You are on page 1of 20

REAL RETURN BONDS AND SUNSPOTS

William Lim1
If investors form rational expectations such that, on average, their inflation forecasts are correct, is there value to indexing the returns on a nominal bond? In this paper, I find dynamic equivalence of the general equilibrium model of McCallum (1984) to one with real return bonds. Furthermore, if Ricardian Equivalence (Barro, 1974) holds, a moneydemand externality results in sunspot equilibria or extrinsic uncertainty, and indexing bond returns does not mitigate sunspots nor reduce nonfundamental volatility. The excess volatility that investors face is consistent with the experience of investing in certain countries described in the extant literature. Field of Research: Finance Finance Theory, Bond Markets

1.

Introduction and Literature Review

According to Robert Shiller (2003), the worlds first inflation-indexed bonds were issued by the Commonwealth of Massachusetts in 1780 during the Revolutionary War. However, the concept of indexed bonds did not reappear until the twentieth century, when Irving Fisher made the advocacy of inflation-indexed bonds a lifelong campaign (Shiller, 2003, p.4). Finland introduced indexed bonds in 1945, Israel and Iceland in 1955, and Brazil in 1964. Later Chile (1966), Columbia (1967), Argentina (1972), the United Kingdom (1975), Mexico (1989) and other countries introduced indexed bonds. Since 1980, indexed bonds have largely been issued by industrialized countries, characterized both by low inflation rates and price-stability-oriented monetary policies: Australia (1985), Canada (1991), Sweden (1994), New Zealand (1995), the United States (1997), France (1998), Italy (2003) and Greece (2003), amongst others. In a speech Christian Noyer (2004), the Governor of the Bank of France, he said that the rationale behind the issuance of these indexed bonds was to develop complete financial markets. The Governing Council of the European Central Bank (ECB) sometimes refers to the information delivered by the changes in the
Associate Professor of Finance, School of Administrative Studies, York University, 4700 Keele Street, Toronto, Ontario M3J 1P3, Canada, Email: limw@yorku.ca. Comments from Chris Ball, Dale Domian and XiaoFei Li and helpful discussions with Gady Jacoby and Peggy Ng are appreciated. All errors remain my own.
1

Electronic copy available at: http://ssrn.com/abstract=2167722

inflation-indexed bond yields both in its introductory statement and in its comprehensive assessment of economic and financial developments, as presented, for instance, in the ECBs Monthly Bulletin. As far as the euro area is concerned, the market for indexed bonds has developed significantly in recent years such that in terms of market capitalization, the overall euro area indexed bond market is now the third biggest in terms of market capitalization. The ECB has used indexed bonds to monitor long-term expected inflation the breakeven inflation rate. However, as Noyer (2004) points out, this break-even inflation rate may be biased due to institutional biases and technical biases (like the inflation lag). There is also the circularity problem inherent in indexed bond yields, where misperceptions from both sides (central bank and investors) could generate aggregate instability, characterized by unsustainable market developments and the emergence of financial imbalances. Although there has been extensive empirical work on inflation-indexed bonds,2 surprisingly, there has not been the same extensive theoretical analysis of these bonds in dynamic general equilibrium frameworks.3 In this paper, I study these bonds in a discrete-time, perfect-foresight model used by McCallum (1984). McCallums model is a version of the Sidrauski (1967) continuous-time money-inutility-function model and the Brock (1975) perfect-foresight monetary model, modified by the inclusion of government bonds as an asset. If the economy is Ricardian4 (Barro, 1974), I find its dynamic equilibrium properties to be almost equivalent to one with nominal bonds, in that the time paths of all variables other than the bond quantity are identical. As perfect foresight implies that expectations are formed rationally, this paper formalizes and extends McCallums (1986, p.410) suggestion that indexation has little or no value in a long-run or steadystate equilibrium when investors form rational expectations. Now what about the short-run? With a money-demand externality (inherent in money-in-the-utility-function models), I show how stationary sunspots could affect asset prices (and rates of return) even in a fundamentally deterministic economy. Stationary sunspot equilibria are multiple equilibrium paths around a steady-state where the actual path, undetermined by fundamentals, is determined by nonfundamentals or sunspots. The sunspot is constructed following Spear (1989, 1991) and Lim (1997, 2009) by first showing that the non-stochastic steady-state is indeterminate, and then bootstrapping a stochastic sunspot variable onto the Euler equations determining the evolution of the state variables and prices, thereby generating nonfundamental volatility. The introduction of real return
See Barr and Campbell (1996), Evans (1998), Kopcke and Kimball (1999), Sack and Elsasser (2002), Kothari and Shanken (2004), Roll (2004) and Weiss (2006). 3 One exception is Geanakoplos (2003). Lioui and Poncet (2004) studied real interest rates but without a real bond. 4 Woodford (1994) defines a Ricardian economy as one where the government is restricted to satisfy a budget constraint in all periods with the amount of government bonds approaching zero at infinity. Barro (1974) then demonstrates the invariance of all other variables to bond-financed changes in tax receipts. This offsetting nature of the present value of taxes and bonds was clearly recognized, as a matter of theory, by Ricardo (1817).
2

2
Electronic copy available at: http://ssrn.com/abstract=2167722

bonds neither mitigates sunspot equilibria nor reduces nonfundamental volatility. That is, the occurrence of nonfundamental or excess volatility (or irrational exuberance) does not vanish nor diminish with real return bonds.5 This result, together with the experience of certain countries, should caution investors that the introduction of real return bonds would not, by itself, reduce asset price volatility. For example, excess volatility of asset prices could arise (as suggested above by Noyer, 2004) due to circularity problems, where misperceptions from both sides (central bank and investors) generate asset price instability. In addition, it would be shown that, even in a steady-state, continuous monetization of the (increasing) interest/coupon payments is possible with real return bonds and thus hyper-inflation could still exist. In an option-pricing framework, Jacoby and Shiller (2005) analyzed the value of the option embedded in inflation-indexed bond yields and found that the value of these zero-coupon inflation-indexed bonds may increase or decrease with the nominal interest rate, real interest rate or expected inflation, depending on the monetary policy practiced. The indeterminacy could be resolved if the monetary policy was defined. In this paper, it would be seen that the embedded option does not affect the Ricardian property of the model, and so it is not explicitly modeled. However, indeterminacy still arises in economies with real return bonds without this option. The contribution of this paper is then to show that if investors form expectations rationally, the widespread benefits of indexing nominal debt suggested by Campbell and Shiller (1996) may be more limited or nonexistent. Campbell and Shiller also discuss using the spread between indexed and nonindexed bond yields as a measure of inflation expectations, but point out that such a measure could be contaminated. I formalize the contamination and circularity problems by showing that the steady-state is indeterminate even with real return bonds, which allows for the existence of sunspot equilibria. The paper proceeds as follows: the analysis of an economy with nominal bonds is presented in Section II, followed by the analysis with real return bonds in Section III which shows that both economies dynamic equilibrium are almost equivalent. The existence of sunspot equilibria is shown in Section IV, and an analysis of these sunspot equilibria is given in Section V with an examination of country experiences with real return bonds. Section VI concludes with a discussion as to why there is excess volatility when there is money.

2.

Review of McCallums Model

In order to focus on the effects of introducing real return bonds vis--vis nominal bonds in a dynamic general equilibrium framework, it will be presumed throughout the analysis that the economy can be represented by a deterministic, aggregative, flexible-price, equilibrium model. For some issues, such a model may be inadequate or misleading. My present concern, however, involves the
Although sunspot prices are martingales and reflect all publicly available information, they also reflect extraneous information and are therefore excessively volatile.
5

3
Electronic copy available at: http://ssrn.com/abstract=2167722

issues of excess volatility and equilibrium statistical properties of an economy with real return bonds, as substantial empirical evidence can be found on these issues (e.g., Barr and Campbell, 1996; Kothari and Shanken, 2004). For this type of concern, the flexible-price general equilibrium model of McCallum (1984) with nominal bonds seems well suited. It is described as follows: Consider an economy composed of a large number of infinitely-lived agents, each of which seeks in competitive markets at period t to maximize: U(ct,mt) + U(ct+1,mt+1) + 2U(ct+2,mt+2) + (1)

Here ct is consumption in period t and mt=Mt/Pt, with Mt the agents nominal money stock at the start of period t and Pt the price of the (single composite) consumption good in period t. The within-period utility function U is assumed to be twice-continuously differentiable, strictly concave, monotonic and bounded such that unique, positive values are chosen for ct and mt each period. The discount factor =1/(1+), with the time-preference parameter strictly positive. In order to keep matters as simple as possible, it is assumed that there is no depreciation/population growth. Instead of a production economy, I will differ from McCallum (1984) by assuming strictly positive constant endowments, yt=y, each period. Each agent has the opportunity in period t of purchasing zero-coupon government bonds at a discount money price of Qt. Each bond is redeemed in period t+1 for one unit of money, so the nominal rate of return on bonds between periods t and t+1 is Rt=(1-Qt)/Qt. Lump-sum taxes in the nominal amount Ptvt are collected from the agent in period t. Consequently, the nominal budget constraint for period t is: Ptct + Mt+1 Mt+ QtBt+1 Bt Ptyt - Ptvt (2)

where Bt0 the number of nominal bonds held at the start of period t. The lefthand side of the equation denotes that the agent can spend the after-tax nominal income (on the right-hand side) on the consumption good, investing in more nominal money balances or investing in more nominal bonds. Now the real rate of return on bonds, rt, is defined by the Fisher equation (1+rt)=(1+Rt)/(1+t) where t=(Pt+1-Pt)/Pt is the inflation rate. Dividing equation (2) by the price of the composite consumption good in period t, Pt, noticing that QtBt+1=(1+Rt)1 Bt+1=(1+rt)-1bt+1 (where bt=Bt/Pt each period), and requiring that equation (2) holds with equality each period (from the assumptions on the utility function), we get, like McCallum (1984): ct + (1+t)mt+1 mt+ (1+rt)-1bt+1 bt = yt - vt (3)

Given this setup, the optimality conditions for the agents problem can be found by considering the Lagrangean expression:

L =
t =0

t {U(ct,mt) + t [yt - vt - ct - (1+t)mt+1 + mt - (1+rt)-1bt+1 + bt}]}

with ct, mt+1 and bt+1 as the choice variables. With the time-preference parameter strictly positive, the discount factor =1/(1+) < 1. Because of the assumptions on U which assure that ct and mt+1 will be strictly positive, the first-order Euler conditions associated with ct, mt+1 and bt+1 can be written as equalities holding for all t = 1, 2, . They are, with Uc denoting the partial derivative of U with respect to the consumption good and Um denoting the partial derivative of U with respect to real money balances: Uc(ct,mt) - t = 0 Um(ct+1,mt+1) - t(1+t) + t+1 = 0 bt+1[-t(1+rt)-1+ t+1] = 0 (4) (5) (6)

From equation (4), assumptions on U imply a positive t in each period. In addition to the first-order conditions, we also have the infinite-horizon transversality conditions for mt+1 and bt+1: lim mt+1tt(1+t) = 0 t lim bt+1tt(1+rt)-1 = 0 t (7)

(8)

Equations (3) to (6) are necessary for optimality while equations (3) to (8) are jointly sufficient. Thus if equations (7) and (8) are satisfied, the agents choice of ct, mt+1, bt+1 and t will be given initial stocks of money balances and bonds, time paths of prices rt and t, transfers vt and endowments yt described by equations (3) to (6).6 Next, consider the governments budget constraint. Expressing all quantities in per capita terms and letting gt denote government purchases of the consumption good, we have the identity: Mt+1 Mt + QtBt+1 Bt + Ptvt = Ptgt (9)

Without loss of generality, let gt = 0 and vt = v for all t periods such that in real terms:
6

McCallum (1984) points out that these could be thought of as households, where each generation cares about the utility of its offspring. Barro (1974) has demonstrated that an economy of finite-lived agents who care about the utility of their offspring may, in general, be treated for analytical purposes as one with infinitely-lived agents.

(1+t)mt+1 mt+ (1+rt)-1bt+1 bt + v = 0

(10)

Note that together equations (3) and (10) imply the composite consumption good market equilibrium for each period t such that only two of equations (3), (10) and (11) are needed: ct = yt = y (11)

Finally, assume that the money supply grows each period at a constant rate . That is, Mt+1/Mt = 1+ each period. With the identities mt=Mt/Pt and t=(Pt+1-Pt)/Pt holding for all periods t as before, this implies, for money market equilibrium in each period t: (1+t) (mt+1/mt) = 1 + (12)

Given the government choices of v0 and >0 (policy variables) and constant endowment y and discount factor , equations (4), (5), (6), (10), (11) and (12) will determine equilibrium paths for t, mt, rt, bt, ct and t respectively. From equation (4), t = Uc(ct,mt), t+1 = Uc(ct+1,mt+1) and so forth. Then from equation (11), ct = ct+1 = y for all periods, which implies that t = Uc(y,mt), t+1 = Uc(y,mt+1) and so forth. Substituting these into the other four equations yields: Um(y,mt+1) - Uc(y,mt)(1+t) + Uc(y,mt+1) = 0 bt+1[-Uc(y,mt) (1+rt)-1+ Uc(y,mt+1)] = 0 (1+t)mt+1 mt+ (1+rt)-1bt+1 bt + v = 0 (1+t) (mt+1/mt) = 1 + (5) (6) (10) (12)

Notice from the above system of four equations that v, bt and rt do not appear in equations (5) or (12). Then rewrite equation (12) as: (1+t) = (1+) (mt/mt+1) Substituting for (1+t) in equation (5) yields: Um(y,mt+1) - Uc(y,mt)(1+)(mt/mt+1) + Uc(y,mt+1) = 0 (14) (13)

This shows that the deterministic equilibrium time path of {mt} is solely determined by equation (14) given the governments choice of and initial money stock m0. Equation (13) then determines the non-stochastic evolution of {t}, and equation (6) the non-stochastic evolution of {rt} (for bt+1>0), given the deterministic time path of {mt}. Equation (10) then determines the deterministic equilibrium time path of {bt} given {mt}, {t}, {rt}, an initial bond quantity b0 and the lump-sum taxes v.

2.2. Steady-State Analysis. Using the foregoing model, let us now consider properties of steady states, i.e., dynamic equilibria in which every variable grows at some constant rate. Under present assumptions, with no technical progress nor population growth, this condition requires that mt and bt be constant over time, i.e., have growth rates of zero (McCallum, 1990). The system of four equations (5), (6), (10) and (12) can then be expressed as (with b>0): Um(y,m) - Uc(y,m)(1+-) = 0 -1 + (1+r) = 0 m - r(1+r)-1b + v = 0 = Um(y,m) - Uc(y,m)(1+-) = 0 m - (1-)b + v =0 (15) (16) (17) (18)

From equation (16), r = 1/ - 1, and with = from equation (18), we have: (19) (20)

Proposition 1 (Existence of Steady-State Equilibrium). With 0<<1, y>0, >0 and v0, there exists a steady-state equilibrium in the economy with nominal bonds where c=y>0, =>0, r=1/ 1>0, 0<m<, 0<b< and >0. Proof of Proposition 1. With b = (m + v)/(1-) > 0 (from equation (20)) and = Uc(y,m) > 0 (from the assumption on U), it suffices to show that a steady-state equilibrium exists if a unique m, 0<m<, could be found from equation (19). By the Implicit Function Theorem, a unique value of m could be found if the following condition holds: Umm(y,m) - Ucm(y,m)(1+-) 0 (21)

With >0 and 0<<1 so 1+->0, if Ucm(y,m)0, then equation (21) holds as strict concavity assumptions on U ensure Umm(y,m)<0. Even if equation (21) is degenerate when Ucm(y,m)<0, all we need to do is to perturb the utility function and it will no longer be degenerate, as in Lim (1997). Thus the set of utility functions where equation (21) does not hold is of measure zero. The next section demonstrates that the steady-state equilibrium of McCallums model (as described above) is similar to an economy with real return bonds, with identical conditions except for equations (17). This would lead to as much excess volatility in an economy with real return bonds as one with nominal bonds.

3. Economy with Real Return Bonds


Jacoby and Shiller (2005) explain an indexed bond as one that pays a constant coupon rate that applies to a principal that is fully adjusted to inflation based on a Consumer Price Index (CPI). Most indexed bonds have an inflation-protection scheme that guarantees that the value of the inflation-adjusted principal is never below its original value. In terms of pricing, the inflation-adjustment scheme implies that in nominal terms, a long position in a pure-discount inflationprotected bond is equivalent to an unadjusted (nominal) Treasury bond and a long position in a European call option written on a fully-adjusted (real) purediscount riskless bond. Instead of an inflation-indexed bond, we introduce a real return bond, which provides a real yield to the bondholder with the possibility of a negative nominal bond yield when the price level or CPI falls (i.e., deflation). With the assumption in the previous section that the money supply grows each period at a positive constant rate , in the steady-state, the inflation rate is always positive by equation (18) and the nominal bond yield is never negative. Also, it must be noted that actual episodes of deflation have been rare in historical data. Therefore, although the model does not explicitly account for the embedded option in indexed bonds, it does not result in much loss of generality. In addition, the embedded option in indexed bonds will reveal itself only in the Euler equation governing the time paths of the bond stock and the bond yield. It will be shown that, as in the economy with nominal bonds, the time paths of the money stock and inflation rate are independent of the bond stock and the bond yield; that is, the Ricardian property continues to hold. The excess volatility in the economy results from the indeterminacy of the money stock and inflation rate and does not depend on whether the bonds are nominal, or real, or inflation-indexed with an embedded option. Therefore, the embedded option in indexed bonds will be irrelevant to the issue of excess volatility if the economy is Ricardian. 3.1. The McCallum Model with Real Return Bonds. Consider an economy composed of a large number of infinitely-lived agents, each of which seeks in competitive markets at period t to maximize: U(ct,mt) + U(ct+1,mt+1) + 2U(ct+2,mt+2) + (22)

The notation and assumptions of the model follow from the economy with nominal bonds in the previous section. Here, each agent has the opportunity in period t of purchasing zero-coupon real return bonds at a discount money price of Qt. Each bond is redeemed in period t+1 for one unit of money adjusted for expost or past inflation from period t to t+1 which is t. So the nominal rate of return on bonds between periods t and t+1 is Rt = (1+t-Qt)/Qt = [(1+t)/Qt] -1. Thus Qt = (1+t)/(1+Rt) = (1+rt)-1 from the Fisher equation. That is, the real return bonds are discounted at the real rate. The agents budget constraint for period t can then be written in nominal terms as:

Ptct + Mt+1 Mt+ QtBt+1 Bt(1+t-1) Ptyt - Ptvt

(23)

where Bt0 the number of nominal bonds held at the start of period t that is fully adjusted to ex-post inflation from the previous period (t-1 to t) such that the nominal value of these bonds in period t is Bt(1+t-1). This is equivalent to adjusting the principal to inflation based on the CPI (which measures past inflation) and explicitly models the inflation lag bias in real return bonds. Following our previous analysis, the agents budget constraint can be written in real terms as: ct + (1+t)mt+1 mt+ (1+t)(1+rt)-1bt+1 bt(1+t-1) = yt - vt (24)

Given this setup, the optimality conditions for the agents problem can be found by considering the Lagrangean expression:

= t{U(ct,mt) + t [yt -vt -ct - (1+t)mt+1 +mt - (1+t)(1+rt)-1bt+1 + (1+t-1)bt}]}


t =0

with ct, mt+1 and bt+1 as the choice variables. As in the previous section, the firstorder Euler conditions associated with ct, mt+1 and bt+1 can be written as equalities holding for all t = 1, 2, . They are, with Uc denoting the partial derivative of U with respect to the consumption good and Um denoting the partial derivative of U with respect to real money balances: Uc(ct,mt) - t = 0 Um(ct+1,mt+1) - t(1+t) + t+1 = 0 bt+1[-t(1+t)(1+rt)-1+ t+1(1+t)] = 0 where multiplying both sides of equation (27) by (1+t)-1 yields: bt+1[-t(1+rt)-1+ t+1] = 0 (28) (25) (26) (27)

From equation (25), assumptions on U ensure a positive t each period. In addition to the first-order conditions, we also have the infinite-horizon transversality conditions for mt+1 and bt+1: lim mt+1tt(1+t) = 0 t lim bt+1tt(1+t)(1+rt)-1 = 0 t (29)

(30)

Equations (24) to (26) and (28) are necessary for optimality while equations (24) to (26) and (28) to (30) are jointly sufficient. Thus if equations (29) and (30) are satisfied, the agents choice of ct, mt+1, bt+1 and t will be given initial stocks of money balances and bonds, time paths of prices rt and t, transfers vt and endowments yt described by equations (24) to (26) and (28). Next, consider the governments budget constraint. Expressing all quantities in per capita terms, we have the identity: Mt+1 Mt+ QtBt+1 Bt(1+t-1) + Ptvt = Ptgt (31)

Without loss of generality, let gt = 0 and vt = v for all t periods such that in real terms: (1+t)mt+1 mt+ (1+t)(1+rt)-1bt+1 bt(1+t-1) + v = 0 (32)

Note that together equations (24) and (32) imply the composite consumption good market equilibrium for each period t such that only two of equations (24), (32) and (33) are needed: ct = yt = y (33)

Finally, assume that the money supply grows each period at a constant rate . That is, Mt+1/Mt = 1+ each period. With the identities mt=Mt/Pt and t=(Pt+1-Pt)/Pt holding for all periods t as before, this implies, for money market equilibrium in each period t: (1+t) (mt+1/mt) = 1 + (34)

Given the government choices of v0 and >0 (policy variables) and constant endowment y and discount factor , equations (25), (26), (28), (32), (33) and (34) will determine equilibrium paths for t, mt, rt, bt, ct and t respectively. From equation (25), t = Uc(ct,mt), t+1 = Uc(ct+1,mt+1) and so forth. Then from equation (33), ct = ct+1 = y for all periods, which implies that t = Uc(y,mt), t+1 = Uc(y,mt+1) and so forth. Substituting these into the other four equations yields: Um(y,mt+1) - Uc(y,mt)(1+t) + Uc(y,mt+1) = 0 bt+1[-Uc(y,mt) (1+rt)-1+ Uc(y,mt+1)] = 0 (1+t)mt+1 mt+ (1+t)(1+rt)-1bt+1 bt(1+t-1) + v = 0 (1+t) (mt+1/mt) = 1 + (26) (28) (32) (34)

Notice from the above system of four equations that v, bt and rt do not appear in equations (26) or (34). Then rewrite equation (12) as:

10

(1+t) = (1+) (mt/mt+1) Substituting for (1+t) in equation (5) yields: Um(y,mt+1) - Uc(y,mt)(1+)(mt/mt+1) + Uc(y,mt+1) = 0

(35)

(36)

This shows that the deterministic equilibrium time path of {mt} is solely determined by equation (36) given the governments choice of and initial money stock m0. Equation (35) then determines the non-stochastic evolution of {t}, and equation (28) the non-stochastic evolution of {rt} (for bt+1>0), given the deterministic time path of {mt}. Equation (32) then determines the deterministic equilibrium time path of {bt} given {mt}, {t} (with -1=0) and {rt}, an initial bond quantity b0 and the lump-sum taxes v. Notice from the above system that except for equation (32), all the other conditions for dynamic equilibrium are exactly identical to the corresponding conditions for dynamic equilibrium of an economy with nominal bonds! Equation (36) is identical to equation (14), equation (35) is identical to equation (13) and equation (28) is identical to equation (6). With ct = y for all periods, the dynamic equilibrium properties of an economy with real return bonds are therefore almost equivalent to one with nominal bonds with identical time paths of {mt}, {t}, {rt} and {t} (with t = Uc(y,mt)). Any excess volatility in the economy will result from the indeterminacy of the money stock and inflation rate and will not depend on whether the bonds are nominal, or real, or inflation-indexed with an embedded option. Therefore, the embedded option can be ignored without loss of generality. Next, I show that the steady-state values of all the variables except for the real bond stock are also identical. 3.2. Steady-State Analysis. Let us now consider properties of steady states in the economy with real return bonds, i.e., dynamic equilibria in which every variable grows at some constant rate. Under our assumptions, this condition requires that mt and bt be constant over time, i.e., have growth rates of zero. The system of four equations (26), (28), (32) and (34) can then be expressed as (with b>0): Um(y,m) - Uc(y,m)(1+-) = 0 -1 + (1+r) = 0 m - r(1+r)-1(1+)b + v = 0 = Um(y,m) - Uc(y,m)(1+-) = 0 (37) (38) (39) (40)

From equation (38), r = 1/ - 1, and with = from equation (40), we have: (41)

11

m - (1-)(1+)b + v = 0

(42)

Proposition 2 (Existence of Steady-State Equilibrium with Nominal Bonds). With 0<<1, y>0, >0 and v0, there exists a steady-state equilibrium in the economy with nominal bonds where c=y>0, =>0, r=1/ 1>0, 0<m<, 0<b< and >0. Proof of Proposition 2. With b = (m + v)/(1-)(1+) > 0 (from equation (42)) and = Uc(y,m) > 0 (from the assumption on U), it suffices to show that a steady-state equilibrium exists if a unique m, 0<m<, could be found from equation (41). By the Implicit Function Theorem, a unique value of m could be found if the following condition holds: Umm(y,m) - Ucm(y,m)(1+-) 0 (43)

With >0 and 0<<1 so 1+->0, if Ucm(y,m)0, then equation (43) holds as strict concavity assumptions on U ensure Umm(y,m)<0. Even if equation (43) is degenerate when Ucm(y,m)<0, all we need to do is to perturb the utility function and it will no longer be degenerate, as in Lim (1997). Thus the set of utility functions where equation (43) does not hold is of measure zero. Notice that as equation (41) is identical to equation (19), the steady-state real money balance in an economy with real return bonds is identical to the steadystate real money balance in an economy with nominal bonds. From equation (38), r = 1/ - 1, and from equation (40), = , which are the same steady-state values for r and in an economy with nominal bonds. With c = y and = Uc(y,m), this means that the only steady-state value different in an economy with real return bonds is the steady-state value of b = (m + v)/(1-)(1+), which is less than the steady-state value of b in an economy with nominal bonds by a factor of 1/(1+). The real value of the bond stock has to be greater if the bonds were nominal, as the real value of nominal bonds decreases with inflation. With real return bonds, however, the real value remains constant and thus a smaller stock of real return bonds is required in the steady-state. Finally, it is evident from equation (42) that in the steady-state, with v=0, continuous monetization of the (increasing) interest/coupon payments is possible even with real return bonds.

4.

Existence of Sunspot Equilibria

It has been established that the equations governing the determinate equilibrium time paths of {mt}, {t}, {rt} and {t} are identical in both economies (McCallums model with nominal and the last sections model with real return bonds). It has also been established that, for both economies, the equilibrium time path of {mt} is solely determined by one equation where {mt} is the only unknown variable, and this equation is identical in both economies.7 With the determinate solution of
7

See equations (14) and (36).

12

{mt} found from this equation, the solutions for {t}, {t} and {rt} could then be found from three other equations (identical in both economies8), where given {mt}, these three other equations have only one unknown variable, and the solution of each equation for that one unknown yields determinate equilibrium time paths for {t}, {t} and {rt}. Despite the fact that both economies are nonstochastic, the following proposition shows that there exists a nonempty, open set of economies where it is possible to construct an equilibrium time path around the steady-state value of m such that mt = m + t, mt+1 = m + t+1, for sufficiently small {t}. This means that the equilibrium time path of m is locally nonunique or indeterminate. Denote Ucm as the partial derivative of Uc with respect to m and Umm as the partial derivative of Um with respect to m. Proposition 3 (Indeterminacy of Equilibria in McCallums Model with Nominal Bonds). There exists a function mt+1 = g(mt) which satisfies equation (14). If: (a) (1+-)Ucm < Umm < 0 AND (1++)Ucm + Umm + 2(1+)Uc/m > 0; or (b) Umm < (1+-)Ucm AND (1++)Ucm + Umm + 2(1+)Uc/m < 0, then there exist neighborhoods of m and of zero such that for all {t} , the mapping g is a contraction on and the steady-state equilibrium of Proposition 1 is indeterminate. 9 Proof of Proposition 3. See Appendix 1. Proposition 4 (Indeterminacy of Equilibria with Real Return Bonds). There exists a function mt+1 = g(mt) which satisfies equation (36). If: (a) (1+-)Ucm < Umm < 0 AND (1++)Ucm + Umm + 2(1+)Uc/m > 0; or (b) Umm < (1+-)Ucm AND (1++)Ucm + Umm + 2(1+)Uc/m < 0, then there exist neighborhoods of m and of zero such that for all {t} , the mapping g is a contraction on and the steady-state equilibrium of Proposition 2 is indeterminate. Proof of Proposition 4. See Appendix 2. Notice that the conditions for indeterminacy are exactly the same for nominal bonds and real return bonds. The introduction of real return bonds does not reduce the set of indeterminate economies. Following the bootstrapping technique of Spear, Srivastava and Woodford (1990) and Lim (1997), the indeterminacy of equilibria allows one to construct a large family of equilibria by replacing identical equations (14) and (36) with: Um(y,mt+1) - Uc(y,mt)(1+)(mt/mt+1) + Uc(y,mt+1) - t = 0
8

(44)

See equations (4) and (25) for {t}, equations (13) and (35) for {t}, equations (6) and (28) for {rt}. 9 The difference between a unique steady-state value and a unique equilibrium time path is worth emphasizing. The indeterminacy leading to sunspot equilibria is due to multiple equilibria where coordination failures or misperceptions could result in a time path that appears (nonfundamentally) stochastic.

13

where the sunspot variable t is drawn from a stationary, i.i.d. distribution such that d() = 0. Notice that the identical equilibrium conditions (14) and (36) have the form G(mt,mt+1,t) = 0. Since {t} is exogenous and thus independent of {mt}, the Implicit Function Theorem allows me to write mt+1 = g(mt,t) if and only if G(mt,mt+1,t)/mt+1 0 which must be true for Propositions 3 and 4 to hold. mt+1 = g(mt,t) is then the nonfundamental stochastic equilibrium path of {mt}. Proposition 5 (Existence of Sunspot Equilibria). There exists a nonempty, open set of economies with nominal or real return bonds exhibiting nontrivial stationary sunspot equilibria. Proof of Proposition 5. See Appendix 3. With {mt} stochastic, it must be true that {t}, {t} and {rt} are stochastic as their equilibrium time paths are derived from {mt}.10 Then with {mt}, {t}, {t} and {rt} stochastic, {bt} must be stochastic as well, in both economies with nominal or real return bonds. Hence although both economies are fundamentally deterministic, they would exhibit extrinsic volatility. That is, even with real return bonds, sunspot equilibria exist. The occurrence of nonfundamental or excess volatility (or irrational exuberance) does not vanish nor diminish with real return bonds.

5. Analysis of Excess Volatility and Empirical Evidence


To analyze the excess volatility, we first need to ask what do the indeterminacy conditions (leading to sunspot equilibria) in Propositions 3 and 4 mean? Condition (a) requires: (1+-)Ucm < Umm < 0 AND (1++)Ucm + Umm + 2(1+)Uc/m > 0 This condition will hold only if Ucm<0, and Uc>0 sufficiently large such that the second part of the condition holds. Now Uc being sufficiently large means that consumption C is relatively low and Ucm<0 means that consumption and real money balances are somewhat substitutable. It is likely that this condition may be satisfied in less wealthy countries. Next, condition (b) requires: Umm < (1+-)Ucm AND (1++)Ucm + Umm + 2(1+)Uc/m < 0 This condition will hold if Umm<0 is sufficiently large to offset the possibly positive values of the other variables. This means that agents are quickly saturated with real money balances m as their positive marginal utility falls by much with just a small increase in m. It is likely that this condition may again be satisfied in less wealthy economies where there are fewer goods for money to facilitate
10

As mt=Mt/Pt and Mt+1=(1+)Mt=(1+) M0, nonfundamental movements in mt reflect nonfundamental movements in Pt and our setup is equivalent to one in which agents forecast prices and the price forecast function is bootstrapped with a sunspot variable.
t+1

14

transactions.11 An example of such an economy may be Brazil. Kopcke and Kimball (1999) report that Brazil issued indexed bonds in 1964. Despite the introduction of indexed bonds, price volatility in Brazil from 1978 to 1996 was measured at 900 percent! Furthermore, during the 1970s, the Brazilian governments ability to hedge the risk of indexed bonds diminished as the rising price of imports, especially oil, and a faltering economy caused the value of its indexed debt to increase more rapidly than its tax revenues. As a result, it altered its formula for indexing the principal of its bonds, allowing only fractional indexation and by 1979, the rate of inflation exceeded the appreciation of principal by 30 percent! Other less wealthy economies with indexed bonds and high price volatility from 1978 to 1996 include Mexico (40 percent volatility), Turkey (25 percent volatility) and Israel (300 percent volatility). These experiences should caution investors that the introduction of real return bonds would not, by itself, ensure price stability or reduce asset price volatility. If these investors are risk-averse, their welfare would also be reduced with sunspot equilibria by Jensens inequality. That is, with real money balances directly affecting utility, sunspot equilibria allocations would not be Pareto-optimal. However, since the sunspots constructed are stationary, inflation defined as a steady and sustained rise in the general price level is still, on average, a monetary phenomenon. Empirical studies which rule out bubble inflation do not rule out the existence of these sunspots.

6. Conclusion
Lim (2001) suggests that the sunspot equilibria arise from money demand externalities. Tobin (1978, p.85) describes the external effect of others money demand as follows: Its (moneys) value to me today depends on its value to you tomorrow, which depends on its value to someone else the next day, and so on into the endless future. As noted by Cass and Shell (1989), an externality may result in multiple equilibria where the eventual outcome is determined by extrinsic or nonfundamental beliefs (like sunspots) due to a failure of the conditions for the Arrow-Debreu classical hypothesis to hold. This externality was modeled in this paper by including money prices as arguments in the utility function (which appear as the denominator of real money balances) since money prices are determined, in part, by others money demand (Samuelson, 1947). For certain economies (defined by preferences), this externality resulted in sunspot equilibria which reduced welfare. In short, money entails an externality because it requires general acceptance for its usefulness. That is, the value of (intrinsically useless paper) money that I carry depends on your demand for it. Real return bonds do not mitigate this externality, as shown by the experiences of several countries.
See Lim (1997) for a detailed analysis of cross-derivatives of a utility function. Using preferences to determine whether an economy would suffer from excess volatility is similar to using genetic traits to determine the risk of certain populations to certain diseases (e.g., Telatar, et al., 1998). However, due to concerns about singling out ethnic groups in disease disposition, there is reluctance to use genetic traits as a factor.
11

15

These countries are typically less wealthy, consistent with theoretical predictions, and their central banks have difficulty maintaining credibility in order to maintain demand for their currencies. Finally, this paper suggests that while austerity measures imposed on some countries in the Euro zone may be painful, perhaps such measures helped to reduce excess volatility by reducing the likelihood of hyperinflation through pegging the value of currency. Also, perhaps a criterion for bond investors could be whether or not a countrys currency is pegged to a stronger currency or a pure investment commodity, as such a country would have a more stable currency and less volatility.

7.

References

Barr, D.G., Campbell, J.Y., 1996. Inflation, real interest rates and the bond market: a study of UK nominal and index-linked government bond prices. Working Paper 5821, National Bureau of Economic Research, Cambridge, Massachusetts. Barro, R.J., 1974. Are government bonds net wealth? Journal of Political Economy 82, 1095-1117. Brock, W.A., 1975. A simple perfect foresight monetary model. Journal of Monetary Economics 1, 133-150. Campbell, J., Shiller, R., 1996. A scorecard for indexed government debt. NBER Macroeconomics Annual, eds. Bernanke, B., Rotemberg, J., Cambridge, MA: MIT Press. Cass, D., Shell, K., 1989. Sunspot equilibrium in an overlapping generations economy with an idealized contingent-commodities market. Economic Complexity: Chaos, Sunspots, Bubbles and Nonlinearity, eds. Barnett, W.A., Geweke, J., Shell, K., London: Cambridge University Press. Evans, M.D.D., 1998. Real rates, expected inflation, and inflation risk premia. Journal of Finance 53 (1), 187-218. Geanakoplos, J., 2003. The ideal inflation indexed bond and Irving Fishers impatience theory of interest in an overlapping generations world. Cowles Foundation Discussion Paper No. 1429, Cowles Foundation for Research in Economics, Yale University, New Haven, Connecticut. Jacoby, G., Shiller, I., 2005. Understanding inflation-protected bonds. Working Paper, I.H. Asper School of Business, The University of Manitoba, Canada. Kopcke, R.W., Kimball, R.C., 1999. Inflation-indexed bonds: The dog that didnt bark. New England Economic Review, Federal Reserve Bank of Boston, 3-24. Kothari, S.P., Shanken, J., 2004. Asset allocation with inflation-protected bonds. Financial Analysts Journal 60 (1), 54-70. Lim, W., 1997. Observing sunspots at home. Journal of Housing Economics 6, 203-222. Lim, W., 2001. A note on local determinacy with inter-temporal demand for money and goods. Working Paper #2001-007, Faculty of Administration, University of New Brunswick, Federicton, New Brunswick.

16

Lim, W., 2009. Contagion And Sunspots Surrounding Speculative Home Equity Lending Losses. Conference Paper, Financial Management Association 2009 International Meeting, Reno, NV, USA Lioui, A., Poncet, P., 2004. General equilibrium real and nominal interest rates. Journal of Banking and Finance 28, 1569-1595. McCallum, B.T., 1984. Are bond-financed deficits inflationary? A Ricardian analysis. Journal of Political Economy 92 (1), 123-135. McCallum, B.T., 1986. On Real and Sticky-Price Theories of the Business Cycle. Journal of Money, Credit and Banking 18(4), 397-414. McCallum, B.T., 1990. Inflation: theory and evidence. Handbook of Monetary Economics, Vol. II, eds. Friedman, B.M., Hahn, F.H., Amsterdam: Elsevier Science Publishers B.V., 963-1012. Noyer, C., 2004. The role of inflation-indexed bonds in the process of setting monetary policy: a central bankers perspective. Speech by Mr. Christian Noyer, Governor of the Bank of France, at the Morgan Stanley seminar on indexed bonds, Paris, 9 June 2004. BIS Review 37, 1-6. Ricardo, D., 1817. Principles of Political Economy and Taxation, London: Murray. Roll, R., 2004. Empirical TIPS. Financial Analysts Journal 60 (1), 31-53. Sack, B., Elsasser, R., 2002. Treasury inflation-indexed debt: a review of the U.S. experience. Working Paper, Federal Reserve Board, Washington, D.C.. Samuelson, P.A., 1947. Foundations of Economic Analysis, Cambridge: Harvard University Press. Shiller, R.J., 2003. The invention of inflation-indexed bonds in early America. Cowles Foundation Discussion Paper No. 1442, Cowles Foundation for Research in Economics, Yale University, New Haven Connecticut. Sidrauski, M., 1967. Rational choice and patterns of growth in a monetary economy. American Economic Review Papers and Proceedings 57, 534-544. Spear, S., 1989. Are sunspots necessary? Journal of Political Economy 97, 965973. Spear, S., Srivastava, S., Woodford, M., 1990. Indeterminacy of stationary equilibrium in stochastic overlapping generations models. Journal of Economic Theory 50, 265-284. Spear, S., 1991. Growth, externalities and sunspots. Journal of Economic Theory 54, 215-223. Telatar, M., Teraoka, S., Wang, Z., Chun, H.H., Liang, T., Castellvi-Bel, S., Udar, N., Borresen-Dale, A.-L., Chessa, L., Bernatowska-Matuszkiewicz, E., Porras, O., Watanabe, M., Junker, A., Concannon, P., Gatti, R.A., 1998. AtaxiaTelangiectasia: Identification and detection of founder-effect mutations in the ATM gene in ethnic populations. American Journal of Human Genetics, 62, 8697. Tobin, J., 1978. Discussion by. Models of Monetary Economics, eds. Karaken, J., Wallace, N., Minneapolis, Minnesota: Federal Reserve Bank of Minneapolis. Weiss, L., 2006. Inflation indexed bonds and monetary theory. Economic Theory 27, 271-275. Woodford, M., 1994. Monetary policy and price level determinacy in a cash-inadvance economy. Economic Theory 4, 345-380.

17

Appendix 1 Proposition 3 (Indeterminacy of Equilibria in McCallums Model with Nominal Bonds). There exists a function mt+1 = g(mt) which satisfies equation (14). If: (a) (1+-)Ucm < Umm < 0 AND (1++)Ucm + Umm + 2(1+)Uc/m > 0; or (b) Umm < (1+-)Ucm AND (1++)Ucm + Umm + 2(1+)Uc/m < 0, then there exist neighborhoods of m and of zero such that for all {t} , the mapping g is a contraction on and the steady-state equilibrium of Proposition 1 is indeterminate. Proof of Proposition 3. We will show that under (a) or (b), for t=0, the steadystate equilibrium of Proposition 1 is a contraction at its fixed point m. By continuity, it will then follow that for sufficiently small neighborhoods of m and of zero, if {t} , then the steady-state equilibrium is a contraction on . From equation (14), let G = Um(y,mt+1) - Uc(y,mt)(1+)(mt/mt+1) + Uc(y,mt+1). Let Gmt+1 denote the partial derivative of G with respect to mt+1 at the steady-state and Gmt the partial derivative of G with respect to mt at the steady-state. By the Implicit Function Theorem, we can write the evolution of {mt} as mt+1 = g(mt) if and only if Gmt+1 0, where: Gmt+1 = Umm(y,mt+1) + Uc(y,mt)(1+)(mt/mt+1)(1/mt+1) + Ucm(y,mt+1) (45) As argued in Proposition 1 and Lim (1997), even if equation (45) is degenerate, all we need to do is to perturb the utility functions and equation (45) will no longer be degenerate. To show that g is a contraction, we need to show that g'(mt)<1 at the steady-state m. That is, by the Implicit Function Theorem, to show that the steady-state equilibrium is a contraction, it suffices to show that the following condition (46) holds at the steady-state, where

g (m )= dm = Gm < 1 dm Gm
t +1 t t t t +1

(46)

Now at the steady-state value of m:


Gmt +1
t

(1+ )U ( y,m) + (1+ )U ( y,m) / m = U ( y,m) + U ( y,m) + (1+ )U ( y,m) / m Gm


cm c mm cm c

(47)

Notice from (47) that if Umm is sufficiently large in absolute value, the denominator is strictly greater than the numerator and (47) holds regardless of the sign of Ucm. To derive more exact sufficient conditions such that (47) holds, we work through the inequalities and these sufficient conditions are: (a) (1+-)Ucm < Umm < 0 AND (1++)Ucm + Umm + 2(1+)Uc/m > 0; or (b) Umm < (1+-)Ucm AND (1++)Ucm + Umm + 2(1+)Uc/m < 0.

18

Appendix 2 Proposition 4 (Indeterminacy of Equilibria with Real Return Bonds). There exists a function mt+1 = g(mt) which satisfies equation (36). If: (a) (1+-)Ucm < Umm < 0 AND (1++)Ucm + Umm + 2(1+)Uc/m > 0; or (b) Umm < (1+-)Ucm AND (1++)Ucm + Umm + 2(1+)Uc/m < 0, then there exist neighborhoods of m and of zero such that for all {t} , the mapping g is a contraction on and the steady-state equilibrium of Proposition 2 is indeterminate. Proof of Proposition 4. We will show that under (a) or (b), for t=0, the steadystate equilibrium of Proposition 2 is a contraction at its fixed point m. By continuity, it will then follow that for sufficiently small neighborhoods of m and of zero, if {t} , then the steady-state equilibrium is a contraction on . From equation (36), let G = Um(y,mt+1) - Uc(y,mt)(1+)(mt/mt+1) + Uc(y,mt+1). Let Gmt+1 denote the partial derivative of G with respect to mt+1 at the steady-state and Gmt the partial derivative of G with respect to mt at the steady-state. By the Implicit Function Theorem, we can write the evolution of {mt} as mt+1 = g(mt) if and only if Gmt+1 0, where: Gmt+1 = Umm(y,mt+1) + Uc(y,mt)(1+)(mt/mt+1)(1/mt+1) + Ucm(y,mt+1) (48) As argued in Proposition 2 and Lim (1997), even if equation (48) is degenerate, all we need to do is to perturb the utility functions and equation (48) will no longer be degenerate. To show that g is a contraction, we need to show that g'(mt)<1 at the steady-state m. That is, by the Implicit Function Theorem, to show that the steady-state equilibrium is a contraction, it suffices to show that the following condition (49) holds at the steady-state, where

g (m )= dm = Gm < 1 dm Gm
t +1 t t t t +1

(49)

Now at the steady-state value of m:


Gmt +1 =

Gm

(1+ )U ( y,m) +(1+ )U ( y,m) / m U ( y,m)+ U ( y,m)+(1+ )U ( y,m) / m


cm c mm cm c

(50)

Notice from (50) that if Umm is sufficiently large in absolute value, the denominator is strictly greater than the numerator and (50) holds regardless of the sign of Ucm. To derive more exact sufficient conditions such that (50) holds, we work through the inequalities and these sufficient conditions are: (a) (1+-)Ucm < Umm < 0 AND (1++)Ucm + Umm + 2(1+)Uc/m > 0; or (b) Umm < (1+-)Ucm AND (1++)Ucm + Umm + 2(1+)Uc/m < 0.

19

Appendix 3 Proposition 5 (Existence of Sunspot Equilibria). There exists a nonempty, open set of economies with nominal or real return bonds exhibiting nontrivial stationary sunspot equilibria. Proof of Proposition 5. With Propositions 1, 2, 3 and 4 and the above discussion (in Section 4), it remains to show that there exists an invariant measure for the equilibrium time path or evolution function mt+1 = g(mt,t). For any continuous function h(mt+1,t), define for t , t i.i.d., (Ph)(mt,t-1) = R h[g(mt-1, t-1),t] d(t) (51)

Since g is a continuous function of its arguments, by the Implicit Function Theorem, the transition operator P plainly takes continuous functions into continuous functions. Hence by Rosenblatts theorem, there exists an invariant distribution for the equilibrium time path or evolution of {mt} based on expectations of (mt,t-1). This distribution, together with the evolution function g(mt,t), constitutes a stationary rational expectations equilibrium. Since the random variable, or sunspot variable t, is nondegenerate, the equilibrium is stochastically nontrivial.

20

You might also like