Fiscal inconsistencies and high inflation

Journal of Development Economics 43 (1994) 85-104. North-Holland

Fiscal inconsistencies and high inflation*

Daniel Heymann

CEPAL, Buenos Aires, Argentina and lnstituto Torcuato Di

Pablo Sanguinetti

Tella, Buenos Aires, Argentina

Universidad Torcuato Di Tella, Buenos Aires, Argentina

Received May 1992, final version received November 1992

This paper presents a simple model of steady-state inflationary financing. The discrepancy between the target level of spending and a (fixed) taxing capacity indicates the degree of fiscal pressures; this variable need not be monotonically related to the measured deficit, since actual expenditures are endogenous. Large liscal inconsistencies generate unique steady states located on the ‘bad’ side of the inflation-tax Laffer curve. There, high inflation may coexist with low observed deficits, and the revenue constraint results in ‘repressed spending’. Such situations correspond qualitatively with the loss of control over fiscal policies typical of high inflation economies.

Key words: High inflation; Fiscal pressures

JEL classification: E31; E61

1. Introduction

High inflation creates large economic and social costs. People who live in unstable economies are vividly aware of those costs. Consequently, governments do not seem to lack motivation to stabilize. Rather, getting rid of inflation is often one of their main concerns. But, although proper hyperinflations are short-lived, the record shows a number of cases where countries have operated under high inflation for sizeable periods of time. This evidence raises the question of whether, after all, stability is so much desired.

There are several, not altogether exclusive, ways to address this puzzle. One line of analysis has tried to identify features of the price- and wage

Correspondence to: Daniel Heymann, CEPAL, Av. Corrientes 2554, 1046 Buenos Aires, Argentina.

*Previous drafts of this paper were presented in seminars at the Instituto Di Tella, the Central Bank of Argentina, the Central Bank of Uruguay, UCLA and the Federal Reserve. We also thank comments received from Aquiles Almansi, Axe1 Leijonhufvud, Sebastian Edwards, Iavier Ortiz, Carlos Rivas, Federico Sturzenegger and Mariano Tomassi. The authors are responsible for the errors and opinions expressed.

0304.3878/94/$07.00 0 1994Elsevier Science B.V. All rights reserved

SSDI 0304-3878(93)EOO38-Y

86 D. Heymann and P. Sanguinetti, Fiscal inconsistencies and high inflation

setting process which can propagate inflationary impulses and perpetuate inflation [cf., for example, Frenkel and Fanelli (1989)]. Another body of literature has focused on the dynamics of expectations in a monetary-fiscal framework, and studied conditions under which an economy may get into a high inflation ‘trap’ even though, with a different behavior of expectations, the observed budget deficit could be financed at lower rates of price growth.’ The game-theoretic literature on policy-making, on its side, has stressed the possibility of suboptimal equilibria in the interaction between governments and the private sector, arising, in particular, from the incentives that the authorities may perceive to extract resources from the public through inflationary ‘surprises’.’

One of the most characteristic images of high inflation is that of governments unable to cope with the demands for spending within the constraints set by their ability to collect regular taxes. In the limit, public finances seem ‘out of control’ [cf. Dornbusch et al. (1991), Leijonhufvud (1990)]. Steady-state arguments can hardly describe those conditions. How- ever, some insight can still be gained even while restricting the analysis to stationary outcomes.

In this paper, we discuss inflationary taxation through a simple policy- game model. We consider the decision problem of a government which has a ‘target’ level of expenditures exceeding the revenue-raising capacity of the tax system, and must finance its deficit by issuing money. We treat the taxing capacity and the demands for spending as exogenous (although, of course, they are themselves the result of a political game); the discrepancy between them defines the ‘pressures’ on fiscal policies. But the actions that the government actually undertakes are determined endogenously: what is given from the outside is the size of those pressures, not the observed budget deficit.

One of the results of the analysis is that both mild and very strong fiscal inconsistencies are associated with unique steady states. Thus, according to the model, self-generated high inflations would not be feasible in economies with more or less solid fiscal policies. And very high inflations would not correspond to expectations-induced ‘traps’, but to an underlying fiscal weakness.

However, the inflation rate is not necessarily a monotonically increasing function of the measured deficit. A government that would want to spend much above the attainable tax yield may be unable to obtain much revenue from the inflation tax, due to the low money demand. In those states, the fiscal situation can be described as being ‘out of hand’: the government faces a tight revenue constraint, real public spending and the actual deficit can be

‘Cf. Bruno and Fischer (1990), Bruno (1989), Kiguel (1989), Dornbusch and Fischer (1986), Evans and Yarrow (1981).

‘See, for example, Barro (1983, 19X6), Bruno (1991), Calvo (1978, 1988), Cukierman (1988), Kiguel and Liviatan (1989). A recent survey of the ‘policy-game’ literature can be found in Persson and Tabellini (1990).

D. Heymann and P. Sanguinetti, Fiscal inconsistencies and high inflation 87

low, but there is a large ‘hidden deficit’ which is not accommodated simply because the authorities have no ways to finance it.

The argument suggests that the pressures on the budget describe better than the observed deficit the ‘fundamental’ fiscal conditions leading to high rates of the inflation tax. From a policy perspective, the model has little to say about the initial transition out of high inflation, but it offers some clues about the pre-requisites of a durable stabilization. In this respect, a fiscal correction is not seen as equivalent to a drop in the measured deficit. Rather, stabilization would entail that the government attacks the ‘basic’ inflationary forces deriving from the latent demands for spending and the nature of the tax system. This, in turn, may need the creation of new fiscal institutions or the strengthening of the existing ones,

The paper is organized as follows. In the next section we develop the model; the results are complemented by numerical simulations commented briefly in section 3. Section 4 includes an extension of the analysis which allows for the existence of public debt. Summary and conclusions are presented in section 5.

2. A simple model

2.1. The husic ji-amework

Our starting point is a very stylized description of a government which has to deal with the demands for public goods and the pressures of various groups that lobby for expenditures in their favor. These determine a ‘target’ or ‘desired’ level of real expenditure, g*; for the purpose of the present discussion we treat g* as given. 3 If actual spending, g, deviates from g* we assume that the government bears a cost, which may be loosely associated with the ‘voice’ of disappointed groups and/or the public’s dissatisfaction with the supply of public services. We also postulate that, as it is observed in practice, inflation hurts the government in place. Hence, the period t preferences of the government are represented by a simple quadratic loss function4

‘The literature provides examples of distributive games behind the determination of g*. See, for example, Alesina and Tabellini (1990) Persson and Svensson (1989) Cukierman et al. (1992) Heymann et al. (1991) and Sanguinetti (1992). One of the main points of that literature is that a variable like g* institutions.

may be strongly influenced by the nature and the performance of fiscal

4We have chosen to represent government preferences as a function of the ‘expenditure gap’, and not just the level of public spending [as in Barre (1983) and Cukierman (1988)]. The choice is quite critical for the results that follow; we believe that postulating a (politically determined) level of desired spending is more descriptive of actual fiscal decisions than the alternative assumptions that governments either seek to maximize spending or to optimize the consumption stream of a ‘representative’ agent as if they were isolated from special interest pressures. The variable g* is not directly observable; nevertheless, various types of indicators may possibly be used in order to guess its approximate value.

88 D. Heymann and P. Sanguinetti, Fiscal inconsistencies and high infation

(1)

where g, is realized spending and pt the inflation rate in t. Since the exact nature of the ‘political costs’ behind L, is left undefined, the parameter CI does not have a precise interpretation; however, for a given supply of ‘basic’ government services, it can be taken as a measure of the ‘relative strength’ of the general public, who bears the cost of inflation, vis-a-vis special interest groups who pressure for transfers and/or expenditures in their favor. The intertemporal objective function of the government takes the usual form?

The authorities can draw revenues from regular taxes and from inflation.6 Tax receipts are subject to a fiscal-lag effect: this corresponds to what is commonly observed in high inflations; it also allows us to study the likely consequences of policy changes such as the indexation of taxes. We assume that there is a fixed maximum ‘capacity’ for collecting regular taxes in real terms.7 Hence, tax revenues are given by the following function:

h, h(p,) = ho + lx:

where ho and h, are the parameters describing the tax system: ho+ h, describes the maximum revenues that the tax system is capable of collecting in real terms at a zero inflation rate, while h, measures the exposure of the system to the Olivera-Tanzi effect. Seigniorage revenues in period t are defined as

S(P,i PP+ 1) P: - )-mt(P:,l)-m,~l(P,e)+m,,(P3i~, f

where m,(p;+,) is a well-behaved money demand function which depends on expected inflation and is such that m’( .)<O, m(p’)-+O when pe-+co, and

-Since we are interested in non-reputational equilibria, the distinction between the cases of finite and infinite decision horizons is not crucial.

‘An extension incorporating debt is sketched in section 4. The model outlined here can accommodate the existence of ‘old debts’; demands from creditors for the servicing of these debts would then be reflected in g*. The analysis of section 4 treats separately the costs of not servicing debts.

‘An alternative assumption, common in the literature, would be that taxes are distortionary, and that there are increasing marginal costs of taxation. We have opted for the present formulation because of its simplicity, and also because we want to concentrate on ‘repressed spending’, which seems to be a standard feature of very high inflations.


p’m(p’) is single peaked.* Combining (3) and (4), the government budget constraint is given by

g, = m,(~:+ 1) -m,- lW) +m,- l(d) 1 +pI l+h,+--- h, 1 +Pt’

(5)

Replacing (5) in (1) and using (3) we can rewrite the problem of the government as

min 2 b Pl t=0 C(

$ m,(PF+l)-m,~,(pr)+m,-,(pr)i~ f

hl +h,+---- 1 +pt -g*

2 1

) 1 +jap? (6)

We will make the usual assumption that the inflation rate, pI, is a choice variable for the government: this is clearly a strong hypothesis, meant for the purpose of simplifying the analysis.

2.2. The one-shot Nash solution

A key issue involved in the solution of the above optimization problem is whether the current value of the inflation rate, pa influences the outcome in future periods, and not only the current loss function I,,. This is equivalent to the question of whether today’s choice of the inflation rate has any effect on current and future values of expected inflation. We assume that the public’s expectations are ‘forward-looking’ and we rule out the possibility that individuals and the government are engaged in a ‘reputation game’.’ On the other hand, the government is supposed to be unable to make binding commitments about either the deticit or the rate of money growth. Thus, the policy-maker would solve the above maximization problem taking current and future expectations as given. Hence, it is clear from (6) that the value of pt has no effect on future losses and that, as a consequence, the solution to the intertemporal problem is equivalent to the solution of a

8We follow usual practice in assuming that the ‘inflation-tax revenue curve’ has a maximum and only one. Admissible money demand functions, however, can be consistent with a variety of different shapes of that curve.

‘Trigger strategies in infinite-horizon games of this sort give rise to the well-known problems of multiple equilibria, and it is not clear how atomistic individuals would coordinate their expectations so as to converge on a strategy that would ‘discipline’ the government’s choices [Rogoff (1989)]. In addition, the single-period Nash equilibrium (if unique) is known to be the single equilibrium of a repeated finite-horizon game. It may be noted that backward-looking expectations also make the decision periods not self-contained: future levels of money demand now depend explicitly on current actions by the government.

90 D. Heymann and P. Sanguinetti, Fiscal inconsistencies and high injlation

sequence of one-period problems where the government chooses pt such that current-period losses L, are minimized. The corresponding first-order condition is given by

(g _g*)(mt-lw-hl)+rp =o, I

(1 +Pt)* f

The above expression implicitly defines the government’s reaction function p,(p:). Imposing the perfect foresight condition pP=~~r’ in (7) we obtain the Nash condition that describes the path for the inflation rate in this discretionary equilibrium:

i

Pt MPt + 1) -4 - l(PJ + m, l(PJ l+p,

+h,+A- -g* l+Pt I (mt-l(Pt)-h,)+ap =o

(l+p,Y f * (7’)

For a fixed g*, the problem admits a steady-state solution in which

P~+I =P~=P and

m(P) __ p +h,+- h, l+P l+P

(m(pFd+ccp=O (1 +P)* .

It is easy to see from (8) that an interior solution for which O<p* < co requires both that g<g* and m(p*) > h,. The reason for the first inequality is straightforward: given that inflation is costly, minimization of total losses implies that, at the margin, the cost due to inflation should be equalized with the loss suffered from not being able to reach the target level for public expenditures. The second inequality can be readily interpreted by noting that

& _m(p*)-h 8P pe (1 +p*)2 .

Hence m(p*) > h, implies that, for given expectations, the total revenue curve will have a positive slope. Intuitively, the equilibrium inflation rate will always be smaller than p, the value for which m(p) = h,, and where the revenues from the inflation tax are exactly compensated by the income loss produced by the fiscal lag effect. Thus, the fact that p*<p opens up the

“We do not interpret the perfect foresight hypothesis in a literal sense: it is meant here to depict a steady state situation in which the public has roughly come to understand the properties of the policy regime.


A- 1--7--r----r

“p lnflatlon (p) P

lH(p)--R(P) 1

Fig. 1

possibility that, given private sector expectations, the government would try to produce surprise inflation.

We can rewrite (8) in the following way:

or alternatively,

where the above condition implicitly defines a relationship between the inflation rate and a ‘target’ deficit d*, which measures the discrepancy between desired spending and the maximum (at zero inflation) yield of standard taxes. Now, the expression R(p) = (m(p) - h,)p/( 1 + p) measures inflationary tax revenues net of the losses produced by the fiscal lag effect; given the assumptions about the demand for money, it is a single-peaked function which has a maximum at p = 6 and reaches zero at p =0 and p =@.‘I The second term in the right-hand side of (9), H(p), is an increasing, convex function, which tends to infinity as p tends to p.

Fig. 1 shows the graph of each term separately; fig. 2a depicts the determination of the equilibrium level of inflation resulting from the intersec- tion of the curve G(p)=R(p)+H(p) with the level of fiscal pressures d* [eq.

“Due to the fiscal-lag effect, the ‘net inflation revenue’ reaches a maximum at a value for p which is lower than that which maximizes the inflation tax.


2a. Alpha large

lnflatlon (p) P’ E

2b. Alpha small

P’

lnflatlon (p) P' b

Fig. 2

(9)]. In the case shown in this figure, the rate of inflation increases monotonically with d*, and tends to the limit p as d* grows large. On the other hand, for a small enough value of c(, the possibility exists that the G(p) curve displays a non-monotonic behavior. This case is illustrated in lig. 2b, where now the relationship between inflation and fiscal pressures is not uniquely determined for all values of d*, opening up the possibility for multiple equilibria. Though this latter case is probably not too relevant in practice (cf. section 3 below), its qualitative results are interesting, since it points to the existence of four ‘regimes’, associated with different types of behavior of fiscal policies and inflation.

2.3. Fiscal pressures and inji’ation

These four regions - whose qualitative features, except for the case where a


multiplicity of equilibria arises, are also illustrative of the solutions when c( is ‘large’ - are indicated in fig. 3. The figure shows the graph for the equilibrium relationship d* =G(p) (for a small) together with a plot for the inflation-tax revenue curve R(p). Four cases are clearly distinguished:

inflation (p) Pb PC

[- G(P) R(P) ’

Fig. 3

(i) g* 5 h, + h, (or d* SO). In this case there is no ‘pressure’ to spend beyond the taxing potential. As a consequence, inflation (due to a fiscal impulse) is zero, and the tax system operates with ‘excess capacity’. The desired deficit is zero; the target level of expenditures can be financed by setting taxes below the maximum feasible value; for d * =O, p* =0 is the unique solution to (9) if tl # 0. (ii) O<d*<d*. Here, the target level for spending exceeds the maximum possible revenues from regular taxation. Nevertheless, the pressure to inflate is small relative to the government’s capacity to fund deficits through money creation; d * is still far from reaching the maximum inflation-tax revenue level R(i). Hence, though the government finds it optimal to set inflation at a positive level (so that actual expenditures get closer to the target value g*), it will not validate high inflationary expectations.

In this region, as the fiscal pressures increase, so do inflation and the actual deficit (g-h,- h,/l +p). This follows from applying the implicit function theorem to expression (9) to find

dP 1 _Z (7d *

~ > 0. R’(p) + H’(p)

The above expression is positive since the economy is located on the


increasing branch of the inflationary tax revenue curve: R’(p) >O (see fig. 3) and H’(p) is always positive. In summary, this case would correspond qualitatively to a moderate inflation regime, where fiscal pressures are partially accommodated by the government by raising the inflation tax. There is a unique discretionary equilibrium p* for each value of the ‘fiscal overhang’ d*, so that ‘inflationary traps’ are ruled out. (iii) d* <d* <8*. As the value of d* increases, the possibility appears of multiple equilibria (for small values of a). l2 As fig. 3 illustrates, in this high inflation region, a given fiscal ‘overhang’, say d& is consistent with several steady-state levels of inflation. Not only inflation is higher at points B and C (the ‘bad equilibria’), but also actual government expenditures are lower at those states (compared to point A). The reason for this, of course, is that at such high levels of inflation the economy is already located on the wrong side of the total revenue curve. i3 We note that, here, high inflation traps are possible and, moreover, if the system is in one of those states the same ‘fundamentals’ can be consistent not only with a lower inflation rate, but also with a larger volume of public spending. (iv) Finally, we have the region for which d* >d*. Here, the behavior of the government resembles that of a revenue maximizer. The discrepancy between the level of spending which would ‘pacify’ political demands (g*) and the limited ability to tax (h,+h,) is so big that the target level of deficit is now beyond the maximum level of inflationary tax revenues R(p”) (see fig. 3). Thus, the strong incentives for the government to inflate imply that private expectations are going to be validated only at a very high level of inflation; as a consequence, the equilibrium is located on the wrong side of the revenue curve. Inflation rises with fiscal pressures, but realized expenditures decline with g*:

ap 1 -= ag*

->o, R’(P) + H’(P)

and

-aK = R'(p) $ < 0, ag*

since R’(p) ~0 and H’(p) > 1 R’(p) 1. Thus, a curious effect appears, because increases in fiscal pressures produce

‘2Recall that this region does not exist if a is large. 13Total revenues can be expressed as

g=(m(p)-h,)L+h,+h,=R(P)+h,+h,, l+P

since Rh)S R(P,), then gb) Sg(P.).


a fall in actual spending. This result bears an analogy with the observed behavior in hyperinflations, where public expenditures typically shrink while prices increase very rapidly. Also, from fig. 3 it can be seen that in this region, the equilibrium is again unique and determined only by d *. In other words, no expectations-driven inflationary traps are possible here. In the limit, as the value of d* keeps increasing, p becomes close to p, where even an ‘unexpected increase in inflation would cause a revenue loss through the fiscal-lag effect which nearly offsets the increase in seigniorage. At this point, the government has clearly lost control over its fiscal policies: it is hard- pressed to spend, but cannot deliver even by bearing higher inflationary costs, and it ends up losing both because of the high level of inflation and the low level of realized expenditures.

2.4. High inflation with low deficits

The equilibrium in this fourth region of very high inflation has other interesting properties. As d* keeps increasing, total revenues approach h,, + h,, that is, the revenue from regular taxes in the absence of the fiscal-lag effect. Therefore, actual expenditures are similar to those which the government could sustain at zero inflation. This corresponds to the intuitive idea that, when inflation reaches extremely high levels ‘it serves no purpose’ as a revenue raising devise.

In such conditions, it may seem that the government could actually ‘do without inflation’. Thus, in principle, it could be argued that, if expectations were somehow brought down, regular taxes would automatically compensate for the loss of inflationary revenues, and the economy would stabilize without any further action needed. But this statement would be misleading. In this type of very high inflation, the government is ‘forced’ to contract spending; if by any chance inflationary expectations were to go to zero, the pressures on fiscal policies would come to the open so that, given its unchanged incentives, the government would again rely on the inflation tax. In other words, the zero-inflation policy is not time-consistent. A corollary of this argument would be that a sustained disinflation requires a shift in the parameters of the problem and, in particular, an increase in the taxing capacity and/or a drop in g*. That is, government incentives would have to be changed permanently.

2.5. The effects I$ tax indexing

Another implication of the model has to do with the effect of policies such as the indexation of the tax system (i.e. a reduction of h, for a given fiscal ‘capacity’ h,+ h,). It is found that the effects of such a policy would be


substantially different depending on the type of fiscal and inflationary regime.r4 Applying the implicit function theorem to eq. (9) we obtain

Consider first the case where the economy is in the low inflation regime (d* <A*). As indicated earlier, in this region the equilibrium is located on the increasing side of the inflation-tax revenue curve so that R’(p) is positive; therefore, the denominator in (10) is also positive. With respect to the numerator, though (p/( 1 +p)) (m(p) - h,) is always smaller than d * (as fiscal pressures will only be partially accommodated), the fact that the fiscal overhang is relatively small enables the government to get fairly close to d * with a relatively low inflation rate, so that the expression in brackets in the numerator of (10) will be positive. I5 Thus, when d* is small, indexation reduces the equilibrium inflation rate. The intuition behind this result is clear: the indexation of taxes enables the government to spend more at a given inflation rate; it will use this opportunity to increase expenditues (so as to close the gap between g ad g*) and to cut its inflationary financing at the same time.

On the other hand, in a regime of very high inflation (d* >d), the inverse result holds. Although the denominator is still positive [cf. point (iv) above], the value of the numerator will now be negative. To see this, recall that as d* keeps increasing, m(p) -h, goes to zero implying that actual expenditures (g) tend to h,+h,, and that the expression between brackets in (10) tends to -d* which is, of course, negative. As a consequence, inflation and h, are negatively associated in this region. The intuition is straightforward: in this case, the fiscal lag acts as a disincentive against producing surprise inflation. If this brake operates with less strength, the government’s ‘scramble for revenues’ would make it accelerate inflation even further.

Thus, when the government is subject to strong pressures, not only public expenditures may fall because of the revenue-constraint, but also, measures that in other contexts would have helped to reduce inflation, could in fact accelerate inflation even further. There is an analogy between tax indexation

14This conclusion is independent of whether a is small or large. Fischer and Summers (1989) have analyzed a policy game where tax indexing is taken as a measure that reduces the marginal cost of inflation; they find that the equilibrium inflation rate always increases when the tax system is less exposed to the fiscal-lag effect. Our result is different, since we model directly the effect of tax indexing on revenues - and thus on ‘spending opportunities’ ~ without modifying the parameters of the loss function. In this case, reducing the fiscal lag effect lowers the steady- state inflation when fiscal pressures are low, but may have the opposite effect when d* is large.

15A formal proof is omitted. However, note that as d* tends to zero R(p) tends to d* so that 2R(p)>d*.


and other policies traditionally used to reduce the distortions produced by high inflation, such as the issue of ‘inflation-proof financial assets: such policies may tend to accelerate the inflationary process and, under certain circumstances, may not be welfare improving [cf. Fischer and Summers (1989)].

3. Some simulation exercises

Although the model is not designed to produce quantitatively precise results, it seems useful to make it subject to numerical simulations, in order to get a feeling of what kind of outcomes it generates, and check whether they bear any resemblance with actual experience. In order to do so, it is of course necessary to impose values to the parameters and to specify the functional form of money demand. We will work with a standard Cagan-type function;16 accordingly, the model has one ‘driving variable’, d* and four parameters: the zero-inflation level of money demand A, the semi-elasticity of the demand for real balances fl, the value of the tax base subject to the fiscal lag effect h,, and CI, the ‘inflation cost’ parameter in the government’s loss function. The time unit is defined as one month. Therefore, inflation will be measured as a monthly rate, and (since we shall use a GDP scaling throughout), real money balances shall be measured as a portion of monthly GDP.

The analysis will be carried out using basic values for the parameters reminiscent to that of the Argentine case. A money demand function with A = 1, /?= 3 implies a level of real balances of around 8% of annual GDP at low inflation and a fraction over 3 points of GDP at 30% per month inflation. The value of 3 for the semi-elasticity fi is similar to that estimated by Kiguel and Neumayer (1990) and Rodriguez (1991). This implies a maximum ‘gross’ inflation tax revenue (i.e. before subtracting losses through the fiscal-lag effect) of 9.5% of GDP, at an inflation rate near 25% pr month. Moreover, we shall assume a ‘tax capacity’ (h,+h,) of 20% of GDP, with h, =0.12, that is, a little more than half of the tax base is subject to a one-month collection lag.

The results of the simulations for different values of the parameter a are depicted in fig. 4. They indicate that:

(i) Multiple equilibria are indeed identified in some cases. However, given the monetary-fiscal parameters used in the exercise, steady-state high inflation traps appear only for very low values of a.

16We shall be using this specification as if it was valid for the whole range of possible inflation rates. However, the empirical evidence suggests that the semi-elasticity B may fall with the rate of inflation. See, for example, Easterly and Schmidt-Hebbel (1991).


4a. Beta=3 hl=O.12

0.90

0.80- 0.70-

b o.EG-

5 0.50

2 0.40- $J 0.30-

0.x)-

O.lO- .“.-=+- kr

o.mc ’ 0.00 0.10 0.20 0.33 0.40 0.50 0.60 0.70 0.80

Monthly Inflation (p)

l- alpha=O.CCKE - alpha=O.Ol ..~~~*..~ alpha=O.B 1

4b. Beta=3 hl=O.12 1 .m-

0.93

0.80-

0.00 0.10 0.M 0.30 0.40 0.50 0.60

Monthly Inflation (p)

I- alpha=O.l - alpha=0.5 *‘*~~*~*~ alpha=1 1

Fig. 4. Simulations.


(ii) The inflation rate that solves the model seems not too sensitive to the parameter a when fiscal pressures are low (in the ‘moderate inflation’ range); by contrast, the behavior at large d * varies considerably with a. (iii) The model generates high inflation solutions for large, but not implau- sible levels of d* when a takes ‘intermediate’ values. For example, with a=O.l, a ‘target’ zero inflation deficit of 15% of GDP would be associated with an inflation rate over 20% per month; the same d* would induce a 30% inflation with a = 0.05.” (iv) Inflation rates in the ‘hyperinflation’ range (say, over 50% per month) are not obtained as solution except for very low levels of a or extremely large fiscal pressures. In other words, while high inflations appear in the model as possible steady states for not extraordinary parameter values, hyperinflations look like aberrant states. In a way, this may suggest that the model does not account well for hyperinflations, but it also reinforces the impression of those episodes as extreme cases where the collapse of fiscal policies leaves little room to contemplate other policy objectives - and as phenomena of a transitory nature, which ‘burn themselves out’ in a relatively short interval of time.

4. Fiscal pressures and the public debt

The argument presented in the previous sections relied on the assumption that money creation is the only source of government finance. A closer look at the behavior of public debt in high and hyper-inflations shows a rather complex picture of old debts not punctually serviced or refinanced with great difficulties, new bond issues of limited sizes with very high promised yields and very short maturities - that the market sometimes willingly absorbs even when outstanding assets remain unpaid - more or less ‘forced’ sales of bonds to the financial sector or to suppliers, arrears in the pre-commited payments and the like. That is, by some means or other, governments do engage in debt financing, and the existing debt is not homogeneous regarding the public’s perceptions of the actual default risk. By and large, however, the image of a credit constrained government appears appropriate in those instances: not only does the ‘monetarist arithmetic’ [in terms of Sargent and Wallace (1981)] operate at high speed, but also the perspective of an outright default (which sometimes materializes) is present in agents’ expectations, so that the government’s access to the loan market is severely restricted.

This section sketches a simple extension of the previous model, incorporating government debt. The exercise is set up so as to stay as close as possible

“It may be useful to recall that the measured, actual deficit has a maximum near 9.5”/” of GDP. The solution with p = 0.30 is just above the rate which maximizes the ‘gross’ inflation tax.

100 D. Heymann and P. Sanguinetti, Fiscal inconsistencies and high injation

to the one-period analysis done before; for that purpose and for general tractability reasons, several simplifying assumptions have to be imposed. The problem we study is a variant of one discussed in recent literature [cf. Calvo and Guidotti (1990a, b)]. The government is supposed to have ‘inherited’ a certain volume of debt, of size b, a portion of which is (perfectly) indexed to the price level. All bonds have a one-period maturity and they become due at the end of period t- 1, after that period’s markets have ‘closed’ and before they open in t. At this moment, the government must refinance the debt, or it can choose to repudiate its obligations. If it does default, we assume that no new debt is issued in the following periods. When the default option is not taken, the government may sell either perfectly indexed bonds (i.e. with returns contingent on the price level at maturity) or nominal bonds. In period t, the government taxes, spends and sets the inflation rate; its expenditures now include (in the non-default case) the service of the debt. As before, the decision-maker cares about the level of inflation and the discrepancy between the level of ‘primary’ spending (i.e excluding debt services) and a ‘desired’ level.

In this setting, the government has two ways of trying to reduce the real value of its debt burden: by inflating away its nominal liabilities or by explicitly declaring its bankruptcy, in which case we assume that it bears a cost, which is represented by adding a fixed term to the loss function.” The formulation corresponds to the view that explicit default is more of an ‘eventful’ decision than a choice at the margin; including the costs in the loss function suggests an interpretation as the ‘voice’ of bond-holders, which somehow hurts the government.

The willingness to hold government bonds can be expected to vary with the level of the debt and with d*, ‘the primary fiscal pressures’. When fiscal pressures are large, the government is bearing high costs due to inflation and to the expenditure gap: in those circumstances, the authorities could be tempted to default on the debt, in order to soften a very tight budget constraint, even if that means incurring a direct cost. If that reaction is incorporated into the public’s expectations, a rise of d* above a certain critical value may trigger a collapse in the demand for government debt. This is indeed what the model indicates.”

It is intuitively clear why such a threshold value for d* should exist. For small primary fiscal pressures, the resulting inflation is also low, so that the government can ‘afford’ to increase it slightly in order to collect enough

“Default costs have been treated in various ways in the literature. Calve (1988) introduces them in the government’s budget constraint, so that their effect is to reduce spending. Prati (1991) assimilates default costs to a sort of distortionary tax by including them in the budget constraint of the representative agent. The approach we follow here is similar to that presented in Alesina et al. (1990).

19A proof is outlined in a longer version of the paper.


seigniorage revenues to meet the interest payments and still be able to partially accommodate the (small) pressures for deficit spending. Clearly, this alternative is cheaper compared to the case in which the government decides not to service the debt and bears the fixed cost of default. By contrast, if d * is large, and consequently the inflation rate is also high, the costs of servicing the debt rise considerably both in terms of an increasing inflation and of lower levels of primary expenditures. In those circumstances, the government may find it preferable to pay the cost of repudiation in order to be relieved

of its debt burden. The critical value d^ is of course a function of the parameters of the

problem. Increases in debt services reduce 2: for a given primary fiscal pressure, there is - ceteris paribus - a maximum sustainable debt; this falls as d* grows larger. In other words, a ‘hard-pressed government finds itself credit-constrained even at low levels of debt.

In addition, it can be verified that d^ grows with the proportion of the debt which is indexed to the price level. Thus, according to the model, the existence of nominal debt stengthens the incentive for explicit default: when the debt consists mainly of indexed bonds, the government gains less from inflationary surprises; this, in turn, reduces the equilibrium inflation rate for a given d* and, therefore, lowers the losses in the debt-servicing case.

It would then seem that, in a literal interpretation of the previous result, the ‘optimal’ mix of indexed and nominal debt would be located at the corner solution, where the government does not issue nominal liabilities at all. However, the assumptions of the exercise are clearly biased towards this conclusion. First, we have worked under the hypothesis that it is feasible to have perfect contemporaneous indexing when, in fact, the definition of an appropriate basket to use in indexation clauses is non-trivial, and there are built-in lags in measuring and reporting price indices. This problem becomes of great importance when inflation grows high.

In addition, our analysis has disregarded the possibility of random shocks. Recent literature has stressed that nominal debt can be a low-cost instrument to absorb those shocks [cf. Bohn (1988)]. We conjecture that an analysis similar to that sketched here, but allowing, say, for disturbances in fiscal pressures, may result in a positive level of nominal debt in the ‘preferred’ mix of government liabilities, and also would suggest that the share of nominal debt should vary negatively with fiscal pressures. The qualitative reasoning behind this conjecture is as follows. For low levels of d*, the incentives to generate unexpected inflation are weak, since the government does not face a large spending gap; inflation is also low. In those conditions, the gains from being able to smooth out shocks in d * would offset with excess the losses from a higher equilibrium inflation. On the contrary, with a large d*, the existence of a sizeable volume of nominal debt has strong effects on equilibrium inflation and the ‘marginal cost’ of inflation is high: this would

102 D. Heymann and P. Sanguinetti, Fiscal inconsistencies and high inj7ation

imply a preference for issuing indexed bonds - as long as d * remains lower than the critical value -, even if it means giving up an option to accommodate fluctuations in desired spending.

5. Concluding remarks

We have developed a very simple model of inflationary finance. In its formal aspects, this model represents a slight departure from the existing literature. Its result, however, seem to capture some of the features of different inflationary regimes. We would like to stress three types of results.

First, the measured deficit may not be a good indicator of the government’s incentives to engage in inflationary financing. A low deficit can mean either that the government ‘lives comfortably within its means’ or, alternatively, that it is subject to very strong pressures, but is unable to finance more than a limited volume of spending at a very high inflation rate. Beyond an initial transitional period, a sustainable stabilization would require diminish- ing the pressures on fiscal policies.

Second, multiple equilibria are typical features of models of high inflation. They do arise in our model, but as a property of certain particular cases. Our analysis suggests that moderate inflation regimes are immune to (at least some type of) multiple equilibria problems; very high inflations, on the other hand, do not appear here as expectational ‘traps’.

Third, some policy measures, i.e. tax indexation, may have quite different effects depending on the inflationary regime, since they may act in quite different ways on the incentives of the government and of the groups that influence fiscal policy.

The main conclusions of the model depend on a particular specification of the policy-maker preferences. The choice of the loss function allows us to incorporate in a single analysis different types of fiscal behavior. It is clear, however, that there is a considerable degree of arbitrariness in the specitica- tion of government preferences. Ideally, a model like this one should be embedded in a more comprehensive description incorporating the distributio- nal games which determine both the capacity of the tax system and the demand for public spending. An analysis of this kind, most probably, would emphasize the importance of fiscal institutions such as those governing the budget process and the asymmetries in the power to influence fiscal policies given by such institutions and by the relative strength of competing social groups.

From the point of view of a narrowly economic description of the inflation process, our model produces steady-state outcomes which do not represent accurately the volatile behavior observed in high inflation episodes. We have down-played informational considerations and issues of price dynamics in order to highlight what we believe are some interesting aspects of the high


inflation problem; a full account of the matter should certainly include those

themes in a prominent place.

References

Alesina, A., A. Prati and G. Tabellini, 1990, Public confidence and debt management: A model and a case study of Italy, in: R. Dornbusch and M. Draghi, Public debt management: Theory and history (Cambridge University Press, Cambridge).

Alesina, A. and G. Tabellini, 1990, A political theory of fiscal deticits and government debt, Review of Economic Studies, July.

Barro, R., 1983, Inflationary finance under discretion and rules, Canadian Journal of Economics, Feb.

Bohn, H., 1988, Why do we have nominal government debt?, Journal of Monetary Economics 21, no. 1.

Bruno, M., 1989, Econometrics and the design of economic reform, Econometrica, Feb. Bruno, M., 1991, High inflations and the nominal anchors of an open economy, Working paper

no. 3518 (NBER, Cambridge, MA). Bruno, M. and S. Fischer, 1990, Seignorage, operating rules and the high inflation trap,

Quarterly Journal of Economics, May. Calve, G., 1978, On the time consistency of optimal policy in a monetary economy,

Econometrica. Calvo, G., 1988, Servicing the public debt: The role of expectations, American Economic Review,

Sep. Calvo, G. and P. Guidotti, 1990a, Indexation and maturity of government bonds: An

exploratory model, in: R. Dornbusch and M. Draghi, eds., Public debt management: Theory and history (Cambridge University Press, Cambridge).

Calvo, G. and P. Guidotti, 1990b, Credibility and nominal debt: Exploring the role of maturity in managing inflation, IMF StaB Papers 37, Sep.

Cukierman, A., 1988, Rapid inflation - deliberate policy or miscalculation?, Carnegie-Rochester Conference Series on Public Policy 29, 1 l-76.

Cukierman, A., S. Edwards and G. Tabellini, 1992, Seignorage and political instability, American Economic Review, June.

Dornbusch, R. and S. Fischer, 1986, Stopping hyperinflations past and present, Weltwirschaft- liches Archiv 122, l-46.

Dornbusch, R., F. Sturzenegger and H. Wolf, 1991, Extreme inflations: Dynamics and stabilization, Brookings Papers on Economic Activity 2, 2-84.

Easterly, W. and K. Schmidt-Hebbel, 1991, The macroeconomics of public sector deficits: A synthesis (World Bank, Washington, DC).

Evans, J. and G. Yarrow, 1981, Some implications of alternative expectation hypotheses in the monetary dynamics of hyperinflations, Oxford Economic Papers, March.

Fischer, S. and L. Summers, 1989, Should governments learn to live with inflation?, American Economic Review, May.

Frenkel, R. and J. Fanelli, 1989, El regimen de alta inflation, Document0 de Trabajo, CEDES. Heymann, D., F. Navajas and I. Warnes, 1991, Conflict0 distributivo y deficit fiscal: Algunos

juegos inflacionarios, El Trimestre Economico, Jan.-March. Kiguel, M., 1989, Budget deficits, stability and the monetary dynamics of hyperinflations,

Journal of Money, Credit and Banking, May. Kiguel, M. and N. Liviatan, 1989, Some implications of policy games for high inflation

economies, WPS 379 (Country Economics Department, World Bank, Washington, DC). Kiguel, M. and A. Neumeyer, 1989, Inflation and seignorage in Argentina, Working paper no.

289 (World Bank, Washington, DC). Leijonhufvud, A., 1990, Extreme monetary instability: High inflations, Working paper no. 53/

1990 (University of Hohenheim). Persson, L. and L. Svensson, 1989, Why a stubborn conservative would run a deficit: Policy

with time inconsistent preferences, Quarterly Journal of Economics.


Persson, T. and G. Tabellini, 1991, Macroeconomic policy, credibility and politics (Harwood Academic Publishers, London).

Prati, A., 1991, Self-fulfilling debt runs and debt management, Mimeo. (UCLA, Los Angeles, CA).

Rodriguez, C., 1991, The macroeconomics of the public sector deficits: The case of Argentina, WPS 632 (World Bank. Washington. DC) March.

Rogoff, K., 1989, Reputation, coo;dination’ and monetary policy, in: R. Barro, ed., Modern business cycle theory (Harvard University Press, Cambridge, MA).

Sanguinetti, P., 1992, The politics of intergovernmental transfers and local government deficits: Theory and evidence, Mimeo. (UCLA, Los Angeles, CA).

Sargent, T. and N. Wallace, 1986, Some unpleasant monetarist arithmetic, in: T. Sargent, Rational expectations and inflation (Harper & Row, New York).

Fiscal inconsistencies and high inflation

Documents