Staff Papers

STAFF PAPERS

PETER HOLE Chair, Editorial Committee IAN S. MCDONALD Editor and Deputy Chair THOMAS WALTER Assistant Editor Editorial Committee William E. Alexander Paul R. Masson F. Charles Adams Donald J. Mathieson David Burton Guy M. Meredith Peter Isard Ratna Sahay G. Russell Kincaid Subhash M. Thakur Malcolm D. Knight Howell H. Zee Jorge Marquez-Ruarte

Among the responsibilities of the International Monetary Fund, as set forth in its Articles of Agreement, is the obligation to "act as a centre for the collection and exchange of information on monetary and financial problems." Staff Papers makes available to a wider audience papers prepared by the members of the Fund staff. The views presented in the papers are those of the authors and are not to be interpreted as necessarily indicating the position of the Executive Board or of the Fund. To facilitate electronic storage and retrieval of bibliographic data, Staff Papers has adopted the subject classification scheme developed by the Journal of Economic Literature.

Subscription: US$54.00 a volume or the approximate equivalent in the currencies of most countries. Four numbers constitute a volume. Single copies may be purchased at $18.00. Individual academic rate to full-time professors and students of universities and colleges: $27 a volume. Subscriptions and orders should be sent to: International Monetary Fund Publication Services 700 19th Street, N.W. Washington, D.C. 20431, U.S.A. Telephone: (202) 623-7430 Telefax: (202) 623-7201 E-mail: [email protected] Internet: http://www.imf.org

©International Monetary Fund. Not for Redistribution INTERNATIONAL MONETARY FUND STAFF PAPERS

Vol. 44 No. 2 JUNE 1997

©International Monetary Fund. Not for Redistribution EDITOR'S NOTE The Editor invites from contributors outside the Fund brief comments (not more than 1,000 words) on published articles in Staff Papers. These comments should be addressed to the Editor, who will forward them to the author of the original article for reply. Both the comments and the reply will be considered for publication.

The term "country," as used in this publication, may not refer to a territorial entity that is a state as understood by international law and practice; the term may also cover some territorial entities that are not states but for which statistical data are maintained and provided internationally on a separate and independent basis.

The U.S. Library of Congress has cataloged this serial publication as follows: International Monetary Fund Staff papers — International Monetary Fund. v. 1- Feb. 1950— [Washington] International Monetary Fund. v. tables, diagrs. 23 cm. Three no. a year, 1950-1977; four no. a year. 1978— Indexes: Vols. 1-27, 1950-80, 1 v. ISSN 0020-8027 = Staff papers — International Monetary Fund. 1. Foreign exchange—Periodicals. 2. Commerce—Periodicals. 3. Currency question—Periodicals. HG3810.15 332.082 53-35483

©International Monetary Fund. Not for Redistribution CONTENTS

Vol. 44 No. 2 June 1997

Pension Reform, Financial Market Development, and Economic Growth: Preliminary Evidence from Chile ROBERT HOLZMANN • 149

Fiscal Policy and Long-Run Growth VITO TANZI and HOWELL H. ZEE • 179

Fiscal Adjustments in OECD Countries: Composition and Macroeconomic Effects ALBERTO ALESINA and ROBERTO PEROTTI • 210

Is the Phillips Curve Really a Curve? Some Evidence for Canada, the United Kingdom, and the United States GUY DEBELLE and DOUGLAS LAXTON • 249

Comments Internal Migration, Center-State Grants, and Economic Growth in the State of India: A Comment on Cashin and Sahay M. GOVINDA RAO and KUNAL SEN • 283 Internal Migration, Center-State Grants, and Economic Growth in the State of India: A Reply to Rao and Sen

PAUL CASHIN and RATNA SAHAY • 289

IMF Working Papers • 292

IMF Papers on Policy Analysis and Assessment • 294

iii

©International Monetary Fund. Not for Redistribution This page intentionally left blank

Pension Reform, Financial Market Development, and Economic Growth: Preliminary Evidence from Chile

ROBERT HOLZMANN*

The Chilean pension reform of 1981, a shift from an unfunded to a funded scheme, is considered to have contributed to this country's excellent economic performance. Positive growth effects allow, in principle, a Pareto- improving shift in pension financing. This paper highlights the theoretical underpinnings of the reform and presents empirical data and preliminary econometric testing of the conjectured reform effects on financial market developments, as well as the impact on total factor productivity, capital formation, and private saving. The empirical evidence is consistent with most but not all claims. In particular, the direct impact of the reform on saving was low, and initially even negative. [JEL G23,O16,O46,O54]

HE REFORM of the public retirement scheme is a standing agenda in Tessentially all countries throughout the world (see Holzmann (1988) and World Bank (1994)). It is of particular importance for the emerging economies of Latin America and Central and Eastern Europe as their current schemes constitute an important drain on the public budget, reducing

* Robert Holzmann is Professor of Economics and Managing Director of the Economics Department of the European Institute at the University of Saarland. He is currently on leave and Director of the Social Protection Team of the Human Devel- opment Department of the World Bank. The first version of this paper was prepared while he was a guest professor at the Instituto de Economia, Pontificia Universidad Catolica de Chile in March-April 1995, and a revised version was prepared while he was a visiting scholar at the Fiscal Affairs Department of the IMF in June-August 1995. The author thanks both institutions for their encouragement and support, and the possibility to present the results at their research seminars, where he received many valuable comments. Special thanks for helpful comments and support go to Friedrich Breyer, Peter Diamond, Philip Gerson, Ross Levine, Sandy Mackenzie, and Salvador Valdes-Prieto. All errors and omissions are, of course, his own doing. 149

©International Monetary Fund. Not for Redistribution 150 ROBERT HOLZMANN national saving, capital formation, and growth. Also, these schemes are held responsible for the main distortions in the labor market. Thus the experience of the Chilean pension reform of 1981 has special attraction for other emerging economies, as this reform was thought to have contributed to the country's excellent economic performance after the shocks of the foreign debt and domestic banking crises of 1981-83 had been absorbed (see Table 1). Very recently, various countries in Latin America, such as Argentina, Peru, Colombia, and Mexico, have begun to imitate to some extent the Chilean approach, and some former centrally planned economies, such as Hungary, Latvia, and Poland, are taking preparatory steps in this direction.1 In a nutshell, the Chilean reform consisted of (1) a shift from a conventional unfunded and defined-benefit plan to a funded, defined-contribution plan, (2) the replacement of public administration of the program with private administration by competing pension funds, and (3) the separation of the social assistance element from the mandated saving element of retirement provisions. The government is still involved with supervising and regulating the new mandatory but funded scheme, guaranteeing minimum benefits, and financing the transition; otherwise, the market is allowed to play its role.2 Various advantages for the reform approach are claimed by domestic and foreign observers. At the political level, three effects stand out. First, the approach provides a break in the deadlock of traditional reform attempts since it suggests a time-consistent and, hence, credible reform (Holzmann, 1994). Second, the approach isolates retirement provisions to a large extent from political interference and risk (Diamond, 1997; and Godoy-Arcaya and Valdes-Prieto, 1997). Last but not least, it sensitizes workers to financial issues and enterprise performance, thus reducing the dichotomy between capital and labor (Pinera, 1991). At the economic level, three main reform effects are claimed. First, the reform establishes a close link between contributions and benefits, thus reducing the labor market distortions with which traditional unfunded programs are considered to be fraught (World Bank, 1994). Second, the reform approach furthers and accelerates financial market developments and thus efficiency of resource allocation. Last but not least, the reform affects positively national saving and capital accumulation and, hence, contributes to economic growth (IMF, 1995).

1 For a survey of the Latin American pension reform attempts and further references, see Queisser (1995); for a review of the current reform approaches in Cen- tral and Eastern Europe, see Holzmann (1997). 2 For a detailed survey and analysis of the Chilean pension reform and further references, see Diamond and Valdes-Prieto (1994).

©International Monetary Fund. Not for Redistribution Table 1. Macroeconomic Indicators and Pension Fund Performance (In percent, unless otherwise indicated)

Macroeconomic indicators Pension fund performance AFP assets (in % of market assets) Real GDP Unemploy- Real exchange Private Real rate of AFP assets Enterprise bonds and Years growth Inflationa ment rateb rate (1977=100)c saving rated return (in % of GDP) commercial papers Shares

1970 2.1 34.9 5.7 48.5 8.9 1971 9.0 22.1 3.9 45.2 11.0 — — — — 1972 -1.2 163.3 3.3 41.9 12.2 — — — — 1973 -5.6 508.4 5.0 62.7 7.7 — — 1974 1.0 375.9 9.5 95.0 17.6 — . 1975 -13.3 340.7 14.9 123.8 -0.6 1976 3.2 174.3 12.7 111.4 9.0 1977 8.3 63.5 11.8 100.0 6.7 — — — — 1978 7.8 30.3 14.2 119.3 7.7 — — — — 1979 7.1 38.9 13.6 122.9 6.1 — — 1980 7.7 31.2 10.4 106.5 6.5 1981 6.7 9.5 11.3 92.6 2.7 21.3 0.9 0.4 1982 -13.4 20.7 19.6 103.3 4.1 28.8 3.6 0.9 1983 -3.5 23.1 14.6 124.0 7.2 21.3 6.4 8.4 — 1984 6.1 23.0 13.9 129.6 4.9 3.5 8.6 10.2 — 1985 3.5 26.4 12.0 159.2 8.7 13.4 10.6 7.7 1986 5.6 17.4 8.8 175.1 10.7 12.3 12.7 11.1 2.0 1987 6.6 21.5 7.9 182.7 15.2 5.4 14.2 27.1 3.2 1988 7.3 12.7 6.3 194.7 17.2, 6.4 15.1 46.0 4.2 1989 9.9 21.4 5.3 190.2 16.5' 6.9 17.7 47.5 4.8 1990 3.3 27.3 5.7 197.4 19.4 11.5 24.3 58.2 5.5 1991 7.3 18.7 5.3 186.3 19.9 29.7 30.4 60.8 8.3 1992 1 1.0 12.7 4.4 165.4 19.4 3.1 30.6 61.0 10.1 1993 6.3 12.2 4.5 168.7 19.2 16.2 37.3 56.7 11.4 1994 4.2 8.9 5.9 160.5 20.7 18.2 41.1 57.2 10.6 1995 8.5 8.2 4.7 156.9 20.8 -2.5 38.8 55.3 10.3 Sources: Central Bank of Chile, Monthly Bulletin; and Superintendency of Pension Fund Administration, Statistical Bulletin. a Based on consumer price index: 12 months ended December. b Based on October-December data. c An increase indicates a real depreciation of the domestic currency. d Private saving in percent of GDP.

©International Monetary Fund. Not for Redistribution 152 ROBERT HOLZMANN

The investigation of these economic effects and first attempts at empirical testing are the core of this paper. If confirmed, the positive effects of such a pension reform approach would go far beyond the successful restructuring of a major public expenditure program, as it would allow the economy of a reforming country to be put permanently on a higher growth path than would otherwise be possible. Placing the reform effects within the framework of endogenous growth theory would also allow, in principle, traditional arguments against a shift from an unfunded to a funded scheme to be overcome since the transition generation would not necessarily be burdened twice. Yet even transitory growth effects of relevant size may substantially reduce or even eliminate the transition costs. To analyze these hypotheses, the paper has been structured as follows. Section I presents the theoretical underpinnings of the claimed welfare economics effects of the pension reform approach, stressing the importance of externalities and putting the approach within the framework of modern endogenous growth theory and a specific, testable model. Section II outlines the research strategy and discusses the problems of empirical testing posed by the short time span of pension reform and the multiple economic reforms. Section III presents empirical data and preliminary econometric testing of the effects claimed for Chile on total factor productivity, capital formation, and saving. The empirical results are, in general, consistent with most but not all hypotheses and underscore the importance of a sound fiscal policy to support such a reform. Some tentative conclusions are drawn in Section IV.

I. Theoretical Background

Shifting from an unfunded to a funded scheme raises the issue of the repayment of the implicit debt of the existing pension scheme and the bur- dening of the transition generation. For this reason, countries have generally rejected this reform option.3 However, the welfare economics issue of a "Pareto-improving transition"—which makes at least one generation better and no other worse off—receives a different assessment once the economic externalities of the reform are taken into account, and the case for reform can be strengthened if these effects are embedded in the theory of endogenous growth. This section outlines these considerations.

3 Merely changing the financing mechanism of an unsustainable, unfunded retirement scheme is not sufficient to put it on a sustainable, funded basis. Any real pension reform essentially has to undertake two changes simultaneously: reduce the commitment of the given pension scheme (given target income replacement rates, essentially by increasing the retirement age) and shift the financing mechanism.

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 153

Pareto-Improving Transition and Economic Externalities

In the conventional world of neoclassical economics, an unfunded pension scheme is Pareto efficient even when the interest rate permanently exceeds the natural growth rate if the scheme does not create economic distortions; for example, if it is financed via lump-sum taxes and provides lump-sum transfers. Although only the first generation gains and all later generations are worse off, there exists no mechanism to reverse the situation without the welfare position of at least one generation deteriorating (Breyer, 1989). The result is intuitively and immediately understandable, as it amounts to an application of the second basic theorem of welfare economics: any lump-sum redistribution of income entails an allocation that is different but also Pareto efficient (Homburg, 1990). Consequently, a Pareto-improving transition requires either that the fully- funded scheme exhibit fewer distortions than the unfunded scheme—for example, if negative externalities are reduced—or that the fully funded scheme introduce positive externalities, such as a shifting outside the intertemporal budget constraint of the economy. These externalities may result from special growth effects.

The Case of Lower Negative Externalities Lower negative externalities can be motivated by eliminating the many distortions that an unfunded scheme may exert on intertemporal consumption or on labor supply decisions, resulting in an excess burden. By shifting from an unfunded to a funded scheme, the reduction or elimination of the excess burden may be used to repay the implicit debt of an unfunded scheme within finite time (Homburg, 1990). Because public pension schemes and the way that they are financed entail numerous distortions, a change in the funding mechanism may thus actually improve welfare and, further on, may diminish the impact of population aging. The conclusion rests, however, on the assumption that the funded scheme is less distortionary for individual saving decisions and labor supply than the unfunded one. Nevertheless, such a result is not necessarily linked with the funding procedure; under the assumption of elastic labor supply, it is typically related to the inadequate link between contributions and benefits in unfunded schemes. Public and earnings-related pension schemes traditionally have a distributional and annuity component, and it is the mingling of both components and the lack of a clear link between contributions and benefits that is claimed to be responsible for the distortions (Schmidt-Hebbel, 1993; and World Bank, 1994). These distortions may also be reduced in an unfunded scheme by separating both components more clearly. For example,

©International Monetary Fund. Not for Redistribution 154 ROBERT HOLZMANN a two-tier scheme can be set up in which a basic tax-financed, flat-rate scheme takes care of distributional and poverty considerations, while a fully earnings-related one, financed by earmarked contributions only, takes care of income replacement considerations. However, it remains unclear at the theoretical level whether such a separation always creates fewer distortions than a well-conceived, traditional social insurance scheme. The basic poverty-oriented component will exist in any alternative scheme, and the incurred distortions are the inevitable consequence of introducing distributional activities. The remaining distortions in the fully earnings-related scheme are reduced to the effects of the funding mechanism. These effects may exist because a nondistortionary pension scheme requires actuarial neutrality, which can be achieved in an unfunded scheme only if the implicit rate of return (the natural rate of growth) equals the rate of interest (that is, if the golden rule of growth holds (Breyer and Straub, 1993)). Put differently, an unfunded pension system will always entail a net tax on labor if the system's implicit rate of return is below market, and the implicit tax burden may be sizable (Valdes-Prieto, 1997). Even if the funded scheme were nondistortionary compared with the unfunded one, new distortions may be introduced through the transition and the financing of the now explicit social security debt. To keep the ratio of debt to GDP constant, the interest rate-economic growth rate differential of this debt has to be financed via taxes, and a new excess burden may emerge. However, as the higher rate of return allows a lower contribution rate to pay the same benefit, a new wage tax may be introduced to keep the take-home pay constant and limit the new explicit debt. Finally, the decision to shift from an unfunded to a funded scheme is also a question of the scope of the expected welfare gains. Simulation studies with overlapping generation models a la Auerbach and Kotlikoff (1987) suggest that the welfare gains resulting from the elimination of labor market distortions are comparatively small. A model calibrated on the German pension system exhibits long-term welfare gains of some 9 percent of lifetime resources of future generations if the transition generation is not compen- sated. With compensation, the long-term welfare gains are reduced to some 1.5-2 percent of lifetime resources of future generations (Raffelhuschen, 1993). Simulations by Kotlikoff (1996) provide larger welfare gains to future generations of 4.5 percent (while compensating the transition generation) when it is assumed that the tax-benefits linkage is weak, that the initial tax structure features a progressive income tax, and that consumption tax is used to finance the transition. However, when the initial tax structure is a proportionate income tax, the tax-benefits linkage is strong, and income taxes are raised to finance the transition, there is a 3.1 percent welfare loss to future generations. Other studies exhibit similar small and distant welfare

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 155

gains. Long-term welfare effects of some 1.5-4.5 percent under a potential Pareto-improving transition are likely to be too small to convince politicians to undertake a transition technically and politically difficult when potential gains emerge only in the distant future.

The Case of Positive Externalities The other argument for a shift toward (fully or partially) funded pensions is based on positive externalities resulting from a change in the financing mechanism. In the framework of traditional neoclassical theory (that is, exogenous technical progress and, thus, a given intertemporal budget constraint), welfare gains can be derived from portfolio considerations. It can be argued that the internal rate of return of an unfunded scheme—the natural growth rate—is a stochastic variable that exposes each pension cohort to an income risk. The same can be claimed for the internal rate of return of a funded scheme—the interest rate. Thus, if the covariance of both returns is lower than one, a mixed financing mechanism reduces the overall income risk and provides positive welfare effects. Potentially larger welfare effects can be achieved by changing the financing mechanism, which can shift outward the intertemporal budget constraint. Such a shift can be derived from endogenous growth considerations with regard to labor market and financial market externalities. At the labor market level, the type of pension scheme (unfunded or funded) and the perceived link between contributions and benefits can determine the distribution of labor supply between the formal and informal sectors. If the latter is less productive, a pension reform that moves labor supply to the formal sector will enhance overall productivity and can in an endogenous growth model lead to a higher growth path (Corsetti, 1994). At the financial market level, a shift from an unfunded to a funded scheme can promote the development of financial markets by making them deeper, more liquid, and more competitive. The resulting enhanced resource allocation may lead not only to a onetime efficiency gain but also to a permanently higher growth path (Holzmann, 1994).

Financial Market Intermediation and Endogenous Growth

The modeling of financial markets and investigation of their welfare economics and growth implications are still in their infancy. The claim that the effectiveness of financial markets and the level (or rate of growth) of real activity are closely related, however, is not new; empirical investigations have been undertaken for decades (for example, Goldsmith, 1969). Against

©International Monetary Fund. Not for Redistribution 156 ROBERT HOLZMANN the background of neoclassical growth theory, however, these studies could argue only for temporary efficiency effects resulting from financial market developments. More recent developments in growth theory allow for level as well as growth-path effects, but these models concentrate on specific aspects of financial markets and their impact on real activity. In one model, for example, financial markets provide liquidity, allowing a shift from current liquid but unproductive assets toward less liquid but more productive assets (for example, Bencivenga, Smith, and Starr, 1996). In another model, financial markets promote the acquisition and the dissemination of information, thus allowing for better resource and risk allocation (for example, Greenwood and Jovanovic, 1990). In a third model, financial markets permit agents to increase specialization, shifting away from less specialized and inefficient technologies (for example, Saint-Paul, 1992). All these models cover important aspects of financial markets and their impact on real activity, providing important analytical insight on issues raised in the literature for decades. However, they all fall short of providing a comprehensive framework of the different effects of financial markets and of empirically testable relationships. Establishment of this framework still awaits future work. To introduce potential growth effects of financial market developments into an endogenous growth model in a simple but testable manner, the following structure is proposed, borrowing from Villanueva (1993):4 dK/dt = s(K, ...)Y- SK, with dsf 3K > 0, and (1) dT/dt = a(K, . . .)K/L + XT, with a > 0. (2)

4The other four equations of this growth model are traditional. The first specifies the output, Y, via a production function with constant returns to scale on capital, K, and labor, N (man-hours in efficiency units):

The second equation specifies an exogenous growth rate, n, of total employment (in man-hours, L): dL/dt = nL. The third specifies a definition equation between N and L via the technical- change multiplier, T: N = TL. The fourth equation defines the capital coefficient: k = K/N. d{.)ldt is the time derivative, and 5 is the rate of depreciation of capital.

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 157

The saving ratio (that is, the investment ratio in a closed economy) is positively related to a variable measuring the depth, liquidity, and maturity of financial markets, summarized in the parameter K. Further variables that may influence the national saving rate are public saving behavior or tax regulations. Also, the change in technical progress, dT/dt, is dependent not only on the exogenously given rate of labor-augmenting technical change, X, but also on an efficiency variable, a, that interacts multiplicatively with the ratio of capital to labor. a(K, . . .) depends on the financial market variable, K, and also on other variables traditionally quoted in the literature (such as the level of export orientation and the share of education expenditure in the budget). X captures other growth effects not explicitly detailed in the model.5 In this model, the steady state growth rate of the economy depends positively on the level of K: [(dY/dt)/Y]* = a(K, . . .)k* + X + n = g* (k*), (3) with k*, the steady state capital intensity, measured in efficiency units of labor. The model leads to the traditional result for a = 0. With a > 0, however, a higher saving rate leads not only to an increase in the optimal capital-labor ratio (as in the traditional growth models) but also to a higher steady state growth rate, which in traditional models is not influenced by the saving rate. A further important property of the model under an optimal consumption plan (that is, dc*/ds = 0) is that both the steady state growth rate and the optimal net return on capital are higher than the exogenous rates of technical progress and population growth:

df(k*)/dk* - 5 = g* (k*) + a(K, . . ,)k* = X + n + 2OC(K, . . ,)k*. (4) Hence, compensating the transition generation by the conventional rate of return of an unfunded scheme only (that is, by X + n, assuming that a was zero prior to the shift to a funded scheme) while using part of the growth differential (up to 2a (K, . . .) /:*) to finance the transition allows, in principle, for the construction of a Pareto-improving transition to a funded scheme.6 One approach could be to pay wages (and pensions) to individuals according to the old growth path until the growth differential allowed the repayment of the implicit debt. As the marginal product of each worker increases at the same rate as his efficiency, g* (k*)—n, capturing this growth

5As the model features an external effect, the solution of the social planner's problem will not necessarily coincide with the competitive equilibrium in a decentralized economy. 6In the decentralized solution, the net rate of return with externalities is f '(k*) - 5 = X + n + a (K, . . .) k*, still leaving a growth differential of a (K, . . .) k* to compensate the transition generation.

©International Monetary Fund. Not for Redistribution 158 ROBERT HOLZMANN differential requires the use of lump-sum taxation to ensure Pareto indif- ference for the transition generation until the social security debt is repaid.

II. Testing the Impact: Conceptual Considerations and Data

Establishing statistically significant links between pension reform, the financial market variable, K, the endogenous variable of technological progress, a (K, . . .), and the saving/investment rate, S(K, . . .), poses at least four main problems: • The multiplicity of reforms undertaken makes it difficult to separate the influence of competing explanations for growth developments, such as strengthened macroeconomic stability and enhanced export orientation. • It is likely that financial market development is only a necessary but not a sufficient condition for higher growth, to which pension reform may contribute via acceleration. • Currently, there is no established methodology to differentiate between (conditional) convergence and transitory and endogenous growth effects. • There are data problems, including the length of usable time series. The Chilean economic data are in better shape than most other emerging markets, but this qualification is true only for the most recent data. Against this background, a research strategy is applied that should strengthen confidence in the hypothesis advanced of the effects of pension reform on financial market development and growth, although a watertight proof may not exist. The strategy consists of three elements. First, appropriate indicators of financial market developments are constructed to determine the impact of pension fund activities, and the relationship between those indicators and a and s (technological progress and the saving/investment rate) is empirically tested. Pension funds should contribute to deeper financial markets, higher liquidity, enhanced competition, and better risk allocation in financial markets. The financial market indicators should be able to measure these effects. The financial interrelation ratio (FIR) compares the range of financial instruments with the net wealth of the economy (approximated by the capital stock). The financial intermediation ratio (FMR) compares the scope of financial instruments with the assets of the financial institutions. Two alternative measures of financial instruments—and thus indices—are considered. FIR-] and FMR-1 follow an instrument concept, as used in some Chilean studies, to present the scope and structure of financial markets (De la Cuadra and Valdes-Prieto, 1992). These instruments measure only financial liabilities, enhanced by the assets (-liabilities) of pension funds,

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 159 mutual funds, and insurance companies. FIR-2 and FMR-2 follow the more traditional approach of measuring financial market instruments (as pio- neered by Goldsmith, 1969). These instruments cover both the asset and liability sides of the financial market and thus deliberate double count, but they are restricted to claims to the financial and nonfinancial sectors plus equity holdings. Traditionally, both indices have been rising with economic development, leveling off only at an advanced stage of an economy (Goldsmith, 1969). A stock market index, the average of three stock market indicators, is intended to capture the impact of that market on economic activity (Levine and Zervos, 1996). The market capitalization ratio equals the value of listed shares divided by GDP. It is assumed that the overall stock market size is positively correlated with the ability to mobilize capital and to diversify risk on an economy-wide basis. The total value-traded ratio equals the total value of shares traded on the stock market exchange divided by GDP. The turnover should reflect liquidity on an economy-wide basis. The turnover ratio equals the value of total shares divided by market capitalization. Since the market capitalization ratio may be high and the value-traded ratio low, or vice versa, the ratio of both—that is, the turnover ratio—complements the prior indicators. Indicators for competitiveness and risk allocation may be derived by asset mispricing, which is the calculation of the systematic deviations of actual returns and those implied by reference models: the capital assets pricing model and the arbitrage pricing model (Korajczyk, 1996). The hypothesis is that rising pension fund activities should make the financial system more efficient, thus contributing to a reduction in the mismatch between the actual returns and those implied by reference models. Second, cross-country results investigating the impact of financial market developments on economic growth, capital accumulation, and productivity improvement for 41 countries for the period 1976-93 are compared, using the stock market index approach (Levine and Zervos, 1996). Con- trolling for initial conditions and various economic and political factors, this relationship turns out to be statistically significant and robust; the predictive power of financial market developments in forecasting economic growth over 18 years implies that (1) financial developments do not simply follow economic activity, and (2) the strong link between both does not merely reflect positive correlation among contemporaneous shocks to financial markets, institutions, and economic activity. These results give confidence in the causality of the influence and allow for a comparison with own results with regard to the sign and magnitude of estimated parameters. Third, the claims for this paper's hypothesis are cross-checked using other methods and data. Given the multiplicity of the reforms in Chile and the often nonquantitative nature of the links between pension reform, financial

©International Monetary Fund. Not for Redistribution 160 ROBERT HOLZMANN

market developments, and changes in macroeconomic variables, no full proof of the hypothesis can be expected. However, applying different methods and checking for consistency against other claims should enhance confidence in the results.

III. Preliminary Empirical Results

This section presents empirical evidence and preliminary econometric results on the conjectured link of pension reform with financial market developments and economic growth. The first link is investigated by assessing the impact of the pension fund's activities on financial market indicators, and the second link by assessing the impact of these indicators on total factor productivity, capital formation, and saving.7 The impact of the pension reform on labor market performance was also investigated, and the results are consistent with the claim that a closer link between contributions and benefits promotes formal market participation. However, the quality of the data did not allow econometric testing, and the qualitative results are not reported for reasons of space.

Pension Reform and Financial Market Development

A central claim about the effects of the Chilean pension reform, echoed internationally, is its contribution to the development of the financial sector (IMF, 1995). The general hypothesis is that the rising investment needs of the pension funds, the instruments thereby created, and the competitive setup of the privately managed pension funds made the financial market deeper, more liquid, and more competitive. An inspection of the data and very simple empirical testing seemingly confirm the claims. Essentially, all investigated financial market indicators moved upward strongly once the banking crisis of 1981-83 had been solved (Figures 1 and 2). The FIRs show a strongly rising tendency that well exceeds the level reported for prior decades.8 The FMRs exhibit a similar development, with FMR-2 reflecting the strong increase in bank credits to the private sector that ultimately led to the banking crisis and the consolidation afterward. The

7 The data draw from a wide range of different sources, mainly official but also private. Long-term time series in official publications are often not available and sometimes require index-type linking to overcome structural breaks. 8 No strictly comparable FIRs could be established for the pre- and post-1975 periods. However, the available data for Chile suggest a long-term decrease in FIR-2 from 63 percent (1940) to 32 percent (1950) and 29 percent (1960), with a slight increase to 39 percent (1971) prior to a major shake-up of the economy (Cerda and Zeballos, 1975).

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 161

Figure 1. Financial Market Indicators

Note: FIR-1 and FMR-1 measure financial liabilities, enhanced by the assets (-liabilities) of pension funds, mutual funds, and insurance companies. FIR-2 and FMR-2 measure both financial assets and liabilities, but they are limited to claims to the financial and nonfinancial sectors and equity holdings. almost linear rise in the stock market index started in 1985, which was the year of the first participation of pension fund in stock market activities. Prior to this date, the investment rules permitted only the purchase of debt instruments. At the end of 1994, pension fund assets constituted almost 40 percent of all outstanding financial instruments, and the total pension fund assets at the end of 1994 amounted to 41 percent of GDP (Figure 3). The correlation of pension fund assets and financial market indicators, as well as of pension fund shares in total traded shares and the market capitalization ratio, is very strong, with coefficients in simple regressions close to 1 and R2 of 0.9 and above (not shown). At a monthly level, there is also a strong correlation between the turnover in asset trade (including bonds and shares) and the level of assets held by the pension funds at the end of the month (as a proxy for turnover since no such data are available), with a break occurring about the end of 1984. Before 1985, the correlation is zero or negative, except for the trade in assets with fixed return (p = 0.65); this was the period when pension funds were restricted to the holding of debt instruments. For the period January 1985 to June 1995, the correlation between the monthly

©International Monetary Fund. Not for Redistribution 162 ROBERT HOLZMANN

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 163 turnover in each asset and the stock of pension fund assets at the end of the month was always above 0.9. This empirical evidence is consistent with the claim that pension funds made the financial markets deeper and more liquid. With regard to the contribution of pension funds to enhanced competitiveness and risk allocation, the available data allow only for a very cursory investigation. The indicators of asset mispricing, based on the arbitrage, capital assets, and international assets pricing models, respectively (Korajczyk, 1996), are compared with the indicators of pension fund assets.9 If the pension fund activities improve the performance of the financial markets, the mispricing should decrease with enhanced fund activities. The correlation between the mispricing and pension fund indicators has the expected sign, is statistically significant at the 5 percent error level, and ranges between minus 0.27 and minus 0.52. Pension fund activities are contributing to a more sophisticated financial market, as evidenced by the role of pension funds in the development of financial instruments, such as indexed annuities, the provision of funds to key sectors, such as mortgage bonds to finance housing, the importance of enterprise bonds (which are mainly held by pension funds), and the increased holding of traded shares by the pension funds. With the gradual relaxation of regulations for pension fund investments, their portfolios also have become more diversified (Table 2). Various evidence suggests that pension funds are operating efficiently and that the selected portfolios—given that the restrictions on asset investments are only gradually being lifted—are on the (restricted) efficiency frontier (Walker, 1993a and 1993b). In a competitive environment, this may constitute indirect proof of the overall efficiency of the financial system. Yet all this evidence does not establish watertight proof that the establishment of pension funds has been the decisive factor in the impressive development of financial markets since the mid-1980s. The empirical evidence is only consistent with the claim. The healthy growth and development of financial markets after 1983 may simply reflect changes in legis- lation and the lessons learned from the experiences and mistakes of the late 1970s and early 1980s.10 Since the counterfactual of the development of financial markets without pension reform cannot be established and empirical evidence from other countries with similar reforms is not at hand, it may be impossible to prove the claim. Hence, in order to increase confidence in the claim, further evidence using different approaches is required.

9 These data are used in Levine and Zervos (1996), and the access granted by Ross Levine is gratefully acknowledged. 10 For an analysis of the experience and mistakes of financial market liberalization during this early period of economic reform, see De la Cuadra and Valdes-Prieto (1992).

©International Monetary Fund. Not for Redistribution Table 2. Investment Portfolios of Pension Funds (In percent of total investments, unless otherwise indicated)

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 Financial institutions 71.3 73.4 53.4 56.0 56.3 48.7 49.7 50.1 39.2 33.4 26.8 25.3 20.6 20.1 23.7 Deposits and bank bonds 61.9 26.6 2.7 12.9 20.9 23.2 28.3 29.5 21.5 17.4 13.4 11.1 7.6 6.5 7.5 Mortgage bonds 9.4 46.8 50.7 43.1 35.3 25.5 21.4 20.6 17.7 16.1 13.4 14.2 12.9 13.7 16.2 Public institutions 28.1 26.0 44.5 42.2 42.6 46.7 41.5 35.4 41.6 44.1 38.3 40.9 39.2 39.7 40.5 Central bank 27.4 4.3 14.1 16.6 20.4 26.0 29.8 30.0 38.1 42.5 37.4 40.1 38.7 39.4 40.4 Treasury 0.7 21.7 30.4 25.7 22.2 20.7 11.7 5.4 3.5 1.6 0.9 0.7 0.5 0.2 0.1 Enterprises 0.6 0.6 2 2 1.8 1.1 4.6 8.8 14.5 19.2 22.4 34.9 33.8 39.7 39.3 35.6 Bonds 0.6 0.6 2.2 1.8 1.1 0.8 2.6 6.4 9.1 11.1 11.1 9.6 7.4 6.3 5.4 Shares — — — — — 3.8 6.2 8.1 10.1 11.3 23.8 24.2 32.3 33.0 30.2 > Foreign investments — — — — — — — — — — — — 0.6 0.9 0.2 Total investment (in millions of Chilean pesos) 11.9 44.8 99.9 161.9 281.9 433.9 644.5 891.5 1,334.6 2,249.4 3,778.1 4,744.2 6,579.8 8,997.9 10,074.0 Sources: Superintendency of AFP, Statistical Bulletin; and PrimAmerica Consultatores (1995).

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 165

Total Factor Productivity and Financial Market Development

The model outlined above assumes that the change in labor productivity, dT/dt, is determined by an exogenous component X and a component a (K, . . .) K/L that is positively dependent on the financial market indicator K. In the selected specification, we use total factor productivity, TFP,11 which in a linear approximation should exhibit a positive depen- dence from K:

TFPt =ao+ a1D(L)Kt + a2Xt + ut, (5) with D (L) an appropriately chosen lag structure for K, since the effects can be expected to be distributed over various periods, and X representing other variables able to capture further impacts, the most important of which are cyclical effects via the use of the unemployment rate and its change. The basic idea behind this specification of TFP—the Solow residual—is that it reflects technological shocks following (perhaps) a stationary autoregressive process (hence, the alternative inclusion of the lagged TFP). Including the unemployment rate and its change is an (imperfect) attempt to capture the cyclical development during the period under investigation, keeping in mind the high likelihood that the stock of factors (capital and labor) does not correctly represent the factor services actually provided. In the steady state, the term with the (now constant) unemployment rate col- lapes with the constant, and the impact of the change in the unemployment rate disappears. This simple specification (which also takes account of the limited number of observations) leads to a satisfying statistical fit for the reduced sample period (Table 3).12 In accordance with the endogenous

11 To estimate TFP, a simple growth-accounting exercise is undertaken (with log- arithmic approximation to account for discrete time and a constant labor share of 0.65). Admittedly, this estimation ignores its well-known limitations, such as the quality of factor inputs and the time-varying factor shares, because of data constraints. 12 For definitions of the variables used in the regressions in Tables 3-5, see notes to Table 3. All variables in Tables 3-5 have been subject to a unit root test (augmented Dickey-Fuller test), which has been passed for all endogenous and most exogenous variables at the 5 percent confidence interval. For the unemployment rate, the unit root could not be rejected. This outcome may cast doubt on the estimated parameters and their statistical significance; although the stability of the parameters under alternative measures of financialmarket s and the Durbin-Watson test statistics of about 2 may give confidence in the results, further investigation is certainly required. The parameter estimates for the AFIR and ASMI variables differ slightly from previously presented estimates, owing to data revisions and the normalization of the respective sample average to one. The normalization allows for a direct parameter comparison between different financial market indicators and a straightforward interpretation of the parameter value.

©International Monetary Fund. Not for Redistribution 166 ROBERT HOLZMANN

Table 3. Total Factor Productivity and Financial Market Developments TFP UER AUER FMI Equation Constant (-0 (-2) (-0 (la) 0.049 -0.512 -0.579 R2 = 0.858 (4.37) (4.76) (4.40) DW = 2.21 (lb) 0.055 -0.125 -0.566 -0.569 R2 = 0.871 (4.38) (1.66) (4.77) (4.42) DW= 1.77 (2a) 0.021 -0.331 0.661 0.011 R2 = 0.891 [FMI(-1) = AFIR-1] (1.16) (2.39) (5.16) (1.85) DW = 2.19 (2b) 0.025 -0.175 -0.374 -0.679 0.013 R2 = 0.916 [FMI(-1) = AFIR-I] (1.47) (1.72) (2.87) (5.73) (2.32) DW = 1.48 (3a) 0.029 -0.407 -0.700 0.010 R2 = 0.894 [FMI(-1) = AFIR-2] (1.92) (3.64) (5.17) (1.93) DW = 2.13 (3b) 0.0349 -0.132 -0.462 -0.703 0.010 R2 = 0.908 [FMI (-1) = AFIR-2] (2.28) (1.53) (3.96) (5.34) (2.02) DW = 1.75 (4a) 0.040 -0.440 -0.428 0.004 R2 = 0.866 [FMI (-1) = ASMI] (2.47) (3.12) (4.65) (0.80) DW = 2.20 (4b) 0.043 -0.171 -0.475 -0.592 0.005 R2 = 0.888 [FMI (-1) = ASMI] (2.78) (1.81) (3.46) (4.66) (1.23) DW = 1.58 (5a) 0.034 -0.427 -0.548 0.009 R2 = 0.916 [FMI(-1) = AFMR-I] (3.21) (4.65) (5.24) (2.73) DW = 2.52 (5b) 0.042 -0.260 -0.509 -0.541 0.012 R2 = 0.964 [FMI(-l) = AFMR-1] (5.54) (3.69) (7.67) (7.58) (5.01) DW = 2.48 (6a) 0.049 -0.562 -0.645 0.005 R2 = 0.914 [FMI (-1) = AFMR-2] (5.40) (6.28) (5.92) (2.67) DW = 2.51 (6b) 0.055 -0.119 -0.613 -0.645 0.005 R2 = 0.926 [FMI(-1) = AFMR-2] (5.48) (1.86) (6.38) (6.07) (2.70) DW = 2.43 Notes: OLS method is used. Endogenous variable is TFP (total factor productivity). UER is the unemployment rate, and FMI is the financial market indicator. Measures of FMI used in the equations are FIR-1, FIR-2, FMR-1, FMR-2 (defined in Section II), and SMI (stock market index). Period of estimation is 1979-94, with lagged variables for FMI estimator starting in 1975. Absolute t-values in parentheses. growth model, the improvement of financial markets is specified to have a permanent level effect on total factor productivity. Indeed, adding the financial market indicators improves the overall fit, yielding for most lagged market indicator variables (with an Almon-lag-type structure)13 coefficients

13 The approach uses the data structure of the Almon lag to calculate a compos- ite variable AFMI(1,s)t = 1sAFMIt +2sAFM/t- 1 + . . . 1sAFMIt-1. Thus AFMI (2,2), = AFMI, + AAFMI1_1 + 9AFMIt_2. This approach results from economic, econometric, and data considerations. One would assume from an economic viewpoint that improvements in the financial market, as measured by changes in the level of FMI, would have little immediate impact but that the impact would grow over time. However, a direct application of the traditional Almon procedure is pre- vented by the small number of observations, with the increasing s under the full Almon approach reducing the degrees of freedom on a pro rata basis. In addition, AFMI (l,s) for 1 and s = 1, . . . , 3 proved to be highly correlated. Finally, unless very strict conditions are met, the Almon procedure will yield biased and inconsistent estimates.

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 167 that are significant at the 5 percent level and below while reducing the significance of the constant. The low t-value for the stock market index may be due to high multicollinearity between the unemployment rate and the stock market index, but alternative specification and estimation techniques were not successful. The estimated parameter values prove to be robust for different specifications and time periods of estimation (not shown), and the lagged impact of financial market indicators (compared with the contemporaneous impact, which proves statistically to be totally insignificant) gives confidence in the causality. Taken at face value, the results suggest that financial market developments strongly affect total factor productivity. Using the point estimates and assuming an equilibrium unemployment rate of 5 percent, the exogenous technical progress would amount to some 1 percent, to which 1+ percent of technical progress generated by financial market developments is added, yielding long-term annual total factor productivity of 2+ percent. The estimated financial market indicator effect of 1+ percent is likely to proxy other effects that may be highly correlated with financial market developments, such as reductions in exchange rate restrictions and increasing openness of the economy. Given data restrictions, the separation of these effects is not possible currently. The magnitude of parameter estimates for the stock market index variable in equation (4) in Table 3 invites comparison with the estimates for the cross-country study quoted above (Levine and Zervos, 1996). Their point estimate is 0.007 (with a t-value of 1.96), compared with our results of 0.004 and 0.005. While the statistical closeness of the point estimates may be spurious, the coincidence is surprising. In summary, the tentative empirical evidence suggests a positive impact of financial market developments on total factor productivity and thus economic growth. The available data and estimates do not allow one to distinguish between permanent and transitory effects. Yet, even if the higher total factor productivity of 1 percent a year were only of a temporary nature, it would if accumulated over 20-40 years provide welfare gains for current (and all future) generations, which should permit a major compensation of the transition generation.

Capital Formation, Saving, and Financial Market Development

In economic policy discussions, both capital formation and national savings are generally claimed to be positively influenced by a deepening of the financial market induced by pension reform (IMF, 1995). In a closed economy, the overall effect should be identical but may lead to a different distribution between public and private sector investments, on the one hand,

©International Monetary Fund. Not for Redistribution 168 ROBERT HOLZMANN

Table 4. Capital Formation and Financial Market Developments Equation Constant K Percent (-1) UER FMl (-1) (7) 0.030 0.79 -0.21 R2 = 0.945 (5.10) (9.11) (5.14) DW = 1.63 (8) 0.026 0.526 -0.175 0.008 R2 = 0.962 [FMI (-1) = AFIR-1] (4.78) (3.74) (4.38) (2.19) DW = 1.92 (9) 0.033 0.565 -0.247 0.007 R2 = 0.963 [FMI (-1) AFIR-2] (6.33) (4.64) (6.38) (2.28) DW = 2.21 (10) 0.028 0.770 -0.195 0.001 R2 = 0.947 [FMI (-1) ASMI] (4.18) (8.47) (3.92) (0.70) DW = 1.68 2 (11) 0.026 0.746 -0.193 0.004 R = 0.966 [FMI(-1) AFMR-1] (5.17) (10.3) (5.55) (2.61) DW = 2.00 (12) 0.038 0.684 -0.295 0.005 R2 = 0.969 [FMI (-1) AFMR-2] (6.95) (8.88) (6.78) (2.85) DW = 2.37 Notes: Endogenous variable is K Percent (change in real capital stock). For definitions of exogenous variables and other regression information, see notes to Table 3. and saving balances, on the other. In an open economy, foreign saving may additionally compensate for sectoral investment or saving imbalances. The traditional wisdom is that the influence of financial market deepening on both capital formation and saving is positive. However, further theoretical considerations and international empirical evidence suggest that the effects can go in either direction or compensate to zero. A positive relationship between financial market indicators and capital formation (as measured by the change in real capital stock) is suggested by considerations of better access to financing (that is, reduced borrowing constraints), enhanced incentives to undertake longer-term investment projects, and better risk allocation. However, a more efficient market, with reduced transaction costs and higher rates of return for all investment projects, may also lead to a redirection of the holding of wealth by economic agents in the form of existing equity claims and to less new capital investment (Bencivenga, Smith, and Starr, 1996). Preliminary econometric testing for Chile suggests that the change in capital stock follows an adjustment process, with the lagged variable figur- ing very significantly. This process is also influenced by cyclical effects, or expectations about future income developments, as proxied by the unemployment rate (Table 4).14 Inserting the lagged financial market indicator variables (again with an Almon-type lag) leads to an improvement in the equation fit and to coefficients that are consistent in sign and significant at

14 In view of the weak public sector statistics for Chile, the use of the unemployment rate is preferred to the use of disposable income.

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 169 the 5 percent level and below. The exception is the stock market index variable, which leads to insignificant coefficients, possibly reflecting the same multicollinearity problems noted above. Again, taking the point estimates for the FIR and FMR variables at face value, the long-term growth rate in real capital stock is some 5-8 percent. This rate is reduced by an assumed long-term unemployment rate of 5 percent by some 1-2 percentage points but increased by the enhanced financial markets by about 1.5 percentage points, or by more than one-fourth of its "natural" level. This result hints at the nonnegligible effects of financial market developments on the formation of capital stock. These effects have to be considered in addition to the effects on total factor productivity in assessing the growth consequences of financial market developments.15 With regard to the relationship between financial market indicators and saving, the same theoretical qualifications about the a priori sign and strength apply. Financial market developments may induce higher (private) saving via the provision of attractive saving outlets, but they may also reduce saving for a number of reasons. First, real interest rates may be higher, and the conventional income effect may prevail over the substitution effect. Second, better risk diversification is likely to change the form of saving and may reduce the level. Third, annuities may be available (allowing for higher old-age consumption and reducing the amount of un- intended bequests). Finally, access to consumer credits may be improved (allowing for better consumption smoothing).16 Preliminary econometric testing suggests that the negative effect of financial market developments on private saving prevails (Table 5). The basic specification of the private saving rate, with the constant and unemployment variables as the explanatory variables (again proxying income expectations), is generally improved if financial market variables are added. These financial market variables enter negatively, with parameters that are statistically significant at an error level of 5 percent and below (except, again, for the stock market index). This finding is consistent with international evidence on the impact of financial market liberalization and private saving behavior (IMF, 1995). Nevertheless, in view of the rising national and private saving in Chile in recent years, the result is surprising and requires a closer look at the components of, and developments in, saving.

15 Our reported long-term point estimate of the effects of financial markets on capital accumulation of about 0.015 is close to the estimate in the Levine and Zervos (1996) cross-country study of 0.011 (with a t-value of 2.38). 16 See Masson, Bayoumi, and Samiei (1995) for empirical references.

©International Monetary Fund. Not for Redistribution 170 ROBERT HOLZMANN

Table 5. Private Savings and Financial Market Developments Equation Constant UER AUER (-2) FMI {-1) (13) 0.274 -1.53 0.662 R2 = 0.904 (18.7) (11.4) (3.94) DW = 1.43 (14) 0.309 -1.75 0.811 -0.013 R2 = 0.914 [FMI (-1) = AFIR-1] (10.0) (7.87) (3.98) (1.31) DW = 1.72 (15) 0.302 -1.66 0.844 -0.013 R2 = 0.924 [FMI (-1) = AFIR-2] (14.7) (10.9) (4.52) (1.86) DW= 1.75 (16) 0.306 -1.78 0.814 -0.010 R2 = 0.913 [FMI (-1) = ASMI] (11.2) (8.12) (4.00) (1.26) DW= 1.76 (17) 0.308 -1.68 0.859 -0.018 R2 = 0.961 [FMI(-1) = AFMR-1] (23.6) (16.3) (6.89) (4.12) DW = 2.96 (18) 0.267 -1.37 0.761 -0.008 R2 = 0.942 [FMI (-1) = AFMR-2] (21.5) (10.5) (5.30) (2.85) DW = 1.89 Notes: Endogenous variable is SPR (private saving as a share of GDP). For definition of exogenous variables and other regression information, see notes to Table 3.

Pension Reform and National Saving

The structure and development of capital formation and its financing over the last 20 years exhibits a strong rise in the private saving rate since the mid-1980s, the substantial importance of public saving, and the fall in importance of foreign saving (Figure 4). Public plus private saving—that is, national saving—has reached levels not achieved since at least 1970. To gauge the impact of pension reform on national saving, a different disaggregation, consistent with the System of National Accounts (SNA), is proposed. A balancing item is calculated that consists of two components: the private saving generated (in an accounting sense) by the new scheme, reduced by the fiscal costs (that is, public dissaving). The first component, the saving generated, consists of two flows: (1) the net contribution payments to the new scheme (essentially the 10 percent premium revenue), which in the SNA is considered as private saving; and (2) the returns on assets generated by the new scheme. These returns are not equivalent to the change in assets by the pension funds (net of contribution payments), as the asset increase comprises both capital gains and flow effects. We are interested only in the latter, which are part of private saving in the SNA. Both flows were estimated from available data.17

17 The net contributions were calculated as premium payments plus other increases minus commissions charged minus benefit payments. The flow returns on pension fund assets were calculated using the average yearly stock and a representative interest rate of the financial market. In principle, the flows ought to be corrected for savings generated in the upstream insurance companies and for dissavings generated by benefit payouts, but these data were not at hand.

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 171

The second component, the financial costs (public dissaving, excluding the interest payments resulting from the larger fiscal deficit), consists also of two flows: (1) the deficit in the old scheme arising from the loss of contributors (which was approximated by the state contribution to the old scheme, excluding assistance pension financing); and (2) expenditure for recognition bonds, which compensate those who switched to the funded scheme for rights acquired under the unfunded scheme, as reported in the social security statistics. Table 6 indicates that the contribution of pension reform to national saving was negative from the inception of reform in 1981 until 1988. The positive balance since 1989 is essentially due to higher returns on capital investment, while the flows from premiums in percent of GDP remain largely constant at about 2.5 percent. Also, the fiscal costs exhibit in 1990 an increase of only some 0.8 percent of GDP (owing to improved pension indexation by the new government). Figure 5 disaggregates national saving according to (1) the saving balance of pension reform; (2) the private sector saving net of the pension system; and (3) the public sector saving net of (that is, plus) the fiscal costs. The striking message of this disaggregation is the parallel rise of each net saving item in the period 1984-87 and their leveling off in the period

©International Monetary Fund. Not for Redistribution Table 6. Pension Reform and Saving Effectsa (In percent of GDP) Year 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Total saving effects (net) -0.2 -2.2 -1.8 -1.1 -0.2 -0.9 -0.2 -0.9 1.6 3.4 3.2 2.6 2.8 2.0 Private saving generated 1.3 2.6 3.0 3.7 4.4 3.9 4.6 3.8 6.0 8.6 8.4 7.7 8.2 7.6 Flows: Net contributionb 0.9 1.7 1.5 1.5 1.5 1.6 1.6 1.6 1.9 1.9 2.4 2.8 2.4 2.2 Flows: Net savingc 0.4 0.9 1.5 2.2 2.9 2.3 3.1 2.2 4.2 6.7 6.1 4.9 5.7 5.4 Fiscal costs of reform -1.5 -4.8 -4.8 -4.8 -4.6 -4.7 -4.8 -4.7 -4.5 -5.3 -5.3 -5.1 -5.4 -5.6 Deficit coveraged -1.5 -4.7 -4.6 -4.6 -4.3 -4.3 -4.4 -4.2 -4.1 -4.8 -4.8 -4.6 -4.8 -5.0 Recognition bonds -0.1 -0.2 -0.2 -0.3 -0.4 -0.5 -0.5 -0.4 -0.5 -0.4 -0.5 -0.6 -0.7 Sources: National statistics; PrimAmerica Consultatores (1995); and author's calculations. a Consistent with system of national accounts. b Pension contributions to pension funds plus other increases minus commissions minus benefit payments. c Change in pension assets net of capital gains. d State contribution to social security pension funds (net of payments for social assistance).

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 173

1988-90. Since then, both private saving and public saving net of the pension system have remained largely constant at slightly above 10 percent of GDP. The positive net contribution of the new pension system to national saving began after both private and public saving net of the pension system had peaked, thereby confirming the econometric evidence that the pension reform had no direct, positive impact on private or national saving. Figure 6 highlights the contribution of fiscal performance and public saving to the transition from an unfunded to a funded pension scheme, as well as the fiscal costs involved in that transition. The figure shows that restrictive fiscal policy that pays off government debt through higher taxes or lower expenditure and thus shifts resources from current to future generations encourages higher saving and capital formation. Reportedly, the Chilean authorities deliberately strengthened the fiscal stance for some years before beginning the pension reform.18 This strategy led in 1980 to a positive budget balance of the general government of 4.4 percent of GDP and a public saving rate—that is, the balance of current public revenue and expenditure in percent of GDP—of 7.4 percent of GDP.

18 The source for this information is Juan Carlos Mendez, Budget Director until 1980, quoted in Diamond and Valdez-Prieto (1994).

©International Monetary Fund. Not for Redistribution 174 ROBERT HOLZMANN

The ensuing decline in both the fiscal stance and public saving triggered by the beginning of the reform in 1981 was quickly reversed; the public saving rate became positive again in 1985, and the fiscal stance was balanced in 1988. The share of current revenue as a percent of GDP declined, owing to various tax reforms and tax rate cuts, but the fiscal stance was strengthened through program reforms and expenditure cuts (see Marshall and Schmidt-Hebbel (1994)). The new, democratic government that came into office in 1989 has stabilized the expenditure level but has done little to reverse the trend of declining revenue. On the surface, this fiscal behavior suggests that pension reform and its transition is being fully financed by the current generation through a lower level of public consumption, in contrast to the alternative of shifting the burden to future generations via larger (explicit) public debt. However, this assessment is partly changed if the counterfactual chosen is the old growth path without reform. If, in an extreme formulation, all the growth differential of real GDP of up to 2 percent a year resulting from higher total factor productivity and capital formation were to be attributed to pension reform, the net burden for the transition generation would amount to about 3 percent a year, given fiscal costs of transition of some 5 percent a year. However, a strengthening of the fiscal stance is also required in this counter-

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 175 factual, as the budget captures only a fraction of the increased output level (some 20 percent) while the remainder flows to the private sector.

IV. Tentative Conclusions

Given the still incomplete state of the data and the tentative, simple econometric testing done, the conclusions reached in this paper can be seen as only very provisional. However, three main conclusions are suggested. First, the empirical findings are seemingly consistent with the claim that financial market developments enhanced economic growth and that pension reform has contributed to this development. The links between financial market indicators and total factor productivity and capital accumulation are surprisingly robust, and a link between pension funds and financial market developments is suggested by the data with regard to the level and scope of their interaction. However, the question of how the financial market in Chile would have developed had the pension reform not taken place may never be answered. Second, contrary to the common belief about the effects of the pension reform, the empirical findings suggest that the direct effect of financial market developments on the private saving rate was negative. These results are supported by an alternative approach that disaggregates the impact of pension reform on the national saving rate along SNA lines. The data indicate that net pension savings were negative until 1989 and small afterward. These approaches independently suggest that the conventionally assumed impact of a Chilean-type pension reform on private (and national) saving may not hold. The suggested alternative causality is twofold: (1) economic growth owing to higher total factor productivity and capital accumulation and better labor market performance is at the heart of the higher private saving rate; and (2) the increase in the private saving rate is strengthened by higher public saving. Including the financing of the transition costs of the reform, public saving increased by some 10 percentage points of GDP, about one-half of which was used to finance the reform. These results also temper the optimism reigning in countries in Latin America and Eastern Europe, where pension reform is seen as an easy vehi- cle to boost national saving, and thus capital accumulation and growth. Receiving the economic benefits of such a reform is likely to require a comprehensive and credible economic reform package and definitely requires a tight fiscal stance. Otherwise, the positive effect of financial market development on total factor productivity may not emerge, or it may be (partially) counteracted by reduced capital formation or require higher foreign saving.

©International Monetary Fund. Not for Redistribution 176 ROBERT HOLZMANN

However, if the fiscal stance is considered weak, higher foreign saving may not materialize. Third, the main message from the Chilean experience for the emerging economies throughout the world is, in principle, encouraging. Pension reform and a shift from an unfunded to a funded scheme may create positive externalities (on labor and financial markets), thereby accelerating the growth rate. Furthermore, these potential (endogenous or transitory) growth effects ease or perhaps even eliminate the double burden for the transition generation (if measured according to the old growth path). Finally, the potentially higher growth path may allow for some back-loading of the fiscal implications via recognition bonds and other devices. However, pension reform as implemented in Chile requires a comprehensive reform package covering essentially all areas of macroeconomic and microeconomic policy and institution building, supported by a tight fiscal policy. Moreover, pension reform must not be left only to social policy specialists; it should be entrusted also to economists of different backgrounds, including industrial economics, public finance, and financial markets.

REFERENCES

Auerbach, Alan J., and Laurence J. Kotlikoff, 1987, Dynamic Fiscal Policy (Cam- bridge and New York: Cambridge University Press). Banco Central de Chile, Boletin Mensual, various issues. Bencivenga, Valerie R., Bruce D. Smith, and Ross M. Starr, 1996, "Equity Markets, Transactions Costs, and Capital Accumulation: An Illustration," World Bank Economic Review, Vol. 10 (May), pp. 241-65. Breyer, Friedrich R., 1989, "On the Intergenerational Pareto Efficiency of Pay- As-You-Go Financed Pension Systems," Journal of Institutional and Theo- retical Economics, Vol. 145 (December), pp. 643-58. , and Martin Straub, 1993, "Welfare Effects of Unfunded Pension Systems When Labor Supply Is Endogenous," Journal of Public Economics, Vol. 50 (January), pp. 77-91. Cerda, J.G., and H.O. Zeballos, 1975, El Mercado de Capitales en Chile (Santiago, Chile: Banco Central de Chile). Corsetti, Giancarlo, 1994, "An Endogenous Growth Model of Social Security and the Size of the Informal Sector," Revista de Analisis Economico, Vol. 9 (June), pp. 57-76. De la Cuadra, Sergio, and Salvador Valdes-Prieto, 1992, "Myths and Facts About Financial Liberalization in Chile: 1974-1983," in If Texas Were Chile: A Primer on Banking Reform, ed. by Phillip L. Brock (San Francisco, California: ICS Press), pp. 11-129. Diamond, Peter, 1997, "Insulation of Pensions from Political Risk," in The Eco- nomics of Pensions: Principles, Policies, and International Experience, ed. by

©International Monetary Fund. Not for Redistribution PENSION REFORM IN CHILE 177

Salvador Valdes-Prieto (Cambridge and New York: Cambridge University Press), pp. 33-57. , and Salvador Valdes-Prieto, 1994, "Social Security Reforms," in The Chilean Economy: Policy Lessons and Challenges, ed. by Barry P. Bosworth, Rudiger Dornbusch, and Raul Laban (Washington: Brookings Institution), pp. 257-328. Godoy-Arcaya, Oscar, and Salvador Valdes-Prieto, 1997, "Democracy and Pen- sions in Chile: Experience with Two Systems," in The Economics of Pensions: Principles, Policies, and International Experience, ed. by Salvador Valdes- Prieto (Cambridge and New York: Cambridge University Press), pp. 58-91. Goldsmith, Raymond W., 1969, Financial Structure and Development (New Haven, Connecticut: Yale University Press). Greenwood, Jeremy, and Boyan Jovanovic, 1990, "Financial Development, Growth, and the Distribution of Income," Journal of Political Economy, Vol. 98 (October), pp. 1076-107. Holzmann, Robert, 1988, Reforming Public Pensions (Paris: Organization for Eco- nomic Cooperation and Development). , 1994, "Funded and Private Pensions for Eastern European Countries in Transition," Revista de Analisis Economico, Vol. 9 (June), pp. 183-210. , 1997, "Starting Over in Pensions: The Challenges Facing Central and East- ern Europe," Journal of Public Policy (forthcoming). Homburg, Stefan, 1990, "The Efficiency of Unfunded Pension Schemes," Journal of Institutional and Theoretical Economics, Vol. 146 (December), pp. 640-47. International Monetary Fund, 1995, World Economic Outlook, May 1995: A Sur- vey by the Staff of the International Monetary Fund, World Economic and Financial Surveys (Washington). Korajczyk, Robert A., 1996, "A Measure of Stock Market Integration for Devel- oped and Emerging Markets," World Bank Economic Review, Vol. 10 (May), pp. 267-89. Kotlikoff, Laurence J., 1996, "Privatization of Social Security: How It Works and Why It Matters," in Tax Policy and the Economy, Vol. 10, ed. by James M. Poterba (Cambridge, Massachusetts: National Bureau of Economic Research and MIT Press), pp. 1-32. Levine, Ross, and Sara Zervos, 1996, "Policy, Stock Market Development and Long-Run Growth," World Bank Economic Review, Vol. 10 (May), pp. 323-39. Marshall, Jorge, and Klaus Schmidt-Hebbel, 1994, "Chile: Fiscal Adjustment and Successful Performance," in Public Sector Deficits and Macroeconomic Per- formance, ed. by William Easterly, Carlos A. Rodriguez, and Klaus Schmidt- Hebbel (Oxford: Oxford University Press for the World Bank), pp. 167-224. Masson, Paul R., Tamim Bayoumi, and Hossein Samiei, 1995, "International Evi- dence on the Determinants of Private Saving," IMF Working Paper 95/51 (Washington: International Monetary Fund). Pinera, Jose, 1991, El Cascabel al Gato: La Batalla por la Reform Previsional (Santiago, Chile: Zig-Zag). PrimAmerica Consultatores, ed., 1995, Sistema de AFP-Antecedentes Estadisticos, 1981-1994 (Santiago, Chile: Editorial Antarctica).

©International Monetary Fund. Not for Redistribution 178 ROBERT HOLZMANN

Queisser, Monika, 1995, "Chile and Beyond: The Second-Generation Pension Reforms in Latin America," International Social Security Review, Vol. 48 (No. 3-4), pp. 23-39. Raffelhiischen, Bernd, 1993, "Funding Social Security Through Pareto-Optimal Conversion Policies," Journal of Economics, Sup. 7, pp. 105-31. Saint-Paul, Gilles, 1992, "Technological Choice, Financial Markets and Economic Development," European Economic Review, Vol. 36 (May), pp. 763-81. Schmidt-Hebbel, Klaus, 1993, "Pension Reform Transitions from State Pay- As-You-Go to Privately Managed, Fully Funded Systems" (unpublished; Washington: World Bank). Superintendencia de Administradoras de Fondos de Pensiones, Boletin Estadistico, various issues. Valdes-Prieto, Salvador, 1997, "Introduction" to The Economics of Pensions: Principles, Policies, and International Experience, ed. by Salvador Valdes- Prieto (Cambridge and New York: Cambridge University Press), pp. 1-30. Villanueva, Delano, 1993, "Openness, Human Development, and Fiscal Policies: Effects on Economic Growth and Speed of Adjustment," IMF Working Paper 93/59 (Washington: International Monetary Fund). Walker, Eduardo, 1993a, "Desempeno Financiero de las Carteras de 'Renta Fija' de Los Fondos de Pensiones en Chile: <^Ha Tenido Desventajas Ser Grandes?" Cuarderno de Economia, Vol. 30 (April), pp. 1-33. , 1993b, "Desempeno Financiero de las Carteras Accionarias de los Fondos de Pensiones en Chile: ^Ha Tenido Desventajas Ser Grandes?" Cuarderno de Economia, Vol. 30 (April), pp. 35-75. World Bank, 1994, Averting the Old Age Crisis: Policies to Protect the Old and Promote Growth (Oxford and New York: Oxford University Press for the World Bank).

Fiscal Policy and Long-Run Growth

VITO TANZI and HOWELL H. ZEE*

This paper discusses in a systematic and comprehensive way the existing literature on the relationship between the growth of countries' economies and various public finance instruments, such as tax policy, expenditure policy, and overall budgetary policy, from the perspectives of allocative efficiency, macroeconomic stability, and income distribution. It reviews both the conceptual linkages between each of the instruments and growth and the empirical evidence of such relationships. The paper broadly concludes that fiscal policy could play a fundamental role in affecting the long- run growth performance of countries. [JEL H00, O00, E60]

ISCAL POLICY has occupied center stage in recent policy deliberations in Fmany developed, developing, and transition economies,1 as concerns with fiscal dimensions, such as high unemployment, inadequate national savings, excessive budget deficits and public debt burdens, and looming crises in the financing of pension and health care systems, have intensified. Timely policy responses to these concerns have taken on a sense of urgency. In these circumstances, issues relating to the appropriate scope, nature, and conduct of fiscal policy, in the context of both mitigating macroeconomic instability in the short run and fostering growth in the long run, have naturally come to the fore in policy debates. The division of fiscal policymaking into three hypothetical but inter- dependent branches—allocation, distribution, and stabilization—as first formalized by Musgrave (1959) almost 40 years ago, remains to this day a useful conceptual framework for discussing, analyzing, and evaluating

*Vito Tanzi is Director of the Fiscal Affairs Department. He holds a doctorate from Harvard University. Howell H. Zee is Advisor, Fiscal Affairs Department. He holds a doctorate from the University of Maryland (College Park). The authors gratefully acknowledge the helpful comments from Martin C. McGuire and useful suggestions from Gian Maria Milesi-Ferretti. 1 See, for example, IMF (1996). 179

©International Monetary Fund. Not for Redistribution 180 VITO TANZI and HOWELL H. ZEE

alternative fiscal policy measures, even though the world today is much more complex than the one that existed at the time when Musgrave wrote on the subject. Musgrave's framework has endured largely because it helps organize one's thoughts, in an elegant way, on the fundamental issues of interest to policymakers—efficiency in resource utilization, equity in income distribution, and cyclical fluctuations. This paper considers the positive and normative aspects of the impact on growth of different instruments of fiscal policy. These instruments fall broadly under the three conventional classifications of taxation, public expenditure, and aggregate budgetary balance. Each is analyzed from the perspective of Musgrave's three economic branches. While a multiplicity of factors, some of which are of a noneconomic nature, could plausibly affect the performance of an economy from period to period, a country's growth over a reasonably long period of time is ultimately determined by three factors: (1) given the state of technical know-how in that country, the efficiency with which any existing stock of resources is utilized (which would depend, among other things, on cultural, institutional, and political, as well as economic, parameters); (2) the accumulation over time of productive resources (which would include human and other forms of intan- gible capital); and (3) technological progress (which for most countries would depend, among other things, on their ability to absorb new technology from abroad).2 This paper traces various channels through which tax policy, expenditure policy, and overall budgetary policy could affect growth through their impact on the above three factors. While for ease of exposition the growth effects of different fiscal policy instruments are discussed separately below, it does not imply that they are independent of each other: the impact on growth of taxes would depend, for example, in addition to their level and structure, on how the tax revenue is spent (the composition of public expenditure), as well as on how taxes affect the overall budgetary balance (for any given total level of public expenditure). This interdepen- dence underscores the danger in policy deliberations of focusing too

2 While the second factor—resource accumulation—has traditionally been the focal point of growth economics, Schumpeter (1934) made the case for the first and third factors, which together imply productivity improvement, as the main ingredients for growth. Tanzi (1995) expanded on the Schumpeterian theme and emphasized the importance of a country's social absorptive capacity (with respect to technology) in determining its development. The resource-accumulation versus productivity-improvement debate has raged in recent years as researchers have tried to understand the factors that have contributed to the impressive growth of a small number of East Asian economies. For an argument supporting the resource- accumulation thesis in this context, see Young (1995).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 181 narrowly on the value of one variable (for example, the budgetary balance) to the neglect of its underlying components—a conceptual equivalent to not properly controlling for relevant variables in empirical estimations. In general, the empirical growth literature has shown that estimation results could be materially affected by the presence or absence of controlled variables. The relevance of considering the growth effects of fiscal policy must be predicated, of course, on the basic proposition that policy matters for long- run growth. Although this may seem intuitively obvious, it is in fact a relatively new idea; it became established in mainstream economic think- ing only with the recent advent of the endogenous growth literature.3 As an alternative paradigm to the neoclassical growth theory of Solow (1956) and Swan (1956),4 in which long-run growth was completely determined by factors exogenous to the theory itself (and, therefore, is invariant to policy),5 the endogenous growth literature has been largely motivated as an attempt to overcome the former's failure to reconcile theory with some of Kaldor's (1961) six celebrated, stylized facts of growth, most notably the seeming absence in the data of any discernible sign of growth convergence along income levels across countries—a fundamental implication of the neoclassical paradigm.6 While this paper is not a survey per se of the variety of models, results, and policy implications of the voluminous endogenous growth literature,7 it takes as given the basic premise of this literature that a country's growth performance in the long run is endogenously determined by a set of variables that are responsive to (and affected by) policy—in this particular case, fiscal policy. A legitimate question that could be raised is the appropriateness of adopting national output, rather than welfare, as the yardstick for evaluating policy, since it is presumably welfare that is the ultimate concern of

3 While the early roots of this literature go back to Arrow (1962) and Uzawa (1965), its present analytics owe much to Romer (1986) and Lucas (1988). 4 The intertemporal utility-maximization-based neoclassical growth models that have become familiar today are actually due to Cass (1965) and Koopmans (1965). 5 The policy invariance implication of the neoclassical growth theory applies only to the steady state, the attainment of which may take a period longer than most would regard as the long run. This point should be borne in mind when future references to the above implication are made in the rest of the paper. 6 For an illuminating discussion, see Romer (1989). There is, however, no consensus among researchers on the question of convergence. Studies by Baumol (1986), Mankiw, Romer, and Weil (1992), Barro and Sala-i-Martin (1995), and Sala-i-Martin (1996a and 1996b) tended to confirm the existence of convergence, provided that variables other than income (such as human capital) are properly controlled. For a dissenting view, see Quah (1996a and 1996b). 7 For a textbook treatment of the literature, see Barro and Sala-i-Martin (1995).

©International Monetary Fund. Not for Redistribution 1 82 VITO TANZI and HOWELL H. ZEE

policymakers. The problem of the gap, and in more extreme circumstances that of changes in opposite directions, between output and welfare is well known: it follows from the omission in national income accounts of imputed values for household production, leisure, and environmental externalities, among other factors.8 Hence, it is theoretically possible for some policy measures to stimulate output and yet reduce welfare (for example, by providing excessive tax incentives to certain industries without properly taking into account the possible pollution costs associated with their activities). While this problem is conceptually important (and, hence, worth noting explicitly), its resolution in practice would present for- midable measurement problems because the more encompassing welfare effects of policy are not directly and objectively observable. Moreover, in discussing issues of long-run growth, the bias in evaluating policy on the basis of output effects is somewhat lessened, as any persistent difference between changes in output and those in welfare is unlikely to be sustainable in the long run.9 Accordingly, this paper will adopt the conventional approach in assuming that output growth is positively correlated with welfare improvement. In what follows, the impact of fiscal policy on growth will be taken up first from the point of view of allocative efficiency, then from that of stability. The relatively new literature on the growth effects of fiscal policy from an income distribution perspective will be considered last.

I. Allocative Efficiency and Growth

Tax Policy

Some of the clearest and most direct conceptual links between fiscal policy and growth have traditionally been associated with tax policy. These links have been made, of course, because the allocative impacts of taxation (for example, on labor-leisure choice, on consumption-saving behavior, and on the relative profitabilities of different industries, among others) are easily appreciated by theorists and policymakers alike and, consequently, have long been one of the best-researched areas in eco-

8 An early, influential investigation into this problem was provided by Nordhaus and Tobin (1973). A different aspect of the problem commonly found in most centrally planned economies is that of the suboptimal mix of outputs—often heavily biased toward the production of capital goods. 9 The use of output as a reference point in policy evaluation was defended by Aiyagari (1990).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 1 83 nomics. The various links between taxation and growth have, however, different conceptual underpinnings, so it will be useful to consider them separately. One link is built on the idea that all taxes are nonneutral, with the singu- lar exception of lump-sum levies (which are largely irrelevant as a practi- cal instrument, and may even be nonneutral in an intertemporal context if planning horizons are finite or in open economies). With nonneutral taxes, private economic agents' allocative decisions will be different from those that would be made in the absence of such taxes. This tax-induced distor- tion in economic behavior results in a net efficiency loss to the whole economy, commonly referred to as the "excess burden of taxation," even if the government engages in exactly the same activities—and with the same degree of efficiency—as the private sector with the tax revenue so raised.10 It then follows that the higher the level of taxation, the larger would be this efficiency loss. Moreover, the loss typically grows disproportionately with increases in the tax level when there are other tax distortions in the economy. This result would hold even if the taxes were optimally structured, in the sense that, the excess burden of each tax is equalized (proportionally) across all taxes.11 It must be pointed out that, while it is straightforward to conceptualize a negative relationship between the level of taxation and the level of output, it is not clear why the level of taxation would adversely affect the long-run growth of output.12 To obtain a growth effect, the appropriate variable to consider should instead be the rate of change in the level of taxation.13 This link between taxation and growth is established on the basis of the former's excess burden in a static context. A second and more conspic- uous link has to do with the impact of taxation on factor accumulation, particularly capital; it relates, therefore, to the excess burden of taxation in a dynamic sense. Because in the neoclassical growth paradigm long-run growth is invariant to policy,14 as noted earlier, the focus of the traditional

10 The concept and measurement of excess burden has a long and controversial history in economics, dating in its modern formulation at least as far back as the work of Dupuit more than a century ago. For a recent comprehensive survey of this literature, see Auerbach (1985). 11 This is (heuristically) the celebrated Ramsey (1927) rule of optimal taxation, a modern reformulation and generalization of which can be found in Diamond and Mirrlees (1971). 12 If the period for output to adjust to any given change in the level of taxation is lengthy, the level of taxation would have, of course, an impact on the measured growth over the period. 13 This is a central result of Engen and Skinner (1992). 14 Policy can, however, have a transitory impact.

©International Monetary Fund. Not for Redistribution 184 VITO TANZI and HOWELL H. ZEE

analyses of capital income taxation employing such a paradigm is on the long-run tax impact on the level rather than on the growth of output.15 Policy implications are entirely different, however, when growth is responsive to policy, as in the case of the endogenous growth theory. Here, all other things being equal, a tax on income from (physical) capital would lower the after-tax return to savings and is, therefore, a disincentive to accumulate (physical) capital. But the ultimate impact of this tax on growth turns out to be a priori ambiguous; it is dependent on how the other factors, such as human capital, that cooperate with physical capital in the production process are affected by the tax (assuming these other factors are not taxed).16 A comparison between two simple cases, both with human capital as the only other factor of production, will illustrate the difference in outcomes in a particularly transparent manner. In one case, assume that the production of human capital requires only human capital (Lucas, 1988). Then the growth-depressing effect of the tax on physical capital will be entirely offset by an increase in human capital accumulation. Hence, the net impact on growth is zero. Alternatively, suppose the production of human capital requires both human and physical capital. In this case, the offset will be only partial, and the net impact on growth is negative (Rebelo, 1991). This simple comparison underscores the important point that the growth effects of income taxation on (physical) capital are sensitive to the specification of production technology.17 In general, however, it can be rea- soned that, the lighter the tax burden on the production of human capital relative to the tax burden on other sectors that are human capital intensive, the smaller will be the adverse impact on growth of taxing physical capital.18

15 The voluminous literature on this subject was succinctly surveyed in Sandmo (1985). Two well-known results from this literature are worth noting. Atkinson and Sandmo (1980) showed that, in a two-period, life-cycle, overlapping-generations model, the optimal capital income tax rate in the long run is not necessarily zero, but instead would generally depend on the relative tax elasticities of labor supply and savings, as well as on their cross elasticities. This outcome is characteristic of the optimal taxation literature. In contrast, Chamley (1986), using an infinite- horizon model, demonstrated that the long-run optimal capital income tax rate is in fact zero. The two results differ because the intergenerational inefficiency resulting from taxing capital income is not fully capturable in a life-cycle framework. 16 If all factors, including human capital, are taxed at the same rate, then long-run factor proportions are unchanged by the tax, in which case long-run growth would be unambiguously lowered as a result. See Rebelo (1991). 17 Many of these issues were surveyed in Xu (1994). Zee (forthcoming) showed that, in addition to the technology of production, the growth effects of income taxation will also depend on the specification of time preference. If time preference is endogenous, that is, if one's valuation of current relative to future consumption is responsive to the current levels of income and consumption, then an income tax would also affect savings through this time preference channel. 18 On this point, see Lucas (1990). A quantitative assessment of the growth effects of taxing both physical and human capital in a nonuniform manner under different technological specifications was provided by Stokey and Rebelo (1995).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 185

The above discussion suggests that the structure of taxation could have important implications for growth. This consideration actually is not limited to simply the area of capital income taxation, or even to income taxation in general; it has, in fact, broad significance for the overall structure of the entire tax system. For a given total tax level, a relative shift from income to consumption taxation would, for example, reduce the disincentive to save and, consequently, boost capital accumulation.19 While a tax on consumption distorts labor-leisure choice, it is neutral with respect to the relative price of consumption today and tomorrow and, thus, can produce only a level (rather than growth) effect.20 In addition to its impact on resource accumulation, the structure of a tax system may have other growth consequences. A heavy reliance on trade taxes could, for example, impede an economy's capacity to absorb or develop new technologies—thus harming its growth prospects—by reducing the exposure of domestic industries to international markets and competition; however, tax administration constraints may pose difficulties for their elimination.21 Another channel through which tax policy could have a significant impact on both resource accumulation and technological progress is the provision of tax incentives (also known otherwise as a form of tax expenditure), which in one form or another exist in almost all (developing and developed) countries, for promoting investment and research and development activities. While there is a broad consensus among economists that general, that is, nontargeted, incentives are of questionable value—relative to other factors such as the stability, simplicity, and neutrality of a tax system with internationally comparable tax rates—in achieving their stated objectives, there is much less agreement on targeted incentives.22 By their very nature, all tax incentives create distortions; however, if targeted incentives are designed to mitigate certain market failures, then the distortions

19 It is common to note that, in the absence of a labor-leisure choice, the intertemporal budget constraint of an economic agent implies that taxing wage income (and inherited wealth) only (leaving interest income untaxed) is equivalent to taxing consumption (and bequests), with national savings unaffected by the choice between these two taxes (unless there are other distortions). Tanzi and Zee (1993) showed, however, that, if consumption requires time, a wage tax would discourage savings in a manner similar to that of a tax on interest income. 20 For an extended discussion on the growth effects (or the lack thereof) of taxing consumption, see Stokey and Rebelo (1995). 21 Trade taxes are frequently the most administratively reliable tax handles and consequently are heavily relied upon to produce revenue in many developing countries. On average, trade taxes (especially import duties) amount to about one-fourth of total tax revenue in a broad group of non-Organization for Economic Coopera- tion and Development (OECD) countries, compared to about 2 percent in OECD countries. See Zee (1996). 22 Many conceptual and analytical aspects of tax incentives, as well as country practices, are covered in OECD (1994) and Shah (1995).

©International Monetary Fund. Not for Redistribution 186 VITO TANZI and HOWELL H. ZEE that they cause may well be outweighed by the benefits that can be reaped from their use. For example, the social benefits from enhanced growth produced by many investments and research and development activities could exceed their private returns. Without corrective public measures, such activities would be below their optimal levels.23 Such arguments usually neglect, however, political economy costs of providing tax incentives, as they tend to encourage rent-seeking behavior, corruption, and the development of special interest groups.24 While these costs are not easy to measure, their adverse impact on growth may be significant.25 The empirical evidence of the impact of various aspects of tax policy on growth has so far been mixed.26 While there is some general indication that the relationship between either the total tax or the income tax level and growth is negative, this relationship is not robust and is sensitive to model specification, particularly with respect to the list of nontax variables that are controlled. Easterly and Rebelo (1993) experimented with thirteen different tax measures and found only one—a marginal income tax rate computed by a time-series regression of income tax revenue on GDP—to be statistically significant in explaining growth variations among their sample countries. By including the initial real GDP per capita as a variable in the regressions, these authors found that the strikingly negative correlation between the ratio of income tax revenue to GDP and growth shown in Plosser (1992)— and subsequently much cited by others—to be statistically insignificant. Mendoza, Milesi-Ferretti, and Asea (forthcoming) constructed more- detailed effective income and consumption tax rates using corresponding data on tax revenues and tax bases for 18 OECD countries, following a

23 DeLong and Summers (1991) argued on just this basis for providing tax incentives to equipment investment, which they find to have strong growth effects. Murphy, Shleifer, and Vishny (1989) showed that intersectoral spillover effects of industrialization would call for the implementation of investment promotion policies in a coordinated manner. 24 The growth-lowering effects of rent-seeking activities have been examined in Baumol (1990) and Murphy, Shleifer, and Vishny (1991) in the context of how entrepreneurship and talent are allocated among alternative activities. Mauro (1995) found cross-country evidence that corruption retards growth. 25 It is common for advocates of tax incentives to point to the extensive use of such incentives in some high-growth Asian economies as evidence of their effectiveness. Tanzi and Shome (1992) speculated, however, that this positive association probably has less to do with the nature of the incentives themselves than with the characteristics of the countries where they are used, such as the quality of the civil servants and the efficiency of the public bureaucracy—characteristics that tend to minimize the political economy costs of providing the incentives. See also Olson (1996) for a discussion of the relationships among economic incentives, institutions, and economic performance. 26 For surveys of this literature, see Levine and Renelt (1991), Easterly and Rebelo (1993), and Xu (1994).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 1 87 methodology developed by Mendoza, Razin, and Tesar (1994); they found that, based on panel regressions, reductions in income (consumption) taxes would have a robust and statistically significant positive (negative) impact on investment. This impact is, however, quantitatively not sufficient to produce any statistically significant long-run growth effects. As noted earlier, Engen and Skinner (1992) found statistically significant relationships between growth and the rate of change in tax levels. The most severe difficulty in isolating the impact of taxation on growth arises because key nontax fiscal variables, such as public expenditure and budget policies, that are often not independent of tax policy can also affect growth (see below); also, the complex interactions among the fiscal and other macroeconomic variables create difficulties.27 For example, some evidence also suggests that the growth effects of fiscal policy variables are dependent on income levels, and the negative relationship between tax levels and growth rates is strongest among the least developed countries.28 Overall, the general conclusion can be drawn that the empirical evidence on the relationship between taxation and growth is much weaker than what the theory would have led one to expect.

Public Expenditure Policy

The financing of any level of public expenditure,29 whether through taxation or borrowing, involves the absorption of real resources by the public sector that otherwise would be available to the private sector.30 From a purely static, allocative point of view, this absorption would improve overall efficiency if the social return (benefit) from public expenditure exceeds its private opportunity cost. While public expenditure may dis- place private sector output (the crowding-out effect), it may also improve private sector productivity (the externality or public good effect). Its total social return must, therefore, be interpreted as the sum of both of these effects.31 The net impact on aggregate output of the crowding-out effect

27 In their sensitivity analysis of cross-country growth regressions, Levine and Renelt (1992) found that the investment share in GDP is the only robust variable in explaining growth. 28 See Martin and Fardmanesh (1990). 29 Public expenditure here refers to the exhaustive type, that is, expenditures of a purely transfer nature (including subsidies and welfare payments) are excluded. This definition is also consistent with the national income accounts data on such expenditure on which most empirical studies are based. Transfers have, however, distributional implications, which are discussed in Section III. 30 The absorption of domestic resources will be delayed, of course, if foreign borrowings or unemployed resources are available. 31 For a clear separation of these two effects, see Ram (1986).

©International Monetary Fund. Not for Redistribution 188 VITO TANZI and HOWELL H. ZEE of public expenditure clearly depends on the relative marginal productiv- ities of the public and private sectors. There is a widespread belief that, absent externalities, public production tends to be less efficient than private production.32 Hence, on account of this effect alone, the higher the level of public expenditure, the greater the inefficiency and the lower the level of output. To relate public expenditure to long-run output growth, however, it should be the rate of change in the level of public expenditure that matters,33 a point that is analogous to the case noted above involving the level of taxation. The externality effect of public expenditure, in contrast, enhances growth by raising private sector productivity. Here, a high growth rate could be achieved by a higher level of such expenditure. In the recent endogenous growth literature, the focus has been on the stock of public infrastructure (or the level of services that flows from it) as a productive input;34 conceptually, however, there is no reason why this effect should be limited to infrastructure spending only. For example, public expenditures, such as those on elementary education and vocational training, that enhance human capital (a key variable in endogenous growth) could have a similar impact. The opposing natures of the crowding-out and externality effects implies that the structure of public expenditure, rather than merely its level, would be of considerable importance. In analyzing the composition of public expenditure, the traditional approach has been to divide it broadly into the categories of public consumption and public investment, with the idea that the former tends to retard, and the latter to promote, growth. While intuitively appealing, this classification can quickly become problematic. Many public investment projects could be wasteful, for example, in the sense that their marginal net present values could be negative for the society as a whole; at the same time, many public consumption expenditures, such as certain kinds of educational training, operations and maintenance spending on existing infrastructure, and even targeted funding for research and development activities, could be enor- mously beneficial for long-run growth. Hence, a more useful classification—one that has gained currency recently—would divide public expenditure into productive (that is, growth-inducing) and unproductive (that is, growth-retarding) categories, taking into consideration the levels and

32 This belief is often the rationale for advocating privatization of public enterprises. See World Bank (1995). 33 Ram (1986) made this point explicit in his model. 34 See, in particular, Aschauer (1989) and Barro (1990). The analytics of endogenous growth models incorporating public expenditure as a productive input were surveyed in Barro and Sala-i-Martin (1992).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 1 89 mixes of both the resources absorbed and the outputs produced by different expenditure programs.35 The usefulness of the productive-unproductive classification for growth analyses is particularly apparent in a dynamic context because it focuses one's attention on the impact of public expenditure on private savings and investment and, hence, capital accumulation. There are three dimensions to this impact. First, public expenditure needs to be financed, and this reduces resources for private savings.36 Second, to the extent that public expenditure improves private productivity, it stimulates private savings. Finally, the degree of complementarity or substitutability between public and private expenditure is important. The lower (higher) the complementarity (substitutability), the smaller its impact on private savings.37 The combined impact of these effects on private savings would suggest that the relationship between the level of public expenditure and growth is typically not monotonic. For a given degree of complementarity or substitutability, growth may be enhanced by public expenditure up to a point, after which the relationship between the two turns negative.38 This relationship has provided a basis for determining the growth-maximizing level of public expenditure, as well as of government intervention, in a decentralized economy.39 As with the case of taxation, the empirical evidence of the growth effects of public expenditure (as a share of GDP) is inconclusive. Using cross- country regressions, Ram (1986) found that, although growth in general is positively correlated with the rate of change in total public expenditure, it is negatively correlated with the level of such expenditure. This latter result was also obtained by Levine and Renelt (1992). When public expenditure is broadly disaggregated, there is a stronger indication that growth is negatively correlated with public consumption net of defense and education spending (Easterly and Rebelo,1993; Barro and Sala-i-Martin, 1995; and, to a certain extent, Levine and Renelt, 1992). One possible explanation for

35 For a recent development of this argument, see Devarajan, Swaroop, and Zou (1996). See also Ghu and others (1995) for a discussion of the various aspects of the productive-unproductive classification of public expenditure. One type of unproductive public expenditure that has received much attention recently is military spending. See, for example, Knight, Loayza, and Villanueva (1996). However, not all public expenditure programs are designed to promote growth. Hence, some public expenditures could be unproductive in the growth sense while effective in the sense of achieving their objectives. 36 Bradford (1975) emphasized the importance of knowing whether the financing source is consumption, private capital formation, or unemployed resources. See also the discussion below on budget policy. 37 There is a voluminous literature on this last aspect of public expenditure, stimulated by the classic analysis of Bailey (1971). For a recent treatment and review, see Karras (1994). 38 Public expenditure may also become increasingly wasteful after a certain point, as argued by Tanzi and Schuknecht (1995). 39 See, in particular, Barro (1990) and Jones, Manuelli, and Rossi (1993).

©International Monetary Fund. Not for Redistribution 190 VITO TANZI and HOWELL H. ZEE this negative relationship is that, in the aggregate, such public consumption is viewed by economic agents as a less-than-perfect substitute (or possibly even a complement) for private consumption, so private savings decline as a result. Karras (1994) found evidence of complementarity between public and private consumption. As regards more specific categories of public consumption, Knight, Loayza, and Villanueva (1996) found a significant adverse impact of military spending on growth,40 while Aschauer (1989) found that the impact of such spending on private sector productivity in the United States, although negative, is insignificant.41 A significantly positive impact on growth of public spending on education was found by Barro and Sala-i-Martin (1995), who interpreted the result to represent the growth effect of improved quality in human capital. However, this positive impact is also consistent with the Tanzi (1995) argument that such spending increases a country's ability to absorb technology from abroad and invent new technologies. Levine and Renelt (1992) found neither military nor public education expenditures as having a robust correlation with growth. The finding by Aschauer (1989) of a strong and positive correlation between nonmilitary public capital stock and private sector productivity in the United States has been widely cited as evidence of the importance of public investment in promoting growth. Of particular interest here is the identification of a subset of core infrastructure (utilities and transportation facilities) as having the greatest impact. In a cross-country setting, Easterly and Rebelo (1993) also obtained strong support for a positive correlation between growth and public investment, especially that in transportation and communications, but Levine and Renelt (1992) found that the growth effects of public investment are not robust. The difficulties noted above in properly estimating the growth effects of taxation clearly apply to public expenditure as well. Even if the correlation between growth and public expenditure (or a subset thereof) is found to be robust (in the sense that other relevant variables have been adequately controlled), the direction of causation underlying the correlation would still be unclear. Higher income growth may well generate higher demands for some or all types of public expenditure, which, in turn, may necessitate higher levels of taxation.42 Hence, it is at least plausible that the direction of

40 One of the first to investigate the relationship between military spending and growth was Benoit (1973 and 1978), who stimulated much follow-up research on the topic. Some of this literature was recently surveyed by Ram (1995). 41 In the United States, military spending has often produced technologies potentially beneficial to the whole economy. This is less likely to happen in other—in particular, developing—countries. 42 For example, higher growth may generate a higher demand for cars, which, in turn, may generate a higher demand for roads.

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 191 causation could run from growth to public expenditure and taxation. To be sure, most researchers are aware of this problem of reverse causation, but the empirical growth literature has so far not dealt with it satisfactorily. A further problem that has not been addressed in this literature is that the relationship between growth and fiscal variables may not be monotonic, either over the levels of the fiscal variables themselves or over income levels, or both. As noted above, it is analytically plausible to argue that increasing levels of public expenditure would first raise and then reduce growth. If countries pursue approximately growth-maximizing public expenditure policies, one would expect little correlation between growth and the level of public expenditure in a cross-country regression. Similarly, a case could be made that the growth effects of fiscal variables, if any, may well change direction as income rises.43 These and other problems suggest that there is much scope for further empirical research in disentangling the complex interactions among different fiscal variables.

Budget Policy

Another broad fiscal variable that could have implications for growth is budget policy, in the sense that the level of public revenue relative to that of public expenditure, that is, the budget balance, may have growth effects that are separate from those related to the absolute level of either taxation or public expenditure, as discussed above. One type of effect stems from the stability implications of budget imbalances; this is considered in Section II below. Another type is related to a possible behavioral response from the private sector triggered by such imbalances. If the private sector regards budget deficits (even if financed by debt) simply as taxes delayed, for example, then it may choose to increase its own savings to neutralize the public dissavings, thus leading to an unchanged level of national savings. Alterna- tively, budget deficits might not induce a response in private sector savings, in which case national savings would be reduced and growth hampered.44 The question of whether there is neutrality between debt and tax financing has been the focus of much recent research.45 A crucial condition for the

43 In an investigation of Wagner's law, Tanzi and Zee (1995) found the correlation between the levels of public wage expenditure and income to be positive for middle- and low-income countries and negative for high-income countries. 44 For the present discussion, assume that the public expenditure giving rise to the budget deficits does not entirely consist of public investment. 45 This neutrality is commonly referred to as the Ricardian equivalence, since the idea can be traced back to the writings of Ricardo, as well as to some early Italian public finance literature (see Buchanan (1958) for an account). Its modern revival is usually credited to Barro (1974). Bailey (1971) contained a clear discussion of its implications.

©International Monetary Fund. Not for Redistribution 192 VITO TANZI and HOWELL H. ZEE neutrality to hold is that, when the planning horizons of economic agents are finite (as would be the case under the intuitively appealing notion of life- cycle savings), there are operative private transfers (gifts and bequests) between generations, so that the implied tax burden of public dissavings on future generations is not ignored by the current generation. It is now widely recognized that strict neutrality would also depend on the absence of a host of other factors, such as tax distortions, income uncertainties, and imperfect credit markets.46 While conceptually intriguing, the importance of the above neutrality result clearly lies in the empirical evidence. Unfortunately, similar to the case of the growth effects of taxation and public expenditure, the empirical support for debt neutrality is mixed.47 On the whole, the evidence, particularly from cross-country data, seems to suggest that the response by private sector savings to public sector dissavings does not completely neutralize the latter. Direct tests of the impact of budget deficits on growth based on cross- country data have also been recently performed by a number of studies: Easterly and Rebelo (1993) found the correlation between the two significant and negative; Martin and Fardmanesh (1990) found the correlation significant and negative only for middle-income countries; and Levine and Renelt (1992) found the correlation fragile.

II. Stability and Growth

Tax and Expenditure Policies

From the point of view of stability, the growth implications of tax and public expenditure policies are similar and, therefore, can be discussed jointly. The most direct link between tax policy and growth has to do with the volatility injected into the returns to an investment project by an uncertain tax regime. In the recent literature on investment under uncertainty, it

46 The literature on the Ricardian equivalence is too voluminous to even attempt a partial survey here. Recent assessments of relevant issues were provided by Leiderman and Blejer (1988), as well as by two of the central debaters, Barro (1989), a proponent, and Bernheim (1989), a critic. In a recent analysis, Bailey (1993) derived the important result that, if taxes are capitalized into property values and properties are part of the bequest from one generation to another, then (approximate) Ricardian equivalence would hold even if generations are not linked by transfers over an infinite horizon. 47 In testing Ricardian equivalence, empirical works have largely focused on the impact of budget deficits on one or more of the following three variables: private consumption-savings; intergenerational transfers; and interest rates. For reviews of empirical evidence, see Bernheim (1987) and the associated comments of discus- sants; Leiderman and Blejer (1988); and Barro (1989).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 1 93 has been established that, since most projects are to some extent irreversible, increased uncertainty about their returns would generally lead to a reduction (postponement) in investment.48 Hence, uncertainty about the tax regime, which, in turn, leads to uncertain after-tax returns, is likely to discourage investment and hamper growth.49 Tax regime uncertainty could be attributable to a number of factors. The difficulty in forecasting the direction of prospective tax reforms under political debate is one obvious example. Another example would be the possible changes to the tax regime necessitated by unexpected shocks to income or interest rates, or both, or by unforeseen public expenditure needs. A tax system that is not indexed or has significant collection lags, or both, would also give rise to uncertain real effective tax rates in an unstable, inflationary environment.50 However, one type of uncertainty unrelated to such unan- ticipated factors could nevertheless arise even in a framework of an optimizing government whose objective coincides with that of the representative economic agent: if the optimal tax regime changes from one period to the next, there would be uncertainty as to whether the government would maintain the same regime over time. This problem, which is generally known as the time inconsistency of optimal policy,51 can best be understood intuitively by considering a simple two-period model with endogenous savings and labor supply. The government in period 1 optimizes and determines the optimal tax rates on labor and capital for period 2. When period 2 comes around, savings undertaken in period 1 have become fixed capital (sunk cost) and, if taxed, would not give rise to any excess burden. Hence, it would be optimal for the government in period 2 to tax only capital, which is a policy that, in general, will

48 This result comes about because it may pay to wait for a favorable state of nature. Depending on the type of the uncertainty, however, uncertain returns could, in some circumstances, stimulate investment. One reason is that the act of invest- ing itself sometimes provides additional information that could act to reduce the uncertainty; another reason is that a mean-preserving spread of variance (that is, an increase in variance with the same mean) with respect to returns would increase the expected value of a project, if the valuation function displays diminishing marginal value of returns. For a recent comprehensive treatment of this literature, see Dixit and Pindyck (1994); for an illustration that the impact of uncertainty on investment is dependent on the way that tax regime uncertainty is modeled, see Hassett and Metcalf (1994). 49 Recent theoretical analyses that lend support to this conclusion include Aizenman and Marion (1993) and Dixit and Pindyck (1994). 50 Growth effects of inflation are considered below in connection with budget policy. For discussions of the impact of inflation on real tax revenue in the presence of collection lags, see Tanzi (1977 and 1978). 51 The vast literature on the time-inconsistency problem had its origin largely in the seminal work of Kydland and Prescott (1977).

©International Monetary Fund. Not for Redistribution 194 VITO TANZI and HOWELL H. ZEE

not be the same as the optimal policy set by the government in period 1.52 While it is tempting to interpret the foregoing result as simply another consequence of investment being irreversible, the time-inconsistency problem is in fact fairly general and can occur even in models without capital.53 The likely adverse impact on growth of tax regime uncertainty, irrespec- tive of its origin, naturally raises questions about the possible ways by which the uncertainty (or at least some types of such uncertainty) could be alleviated. For the time-inconsistency problem, various potential mechanisms have been advanced to precommit the government in a given period to maintaining an optimal policy over time.54 If the uncertainty stems instead from unexpected shocks to income or interest rates, or both, or from unforeseen public expenditure needs, then an appropriate debt-management policy could obviate the need for altering the tax regime as a response to such occurrences.55 Finally, if inflation is the source of uncertainty about the real tax burden, the first-best solution is clearly to implement appropriate policies to reduce macroeconomic instabilities; indexing the tax system and adopting administrative measures to reduce tax collection lags are possible second-best solutions. Based on cross-country regressions of a large sample of developing countries, Aizenman and Marion (1993) presented empirical evidence that

52 See Fischer (1980) for a particularly illuminating discussion of this example. Kydland and Prescott (1980) examined essentially the same example in greater gen- eralities. The same demonstration can be made with respect to human capital investment, where a government could find it optimal to tax such investment lightly in the early periods of an individual's life, but to tax the returns from human capital heavily once the capital has been formed. 53 See, for example, the well-known demonstration by Lucas and Stokey (1983). As it turns out, in a typical intertemporal model with endogenous labor supply but with no capital, whether an optimal tax regime is time inconsistent or not depends critically on the tax instruments at the disposal of the government. If only an income tax (either on wages or on interest income, or both) is available, the outcome is time inconsistent (Turnovsky and Brock, 1980; and Lucas and Stokey, 1983). Rogers (1987) found that a consumption tax is time consistent under a Cobb-Douglas utility function. When the utility function has a general specification, however, Zee (1994) showed that an optimal tax regime would be time consistent only if both the income and consumption taxes are available. Moreover, Zee (1994) also showed that an optimal time-consistent tax structure could be distortive. 54 These mechanisms include imposing on the government reputational constraints (Barro and Gordon, 1983), social contractual obligations (Kotlikoff, Persson, and Svensson, 1988), and particular structures of government debt (Lucas and Stokey, 1983; and Persson, Persson, and Svensson, 1987). 55 This is the intertemporal consumption-smoothing argument of Barro (1979 and 1995b). By varying the level and structure of public debt, tax rates could be smoothed over time and over states of nature to minimize the intertemporal excess burden of distortive taxes. The ability to restructure public debt varies, of course, across different countries; in many developing countries, this ability is often quite limited.

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 195 suggests that, to varying degrees, there is a significant and negative correlation between growth and uncertainty in a number of fiscal variables, such as levels of revenue, public expenditure, and budget deficits.56 Easterly and Rebelo (1993) also found that the standard deviation in the ratio of domestic tax revenue to consumption and investment had a significant and negative impact on growth.

Budget Policy

Assume for the moment that monetary financing of budget imbalances is not available. In such circumstances, the evolution of the stock of real public debt is entirely governed by the path of cumulative real budget imbalances over time. If the economy is dynamically efficient (that is, its long-run real interest rate exceeds its long-run growth rate) and the government is to be solvent, then any indebtedness of the government would have to be eliminated eventually through appropriate budget policy that would bring the present value of the stock of public debt at some future date (which could be infinity) to zero.57 An important implication of this solvency requirement for the conduct of budget policy is that the government would be obligated to accumulate a sufficient level of net primary budget surpluses (in present-value terms) over time to pay off its initial debt.58 This implication, in turn, provides a natural basis for evaluating whether current budget policy, if maintained, is sustainable (Wilcox, 1989) and, if not, to what extent tax rates must be raised (for a given path of public expenditure) to ensure government solvency (Blanchard and others, 1990). The relevance of policy sustainability for growth is twofold. If current policy is deemed to be unsustainable, then either a regime change in tax (and/or expenditure) policy would be expected to occur, or recourse would be had to monetary financing. A regime change would increase policy

56 Aizenman and Marion (1993) measured uncertainty in a variable by the standard deviation of the residuals from a first-order autoregressive process of that variable. 57 This is a widely invoked requirement in the literature. Notable recent examples are Wilcox (1989) and Blanchard and others (1990). 58 Whether an economy is dynamically efficient or not is an empirical question; theory cannot rule out the possibility that its long-run growth rate could exceed the long-run real interest rate (see Diamond (1965)). If that is the case, the solvency requirement is no longer meaningful. This is because the government could sustain some positive stock of public debt forever simply through additional borrowing, without having to run budget surpluses (a "Ponzi finance scheme"), because by assumption the debt-service cost is lower than income growth. The determination of a sustainable positive stock of public debt was examined by Zee (1988). Recently, however, Abel and others (1989) found that most capitalistic economies are dynamically efficient.

©International Monetary Fund. Not for Redistribution 196 VITO TANZI and HOWELL H. ZEE uncertainty, whose impact on growth has already been discussed. Monetary financing would lead to inflation, which raises the important issue of the possible growth effects of inflation. There are a number of conceptual links between inflation and growth. One of the oldest is built on the idea that inflation can be viewed as a distortive tax on real money balances and, therefore, has efficiency consequences in much the same way as the other, more traditional distortive taxes discussed in Section I.59 As pointed out in that discussion, from a purely allocative perspective, any adverse growth impact from distortive taxes would have to stem from increases in the level of taxation, in this case an acceleration in inflation.60 From a stability perspective, however, arguments have been advanced that higher inflation rates would lead to greater uncertainty about future inflation (Okun, 1971; and Friedman, 1977); thus larger efficiency losses would result simply from higher levels of inflation. The impact of inflation on growth has also been examined directly in growth models. In the earlier growth literature, the focus was on the issue of the superneutrality of money, that is, whether inflation could affect the steady state capital-labor ratio, rather than on the growth effects of inflation per se (as the long-run growth in these models is exogenous).61 With endogenous growth models, however, a number of direct channels through which inflation could affect growth open up, such as the potential impacts of inflation on both physical and human capital accumulation, as well as the interactions between inflation and a tax system that is based on nominal rather than real magnitudes.62 On the whole, however, the theoretical results in both the old and new growth models seem to be too dependent on model specifications to render them useful as yet for policy purposes. While further theoretical explorations of the growth effects of inflation are certainly called for, increasing empirical evidence suggests that there

59 The seminal work on measuring the welfare cost of inflation as the excess burden of a tax in a partial equilibrium framework was that by Bailey (1956), from which a vast literature ensued. The integration of the inflation tax into a standard optimal taxation model was first carried out by Phelps (1973). Chari, Christiano, and Kehoe's (1996) recent reexamination of this literature clarified a number of important theoretical points concerning the relationship between the inflation tax and other commodity taxes. 60 This point notwithstanding, the recent study by Lucas (1994) indicated that, employing the Bailey (1956) framework, the welfare cost of inflation in the United States is much higher than commonly believed. A large welfare cost was also found by Dotsey and Ireland (1996), who extended the Bailey-type measure into a general equilibrium framework with endogenous labor supply. 61 The voluminous literature on the superneutrality of money was recently surveyed by Orphanides and Solow (1990). 62 Jones and Manuelli (1995) addressed many of these issues. Inflation can render a previously optimal tax system suboptimal through a variety of channels: different collection lags of different taxes, differential tax impacts on different tax bases, and nonproportional tax rates.

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 197 exists a significant and negative correlation between high inflation and growth.63 Based on panel data, the inflation threshold above which growth effects become significant ranges from 8 percent to 40 percent.64 Further- more, Judson and Orphanides (1996) found that inflation volatility is robustly and negatively correlated with growth at all levels of inflation. Hence, there seems to be a compelling case for believing that an expansionary budget policy resulting in high rates of inflation would most likely exact a growth penalty.65

III. Income Distribution and Growth

While economists may disagree on the relative importance of the allocative and distributional objectives of fiscal policy, most will accept the proposition that some trade-off is involved in pursuing the two policy objectives. The trade-off stems, of course, from the disincentive effects of distortive taxes that are required to finance direct or indirect transfer payments from the rich to the poor. Indeed, in a static framework, it is easy to demonstrate that, under fairly general assumptions about (heterogeneous) individual preferences regarding income and work effort, the efficiency cost of pursuing an egalitarian policy could be prohibitively high.66 Hence, in the traditional view, policies effecting a redistribution of income toward equality would exact an increase in the price of (aggregate) output loss that is likely to be larger than the reduction in income inequality achieved by such policies. When extended to a dynamic context, such a view leads quite naturally to the conclusion that there is an increasing marginal cost, in terms of growth forgone, of income redistribution, on account of the saving- depressing effects of taxation. The validity of this traditional view has, however, been challenged recently by several strands of research. One strand argues that redistributive taxation and the expenditure that it finances are a form of social insurance

63 For recent surveys of the empirical literature, see Briault (1995) and Thornton (1996). 64 The threshold was found to be 8 percent in Sarel (1996), 10 percent in Judson and Orphanides (1996), 15 percent in Barro (1995a), and 40 percent in Bruno and Easterly (1995). 65 If the inflationary effects of an expansionary budget policy are countered by a restrictive monetary policy, then the growth penalty will be exacted through high interest rates. Moreover, even though the statistical relationship between growth and low inflation is weak, Feldstein (1996) showed that the interactions between an existing distortive tax system and inflation would result in substantial welfare losses, even at low inflation rates. 66 For a particularly simple illustration of this result, see Baumol and Fischer (1969).

©International Monetary Fund. Not for Redistribution 198 VITO TANZI and HOWELL H. ZEE over an economic agent's lifetime against certain types of risk for which private insurance may not be available. Consequently, redistributive policies could stimulate productive risk taking and output growth, although such behavior does not necessarily result in greater equality in the after-tax distribution of income.67 A second strand emphasizes the importance of various aspects of financial market imperfections for growth. A central idea here is that the potential productivity of the poor cannot be fully realized unless they are given the opportunity to participate in financial markets. If financial markets were perfect, the poor would be able to borrow against their future earnings to acquire, for example, basic needs (including nutrition, health care, and education) and human capital. In the absence of such markets, however, redistributive policies are needed to raise the poor's standard of living at least beyond some threshold so that they can become productive members of society and, consequently, contribute to output growth.68 Once gainfully employed, the poor could then begin to acquire assets, accumulate human capital, and gain access to financial markets to further raise their earnings potential. The financial markets, in turn, by benefiting from the increased participation in the intermediation process by economic agents, would become more developed, and the growth prospects for the whole economy would be enhanced as a result.69 An implication for fiscal policy from this strand of literature is clearly that redistributive policies that result in less income inequality could well promote growth. Yet another strand of research focuses on the impact of various political economy factors on growth. While model structures differ across different studies, at the core of this literature is the idea that income distribution affects political outcomes, which, in turn, affect the kind of policies that are actually implemented through the voting process.70 By invoking the stan-

67 For this argument, see Sinn (1995 and 1996). The link between redistributive taxation and social insurance was explored earlier in Eaton and Rosen (1980) and Varian (1980). While the connection between taxation and risk taking is not new, the existing literature on it by and large focuses on the impact of taxation on portfolio investment decisions (see Atkinson and Stiglitz (1980) for a review) rather than on issues of income redistribution. 68 There is a large basic-needs-related literature in development economics. See, for example, Streeten and others (1981). A recent analytical treatment of the linkage between such needs and redistribution and growth was that by Dasgupta (1993). 69 A notable recent study on the growth effects of income distribution in a framework of human capital accumulation constrained by imperfect financial markets was that by Galor and Zeira (1993). Greenwood and Jovanovic (1990) stressed the importance of the interrelationship among income distribution, financial market development, and growth. 70 For various surveys of this literature, which cover issues that go beyond fiscal policy in a number of directions, see Perotti (1992 and 1994), Persson and Tabellini (1992), Alesina and Perotti (1994), and Verdier (1994).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 199 dard median-voter theorem, this literature is able to demonstrate that the greater the inequality of income, the higher will be the voted level of taxation, either for the provision of public goods (Alesina and Rodrik, 1994) or for purely redistributive transfers (Persson and Tabellini, 1994), as a poorer median voter faces a lower tax price of public expenditure than a richer one. Since higher taxes, in turn, lower growth by depressing either physical or human capital accumulation, or both, a direct causal effect of income distribution on growth is thus established. This political economy approach takes as given the initial distribution of income (or wealth); consequently, it cannot be used to explain how such a distribution is arrived at in the first place. A potential solution to this limi- tation can be found in an older but voluminous literature, associated with the seminal work of Kuznets (1966), that focused on just the reverse causation, that is, the impact of growth on income distribution. Kuznets (1966) argued that growth would first increase income inequality and then reduce it after some level of income is reached. This relationship can often be derived from a two-sector economy setting comprising, say, a high-growth urban sector and a low-growth rural sector. As labor migrates from the rural to the urban sector with economic development, various conventional measures of inequality would first rise and then fall.71 By combining these two bodies of literature, and perhaps in conjunction with elements of the literature on financial market imperfections noted above, it is possible to derive a two-way causal relationship between income distribution and growth.72 Hence, the trade-off between the allocative and distributional objectives of fiscal policy is not absolute: growth with redistribution is possible.73 Most of the above-cited studies employing the political economy approach present cross-country empirical evidence, based on various samples of developed and developing countries. This evidence supports to varying degrees a negative correlation between income inequality (measured in some base year close to the beginning of the sample period over which growth rates are computed) and growth. Clarke (1995) recently confirmed that this negative correlation is robust across a broad sample of countries

71 For recent surveys of this literature, see Adelman and Robinson (1989) and Anand and Kanbur (1993). Bourguignon (1990) recently found, however, that the Kuznets relationship does not hold up well under a more general two-sector specification with different classes of agents and endogenous terms of trade between the two sectors. 72 A recent attempt in this direction was made by Perotti (1993b), who considered tax and transfer policies explicitly as voting outcomes in a model with imperfect financial markets. He obtained versions of a Kuznets-like inverted-U relationship between degrees of income inequality and income levels. 73 This is the central policy conclusion reached by Bruno, Ravallion, and Squire (1996).

©International Monetary Fund. Not for Redistribution 200 VITO TANZI and HOWELL H. ZEE and with alternative measures of inequality, after controlling for other variables that are standard in the endogenous growth literature. While the evidence on the adverse impact of initial income inequality on growth seems compelling, what remains unclear is the precise channels through which this impact operates. In the models constructed by Alesina and Rodrik (1994) and Persson and Tabellini (1994), for example, a high degree of income inequality generates heavy taxation for high public investment expenditure or large public transfers, but Perotti (1993a) found rather weak support in the data for these chains of events. Recently, Alesina and Perotti (1996) identified an alternative transmission mechanism: income inequality creates social unrest and political instability, which, in turn, depress investment and growth. Their empirical analyses, which involved the construction of an index of social and political instability (SPI), found cross-country evidence for negative correlations both between income equality and SPI and between SPI and investment. This latter finding is consistent with other empirical studies that found a negative correlation between political instability and growth, for example, Barro (1991) and Mauro(1995).

IV. Concluding Remarks

Economists working in public finance have always believed that fiscal policy, interpreted as the manipulation of fiscal instruments to achieve specific objectives, can affect economic growth. This belief is reflected in the title of many books and articles that refer to the assumed connection between fiscal policy and economic growth. This connection has been thought to originate from various channels, such as the negative effect of distortive taxes, the negative effect of progressive taxes on the propensity to save, and the scope for mobilizing resources through use of additional resources from higher taxation to increase the level of public investment. While public finance economists seemed to have no doubts that they could influence growth through the policy changes that they recommended, the prevailing neoclassical growth theory did not leave much role to policy, except for relatively short-run effects on growth. This dichotomy resulted in part from various assumptions implicit in the theory and in part from the different time horizons contemplated by the public finance economists and the growth theorists. For example, while the neoclassical growth theory gives no role to policy for long-run growth, its definition of the long run could be long enough to leave ample scope for the effect of policy over the time horizon of interest to most governments and individuals. The present paper has attempted to consider in a systematic and comprehensive way the relationship between various public finance instruments

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 201 and the growth of countries' economies. It has surveyed a large body of literature, both theoretical and empirical, in an attempt to reach conclusions as to the way in which taxes, public spending, and budgetary policy can influence growth by affecting the allocation of resources, the stability of the economy, and the distribution of income. The literature is very extensive and very rich, and, at times, it is hard to interpret. Yet it is much less defin- itive on some of these issues than one would have thought. In particular, the empirical literature is somewhat disappointing in its support of theoretically reached conclusions. Despite the lack of robust results in the empirical literature, the conclusion of this paper has been that, when interpreted from the perspective of the new endogenous growth theory, fiscal policy could play a fundamental role in affecting the long-run growth performance of countries. Thus, economists should not hesitate to recommend changes in the instruments of public finance in the direction that theory has deemed important for enhancing growth, such as the adoption of policies to improve the neutrality of taxation, promote human capital accumulation, and lessen income inequality.

REFERENCES

Abel, Andrew B., and others, 1989, "Assessing Dynamic Efficiency: Theory and Evidence," Review of Economic Studies, Vol. 56 (January), pp. 1-20. Adelman, Irma, and Sherman Robinson, 1989, "Income Distribution and Develop- ment," in Handbook of Development Economics, Vol. II, ed. by Hollis Chenery and T.N. Srinivasan (Amsterdam and New York: North-Holland), pp. 949-1003. Aiyagari, S. Rao, 1990, "Deflating the Case for Zero Inflation" Quarterly Review, Federal Reserve Bank of Minneapolis, Vol. 14 (Summer), pp. 2-11. Aizenman, Joshua, and Nancy P. Marion, 1993, "Policy Uncertainty, Persistence and Growth," Review of International Economics, Vol. 1 (June), pp. 145-63. Alesina, Alberto, and Roberto Perotti, 1994, "The Political Economy of Growth: A Critical Survey of the Recent Literature," World Bank Economic Review, Vol. 8 (September), pp. 351-71. , 1996, "Income Distribution, Political Instability, and Investment," Euro- pean Economic Review, Vol. 40 (June), pp. 1203-28. Alesina, Alberto, and Dani Rodrik, 1994, "Distributive Politics and Economic Growth," Quarterly Journal of Economics, Vol. 109 (May), pp. 465-90. Anand, Sudhir, and S.M. Ravi Kanbur, 1993, "The Kuznets Process and the Inequality-Development Relationship," Journal of Development Economics, Vol. 40 (February), pp. 25-52. Arrow, Kenneth J., 1962, "The Economic Implications of Learning by Doing," Review of Economic Studies, Vol. 29 (June), pp. 155-73. Aschauer, David A., 1989, "Is Public Expenditure Productive?" Journal of Mone- tary Economics, Vol. 23 (March), pp. 177-200.

©International Monetary Fund. Not for Redistribution 202 VITO TANZI and HOWELL H. ZEE

Atkinson, Anthony B., and Agnar Sandmo, 1980, "Welfare Implications of the Tax- ation of Savings," Economic Journal, Vol. 90 (September), pp. 529-49. Atkinson, Anthony B., and Joseph E. Stiglitz, 1980, Lectures on Public Economics (New York: McGraw-Hill). Auerbach, Alan J., 1985, "The Theory of Excess Burden and Optimal Taxation," in Handbook of Public Economics, Vol. I, ed. by Alan J. Auerbach and Martin Feldstein (Amsterdam and New York: North-Holland), pp. 61-127. Bailey, Martin J., 1956, "The Welfare Cost of Inflationary Finance," Journal of Political Economy, Vol. 64 (April), pp. 93-110. , 1971, National Income and the Price Level: A Study in Macroeconomic Theory (New York: McGraw-Hill). 1993, "Note on Ricardian Equivalence," Journal of Public Economics, Vol. 51 (July), pp. 437-46. Barro, Robert J., 1974, "Are Government Bonds Net Wealth?" Journal of Political Economy, Vol. 82 (November/December), pp. 1095-117. , 1979, "On the Determination of the Public Debt," Journal of Political Economy, Vol. 87 (October), pp. 940-71. , 1989, "The Ricardian Approach to Budget Deficits," Journal of Economic Perspectives, Vol. 3 (Spring), pp. 37-54. , 1990, "Government Spending in a Simple Model of Endogenous Growth," Journal of Political Economy, Vol. 98 (October), pp. S103-25. , 1991, "Economic Growth in a Cross Section of Countries," Quarterly Journal of Economics, Vol. 106 (May), pp. 407-43. , 1995a, "Inflation and Economic Growth," Bank of England Quarterly Bul- letin, Vol. 35 (May), pp. 166-76. , 1995b, "Optimal Debt Management," NBER Working Paper No. 5327 (Cambridge, Massachusetts: National Bureau of Economic Research). , and David B. Gordon, 1983, "Rules, Discretion and Reputation in a Model of Monetary Policy," Journal of Monetary Economics, Vol. 12 (July), pp. 101-21. Barro, Robert J., and Xavier Sala-i-Martin, 1992, "Public Finance in Models of Economic Growth," Review of Economic Studies, Vol. 59 (October), pp. 645-61. —, 1995, Economic Growth (New York: McGraw-Hill). Baumol, William J., 1986, "Productivity Growth, Convergence, and Welfare: What the Long-Run Data Show," American Economic Review, Vol. 76 (December), pp. 1072-85. , 1990, "Entrepreneurship: Productive, Unproductive, and Destructive," Journal of Political Economy, Vol. 98 (October), pp. 893-921. , and Dietrich Fischer, 1969, "The Output Distribution Frontier: Alterna- tives to Income Taxes and Transfers for Strong Equality Goals," American Economic Review, Vol. 69 (September), pp. 514-25. Benoit, Emile, 1973, Defense and Economic Growth in Developing Countries (Boston: D.C. Heath).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 203

, 1978, "Growth and Defense in Developing Countries," Economic Devel- opment and Cultural Change, Vol. 26 (January), pp. 271-80. Bernheim, B. Douglas, 1987, "Ricardian Equivalence: An Evaluation of Theory and Evidence," in NBER Macroeconomics Annual, ed. by Stanley Fischer (Cambridge, Massachusetts: MIT Press), pp. 263-304. , 1989, "A Neoclassical Perspective on Budget Deficits," Journal of Eco- nomic Perspectives, Vol. 3 (Spring), pp. 55-72. Blanchard, Olivier, and others, 1990, "The Sustainability of Fiscal Policy: New Answers to an Old Question," OECD Economic Studies, Vol. 15 (Autumn), pp. 7-36. Bourguignon, Francois, 1990, "Growth and Inequality in the Dual Model of Devel- opment: The Role of Demand Factors," Review of Economic Studies, Vol. 57 (April), pp. 215-28. Bradford, David F., 1975, "Constraints on Government Investment Opportunities and the Choice of Discount Rate," American Economic Review, Vol. 65 (December), pp. 887-99. Briault, Clive B., 1995, "The Costs of Inflation," Bank of England Quarterly Bul- letin, Vol. 35 (February), pp. 33-45. Bruno, Michael, and William Easterly, 1995, "Inflation Crisis and Long-Run Growth," NBER Working Paper No. 5209 (Cambridge, Massachusetts: National Bureau of Economic Research). Bruno, Michael, Martin Ravallion, and Lyn Squire, 1996, "Equity and Growth in Developing Countries: Old and New Perspectives on the Policy Issues," World Bank Policy Research Working Paper No. 1563 (Washington: World Bank). Buchanan, James M., 1958, Public Principles of Public Debt: A Defense and Restatement (Homewood, Illinois: R.D. Irwin). Cass, David, 1965, "Optimal Growth in an Aggregative Model of Capital Accu- mulation," Review of Economic Studies, Vol. 32 (July), pp. 233-40. Chamley, Christophe, 1986, "Optimal Taxation of Capital Income in General Equi- librium with Infinite Lives," Econometrica, Vol. 54 (May), pp. 607-22. Chari, V.V., Lawrence J. Christiano, and Patrick J. Kehoe, 1996, "Optimality of the Friedman Rule in Economies with Distorting Taxes," Journal of Monetary Economics, Vol. 37 (April), pp. 203-23. Chu, Ke-young, and others, 1995, Unproductive Public Expenditures: A Pragmatic Approach to Policy Analysis, IMF Pamphlet Series, No. 48 (Washington: International Monetary Fund). Clarke, George R.G., 1995, "More Evidence on Income Distribution and Growth," Journal of Development Economics, Vol. 47 (August), pp. 403-27. Dasgupta, Partha, 1993, An Inquiry into Weil-Being and Destitution (Oxford: Clarendon Press). DeLong, J. Bradford, and Lawrence H. Summers, 1991, "Equipment Investment and Economic Growth," Quarterly Journal of Economics, Vol. 106 (May), pp. 445-502. Devarajan, Shantayanan, Vinaya Swaroop, and Heng-fu Zou, 1996, "The Compo- sition of Public Expenditure and Economic Growth," Journal of Monetary Economics, Vol. 37 (April), pp. 313-44.

©International Monetary Fund. Not for Redistribution 204 VITO TANZI and HOWELL H. ZEE

Diamond, Peter A., 1965, "National Debt in a Neoclassical Growth Model," Amer- ican Economic Review, Vol. 55 (December), pp. 1126-50. , and James Mirrlees, 1971, "Optimal Taxation and Public Production II: Tax Rules," American Economic Review, Vol. 61 (June), pp. 261-78. Dixit, Avinash K., and Robert S. Pindyck, 1994, Investment Under Uncertainty (Princeton, New Jersey: Princeton University Press). Dotsey, Michael, and Peter Ireland, 1996, "The Welfare Cost of Inflation in General Equilibrium," Journal of Monetary Economics, Vol. 37 (February), pp. 29-47. Easterly, William, and Sergio Rebelo, 1993, "Fiscal Policy and Economic Growth: An Empirical Investigation," Journal of Monetary Economics, Vol. 32 (December), pp. 417-58. Eaton, Jonathan, and Harvey S. Rosen, 1980, "Optimal Redistributive Taxation and Uncertainty," Quarterly Journal of Economics, Vol. 95 (September), pp. 357-64. Engen, Eric M., and Jonathan Skinner, 1992, "Fiscal Policy and Economic Growth," NBER Working Paper No. 4223 (Cambridge, Massachusetts: National Bureau of Economic Research). Feldstein, Martin, 1996, "The Costs and Benefits of Going from Low Inflation to Price Stability," NBER Working Paper No. 5469 (Cambridge, Massachusetts: National Bureau of Economic Research). Fischer, Stanley, 1980, "Dynamic Inconsistency, Cooperation, and the Benevolent Dissembling Government," Journal of Economic Dynamics and Control, Vol. 2 (February), pp. 93-107. Friedman, Milton, 1977, "Nobel Lecture: Inflation and Unemployment," Journal of Political Economy, Vol. 85 (June), pp. 451-72. Galor, Oded, and Joseph Zeira, 1993, "Income Distribution and Macroeconomics," Review of Economic Studies, Vol. 60 (January), pp. 35-52. Greenwood, Jeremy, and Boyan Jovanovic, 1990, "Financial Development, Growth, and the Distribution of Income," Journal of Political Economy, Vol. 98 (October), pp. 1076-107. Hassett, Kevin, and Gilbert E. Metcalf, 1994, "Investment with Uncertain Tax Policy: Does Random Tax Policy Discourage Investment?" NBER Working Paper No. 4780 (Cambridge, Massachusetts: National Bureau of Economic Research). International Monetary Fund, 1996, World Economic Outlook, May 1996: A Sur- vey by the Staff of the International Monetary Fund, World Economic and Financial Surveys (Washington). Jones, Larry E., and Rodolfo E. Manuelli, 1995, "Growth and the Effects of Infla- tion," Journal of Economic Dynamics and Control, Vol. 19 (November), pp. 1405-28. , and Peter E. Rossi, 1993, "Optimal Taxation in Models of Endogenous Growth," Journal of Political Economy, Vol. 101 (June), pp. 485-517. Judson, Ruth, and Athanasios Orphanides, 1996, "Inflation, Volatility and Growth," Finance and Economics Discussion Series, No. 96/19 (Washington: Board of Governors of the Federal Reserve System).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 205

Kaldor, Nicholas, 1961, "Capital Accumulation and Economic Growth," in The Theory of Capital, ed. by Friedrich A. Lutz and Douglas C. Hague (New York: St. Martin's Press), pp. 177-222. Karras, Georgios, 1994, "Government Spending and Private Consumption: Some International Evidence," Journal of Money, Credit, and Banking, Vol. 26 (February), pp. 9-22. Knight, Malcolm, Norman Loayza, and Delano Villanueva, 1996, "The Peace Dividend: Military Spending Cuts and Economic Growth," Staff Papers, Inter- national Monetary Fund, Vol. 43 (March), pp. 1-37. Koopmans, Tjalling C., 1965, "On the Concept of Optimal Economic Growth," in Study Week on the Econometric Approach to Development Planning, by Pontificia Accademia delle Scienze (Amsterdam and Chicago: North- Holland), pp. 225-87. Kotlikoff, Laurence J., Torsten Persson, and Lars E.O. Svensson, 1988, "Special Contracts as Assets: A Possible Solution to the Time-Consistency Problem," American Economic Review, Vol. 78 (September), pp. 662-77. Kuznets, Simon S., 1966, Modern Economic Growth: Rate, Structure, and Spread (New Haven, Connecticut: Yale University Press). Kydland, Finn E., and Edward C. Prescott, 1977, "Rules Rather Than Discretion: The Inconsistency of Optimal Plans," Journal of Political Economy, Vol. 85 (June), pp. 473-91. , 1980, "Dynamic Optimal Taxation, Rational Expectations and Optimal Control," Journal of Economic Dynamics and Control, Vol. 2 (February), pp. 79-91. Leiderman, Leonardo, and Mario I. Blejer, 1988, "Modeling and Testing Ricardian Equivalence: A Survey," Staff Papers, International Monetary Fund, Vol. 35 (March), pp. 1-35. Levine, Ross, and David Renelt, 1991, "Cross-Country Studies of Growth and Policy: Methodological, Conceptual, and Statistical Problems," World Bank Policy Research Working Paper No. 608 (Washington: World Bank). , 1992, "A Sensitivity Analysis of Cross-Country Growth Regressions," American Economic Review, Vol. 82 (September), pp. 942-63. Lucas, Robert E., Jr., 1988, "On the Mechanics of Development Planning," Jour- nal of Monetary Economics, Vol. 22 (July), pp. 3—42. , 1990, "Supply-Side Economics: An Analytical Review," Oxford Economic Papers, Vol. 42 (April), pp. 293-316. , 1994, "On the Welfare Cost of Inflation," CEPR Technical Paper No. 394 (London: Centre for Economic Policy Research). -, and Nancy L. Stokey, 1983, "Optimal Fiscal and Monetary Policy in an Economy Without Capital," Journal of Monetary Economics, Vol. 12 (July), pp. 55-93. Mankiw, N. Gregory, David Romer, and David N. Weil, 1992, "A Contribution to the Empirics of Economic Growth," Quarterly Journal of Economics, Vol. 107 (May), pp. 407-37. Martin, Ricardo, and Mohsen Fardmanesh, 1990, "Fiscal Variables and Growth: A Cross-Sectional Analysis," Public Choice, Vol. 64 (March), pp. 239-51.

©International Monetary Fund. Not for Redistribution 206 VITO TANZI and HOWELL H. ZEE

Mauro, Paolo, 1995, "Corruption and Growth," Quarterly Journal of Economics, Vol. 110 (August), pp. 681-712. Mendoza, Enrique G., Gian Maria Milesi-Ferretti, and Patrick Asea, forthcoming, "On the Ineffectiveness of Tax Policy in Altering Long-Run Growth: Harberger's Superneutrality Conjecture," Journal of Public Economics. Mendoza, Enrique G., Assaf Razin, and Linda L. Tesar, 1994, "Effective Tax Rates in Macroeconomics: Cross-Country Estimates of Tax Rates on Factor Incomes and Consumption," Journal of Monetary Economics, Vol. 34 (December), pp. 297-323. Murphy, Kevin M., Andrei Shleifer, and Robert W. Vishny, 1989, "Industrializa- tion and the Big Push," Journal of Political Economy, Vol. 97 (October), pp. 1003-26. , 1991, "The Allocation of Talent: Implications for Growth," Quarterly Journal of Economics, Vol. 106 (May), pp. 503-30. Musgrave, Richard A., 1959, The Theory of Public Finance: A Study in Public Economy (New York: McGraw-Hill). Nordhaus, William D., and James Tobin, 1973, "Is Growth Obsolete?" in The Mea- surement of Economic and Social Performance, Vol. 38, Studies in Income and Wealth, ed. by Milton Moss (New York: Columbia University Press), pp. 509-32. Okun, Arthur M., 1971, "The Mirage of Steady Inflation," Brookings Papers on Economic Activity: 2, Brookings Institution, pp. 485-98. Olson, Mancur, Jr., 1996, "Big Bills Left on the Sidewalk: Why Some Nations Are Rich, and Others Poor," Journal of Economic Perspectives, Vol. 10 (Spring), pp. 3-24. Organization for Economic Cooperation and Development, 1994, Taxation and Investment Flows: An Exchange of Experiences Between OECD and the Dynamic Asian Economies (Paris: Organization for Economic Cooperation and Development). Orphanides, Athanasios, and Robert M. Solow, 1990, "Money, Inflation and Growth," in Handbook of Monetary Economics, Vol. 1, ed. by Benjamin M. Friedman and Frank H. Hahn (Amsterdam and New York: North-Holland), pp. 223-61. Perotti, Roberto, 1992, "Income Distribution, Politics, and Growth," American Economic Review, Papers and Proceedings, Vol. 82 (May), pp. 311-16. , 1993a, "Fiscal Policy, Income Distribution, and Growth" (unpublished; New York: Columbia University). — , 1993b, "Political Equilibrium, Income Distribution, and Growth," Review of Economic Studies, Vol. 60 (October), pp. 755-76. , 1994, "Income Distribution and Investment," European Economic Review, Vol. 38 (April), pp. 827-35. Persson, Mats, Torsten Persson, and Lars E.O. Svensson, 1987, "Time Consis- tency of Fiscal and Monetary Policy," Econometrica, Vol. 55 (November), pp. 1419-31. Persson, Torsten, and Guido Tabellini, 1992, "Growth, Distribution, and Politics," European Economic Review, Vol. 36 (April), pp. 593-602.

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 207

, 1994, "Is Inequality Harmful for Growth?" American Economic Review, Vol. 84 (June), pp. 600-21. Phelps, Edmund S., 1973, "Inflation in the Theory of Public Finance," Swedish Journal of Economics, Vol. 75 (March), pp. 67-82. Plosser, Charles I., 1992, "The Search for Growth," in Policies for Long-Run Eco- nomic Growth: A Symposium Sponsored by the Federal Reserve Bank of Kansas City (Kansas City, Missouri: Federal Reserve Bank of Kansas City), pp. 57-86. Quah, Danny T., 1996a, "Empirics for Economic Growth and Convergence," European Economic Review, Vol. 40 (June), pp. 1353-75. , 1996b, "Twin Peaks: Growth and Convergence in Models of Distribution Dynamics," Economic Journal, Vol. 106 (July), pp. 1045-55. Ram, Rati, 1986, "Government Size and Economic Growth: A New Framework and Some Evidence from Cross-Section and Time-Series Data," American Economic Review, Vol. 76 (March), pp. 191-203. , 1995, "Defense Expenditure and Economic Growth," in Handbook of Defense Economics, Vol. 1, ed. by Keith Hartley and Todd Sandler (Amster- dam and New York: Elsevier), pp. 251-73. Ramsey, Frank P., 1927, "A Contribution to the Theory of Taxation," Economic Journal, Vol. 37 (March), pp. 47-61. Rebelo, Sergio, 1991, "Long-Run Policy Analysis and Long-Run Growth," Jour- nal of Political Economy, Vol. 99 (June), pp. 500-21. Rogers, Carol Ann, 1987, "Expenditure Taxes, Income Taxes, and Time- Inconsistency," Journal of Public Economics, Vol. 32 (March), pp. 215-30. Romer, Paul M., 1986, "Increasing Returns and Long-Run Growth," Journal of Political Economy, Vol. 94 (October), pp. 1002-37. , 1989, "Capital Accumulation in the Theory of Long-Run Growth," in Modern Business Cycle Theory, ed. by Robert J. Barro (Cambridge, Massachusetts: Harvard University Press), pp. 51-127. Sala-i-Martin, Xavier, 1996a, "The Classical Approach to Convergence Analysis," Economic Journal, Vol. 106 (July), pp. 1019-36. , 1996b, "Regional Cohesion: Evidence and Theories of Regional Growth and Convergence," European Economic Review, Vol. 40 (June), pp. 1325-52. Sandmo, Agnar, 1985, "The Effects of Taxation on Savings and Risk Taking," in Handbook of Public Economics, Vol. I, ed. by Alan J. Auerbach and Martin Feldstein (Amsterdam and New York: North-Holland), pp. 265-311. Sarel, Michael, 1996, "Nonlinear Effects of Inflation on Economic Growth," Staff Papers, International Monetary Fund, Vol. 43 (March), pp. 199-215. Schumpeter, Joseph A., 1934, The Theory of Economic Development (Cambridge, Massachusetts: Harvard University Press). Shah, Anwar, ed., 1995, Fiscal Incentives for Investment and Innovation (Wash- ington: World Bank). Sinn, Hans-Werner, 1995, "A Theory of the Welfare State," Scandinavian Journal of Economics, Vol. 97 (No. 4), pp. 495-526. , 1996, "Social Insurance, Incentives and Risk Taking," International Tax and Public Finance, Vol. 3, pp. 259-80.

©International Monetary Fund. Not for Redistribution 208 VITO TANZI and HOWELL H. ZEE

Solow, Robert M., 1956, "A Contribution to the Theory of Economic Growth," Quarterly Journal of Economics, Vol. 70 (February), pp. 65-94. Stokey, Nancy L., and Sergio Rebelo, 1995, "Growth Effects of Flat-Rate Taxes," Journal of Political Economy, Vol. 103 (June), pp. 519-50. Streeten, Paul, and others, 1981, First Things First: Meeting Basic Needs in Devel- oping Countries (New York: Oxford University Press for the World Bank). Swan, Trevor W., 1956, "Economic Growth and Capital Accumulation," Economic Record, Vol. 32 (November), pp. 334-61. Tanzi, Vito, 1977, "Inflation, Lags in Collection, and the Real Value of Tax Rev- enue," Staff Papers, International Monetary Fund, Vol. 24 (March), pp. 154-67. , 1978, "Inflation, Real Tax Revenue, and the Case for Inflationary Finance: Theory with an Application to Argentina," Staff Papers, International Mone- tary Fund, Vol. 25 (September), pp. 417-51. , 1995, "Long-Run Growth and Public Policy," in Social Capability and Long-Term Economic Growth, ed. by Bon Ho Koo and Dwight H. Perkins (New York: St. Martin's Press), pp. 142-56. , and Ludger Schuknecht, 1995, "The Growth of Government and the Reform of the State in Industrial Countries," IMF Working Paper 95/130 (Washington: International Monetary Fund). Tanzi, Vito, and Parthasarathi Shome, 1992, "The Role of Taxation in the Devel- opment of East Asian Economies," in The Political Economy of Tax Reform, ed. by Takatoshi Ito and Anne O. Krueger (Chicago: University of Chicago Press), pp. 31-61. Tanzi, Vito, and Howell H. Zee, 1993, "Time Constraints in Consumption and Savings Behavior," Journal of Public Economics, Vol. 50 (February), pp. 253-59. , 1995, "Human Capital Accumulation and Public Sector Growth," IMF Working Paper 95/95 (Washington: International Monetary Fund). Thornton, Daniel L., 1996, "The Costs and Benefits of Price Stability: An Assess- ment of Howitt's Rule," Federal Reserve Bank of St. Louis Review, Vol. 78 (March/April), pp. 23-38. Turnovsky, Stephen J., and William A. Brock, 1980, "Time Consistency and Opti- mal Government Policies in Perfect Foresight Equilibrium," Journal of Pub- lic Economics, Vol. 13 (April), pp. 183-212. Uzawa, Hirofumi, 1965, "Optimal Technical Change in an Aggregative Model of Economic Growth," International Economic Review, Vol. 6 (January), pp. 18-31. Varian, Hal R., 1980, "Redistribute Taxation as Social Insurance," Journal of Public Economics, Vol. 14 (August), pp. 49-68. Verdier, Thierry, 1994, "Models of Political Economy of Growth: A Short Survey," European Economic Review, Vol. 38 (April), pp. 757-63. Wilcox, David W., 1989, "The Sustainability of Government Deficits: Implications of the Present-Value Borrowing Constraint," Journal of Money, Credit, and Banking, Vol. 21 (August), pp. 291-306. World Bank, 1995, Bureaucrats in Business: The Economics and Politics of Gov- ernment Ownership (Washington: Oxford University Press for the World Bank).

©International Monetary Fund. Not for Redistribution FISCAL POLICY AND LONG-RUN GROWTH 209

Xu, Bin, 1994, "Tax Policy Implications in Endogenous Growth Models," IMF Working Paper 94/38 (Washington: International Monetary Fund). Young, Alwyn, 1995, "The Tyranny of Numbers: Confronting the Statistical Real- ities of the East Asian Growth Experience," Quarterly Journal of Economics, Vol. 110 (August), pp. 641-80. Zee, Howell H., 1988, "The Sustainability and Optimality of Government Debt," Staff Papers, International Monetary Fund, Vol. 35 (December), pp. 658-85. , 1994, "Time-Consistent Optimal Intertemporal Taxation in Externally Indebted Economies," Public Finance, Vol. 49 (No. 1), pp. 113-25. , 1996, "Empirics of Cross-Country Tax Revenue Comparisons," World Development, Vol. 24 (October), pp. 1659-71. -, forthcoming, "Endogenous Time Preference and Endogenous Growth," International Economic Journal.

Fiscal Adjustments in OECD Countries: Composition and Macroeconomic Effects

ALBERTO ALESINA and ROBERTO PEROTTI *

This paper studies how the composition of fiscal adjustments influences their likelihood of "success, " defined as a long-lasting deficit reduction, and their macroeconomic consequences. We find that fiscal adjustments that rely primarily on spending cuts in transfers and the government wage bill have a better chance of success and are expansionary. On the contrary, fiscal adjustments that rely primarily on tax increases and cuts in public investment tend not to last and are contractionary. We discuss alternative explanations for these findings by studying a full sample of members of the Organization for Economic Cooperation and Development and by focusing on three case studies: Denmark, Ireland, and Italy. [JEL H1, H5, E62]

N THE LAST two decades, the debt-to-GDP ratios of many members of the IOrganization for Economic Cooperation and Development (OECD) have increased to levels historically observed only in the aftermath of major wars, as Table 1 documents. The policymakers of countries with fiscal problems face several critical questions:

• How large should the fiscal adjustment be? • Should one cut expenditures or raise revenues, and, more specifically, which components of spending and revenues should one adjust?

* Alberto Alesina is Professor of Economics and Government at Harvard Uni- versity. This paper was written while he was a visiting scholar in the European I Department of the IMF. He thanks Massimo Russo and the entire department for their hospitality. Roberto Perotti is Assistant Professor of Economics at Columbia University. For very helpful comments, the authors thank Massimo Russo, Michael Deppler, Alessandro Leipold, Alessandro Prati, Vito Tanzi, and participants in two seminars in the European I Department and the National Bureau of Economic Research. This research is partially supported by a grant from the National Science Foundation. 210

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 211

Table 1. Public Debt in OECD Countries (Gross debt as a percent of GDP) 1965 1975 1990 1994 Australia ...... 23.5 36.1 Austria 19.4 23.9 58.3 65.7 Belgium 67.5 61.1 128.5 135.0 Canada 58.8 43.1 73.1 95.6 Denmark 11.3 11.9 68.0 81.1 Finland 17.7 8.6 16.8 62.3 France 53.la 41.1 43.4 54.7 Germany 17.3 25.1 43.4 51.5 Greece 14.1 22.4 77.7 119.0 Ireland ... 64.4 97.4 92.3 Italy 35.4 60.4 106.4 123.9 Japan — 22.4 66.0 75.6 Netherlands 52.2 41.4 78.8 79.1 Norway 47.0a 44.7 32.5 43.5 Portugal ...... 68.6 70.5 Spain ...... 50.3 68.2 Sweden 30.5 29.5 44.3 79.5 United Kingdom 81.8a 63.7 39.3 54.5 United States 52.1 42.7 55.7 63.0 Source: OECD. a1970.

• Will the fiscal consolidation last, or will it be reversed and will larger deficits soon reappear? and • Will the fiscal adjustment cause a recession? The critical point that we stress in this paper is that all these questions are deeply interconnected.1 For instance, the composition of the fiscal adjustment influences both the likelihood of achieving a permanent consolidation of the budget and the macroeconomic consequences of the fiscal consolidation. We identify two different types of fiscal adjustments. "Type 1" adjustments rely primarily on expenditure cuts, in particular, cuts in transfers, social security, government wages, and employment. Tax increases are a small fraction of the total adjustment, and, in particular, taxes on households are not raised at all or are even reduced. On the contrary, "Type 2" adjustments rely mostly on broad-based tax increases, and often the largest increases are in taxes on households and social security contributions. On the expenditure side, almost all the cuts are in public investment, while government wages, employment, and transfers are completely untouched or only slightly affected. We find that, even when the two types of adjustments are the same size in terms of reducing primary deficits, Type 1 adjustments

1 For earlier results on related issues, see Alesina and Perotti (1995a and 1996).

©International Monetary Fund. Not for Redistribution 212 ALBERTO ALESINA and ROBERTO PEROTTI induce a more lasting consolidation of the budget and are expansionary, while Type 2 adjustments are soon reversed by further deteriorations of the budget and have contractionary consequences for the economy. Type 1 adjustments are more permanent because they tackle the two items of the budget, government wages and welfare programs, that have the strongest tendency to automatically increase; in fact, these items have been increasing in the last three decades as a share of total government spending.2 Concerning the macroeconomic consequences of the two types of adjustments, the literature, which we review below, has generally focused on credibility and wealth effects of fiscal adjustments on consumption. In this paper, we also emphasize the effects of fiscal policy on unit labor costs and competitiveness. In fact, we suggest that the unit labor cost channel may even be more empirically relevant for consumption than the wealth effects and credibility channels. This paper is organized as follows. Section I critically reviews several theoretical arguments on the contractionary or expansionary effects of fiscal adjustments. Section II discusses problems affecting cyclical adjustments of fiscal variables and presents the procedure that we use. Section III provides the empirical evidence on fiscal adjustments in a sample of 20 OECD countries for the period 1960-1994. Section IV analyzes three countries—Ireland, Denmark, and Italy—as examples of Type 1 and Type 2 adjustments. The last section concludes.

I. Contractionary or Expansionary Fiscal Consolidations: The Theory

Keynesian Effects

The standard Keynesian argument is that a fiscal contraction has a temporary contractionary effect through an aggregate demand channel in a model with sticky prices and wages. A standard multiplier effect implies that spending cuts are more recessionary than tax increases.

Expansionary Effects of Fiscal Contractions: Demand Side

Wealth Effects on Consumption A cut in government spending, if perceived as long lasting, implies a permanent reduction in the future tax burden of consumers, generating a positive wealth effect. However, even tax increases can have expansionary

2 See Tanzi and Schuknecht (1995) for a discussion of the transformation of the composition of government budgets in OECD countries in the last century.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 213 effects on consumption. Blanchard (1990a) argues that a tax increase today can have expansionary effects if it generates the expectations of less dra- matic and disruptive tax increases tomorrow. Also, by resolving uncertainty about the course of future fiscal policy it may reduce precautionary savings. Blanchard's argument is an example of what Bertola and Drazen (1993) characterize as the "expectation view of fiscal policy." That is, in an intertemporal model of consumption behavior, the effects of current fiscal policy depend on what expectations it generates for the course of future fiscal policy. In their words, "a policy innovation that would be contractionary in a static model may be expansionary if it induces sufficiently strong expectations of future policy changes in the opposite direction" (Bertola and Drazen, 1993, p. 12). They consider a model in which government spending follows a random walk with a positive drift and national income is (for simplicity) constant. Bertola and Drazen argue that stabilizations often have a discrete character. That is, even if public debt accumulates rapidly, political constraints often delay the adoption of the appropriate stabilization policies.3 However, because of the feasibility constraint, sooner or later a stabilization will have to occur. Call gc the ratio of government spending to GDP at which with probability p a stabilization occurs, where stabilization is defined as a sharp drop in spending. If the stabilization does not occur at gc, it will occur with certainty later, when the ratio of spending to GDP reaches a much higher level, g. At that point, spending is set below gc so that a stabilization may again occur at gc with probability p. The Bertola-Drazen model implies that, at low levels of government spending, private consumption falls at less than one-to-one ratio to increases in government spending: this is because an increase in spending implies that a stabilization may come soon. Thus, higher government spending only partially crowds out private demand—a typical Keynesian result, obtained in a fully neoclassical model. When the economy reaches gc, and if the stabilization occurs, private consumption jumps up, reflecting the wealth effect or reduced expected future taxes. With probability 1 —p, however, the stabilization does not occur. In this case, consumption jumps down because the public realizes that political constraints will delay the stabilization. Thus, a failed stabilization has a contractionary effect because it signals the lack of political commitment to a "serious" attempt to reduce spending. Bertola and Drazen's point is consistent with arguments put forward by Giavazzi and Pagano (1996). The latter argue that large fiscal contractions can be expansionary precisely because they signal a permanent and decisive

3 For instance, Alesina and Drazen (1991) model this delay as a result of a war of attrition among groups disagreeing on how to share the burden of the stabilization, so that time is necessary to resolve the political stalemate.

©International Monetary Fund. Not for Redistribution 214 ALBERTO ALESINA and ROBERTO PEROTTI change in the stance of fiscal policy, while small adjustments may have the opposite effect for the opposite reason. The "small" adjustments (in Gia- vazzi and Pagano's words) can be interpreted as an example of the failed stabilization at trigger point gc, using the terminology of Bertola and Drazen. Sutherland (1995) considers the effects of a stabilization in a model in which consumers have finite lives, as in Blanchard (1985). Each consumer faces a constant probability of death in every period of his life; thus the model does not imply Ricardian properties even with nondistortionary taxes and transfers. As in Bertola and Drazen (1993), the government is on an unstable fiscal path and a stabilization sooner or later has to occur. At a low level of debt, the model displays Keynesian features. This is because the stabilization is expected to occur in the distant future, when many of the current consumers will be dead. These consumers do not internalize fully the future increase in taxation. As debt increases, the model displays non- Keynesian features, namely, an increase in the fiscal deficit that is contractionary. This happens because current consumers expect that a stabilization will occur relatively soon, when they are still alive. One problem with the expectation view of fiscal policy is that the critical variable driving the results is the effect of current policies on the public's expectation about future policy changes, a variable that is intrinsically unobservable. A devil's advocate might argue that any behavior of private consumption following any type of fiscal policy can be rationalized by an appropriate assumption about what the current fiscal policy signals about unobservable future policies. A second channel for wealth effects arises from the fall in interest rates that may accompany a fiscal adjustment. Lower interest rates imply a higher market value of private wealth.

Credibility Effects A fiscal consolidation, particularly a strong one in a high-debt country, may have important credibility effects on interest rates by reducing risk premiums. The latter can be of two types: (1) inflation risk premiums or (2) default or consolidation risk premiums. Default risk may be trivial for low-debt countries but may become significant for high-debt ones.4 A discrete change in the fiscal policy stance may have a significant credibility effect on interest rates, which would crowd in private investment and consumption of durable foods. For instance, Miller, Skidelsky, and Weller

4 For some empirical discussion of default risk premiums in high-debt OECD countries, see Alesina and others (1992). They show that, although a rough measure of default risk is not influenced by the level of the ratio of debt to GDP at low levels of debt, it is affected by the debt level at high levels.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 215

(1990) consider a model with a threshold above which the government is forced to impose a tax on bond holders. With random shocks on the level of debt, the risk premium increases as the debt level approaches that threshold, leading to a fall in private demand. A decisive stabilization that reduces the debt level to well below the threshold eliminates the risk premium, crowding in the components of private demand particularly sensitive to interest rates. Along similar lines, Alesina, Prati, and Tabellini (1990) show that, at a high level of debt, particularly if the debt has short maturity, self-fulfilling confidence crises may materialize. If a crisis occurs and the public is not willing to roll over the debt, the government is forced to tax bond holders or default. This model displays multiple equilibria for some levels of debt- to-GDP ratios. Thus, a fiscal adjustment that reduces the debt below the level at which multiple equilibria are possible may have large discrete effects on the risk premiums. As for the "expectation theory" of consumption behavior, however, this argument also relies on a variable, "credibility of the adjustment," that is ex ante unobservable. That is, it is not a priori obvious what makes an adjustment credible or not.

Supply-Side Effects

Labor Supply: Neoclassical Effects While wealth effects on consumption from permanent reductions in government spending are expansionary on the demand side, the same wealth effects may reduce labor supply (Barro, 1981). If both consumption and leisure are normal goods, a wealthier individual will want to consume more of both, and he will therefore work less. In addition, the higher real wages induced by reducing Tobin's q may crowd out investment (Baxter and King, 1993). The standard substitution effect suggests that tax increases should reduce work effort and labor supply. Higher labor income taxes reduce labor supply. Thus, a permanent spending cut financed by a tax cut has two opposite effects on labor supply: the wealth effect reduces it while the substitution effect increases it. In the case of a temporary cut in government spending, the wealth effect should be small and the substitution effect relatively large. Thus, for temporary spending cuts the substitution effect should predominate, while for permanent spending cuts the wealth effect should dominate. However, Baxter and King (1993) argue that the financing side of the government budget (that is, whether or not the spending cut is financed with distortionary taxes) is more important for its macroeconomic impact on supply

©International Monetary Fund. Not for Redistribution 216 ALBERTO ALESINA and ROBERTO PEROTTI than for its duration. Empirically, however, both the wealth effect and the substitution effect on individual labor supply are likely to be small.5

Labor Market Structure While the effect of taxes on individual labor supply may be small, their effects on aggregate labor supply in unionized labor markets may be much larger. With unionized labor markets, a permanent increase in labor taxation shifts the union's aggregate supply of labor because it decreases the after-tax income of employed union members at any before-tax wage. In other words, an increase in labor taxation leads the union to demand higher real wages to compensate for the decrease in after-tax income. Alesina and Perotti (1997) show that this effect depends on the structure of labor markets. The effect is weak in countries with decentralized labor markets (like the United States or Canada) and is also relatively weak in countries with highly centralized unions, where comprehensive union- government negotiations internalize the entire fiscal maneuver. That is, a centralized union will take into consideration at the bargaining table the spending side of the government budget as well; thus, for instance, the union will internalize the effects of higher taxes on more public goods or larger transfers. On the contrary, the effect of labor taxation on unit labor cost is greatest in countries where unions are strong enough to pass on tax increases to wages but not encompassing enough to internalize the connection between taxes and benefits of the fiscal maneuver. For example, Alesina and Perotti (1997) calculate that in the countries in the intermediate group an increase of the income tax of 1 percent of GNP causes an increase in relative unit labor costs of about 2 percent. In summary, according to this channel, a tax increase affects unit labor costs of firms, influencing their competitiveness.

Why Composition Matters

The previous discussion has highlighted that tax increases and spending cuts may have very different effects. Several additional reasons suggest that even the composition of spending cuts may have important consequences on how permanent the fiscal adjustment is and on its macroeconomic consequences.6 One can identify at least three reasons why the composition of cuts may matter.

5 See, for instance, Pencavel (1986). For a general treatment of the neoclassical approach to fiscal policy, see Barro (1989). 6 For a discussion of composition effects on fiscal adjustments, see also Perotti (1996a).

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 217

Expectation Effect Different types of spending cuts may be more or less permanent. Con- sider two types of spending cuts of the same magnitude. The first one relies only on a reduction in public investment, for instance in maintenance of public infrastructure. The second one includes cuts in welfare obtained by changing eligibility criteria for transfer programs and by cutting government employment. Even though the two types of spending cuts may have the same magnitude of impact, clearly the second one has more lasting effects than the first. In fact, maintenance of public infrastructure cannot be postponed forever; on the contrary, structural changes in the parameters that determine the extent of coverage of the "welfare state" by influencing the dynamics of entitlement have long-lasting effects. Thus, according to this argument, the composition of spending cuts, much more than the size per se, influences expectations about the future stance of fiscal policy, a critical point of the expectation view.

Political Credibility Effect Governments that are willing to tackle the politically more delicate components of budgets, such as public employment, social security, and welfare programs, may signal that they are really "serious" about the fiscal adjustment. Not every government has the necessary strength to tackle these politically difficult issues. Typically, coalition governments lack the necessary cohesion to implement this type of adjustment, as shown by Alesina and Perotti (1995a).7 In fact, coalition governments succumb to intercoalition conflicts concerning the distributional consequences of the adjustment.8 However, single-party governments may have the necessary strength to cut transfers, social security programs, and the government wage bill.9

Labor Market Effect Cuts in the government wage bill may have different effects than cuts in nonwage government consumption. Lane and Perotti (1995) show that a fall in government employment shifts the aggregate demand for labor facing the

7 This result is consistent with previous empirical results by Roubini and Sachs (1989) and Grilli, Masciandaro, and Tabellini (1991), which suggest that coalition governments are less fiscally responsible. 8 A model of a war of attrition among coalition members may rationalize this effect; see Alesina and Drazen (1991) and Spolaore (1993). For a similar argument based upon a "tragedy of the common" game, see Velasco (1996). 9 According to the results obtained by Alesina and Perotti (1995a), both left- leaning and right-leaning governments have been able in about equal proportion to achieve successful fiscal stabilization.

©International Monetary Fund. Not for Redistribution 218 ALBERTO ALESINA and ROBERTO PEROTTI union, thus improving profitability through two channels: a fall in unit labor costs, and a depreciation of the exchange rate in a flexible exchange system. On the contrary, a cut in nonwage government consumption does not have these effects because, at least up to a first degree of approximation, the private and public sectors have the same propensity to spend on the goods and services that enter into the definition of nonwage government consumption. Lane and Perotti (1995) present empirical evidence on a sample of OECD countries that shows that the composition of spending cuts strongly influences labor market variables in the direction described above. The composition of tax increases may also influence the macroeconomic consequences of fiscal adjustment. For instance, Alesina and Perotti (1995b and 1997) show that taxes on households and social security contributions have the largest impact on relative unit labor costs, via union behavior. In this paper, we will refer to adjustments that rely primarily on cutting transfers and government employment and wages while keeping taxes on households constant or even reducing them as "Type 1" adjustments. We will refer to adjustments that rely primarily on tax increases, particularly on households, while cutting public investment spending as "Type 2" adjustments.

II. Fiscal Impulse: Cyclical Adjustment

We are interested in discretionary changes in the fiscal position of a country. Thus we need a measure of fiscal impulse, defined as the discretionary change in the government budget balance. For this reason, we focus upon primary deficits rather than total deficits, since fluctuations in interest payments cannot be considered discretionary. The second, more difficult, issue concerns the cyclical correction. One needs to isolate the discretionary change in the primary deficit, defined as the difference between the actual change in the deficit and the change that would have occurred had the policymakers done nothing. Clearly, the problem is to define what doing nothing means. For instance, it may mean, among other things, that certain spending programs remain constant in nominal terms at last year's level, or constant in real terms, or constant in their share of GDP. The following brief discussion highlights some possible methodologies; it is beyond the scope of the present paper to provide a comprehensive discussion of the issue of cyclical adjustment.10 The simplest approach to cyclical corrections is simply to ignore the problem and consider changes in the primary balance as a measure of fiscal

10 For overviews, see McKenzie (1989) and Perotti (1996b).

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 219 impulse. A second measure, typically used by the OECD, defines the fiscal impulse as the difference between the current primary deficit and the primary deficit that would have prevailed if expenditures in the previous year had grown with potential GDP and revenues with actual GDP. This approach, however, relies on measures of potential output that might be questionable. A third measure, used by the International Monetary Fund, is similar to the OECD's but assumes that the benchmark year is not the previous one but a reference year when output was supposed to be at its potential level. In this paper, as in Alesina and Perotti (1995a), we use a fourth measure proposed by Blanchard (1990b), which maintains much simplicity and transparency while providing an attractive cyclical correction. Essentially, this measure uses a calculation of what the government budget would be if unemployment had not changed from the previous year. Specifically, this cyclical adjustment is an attempt to eliminate from the budget changes in taxes and transfers associated with changes in the unemployment rate. We use a simple implementation of Blanchard's measure, constructed as follows. For each country, we regress transfers as a share of GDP (TRANSF) on two time trends (1960-75) and (1976-94) and the unemployment rate (U):

TRANSFt = a0 + a1 TREND1 + a2 TREND2 + a3Ut + et. (1)

Defining a0, a1, a2, and a3 as the estimated parameters of equation (1) and et as the estimated residual, we then compute what the variable TRANSF would have been in period t if the unemployment rate had remained constant between t - 1 and t:

TRANSFt(Ut-1) = a0 + a1TREND1 + a2 TREND2 + a3Ut-1 + et. (2)

We follow the same procedure to adjust tax revenues, defined as Tt(Ut-1). Using TRANSFt(Ut-1), together with all the other components of spending and Tt(Ut-1), we compute our measure of cyclically adjusted primary deficits.11 This measure of the fiscal impulse is then constructed as the difference between the cyclically adjusted primary deficit in period t and the same variable in period t - 1, as a share of GDP. We sometimes refer to this fiscal impulse variable as the "Blanchard fiscal impulse." Several reasons justify our choice for a cyclical correction. First, we find its simplicity attractive. This measure does not rely on possibly questionable and sometimes obscure measures of potential output or base years. Second, this simplicity does not come at a high price for us; in fact, since our focus is on large changes in the fiscal stance, cyclical factors typically should not play a major role. Clearly, one can think of exceptional circumstances in which major exogenous shocks have caused large changes in budget deficits,

11 See the appendix in Alesina and Perotti (1995a) for details.

©International Monetary Fund. Not for Redistribution 220 ALBERTO ALESINA and ROBERTO PEROTTI but these are probably more the exception than the rule. Third, as McDermott and Wescott (1996), who are pursuing a similar research strategy, adopt the OECD cyclical correction, it is useful to check on the robustness of results through a different approach. Finally, sensitivity analysis using different cyclical corrections suggests that our results are quite robust.

III. Fiscal Adjustments in OECD Countries

Definitions, Sample, and Basic Statistics

We consider a sample of 20 OECD countries for the period 1960-94. The countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States. Our definition of government is general government as defined by the OECD, which is our data source. We have 378 usable observations; all the country years lacking the complete set of data needed for our tests were excluded. We focus upon periods of very tight fiscal policy, considering, in particular, relatively large budget adjustments, for two reasons. First, by considering large policy changes, we are less likely to be unduly influenced by cyclical factors. Second, macroeconomic and composition effects are more likely to be detectable in the case of large adjustments. We use the following definition of a tight fiscal policy: Definition 1: A period of tight fiscal policy is a year in which the cyclically adjusted primary deficit falls by more than 1.5 percent of GDP or a period of two consecutive years in which the cyclically adjusted primary deficit falls by at least 1.25 percent a year in both years. This definition allows for both a yearly definition of an adjustment and a two-year definition. While there is a certain degree of arbitrariness in this choice, as in any alternative, we show below, by reviewing related literature and performing sensitivity tests, that our results are not unduly sensitive to this particular definition. Table 2 shows that, on average, the fiscal impulse is close to zero. How- ever, this is the result of significant increases in primary expenditures and revenues of more than 0.3 percent of GDP a year. This observation reflects the well-known "growth of government" that has occurred in the last several decades in OECD countries. Our sample provides 62 years of tight fiscal policy, with an average improvement in the Blanchard fiscal impulse of more than 2.5 percent of GDP. This reduction in deficits is about equally distributed between higher taxes and lower spending. We now need a definition of success.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 221

Table 2. Tight Fiscal Policies in OECD Countries: Changes in Primary Expenditures and Revenues (In percent of GDP) Change Number Fiscal in primary Change of observations impulse expenditures in revenues All periods 378 -0.07 0.32 0.38 (0.09) (0.09) (0.06) Periods of tight 62 -2.57 -1.34 1.22 fiscal policy (0.20) (0.23) (0.15) Source: OECD. Note: Standard deviations in parentheses. Table 3. Successful Fiscal Adjustments in OECD Countries Country Yeara Australia 1987 Denmark 1984 (2 years) United Kingdom 1977 Ireland 1988 (2 years) Ireland 1989 (2 years) Norway 1980 Portugal 1977 Portugal 1982 Sweden 1984 (2 years) Sweden 1987 (2 years) United States 1976 a For cases of two-year adjustments, the second year is indicated.

Definition 2: A period of tight fiscal policy is successful if (1) in the three years after the tight period the ratio of the cyclically adjusted primary deficit to GDP is on average at least 2 percent of GDP below the last year of the tight period or (2) three years after the last year of the tight period the ratio of debt to GDP is 5 percent of GDP below the level of the last year of the tight period. This definition allows for a measure of success on both the stock of debt and the flow of cyclically adjusted primary deficits. This definition is quite demanding, and about one-fourth of tight policies are successful. Table 3 lists all the cases of successful adjustments: 11 episodes, for a total of 16 observation years. As we discuss below, our results are not unduly sensitive to rea- sonable changes in this definition; in particular, we illustrate results obtained with a much more lenient definition of success.12 12 One may argue that a definition of success should allow for differences across countries based upon initial conditions. However, the data would be too diffuse and impossible to analyze if different definitions of success were allowed for different initial conditions. We leave this problem to future research.

©International Monetary Fund. Not for Redistribution 222 ALBERTO ALESINA and ROBERTO PEROTTI

Success and Composition

Table 4 shows that successful adjustments are slightly larger in terms of fiscal impulse than unsuccessful ones: the difference is about 0.5 percent of GDP. While this difference is not trivial, more striking differences appear in the composition of adjustments. In successful cases (16 observation years), about 73 percent of the adjustment is on the spending side; in unsuccessful cases (46 observation years), about 44 percent of the adjustment is on the expenditure side. Even more striking are the differences in composition of types of spending and sources of revenue. Tables 5 and 6 consider a breakdown of the major components of the spending side. In unsuccessful cases, more than two-thirds of the cuts are in capital spending (public investment) while everything else, particularly government wages, is virtually untouched. In successful cases, cuts in capital expenditures are actually much lower in terms of GDP share than in unsuccessful cases, despite the larger amount of total spending cuts. In fact, only one-fifth of total spending cuts in successful cases are in public investment. The critical difference is that in successful adjustments the largest cuts are in transfers and government wages, which together account for about 50 percent of the total spending cuts. In successful adjustments, transfers and government wages are reduced by almost 1.1 percent of GDP a year while in unsuccessful cases the sum of these two components is less than 0.2 percent of GDP a year. Cuts in the government wage bill may arise from lower wages and from lower employment. Table 7 shows that, while during successful adjustments the growth of government employment significantly drops, the rate of growth of this variable does not change either during or after unsuccessful adjustments. This outcome holds both for government employment alone and for the same variable as a share of the labor force.

Table 4. Successful and Unsuccessful Adjustments in OECD Countries: Changes in Primary Expenditures and Revenues (In percent of GDP) Changes Number Fiscal in primary Change of observations impulse expenditures in revenues Successful 16 -2.92 -2.12 0.83 adjustments (0.28) (0.29) (0.36) Unsuccessful 46 -2.44 -1.07 1.36 adjustments (0.24) (0.28) (0.16) Source: OECD. Note: Standard deviations in parentheses.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 223

Table 5. Successful and Unsuccessful Adjustments in OECD Countries: Composition of Expenditure Cuts as Share of GDP (In percent) Nonwage govern- Number of Primary Public Govern- ment observa- expen- investment consump- Sub- tions ditures ment Transfers wages tion sidies Successful 16 -2.12 -0.43 -0.48 -0.58 -0.30 -0.27 adjustments (0.29) (0.16) (0.18) (0.08) (0.06) (0.18) Unsuccessful 46 -1.07 -0.68 -0.14 -0.05 -0.06 -0.08 adjustments (0.29) (0.30) (0.08) (0.06) (0.03) (0.05) Source: OECD. See Appendix for a precise definition of variables. Note: Standard deviations in parentheses.

Table 6. Successful and Unsuccessful Adjustments in OECD Countries: Composition of Expenditure Cuts as Share of Total Expenditure Cuts (In percent) Nonwage govern- Public Govern- ment Number of investment consump- Sub- observa- ment Transfers wages Columns tion sidies tions (1) (2) (3) (2) + (3) (4) (5) Successful 16 -0.20 -0.23 -0.28 -0.51 -0.14 -0.13 adjustments (0.07) (0.09) (0.04) (0.03) (0.08) Unsuccessful 46 -0.63 -0.13 -0.04 -0.17 -0.06 -0.07 adjustments (0.28) (0.07) (0.05) (0.03) (0.05) Source: OECD. Computations from Table 5. Note: Standard deviations in parentheses.

Table 7. Successful and Unsuccessful Adjustments in OECD Countries: Changes in Government Employment (In percent) Successful Unsuccessful Before During After Before During After Rate of growth of government 3.01 0.91 2.52 2.81 2.18 2.32 employment (0.36) (0.79) (1.19) (0.40) (0.32) (0.46) Rate of growth of government employment as share of total 1.41 0.29 1.04 1.79 1.22 1.34 labor force (0.34) (0.65) (1.04) (0.38) (0.27) (0.24) Source: OECD. Notes: "Before" is the yearly average of the two-year period before the fiscal adjustment; "during" is the one-year adjustment period or yearly average of the two-year adjustment period; "after" is the yearly average of the two-year period after the adjustment. Standard deviations in parentheses.

©International Monetary Fund. Not for Redistribution 224 ALBERTO ALESINA and ROBERTO PEROTTI

Table 8. Successful and Unsuccessful Adjustments in OECD Countries: Composition of Revenue Increases as Share of GDP (In percent) Number Taxes of on Taxes Social observa- house- on Indirect security tions Revenues holds business taxes contributions Successful 16 0.83 0.51 0.20 0.09 adjustments (0.36) (0.23) (0.22) (0.12) (0.19) Unsuccessful 46 1.36 0.35 0.29 0.43 0.31 adjustments (0.16) (0.09) (0.09) (0.08) (0.09) Source: OECD. See Appendix for a precise definition of variables. Note: Standard deviations in parentheses.

Tables 5, 6, and 7 thus show that successful adjustments are based on broad-based spending cuts that do not spare the most politically sensitive parts of the budget, namely, transfers, social security, government wages, and employment; in fact, these components receive the largest share of expenditure reductions. As mentioned above, these are Type 1 adjustments. On the contrary, unsuccessful adjustments concentrate most of their cuts in capital expenditures (Type 2 adjustments), probably for two reasons. First, the effects of cuts in public investment, such as postponing the maintenance of infrastructure or delaying new capital projects, are less immediately visible to voters than cuts in their salaries or pensions checks.13 Second, "creative accounting" is probably easier in the capital accounts.14 Tables 8 and 9 display the composition of tax increases. In successful, Type 1 adjustments, tax increases are concentrated in business and indirect taxes. The increase in business taxes may be due to a larger base rather than to higher rates. In fact, we show below that the profit share tends to increase in successful adjustments. Taxes on households do not increase at all, and social security contributions are also spared almost completely. However, the tax increase in unsuccessful, Type 2 adjustments is widely spread across all components. The contrast between cases of success and failure is particularly striking for taxes on households and social security. The total share of tax increases of these two components is only 10 percent in successful cases and almost 50 percent in unsuccessful cases. As shown in Alesina and Perotti (1995b and 1997) and discussed above, these two types

13 See Rogoff (1990) for an insightful discussion of the impact of political cycles on budget composition; this explains how opportunistic policymakers cut public investment before elections because it is less visible to voters than transfers. 14 See our discussion on Italy below, Tanzi (1994), and Alesina, Mare, and Perotti (1996).

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 225

Table 9. Successful and Unsuccessful Adjustments in OECD Countries: Composition of Revenue Increases as Share of Total Revenue Increases (In percent) Number of Taxes on Taxes on Indirect Social security observations households business taxes contributions

Successful 16 0.62 0.24 0.10 adjustments (0.27) (0.26) (0.14) (0.23) Unsuccessful 46 0.25 0.21 0.32 0.23 adjustments (0.07) (0.06) (0.06) (0.07) Source: OECD. Computations from Table 8. Note: Standard deviations in parentheses.

of taxes—taxes on households and social security contributions—have the strongest effects on unit labor costs, via union behavior. As we show in the next subsection, the movement of unit labor costs is, in fact, very different in successful and unsuccessful adjustments, and may help explain differences in the macroeconomic consequences of fiscal adjustments.

Macroeconomic Consequences of Fiscal Adjustments

Table 10 summarizes some basic statistics on macroeconomic conditions before, during, and in the immediate aftermath of successful and unsuccessful adjustments. The term "before" refers to the two-year period before the beginning of the tight fiscal policy. The term "after" refers to the two- year period after the last year of the tight policy. The term "during" is the year or the two-year period of the fiscal adjustment. All the variables in the table are yearly averages. The rate of GDP growth, as measured in differences from the average of the seven major industrial countries, shows large differences between successful and unsuccessful cases. During successful adjustments, growth is more than 1 percent above the industrial country average, and this difference is statistically significant; afterward, growth falls but still is above the industrial country average. On the contrary, during and after unsuccessful adjustments, growth remains below the industrial country average. Before successful adjustments, growth is not higher (relative to the industrial countries) than before unsuccessful adjustments; in fact, it is slightly lower. This observation suggests that adjustments are not successful simply because they are started in periods of high growth. Unemployment relative to the industrial country average after a successful adjustment is at about the same level as before it. After an unsuccessful

©International Monetary Fund. Not for Redistribution 226 ALBERTO ALESINA and ROBERTO PEROTTI

Table 10. Successful and Unsuccessful Adjustments in OECD Countries: Macroeconomic Conditions (In percent) Successful Unsuccessful Before During After Before During After GR -0.08 1.05 0.28 0.27 -0.29 -0.23 (0.49) (0.45) (0.57) (0.21) (0.34) (0.25) U 1.20 1.44 1.08 0.82 1.49 1.74 (1.09) (1.36) (0.93) (0.44) (0.61) (0.43) AI -1.05 4.93 9.14 2.58 2.04 1.35 (2.80) (1.43) (2.59) (0.94) (1.56) (1.08) AC 1.75 2.98 2.71 2.36 2.70 2.34 (0.53) (0.60) (0.61) (0.28) (0.34) (0.32) i 3.81 3.26 4.80 2.95 3.58 3.93 (0.65) (0.73) (1.14) (0.46) (0.77) (0.57) r -0.65 -0.28 -0.05 -0.84 -0.65 0.66 (0.78) (0.90) (0.52) (0.34) (0.48) (0.34) ULC -3.15 3.60 0.77 -0.77 -0.65 0.34 (2.57) (1.97) (2.00) (1.03) (1.29) (1.03) EXCH RATE -4.15 -3.78 -2.30 -2.26 -2.12 -1.82 (1.46) (1.68) (1.98) (0.72) (1.06) (0.76) TB 0.33 1.72 0.13 0.46 0.21 0.21 (0.86) (0.52) (0.37) (0.22) (0.25) (0.27) VAULC 0.87 2.63 -0.27 0.10 0.53 -0.08 (0.45) (0.69) (0.63) (0.27) (0.36) (0.28) WSH 57.30 54.91 54.32 52.46 51.97 52.39 (1.26) (1.23) (1.40) (0.74) (1.08) (0.77) PSH 28.08 30.28 30.01 32.42 32.35 32.85 (1.39) (1.25) (1.16) (0.58) (0.84) (0.63) Source: OECD. Notes: "Before" is the two-year period before the adjustment; "during" is the adjustment period; "after" is the two-year period after the adjustment. GR is the yearly average growth of GDP relative to the average for the seven major industrial countries, weighted by GDP. U is the unemployment rate relative to the average for the seven major industrial countries. AI is the yearly rate of growth of private investment. AC is the average yearly rate of growth of consumption. i is a nominal long-term interest rate (ten-year government bond) relative to the average for the seven major industrial countries. r is the real long-term interest rate computed as i — inflation, relative to the seven major industrial countries. ULC is the yearly rate of change of relative unit labor costs. EXCH RATE is the rate of change of the nominal effective exchange rate. TB is the trade balance. VAULC is the rate of change of the value-added deflator over unit labor costs; WSH is the wage share over GDP, and PSH is the profit share over GDP. See the Appendix for a precise definition of these variables. Standard deviations in parentheses. adjustment, unemployment relative to the industrial countries doubles from less than 1 percent to almost 2 percent above the industrial country average. The rate of growth of private investment shows a sizable difference between the two types of adjustments. An investment boom occurs during

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 227 and after the successful adjustments. In unsuccessful cases, the rate of growth of investment falls during and after the adjustment, and it is much lower than in successful cases. The cumulative growth rate of investment during and after successful adjustments is about 25 percent, while it is only about 8 percent during and after unsuccessful cases. The rate of growth of consumption does not show large differences between the two types of adjustments. These observations on investment and consumption are intriguing, since the academic literature (theoretical and empirical) has typically focused on consumption much more than on investment. Relative to the seven major industrial countries, nominal and ex post real long-term interest rates do not show strong differences between the two types of adjustments, although nominal rates do fall during successful adjustments and increase during unsuccessful ones. Very interesting observations emerge from the evidence on unit labor costs and measures of competitiveness. The data on relative unit labor costs show that these costs significantly fall before and during successful cases while they are about constant in unsuccessful ones. The cumulative fall before and during successful cases is more than 10 percent. The behavior of relative unit labor costs can be influenced by two factors: a depreciation of the nominal exchange rate in an economy with nominal rigidities, and a containment of wage pressure. As Table 10 shows, both successful and unsuccessful adjustments have been accompanied and preceded by nominal depreciations. The depreciations have been somewhat larger in successful cases but significant as well in unsuccessful adjustments. How- ever, while in successful cases the nominal depreciations have had an impact on competitiveness (unit labor costs), in unsuccessful cases they have not. These observations suggest that the behavior of real wages is significantly different in the two types of adjustments. As argued above, this difference may be linked to the composition of the fiscal adjustments, and in particular to the difference between the two types of adjustments in the behavior of government wages and employment, taxes on households, and social security contributions. The evidence on the trade balance confirms that net exports have performed better in successful than in unsuccessful adjustments. The last three rows of Table 10 display indices of profitability and distributional shares. An index of the value-added deflator over unit labor costs (VAULC) shows a rather different behavior in successful and unsuccessful adjustments. In the former group, one observes a significant increase in this index of profitability during the adjustments, while in unsuccessful cases the same index is constant. The behavior of real wages, unit labor costs, and profitability also has clear implications for the distributional shares. While the wage share (WSH) goes down by about 3 points and the profit share

©International Monetary Fund. Not for Redistribution 228 ALBERTO ALESINA and ROBERTO PEROTTI

increases by about 2 points during successful adjustments, these two shares are virtually constant in unsuccessful cases. Table 10 suggests that perhaps the strongest channel influencing the macroeconomic consequences of fiscal adjustment goes through unit labor costs, via their effects on investment and exports. This channel may be at least as important, if not more so, than the channels typically emphasized in the literature, which are based upon "wealth effects cum expectations" on consumption and credibility effects on consumption. The discussion in the next section on three case studies sheds more light on this issue. Needless to say, Table 10 is far from conclusive, and much more statistical evidence is necessary to disentangle the various channels through which fiscal adjustments can influence the economy. In particular, several critical issues need more attention. First, more work is needed to disentangle the effects of exchange rate depreciations and wage moderation on the likelihood of success of fiscal adjustments. One interpretation of the evidence presented above is that the composition of successful adjustments induces wage moderation while the composition of unsuccessful ones does not. A critical exhibit in favor of this interpretation is that, during successful, Type 1 adjustments, government wages are cut, government employment does not increase much, and taxes on households are constant. According to the models and the empirical evidence presented by Alesina and Perotti (1995b and 1997) and Lane and Perotti (1995), these features should imply wage moderation on the part of the unions. On the contrary, the features of unsuccessful, Type 2 adjustments, namely, higher increases in taxes and no cuts in government wages or employment, have the opposite effects on unit labor costs. However, several major multiyear fiscal adjustments (see next section) are preceded by a devaluation of the exchange rate. Disentangling the effects of wage moderation and fiscal variables from the effects of exchange rate depreciations on the supply side and the cost of firms is the critical next step to understanding the dynamics of fiscal adjustments. Also, Table 10 shows that growth is significantly higher during successful adjustments than during unsuccessful ones. Our favorite interpretation of this finding is that the composition of the adjustment, through its credibility, wealth, and, especially, unit labor cost effects, influences growth. An alternative interpretation is that adjustments are successful because growth is, for some exogenous reason, particularly high during these episodes. While our evidence cannot be conclusive on this point, the alternative interpretation, which argues that growth explains success, fails to explain the difference in composition between successful and unsuccessful adjustments. If everything is driven by exogenous growth effects, successful and unsuccessful adjustments should look approximately the same in terms of their composition.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 229

Sensitivity Analysis and Comparisons with Previous Results

Sensitivity analysis of our results shows that they are robust to changes in the definitions. For instance, we have relaxed the definition of success as follows: Definition 3: A period of tight fiscal policy is successful if (1) the cyclically adjusted primary deficit as a share of GDP is on average lower than in the last year of the tight policy, or (2) the debt-to-GDP ratio three years after the last year of the adjustment is below the level of the last year of the adjustment. With this new definition, we now have 38 observation years of success, corresponding to 28 episodes, and 24 observation years of failures. Table 11 considers differences in the composition of expenditure cuts and tax increases and should be compared with Table 4 above. The differences between successful adjustments and unsuccessful ones in Table 11 are very similar to those in Table 4; in fact, they are slightly larger. The size of the adjustment in terms of fiscal impulse is virtually identical. Table 12 considers the composition of spending cuts and should be compared with Table 6. The difference in the composition of successful and unsuccessful adjustments remains striking. Virtually all the adjustment in unsuccessful cases (84 percent of total spending cuts) is on public investment, while transfers and government wages slightly increase. In successful cases, cuts of transfers and government wages are almost one-half of the total expenditure reductions. The differences between the two cases in the composition of tax increases are qualitatively similar to those reported in Tables 8 and 9. As for the macroeconomic consequences, this alternative, more lenient definition of success produces results that are quite similar to those

Table 11. Successful and Unsuccessful Adjustments in OECD Countries (Relaxed Definition): Changes in Primary Expenditures and Revenues (In percent of GDP) Changes Number Fiscal in primary Changes of observations impulse expenditures in revenues Successful 38 -2.61 -1.66 0.96 adjustments (0.17) (0.21) (0.19) Unsuccessful 24 -2.50 -0.84 1.64 adjustments (0.43) (0.46) (0.24) Source: OECD. Note: Standard deviations in parentheses.

©International Monetary Fund. Not for Redistribution 230 ALBERTO ALESINA and ROBERTO PEROTTI

Table 12. Successful and Unsuccessful Adjustments in OECD Countries (Relaxed Definition): Composition of Expenditure Cuts as Share of Total Expenditure Cuts (In percent) Nonwage government Number Public Govern- of invest- Trans- ment Consump- observa- ment fers wages Columns tion Subsidies tions (1) (2) (3) (2) + (3) (4) (5) Successful 38 -0.33 -0.24 -0.19 -0.43 -0.10 -0.09 adjustments (0.09) (0.06) (0.04) (0.02) (0.05) Unsuccessful 24 -0.84 0.03 0.04 0.07 -0.08 -0.11 adjustments (0.64) (0.12) (0.11) (0.06) (0.10) Source: OECD. Note: Standard deviations in parentheses. reported in Table 10; however, as should be expected, the differences between the successes and the failures are a bit smaller. The behavior of unit labor costs remains the most striking difference between the two types of adjustments. In a separate test, we also have relaxed our definition of a tight policy, reducing (in absolute value) the threshold level for a tight policy. For instance, we have considered as a tight episode any two-year period in which the Blanchard fiscal impulse is lower than minus 1 percent, as opposed to minus 1.25 percent, as in definition 1. Our results do not change qualitatively. Further experiments combining different definitions of tight episodes and success confirm the general robustness of our results.15 Also, a comparison with our previous results (Alesina and Perotti, 1995a and 1996) confirms that the basic picture painted in this section is quite robust. In our previous papers, we have used a definition of tight policy that included only one-year adjustments. What we have done here improves upon this definition by broadening it to include a two-year period of "moderate adjustment" rather than accepting only one year of sharp adjustment. We have also improved upon our previous definition of success by focusing on not only the stock of debt but also the flow of deficits as criteria of success. Also, we use an updated data set in this paper that includes only series with a full set of usable observations. Finally, we consider many more macroeconomic variables and make some progress toward disentangling the macroeconomic effects of adjustments, particularly those related to the external sector and exchange rate movements.

15 All these results, including the macroeconomic consequences of a more lenient definition of success, are available upon request.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 231

McDermott and Wescott (1996) use a similar methodology to ours. Their definitions are a combination of what we use in this paper and what we have used in our previous work. Their definition of tight fiscal policy allows for two- or three-year periods of consecutive tightening. Their definition of success is based only on improving the ratio of debt to GNP. They also use a different methodology for the cyclical adjustment. As far as the composition of fiscal adjustment is concerned, McDermott and Westcott confirm our results on the spending side. This outcome is quite reassur- ing, given the several differences between their definitions and procedures and ours. McDermott and Westcott also make similar findings concerning the size of the adjustment in successful and unsuccessful cases. They interpret their findings to mean that the size of the adjustment is very important. We still find the differences in the composition of adjustment more impressive than the differences in size, particularly in light of the results of Table 11. Using a rather different methodology based upon the estimation of consumption functions, Giavazzi and Pagano (1996) argue that large and persistent fiscal adjustments are expansionary while smaller ones are not because of credibility and wealth effects. Meanwhile, Bartolini, Razin, and Symansky (1995) use the multicountry model (MULTIMOD) to study the effects of fiscal adjustments in the major industrial countries. They find that adjustments have short-run output costs and long-run benefits. However, the recovery time for adjustments relying on increases in indirect taxes and expenditures is quicker than for other types of adjustments, a result that is quite consistent with our evidence presented above.

IV. Three Major Fiscal Adjustments

In the previous sections, we have argued that one can identify two types of fiscal adjustments. Type 1 adjustments rely primarily on spending cuts; the components of spending that receive the largest cuts are transfer programs and government employment and wages. Furthermore, in these adjustments, taxes on households are kept constant or even reduced. Type 2 adjustments rely primarily on tax increases, particularly on households, and spending cuts in public investment. We have shown that Type 1 adjustments are more permanent and expansionary, while Type 2 adjustments tend to be reversed by further deteriorations of the budget and have worse macroeconomic consequences. In this section, we look at recent fiscal adjustments in Ireland, Denmark, and Italy. We argue that the Irish adjustment in the late 1980s (1986- 1989) is a Type 1 adjustment. The Danish adjustment of 1983-86 has substantially different features from the Irish one and lies somewhere between

©International Monetary Fund. Not for Redistribution 232 ALBERTO ALESINA and ROBERTO PEROTTI

Type 1 and Type 2. Finally, we suggest that the current Italian adjustment is a Type 2 adjustment, at least up until 1995.

The Irish Adjustment 1987-89

Ireland entered the 1980s with a serious fiscal problem: the borrowing requirement was nearly 16 percent of GNP. In 1982-84, a weak and divided coalition government engaged in a fiscal adjustment focusing almost completely on the revenue side. In particular, taxes on households were sharply raised. The only modest spending cuts were in public investment. The government's actions look like a textbook case of Type 2 adjustment. The Irish pound was pegged to the currencies of the European Monetary Union (EMU), leading to a rapid disinflation with falls in interest rates. However, the stabilization failed to permanently consolidate the budget and had very strong negative effects on domestic demand. The fiscal balance continued to deteriorate in 1984-86, mostly owing to increases in the interest burden and primary spending; the latter grew by about 2 percentage points of GDP in those three years. In 1987, the ratio of gross debt to GDP peaked at almost 120 percent. A new government elected in February of that year launched an adjustment program with totally different features from the failed one of the early 1980s. This adjustment led to a sharp drop in the debt-to-GDP ratio in the following five years (from almost 120 percent to slightly more than 90 percent) and to rates of growth well above OECD or major industrial country averages. The turnaround of the Irish economy that started in 1987 is remarkable.

Size and Composition of Adjustment Table 13 summarizes several features of the Irish adjustment. Our measure of cyclically adjusted primary surplus, the Blanchard fiscal impulse, improved by about 8 percent of GDP from 1985-86 to 1990-91. The entire adjustment was on the spending side. The only moderate increase in household taxes during the adjustment was due entirely to a one-off tax amnesty in 1988. Total revenues (including taxes on households) as a share of GDP were lower in 1989-90 than in 1986. The spending cuts were broad ranging, but the largest in terms of share of GDP were in transfers, which fell by more than 2.5 percent of GDP. The government wage bill and public investment received the second-largest reduction. Reductions in government wages were obtained by an agreement with the unions in 1987 to limit pay increases to 2.5 percent, well below inflation. Even more important was the large reduction in public employment. Between 1986 and 1989, total public employment was cut by about 10 percent, from 300,000 to 270,000, as the result of a hiring freeze instituted in 1987 and early

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 233

Table 13. Fiscal Adjustmentin Ireland: Size and Composition,1985-91 (In percent of GDP) Change Change during during and after Before During After adjustment adjustment (1) (2) (3) (2) - (1) (3) - (1) Fiscal impulse -3.5 2.1 4.4 5.6 7.9 Primary expenditures 40.6 35.3 33.8 -5.3 -6.8 Transfers 17.4 16.0 14.8 -1.4 -2.6 Government wages 12.2 11.0 10.7 -1.2 -1.5 Nonwage government consumption 7.4 6.2 6.3 -1.1 -1.0 Public investment 3.6 2.1 2.1 -1.5 -1.5 Revenues 35.9 35.9 34.5 — -1.4 Taxes on households 12.9 13.4 12.2 0.5 -0.7 Taxes on business 1.3 1.3 2.0 — 0.7 Indirect taxes 16.3 16.0 15.0 -0.3 -1.3 Social security contributions 5.4 5.2 5.3 -0.2 -0.1 Source: OECD. Note: "Before" is the two-year period before the adjustment, 1985-86; "during" is the period of the adjustment, 1987-89; "after" is the two-year period after the adjustment, 1990-91. retirement and voluntary redundancy schemes.16 Social spending was cut, mostly in the health sector.

Tax Reform and Wage Bargaining Broad tax reform was introduced in Ireland in 1988 and took effect in 1989. A cornerstone of the reform was a cut in marginal tax rates on the income of households. The top rate, which had been as high as 65 percent in 1984-85, was reduced to 56 percent; also, the standard rate was cut from 35 percent to 32 percent. These cuts were accompanied by an increase in standard allowances. As the Irish tax system is quite progressive, these tax cuts reached very low on the income ladder. The corporate income tax was also reduced from 47 percent to 43 percent. Wage bargaining was decentralized throughout the 1980s. However, a centralized wage agreement was reached in 1987 in the context of the Pro- gram for National Recovery. The accord, which covered 1988-90, insured wage moderation both in the private and in the public sector. Most firms'

16 In order to facilitate voluntary resignations, the government sponsored retrain- ing programs to help individuals leave the public sector. These relatively cheap programs were quite successful.

©International Monetary Fund. Not for Redistribution 234 ALBERTO ALESINA and ROBERTO PEROTTI wage agreements were in line with the guidelines established by this accord. We claim that this tax reform and the wage bargaining agreements are related. As argued in the previous section, wage moderation was achieve probably because the unions internalized the increase in the after-tax dis posable income owing to the tax cuts. In addition, the effects of a soft labc market with high unemployment played a part, and the more or less volun tary layoffs in the public sector probably increased the unions' moderatio at the bargaining table.

Macroeconomic Consequences of Adjustment The year 1987 marked a remarkably positive turnaround for the Iris economy. Unemployment started to decrease in 1988 after a rising trend that had lasted 25 years. Table 14 shows that GDP growth, which wa almost 2 percent below the major industrial country average before th adjustment, was well above that average during and after the adjustment Growth was sustained by both domestic and foreign factors. Domestic con sumption grew at steady rates during and after the adjustment, mostly owing to consumer durables, particularly automobiles. Private investment increased, picking up the slack left by the contraction of public investment The external sector was sustained by the devaluation of the Irish pound in 1987 and by the policy of wage moderation adopted in the same year. Nominal and real interest rates fell dramatically between 1987 and 1985 both in absolute terms and relative to Germany. Short-term real rates (defined

Table 14. Fiscal Adjustment in Ireland: Macroeconomic Conditions, 1985-91 (In percent) Change Change during during and after Before During After adjustment adjustment (1) (2) (3) (2) - (1) (3) - (1) GR -1.72 1.89 3.86 3.61 5.58 U 9.80 10.80 8.63 1.00 -1.17 AI -7.71 8.31 1.31 16.02 9.02 AC 2.47 5.78 1.93 3.31 -0.54 i 3.41 2.14 1.43 -1.27 -1.98 ULC 8.67 -2.73 2.16 EXCH RATE 3.30 -1.03 3.31 TB 1.51 2.27 -1.91 WSH 54.23 51.58 50.41 -2.65 -3.82 PSH 21.75 25.32 26.48 3.57 4.73 Source: OECD. Notes: For the definition of the variables, see Table 10. For the definition of "before, "during," and "after," see Table 13.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 235 as nominal minus actual inflation) fell from about 11 percent at the end of 1986 to about 7 percent in 1989. The differential of these rates with Germany's fell from about 7 percent in 1986 to about 3 percent in 1989. Long-term rates also declined, with the differential of real rates with respect to Germany's falling from about 5 percent in 1986 to about 2 percent in 1989. These developments are an indication of credibility effects in asset markets, as pointed out by Dornbusch (1989). The increase in market value of private wealth owing to the fall in interest rates generated a positive wealth effect in consumption. As noted by Giavazzi and Pagano (1990), an effect on liquidity-constrained individuals, who obtained an increase in disposable income from the tax cuts, was a second source of the increase in private consumption. Interestingly, most of the increase in consumption came from durables, perhaps implying a strong interest rate effect. All indicators of competitiveness show marked improvements in Ireland between 1986 and 1989. For instance, IMF staff calculations of relative unit labor costs fell by a cumulative 15 percent in this period. Our measure of relative unit labor costs fell by almost 12 percent in the same period. These developments were the result of the combined effect of wage moderation and the devaluation of the Irish pound. The trade balance improved during the adjustment despite the booming domestic demand (Table 14). The effect of wage moderation is apparent in the evolution of distributional shares (see Table 14). The wage share fell from about 54 percent before the adjustment to about 50 percent afterward, while the profit share gained about 5 points in the same period. IMF staff calculations of profitability also show a sharp improvement. For instance, the ratio of whole- sale prices to unit labor costs increased by more than 20 percent between 1986 and 1989. The ratio of export unit values to unit labor costs displays a similar pattern.

The Danish Adjustment 1983-86

The fiscal position of Denmark deteriorated rapidly in the late 1970s. The largest deficits were registered in 1982, as the central government deficit exceeded 11 percent of GDP. The worsening of the fiscal balance was entirely due to a sharp increase in expenditures, from about 44 percent of GDP in 1978 to almost 54 percent in 1982. The ratio of gross debt to GDP reached almost 80 percent in 1982, and a large fraction of it (by OECD stan- dards) was foreign debt.17 Average bond yield peaked at about 20 percent in 1982, when inflation was about 9 percent. The real interest rate differential

17 External debt was about 35 percent of GDP.

©International Monetary Fund. Not for Redistribution 236 ALBERTO ALESINA and ROBERTO PEROTTI between Denmark and Germany was about 8 percent in 1982. In October 1982, Standard & Poor added a credit watch to the AAA rating of Danish government bonds. Also in October 1982, a convincing electoral victory established a cohe- sive conservative coalition in office in Denmark. This government launched a major fiscal adjustment program and implemented it in the following four years. At the same time, the government announced that the exchange rate was irrevocably fixed and that the series of devaluations in previous years had ended. As pointed out by Giavazzi and Pagano (1990), the credibility of this announcement was cemented by the lack of a Danish devaluation in March 1983, when the EMS underwent a general realignment.

Composition of Adjustment Table 15 shows that the turnaround of Denmark's fiscal position was remarkable: the largest in the recent history of any OECD economy. By 1989, our cyclically adjusted measure of budget deficit improved by more than 11 percent of GDP relative to 1981-82. The adjustment was divided about equally between expenditure cuts and tax increases.

Table 15. Fiscal Adjustment in Denmark: Size and Composition, 1981-89 (In percent of GDP) Change Change during during and after Before During After adjustment adjustment (1) (2) (3) (2) - (1) (3) - (1) Fiscal impulse -5.7 2.7 5.7 8.4 11.4 Primary expenditures 50.4 46.4 46.6 -4.0 -3.8 Transfers 19.7 18.6 19.0 -1.1 -0.7 Government wages 19.9 18.3 18.2 -1.6 -1.7 Nonwage government consumption 8.0 7.3 7.2 -0.7 -0.8 Public investment 2.8 2.2 2.2 -0.6 -0.6 Revenues 45.7 49.3 50.4 3.6 4.7 Taxes on households 24.4 25.6 28.1 1.2 3.7 Taxes on business 1.1 2.4 2.3 1.3 1.2 Indirect taxes 18.0 18.5 19.2 0.5 1.2 Social security contributions 2.2 2.8 2.8 0.6 0.6 Source: OECD. Note: "Before" is the two-year period before the adjustment, 1981-82; "during" is the period of the adjustment, 1983-86; "after" is the two-year period after the adjustment, 1987-89.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 237

On the expenditure side, most of the cuts were in transfer programs and government wages (see Table 15). Cuts in transfer programs were broad ranging, focusing on unemployment insurance (freezing the maximum rate and reducing abuses), changes in the parameter of pension schemes, particularly for public employees, and cuts in sickness insurance funds. Trans- fers to local governments were also reduced in 1983-86. As shown in Table 15, transfers fell by more than 1 percent of GDP during the adjustment. Public employment, which had been growing rapidly before 1982, was frozen. In the context of the income policies introduced in 1982, government sector wages grew by less than inflation. Wage indexation was sus- pended until 1987, and the automatic link of public sector wages to wage increases in the private sector was eliminated. As a result of all these measures, the government wage bill was substantially reduced. On the revenue side, most of the increase was due to direct taxes on households and businesses, with more modest increases in indirect taxes (Table 15). In the context of the policy of reducing net transfers, several social security contributions were increased, including contributions from employees to the unemployment fund and tax subsidies to private pension schemes. The result was a substantial increase in the tax burden for families during the adjustment.

Macroeconomic Consequences of Fiscal Adjustment The average growth rate during Denmark's adjustment in 1983-86 was about 3.3 percent a year, well above the major industrial country average (Table 16). Consumption and especially private investment both boomed. The latter grew by more than 13 percent a year and the former by more than 4 percent. The share of private investment grew by almost 2.5 percentage points of GDP relative to 1981-82. Exports increased steadily as a result of improved competitiveness (the reduction of relative unit labor costs), owing in part to the depreciation of the exchange rate in the early 1980s and to the wage moderation accord of 1982, coupled with the suspension of indexation clauses. The effect of this wage policy is reflected in Table 16 in the sharp drop of the wage share and the even larger increase in the profit share in the period 1983-86 relative to the previous two years. Giavazzi and Pagano (1990) emphasize the expansionary consequences of the credibility and wealth effects in these four years. Interest rates dramatically fell immediately after the fiscal adjustment was announced and the implementation began. Giavazzi and Pagano (1990) calculate that their measure of an ex ante real interest rate fell from 6.7 percent in the 1979-82 period to 3.3 percent for the 1983-86 period of fiscal adjustment. Differentials with real German interest rates were at least halved, dropping from about 8 percent to less than 4 percent. Table 16 also shows a sharp drop

©International Monetary Fund. Not for Redistribution 238 ALBERTO ALESINA and ROBERTO PEROTTI

Table 16. Fiscal Adjustment in Denmark: Macroeconomic Conditions, 1981-89 (In percent) Change Change during during and after Before During After adjustment adjustment (1) (2) (3) (2) - (1) (3) - (1) GR 0.40 0.38 -3.06 -0.02 -3.46 U 2.21 1.55 1.36 -0.66 -0.85 AI 1.69 13.42 -6.36 11.73 -8.05 AC -0.44 4.17 -1.26 4.61 -0.82 i 8.55 3.36 2.43 -5.19 -6.12 r 4.94 1.99 1.46 -2.95 -3.48 ULC -13.85 0.29 8.25 EXCH RATE -4.53 1.16 0.92 TB 0.80 0.03 1.50 -0.77 0.70 WSH 56.20 54.17 55.56 -2.03 -0.64 PSH 31.57 35.14 33.09 3.57 1.52 Source: OECD. Notes: For the definition of the variables, see Table 10. For the definition of "before," "during," and "after," see Table 15. of ex post real rates relative to the major industrial countries, from almost 5 percent before the adjustment to less than 2 percent during the adjustment, to less than 1.5 percent afterward. Another indicator of the credibility of the adjustment is the jump upward in indicators of consumer confidence. The sharp drop in real interest rates, accompanied by the wage moderation achieved in 1982, the increase in the profit share, and the reduction in unit labor costs, all contributed to the boom in private investment.

Immediate Aftermath of Adjustment Giavazzi and Pagano (1990) interpret the Irish and the Danish adjustments as two cases of expansionary fiscal consolidations and emphasize their simi- larities. However, the immediate aftermath of the two adjustments shows that they were quite different. While the Irish economy continued booming after the adjustment, the Danish economy moved into a recession in 1987-88, with growth rates of minus 0.6 percent and minus 0.2 percent, respectively. In fact, average growth in Denmark was more than 3 percent lower than in the major industrial countries in both 1987 and 1988 (Table 16). In 1989, growth was only 1.1 percent. Unemployment increased in 1988 from 7.9 percent to 8.7 percent and continued to climb in the following two years. The underlying cause of the recession in Denmark lies in the fall in competitiveness. The real effective exchange rate (relative unit labor costs) as measured by the IMF fell by 8 percent in 1986 and almost 11 percent in 1987.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 239

The current account deficit increased to almost 6 percent of GDP in 1986. The trade balance worsened during the adjustment, as shown in Table 16, particularly in 1985-86. The improvement of the trade balance after the adjustment was largely driven by the recession. The decrease in competitiveness can be only partially explained by the appreciation of the real exchange rate (triggered by the combination of rapidly falling inflation and a fixed nominal rate). An important additional factor was the wage agreement of early 1987, which introduced sizable wage increases for 1987 and 1988. Their effects, after the restraint exercised during the adjustment period, is shown in Table 16 by the rebound of the wage share of more than 1 point of GDP from the average of 1983-86. The profit share, on the contrary, fell by about 2 percentage points. Our interpretation is that the wage accord of 1987 incorporated union demands for compensation for the increases in income taxes of the previous four years, which had reduced workers' after-tax income. Thus, the tax increases might have been the originating cause of the recession, via wage negotiations, as implied by Alesina and Perotti (1997).

The Italian Adjustment 1989-95

The beginning of fiscal adjustment in Italy can be dated to 1989. How- ever, despite the reduction in the primary deficit that turned into a surplus in 1991, the debt-to-GNP ratio continued to increase, reaching 125 percent in 1994. Only in 1995 did this ratio decline slightly. Until 1993, the adjustment was completely on the revenue side. In fact, the ratio of primary spending to GNP increased between 1989 and 1993. Thus, the adjustment between 1989 and 1993 looks clearly to be a Type 2 adjustment. From 1993 to 1995, primary spending was cut and revenues fell; at first sight, therefore, one may be tempted to conclude that from 1993 onward the Ital- ian adjustment has been a Type 1 adjustment. However, a more careful examination of the evidence calls into question this interpretation.

Size and Composition of Adjustment Table 17 considers the fiscal adjustment of Italy's general government, the definition which is most compatible with the one used for the other countries in this study.18 First, the overall size of the adjustment from 1989 to 1994-95

18 Fiscal data for Italy are drawn from the IMF, specifically from the March 1996 background paper and appendices on recent economic developments in Italy. Since data for 1995 are particularly important for the Italian adjustment, we could not use our OECD data set, which we adopted above, as those data end in 1994. Also, for Italy, it is particularly useful to contrast the central government data with the general government data.

©International Monetary Fund. Not for Redistribution 240 ALBERTO ALESINA and ROBERTO PEROTTI

Table 17. Fiscal Adjustment in Italy (General Government): Size and Composition, 1989-95 (In percent of GDP) Change Change Change from from 1989 from 1989 1993 to 1989 1993 1994-95 to 1993 to 1994 1994-95 (1) (2) (3) (2) - (1) (3) - (1) (3) - (2) Revenues 42.0 48.3 45.5 6.3 3.5 -2.8 Expenditures 51.9 57.8 53.5 5.9 1.6 -4.3 Interest payments 8.9 12.1 10.8 3.2 1.9 -1.3 Primary expenditures 43.0 45.7 42.7 2.7 -0.3 -3.0 Wages 11.9 12.5 11.7 0.6 -0.2 -0.8 Social security 17.6 19.5 19.1 1.9 1.5 -0.4 Subsidies 2.5 2.3 2.1 -0.2 -0.4 -0.2 Other current expenditures 6.2 6.6 6.1 0.4 -0.1 -0.5 Fixed investment 3.3 2.7 2.2 -0.6 -1.1 -0.5 Other capital expenditures 1.5 2.1 1.5 0.6 — -0.6 Overall balance -9.9 -9.6 -8.1 0.3 1.8 1.5 Primary balance -1.0 2.6 2.8 3.6 3.8 0.2 Source: IMF.

was much smaller than those of Denmark and Ireland, even though Italy's adjustment stretched over a longer period. In terms of the share of GDP, the adjustment in the primary surplus was about one-half of the Irish adjustment and no more than one-third of the Danish adjustment, although Italy had a much higher ratio of debt to GDP than Denmark and a much heavier interest burden than Denmark or Ireland. Until 1993, the adjustment was totally on the revenue side, with revenues increasing by 6.3 percent of GDP to 48.3 percent. Meanwhile, primary expenditures rose by 2.7 percent of GDP, and total expenditures reached 57.8 percent of GDP in 1993. After the spending and tax cuts of 1994 and 1995, primary expenditures returned to their 1989 level as a share of GDP, while revenues were still 3.5 percent of GDP above the 1989 level. The reduction in revenues in the most recent two years was largely due to the one-off nature of many of the tax increases in 1992-93. The composition of the spending cuts in 1994-95 is revealing. The largest share of the cuts came from capital expenditures (1.1 percent of GDP). The remaining cuts in 1993-95 were mostly from government wages (almost 1 percent of GDP), with modest cuts on all the other items. However, despite these recent cuts, government wages in 1995 had about the same share of GDP as in 1989, and, over that same period, social security increased by 1.5 percent of GDP.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 241

The fall in public investment was partly (or largely) due to the effects of criminal investigations; a good portion of this reduction was one-off in nature. A sizable part (0.6 percent of GDP) of the total cuts in the capital spending account fell into the category of "other capital spending." This entry includes refunds of tax credits of the public. In 1993, this entry was unusually high (2.1 percent of GDP). In 1995, although tax refunds total- ing almost 1 percent of GDP were planned (Lit 16,000 billion), only Lit 700 billion were disbursed. On average, about Lit 5,000 billion are disbursed every year. Thus, the postponement of tax refunds in 1995 was largely responsible for the so-called cut in capital spending. The reduction of the wage bill resulted not only from a hiring freeze but especially from a less than full adjustment of public wages to inflation. An agreement with the unions in 1992 linked wages to planned rather than actual inflation. Since in the following three years actual inflation was higher than planned, real wages fell. The unions asked for an adjustment to wages to compensate for this difference and obtained large wage increases in the early months of 1996. Whether this adjustment will occur in full in the next few years is unclear. The point is that the reduction in real wages is the result of a rather convoluted and strategic use of projected and actual inflation, rather than a clear and permanent agreement with the unions on wage restraint.19 Social security was reduced more in 1995 than in 1994. As pointed out by the Bank of Italy (1996), these savings were temporary and due mostly to two one-off measures: (1) some temporary suspensions of "liquidazioni" (severance pay) of public employees and other postponements of pension payments, and (2) postponement of indexation payments from November 1995 to January 1996. Pension Reform In 1995, the Italian government adopted a comprehensive reform of the pension system, which had been one of the least solvent in the OECD. This reform, which followed a previous adjustment in 1992, replaced an incomes- based system with one that links benefits to contributions, although the system retains a "pay-as-you-go" nature. The reform also eliminated senior- ity pensions, which essentially permitted early retirement without penalties. In this paper, we are interested only in the effects of this reform on the current fiscal adjustment and on long-run cuts in spending. The estimates for the budget impact of the reform in the short, medium, and long run are very difficult to make because they are based on uncertain

19 See Alesina, Mare, and Perotti (1996) for a discussion of the persistent bias in the government's forecasts of inflation.

©International Monetary Fund. Not for Redistribution 242 ALBERTO ALESINA and ROBERTO PEROTTI predictions of macroeconomic and demographic variables. However, even the most optimistic ones (from the Italian Treasury) suggest that the savings both in the short run and in the long run are not very large. According to the Italian Treasury, in the medium run (1995 to 2005) the reform would save about 0.4 - 0.5 percent of GDP a year, split equally between reduced spending and increased contributions. Peracchi and Rossi (1996) argue that these estimates are overoptimistic because they rely on forecasts that are not credible. The most pessimistic estimates, from the Bank of Italy, suggest that the reform may even have a negative impact on the government budget, even after the transition phase. A recent ruling of Italy's Constitutional Court concerning large pension arrears implies a further negative fiscal shock for the pension system. These payments will be financed by debt issues and should be about equal to the amount of savings generated by the pension reform over the next five years, as calculated by the Italian Treasury. In summary, the effects of the reform on the budget are highly uncertain and even in the most optimistic scenarios not very large.20

Contractionary or Expansionary Adjustment? Table 18 clearly shows that the Italian fiscal adjustment has strong contractionary consequences for domestic demand, with the external sector containing the negative effect of domestic demand on growth. Italy's growth rate shows a declining trend from 4.1 percent in 1988 to minus 0.7 percent in 1993. In the period 1989-94, the Italian growth rate was on average 0.7 percent below the average for the major industrial countries. As shown in Table 18, private investment literally collapsed in this period: in 1993, it fell by about 18 percent! This development contrasts sharply with the cases of Ireland and Denmark, where private investment boomed during the adjustment. Growth in private consumption was very sluggish and turned negative in 1993. There was no sign—at least until 1993—that interest rates incorporated a "credibility gain bonus." The only good news for the economy in the period 1989-93 was the reduction in unit labor costs. In summary, the Italian adjustment had until 1993 all the features of a contractionary adjustment, driven completely by tax increases, rather than spending cuts.

20 Concerning the short-term fiscal adjustment, one of the problems of the reform is the widely acknowledged very slow phasing-in period. IMF staff calculations suggest that several relatively simple and politically feasible measures to accelerate the adjustment could have led to additional savings of about .50 percent of GDP in 1995 and slightly more in the following year. Although not huge, this amount is far from trivial: it is larger than the actual cuts in social security in 1994-95.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 243

Table 18. Fiscal Adjustment in Italy (General Government): Macroeconomic Conditions, 1989-94 (In percent) Change from Before During 1989 to 1994 (1) (2) (2) - (1) GR 0.21 -0.70 -0.91 U 6.60 5.63 -0.97 AI 10.6 -1.50 -12.10 AC 4.1 1.49 -2.61 i 3.03 5.13 2.10 r 0.57 2.66 2.09 ULC 5.16 -2.25 EXCH RATE -1.76 -3.13 TB -0.44 0.60 1.04 WSH 44.40 44.47 0.07 PSH 38.60 37.61 -0.99 Source: OECD. Note: For the definition of the variables, see Table 10.

Recent developments in Italy in 1994-95 are also quite instructive. The recession of 1993 was contained by the growth of almost 9 percent in export volume. The rate of growth in 1994-95 was also almost exclusively driven by the export sector. In 1994 and 1995, export volume grew by 10.7 percent and 13.6 percent, respectively, with a trade surplus of about 4 percent of GNP. These external developments were fueled mostly by the large real devaluation—about 25 percent—of the exchange rate, which began in Sep- tember 1992 when the lira left the EMS. A second factor influencing competitiveness was the relative wage moderation achieved after 1993. Note that this wage moderation coincided with the end of major tax increases. All the measures of competitiveness indicate substantial gains on the order of 20 percent for Italy in the last three years—the largest such improvement, along with Sweden's, in the OECD. The interest rate differential relative to Germany's ten-year government bonds fell in 1993 from more than 6 percent to less than 3 percent at the end of the year. By the beginning of 1995, this differential was back to almost 5 percent. The differential on 3-month treasury bills in 1995 was about 4 percent, the same level as at the beginning of 1993. This pattern points to the market's doubts about the credibility and persistence of the fiscal adjustment, even after the 1994-95 spending cuts. This behavior of interest rate differentials is in sharp contrast to the Danish and Irish experiences. This evidence points toward a clear conclusion: not only did the 1989-93 fiscal adjustment negatively affect private demand, but the 1994-95 adjust-

©International Monetary Fund. Not for Redistribution 244 ALBERTO ALESINA and ROBERTO PEROTTI ment also had the same effect and would have had negative growth consequences if it had not been accompanied by a massive depreciation of the real exchange rate. In summary, the Italian adjustment until 1993 was certainly a Type 2 adjustment. Some of the spending cuts in 1995 appear to be Type 1 adjustments, but they were accompanied by too many one-off measures and are of uncertain duration.

V. Conclusion

The single most important idea that we hope to communicate in this paper is that the composition of fiscal adjustments matters for their likelihood of success and for their macroeconomic consequences. In our view, the academic literature focuses too much on aggregate models that disregard composition and distributional effects. Also, the emphasis of the Maastricht criteria has shifted the discussion too much toward the simple arithmetic of primary surpluses, deficits, and debt-to-GDP ratios and away from the "how" of cutting deficits to meet these criteria, as if the how did not really matter. Nonetheless, it does. Rather than reviewing in detail our results, we conclude by discussing several issues open for further research:

• A fair amount of evidence suggests that, in some cases, fiscal contractions can be expansionary. An important question is through which channel the expansion occurs. We have argued that models that emphasize the effect of the composition of the adjustment on unit labor costs in unionized and open economies are at least as relevant empirically as models that focus upon credibility and wealth effects. This argument needs a more thorough empirical investigation. • Even though we have taken some steps toward linking the composition of the budget to its macroeconomic effects, one should disag- gregate even more the government accounts. For instance, our variable "transfers," which is quite important for our argument, still includes social security, unemployment compensation, and other social assistance programs. A more disaggregated study in a multicountry sample should clarify which components of the "transfers" variable are more important in determining the success or failure of adjustments. • A very important policy decision concerns the policy mix, particularly exchange rate policy, that should accompany a major fiscal adjustment. Several major successful adjustments have been preceded by devaluations, but devaluations have also accompanied some

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 245

unsuccessful adjustments. The question is whether a devaluation helps to determine the success of the adjustment and its macroeconomic consequences. We have suggested in our paper that a devaluation might help, but that it would not do the trick if the composition of the adjustment did not have the features that we have discussed at length. Further research on this topic not only is fascinating from an academic perspective but also has important policy implications, for example, in connection with the effects of fiscal adjustments before or after a monetary union. • In our statistical analysis of the sample of OECD countries, we have focused on the general government. An intriguing issue is the role of local governments during major fiscal adjustments. Are the adjustments typically carried out by the central government? How do local governments participate in the effort? Is there a shifting of responsi- bility between the two layers of government? Hints made in the discussion of the Italian case suggest that this might be an important area of exploration. • Finally, one may wonder why policymakers are often hesitant to engage in vigorous Type 1 fiscal adjustments, considering that they do not seem to have contractionary effects. Perhaps the answer lies in their distributional effects. We have hinted above that the functional distribution of income is tilted in favor of profits during successful adjustments. A more careful study of the distributional consequences of major fiscal adjustments is an excellent topic of research.

In summary, the fiscal adjustment efforts that several OECD countries are currently undertaking are strictly linked to the problem of reforming the welfare state, whose weight has been increasing in the last few decades. Any fiscal adjustment that avoids dealing with the problems of social security, welfare programs, and inflated government bureaucracies is doomed to failure.

APPENDIX

Definition of Variables

Fiscal Variables All variables used are from the OECD, unless otherwise indicated in the text. Transfers = the sum of social security benefits, social assistance grants, unfunded employee pension and welfare benefits, transfers to the rest of the world, trans-

©International Monetary Fund. Not for Redistribution 246 ALBERTO ALESINA and ROBERTO PEROTTI fers to private nonprofit institutions serving households, net casualty insurance premiums, and other transfers. Public investment = government gross fixed capital formation, that is, the outlays, purchases, and own-account production of producers of government services on additions of new durable goods (commodities) to their stocks of fixed assets. Excluded are the outlays of government services on durable goods for military use. Government consumption (divided into its wage and nonwage components) = the value of goods and services produced for their own use on current account, that is, the value of their gross output less the sum of the value of their commodity and noncommodity sales and the value of their own-account capital formation not seg- regated to an industry. The value of their gross output is equal to the sum of their intermediate consumption of goods and services, compensation of employees, consumption of fixed capital, and indirect taxes. Subsidies = all grants on the current account made by the government to private industries and public corporations. Direct taxes on income = levies by public authorities at regular intervals, except social security contributions, on income from employment, property, capital gains, or any other source. Indirect taxes = taxes assessed on producers in respect of their production, sale, purchase, or use of goods and services that they charge to the expenses of production. Also included are import duties and the operating surplus, reduced by the normal mar- gin of profits of business units and of fiscal and similar monopolies of government. Social security contributions = social security contributions received by the government. For a more detailed discussion of the construction of a measure of primary deficit and the cyclical adjustment, see the appendix of Alesina and Perotti (1995a).

Other Variables GR = (yearly rate of GDP growth of each country year) - (average GDP growth of seven major industrial countries, with GDP weights). U = (unemployment rate of each country year) - (average unemployment of seven major industrial countries, with GDP weights). AI = rate of growth of private business investment. AC = rate of growth of private consumption. i = (nominal interest rate on ten-year government bonds) - (average nominal interest rate on ten-year government bonds in seven major industrial countries, with GDP weights). r = (nominal interest rate on ten-year government bonds — inflation) - (average real interest rate in seven major industrial countries, with GDP weights). The real interest rate in each of the seven industrial countries is obtained as the nominal rate minus inflation. ULC = rate of change of unit labor costs in manufacturing, as calculated by Alesina and Perotti (1997) on OECD data. EXCH RATE = rate of change of the nominal effective exchange rate. TB = trade balance. VAULC = rate of change of the value-added deflator over unit labor costs in manufacturing. WSH = wage share. PSH = profit share.

©International Monetary Fund. Not for Redistribution FISCAL ADJUSTMENTS IN OECD COUNTRIES 247

REFERENCES

Alesina, Alberto, and Allan Drazen, 1991, "Why Are Stabilizations Delayed?" American Economic Review, Vol. 81 (December), pp. 1170-88. Alesina, Alberto, Mauro Mare, and Roberto Perotti, 1996, "Le Procedure di Bilancio in Italia: Analisi e Proposte," Swedish Economic Policy Review, No. 3 (Spring). Alesina, Alberto, and others, 1992, "Default Risk on Government Debt in OECD Countries," Economic Policy, Vol. 7 (October), pp. 427-63. Alesina, Alberto, and Roberto Perotti, 1995a, "Fiscal Expansions and Adjustments in OECD Countries," Economic Policy, No. 21 (October), pp. 205-40. , 1995b, "Taxation and Redistribution in an Open Economy," European Economic Review, Vol. 39 (May), pp. 961-79. , 1996, "Reducing Budget Deficits," Swedish Economic Policy Review, No. 3 (Spring), pp. 113-34. , 1997, "The Welfare State and Competitiveness," American Economic Review (forthcoming). Alesina, Alberto, Alessandro Prati, and Guido Tabellini, 1990, "Public Confidence and Debt Management: A Model and Case Study of Italy," in Public Debt Management: Theory and History, ed. by Rudiger Dornbusch and Mario Draghi (Cambridge and New York: Cambridge University Press), pp. 94-118. Bank of Italy, 1996, Bollettino Economico, No. 22 (February). Barro, Robert J., 1981, "Output Effects of Government Purchases," Journal of Political Economy, Vol. 89 (December), pp. 1086-121. , 1989, "The Neoclassical Approach to Fiscal Policy," in Modern Business Cycle Theory, ed. by Robert J. Barro (Cambridge, Massachusetts: Harvard University Press), pp. 178-235. Bartolini, Leonardo, Assaf Razin, and Steve Symansky, 1995, "G-7 Fiscal Restruc- turing in the 1990s: Macroeconomic Effects," Economic Policy, No. 20 (April), pp. 111-46. Baxter, Marianne, and Robert G. King, 1993, "Fiscal Policy in General Equilib- rium," American Economic Review, Vol. 83 (June) pp. 315-34. Bertola, Giuseppe, and Allan Drazen, 1993, "Trigger Points and Budget Cuts: Explaining the Effects of Fiscal Austerity," American Economic Review, Vol. 83 (March), pp. 11-26. Blanchard, Olivier J., 1985, "Debt, Deficits, and Finite Horizons," Journal of Poli- tical Economy, Vol. 93 (April), pp. 223-47. , 1990a, "Comment on Giavazzi and Pagano," in NBER Macroeconomics Annual 1990, ed. by Olivier J. Blanchard and Stanley Fischer (Cambridge, Massachusetts: MIT Press), pp. 111-16. , 1990b, "Suggestions for a New Set of Fiscal Indicators," OECD Econom- ics and Statistics Department Working Papers No. 79 (Paris: Organization for Economic Cooperation and Development). Dornbusch, Rudiger, 1989, "Credibility Debt and Unemployment: Ireland's Failed Stabilization," Economic Policy, No. 4 (April), pp. 173-209. Giavazzi, Francesco, and Marco Pagano, 1990, "Can Severe Fiscal Adjustments Be Expansionary?" in NBER Macroeconomics Annual 1990, ed. by Olivier J. Blan- chard and Stanley Fischer (Cambridge, Massachusetts: MIT Press), pp. 75-111.

©International Monetary Fund. Not for Redistribution 248 ALBERTO ALESINA and ROBERTO PEROTTI

, 1996, "Non-Keynesian Effects of Fiscal Policy Changes: International Evidence and the Swedish Experience," Swedish Economic Policy Review, No. 3 (Spring), pp. 135-65. Grilli, Vittorio, Donato Masciandaro, and Guido Tabellini, 1991, "Political and Monetary Institutions and Public Finance Policies in the Industrial Countries," Economic Policy, Vol. 6 (October), pp. 342-92. Lane, Philip R., and Roberto Perotti, 1995, "Profitability, Fiscal Policy, and Exchange Rate Regimes" (unpublished; Cambridge, Massachusetts: Harvard University). McDermott, C. John, and Robert F. Wescott, 1996, "The Economics of Fiscal Con- solidation" (unpublished; Washington: International Monetary Fund). McKenzie, G. A., 1989, "Are All Summary Indicators of the Stance of Fiscal Policy Misleading?" Staff Papers, International Monetary Fund, Vol. 36 (December), pp. 743-70. Miller, Marcus, Robert Skidelsky, and Paul Weller, 1990, "Fear of Deficit Financing—Is It Rational?" in Public Debt Management: Theory and History, ed. by Rudiger Dornbusch and Mario Draghi (Cambridge and New York: Cambridge University Press), pp. 293-310. Pencavel, John, 1986, "Labor Supply of Men: A Survey," in Handbook of Labor Economics, Vol. 1, ed. by Orley Ashenfelter and Richard Layard (Amsterdam and New York: North-Holland), pp. 3-102. Peracchi, Franco, and Nicola Rossi, 1996, "Nonostande Tutto e Una Riforma," in Le Nuove Frontiere Della Politica Economica, ed. by F. Galinoberti and others (Milan, Italy: Il Sole 24 Ore). Perotti, Roberto, 1996a, "Fiscal Consolidation in Europe: Composition Matters," American Economic Review, Papers and Proceedings, Vol. 86 (May), pp. 105-10. , 1996b, "Do We Know How to Estimate Discretionary Fiscal Policy?" (unpublished; New York: Columbia University). Rogoff, Kenneth, 1990, "Equilibrium Political Budget Cycles," American Eco- nomic Review, Vol. 80 (March), pp. 21-36. Roubini, Nouriel, and Jeffrey Sachs, 1989, "Political and Economic Determinants of Budget Deficits in the Industrial Democracies," European Economic Review, Vol. 33 (May), pp. 903-33. Spolaore, Enrico, 1993, "Policy Making Systems and Economic Efficiency: Coali- tion Governments Versus Majority Governments" (unpublished). Sutherland, Alan, 1995, "Fiscal Crises and Aggregate Demand: Can High Public Debt Reverse the Effect of Fiscal Policy?" CEPR Discussion Paper No. 1246 (London: Centre for Economic Policy Research). Tanzi, Vito, 1994, "International Systems of Public Expenditure Auditing: Lessons for Italy," in Nuovo Sistema di Controlli Sulla Spesa Publica, by the Bank of Italy (Rome: Bank of Italy), pp. 303-24. , and Ludger Schuknecht, 1995, "The Growth of Government and the Reform of the State in Industrial Countries," IMF Working Paper 95/130 (Washington: International Monetary Fund). Velasco, Andres, 1996, "A Model of Fiscal Deficits and Delayed Fiscal Reforms" (unpublished; New York: New York University).

Is the Phillips Curve Really a Curve? Some Evidence for Canada, the United Kingdom, and the United States

GUY DEBELLE and DOUGLAS LAXTON*

Previous tests for convexity in the Phillips curve have been biased because researchers have employed filtering techniques for the nonaccelerating inflation rate of unemployment (NAIRU) that have been fundamentally inconsistent with the existence of convexity. This paper places linear and nonlinear models of the Phillips curve on an equal statistical footing by estimating model-consistent measures of the NAIRU. After imposing plausible restrictions on the variability in the NAIRU, we find that the nonlinear model fits the data best. The implications for the macroeconomic policy debate are that policymakers who are unsuccessful in stabilizing the business cycle will induce a higher natural rate of unemployment. [JEL C51, E31, E52]

SSESSMENTS OF THE RATE of unemployment that is consistent with A holding inflation stable (at a low rate) represent an integral part of the monetary policy framework in most industrial countries. Whether the current rate of unemployment lies above or below the rate that is consistent with inflation stability is a key input into the monetary policy decision- making process. Unfortunately, the rate of unemployment that is consistent with inflation stability is not directly observable; policymakers can estimate it only by using other, more observable pieces of information about the state of the economy. *Guy Debelle was an Economist in the IMF's Research Department when this paper was completed. He is now with the Reserve Bank of Australia. He holds a Ph.D. from MIT. Douglas Laxton is a Senior Economist in the Research Depart- ment. He did his graduate studies at the University of Western Ontario. The authors would like to thank Irene Chan for excellent research assistance and Helen Hwang for secretarial assistance. They also thank Richard Black, Peter Clark, Hamid Faruqee, Joshua Felman, Peter Isard, Eswar Prasad, David Rose, Hossein Samiei, Demos Tombakis, and Alun Thomas for helpful discussion and comments. 249

©International Monetary Fund. Not for Redistribution 250 GUY DEBELLE and DOUGLAS LAXTON

For the United States, estimates of Phillips curves and the nonaccelerating inflation rate of unemployment (NAIRU) were regularly published in the Brookings Papers on Economic Activity during the 1970s and early 1980s by Robert Gordon (for example, 1970, 1975, and 1977). After a period of inactivity in the 1980s when Phillips curves were generally assumed to have broken down, a number of recent papers (Gordon, 1994; and Fuhrer, 1995) have sought to reestimate the Phillips curve and derive new estimates of the natural rate of unemployment. The general approach has been to regress inflation on a measure of expected or lagged inflation, unemployment gaps, and dummy variables to control for various supply shocks, such as the oil shocks and the Nixon price controls. Despite the variety of techniques used, the common feature of these recent estimates is their use of a linear Phillips curve. A separate strand of the literature has presented evidence supporting the concept of a nonlinear Phillips curve (Macklem, forthcoming; Clark, Laxton, and Rose, 1996; Turner, 1995; and Laxton, Meredith, and Rose, 1994). A convex nonlinear Phillips curve implies that a given fall in the unemployment rate below the NAIRU causes a larger rise in inflation than does a rise of the same magnitude produce a fall in inflation. The literature on the relationship between inflation and unemployment frequently refers to both the NAIRU and the natural rate of unemployment. With linear specifications of the Phillips curve, the two terms are often used interchangeably. More generally, when allowing for the possibility of a nonlinear Phillips curve and stochastic variability in demand and supply, it is useful to distinguish the NAIRU at a point in time from the expected value over time of the unemployment rate that would be consistent with nonaccelerating inflation, given the stochastic distribution of shocks. We define the former as the deterministic NAIRU because this is the unemployment rate that is consistent with nonaccelerating inflation in the absence of shocks. In the remainder of this paper, we will refer to the deterministic NAIRU as either "D-NAIRU" or "u*." AS shown below, the true stochastic NAIRU must be greater than u* when there is convexity in the Phillips curve. Following Friedman (1968), this paper uses the term "natural rate of unemployment" to refer to the NAIRU because this is the only sustainable average level of unemployment that would be consistent with nonaccelerating inflation. The nonlinearity in the model implies that policymakers who are less successful in stabilizing the business cycle will induce a larger natural rate of unemployment in their economies (De Long and Summers, 1988; and Laxton, Rose, and Tetlow, 1993b). Laxton, Rose, and Tetlow (1993a) demonstrate that previous tests for nonlinearity have been severely compromised because researchers have tended to rely upon ad hoc prefiltering techniques that are grossly inconsistent with

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 251

the key implications of a nonlinear Phillips curve. In this paper, we derive estimates of u* and the natural rate of unemployment for three countries— Canada, the United Kingdom, and the United States—by extracting information from a nonlinear model of the Phillips curve. As in Kuttner (1992 and 1994), we use a Kalman filter and a maximum likelihood procedure to simultaneously estimate the parameters of the model along with model-consistent estimates of u*. We also estimate a linear model of the Phillips curve using the Kalman filter. We show that the nonlinear model fits the data better than the linear model for all three of these countries when plausible restrictions are imposed on the variance of the natural rate. The first section describes the linear and the nonlinear models that we estimate and describes the Kalman filtering technique that we use in the estimation. The derivation of the inflation expectations series, which is a key component of our model, is described in Section II. Section III presents the results, while Sections IV and V discuss the uncertainty and sensitivity of the results. Section VI concludes.

I. Models and Estimation Technique

This section describes the models and the procedure that we use to estimate the NAIRU. The standard linear expectations-augmented Phillips curve has the following functional form:

e nt = n t + y(u* - ut) + et, (1) where n is inflation, u is the unemployment rate, and u* is the D-NAIRU. The inflation expectations ne are a linear combination of a backward- and forward-looking component (see the discussion in Section II). The backward-looking component could also reflect inertia in the inflation process. For example, an overlapping-contracts model, such as Fischer (1977), could motivate such a result:

e n t = X A-1(L)nt-i + (1 - X)B(L)nt-i, (2) where A(L) and B(L) are polynomial lag operators. Equations such as (1) have been estimated by Gordon in the 1970s and early 1980s (Gordon, 1970, 1975, and 1977). To proxy inflation expectations, Gordon used lags of the inflation rate of up to two years and also controlled for a number of supply-side influences, such as price controls, relative price changes, and real exchange rate changes. Furthermore, Gordon

©International Monetary Fund. Not for Redistribution 252 GUY DEBELLE and DOUGLAS LAXTON imposed the constraint that the coefficients on lagged inflation sum to one. Evidence in favor of this restriction is still sometimes interpreted as support for the long-run natural rate hypothesis despite Sargent's (1971) explanation that this restriction has nothing to do with the long-run natural rate hypothesis. The restriction is inappropriate because it implies that agents always forecast inflation as if it contained a unit root. In the case of the United States, studies either have assumed that the natural rate of unemployment has been constant over the sample period or have estimated some small shifts to control for changes in the composition of the labor force. More recently, Gordon (1994), Tootell (1994), and Fuhrer (1995) have estimated similar models, which also assume that the natural rate is constant over the sample period. This assumption is supported by both within-sample and out-of sample tests that fail to reject the hypothesis of a constant natural rate. While this assumption may be difficult to reject in the United States, it is likely to be rejected in countries where the natural rate has clearly moved over time (such as the United Kingdom and Canada). Staiger, Stock, and Watson (1996) employ a variety of techniques to derive estimates of the natural rate based on Gordon's approach and on univariate analysis. Their results highlight the uncertainty associated with the estimates. They obtain estimates of the natural rate by assuming, alternatively, that it is constant over the sample period, a constant with occasional shifts, an unobserved random walk, and a specific function of labor market variables. Under one of these assumptions, for example, the degree of uncertainty associated with a point estimate of 6.2 percent in 1990 is a 95 percent confidence interval of 5.1-7.7 percent. In addition, the point estimates vary quite substantially across the different techniques. Kuttner (1992 and 1994) adopts a strategy that is the closest to the one adopted in this paper. He estimates a model of the Phillips curve allowing for time variation in the level of potential output. He employs a Kalman filter to extract an estimate of the level of potential output, where potential output is assumed to be a random walk with positive drift. However, as in the other models, Kuttner assumes that the Phillips curve has a linear specification. The key difference between this paper and the previous literature is that we estimate u* in the context of a nonlinear Phillips curve. The nonlinear Phillips curve used here is assumed to have a simple structure of the form

(u*t - ut) e nt = n t + y + et, (3)

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 253 where u is the observed unemployment rate and u* is the time-varying unemployment rate at which inflation is constant.1 A nonlinear Phillips curve may be motivated by the traditional concept of an upward-sloping aggregate supply curve. As the unemployment rate falls below the NAIRU, bottlenecks start to develop that cause further increases in demand to have even larger inflationary consequences. Once the unemployment rate reaches some lower bound, inflation will increase at an almost infinite rate. As mentioned above, a distinction can be made in the nonlinear model between u* and the natural rate that cannot be made in the linear model. If one defines the natural rate as the expected value of unemployment in the stochastic steady state, the convexity of the nonlinear Phillips curve implies that the natural rate of unemployment will lie above u* by a constant a that embodies the degree of convexity and the nature of the stochastic shocks. Meanwhile, in the traditional Phillips curve model, the linearity ensures that u* and the natural rate are the same. This distinction is most easily demonstrated if we assume that the Phillips curve takes a slightly different functional form:

7t, = et + exp[Y(u* - ut)] - 1 + et. (4)

The D-NAIRU is given by u*: when the rate of unemployment is equal to u*, inflation is equal to inflation expectations and is neither rising nor falling. If we assume that the error term is normally distributed with zero mean, the average rate of unemployment is given by u* + y var(ut)/2. That is, the stochastic steady state rate of unemployment, which we interpret as the natural rate (in the sense of Friedman, 1968), is greater than u*. Fur- thermore, policies that reduce the variance of unemployment will reduce the natural rate of unemployment. That is, a policymaker who is effective in stabilizing the business cycle will reduce the gap between u* and the natural rate. Figure 1 is a useful device for illustrating the implications that such asymmetry has for stabilization policy. The nonlinear curve in the figure, the "Phillips curve," depicts the short-run relationship between inflation adjusted for inflation expectations, n - ne, and the unemployment rate u. The key assumption underlying the Phillips curve is that the slope of the curve, or the trade-off between unemployment and inflation, worsens as unemployment falls significantly below u*. For illustrative purposes, we

1 This functional form has been used to model the unemployment-inflation process at the Australian Treasury (Downes and Johnson, 1994). It is approximately linear in the region where unemployment is equal to u*bu t allows for the intuitively appealing restriction that the unemployment rate cannot fall below zero.

©International Monetary Fund. Not for Redistribution 254 GUY DEBELLE and DOUGLAS LAXTON

Figure 1. The Phillips Curve, the D-NAIRU, and the Natural Rate of Unemployment (In percent)

Range of variability in unemployment 4*

O=l

1 2 10 11 12 r Minimum Unemployment rate (u) short-run unemployment rate

D-NAIRU (u *) Natural rate of unemployment (NAIRU)

Note: a is the difference between the average historical rate of unemployment and the D-NAIRU. assume that u* is equal to 5 percent. Figure 1 illustrates this point by incorporating the incontrovertible assumption that, even in the short run, there is some minimum level of unemployment that cannot be obtained through expansionary policies to manage aggregate demand. For purely illustrative purposes, it is assumed that this minimum level of unemployment is equal to 1 percent of the labor force. Because excess demand conditions are more inflationary than excess supply conditions are disinflationary, allowing the economy to enter the region of excess demand implies that the economy will have to operate longer in the region of excess supply to prevent inflation from drifting upward over time. Thus, if disturbances to the economy cause the unemployment rate to vary over time, the natural rate of unemployment—the average unemployment rate that is consistent with stable inflation—will be higher than the u* that enters the Phillips curve because unemployment will have to spend more time above u* in order to offset the greater inflationary tendencies that will be associated with periods when it falls below u*.

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 255

For illustrative purposes, it is assumed in Figure 1 that the unemployment rate varies between 4 percent and 8 percent, and that the natural rate of unemployment is 6 percent, or 1 percentage point above u*. This asymmetry in the unemployment-inflation process has important policy implications. Stabilization policy that is not successful in reducing the variability in the business cycle can have undesirable consequences, not only for the variance of unemployment but also for the natural rate of unemployment. Figure 2 illustrates this point by considering the alternative case in which unsuccessful stabilization policy allows the unemployment rate to vary over a wider range than in the top panel, which periodically subjects the economy to serious overheating. In this case, the natural rate (shown as 7 percent) will be even higher than in the top panel because it will take larger excess supply or recessionary states to offset the greater inflationary impetus caused by periodically subjecting the economy to serious overheating.

Figure 2. Implications of Greater Unemployment Variability for the Natural Rate of Unemployment (In percent) k, Range of variability in unemployment

oc=2

1 2 10 12 r Minimum Unemployment rate (u) short-run unemployment rate

D-NAIRU (u*) Natural rate of unemployment (NAIRU)

Note: a is the difference between the average historical rate of unemployment and the D-NAIRU.

©International Monetary Fund. Not for Redistribution 256 GUY DEBELLE and DOUGLAS LAXTON

The concept of the NAIRU used in this paper is the one generally used in discussions of the "natural rate." The NAIRU will be affected by the operation of the labor market and embodies the "actual structural characteristics of the labor and commodity markets, including market imperfections" (Friedman, 1968, p. 8). Labor market policy will thus affect the level of the NAIRU through time. However, the distinction that we are focusing on here is that macroeconomic policy will have a further impact on the natural rate through its effect on the variability of the macroeconomy. Previous work has tried to estimate the determinants of the NAIRU directly with mixed success. For example, Lilien (1982) examines the effect of sectoral changes on the structural unemployment rate. Blanchard and Katz (1996) provide a recent summary of this research. While acknowl- edging that changes in labor market policy and institutional factors will affect the NAIRU, we do not directly identify the impact of these factors on the NAIRU in our model. The nonlinear functional form that we employ in our empirical work allows for a hyperbolic shape that imposes an asymptote at zero. In economic terms, this implies that, as the unemployment rate approaches zero, inflation increases at a higher and higher rate. Allowing the asymptote to be at zero is perhaps being overly conservative. It is likely that the economy will run up against insurmountable supply constraints before the unemployment rate reaches zero. Nevertheless, a zero unemployment rate provides an uncontroversial lower limit for the asymptote. u* is allowed to be time varying in both the linear and the nonlinear models. To estimate the Phillips curve, we need estimates of u*, which is not directly observable. The Kalman filter allows us to estimate the model while simultaneously providing a time-series estimate of the NAIRU. The Kalman filter estimates models of the general form:2

2 yt = Xt 'Bt + €t et ~ N(0, Q H), and (5)

Bt = T*Bt-1 +ut ut ~ N(0, O2Q). (6)

The parameter vector B is time varying in a manner determined by the transition equation (6). In our models, we allow u* to be time varying and, more particularly, a random walk. As mentioned above, u* will be affected by structural changes in the labor market, including those induced by labor market policy. However, for estimation purposes, we have not directly modeled the effects of such changes.

2 For more information on the Kalman filter, see Harvey (1981).

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 257

As Kuttner (1994) points out, estimating the natural rate using a Kalman filter has three advantages. First, it allows us to use the information present in the difference between inflation and inflation expectations, rather than relying solely on the univariate properties of the unemployment rate. Sec- ond, it allows for smooth, continuous updating of the estimate in real time as new data become available. This will be a particular advantage in the context of policymaking decisions. Third, we can derive estimates of the uncertainty about the natural rate, as in Staiger, Stock, and Watson (1996). Operationally, we estimate the linear model with the Kalman filter, allowing the constant term yu* in equation (1) to be time varying. We assume that the matrices H and T are identity matrices, and the matrix Q is constructed so that only the constant term is time varying. In the nonlinear model, the coefficient on the inverse of the unemployment rate is time varying. In both models, the time series for u* is obtained by dividing the time varying parameter obtained from the Kalman filter by the (time-invariant) estimated coefficient y. Application of the Kalman filter generates two time series of u*. The first comes from the recursive estimation of the model, which uses data that are available only up to the current period (referred to by Kuttner (1992) as "one-sided" estimates). The second (smoothed or "two-sided") series uses data from the whole sample to estimate a time series for u* and the parameters of the model to maximize the likelihood function. The one-sided estimates allow an assessment of how the model performs in "real time." The information that we use to identify movements in u* is the difference between inflation and inflation expectations. We proxy the backward- looking component of inflation expectations by the first lag of inflation. The estimates of the forward-looking component of inflation in this paper are derived from information present in long-term bond rates. The technique used in this latter step is described in the next section.

II. Data

All data are quarterly data for Canada, the United Kingdom, and the United States; the data are drawn from the Analytical Data Bank of the Orga- nization for Economic Cooperation and Development (OECD). The unemployment and consumer price series are based on official seasonally adjusted estimates for each country. Our approach to measuring inflation expectations is based on the simple idea that long-term interest rates may embody valuable information about policy credibility—for example, see Goodfriend (1993) and McCallum (1996). We use information from bond markets to develop a measure for long-term inflation expectations, nLTEt, which, in turn, is used to

©International Monetary Fund. Not for Redistribution 258 GUY DEBELLE and DOUGLAS LAXTON

e help identify the short-term measure of inflation expectations, n t, that enters the Phillips curve. This approach assumes that the inflation expectations of wage and price setters are related to the inflation premiums that participants in the bond market demand to hold long-term, fixed-income securities. Specifically, it is assumed that short-term inflation expectations are a weighted average of long-term inflation expectations and actual inflation in the previous period, dnLTEt + (1 - 8)t-1, where 5 represents the weight attached to information from the bond market. The measure of long-term inflation expectations for each country is constructed by subtracting an alternative measure of the equilibrium world real interest rate from a measure of long-term interest rates. After constructing these proxies for long-term inflation expectations, we then test to see whether these proxies provide significant explanatory power for identifying movements in the Phillips curve. The estimates of the equilibrium world real interest rate that we employ are based on recent empirical work that suggests that the equilibrium real interest rate has been gradually rising over time in response to the large buildup in world government debt—for example, see Ford and Laxton (1995), Helbling and Wescott (1995), and Tanzi and Fanizza (1995). The basic approach that we follow to derive estimates of the world equilibrium real interest rate is quite simple. First, in order to provide a reason- able benchmark for the equilibrium world real interest rate at the end of the sample, we use data on indexed bonds. Second, after obtaining this benchmark for the end of the sample, we rely on estimates of the effects of government debt on real interest rates to construct a time series that extends back to the early 1970s. Our preferred set of estimates is based on the empirical work by Tanzi and Fanizza (1995), which suggests that a 1 percentage point increase in the gross government debt-to-GDP ratio increases the world real interest rate by about 7 basis points. Based on a benchmark estimate of 4.5 percent for the real interest rate at the end of the sample, the upward trend in gross public debt since the late 1970s would suggest that the equilibrium real interest rate has gradually shifted up from an average value of 2.4 percent in the 1970s to 4.5 percent by the first quarter of 1994.3 Other estimates were also constructed using alternative values of 3.5 percent and 5.5 percent for the end-of-sample benchmark. In addition, because the effects of government debt on real interest rates vary significantly across different empirical studies, we also calculated

3 The benchmark at the end of the sample was estimated principally on the basis of data on indexed government bonds in Canada.

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 259 estimates that were based on both smaller and larger estimates of the effect of debt buildup on the equilibrium world real interest rate. First, to be consistent with the estimates reported by Ford and Laxton (1995) and Helbling and Wescott (1995), we doubled the assumed effect of world government debt on the equilibrium real interest rate. Second, we assumed that government debt had no effect on the equilibrium real interest rate and that the equilibrium real interest rate has thus been constant since the early 1970s. This last assumption is more consistent with earlier empirical work that suggested that government debt has had no discernible effect on real interest rates—for example, see Evans (1985). We found that, despite the different assumptions that were used to construct the equilibrium real interest rate, the resultant measures of long-term inflation expectations were highly significant in all the regressions that were considered.4 In fact, in not one of the alternative specifications described above could we reject the hypothesis that long-term bond yields provide significant explanatory power for identifying historical shifts in the Phillips curve. This result is encouraging because, in theory, there should be some relationship between long-term inflation expectations of participants in the bond market and the more short-term inflation expectations that influence the decisions of wage and price setters. In this vein, our results tend to support Goodfriend's (1993) conclusion that most of the high-frequency variations in long-term bond yields are driven by inflation scares rather than by historical movements in the ex ante real rate of interest.

III. Estimation Results

This section presents empirical estimates of Phillips curves and Phillips lines for Canada, the United Kingdom, and the United States. The basic empirical strategy followed in this paper is to derive model-consistent measures of u* under the assumption that the Phillips curve is really a curve, and then to compare the results of this model with an alternative model that assumes that the curve is linear. As seen at the top of Table 1, the first model that we estimate assumes that the amount of the actual change in inflation relative to expected rates, n - ne, is related to the proportional difference between u* and the actual unemployment rate, (u* - u)/u. This specification hypothesizes that the relationship between inflation and unemployment is approximately linear and symmetric in the region where unemployment is close to u* but

4 We do not report all of the results in this paper, but they can be obtained from the authors.

©International Monetary Fund. Not for Redistribution 260 GUY DEBELLE and DOUGLAS LAXTON

Table 1. Estimates of Phillips Curves with Model-Consistent D-NAIRUs (t-statistics in parentheses)

Estimated

equation: nt = net + y (U* - u)/u + eIIt, where n = year-on-year percent change in the CPI, ret = inflation expectations, 5 rLTEt + (1 - 5)rt-1 (in percent), rLTEt = long-term inflation expectation proxy from bond market (in percent), 5 = estimated weight on inflation proxy from bond market, y = parameter for measure of labor market tightness (in percent), u = unemployment rate (in percent), u* = D-NAIRU estimates derived from Kalman filter (in percent), LLF = value of the likelihood function, a2 = variance of residuals, a = average historical value of u - u* (in percent), Q. = maximum absolute value of u* - u (in percent), and x = maximum absolute quarterly change in u* (in percent).

Estimation period: 1971:Q2 to 1995:Q2. 2 Country y 5 LLF a a Q T Canada 2.37 0.15 -104.03 0.41 0.86 5.10 0.21 (4.65) (3.47) United Kingdom 2.43 0.20 -176.41 1.81 0.57 3.79 0.65 (3.70) (4.48) United States 3.55 0.08 -104.85 0.36 0.33 4.01 0.15 (7.32) (2.45) becomes nonlinear as the unemployment rate moves further away from u*. The results of the assumption of a linear structure, in which the (u* - u)/u term is replaced by (u* - u), are reported in Table 2. The estimated parameters were obtained using the maximum likelihood Kalman filter routine in TSP. This technique builds up model-consistent estimates of u* under the assumption that we can approximate historical shifts in u* by a random walk.5 In both tables, we include the value of the likelihood function (LLF), the estimated parameter values, and their associated standard errors. We also report the maximum absolute gap between u* and the actual unemployment rate u, as well as the maximum absolute quarterly change in u*, to provide some indication of how jumpy the u*'s have to be to explain inflation in these countries.

5 u* does not literally have to follow a random walk. This simple, parsimonious process was chosen because it is flexible enough to allow for permanent shifts in u* in finite samples.

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 261

Table 2. Estimates of Phillips Lines with Model-Consistent D-NAIRUs (t-statistics in parentheses)

Estimated

equation: nt - net + y (u* - u) + eIIt, where n = year-on-year percent change in the CPI, ret = inflation expectations, 5 nLTEt + (1 - 5)T-1_I (in percent), nLTEt - long-term inflation expectation proxy from bond market (in percent), 5 = estimated weight on inflation proxy from bond market, y = parameter for measure of labor market tightness (in percent), u = unemployment rate (in percent), U* = D-NAIRU estimates derived from Kalman filter (in percent), LLF = value of the likelihood function, a2 = variance of residuals, a = average historical value of u - u* (in percent), Q, = maximum absolute value of u* - u (in percent), and x = maximum absolute quarterly change in u* (in percent).

Estimation period: 1971:Q2 to 1995:Q2. 2 Country y 5 LLF a a Q I Canada 0.09 0.21 -103.49 0.18 _ 17.47 7.66 (0.77) (3.23) United Kingdom 0.63 0.33 -173.40 0.80 — 8.99 3.06 (2.36) (5.20) United States 0.55 0.25 -99.73 0.15 — 4.06 1.33 (0.13) (0.06)

Phillips Curve with Model-Consistent u*'s

As can be seen in Table 1, the estimated parameters 8 for the forward- looking proxy for long-term inflation expectations, as well as the estimated parameters y for the measure of labor market tightness, are statistically significant. The estimates of 8 can be interpreted as implying that wage and price setters place a weight of between 8 percent and 20 percent on forward- looking inflation expectations. A higher value of y implies not only that a given amount of excess demand has a larger inflationary impact but also that a given amount of excess supply has a larger disinflationary impact. Table 1 also shows estimates of a,6 the difference between the average historical rate of unemployment and u* (recall Figures 1 and 2).7 As emphasized

6 Standard errors are not reported for the estimate of a. Standard errors could be obtained in an extended system that included an equation for the unemployment gap, in addition to the Phillips curve. 7 Our estimate of a for the United States is fairly close to that obtained by Mankiw (1988).

©International Monetary Fund. Not for Redistribution 262 GUY DEBELLE and DOUGLAS LAXTON

earlier, in economies where policymakers have been less successful in avoiding boom-and-bust cycles (which may depend on the nature of the shocks hitting the economy), we should expect to observe a larger difference between the average rate of unemployment and u*. The estimates of a shown in Table 1 suggest either that policymakers in the United States have been more successful in stabilizing their business cycle than their counterparts in the other two countries, or that the shocks that have hit the U.S. economy have been smaller. Other things being equal, this would have contributed to a lower natural rate of unemployment (or average rate of unemployment) in the United States. The value of a for each country is assumed to be constant here. However, in reality it is likely to be time varying. The value depends on how successful past policymakers were in stabilizing the business cycle. If policymakers have become more successful over time at avoiding large boom-and-bust cycles, one should expect that a would shift down over time as the economy moves to a new stochastic steady state. For example, after the recent disinflation episodes in Canada and the United Kingdom and the subsequent relatively stable inflation process, one might expect that the value of a might decline, reflecting the attainment of a new regime with an improved stabilization performance. However, as suggested in Section IV, a might decline very gradually if it takes a long time for the new regime to develop credibility. The model-consistent estimates of u* from the Phillips curve are plotted in Figure 3. This methodology for measuring u* produces estimates that are fairly smooth, even though the random walk assumption in the measurement equation is capable of producing fairly jumpy measures of u* if this were necessary to explain inflation in these countries. In fact, the estimates suggest that, from the early 1970s to 1995, the maximum quarterly change in u* in the United States was only 0.15 percentage point. Although the jumps in u* are somewhat larger for Canada and the United Kingdom, the estimated u* series are also fairly smooth in these countries. According to these estimates, the increase in u* from the low levels measured for the 1950s and 1960s started sooner in Canada and the United States than in the United Kingdom.8 In particular, based on this methodology, u* in the United Kingdom, although fairly low at the beginning of the sample (for example, 3.3 percent in 1973 :Q1), shows a much stronger upward trend than the other two countries over this estimation period. Indeed, these estimates indicate that there has been a slight tendency for u* to decline over the last decade in the United States.

8 We do not report measures for this earlier period using our methodology because data for aggregate government debt in the OECD countries were not readily available.

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 263

Figure 3. Unemployment, the D-NAIRU, and the Natural Rate of Unemployment, Based on the Nonlinear Model (In percent)

— D-NAIRU (u*) — Natural rate of unemployment ----- Unemployment rate 4 14 Canada 2 - 12

0 — 10

6 — 6

4 - - 4

2 — — 2 0 0 i i i 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994

Source: OECD, Analytical Data Bank.

©International Monetary Fund. Not for Redistribution 264 GUY DEBELLE and DOUGLAS LAXTON

The estimates of the gap between u* and the unemployment rate are broadly consistent with a classical characterization of business cycles in these economies. For example, the maximum absolute unemployment gap is 5.0 percent in Canada, 3.8 percent in the United Kingdom, and 4.0 percent in the United States. All three of these estimates are excess supply gaps and, as can be seen in Figure 3, occurred in the early 1980s—about the time of the business cycle troughs that were related to the large disinflationary episode. More precisely, the peak excess supply gaps are dated as 1982:4 in Canada and the United States and 1983:2 in the United Kingdom. The estimates of u*'s at the end of the sample (1995:2) are 8.8 percent in Canada, 8.1 percent in the United Kingdom, and 6.1 percent in the United States. Comparing these estimates with the actual unemployment rates at the time of 9.5 percent, 8.3 percent, and 5.6 percent, respectively, suggests that there was still some disinflationary pressure in the Canadian economy, the potential for some inflationary pressure in the U.S. economy, and no pressure in the U.K. economy.9 Figure 4 demonstrates the fit of the model in terms of the ability of the estimated excess demand or supply term to explain historical movements in inflation. This figure was constructed by comparing the deviation of infla- LTE tion from expected inflation, nt - 8nt - (1 - 5)rt-1, with our measure of the effect of labor market tightness, y (u* - u)/u. The fit of the simple Phillips curve is remarkably good in all three of these countries, considering the highly parsimonious functional form.10 The figure illustrates the strong inflationary pressure about the time of the two oil shocks in the 1970s and the disinflation in 1981-82. It also illustrates the disinflation in Canada in the early 1990s.

Phillips Line with Model-Consistent u*'s

As noted above, the estimation results for the linear model with model- consistent u*'s are reported in Table 2. The problem with the linear model is evident in the final two columns of Table 2. The maximum absolute value of the gap takes on values that are excessively large in Canada and the United Kingdom. Furthermore, u* changes by as much as 7.7 percent and 3 percent in one quarter in Canada and the United Kingdom, respectively.

9 Although these estimates of u* may be useful for predicting short-term inflationary pressures, they cannot be used to predict future shocks for inflation. The uncertainty of any estimates of u* strengthens the case for cautionary policies that attempt to avoid serious overheating. 10 For example, in Gordon's latest (1994) estimate of the U.S. Phillips curve, he includes lags of up to five years on past inflation developments and a host of other variables to control for supply shocks.

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 265

Figure 4. Historical Performance of the Nonlinear Model (In percent) Deviation of inflation from expected inflation a ..... Effect of labor market tightness b 2.0

1974 1976 1978 1980 1982 1984 1986 1988 1990 1992

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 Source: OECD, Analytical Data Bank.

Clearly, such estimates are not economically sensible. This problem is illus- trated by Figure 5. The series for u* are excessively volatile in Canada and the United Kingdom, and volatile to a lesser extent in the United States. The estimates for u* for Canada, which range from about minus 5 percent to over 20 percent, are not plausible. The nonlinear and linear models are not nested; however, because each model has the same number of parameters, we can simply compare the values of the likelihood function in Tables 1 and 2 to determine which model fits the databest.1 1 This comparison shows that the linear model has a significantly better fit in all three of these countries. However, unlike the nonlinear model, the fit in the linear model is achieved by allowing extreme volatility in the u* series for each country, as shown in Figure 5. Indeed, the u* series generated by the linear model are considerably more volatile than the actual unemployment series. One advantage of the Kalman filter approach is that it is fairly straightforward to impose prior restrictions on volatility in u* series. Because the linear model produced such implausible estimates of u*'s for these countries, we estimated an alternative model in which we imposed some prior restrictions on large jumps in u*'s. In order to make the results comparable to the results for the nonlinear model, we reestimated the linear model subject to the constraint that the maximum absolute quarterly change in u* was equal to that obtained in the nonlinear model (that is, 0.21 percent, 0.65 percent, and 0:15 percent for Canada, the United Kingdom, and the United States, respectively). The results are reported in Table 3. When this restriction is imposed, there is a significant deterioration in the fit of the equation. Indeed, in this case a comparison of the values of the likelihood function indicates that the nonlinear model is the preferred model.

IV. Uncertainty and Recursive Estimates of u*

A fundamental problem that policymakers face in attempting to stabilize the business cycle is the considerable uncertainty in their inferences about current excess demand pressures. Measures of the natural rate of unemployment and, hence, labor market tightness are especially uncertain for countries that have undergone major structural reforms or are regularly subjected to significant supply shocks that require a continuous reallocation of labor resources across sectors. An important lesson from history is that it may be counterproductive to place too large a weight on stabilizing

11 The two models have the same number of estimated parameters, so the values of the likelihood functions are directly comparable.

Figure 5. Unemployment and the D-NAIRU with a Linear Phillips Curve (In percent)

D-NAIRU (u*) Unemployment rate (u)

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994

Source: OECD, Analytical Data Bank.

Table 3. Estimates of Phillips Lines with Restrictive-Model-Consistent D-NAIRUs (t-statistics in parentheses) Estimated equation: nt = ret + y (u* - u) + eIIt, where

K = year-on-year percent change in the CPI, 7et = inflation expectations, 5 KLTEt + (1 - 8)7t-1 (in percent), 7LTEt = long-term inflation expectation proxy from bond market (in percent), 8 = estimated weight on inflation proxy from bond market, 7 = parameter for measure of labor market tightness (in percent), u - unemployment rate (in percent), u* = D-NAIRU estimates derived from Kalman filter (in percent), LLF = value of the likelihood function, a2 = variance of residuals, a = average historical value of u - u* (in percent), Q. = maximum absolute value of u* - u (in percent), and T = maximum absolute quarterly change in u* (in percent).

Estimation period: 1971:Q2 to 1995:Q2. 2 Country Y 5 LLF a a T Canada 0.25 0.14 -104.81 0.41 _ 4.76 0.21 (4.35) (3.33) United Kingdom 0.46 0.22 -179.75 1.83 — 5.28 0.65 (3.56) (4.52) United States 0.47 0.08 -106.61 0.38 — 3.68 0.15 (7.00) (2.27) unemployment if there is considerable uncertainty about the level around which it should be stabilized, that is, the underlying natural rate. The difficulties invovled in measuring the natural rate of unemployment can contribute to significant policy errors. Indeed, one popular interpretation of overheating in the 1970s is that policymakers significantly underestimated the increases in the NAIRU in their countries. This type of policy error can have deleterious implications for the economy if the monetary authorities take considerable time to reestablish the credibility of their com- mitments to low inflation. In this sense, the asymmetric model predicts that the seeds of large contractions are sown when the monetary authority defers dealing with rising inflation and allows excess demand conditions to become entrenched in inflation expectations (Clark and Laxton, 1997). In order to truly assess the inflationary risks of overheating, it is important to develop measures of uncertainty so that confidence bands can be established around any particular point estimate. As mentioned earlier, an advantage of the Kalman filter is that it allows the calculation of real-time estimates of the model. Each period, the filter uses the new information to

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 269 revise its estimates of the model's parameters and the estimate of u*. This exercise replicates to an extent the process that a policymaker would undertake in using this framework to determine inflationary pressures. One can then assess with the advantage of hindsight whether the recursive estimates give a significantly different picture of the degree of inflationary pressure from the full-sample estimates, which incorporate more information. In the framework used in this paper, there are two sources of uncertainty: parameter uncertainty and uncertainty about the natural rate. The solid lines in Figures 6a, 7a, and 8a show the recursive or period-by-period estimates of u* and the model parameters for the nonlinear model. Figures 6b, 7b, and 8b display the same information for the linear model. The dotted lines in the upper panels of Figures 6-8 provide confidence bands of one standard error around our full-sample estimates of u* (which are shown by the bars). These standard error bands were obtained by imposing the full-sample parameter estimates for y and 5 for the whole sample, thus removing the effect of parameter instability, and then reestimating the model. This estimation produces an end-of-sample standard error on the estimate of u* that is applied for the whole sample.12 The parameter estimates are generally stable and approach their full- sample values relatively quickly. Each period, the filter assigns variation in the left-hand-side variable (in effect, the difference between inflation and inflation expectations) among variation in u*, variation in the parameters, and the error term. Thus there is likely to be a negative correlation between movements in the recursive estimates of u* and movements in the recursive parameter estimates. The recursive estimates of u* fluctuate generally within one standard error of the full-sample estimate. In particular, there is little difference in the United States between the recursive estimates of u* and the full-sample estimates. When the recursive estimate lies above the full sample estimate, the policymaker using this framework is overestimating the excess demand in the economy and thus may be running an overly restrictive policy. Such a situation can arise when a negative shock to inflation occurs, as this will tend to pull up the recursive estimate of u* until more observations can make clear whether the shock is temporary or permanent, and until the curvature of the Phillips curve can be more exactly estimated. Nevertheless, the standard error bands are wide in all three countries. Even in the United States, a confidence interval of approximately 66 percent (one standard error) for u* is 1 percentage point wide. This outcome is consistent

12 Technically, the standard error varies recursively over the sample. However, as the variation is not great, we have for expositional purposes used the final-period standard error.

Figure 6a. Recursive Estimates of the D-NAIRU and Parameters for Canada: Nonlinear Phillips Curve Model

— Recursive estimates E3 Full-sample estimates •• • • 'Standard error bands 12 12 D-NAIRU (u*; in percent) 10 10

8 8

6 6

4 4

0 1980 1982 1984 1986 1988 1990 1992 1994

3.6 3.6 Coefficient on measure of labor market tightnessa 3.0 3.0

2.4 2.4

1.8 1.8

1.2 1.2

0.6 0.6

0 0 1980 1982 1984 1986 1988 1990 1992 1994

0.25 0.25 Coefficient on expectation proxy from bond market 0.20 0.20

0.15 0.15

0.10 0.10

0.05 0.05

0 1980 1982 1984 1986 1988 1990 1992 1994 Source: OECD, Analytical Data Bank. a (u*t-ut)/ut.

Figure 6b. Recursive Estimates of the D-NAIRU and Parameters for Canada: Linear Phillips Curve Model

------Recursive estimates • Full-sample estimates — Standard error bands

1980 1982 1984 1986 1988 1990 1992 1994

0.40 0.40 Coefficient on measure of labor market tightnessa

1980 1982 1984 1986 1988 1990 1992 1994 0.25 0.25 Coefficient on expectation proxy from bond market

0.20 0.20

0.15 0.15

0.10 0.10

0.05 0.05

0 1980 1982 1984 1986 1988 1990 1992 1994

Source: OECD, Analytical Data Bank.

Figure 7a. Recursive Estimates of the D-NAIRU and Parameters for United Kingdom: Nonlinear Phillips Curve Model

• Recursive estimates • Full-sample estimates - - - Standard error bands 11 D-NAIRU (U*; in percent) 10 9 8 7 6 5 4 3 2 1 0 1980 1982 1984 1986 1988 1990 1992 1994

4.2 4.2 Coefficient on measure of labor market tightnessa 3.6 3.6

3.0 3.0

2.4 2.4

1.8 1.8

1.2 1.2

0.6 0.6

0 0 1980 1982 1984 1986 1988 1990 1992 1994

0.25 0.25 Coefficient on expectation proxy from bond market 0.20 0.20

0.15 0.15

0.10 0.10

0.05 0.05

0 0 1980 1982 1984 1986 1988 1990 1992 1994

Source: OECD, Analytical Data Bank.

a (u*t-ut)/ut.

Figure 7b. Recursive Estimates of the D-NAIRU and Parameters for United Kingdom: Linear Phillips Curve Model

— Recursive estimates • Full-sample estimates • • • Standard error bands 12 D-NAIRU (u*; in percent) 10

1984 1986 1988 1990 1992 1994 2.4 Coefficient on measure of labor market tightnesa 2.0

1.6

1.2

0.8

0.4

0 1980 1982 1984 1986 1988 1990 1992 1994

0.35 0.35 Coefficient on expectation proxy from bond market

1980 1982 1984 1986 1988 1990 1992 1994

Source: OECD, Analytical Data Bank. a (u*t-ut)/ut.

Figure 8a. Recursive Estimates of the D-NAIRU and Parameters for United States: Nonlinear Phillips Curve Model

— Recursive estimates gj Full-sample estimates .... Standard error bands

D-NAIRU (U*; in percent)

7 7 6 6 5 5 4 4 3 3 2 2 1 1 0 0 1980 1982 1984 1986 1988 1990 1992 1994 5.4 5.4 Coefficient on measure of labor market tightnessa 4.8 4.8 4.2 4.2 3.6 3.6 3.0 3.0 2.4 2.4 1.8 1.8 1.2 1.2 0.6 0.6 0 0 1980 1982 1984 1986 1988 1990 1992 1994 0.10 0.10 Coefficient on expectation proxy from bond market

0.05 0.05

-0.05 -0.05 1980 1982 1984 1986 1988 1990 1992 1994 Source: OECD, Analytical Data Bank. a (u*t-ut)/ut.

Figure 8b. Recursive Estimates of the D-NAIRU and Parameters for United States: Linear Phillips Curve Model

• Recursive estimates • Full-sample estimates —Standard error bands 12 12 D-NAIRU (u*; in percent) 10 10

8 8 - - \ - 6 6 trk'J • • T. * i 4 4

2 2

0 0 1980 1982 1984 1986 1988 1990 1992 1994 0.9 0.9 Coefficient on measure of labor market tightnessa 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 1980 1982 1984 1986 1988 1990 1992 1994 0.30 0.30 Coefficient on expectation proxy from bond market 0.25 0.25

0.20 0.20

0.15 0.15

0.10 0.10

0.05 0.05

0 0 1980 1982 1984 1986 1988 1990 1992 1994 Source: OECD, Analytical Data Bank.

©International Monetary Fund. Not for Redistribution 276 GUY DEBELLE and DOUGLAS LAXTON with the uncertainty surrounding other estimates of u*; the estimates of u* by Staiger, Stock, and Watson (1996), for example, are associated with a 66 percent confidence interval that is 1.3 percentage points wide.13 Turning to the linear model estimates, we see that the recursive estimates of u* are at least as variable as the volatile full-sample estimates. The 66 percent confidence interval for the Canadian u* is almost 8 percentage points wide. Furthermore, in Canada and the United States, the recursive estimates often lie outside the standard error bands of the full sample estimates, implying that a policymaker using the linear model is more likely to misread the extent of excess demand or supply in the economy. In the nonlinear model, the standard error bands also suggest that it is considerably easier to measure u* in the United States than it is in Canada or the United Kingdom. Indeed, perhaps one reason why policymakers in the United States have been more successful in avoiding boom-and-bust cycles is because they have had less difficulty in obtaining reliable measures of the natural rate of unemployment. That being said, the enormous uncertainty in these estimates reinforces the view that it may be desirable for policymakers to exercise caution by setting monetary conditions in a way that guards against the serious overheating that was allowed in the 1970s.

V. Information Content of Long-Term Inflation Expectations

As indicated above, our basic methodology in modeling inflation expectations ne involves using information from bond markets to develop proxies for long-term inflation expectations nLTE, and then testing to see whether these proxies are useful for identifying shifts in the Phillips curve. This approach to measuring inflation expectations may have some important advantages over reduced-form, distributed lag models if these measures of long-term inflation expectations embody information about how wage and price setters revise their expectations in response to shocks or changes in policy regimes. Because there is nothing really fundamental to tie down the distribution of future monetary policies—beyond the reputation of today's policymakers—it may take a considerable amount of time for agents to become convinced that governments are committed to low inflation. Furthermore, it may be entirely rational for market participants, when confronted with a new regime, to discount recent inflation performance along the transition

13 Staiger, Stock, and Watson report substantially larger bands that are based on a confidence level of 90 percent. We prefer to report estimates at the 66 percent level because in many situations policymakers simply cannot afford to be so certain when making decisions that could result in potentially larger policy errors.

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 277 path under the new regime and to weight heavily long moving averages of past inflation performance until it is evident that policymakers are committed to living with any adverse consequences of low inflation. Laxton, Ricketts, and Rose (1993) show that, where the monetary authorities are gaining credibility along the transition path, one should observe persistent excess supply gaps. This explanation is consistent with the results reported in Figure 4, which suggest that Canada and the United Kingdom have been operating mainly in the region of excess supply since the great disinflationary episode that started in the early 1980s. The three panels in Figure 9 provide plots of our measures of inflation 71, inflation expectations ne, and long-term inflation expectations nLTE. These panels are useful for identifying the role of the long-term inflation proxy in generating the statistical properties of the n - ne measure reported above. As can be seen in Figure 8, because of the larger estimated weight on the lagged inflation component (8 < .2 for all three countries), the quarterly variation in the short-term inflation expectations that enter the Phillips curve is motivated by information about recent inflation performance nt-1. However, the persistent deviation between n and ne over longer periods of time is influenced to a significant extent by the long-term inflation proxy. According to these estimates, market participants revise their expectations of long-term inflation very slowly in response to observed inflation performance. This interpretation of our results seems to reconcile the findings of our simple Phillips curves with other empirical work in the literature. First, as Gordon and many others have found, it takes fairly long lags on past inflation and a host of other "supply-shock" variables to save the Phillips curve. Obviously, in our highly simplistic model of the inflation process in these countries, the proxy for long-term inflation expectations fulfills this role. Indeed, our empirical results are consistent with some recent evidence that suggests that trends in long-term interest rates have long-term memory components. For example, Gagnon (1996) shows that the Fisher equation holds surprisingly well if long moving averages of past inflation are used to measure long-term inflation expectations. Because a traditional interpretation of the demand-determined view of the business cycle requires that measures of the business cycle, such as u* - u, must mimic measures of n - ne, the trick of reduced-form modelers has been to search distributed lag models in order to find specifications that fit both the story and the data. In Figure 4, we have shown that, if information from the bond market is used to identify inflation expectations, one can obtain a parsimonious nonlinear specification of the unemployment-inflation process in these countries without appealing to implausible estimates of u*. In other words, our measures of n - ne are consistent with both convexity in the Phillips curve and a fairly traditional interpretation of the business cycle,

Figure 9. Consumer Price Inflation and Inflation Expectations (In percent) - Consumer price inflation a • • • • Short-term inflation expectations • Long-term inflation expectations c

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 15, ,15

1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 Source: OECD, Analytical Data Bank.

©International Monetary Fund. Not for Redistribution IS THE PHILLIPS CURVE REALLY A CURVE? 279 namely, that it is driven principally by demand shocks, with moderate changes in the underlying u*. Given the problems associated with modeling inflation expectations with fixed-parameter, reduced-form models, we are somewhat surprised that more research has not focused on using information from the bond market to help identify shifts in the Phillips curve. As McCallum (1996) points out, without reliable survey measures of inflation expectations in most countries, information from long-term bond yields is probably the only objective indicator of policy credibility. One possibility why researchers have not utilized this information more in the past may be related to the problems with measuring the ex ante real interest rate. Although our procedure of developing proxies for the ex ante real interest rate is admittedly crude, it is interesting that many of our results are not overly sensitive to alternative assumptions. For example, one might argue that the benchmark for the end-of-sample estimate of 4.5 percent may be too high for the United States. When we use a lower estimate of the equilibrium real interest rate—3.5 percent—there is only a slight change in our estimate of y, and a rises from 0.33 percent to 0.48 percent in the United States.14 Similar results are obtained for the other two countries. As we mentioned above, several alternatives were employed in which we changed the assumed effect of government debt on the equilibrium world real interest rate. In fact, in one polar case we doubled the effect to be more consistent with some recent estimates that suggest that the effects have been large in some countries. In another case, we assumed that the world ex ante real interest rate has been unaffected by world government debt. In addition, we used alternative assumptions about the end-of- sample benchmark ranging from 3.5 percent to 5.5 percent. In all cases, it was impossible to reject the hypothesis that valuable information in the bond market will help identify historical shifts in the Phillips curve.

VI. Conclusions and Policy Implications

Previous tests for convexity in the Phillips curve have been biased because researchers have employed filtering techniques for u* that have been fundamentally inconsistent with the existence of convexity. A preferred statistical methodology would place both the linear and nonlinear models on an equal statistical footing by estimating model-consistent measures of u*. This paper proposes a simple method for estimating

14 These results of alternative assumptions for the ex ante log-term real interest rate can be obtained from the authors.

©International Monetary Fund. Not for Redistribution 280 GUY DEBELLE and DOUGLAS LAXTON model-consistent measures of u* that allows for either convexity or linearity in the unemployment-inflation process. After imposing plausible restrictions on variability in u*, we find that the nonlinear model fits the data best in Canada, the United Kingdom, and the United States. The implications of convexity for the macroeconomic policy debate are that policymakers that are unsuccessful in stabilizing the business cycle will induce a higher natural rate of unemployment in their economies. Uncertainty about the true level of u* reinforces the case for a cautionary strategy of raising interest rates before the economy reaches potential. Thus, rather than trying to fine-tune policy with a view to ensuring that all resources are fully employed at all points in time, it may be optimal in the context of uncertainty to develop a strategy that attempts to avoid large boom-and-bust cycles. Indeed, in order to avoid the necessity of generating large recessions to reign in inflationary forces and reestablish the credibility of the commitment to low inflation, it may be optimal for policymakers to provide a buffer zone to guard against the possibility of serious overheating. The width of the buffer zone could be positively related both to the degree of uncertainty about the natural rate of unemployment and the degree of asymmetry in the unemployment-inflation process. As noted above, in the presence of an asymmetry in the unemployment-inflation trade-off, such a buffer zone strategy could raise the average level of output and reduce the average level of unemployment over time if it is successful in avoiding boom-and-bust cycles.

REFERENCES

Blanchard, Olivier J., and Lawrence F. Katz, 1996, "The Natural Rate of Un- employment" (unpublished). Clark, Peter B., and Douglas M. Laxton, 1997, "Phillips Curves, Phillips Lines, and the Unemployment Costs of Overheating," IMF Working Paper 97/17 (Washington: International Monetary Fund). , and David Rose, 1996, "Asymmetry in the U.S. Output-Inflation Nexus," Staff Papers, International Monetary Fund, Vol. 43 (March), pp. 216-51. De Long, J. Bradford, and Lawrence H. Summers, 1988, "How Does Macro- economic Policy Affect Output?" Brookings Papers on Economic Activity: 2, Brookings Institution, pp. 433-80. Downes, Peter, and Andrew Johnson, 1994, "The Impact of a Lower NAIRU on the Australian Macroeconomy: Responses in the TRYM Model," paper presented at 1994 Australian Conference of Economists, Surfers' Paradise, Australia. Evans, Paul, 1985, "Do Large Deficits Produce High Interest Rates?" American Economic Review, Vol. 75 (March), pp. 68-87.

Fischer, Stanley, 1977, "Long-Term Contracts, Rational Expectations, and the Optimal Money Supply," Journal of Political Economy, Vol. 85 (February), pp. 191-205. Ford, Robert, and Douglas M. Laxton, 1995, "World Public Debt and Real Interest Rates," IMF Working Paper 95/30 (Washington: International Monetary Fund). Friedman, Milton, 1968, "The Role of Monetary Policy," American Economic Review, Vol. 58 (March), pp. 1-17. Fuhrer, Jeffrey C., 1995, "The Phillips Curve Is Alive and Well," New England Economic Review, Federal Reserve Bank of Boston (March/April), pp. 41-56. Gagnon, Joseph E., 1996, "Long Memory in Inflation Expectations: Evidence From International Financial Markets," International Finance Discussion Papers, No. 538 (Washington: Board of Governors of the Federal Reserve System). Goodfriend, Marvin, 1993, "Interest Rate Policy and the Inflation Scare Problem: 1979-1992," Economic Quarterly, Federal Reserve Bank of Richmond, Vol. 79 (Winter), pp. 1-24. Gordon, Robert J., 1970, "The Recent Acceleration of Inflation and Its Lessons for the Future," Brookings Papers on Economic Activity: 1, Brookings Institution, pp. 8-41. , 1975, "The Impact of Aggregate Demand on Prices," Brookings Papers on Economic Activity: 3, Brookings Institution, pp. 613-62. , 1977, "Can the Inflation of the 1970s Be Explained?" Brookings Papers on Economic Activity: 1, Brookings Institution, pp. 253-77. 1994, "Inflation and Unemployment: Where Is the NAIRU?" paper presented at Board of Governors of the Federal Reserve System Meeting of Academic Consultants, Washington, December. Harvey, Andrew C, 1981, Time Series Models (New York: Wiley). Helbling, Thomas, and Robert Wescott, 1995, "The Global Real Interest Rate," in Staff Studies for the World Economic Outlook, World Economic and Finan- cial Surveys, by the Research Department of the International Monetary Fund (Washington), pp. 28-51. Kuttner, Kenneth N., 1992, "Monetary Policy with Uncertain Estimates of Potential Output," Economic Perspectives, Federal Reserve Bank of Chicago (January/February), pp. 2-15. , 1994, "Estimating Potential Output as a Latent Variable," Journal of Business and Statistics, Vol. 12 (July), pp. 361-8. Laxton, Douglas M., Guy Meredith, and David Rose, 1994, "Asymmetric Effects of Economic Activity on Inflation: Evidence and Policy Implications," IMF Working Paper 94/139 (Washington: International Monetary Fund). Laxton, Douglas M., Nicholas Ricketts, and David Rose, 1993, "Uncertainty, Learning and Policy Credibility," in Economic Behavior and Policy Choice Under Price Stability (Ottawa: Bank of Canada). Laxton, Douglas M., David Rose, and Robert Tetlow, 1993a, "Problems in Identifying Non-Linear Phillips Curves: Some Further Consequences of Mismeasuring Potential Output," Bank of Canada Working Paper 93-6 (Ottawa: Bank of Canada).

, 1993b, "Monetary Policy, Uncertainty and the Presumption of Linearity," Bank of Canada Technical Reports, No. 63 (Ottawa: Bank of Canada). Lilien, David M., 1982, "Sectoral Shifts and Cyclical Unemployment," Journal of Political Economy, Vol. 90 (August), pp. 777-93. McCallum, Bennett T., 1996, "Inflation Targeting in Canada, New Zealand, Sweden, the United Kingdom, and in General," NBER Working Paper No. 5579 (Cambridge, Massachusetts: National Bureau of Economic Research). Macklem, Tiff, forthcoming, "Asymmetry in the Monetary Transmission Mecha- nism: What Can We Learn from VARs?" Bank of Canada Working Paper (Ottawa: Bank of Canada). Mankiw, N. Gregory, 1988, Comments on "How Does Macroeconomic Policy Affect Output?" by Bradford De Long and Lawrence Summers, Brookings Papers on Economic Activity: 2, Brookings Institution, pp. 481-85. Sargent, Thomas J., 1971, "A Note on the 'Accelerationist' Controversy," Journal of Money, Credit and Banking, Vol. 3 (August), pp. 721-25. Staiger, Douglas, James H. Stock, and Mark W. Watson, 1996, "How Precise Are Estimates of the Natural Rate of Unemployment?" NBER Working Paper No. 5477 (Cambridge, Massachusetts: National Bureau of Economic Research). Tanzi, Vito, and Domenico Fanizza, 1995, "Fiscal Deficit and Public Debt in Industrial Countries," IMF Working Paper 95/49 (Washington: International Monetary Fund). Tootell, Geoffrey M. B., 1994, "Restructuring, the NAIRU, and the Phillips Curve," New England Economic Review, Federal Reserve Bank of Boston (Septem- ber/October), pp. 31-44. Turner, Dave, 1995, "Speed Limit and Asymmetric Inflation Effects from the Output Gap in the Major Seven Economies," OECD Economic Studies No. 24 (1995/I), pp. 57-87.

Internal Migration, Center-State Grants, and Economic Growth in the States of India A Comment on Cashin and Sahay

M. GOVINDA RAO and KUNAL SEN *

EOCLASSICAL growth theory predicts that, in the absence of institutional Nbarriers to the mobility of capital and labor across national or regional boundaries and for economies with similar technology and preferences, per capita incomes tend to converge over a period of time during economic growth. The study by Cashin and Sahay (1996) is one of the few attempts to test this prediction at the regional level for a particular country. The study analyzes the growth trends in 20 states in the Indian Union over the period 1961-91 and concludes that (1) "there has indeed been convergence in real per capita incomes across the states of India during the period 1961-91" (Cashin and Sahay, 1996, pp. 163-64); (2) while the cross-sectional dispersion in real per capita incomes increased over the years, the center-state grants ensured that the dispersion in real per capita disposable incomes remained broadly constant over time; and (3) labor mobility was not an important factor contributing to convergence, particularly because of the existence of barriers. A close examination of the paper, however, brings out a number of conceptual and methodological issues, both in the tests for convergence and in the analysis of the equalizing impact of intergovernmental transfers. First, the convergence is observed only when the manufacturing output share variable is included in the regression. We argue that, for controlling sectoral

* M. Govinda Rao is a Research Fellow in the Australia South Asia Research Centre (ASARC), Research School of Pacific and Asian Studies, the Australian National University. Kunal Sen was a Postdoctoral Fellow in the ASARC, Research School of Pacific and Asian Studies, the Australian National University when this paper was written; he is now a Lecturer at Massey University. 283

©International Monetary Fund. Not for Redistribution 284 M. GOVINDA RAO and KUNAL SEN shocks, the inclusion of this variable is irrelevant. However, if the manufacturing output variable is taken as a policy variable, the results point toward conditional rather than absolute convergence. Second, the inclusion of four "special category" states makes the sample heterogeneous. Third, as net state domestic product includes the effect of federal grants, including them again to analyze the equalizing impact is conceptually erroneous. Finally, it is incorrect to exclude tax devolution from the analysis of equalization. Each of these points is discussed below.

I. Tests for Absolute Convergence

The absolute convergence hypothesis of the neoclassical growth model is that, for economies with the same steady state, poorer economies grow faster than richer ones. Cashin and Sahay test whether states in India exhibit absolute convergence by estimating a regression of the form

(1 / T) 1n(yit / yi,t-T) = C +1 n(yi,t-T)(1 - e-BT)/ T + uit. (1)

Absolute convergence occurs if (3 is significant and positive. The authors do not find evidence of absolute convergence with the regression as estimated in the form of equation (1) above. They then augment the basic regression with two additional explanatory variables: the share of agriculture and manufacturing in each state's net domestic product (NDP) in the initial year (AGRi,t-T and MANi,t-T, respectively). The inclusion of these variables is justified on the grounds that, as sectoral shocks may affect individual states or groups of states differentially, failure to introduce these additional variables may lead to biased estimates of (3, contingent on the realization of a particular shock. Using these variables, the authors now find evidence of a positive and significant (3 to confirm absolute convergence among the 20 Indian states. It is important to note that absolute convergence is found only when MANi,t-T is included in the basic regression along withAGRi,t-T .However , if the purpose of including these variables is to control for sectoral shocks, the inclusion of one of these variables renders the other superfluous in the regression model. In the Indian context, there is a strong correlation between the relative price of agriculture and that of manufacturing (the correlation coefficient between the two in the period 1961-91 is minus 0.94), so that a positive shock to the relative price of agriculture would imply a negative shock to the relative price of manufacturing (and vice versa). Therefore, the inclusion of AGRi,t-Twoul d be sufficient to control for the differential effect of sectoral shocks on states with differing sectoral

©International Monetary Fund. Not for Redistribution COMMENT ON CASHIN AND SAHAY 285 compositions. A similar argument can be made against the inclusion of MANi,t-T in the regression if its inclusion is meant to capture the sectoral shifts in employment from low-productivity agriculture to relatively high- productivity manufacturing; here again, the inclusion ofAGRi,t- Tcontrol s for the effect of changes in the industry mix on initial state per capita income. Therefore, MANi,t-T is an irrelevant variable: its inclusion should make no difference to the regression results. Yet when MANi,t-T is included in the regression, (3—the convergence coefficient—is found to be positive and significant. One important reason for this could be that differences in the degree of industrialization across states in India have largely been due to a policy bias in the planning process aimed at directing the regional spread of industrialization by, among other things, directing the flow of credits, controlling the location of public sector enterprises, and issuing industrial licenses. Because differences in policies across economic regions in the neoclassical growth model would be reflected in differences in steady states, the variable MANi,t-T may be proxying in part for steady state differences across the 20 Indian states. In this case, the significant and positive (3 found in the augmented regression when the share of manufacturing in state NDP is included may indicate conditional convergence rather than absolute convergence, contrary to what has been argued by Cashin and Sahay.

II. Inclusion of Special Category States in the Sample

Cashin and Sahay's pooling of the 20 Indian states in the regression analysis rests on the assumption that the states are homogeneous in technology and preferences, which is certainly not valid in the case of the four hill states of the north and northeast. These states are considered to be distinctly different in both official treatment and academic research; their critical distinguishing feature is the virtual absence of a production base. The only major determinant of net state domestic product in them is government expenditure, as may be seen from Table 1. Not surprisingly, the tax bases in these states are narrow, and much of the economic activity is triggered by central transfers. Therefore, these are considered "special category" states and are treated preferentially in dispensing the plan funds.1 It would have been more meaningful at least to have an exercise excluding these states and to compare the results.

1 The "Gadgil" formula used to distribute plan transfers does not apply to special category states. While for the general category states 30 percent of plan assistance is given as grants and 70 percent as loans, the grant-loan ratio is 90:10 for the special category states.

Table 1. Government Expenditure and Central Transfers in Selected Special Category States, 1991-92 (In percent)

Ratio of government expenditures to net state Ratio of central transfers to States domestic product government expenditures Himachal Pradesh 80.0 31.1 Jammu and Kashmir 77.6 59.0 Manipur 61.3 78.1 Tripura 65.9 76.4 Sources: Reserve Bank of India (1994); and Government of India (1995).

III. Equalizing Impact of Federal Grants: A Conceptual Issue

Section V of Cashin and Sahay's paper deals with the impact of intergovernmental grants on interstate disparities in real per capita incomes. This impact is measured by estimating the unweighted cross-sectional standard deviations in ln Yit (cNDP) for the states in each of the years and comparing them with the dispersion of state per capita disposable incomes (aSDI), defined as NDP plus central grants. An important conceptual issue in this analysis pertains to the definition of disposable income. NDP for the states in India is estimated using the income-originating approach, which includes the effect of federal grants. To add it again to get a measure of disposable income would amount to double counting. Take the simple identity

Yi = Ci + Ii, (2)

where Y, C, and I, respectively, denote income, consumption, and investment of the ith state. This can be shown as

Yi = C (p + g)i+ 1 (p + g)i, (3)

because Ci = C (p + g)i and Ii = I(p + g)i, where g and p represent government and nongovernment sectors. Assuming that taxes and expenditure spillovers cancel out, the state government budget will be

Ti + Fi + Di = Cgi + Igi + Gpi, (4)

©International Monetary Fund. Not for Redistribution COMMENT ON CASHIN AND SAHAY 287 where T, F, D, and Gp refer respectively to tax collections, federal transfers, borrowing, and transfer payments to the private sector, which is a part of Cpi. Substituting (4) in (3), we have

Yi = Cpi + Ipi + Ti + Fi + Di - Gpi. (5)

Thus, states' NDP estimates include the impact of federal grants, and the concept of estimating state disposable income by adding federal transfers to NDP is erroneous.2 Hence, the conclusion ought to be that despite the equalizing effect of federal transfers the dispersion in per capita NDP has widened over the years, and not that central grants ensured that the dispersion in real per capita disposable income was narrower.

IV. Noninclusion of Shared Taxes in the Analysis of Equalization

Another issue pertains to the exclusion of shared taxes in the analysis on the grounds that the transfers are given for the "purpose of reducing regional income disparities" (Cashin and Sahay, 1996, p. 128), and that the determination of the magnitude of the income-equalizing component lacks transparency. This raises two questions: What are the objectives of transfers? and What items should be included in the analysis of equalization? In the literature, the economic rationale for federal transfers is given in terms of (1) closing the fiscal gap or offsetting vertical fiscal imbalances arising from tax and expenditure assignments to different levels of government; and (2) providing horizontal equity transfers to enable the residents in every state to enjoy a given level of public service at a given tax price by offsetting revenue (fiscal capacity) and cost disabilities of the states (Boadway and Flatters, 1982). Thus, reducing interstate income disparities is not one of the objectives of federal transfers, although the transfers given to offset disabilities in taxable capacity would be equalizing. In the Indian context, it is conceptually incorrect to exclude tax devolution from the analysis of equalization because, like grants, shared taxes are used to offset fiscal disabilities. In fact, the Finance Commissions have varied the amount of tax devolution and grants to achieve the intended equalization. The methodology of the commissions consisted of filling the projected budgetary gaps of the states first through tax devolution (based mainly on population and backwardness), and then, for those states still left

2 In their study of the U.S. economy, Barro and Sala-i-Martin (1991) exclude federal transfers from the personal incomes of states to measure the equalizing impact of federal transfers.

©International Monetary Fund. Not for Redistribution 288 M. GOVINDA RAO and KUNAL SEN with gaps, through grants. Stung by the criticism that the commissions acted like "fiscal dentists filling in budgetary cavities," the Seventh Finance Commission doubled the states' share of excise duty from 20 percent to 40 percent to leave most of the states in surplus, obviating the need for grants. However, the increased tax devolution had to be targeted to poorer states by increasing the weight assigned to per capita NDP.3 The empirical analysis shows that the equalization impact of tax devolution has been greater than all other forms of transfers, particularly during the period covered by the awards of the Eighth (1984-89) and Ninth (1989-94) Finance Commissions (Rao, 1996). It must also be noted that the commissions explicitly stated the distribution formula, the weights assigned to different factors, and the relative shares of the states.4 Thus, it is conceptually unsound to exclude shared taxes from the analysis of equalization; it is also factually incorrect to assert that there is no transparency in the equalizing component of shared taxes and to exclude them from the analysis on that basis.

REFERENCES

Barro, Robert J., and Xavier Sala-i-Martin, 1991, "Convergence Across States and Regions," Brookings Papers on Economic Activity: 1, Brookings Institution, pp. 107-58. Boadway, Robin W., and Frank Flatters, 1982, Equalization in a Federal State: An Economic Analysis (Ottawa, Canada: Canadian Government Publishing House). Cashin, Paul, and Ratna Sahay, 1996, "Internal Migration, Center-State Grants, and Economic Growth in the States of India," Staff Papers, International Monetary Fund, Vol. 43 (March), pp. 123-71. Government of India, 1995, Indian Public Finances 1995 (New Delhi, India: Government of India Press). Rao, M. Govinda, 1996, "Intergovernmental Fiscal Relations in a Planned Economy—The Case of India," paper presented at the conference on "Fiscal Decentralization in Developing Countries," Montreal, Canada, September 19-20. , and Raja J. Chelliah, 1996, Survey of Research in Indian Fiscal Federal- ism (New Delhi, India: Indian Council of Social Science Research). Reserve Bank of India, 1994, Reserve Bank of India Bulletin, Vol. 48 (February).

3 Tax devolution was made progressive also by choosing the inverse [(Pi I Yi) / Xi Pi Yi ] and the distance [(Yh - Yi ) Pi / I(Yh - Yi )Pi ] formulas, where Y and Yh represent per capita net state domestic product of the ith and the highest per capita income state, respectively, and Pi is the population of the ith state. 4 For the details of the weights assigned to different factors in the distribution of income tax and union excise duties, see Rao and Chelliah (1996).

Internal Migration, Center-State Grants, and Economic Growth in the States of India A Reply to Rao and Sen

PAUL CASHIN and RATNA SAHAY*

N THEIR COMMENTS on our recent paper (Cashin and Sahay, 1996), Rao Iand Sen have concerns about several issues: our test of convergence; the sample of states included in our analysis; our measure of state disposable income; and our measure of center-state equalization payments. We respond below to each of these issues in turn. First, Rao and Sen are incorrect in asserting that the neoclassical growth model predicts absolute convergence as it, in fact, predicts conditional convergence. What drives the speed of convergence is the level of initial income for each economy relative to its own steady state level of income and steady state income growth rate. Absolute and conditional convergence are likely to be similar for regions of a given country (such as the states of India) that are reasonably homogeneous with respect to their steady state income levels and growth rates (that is, for regions that share similar preferences and technology). As to our examination of cross-state convergence, Rao and Sen seem to be unaware that the use of multiple sectoral variables to control for differential sectoral shocks (which may temporarily affect the growth performance of a state in a manner that is correlated with initial income) is common in the empirical growth literature (Barro and Sala-i-Martin, 1991, 1992, and 1995; and Sala-i-Martin, 1996a and 1996b). Notwithstanding this, Rao and Sen's comment that the share of manufacturing output may be regarded in the Indian context as a policy variable is an interesting one. If true, they are correct in asserting that this "may indicate conditional con-

* Paul Cashin is Assistant Professor at the University of Melbourne, Australia. Ratna Sahay is a Deputy Division Chief in the IMF's Research Department. 289

©International Monetary Fund. Not for Redistribution 290 PAUL CASHIN and RATNA SAHAY vergence rather than absolute convergence. . . ." However, if this assertion were true, then the absence of MAN from our growth regression, implying a failure to control for differing steady states, should reveal itself in a rejec- tion of the restriction of a common C in our equation (1) in favor of Ci. As pointed out below, we find that the restriction of a common C cannot be rejected, which implies that, as we argued in our original paper, MAN's role in our regressions is to control for differential sectoral shocks, not to control for differential steady states. As by definition our cross-sectional regressions impose a common constant term (implying common technology and preferences) for all states, our finding of (3 > 0 implies absolute convergence. Second, Rao and Sen are quite within their rights to argue on a priori grounds that the "special category" states of the north and northeast do not share a common steady state with other states and so should be excluded from any regional growth analysis. However, we tested this contention in our original paper (Cashin and Sahay, 1996, p. 149) and could not reject the null hypothesis of a common steady state for all regions of India. Indeed, we have in all likelihood underestimated the extent of cross-state convergence in India by including the low-income, slow-growing special category states and high-income, fast-growing Delhi in our data set, as we wanted to avoid biases that might arise from arbitrarily excluding regions from the analysis. Third, for most years of analysis, our state net domestic product figures are derived from the production side of the state accounts. As such, they refer to the value of production before intergovernmental transfers, which should be added to derive our measure of state disposable income (Gov- ernment of India, 1986). The exclusion by Barro and Sala-i-Martin (1991) of personal transfers derived from the federal government (as cited by Rao and Sen) was appropriate, given that, unlike us, Barro and Sala-i-Martin were using state personal disposable income in their convergence analysis. Our measure of state disposable income is, as stated in our original paper (Cashin and Sahay, 1996, p. 141), analogous to the national accounts concept of national disposable income (which represents the total income available to residents of a given country for consumption and saving). This concept is quite different from personal disposable income. Finally, the commentators fault us for not doing something that we explicitly stated we would not do, namely, include shared taxes in our definition of grants. We have excluded shared taxes, because, with the lack of meaningful data for the 1961-91 period for all states, we were unable to isolate the income-equalizing component of these transfers. Accordingly, our grants measure (which is based on data published in state budgets) could under- state the contribution of transfers from the center, as acknowledged in our

original paper (Cashin and Sahay, 1996, pp. 129 and 141). More important, we disagree with the commentators' view that all tax devolution from the center should be included in the concept of income-equalizing transfers. Since most taxes are raised by the center, the intended purpose of at least some transfers is merely to make available more resources to all states, independent of equity issues. However, we also readily accept (as we stated in the paper) that some component of these transfers could well be intended for equalizing incomes. In summary, Rao and Sen have confused the concepts of conditional and absolute convergence, and of state and personal disposable income. They also recommend the arbitrary exclusion of certain states from our sample set and criticize us for failing to include taxation elements of center-state equalization (which we expressly stated in our original paper that we would not include). In short, there is little in the comments of Rao and Sen that persuades us to alter the methodology and conclusions contained in our original paper.

REFERENCES

Barro, Robert J., and Xavier Sala-i-Martin, 1991, "Convergence Across States and Regions," Brookings Papers on Economic Activity: 1, Brookings Institution, pp. 107-58. , 1992, "Convergence," Journal of Political Economy, Vol. 100 (April), pp. 223-51. , 1995, Economic Growth (New York: McGraw-Hill). Cashin, Paul, and Ratna Sahay, 1996, "Internal Migration, Center-State Grants, and Economic Growth in the States of India," Staff Papers, International Monetary Fund, Vol. 43 (March), pp. 123-71. Government of India, 1986, Estimates of State Domestic Product, 1960-61 to 1983-84 (New Delhi, India: Government of India Press, Central Statistical Organisation, Department of Statistics, Ministry of Planning). Sala-i-Martin, Xavier, 1996a, "Regional Cohesion: Evidence and Theories of Regional Growth and Convergence," European Economic Review, Vol. 40 (June), pp. 1325-52. , 1996b, "The Classical Approach to Convergence Analysis," Economic Journal, Vol. 106 (My), pp. 1019-36.

IMF Working Papers

Staff Papers draws on IMF Working Papers, which are research studies by members of the Fund's staff. A list of Working Papers issued in 1997:1 follows.

"Fiscal Imbalances, Capital Inflows, and the Real Exchange Rate: The Case of Turkey," by Pierre-Richard Agenor, C. John McDermott, and E. Murat Ucer [97/1] "Labor Market Adjustment in Canada and the United States," by Eswar S. Prasad and Alun Thomas [97/2] "Do Labor Market Policies and Growth Fundamentals Matter for Income Inequal- ity in OECD Countries? Some Empirical Evidence," by Patrick Vanhoudt [97/3] "Improving India's Saving Performance," by Martin Muhleisen [97/4] "Adjusting to New Realities: MENA, the Uruguay Round, and the EU-Mediter- ranean Initiative," by Patricia Alonso-Gamo, Susan Fennell, and Khaled Sakr [97/5] "Long-Horizon Exchange Rate Predictability?" by Jeremy Berkowitz and Lorenzo Giorgianni [97/6] "The Exchange Rate in a Dynamic-Optimizing Current Account Model with Nom- inal Rigidities: A Quantitative Investigation," by Robert Kollman [97/7] "External Stability Under Alternative Nominal Exchange Rate Anchors: An Appli- cation to the GCC Countries," by Zubair Iqbal and S. Nuri Erbas [97/8] "Are Business Cycles Different in Asia and Latin America?" by Alexander W. Hoffmaister and Jorge E. Roldos [97/9] "From Generosity to Sustainability: The Austrian Pension System and Options for Its Reform," by Manfred Koch and Christian Thimann [97/10] "Towards a Market Economy: Structures of Governance," by Pierre Dhonte and IshanKapur [97/11] "The Rationale and Design of Inflation-Indexed Bonds," by Robert Price [97/12] "Patterns of Capital Flows to Emerging Markets: A Theoretical Perspective," by Zhaohui Chen and Mohsin S. Khan [97/13] "Financial Innovations Involving the Greek Drachma," by Michael Papaioannou and E. K.Gatzonas [97/14] "Stock Market Equilibrium and Macroeconomic Fundamentals," by Lamin Leigh [97/15] "Policy Implications of 'Second-Generation' Crisis Models," by Robert P. Flood and Nancy P. Marion [97/16]

292

"Phillips Curves, Phillips Lines, and the Unemployment Costs of Overheating," by Peter B. Clark and Douglas Laxton [97/17] "The Reform of the Pension System in Italy," by A. Javier Hamann [97/18] "Cyclical Effects of the Composition of Government Purchases," by Jahangir Aziz and Luc Leruth [97/19] "Fiscal Policy Management in an Open Capital Regime," by Peter S. Heller [97/20] "What Determines Real Exchange Rates? The Long and Short of It," by Ronald McDonald [97/21] "Broad Money Demand and Monetary Policy in Tunisia," by Volker Treichel [97/22] "Possible Effects of European Monetary Union on Switzerland: A Case Study of Policy Dilemmas Caused by Low Inflation and the Nominal Interest Rate Floor," by Douglas Laxton and Eswar Prasad [97/23] "Debt Relief for Low-Income Countries and the HIPC Initiative," by Anthony R. Boote and Kamau Thugge [97/24] "The Tax Treatment of Government Bonds," by John Norregaard [97/25] "Informational Efficiency, Interest Rate Variability, and Central Bank Operations," by Daniel C. Hardy [97/26] "Recent Export Credit Market Developments," by Paulo F. N. Drummond [97/27] "Market Information and Signaling in Central Bank Operations, or, How Often Should a Central Bank Intervene?" by Daniel C. Hardy [97/28] "International Currencies and Endogenous Enforcement—An Empirical Analysis," by Roohi Prem [97/29] "Designing a Tax Administration Reform Strategy: Experiences and Guidelines," by Carlos Silvani and Katherine Baer [97/30] "Aspects of Fiscal Performance in Some Transition Economies Under Fund- Supported Programs," by William H. Buiter [97/31] "Bank Credit in Argentina in the Aftermath of the Mexican Crisis: Supply or Demand Constrained?" by Luis Catao [97/32] "The Effect of Increasing Government Employment on Growth: Some Evidence from Africa," by James Gordon [97/33] "Monetary Policy and Leading Indicators of Inflation in Sweden," by Josef Baumgartner, Ramana Ramaswamy, and Goran Zettergren [97/34] "Inflation Targeting in Practice," by Guy Debelle [97/35] "Financial Liberalization and Money Demand in ASEAN Countries: Implications for Monetary Policy," by Robert Dekle and Mahmood Pradhan [97/36]

IMF Papers on Policy Analysis and Assessment

Papers on Policy Analysis and Assessment are intended to make staff work in the area of policy design available to a wide audience. A list of all PPAAs issued in 1997:1 follows. Papers may also be considered for inclusion in the journal.

"Designing Disinflation Programs in Transition Economies: The Implications of Relative Price Adjustment," by Sharmini Coorey, Mauro Mecagni, and Erik Offerdal [97/1] "Conditionally as an Instrument of Borrower Credibility," by Pierre Dhonte [97/2] "Are Europe's Social Security Finances Compatible with EMU?" by George Kopits [97/3]

An annual subscription to IMF Working Papers is available for US$210, and an annual subscription to Papers on Policy Analysis and Assessment, for US$80. Each subscription includes 12 monthly shipments and priority mail delivery. Individual Working Papers and Papers on Policy Analysis and Assessment are available on request for US$7 a copy. Prepayment is required. Please specify the series name, issue date, paper number, and complete title. American Express, MasterCard, and Visa credit cards are accepted, as well as checks in US dollars. Request should be made to

International Monetary Fund Publication Services 700 19th St., N.W. Washington, DC 20431, U.S.A.

Telephone: (202) 623-7430 Telefax: (202) 623-7201 E-mail: [email protected] Internet: http://www.imf.org

294

dots (. . .) indicate that data are not available;

a dash (—) indicates that the figure is zero or less than half the final digit shown, or that the item does not exist;

a single dot (.) indicates decimals;

a comma (,) separates thousands and millions;

"billion" means a thousand million, and "trillion" means a thousand billion;

a short dash (-) is used between years or months (for example, 1992-94 or January-October) to indicate a total of the years or months inclusive of the beginning and ending years or months;

a stroke (/) is used between years (for example, 1993/94) to indicate a fiscal year or a crop year;

a colon (:) is used between a year and the number indicating a quarter within that year (for example, 1994:1);

components of tables may not add to totals shown because of rounding.