HOW EFFECTIVE IS FISCAL POLICY IN RAISING NATIONAL SAVING?(Statistical Data Included)

Publication: Review of Economics and Statistics

Publication Date: 01-MAY-00

Bookmark this article Print this article Link to this article Email this article Digg It! Add to del.icio.us RSS

COPYRIGHT 2000 MIT Press Journals

I. Introduction

IN RECENT YEARS, the world has followed a generalized trend towards fiscal orthodoxy. In industrial countries, the steady rise in fiscal deficits since the 1980s (IMF, 1995) has been recently reversed both in the U.S. (with the balanced-budget agreement of 1997) and in the European union (under the fiscal constraints imposed by the Maastricht criteria). Developing countries have also joined the trend: Since the late 1980s, many countries traditionally suffering high or even hyperinflation have succeeded in radical stabilization based on fiscal orthodoxy (with Argentina perhaps as the clearest example), while most other developing countries have been able to strengthen their fiscal stance, with some of them showing consistent fiscal surpluses for a decade or more (Singapore and Chile, for example).

While fiscal adjustment is commonly regarded as the cornerstone of macroeconomic stabilization(1) the actual impact of lower public deficits on national saving and the current account balance remains both theoretically and empirically controversial. Yet this question is essential in assessing the ability of fiscal policy to contribute to the economy's external equilibrium and to reduce its vulnerability to external shocks, such as changes in terms of trade and capital inflows. Further, the relative effectiveness of alternative fiscal instruments (such as taxes, government consumption, and transfers) is still poorly understood.

From the analytical perspective, alternative consumption theories offer very different predictions about the effects of fiscal policy on national saving. While under the simple Keynesian hypothesis, only current levels of taxation matter for private consumption and saving decisions; under the permanent-income hypothesis, future taxes--over an infinite horizon--matter as well. The latter notion is taken to its logical limit by the Ricardian equivalence proposition which states that rational, forward-looking consumers react only to permanent government spending, implying that how government expenditure is financed--taxation, inflation, or debt--matters neither for private consumption nor for national saving (Barro, 1974). In turn, the lifecycle approach with finite lives offers a different prediction: If intergenerational transfers are not fully operative, fiscal policy affects national saving as long as it shifts income across different cohorts.

In reality, however, consumption reacts more strongly to current income (net of taxes or government spending) than these forward-looking theories would predict (for example, Deaton, 1992). One plausible explanation for this fact concerns the borrowing constraints faced by a significant share of consumers, who are forced to a corner solution to their intertemporal optimization problem. As a result, they react strongly to higher current disposable income (and, hence, to lower current taxes) by raising their consumption significantly--just as if they were Keynesian or myopic.

Distortionary taxation gives rise to further effects of fiscal policy on consumption and saving. Changes in the time profile of distorting taxes can affect the consumption plans of forward-looking agents, even when the present value of taxes is left unchanged. Finally, private consumers may also react to public consumption because their utility--not their budget constraints--is directly affected (Bailey, 1971). If public consumption is perceived as a substitute (complement) for private consumption--as in the case of public education, for example--then an increase in the former will be accompanied by a decline (increase) in the latter.

In turn, the empirical literature has devoted considerable attention to the effects of fiscal policy on private consumption or saving during the last decade. Most studies belong to either of two types. The first category is based on the empirical implementation of private consumption functions derived from first principles (and therefore including a limited set of regressors) that are used to test for specific departures from Ricardian equivalence and/or for substitutability/ complementarity between private and public consumption. The second group of studies is based on general reduced-form equations for private consumption or saving (including a broad number of regressors) that used to measure directly the effect of fiscal variables, providing estimates of the offset coefficients for private-public saving. Estimations are typically based on time-series macroeconomic data for one or several industrial and developing countries.

What does this empirical literature find? As we document in more detail below, the results are very mixed. A majority of studies rejects strict Ricardian equivalence, but this rejection is far from unanimous. Estimated offset coefficients for public-private saving are typically significantly different from both 1 and 0, but the range of empirical results is still very broad. In turn, estimates of public-private consumption substitution range from positive to negative in different countries and across different studies. Behind these large discrepancies in parameter estimates and policy inferences are large differences in specification, sample coverage and quality, and estimation techniques.

The purpose of this paper is to offer a fresh look at the; world evidence on the saving impact of fiscal policy. In doing this, we extend the preceding literature in four dimensions. First, we develop a consumption model that aggregates over two representative individuals and encompasses three consumption hypotheses. Second, we estimate: the model on a large macroeconomic panel data set (derived from the World Bank Saving Project database and including 19 industrial and 22 developing countries) that offers significant advantages over previous studies in terms of data quality and coverage. Third, in contrast with all preceding work which is based on individual country time-series estimations, we make use of panel estimation methods that allow us to correct for simultaneity and heterogeneity by using instrumental-variable procedures based on "internal" instruments. Fourth, our large data set allows us to estimate the model separately on the samples for industrial and developing countries, thus making it possible to identify any significant differences between them. The paper is organized as follows. The next section reviews briefly the previous empirical literature. Section III presents the detailed derivation of the model. Next, in section IV, we discuss the data and econometric techniques. The estimation results are reported in section V. Finally, section VI concludes.

II. Overview of Previous Results

As stated above, the recent empirical literature on the effects of fiscal policy on private consumption and saving falls into one of two categories: narrow models derived from first principles or broad reduced-form specifications. We first summarize briefly the former group of studies that are based on private consumption functions derived from intertemporal optimization and whose empirical implementation typically includes a limited number of regressors. These studies are aimed at testing for specific deviations of Ricardian equivalence and/or for substitutability/complementarity between private and public consumption.

Stringent assumptions are required for Ricardian equivalence to hold. (For a detailed review, see Seater (1993).) Among the most important are the following:

(i) full intergenerational caring (equivalent to assuming that the public and private sectors share the same discount rate or that the private sector has an infinite horizon),

(ii) perfect capital markets (no borrowing constraints),

(iii) far-sighted, rational consumers who internalize the government's intertemporal budget constraint,

(iv) the absence of uncertainty (unless complete contingent/insurance markets exist that render ineffective any market-completing efforts of fiscal policy), and

(v) nondistortionary taxes, subsidies, and transfers.

Empirical consumption studies testing Ricardian equivalence have concentrated mostly on the first and second assumptions.

To test the assumption of infinite horizons, most empirical studies follow Blanchard's (1985) model of overlapping generations with finite horizons, reflected by a survival probability ([Gamma]) smaller than 1. Some studies test separately for the existence of heterogeneous consumers by introducing a parameter ([Lambda]) measuring the proportion of consumers who face binding borrowing constraints and therefore find themselves at a corner solution of their desired intertemporal consumption plan.

The Ricardian equivalence proposition has spawned a large industry of empirical studies--most on the U.S., some on industrial countries, and a few on developing countries--and a related cottage industry that surveys them. Among the latter surveys, Bernheim (1987), Leiderman and Blejer (1988), and Elmendorf and Mankiw (1998) tend to conclude that the empirical evidence rejects Ricardian equivalence. However "a small but prominent minority of economists, including Robert Barro, have argued that Ricardian equivalence does in fact describe the world, at least as a first approximation" (Elmendorf and Mankiw, 1998, p. 43), which is a view also reflected in Seater's (1993) survey.

Because in this paper our interest is in multicountry studies, in table 1 we present a brief summary of empirical results that are limited to a selected number of studies that include a sizable number of countries. As the table shows, the empirical literature almost invariably fails to reject the assumption of infinite horizons. Only in Khalid (1996)--and only in two out of 21 countries covered in that study--the annual survival probability [Gamma] is significantly lower than 1. Regarding borrowing constraints (as measured by [Lambda]), the opposite result obtains; this type of capital market imperfection is a frequent cause of rejection of Ricardian equivalence in empirical studies. However, the different studies are not unanimous regarding the extent of this rejection. They range from rejection for each of the sixteen countries studied by Haque and Montiel (1989) and five out of the six countries studied in Campbell and Mankiw (1991), to rejection in only eight out of 21 countries in Khalid (1996). In those countries for which evidence is found of borrowing constraints, the estimated share of constrained consumers varies widely (from 18% to 100%).(2)

TABLE 1.--PREVIOUS RESULTS ON RICARDIAN EQUIVALENCE AND PRIVATE/PUBLIC CONSUMPTION SUBSTITUTION

Study Sample Haque (1988) 16 Develop. Haque and Montiel (1989) 16 Develop. 1960-1985 Campbell and Mankiw (1991) 6 Indust. Corbo and Schmidt-Hebbel (1991) 13 Develop. Karras (1994) 17 Indust. 13 Develop. 1950-1980s Evans and Karras (1996) 33 lndust. 32 Develop. 1950-1990 Khalid...

Read the full article for free courtesy of your local library.