “q” average value of the capital stock.

2

homogeneous investment costs into open economy versions of the Ramsey model, and many

other papers soon followed.4

By the 1990s the open economy Ramsey model with convex investment costs had been

incorporated into graduate textbooks and became one of the canonical models of international

macroeconomics.5 Nevertheless, there has remained the question of how to close this model

with a borrowing constraint. Any infinite-horizon Ramsey model must impose a no-Ponzi-

game condition on the representative agent, which in the presence of complete markets is

equivalent to a solvency condition requiring that the value of debt cannot exceed the present

value of the economy’s net capital income plus wage income.6 The solvency condition

generates an upper bound on debt and a lower bound of zero on consumption. For standard

utility functions that imply positive consumption (e.g., ones that satisfy the Inada conditions),

assumptions in addition to the solvency condition must be incorporated to solve for the

model’s steady state equilibrium.

As is well known, if both the rate of time preference ( ρ ) and the world interest rate (

are taken as constants, then the existence of a steady state for the small open economy model

requires, in addition to the solvency condition, that ρ =

in the dynamics, so that the steady state depends on initial debt and consumption equals

4 See, e.g., Matsuyama (1987), Brock (1988), and Turnovsky and Sen (1991).

5 See, e.g., Blanchard and Fischer (1989), Obstfeld and Rogoff (1995), Turnovsky (1997), and Barro and Saleh-i-

Martin (2004). Convex investment costs also have an increasingly important role in dynamic stochastic general

equilibrium models such as Christiano, Eichenbaum and Evans (2005) and Adolfson et. al (2007).

6 See Levine and Zame (1996). This condition can be expressed algebraically as

∞

− ( − )

(

∫

where

rate. With convex, linearly homogenous investment costs, the value of the capital stock equals the present

discounted value of future rentals on capital net of adjustment costs.

3

permanent income.7 For many purposes modelers may wish to have steady states that are

invariant to initial levels of debt. One alternative way to close the model is to pin down long-

run consumption (

(e.g. Obstfeld 1981). A second alternative is to specify the borrowing rate as an increasing

function of the amount borrowed where the steady-state level of debt (

the solvency constraint (see, e.g., Schmidt-Grohé and Uribe 2003). The assumption I make in

this paper is that debt must be collateralized by a fraction

Current and future labor income cannot be pledged against repayment of a loan.8

Evans and Jovanovic (1989) were among the first to use a collateral requirement on

capital in an optimizing model.9 Cohen and Sachs (1986) were the first to introduce a

collateral constraint in an open economy Ramsey model with convex investment costs, but

their model proved to be analytically tractable only for linear production functions. Mendoza

(2008) has recently used a collateral constraint, but focuses on whether or not the constraint is

7 Three papers in the early 1980s that developed techniques which permitted the analytical characterization of

open economy models with zero roots are Blanchard and Kahn (1980), Buiter (1984), and Giavazzi and Wyplosz

(1985).

8 The four alternative closure assumptions generate the following alternative steady-state borrowing constraints,

where ρ is the rate of time preference,

debt, capital, and consumption:

(

0

.

⇒

(

.

⇒

(

⇒

.

⇒

0 ≤

Given the steady-state capital stock (

function of the initial stock of debt (

0

steady-state consumption via the exogenous world interest rate, while the third assumption pins down steady-state

debt by equating the exogenous rate of time preference with the endogenous debt-determined interest rate.

If

9 The paper by Aiyagari and Gertler (1999) is also an early closed-economy model that employs a collateral

constraint of the type used in this paper.

4

binding rather than on the effect of fractional changes in the collateral requirement

constraint is binding. Barro, Mankiw and Sala-i-Martin (1995) and Lane (2001) modify the

Cohen-Sachs model by replacing convex investment costs with an additional factor of

production that requires nontraded investment. Caballero and Krishnamurthy (2001),

Devereux and Poon (2004), Chari, Kehoe, and McGrattan (2005), and Braggion, Christiano,

and Roldós (2007) also employ collateral constraints, but like Mendoza (2008) restrict their

analyses to the change in model dynamics when there is a switch from a binding to a non-

binding constraint.

Several papers motivate the collateral constraint by the ability of foreign debtors to

seize the capital stock in the case of default (see, e.g., Barro et. al. 1995, Lane 2001, and Chari

et. al. 2005). In this paper, like that of Cohen and Sachs, international lenders cannot seize the

capital stock due to adjustment costs that fix capital in the short run, so the borrowing

constraint refers to the amount of present and future capital income that can be pledged against

the debt. This limit on pledgeable income could reflect the ability of foreign creditors to

impose a penalty equal to a fraction of the economy’s output in the case of default, as in Cohen

and Sachs, or it could reflect the inability of agents to pledge wage income because of the

inalienability of human capital as in Kiyotaki and Moore (1997).10 Following Matsuyama

(2008) and Mendoza (2008), I assume that the collateral requirement captures the outcome of

unmodeled agency problems or legal restrictions that prevent the representative agent from

pledging more than a fraction of capital income when incurring debt.

10 There are, of course, other reasons related to adverse selection and moral hazard. Similarly, Holmstrom and

Tirole (1998) posit a private benefit that cannot be pledge by an entrepreneur to creditors.

5

2. The Model

The open economy model with convex investment costs is well-known and is exposited by a

number of graduate-level textbooks, such as Blanchard and Fischer (1989, 58-69) and

Turnovsky (1997, 57-77). The canonical form of the model is the following, where

neoclassical production function, ψ (

world interest rate, ρ is the rate of time preference, and

∞

−ρ

max

∫

0

0

0

(1)

With the rate of time preference set equal to the interest rate, the model is separable in

production and consumption decisions. With a linearly homogenous investment costs, the

value of the capital stock (

capital net of investment costs:

∞

−

−

∫

[

(

)

( )

( ,

)

(2)

Consumption is constant and is equal to permanent income:

∞

∞

−

−

−

−

(

∫

∫

0

[

] ( )

(

)

(3)

0

0

0

Consumption is constant because the agent can use the world capital market to create an

annuity from current debt and future income.

6

In this paper the agent has more limited access to the world capital market. This

limited access takes the form of a borrowing constraint:

The agent can pledge only a fraction

income. The capital stock cannot literally serve as collateral since adjustment costs prevent its

direct pledging to creditors. Since the value of the capital stock equals the present value of

capital income net of investment costs, the better interpretation of the borrowing constraint is

that at most a fraction

This section looks at the effect of the constraint when all net capital income can be

pledged (

values. The assumption

signifies that capital income can be pledged to lenders, but that borrowers cannot pledge their

wage income. The borrowing constraint

given by (1). The present value Hamiltonian associated with the maximization problem) is:

ρ

where −λ is the shadow value of debt,

the shadow value attached to the borrowing constraint. The first-order necessary conditions

are:

′

ψ (

where

7

(

σ

(8)

where σ = −

′′

The evolution of the price of capital is given by

The state equation for capital is derived from the first order condition ψ (

These latter two equations are identical to the canonical model. The consumption equation

differs from the canonical open economy Ramsey model by the term γ in the consumption

Euler equation. Thus, when

equation while leaving the dynamics of capital stock adjustment unchanged from the canonical

model. Capital accumulation with the borrowing constraint

speed of adjustment as the canonical model:

2

µ

⎛

′

1

=

−

0

m

(11)

1,2

⎜ ⎟

2

⎝ 2 ⎠

ψ

are the characteristic roots of equations (9) and (10). The adjustment dynamics are the same

for the canonical and borrowing-constrained economy because with

investment is completely collateralized by future capital income.

With a binding constraint (

consumption in the borrowing-constrained economy is equal to the real wage:

⇓

(12)

(

8

The borrowing constraint will bind in equilibrium if γ = ρ −

shadow value of the borrowing constraint in steady-state equilibrium. As distinct from the

canonical model, where ρ =

constrained model will have a stationary steady state provided that

autarchic interest rate ( ρ ) exceeds the world interest rate (

is an alternative assumption to imposing equality between the world interest rate and the rate of

time preference in the steady state, including the assumptions that the interest rate is an

increasing function of the stock of debt or that the rate of time preference is a decreasing

function of consumption.

In summary, with a binding collateral constraint

by borrowing from foreigners. Agents cannot borrow against future wage income so that

consumption is equal to the real wage rather than to a constant value as in the canonical model.

3. Nonbinding borrowing constraint

The borrowing constraint may be non-binding (

initial capital stock (

0

smooth consumption, assuming that ρ =

the long-run capital stock, labor income will decline over time. In this case the agent will

accumulate foreign assets to smooth consumption so that the borrowing constraint will not

bind. Once again, a stationary steady state will exist provided that ρ =

consumption will equal permanent income and the model’s dynamics will be the same as those

of the canonical model.

9

In addition, there is a third possibility. The initial capital stock may not be large

enough to allow the agent to borrow against it to perfectly smooth consumption, but it may be

large enough to allow the agent to temporarily smooth consumption by borrowing against it.

In this case, the dynamics are described by a temporary period in which consumption exceeds

wage income before the borrowing constraint becomes binding and consumption equals wage

income.11

The canonical model is described by four dynamic equations in

first two eigenvalues, which correspond to capital and the price of capital, are given by

equation (11). Debt accumulation and consumption dynamics are governed by the third and

fourth eigenvalues,

µ = 0, µ =

3

4

During the temporary period in which the borrowing constraint is not binding, the

trajectory of consumption will be governed by the third and fourth eigenvalues as follows:

µ

µ

3

4

+

1

2

=

1

2

where

1

2

is never binding, adjustment is saddlepath so that

µ3

1

11 See Turnovsky (1997), pp. 94-98 for a discussion and several mathematical examples of this kind of temporary

dynamics.

12 The linearized system (with ρ =

⎛

−

′′

⎜ ⎟ ⎜

⎟⎜

⎟

1 ψ

0

0

0

⎜ ⎟

⎜

⎟⎜

⎟

=

⎜ ⎟ ⎜

&

− ′

⎟⎜

⎟

1 ψ

1

−

⎜ ⎟ ⎜

⎟⎜

⎟

⎜ ⎟

⎜

⎟

⎝

0

0

0 ⎠⎝

## Add New Comment