Fiscal Expansion and External Current Account Balance

This paper analyzes, in a general equilibrium framework, how fiscal policy transmits its effects to the current account of the balance of payments. In discussing the role of fiscal policies in the Fund’s stabilization programs, the paper compares different approaches to explain the determination of the external balance. The model presented here is based on intertemporal optimization and focuses on the behavior of private and public agents. Some empirical evidence is also presented, based on data from ten Latin American countries. Finally, an appendix presents an integrated system of financial, external, and government accounts.

Abstract

This paper analyzes, in a general equilibrium framework, how fiscal policy transmits its effects to the current account of the balance of payments. In discussing the role of fiscal policies in the Fund’s stabilization programs, the paper compares different approaches to explain the determination of the external balance. The model presented here is based on intertemporal optimization and focuses on the behavior of private and public agents. Some empirical evidence is also presented, based on data from ten Latin American countries. Finally, an appendix presents an integrated system of financial, external, and government accounts.

Summary

This paper analyzes, In a general equilibrium framework, how fiscal policy transmits its effects to the current account of the balance of payments. Some empirical evidence is also presented, based on data from ten Latin American countries.

The main findings of the empirical estimates are three. First, the inflation tax, that is, one way in which the government deficit is financed, exerts a large negative effect on private savings and, hence, on the current account. Second, the equivalence of debt and tax financing of government expenditure (the so-called Barro neutrality hypothesis) cannot be substantiated by the data. This implies a critical role of fiscal policy in the determination of external balances since a change in the taxation-borrowing mix appears to have a major influence on the current account through its effect on savings. Third, government capital expenditure crowds in private investment: it may do so because it increases productivity (when it provides infrastructure and services) or because it provides financial resources to the private sector. Therefore, government capital expenditure seems to exert a major influence on private investment that was found largely independent of foreign and domestic interest rates (providing an additional increase of 1.26 in total investment for every unit increase of public capital formation). But, since it also increases absorption, this expenditure will tend to worsen the current account, other things being equal. However, this would tend to be offset by the rise in profits, which, by increasing private savings, would improve the current account. Therefore, if projects show adequate returns, the initial negative position of the current account will be sustainable.

Various competing approaches explain the determination of the external balance: the monetary, the absorption, and the fiscal approach, as well as the traditional elasticities approach. Fund stabilization programs are seen to rely for their theoretical background on a mixture of the monetary and the absorption approaches.

In recent years, a more comprehensive general equilibrium approach based on intertemporal optimization has been developed. Several models based on this approach concentrate on the role of private agents’ saving and investment decisions and analyze whether present generations expand their budget constraint by taxing future generations via government budget deficits, which, in turn, create deficits in the current account. The model presented here is based on a similar theoretical framework, but it focuses on the behavior of private and public agents. An appendix presents an integrated system of financial, external, and government accounts.

I. Introduction

This paper analyzes, in a general equilibrium framework, how fiscal policy transmits its effects to the current account of the balance of payments. Some empirical evidence is also presented, based on data from ten Latin American countries.

The main findings of this paper are that (1) the inflation tax, that is, one way in which the government deficit is financed, exerts a large negative effect on private savings and, hence, on the current account; (2) the equivalence of debt and tax financing of government expenditure (the so-called Barro neutrality hypothesis) cannot be substantiated by the data. This implies a critical role of fiscal policy in the determination of external balances since a change in the taxation-borrowing mix appears to have a major influence on the current account through its effect on savings; and (3) government capital expenditure crowds in private investment: this may be because it increases productivity (when it provides infrastructure and services) or because it provides financial resources to the private sector. This latter variable seems to exert a major influence on private investment that was found largely independent of foreign and domestic interest rates.

There are various competing approaches to explain the determination of the external balance: the monetary, the absorption, and the fiscal approach, as well as the traditional elasticities approach. Stabilization programs designed by the International Monetary Fund have been characterized as relying for their theoretical background on a mixture of the monetary and the absorption approaches. 1/ In addition to the fact that some aspects of the two approaches are difficult to reconcile, neither can explain why the fiscal variables—not only the financing of the deficit—are a crucial component of the Fund’s stabilization programs. In other cases, the fiscal approach has been seen as the theoretical support of the fiscal component of Fund programs, 2/ but, although it can be represented as the real counterpart of the monetary approach, 3/ its reduced form is, in fact, an identity between the fiscal balance and the current account of the balance of payments, and, hence, the approach fails to identify the endogenous causes of changes in the current account.

In recent years, a more comprehensive general equilibrium approach based on intertemporal optimization has been developed. Several models based on this approach concentrate on the role of private agents’ saving and investment decisions and analyze whether present generations expand their budget constraint by taxing future generations via government budget deficits, which, in turn, create deficits in the current account. 4/ The model presented here is based on a similar theoretical framework, but it focuses on the behavior of private and public agents. The saving behavior of households is determined according to the intertemporal optimizing model of Yaari (1965) and Blanchard (1985), and that of investment by firms is determined by a stock adjustment model. Government current revenue is determined by a “tax-smoothing” approach and government current expenditure is also explicitly modeled. Output is taken as given and can be formalized as supply-determined output with the quantity of factors given in the short to medium term. Only one good is considered; consequently, terms-of-trade changes are not considered in the present model specification.

The intertemporal optimization models usually lack consistent empirical estimates; at the econometric level the general equilibrium approaches become very “partial” or the estimates are not well founded on the behavior of agents. This paper tries to fill the gap: its focus is on the empirical application of the model, namely, on (i) the investigation of the forces that account for current account imbalances at a microeconomic level; (ii) the direct and indirect roles played by different kinds of fiscal policies in determining the current account imbalances; and (iii) the application of this kind of model to Fund-supported stabilization programs.

The empirical model estimated here is a five-equation system of government current expenditure, government current revenue, total investment, private saving, and the current account. These equations form a recursive model that clearly spells out both the direct effect of private behavior and of fiscal policy on the current account and the indirect effect of fiscal policy through changes induced in private sector behavior. 5/

Section II contrasts the monetary and fiscal approaches to the balance of payments at the analytical level. It describes the assumptions upon which they rely and presents the structure of the two models. It concludes by showing how the two models can be reconciled in a more general framework. Section III presents the empirical estimates of the five-equation model based on that general framework and shows the effect of specific changes in the explanatory variables on the current account of the balance of payments. Section IV draws some conclusions regarding the policy issues raised by the empirical estimates of the model. Appendix II presents an integrated system of financial, external, and government accounts. This accounting framework, which starts with the current account of the balance of payments identity—on which the fiscal approach was founded—is transformed to provide the basis for the absorption approach and, with further elaboration, also for the monetary approach.

II. Comparison Between Monetary and Fiscal Approaches to the Balance of Payments

This section analyzes the basic characteristics of the monetary and the fiscal approaches to the balance of payments in a context of a fixed and a flexible exchange rate system in the short and long run. It also compares the two models to the more recent portfolio balance models of exchange rate determination.

1. The monetary approach

The monetary approach to the balance of payments, as was largely developed in the Fund and the University of Chicago at the end of the 1950s, stresses the essentially monetary nature of balance of payment imbalances. “its essence is to put at the forefront of analysis the monetary rather than the relative price aspects of international adjustment ….” 6/ “Accordingly, surpluses in the trade account and the capital account respectively represent excess flow supplies of goods and of securities, and a surplus in the money account reflects an excess domestic flow demand for money. Consequently, in analyzing the money account, or more familiarly the rate of increase or decrease in the country’s international reserves, the monetary approach focuses on the determinants of the excess domestic flow demand for, or supply of, money ….” 7/ In the original formulation, the theory is framed in a long-run perspective with the crucial assumption that the monetary authorities cannot sterilize balance of payments surpluses or deficits, and, therefore, they will be channeled into the money supply. 8/ 9/

The fundamental behavioral equation of the model is the money demand function:

Md=p.f(Y,i)(1)

where money demand (Md) is equal to the domestic price level (p), times a function of real income (Y) and nominal interest rate (i). The money demand function is assumed to be stable, and the incremental demand for money is a function of the growth of nominal income. If the economy considered is a small, open economy it would face prices and interest rates determined at the world level. Perfect international capital mobility and perfect substitutability between domestic and foreign bonds are also usually assumed. Perfect international capital mobility implies that the interest rate on domestic bonds is equalized with the interest rate on foreign bonds, plus the forward premium on the exchange rate—that is, covered interest parity; it should not be overlooked that perfect capital mobility is based on the assumption that there is no differential risk of default nor differential risk regarding possible changes in financial markets rules and the exchange rate regimes. 10/ Perfect bond substitutability implies that the agent—behaving according to rational expectations—will allocate their portfolio shares indifferently between domestic and foreign bonds with the same expected rate of return, when expressed in the same currency.

Finally, it is assumed that purchasing power parity (PPP) holds. PPP relies, in turn, on the assumption that prices are perfectly flexible, and, therefore, any change in the nominal exchange rate will not be reflected in the real exchange rate. Perfect flexibility of prices and, in particular of the price of labor, ensures that output will be at its full employment level.

In a system of a fixed exchange rate, we can summarize the assumptions in the model as follows:

i=i*(2)
p=p*(3)
y=y¯(4)

where i (i*) refers to domestic (world) nominal interest rate, p (p*) to domestic (world) price level, and y (y¯) to domestic (full employment) output.

Given that the money supply equals the sum of international reserves (R) and domestic credit (DC), 11/ equilibrium in the money market requires

R=MdDC.(5)

Thus,

ΔR=ΔMd(p,Y,i)ΔDC.(6)

An increase in the demand for money raises international reserves, while an increase in domestic credit by the central bank reduces international reserves.

Therefore, in a fixed exchange rate system, balance of payments disequilibria can be adjusted through a reduction of domestic credit to a level consistent with the evolution of the money demand.

In a flexible exchange rate system, the uncovered interest parity, i - i*, is equal to the expected depreciation of domestic currency:

ii*=E(Δee),(7)

where e is the nominal exchange rate, defined as units of domestic currency in terms of foreign currency.

The depreciation expectation is formed rationally and corresponds to the expected inflation differential, if PPP holds:

E(Δee)=E(Δpp)E(Δp*p*).(8)

For the monetary approach to be able to explain the determination of the balance of payments or the exchange rate in its original formulation, all the above assumptions should hold, together with a stable demand-for-money function. Available empirical evidence has shown, however, that all these assumptions generally do not hold. Moreover, the relationship between the exchange rate and the various components of the balance of payments, and in particular the current account, cannot be easily explained by the monetary approach. Within the narrow version of the monetary approach one can explain that relationship only through price shocks (such as the two oil shocks of 1973 and 1979, which raised the world demand for dollars) or through announcement effects of unexpected outturns in the trade balance. 12/

Therefore, a new generation of models has been originated, still in the tradition of asset-demand models of the balance of payments, but some of the original assumptions of the monetary approach have been relaxed. These are the so-called portfolio-balance models, which are able to offer explanations of the relationship between the current account and the exchange rate by taking into consideration the capital account and, thus, the real (wealth) effects of the current account imbalances. The counterpart of current account surpluses is a shift of wealth from foreigners to residents. The increase in wealth for residents raises money demand and also the demand for domestic bonds if these are not perfect substitutes for foreign bonds. Preferences for domestic bonds can derive from special tax treatment of government bonds as well as from political risk attached to the foreign bonds.

The portfolio-balance models allow for a more complex view of the adjustment mechanism. First, they take into account that the assets market reacts more quickly than the goods market, thus generating risks of overshooting and cumulative destabilizing effects. Second, they take into account the existence of nontraded goods and securities, that is, imperfect goods and financial markets. 13/ This implies, in turn, that any change in financial policy affects not only the balance of payments but also prices and the domestic interest rate. Under these circumstances, the financial crowding out of private firms can be brought about by fiscal expansion by affecting domestic interest rates. This mechanism was not an option in the original version of the monetary approach applied to small, open economies.

The monetary approach’s focus on “the direct connection between the money market and the balance of payments, rather than working through the implied changes in the goods or financial assets markets” 14/ gives simple and comprehensive indicators for the economic stance; however, it gives few instruments to suggest and monitor adjustment policies where the intermediate targets often move in the short period in a direction opposite to the final result. Therefore, in devising a theoretical underpinning for the use of fiscal targets in adjustment programs such as those of the Fund, a so-called fiscal approach has been proposed. 15/

2. The fiscal approach

In contrast to the absorption approach, the fiscal approach, developed by the Cambridge Economic Policy Group (CEPG) in the mid-1970s, focuses on the public sector saving as the only variable relevant to determining the current account of the balance of payments. In common with the monetary approach, the fiscal approach extends the theories of the balance of payments of the 1960s in order to consider stocks demand for assets together with expenditure decisions. The fiscal approach lumps together the private expenditure for consumption and for investment:

E=(1α)YtP+αYt1P(9)

where the nominal private expenditure, E, at time t, is a stable proportion of the nominal disposable income at t and one year lagged, YtP and Yt1P. From their econometric work on the U.K. economic model, Cripps, Fetherston, and Godley (1976) found the coefficients of disposable income to sum to almost unity; thus, they decided to disregard any multiplier/accelerator mechanism in favor of a pure stock adjustment model: the private sector holds a given assets portfolio, determines expenditure to maintain it at the desired level, 16/ and adjusts quickly its expenditure to changes in income. The stock adjustment can be modeled as follows:

Wt*=βYtP(10)

where W*t is the desired wealth of time t and YtP is the private disposable income in the same period. The actual increases in wealth are

ΔWt=γ(Wt*Wt1)(11)

As the wealth increase is equal to the period saving, we can also write

St=γ(Wt*Wt1)(12)

or

St=γβ(YtPWt1).(13)

The demand for the net stock of financial assets is assumed to be a “small and stable” proportion of the disposable income of the private sector. 17/ Interest rates are fixed and investment demand is totally interest-inelastic. Hence, the fiscal approach (which ignores net income and transfers from abroad) models the current account of the balance of payments, X - Z, as determined by the fiscal balance, T - G, and private balance, S - I, as follows:

XZ=(SI)+(TG).(14)

Under certain conditions, the fiscal and the monetary approaches can be considered mirror images of each other. In their simplest version—with only one financial asset and the private expenditure depending solely upon asset stock disequilibrium—the monetary approach concentrates on the official settlement accounts and lumps everything else into the category of “items above the line.” The fiscal approach concentrates on the current account and lumps everything else into the category of “items below the line.” 18/

The flow equilibrium conditions for the commodity and money markets can be written as:

article image

With capital movements equal to zero, the sums by column equal zero, showing the perfect similarity of the theoretical form of the fiscal and monetary approach.

The fiscal and the monetary approaches, however, are rooted in substantially different views about the working of the labor market and of price and output flexibility. While most versions of the monetary approach usually assume a situation of continuous full employment, output and employment are considered flexible by the fiscal approach. Therefore, a fiscal expansion raises output, which is not assumed at its full employment level, and thus tax revenue. The fiscal deficit will be less than the initial fiscal expansion. In addition, there will be no crowding out of the private sector because of the assumption of perfect open financial markets, which implies i = i*, and of a marginal private propensity to spend that is assumed to be unity for both consumption and investments.

Accordingly, the policy recommendations of the two approaches to achieve equilibrium of the external balance are far apart: the key policy recommendations of the New Cambridge school for the U.K. economy was the introduction of import restrictions. This would offset the government expenditure that should continue to play a role in supporting domestic demand. The monetary approach, instead, puts the burden of adjustment on the domestic credit creation and on the government deficit, which is considered the main cause of the domestic credit increases.

The differences in the policy recommendations depend on the sensitivity of exports to changes in domestic prices and on the sensitivity of prices to changes in demand. In the monetary approach framework, the parameter that measures the price sensitivity of exports tends toward plus infinity because of the assumption of a small, open economy for which PPP holds and the demand for exports is infinite. In the fiscal approach, instead, this parameter has a positive but finite value. In the monetary approach, the parameter that measures the price effect of changes in demand in the price equation is equal to plus infinity, as the labor supply curve is vertical; for the fiscal approach, the same parameter is equal to zero, because any change in demand will change the output—not at the full employment level—rather than the prices. The sensitivity of prices to changes in the exchange rate is assumed equal to unity by both schools because they exclude money illusion. However, the fiscal approach allows for lags in the real wage resistance (that is, the exchange rate elasticity of prices is less than unity in the short run) so that, to offset the effects of government expenditure on the current account of balance of payments, the CEPG proposed import quotas or an increase in import tariffs, arguing that increasing tariffs would have the same result as an autonomous reduction in the import propensity of the U.K. economy. 19/

Despite these unorthodox policy recommendations, which depend on the assumption of (stationary) quasi-steady state where private saving equals zero and interest rates are fixed, the model was extended by Milne (1977) and Kelly (1982) from the industrial countries’ framework in order to be applied to the analysis of Fund-supported stabilization programs in developing countries. The fiscal approach to the balance of payments determination is based upon the national accounts identity, which states that the current account of the balance of payments is equal to the government balance and the private sector balance between investment and saving (equation 14). However, Milne’s estimates of equation (14), quoted also by Kelly, 20/ are not very different from estimating a national accounts identity; given the current account identity and treating the private balance as a constant, the parameter of the public balance is bound to be not significantly different from unity.

3. An alternative fiscal approach model

A model that can explain changes in the current account of the balance of payments can be built within the fiscal approach framework once the behavioral content of private and public saving and of investment is specified and once the criticisms to the rather simple behavioral relation between income and private expenditure of the private sector that characterize the fiscal approach are taken into account. The direct effect of government spending and of different kinds of taxes on aggregate demand, and therefore on the current account, should not be overlooked nor the effect on private consumption of an increase in debt, in the monetization of debt (inflation tax), and in taxes. The full exogeneity of the private sector on the determination of the current account stated by the fiscal approach can be relaxed without swinging to the full exogeneity of the public sector on determining the same as in a neutral framework à la Barro.

Barro’s (1974) neutrality hypothesis states that financing government expenditure by taxes or by debt has the same impact on the intertemporal allocation of national consumption because, to fulfill the intertemporal budget constraint, the agents discount the value of the present government debt with the equivalent future tax liabilities necessary to service such a debt. 21/ Assumptions necessary for this hypothesis to hold are the absence of distortionary taxes (i.e., only lump-sum taxes are allowed), perfect capital markets, and infinitely-lived agents (or perfect intergenerational chains). Infinite horizons of the agents make the intertemporal budget constraint of individuals equal to that of the government; with perfect financial markets this amounts to the same discount rate for the individual and the government. In this framework, financing policy has no consequences on the current account of the balance of payments because every fiscal expansion is offset one-to-one by the private sector. It has been demonstrated that once other than lump-sum taxes are introduced in the model (Barro (1979)), or allowances are given for borrowing constraints (Tobin and Buiter (1976)), or myopia or finite horizons of the agents are introduced, the neutrality hypothesis does not hold.

To obtain real effects from current and anticipated financing policy, one or more of the equivalence assumptions must be relaxed. The three sectors considered in the model are households, firms, and government.

The private saving function is based on the Blanchard-Yaari model, 22/ which allows for finite horizons of the agents and thus keeps the distinction between individual and government discount rates. In this model, agents face a probability of death, w, which is constant throughout the agent’s life. The existence of insurance companies allows a transfer of wealth from those who die: thus total financial wealth, W, accumulates at the rate r, the interest rate, while individual—financial and human—wealth accumulates at a rate r + w.

The aggregate consumption function is a linear function of aggregate financial and human wealth:

C=(ω+θ)(H+W),(17)

where ω is the constant probability of death and θ is the discount factor (pure time preference rate). Accumulation of human wealth is given by

H=(r+ω)HY,(18)

where Y is noninterest income; financial wealth is given by

W=rW+YC.(19)

If the government and the foreign sectors are introduced, where government spending is financed either by lump-sum taxes, T, or by debt, D, the consumption function becomes

C=(ω+θ)(YTr+ω+D+F),(20)

where nonhuman wealth has been substituted by government debt and net foreign assets. 23/ The dynamic foreign budget constraint is

F=rF+YCG,(21)

which amounts to net income from foreign assets and the total saving of the economy. The government dynamic budget constraint is

D=rD+GT(22)

D equals zero because D, G, and T are assumed to be constant. Only at time to do D and T increase permanently.

In a steady state, a change in foreign assets is a decreasing function of government debt as well as of consumption:

dFdD=1(ωθr)1(r+ω)(ω+θ)ω(23)=(ω+θ)ωω+θr1r+ω><1becauser><θ.(24)

When r = θ the change in foreign assets will offset one-to-one the change in debt, that is, F = -D.

When agents have infinite horizons, w = 0, the foreign assets are independent from changes in D—that is, we are in the case of debt neutrality.

This result is quite different from that predicted by Barro’s neutrality hypothesis. According to the hypothesis, a zero increase in consumption is to be expected when debt increases. This corresponds to an increase in private savings equal to -D, with no effects on foreign assets in an open economy framework. Indeed, with ω = 0, that is, infinite lives, the individual has the same budget constraint—the same horizon and the same interest rate—of government, and is thus indifferent to the timing of taxes.

In a two-period framework, households maximize a two-period utility function subject to an intertemporal budget constraint. 24/

Max[U(C1)+U(C)2(1+δ+ω)],(25)

where

U(C)=C1σ(26)
U(C)=Cσ.(27)

As usual, the first-order condition for intertemporal utility maximization requires that the marginal rate of substitution between consumption in two consecutive periods equals the reciprocal of the market discount rate to the private sector.

U(C1)U1(C2)=(1+r1+δ+ω)(28)
(C1σC2σ)=(1+r1+δ+ω)(29)
(C1C2)σ=(C2C1)σ=(1+r1+δ+ω)(30)
(C2C1)=(1+r1+δ+ω)1σ(31)

Whether consumption will be rising over time will depend on the ratio between the real interest rate, on the one hand, and the sum of the discount rate and the probability of death, on the other.

Human and nonhuman wealth must be equal to the discounted value of household consumption:

C1+C21+r+ω=H+W,(32)

where human wealth, H, is equal to the discounted value of labor income minus taxes:

H=Y1T1+Y2T21+r+ω.(33)

The hypothesis of utility maximization implies that the consumer at any age allocates resources according to his life resources, that is, the present value of his labor income and the stock of wealth in his possession. 25/

Private savings are equal to private disposable income less private consumption:

Sp=Y1T1C1.(34)

Thus, from equation (31) and equation (32), the two-period private saving function will be

Sp=Y1T1[11+(1+r1+δ+ω)1σ(11+r+ω)]*(H+W).(35)

Changes in the interest rate on financial assets will change the intertemporal allocation of resources because of the intertemporal substitution effect as well as the wealth effect.

The investment function is a stock adjustment one: firms invest to achieve the optimal, desired capital stock, K*. Net investment will then be used to partially adjust the actual to the desired capital stock:

netIt=(1λ)(Kt*Kt1)t=1,2(36)

where λ is the adjustment coefficient to be comprised between 0 and 1.

Government expenditure is given and taxes are set to comply with the government budget constraint:

G1+G21+r=T1+T21+r,(37)

where the implicit initial stock of bonds is assumed to be zero. Government saving is defined as

SG=T1G1.(38)
The private saving equation (35), together with the investment equation (36) and the government saving equation (38), constitute the three building blocks for the following current account equation:
Sp1+Sg1I1=CA1.(39)
26/

The macroeconomic equilibrium is achieved when private and public savings minus total domestic investment equal the current account.

In this framework, the role played by intertemporal substitution effects suggest that fiscal policy can modify the current account balance indirectly, through its effects on investment and saving behavior. Therefore, a cut in the budget deficit through an increase in taxes can improve or deteriorate the current account according to the substitution effect. The same uncertain effect will result from a cut of public investment expenditure, once the assumption of fixed output is relaxed. The following section estimates empirically the model discussed here, specifying the above equations to take into account historical and institutional characteristics of the countries examined. Moreover, the hypothesis of a direct effect of government expenditure, revenue, and deficit on the current account will be tested.

III. The Empirical Model

1. Specification and estimates

An empirical approximation of the equations described above are estimated for ten South American countries, using annual data for the period 1973-83. 27/. The period chosen includes the two oil shocks, the increase in world interest rates, and the appearance of the debt crisis in the developing countries. The limited availability of data prompted the adoption of pooled time-series and cross-sectional data to estimate the model. To avoid the heteroskedasticity problems often connected with pooled time series, all variables are deflated by a measure of size: most of the variables used in the regressions are scaled to gross domestic product (GDP) at market prices and some variables are scaled to total population. To take into account the different institutional characteristics, ten country dummies replace the constants in the estimates. The sources of the data are the United Nations’ National Accounts, the World Bank Debt Tables, the International Financial Statistics Yearbook and the General Finance Statistics Yearbook. To ensure the consistency between public revenue and expenditure and private investments and savings with the current account of the balance of payments, the United Nations’ National Accounts are the source also of fiscal data, whenever possible. An attempt has been made to construct a coherent set of information and of classification rules, which are especially important when working with cross-country data. 28/

The model estimated in this section is a five-equation system of government current expenditure, government current revenue, total investment, private saving, and the current account of the balance of payments. The equations have been estimated with ordinary least squares because the model is recursive. The following notations are used:

  • CE = government current expenditure

  • CR = government current revenue

  • I = total investment

  • SP = private saving

  • SG = government saving = CR - CE

  • CA = current account balance of the balance of payments

  • GDPPC = GDP per capita

  • RGDPG = real GDP growth

  • RFIR = real foreign interest rate

  • RIR = real domestic interest rate

  • INF = inflation rate

  • XZ = exports plus imports (that is, the trade component of total output)

  • GCF = government capital formation

  • INFTAX = inflation tax computed on government outstanding domestic debt at the end of the year.

A bar over a variable denotes ratio to GDP, and the suffix t-1 denotes a year lag. According to the definition of a recursive model, the structural equations can be ordered in the following way:

CE¯=f1(GDPPC,RFIR,RIR,CEt1¯,INF)(40)
CR¯=f2(GDPPC,XZ,CRt1¯)(41)
I¯=f3(RGDPG,GCF¯,It1¯)(42)
SP¯=f4(RGDPG,CR¯,SG¯,SPt1¯,INFTAX)(43)
CA¯=(CR¯CE¯)+(SP¯I¯)(44)

Equations (40)-(42) consist of only exogenous variables on the right-hand side. Equation (43) includes two previously estimated endogenous variables, and all the previous equations enter in the current account equation. In this recursive system, the error terms are assumed to be independent. Thus, each equation has been estimated with ordinary least squares without incurring a simultaneous bias (see Summary Table).

Summary Table.

Specifications and Estimates for the Model

article image
article image
article image
article image
The t-tests are in parentheses.

Equation chosen for derivation of the reduced form for the current account of the balance of payments on page 26.

a. The expenditure equation

The structural equation of current government expenditure depends, on the one hand, on the population structure, the size of the outstanding—domestic and foreign—government debt, the interest rate on government debt, and the inflation rate, and, on the other hand, on cyclical elements such as unemployment benefits. The higher the ratio of old-age population out of the total, the higher are expected to be health expenditures and current transfers for pensions; the higher the proportion of population of age 0-14 years—which is more typical of developing countries—the higher is the education expenditure.

From the estimates, the coefficient of the GDP per capita variable shows that the current expenditure is not countercyclical. The reason is possibly linked to the fact that expenditures are very inelastic as it is shown by the positive and significant coefficient of the lagged dependent variable. Both the domestic and foreign interest rate coefficients are positive, but only the first one is significant at 1 percent confidence level. Inflation increases current expenditure, even if with a very small impact and only at a 10 percent level of significance. Neither of the population variables—the percentages of the population over 65 years and below 14 years over the total—turned out with coefficients significantly different from zero; therefore, equations including these variables are not shown.

b. The current revenue equation

The current revenue equation is based on previous studies of taxable capacity, 29/ which estimate the ratio of taxes to GDP by regressing it on economic variables 30/ that are used as proxies to the base to which the tax rates are applied. A theoretical shortcoming of these tax-handle models—which, in turn, overcome basic shortcomings of the simple “tax-ratio” approach prevailing previously—is that they implicitly assume that revenue determination is the first step in determining the fiscal balance. In other words, the tax revenue equation is estimated as the first equation of a recursive simultaneous equation model 31/ on the basis of which expenditure and then the deficit are decided. If this implicit assumption is removed, expenditure and outstanding debt can enter as explanatory variables in the determination of tax revenue, and taxes are determined within the government budget constraint. Moreover, the “tax-smoothing principle” proposed by Barro (1979) can be taken into account. According to this principle, smoothing tax revenues over time minimizes collection costs and any excess burden of taxation. To test for this hypothesis, some approximation for the present value of future public expenditure should be introduced in the equation.

The results reported in the Summary Table are more favorable to the tax-handle approach than to the tax-smoothing principle: the coefficients of GDP per capita and the trade component of total output are positive and significant at 1 percent level; this trade component, XZ, was remarkably stable throughout different estimates, confirming the relevance for developing countries of taxes based on international trade. 32/ The coefficient of lagged tax revenue, a proxy for historical administrative capacity, is significant at 2.5 percent level of confidence.

In equation (b.2) of the Summary Table, the inflation coefficient showed a consistently negative sign that can be interpreted as a further confirmation of the “Tanzi effect” of high inflation on revenue collection. 33/ However, the coefficient was not significantly different from zero. The tax-smoothing principle was tested adding public expenditure, domestic debt (lagged one year), and foreign debt (lagged one year) to the explanatory variables. None of these variables proved to be significant and the lagged foreign debt appeared with the wrong—negative—sign.

c. The investment equation

The investment equation (36) lies within the framework of the stock adjustment models. As has been shown by Blejer and Kahn (1985), the market imperfections in the developing countries, such as the lack of developed financial markets, institutional constraints in the labor and foreign exchange markets, together with the large share of public investment on the total, make it difficult to apply other optimizing investment theories to estimate investment in developing countries. To these general problems due to the necessary assumptions of the neoclassical models of investment, problems of data collection are added: no data exist on capital stocks or on the user cost of capital.

The agent behavior aiming to achieve optimal capital stock can be modeled with the accelerator hypothesis, which states that given a constant capital/output ratio and full utilization of the capital equipment, a proportional change in capital stock corresponds to any change in output, that is,

KtYt=α;Kt=αΔYt.(45)

The actual capital stock becomes a function of past levels of desired capital, which depend on the past level of output:

netIt=(1λ)(α0tKt1).(46)

Gross investment is obtained by adding the depreciation, which is assumed to be a constant proportion, δ, of previous period capital, to net investment:

It=(αYtKt1)+δKt1(47)

To avoid data availability problems connected with the capital stock, the equation can be transformed as follows:

It(1δ)It1=(1λ)[Kt*(1δ)Kt1*](1λδ)[Kt1(1δ)Kt2](48)
It(1δ)It1=(1λ)[kt*(1δ)Kt1*](1λδ)It1(49)
It=(1λ)Kt*(1λ)(1δ)Kt1*+λIt1(50)
It=(1λ)αYt(1λ)(1δ)αYt1+λIt1(51)

The ratio of investment to GDP was estimated, giving the equation the following form:

ItYt=(1λ)αYt(1δ)αYt1Yt+λIt1Yt.(52)

In the framework of the flexible accelerator, the adjustment coefficient is assumed to vary systematically with underlying economic conditions, like the stage of the cycle and the cost and availability of financial resources. In the estimates, the cyclical component was approximated by GDP growth, and several financial variables were taken into account: it is well known that the levels of current and past profits and the rate of return play a crucial role in determining the desired level of capital stock.

The availability of financial resources plays a major role compared to their cost because of the serious limits of financial markets in developing countries. The three main sources of finance for private investment in developing countries are retained profits, the flow of domestic credit to the private sector, and foreign loans. Accordingly, private domestic credit and private long-term foreign loans, together with two series of real interest rates, were taken into account in the estimates as variables affecting λ. 34/

Finally, government investment is included as an explanatory variable to test the hypothesis of crowding out against the hypothesis of complementarity of public and private investment. The insertion of this variable has proved to be crucial because it represents the only direct effect of fiscal policy over investments. In fact, the more institutional channel of influence of fiscal policy—that is, through taxes on profits—is ruled out in this case. Moreover, government investment in developing countries is largely financed by foreign loans and grants, which bring into the picture the availability of foreign capital.

The results of the equations estimated reject the crowding out of private investment by government investment: the coefficient of government capital expenditure is larger than 1, at 1 percent level of significance, thus showing a multiplicative effect on private investments. The stock adjustment model confirms its explanatory power with the coefficient of the one-year-lagged investment consistently above 0.5 at 1 percent level of significance. The real GDP growth coefficient is significant and presents the expected positive sign.

Domestic costs of borrowing are very difficult to measure because of administered interest rates and selective credit policies; two series of interest rates have been used in the estimates—a real rate on deposits, RIR, and an actual rate of interest on foreign debt, RF1R, to approximate the world interest rate (actual i+ep). However, introducing RIR and RFIR will not change substantially the results and both show insignificant t tests (-1.61 and 1.17, respectively). It is worth noticing that the real deposit rate appears with the expected negative sign, the real foreign interest rate shows a positive sign. Even more disappointing are the effects of financial variables representing the quantity of financial resources made available for investment finance. 35/ This is due to the lack of data on short-term loans, which usually go to the private sector. Still, credit availability influences total investment through government investment, because public investment in developing countries largely corresponds to the sum of foreign grants and loans.

d. The saving equation

The saving equations estimated here draw from equation (35). A number of additional variables have been included in an attempt to capture life-cycle aspects of the saving decisions. The specification also reflects elements of a partial adjustment process to take into account the presence of habit formation in savings behavior.

If the probability of death, w, is zero, the present value of future taxes should equal the market value of government debt. In this case, any increase in debt should be offset by an equivalent increase in wealth. A serious problem facing anyone trying to test the above propositions for the developing countries is the absence of a reliable series for financial wealth of the household. In fact, the only available proxy for household wealth is the government debt. Thus, the proposition that government bonds are not net wealth cannot be tested, except through the indirect effect upon private savings. The same relationship, however, should hold—in the case of infinite lives—between fiscal deficit and private saving: the coefficient of the deficit in the private saving equation should be positive and equal to 1.

This hypothesis does not appear to be supported by the estimates that yield 0.16 for the coefficient of fiscal deficit: an increase in expenditure financed by taxes will, therefore, increase the propensity to save, compared with a debt-financed deficit. The latter discourages saving effectively because the government offers better terms of trade between current and future consumption than do the financial markets. Hence, the coefficient of the deficit should be positive, because agents still expect future taxes to increase, but less than 1, with the degree of offsetting depending, among other things, on the age structure of the population.

The coefficient of the ratio of taxes to GDP, which measures the influence of disposable income on private savings, has the expected negative sign. 36/ But, even more important is that the coefficient of the tax variable in the saving equation shows that an increase of 1 percent in taxes will reduce savings by 0.67: it is interesting to note that this coefficient, derived from actual data, lies between the parameters of 1 and 1/2 proposed by Barro and Blanchard, respectively. Barro deduces his parameter directly from his theoretical model; Blanchard’s is derived from a separate study by Hayashi. This result rejects Barro’s neutrality, leaving room for fiscal policy to affect the current account of the balance of payments through its effects on private saving.

A test was also included for money illusion by agents. From equation (d.3) of the Summary Table, it was found that the coefficient of capital losses on debt, owing to the inflation rate, more than offset the coefficient of nominal interest payments; if agents have a target wealth, one should expect an opposite sign and similar magnitude of the two coefficients. Moreover, nominal interest payments are already included in the disposable income and the current government savings so that one should expect a coefficient close to zero. The uncertainty deriving from high inflation, however, also justifies a negative effect on savings, partly explaining capital flight. The inflation tax calculated on government debt, 37/ therefore, becomes an additional important explanatory variable, but further testing was felt to be advisable given that half of the countries for which the domestic debt series was available suffered from hyperinflation and one may assume that some considerable degree of indexation was present. To cope with this problem, an inflation tax measured on the stock of money was included as a variable, but the resulting coefficient proved to be insignificant. This may be determined by the very well-known phenomenon of currency substitution and reduction in money balances holding in high inflation countries, where, on the other side, any lack of perfect indexation of government bonds can give rise to substantial capital losses.

Equation (d.l) of the Summary Table was chosen for several reasons. The positive coefficient on government saving accorded well with the implications of the life-cycle hypothesis, and the estimates showed it to be closer tc zero than to one, as recently suggested by Modigliani. 38/ The t-test showed that the coefficient of government current spending, which was tried in order to test for the substitutability of private sector consumption and government spending, was not significant. Real GDP growth turns out to have a positive, and highly significant, coefficient that agrees well with the predictions of the life-cycle hypothesis, which states that the wealth/income ratio is a decreasing function of the growth rate and that “between countries with identical individual behavior, the aggregate saving rate will be higher, the higher the long-term growth rate of the economy.” 39/ Finally, the lagged dependent variable shows the expected positive sign and carries some explanatory power, tending to confirm the presence of past habit formation in saving behavior. 40/

The unsatisfactory results of population and financial variables should be pointed out. The ratio of the population over 65 years old did not show a significant coefficient. The same applies to the real rate on deposits and to total—domestic and foreign—government debt which has been inserted in the equation.

e. The current account equation

The current account balance is therefore determined making use of the National Accounts identity, CA = (CR - CE) + (SP - I), and substituting the estimated values of the equations discussed above. 41/

CA¯=.04GDPPC+.15XZ¯+.22CRt1¯.02RFIR.06RIR.36CEt1¯.02INF.67CR¯1.34INFTAX+.16SG¯+.11RGDPG+.19St1P¯1.26Ig¯.55It1¯.

The inflation tax and government capital expenditure affect negatively the two important determinants of the short-run movements of the current account, because the former affects private saving negatively and the latter exerts a multiplicative effect on investment. These results support the view that it is capital expenditure and the way in which the deficit is financed through taxes, debt, and its degree of monetization that influence the current account of the balance of payments, rather than the balance of current government spending.

Among the financial variables determining the current account, any increase in the real domestic and foreign interest rates appears to worsen the current account, through its effect on current government expenditure. In the short term also, the availability of financial resources from abroad worsens the current account, raising investment through government capital expenditure.

The present estimates confirm the role of imperfections in capital markets in the determination of investments in developing countries and the low elasticity of current revenue to GDP growth shown by previous studies.

Finally, the lagged variables for saving, current revenue, and current expenditure, illustrate the important role played by history and the institutional and administrative framework (though the structure of the population does not influence the results as was expected).

IV. Concluding Remarks

The empirical results show that fiscal choices relating to the composition of public expenditure and the structure of taxation have crucial consequences for the current account of the balance of payments. In particular, the inflation tax appears to have a large negative impact on private savings and, hence, on the current account. This effect could well explain the capital flight experienced in Latin American countries, which combine an unsustainable fiscal stance with large foreign and domestic official debt. 42/

One may note that the failure of debt neutrality, shown by the results of the empirical estimates, creates a critical role for fiscal policy since a change in the taxation-borrowing mix (for given government expenditure) appears to have a major influence on the current account through its effect on savings. The absence of debt neutrality does not appear, however, to imply any significant financial crowding out of private investment., since investment seems to be largely independent of foreign and domestic real interest rates, as is private saving. However, interest rates do contribute to the decrease of public saving through their effect on current government expenditure.

Government capital expenditure has a crowding-in effect on private investment (providing an additional increase of 1.26 in total investment for every unit increase of public capital formation). But, since it also increases absorption, it will tend to worsen the current account, other things being equal. However, this would tend to be offset by the rise in profits, which, by increasing private savings, would improve the current account. Therefore, if projects show adequate returns, the initial negative position of the current account will be sustainable.

It would be interesting to see whether the results presented here for Latin American countries were true for other geographical areas. In any case, further research might include the study of the effect of terms of trade changes, by including the distinction between tradable and nontradable goods in the model.