VI Private Saving in IMF Arrangements
Author: Mr. Miguel A Savastano
Language:
English
Keywords:
OP; wage control; saving; wage; real interest rate; saving rate; saving objective; public saving; saving data; capture development; saving equation; Private savings; Real interest rates; Income; Terms of trade; Inflation; Middle East; national saving
Download PDF (4 MB)

Abstract

Raising national saving is a critical goal of many IMF-supported programs. Higher levels of national saving provide more resources for domestic investment—resources that are needed when increased reliance on foreign saving is either unlikely or undesirable. Yet raising national saving is often one of the most elusive goals of IMF-supported programs. The central problem is that, while ways to influence public saving are clear, public saving generally accounts for a small share—on average about 15 percent in the countries under review—of total national saving.1 Thus, sizable increases in total saving frequently must emanate from the private sector. However, because both the theoretical underpinnings and the empirical evidence on the effectiveness of policies to affect private saving are ambiguous, it has been difficult to identify and predict the effects of measures to meet this goal.

Raising national saving is a critical goal of many IMF-supported programs. Higher levels of national saving provide more resources for domestic investment—resources that are needed when increased reliance on foreign saving is either unlikely or undesirable. Yet raising national saving is often one of the most elusive goals of IMF-supported programs. The central problem is that, while ways to influence public saving are clear, public saving generally accounts for a small share—on average about 15 percent in the countries under review—of total national saving.1 Thus, sizable increases in total saving frequently must emanate from the private sector. However, because both the theoretical underpinnings and the empirical evidence on the effectiveness of policies to affect private saving are ambiguous, it has been difficult to identify and predict the effects of measures to meet this goal.

This problem is put in sharp relief by the record on saving during the arrangements under review. On average, programs for which private saving data are available aimed at increases in national saving relative to GDP of almost 1 percentage point a year—equivalent to some 6 percent of initial savings: on average, public saving was targeted to rise by over 75 percent more than the increase in total saving, while private saving was projected to fall during the programs.

The purpose of this study is to examine the experience with private saving during stand-by and extended arrangements. Because data on private saving are not available for all the 36 countries covered in other parts of this review, this study looks at the experiences of 22 countries only. The objectives are to provide a brief discussion of the analytical factors that influence short-run changes in private saving, to record the objectives and outcomes for private saving in IMF-supported programs, and to search for any lessons on how private saving is likely to respond to adjustment and reform policies.

What Do We Know from the Literature?

A vast literature has examined, on theoretical and empirical grounds, the behavior and determinants of private saving in both industrial and developing countries.2 Much of this literature, founded on the life-cycle and permanent income theories of consumption, views the role of private saving as that of smoothing out private consumption over time on the basis of anticipated lifetime income. In this function, private saving is seen as being influenced by five types of determinants: current and expected income and wealth; public saving; interest rates and inflation; external factors (terms of trade and foreign saving); and demographic factors. In spite of the common framework, however, the literature has been plagued by conflicting hypotheses regarding the role played by some of these determinants. Thus, depending on the specification chosen for any given transmission mechanism and on the assumption regarding the permanence of the change in the determinant in question, studies have shown on theoretical grounds that private saving can be either positively or negatively related to the real interest rate, the rate of growth of income, the rate of inflation, financial savings (proxied by the ratio of broad money to output), and the external terms of trade. It has also been shown that changes in public saving can elicit either fully or partially offsetting changes in private saving, depending, among other factors, on whether the change in public saving is perceived as permanent or temporary or reflects changes in tax revenues or in current expenditures.3 The effect of the real exchange rate on private saving has not received much attention in the literature, although, analytically, real depreciations, by shifting income from relatively low-saving wage earners to higher-saving profit earners and through wealth effects, have been linked to increases in saving.4 Finally, studies have established that the effects of fully funded private pension plans on private and national saving hinge crucially on the extent to which individuals regard their (voluntary or mandatory) contributions as perfect substitutes for other forms of saving.

Abundant empirical evidence has not settled all these controversies. Perhaps the most robust findings are that the level of private saving in real terms is positively related to a country’s current level of real income, inversely related to the level of real private wealth and to the share of young and elderly in the total population (the dependency ratio), and subject to a measurement bias that is directly proportional to the rate of inflation.5 For almost all the other determinants of private saving, the empirical results are ambiguous. The weight of the evidence, however, points to a few empirical regularities. First, private saving tends to be positively related to the rate of output growth and to the external terms of trade. Second, changes in public saving are likely to elicit opposite but less-than-equal changes in private saving, the size of the offset being a direct function of the perceived permanence of the change in the fiscal stance.6 Third, changes in real interest rates are likely to have strong effects on financial saving but not on total private saving. Fourth, private saving rates are likely to be inversely related to foreign saving rates when capital mobility and access to foreign borrowing are limited. Fifth, private saving is likely to increase by only a fraction of any new contribution to private pension plans. Little empirical work has been done to test the link between changes in the real exchange rate and private saving.

The above findings appear recurrently in studies that examine the behavior of private saving in both industrial and developing countries, although the results for the latter tend to be less robust. This is in part a consequence of the notorious weakness of the data on saving and some of its determinants in developing countries, but also of several distinctive conditions in those countries that tend to impair the private sector’s ability to use savings for consumption-smoothing purposes: household incomes that are low or even close to subsistence levels; relatively uncertain and variable lifetime income streams that make it difficult to set a long-term path for private consumption; strong intergenerational links; and limited access of domestic residents to well-organized financial markets.

There is a consensus in the literature that the more binding the borrowing constraints faced by firms and individuals, the weaker the response of private saving to changes in variables that ought to guide consumption and saving decisions over time.7 A corollary is that an easing of restrictions on domestic and foreign borrowing in economies where the scope for intertemporal substitution is severely constrained is likely to result, on impact, in a fall in private saving. Two crucial policy implications follow. First, the effectiveness of policies aimed at raising total national saving—for example, reducing the fiscal imbalance—depends on the severity of borrowing constraints and therefore is likely to vary across countries and over time. Second, private saving is unlikely to react favorably in the short run to structural reforms that succeed in lessening the country’s borrowing constraints either directly, such as liberalization of the domestic financial system or the capital account, or indirectly, such as trade reform, privatization, or other policies that increase the economy’s attractiveness to foreign investors.

How Is Private Saving Measured?

The measurement of saving in developing countries suffers from serious shortcomings.8 Because independent measures of saving—for example, household and enterprise surveys—are either nonexistent or unreliable, gross national saving is commonly calculated as the difference between gross domestic investment and an estimate of foreign saving derived from balance of payments data. Thus, conventional measures of national saving are bound to be affected by any deficiency or inconsistency in the figures for investment and foreign saving: those most frequently noted are the classification of purchases of consumer durables as consumption rather than investment, the inadequate computation and valuation of inventories, the distortions stemming from overvalued exchange rates, and the problems in recording private capital flows—particularly capital flight. The net effect of these measurement problems on saving estimates is uncertain and country-specific, although it is commonly argued that they tend to impart a downward bias to national saving.

Measures of private saving are, de facto, a residual among residuals, as they are usually computed as the difference between gross national saving and an estimate of public saving derived from public finance data. In principle, this procedure is likely to add two more biases to the resulting measures of saving: one, already mentioned, stems from inflation-induced distortions to public sector accounts; the other arises from well-known deficiencies in recording fully all public sector operations in developing countries. When significant, these problems tend to bias upward the measure of private saving.

These shortcomings tend to weaken the link between measured national and private saving and their theoretical counterparts. They, therefore, need to be kept in mind when appraising any evidence on the behavior of saving in developing countries. As noted before, in this study the sample was restricted to the 22 countries under review for which estimates of the breakdown of total saving into public and private components were available. This criterion, however, did not ensure full compatibility in the computation and coverage of saving rates across countries. Thus, in three countries (Cameroon, Côte d’Ivoire, and Romania) the breakdown of public and private saving is derived from domestic rather than national saving; in two countries (Brazil and Guyana) saving rates are expressed in constant rather than current prices; in two countries (Brazil and Mexico) the measures of private and public saving are adjusted for the inflation component of interest payments on domestic debt; and in one country (the Philippines) the saving rate is expressed as a ratio to GNP rather than GDP (Appendix Table 6-A1). Moreover, in only 12 of the 22 countries does the measure of public saving capture developments in the public sector broadly defined, rather than solely in the central or general government.9 These methodological differences complicate even more the tasks of understanding and drawing firm policy conclusions on saving behavior. Partly because of these problems, results presented here, although indicative of certain common trends in private saving during programs, are tentative and require further investigation, ideally with a larger and more homogeneous data base.

Private Saving in Recent Arrangements

Several salient features emerge from reviewing developments in private, public, and national saving during the arrangements in the 22 countries for which data are available.10

First, on average and regardless of the length of the arrangement, national and public saving rates rose while private saving rates fell when IMF-supported programs were in place. Looking at saving by program years, public saving rates rose on average by 0.8 percentage point of GDP a year while private saving rates fell on average by 0.5 percentage point; this resulted in a moderate increase, on average, in national saving rates (Table 6-1, top). Over the whole length of IMF involvement in the period under review, public saving rates rose cumulatively on average by 1.7 percentage points while private saving rates fell by 1 percentage point; national saving rates, therefore, rose on average by 0.7 percentage point (Table 6-1, bottom). The average recorded fall in private saving is somewhat larger when the sample excludes Poland and Romania, countries where saving estimates are particularly unreliable.11 In that case, the average decline in the rate of private saving reaches 0.7 percentage point annually and 1.5 percentage points during the length of IMF involvement. A similar pattern emerges when the averages control for the differences in coverage of the public saving rate in the countries of the sample.

Second, declines in private saving rates were slightly more frequent but considerably larger than rises in private saving. Excluding Poland and Romania, private saving rates fell relative to GDP in 26 of the 50 program years by an average of about 3 percentage points; in four countries (Côte d’Ivoire, Honduras, Jamaica, and Venezuela) the one-year decline in private saving reached or exceeded 6 percentage points, while in no country did the annual increase in private saving reach that level (Chart 6-1). Over longer periods of IMF involvement, cumulative falls in private saving also occurred in over half of the countries and exceeded the cumulative rises in private saving.

Table 6-1.

Changes in Saving Rates in Recent IMF Arrangements

(In percentage points of GDP unless otherwise indicated)

article image
Sources: Appendix Tables 6-A1, 6-A2, 6-A3, and IMF staff estimates.

Public saving rate includes operating results of public enterprises.

Third, as suggested by both theory and other empirical studies, within each country, changes in private saving rates tended to offset partially changes in public saving. Of the 50 program years, excluding Poland and Romania, the private saving rate fell in 20 of the 34 years when public saving increased, and rose in 10 of the 16 years when public saving fell. Looking at the cumulative change during IMF involvement in this period, private saving rates fell in 12 of the 16 countries where public saving increased, and rose in 6 of the 8 countries where public saving fell (Chart 6-2). This inverse relation between changes in private and public saving was robust to the differences in coverage of the public saving rate and is likely to reflect several factors, including the transmission mechanisms of fiscal policy discussed above. (The results were also little affected when the program year and cumulative changes were calculated taking as the base period the average of the saving rates prevailing in the two years prior to each annual arrangement—a procedure that lessened the influence of transitory factors on saving rates in the immediate pre-program year.)

To what extent were these developments in saving anticipated in the design of programs? Projections for private saving are often not reported in IMF staff documents, and therefore it is not possible to compare programmed with actual saving developments in all the programs reviewed.12 In the 15 countries (32 program years) for which projections are available, saving objectives were fairly ambitious. On average, national saving rates were expected to increase by 0.8 percentage point of GDP in each program year (Table 6-2). This improvement was expected to stem mostly from a strong performance of public saving, as private saving rates were projected to fall on average by 0.6 percentage point.

Chart 6-1.
Chart 6-1.

Changes in Saving Rates by Program Years

(In percentage points of GDP)

Source: Appendix Table 6-A2.

In the event, changes in private saving proved quite difficult to predict, and deviations from program projections on private saving account for most of the deviations from projected national saving. In fact, program years for which changes in private saving were overprojected were equal in number to those for which changes in private saving were underprojected; in both groups, the difference between projected and actual changes in private saving rates averaged about 2 percentage points, and in seven program years it exceeded 3 percentage points, far larger than the deviations from program targets of public saving. Moreover, private saving rates moved in the direction envisaged in the program in only 20 of the 32 program years (Appendix Table 6-A4). Programs also envisaged that private and public saving rates would change in opposite directions in about half of the program years and that, on average, the drop in private saving would equal between a third and a half of the targeted rise in public saving. However, a partial offset actually took place in 20 of the 32 program years, and, when fiscal adjustment was strong, the offset was much larger than projected.

Chart 6-2.
Chart 6-2.

Cumulative Changes in Saving Rates During IMF Involvement

(In percentage points of GDP)

Source: Appendix Table 6-A3.
Table 6-2.

Saving Rates: Projected and Actual Changes by Program Year

(Changes in saving rates: in percentage points of GDP)

article image
Source: Appendix Table 6-A4.

What Explains Changes in Private Saving?

Changes in private saving are difficult to predict and explain because they reflect the interaction of many determinants. A simple regression analysis, therefore, can help sort out the relative role of major quantifiable determinants. The regressions examine the relationship between program year changes in private saving rates and the explanatory variables identified above for which data are available.13 Specifying the dependent variable as changes in private saving rates differs markedly from the approach in most cross-country studies on saving and, together with the short time series available for each country, imposes several restrictions on the estimation. For instance, using changes in private saving rates by program year ruled out the common practices of using as regressors the average of several annual observations of the explanatory variables or of pooling together all time-series data available for each country to estimate a single saving equation, because those procedures would have prevented the isolation of program-specific effects. Using changes in saving rates also meant that demographic variables or per capita income figures, which usually account for differences in the levels of private saving rates across countries, did not need to be included as regressors. Moreover, it was not obvious a priori whether the explanatory variables should be expressed in levels or first differences to conform with the dependent variable. To address the latter problem, the equation was estimated in three different functional forms: the first expressed all explanatory variables in their original units—percentage changes and ratios to GDP; the second expressed all explanatory variables as first differences; and the third—a hybrid—used as regressors the levels of all the variables originally expressed in percentage terms and the first differences of all those expressed in ratios to GDP. In each, the estimation procedure started with an equation that included all explanatory variables as regressors; variable-deletion (backward-elimination) tests were then used sequentially to discard regressors that were not statistically significant, and thereby to select the subset of variables that best explain the changes in private saving.

Table 6-3 presents the results for the functional form that expresses all regressors in first differences—except for output growth, which is also expressed in levels.14 When all explanatory variables are included in the equation, most of the estimated coefficients have the expected sign but not all are statistically significant. The backward-elimination procedure selected as the most important regressors changes in public saving, the growth of real GDP (in levels), changes in foreign saving, and changes in the ratio of broad money to GDP; together, these four variables account for about half of the variation in private saving rates.

Table 6-3.

Changes in Private Saving: Cross-Program Regressions

(Dependent variable: program year changes in private saving/GDP; Estimation method: ordinary least squares)

article image
Source: IMF staff estimates.

All explanatory variables expressed in first differences, plus GDP growth expressed in levels.

All t-ratios in parentheses; an asterisk indicates that the null hypothesis of a zero coefficient is rejected at the 10 percent significance level.

All regressions exclude the observations for Poland (1991) and Romania (1991).

F-test of joint zero restrictions on the coefficients of the (five) excluded variables; an asterisk indicates that the null hypothesis of zero coefficients is not rejected at the 10 percent significance level.

F-test on linear restriction on the coefficients of revenues and expenditures; an asterisk indicates that the null hypothesis of equality of coefficients (in absolute value) is not rejected at the 10 percent significance level.

The equation selected suggests that changes in private saving were inversely related to changes in public and foreign saving rates and directly related to output growth and (less strongly) to changes in financial saving. Changes in the terms of trade appear to have had a direct but weak independent effect on private saving. Under certain assumptions, the selected equation may resemble the saving-investment balance identity. This resemblance, however, does not invalidate the behavioral content of the equation for at least two reasons: first, the variables selected have a similar effect on private saving changes when the regression controls for other quantifiable determinants of saving; and second, the assumptions needed to transform the selected equation into an identity (for example, a stable relationship between output and investment) are theoretically and empirically disputable and, thus, would need to be tested independently.

The inverse relationship between changes in private and public saving rates stands out as perhaps the most robust finding of the regression analysis. The estimates reported indicate that, on average, about two thirds of any variation in the change of the public saving rate was offset by a change in the private saving rate (similar point estimates for the offset coefficient were obtained in all regressions). The response of private saving appears to have been equally strong regardless of whether the change in public saving was driven by changes in total revenues or in current expenditures. When the two are included separately in the selected equation, the coefficient for changes in revenues is higher (in absolute value) than that for changes in current expenditures, but an F-test suggests that the coefficients are not significantly different. The response of private saving also appears robust to the differences in coverage of the public saving rate within the sample. Binary variables accounting for this difference—interacting with either the intercept or the coefficient of the public saving rate—were added to the regressions, but their estimated coefficients are not statistically significant.

These results highlight the difficulty of designing policies to raise private saving. The effects on national saving of fiscal adjustments that raise public saving—either through higher revenues or lower current expenditures—tend to be partially offset by a decline in private saving rates, while other factors that might counter this effect—changes in interest rates, the real exchange rate, and inflation—appear to have negligible influences on the short-run behavior of private saving.

Two factors not incorporated explicitly in the regressions are the constraints on domestic and foreign borrowing and the effects that structural reforms might have had on easing these constraints. Implicitly, however, the estimates suggest that borrowing constraints may have been important. In particular, the significant response of private saving rates to current output growth and changes in foreign saving, the less-than-full offsetting of changes in public saving, and the weak influence of real interest rates are commonly interpreted by other empirical studies as indirect evidence of the presence of binding domestic and foreign borrowing constraints.15 In any event, the regressions still leave unexplained a considerable fraction of the changes in private saving rates during IMF programs. A better sense of the factors that account for this unexplained variation must, therefore, come from a detailed assessment of developments in countries that experienced large unexplained changes in private saving.

Private Saving in Mexico and Tunisia

Mexico and Tunisia present a number of contrasts for an assessment of nonquantifiable as well as quantifiable determinants of private saving. In both countries, four-year extended arrangements had a sizable structural component, private saving behaved quite differently from what was envisaged in the program, and the regression estimates underpredict the changes in private saving rates—by about 60 percent in Mexico and 50 percent in Tunisia. Yet private saving rates fell sharply in Mexico and rose in Tunisia during the program period. The aim here is to address these differences by focusing on three main questions: (i) what was the behavior of private saving before each arrangement and to what extent did it influence saving developments in the program? (ii) how was private saving expected to behave during the arrangement and why? and (iii) what factors seem to account for the actual behavior of private saving and how do they differ from what was envisaged in the program and from what was captured by the cross-program regressions?

The main message that emerges is that the inverse relationship between public and private saving rates identified in the cross-program analyses is probably part of a wider pattern of adverse short-term responses of private saving to adjustment and reform programs. The size and duration of this response seem to depend on the country’s initial conditions and on the degree to which structural reforms ease the constraints on domestic and foreign borrowing faced by domestic residents and succeed in attracting large inflows of foreign capital; on these counts, Mexico’s reforms went further than those of Tunisia and its financial situation improved from a much worse starting position. This response of private saving to reform programs represents a vexing damper on the external adjustment process that has been observed in other episodes.16 It seems likely, however, that this is a relatively short-term phenomenon reflecting the private sector’s attempt to increase spending when permanent income is expected to rise, consumption opportunities are expanded, and access to financial markets improves. Following a transition period, the benefits from securing and preserving a stable macroeconomic environment and removing obstacles to growth through continued structural reforms, including the creation of funded private pension plans and the removal of labor market distortions, are likely to lead to a phase of sustained growth and rising saving rates. These propositions for the medium term cannot be examined with the short period of data available in this study, although they are strongly supported in long-term studies of saving behavior.17

Mexico: Falling Private Saving Amid Large Disinflation and Decisive Reform

From 1970 to 1981, the rate of national saving in Mexico hovered around 20 percent of GDP (Chart 6-3). With the debt crisis of 1982, national saving fell to less than 18 percent of GDP, but rebounded to an average of around 23 percent of GDP during 1983–85 only to fall again after the sharp drop in oil prices of 1986. This pattern was strongly influenced by developments in private saving. During most of the 1980s, private saving fluctuated widely around 15 percent of GDP. (Consistent data for private saving, with adjustments to private and public saving for the inflation component of interest payments on domestic public sector debt, are available only since 1983.) Private saving also fell sharply with the drop in oil prices in 1986, but rose steadily thereafter to a peak of almost 18 percent of GDP in 1988, the year prior to the extended arrangement. This increase is widely attributed to high real interest payments on public debt held by the private sector and to measurement distortions stemming from rising inflation.

The extended arrangement approved in May 1989 had ambitious saving objectives. Both initial medium-term projections and the one-year-ahead projections made for each annual segment of the arrangement envisaged a drop in private saving in 1989, but thereafter a sustained rise of national and private saving rates (Chart 6-4). Private saving was projected to decline in 1989 because of lower real interest income on public debt holdings, large interest payments on foreign private debt, and high—but declining—consumption growth induced by the import liberalization measures of 1988. Private saving rates were projected to rise over the medium term because of an improved environment brought about by a strong fiscal adjustment, a drop in inflation, and wide-ranging structural reforms. Underlying this expected response were two crucial assumptions: (i) the ongoing trade and financial liberalizations would not give rise to long-lasting excess demand pressures; and (ii) except for a short-lived repatriation of flight capital following the debt agreement with commercial banks, private capital inflows would remain moderate.

Chart 6-3.
Chart 6-3.

Mexico and Tunisia: Saving Rates, 1971–92

(In percent of GDP)

Sources: IMF, International Financial Statistics, various issues, and staff estimates.Note: EFF = extended Fund facility.1Public and private saving series are adjusted for the inflation component of interest payments on the public sector debt denominated in domestic currency.2Data on saving of public enterprises are not available separately and are therefore included in the private saving series.

In the event, private saving fell by over 9 percentage points of GDP during the arrangement, with almost three quarters of the fall in the first two years (Table 6-4). Thanks to the strong increase in public saving, also concentrated in the first two years, the rate of national saving fell by slightly less than 3 percentage points of GDP during 1989–92. As it turned out, the two assumptions that were to ensure that private saving rates rose as envisaged were not borne out: aggregate spending—especially on imports—grew steadily despite a stronger-than-expected fiscal adjustment, and private capital inflows surged.

Chart 6-4.
Chart 6-4.

Mexico and Tunisia: Projected and Actual Changes in Saving Rates

(In percentage points of GDP)

Sources: Appendix Table 6 - A4 and IMF staff estimates.1Initial medium-term projections. For Tunisia, initial projections covered only the three-year period originally envisaged for the arrangement.2Projections made at the time of approval of each annual segment of the arrangement.3Projected changes were zero.4Actual change was zero.5One-year projection was zero.

The most important factors contributing to this outcome appear to have stemmed from the decisive implementation of adjustment and structural reforms, which increased the economy’s attractiveness to foreign capital and presumably, raised permanent income and potential growth. These developments are likely to have dampened, on impact, the incentives for private saving. Four influences appear to have been critical:

1. Comprehensive structural reforms. Structural reforms comprised the liberalization of the trade and financial systems, the easing of foreign investment regulations, the negotiation of free trade agreements, and large-scale privatizations.18 Combined with a large decline in inflation and the alleviation of the external debt burden achieved through a financing agreement with commercial banks in 1990,19 these reforms are likely to have increased expected private wealth and reduced the uncertainty of expected future income, thereby depressing current private saving rates.

Table 6-4.

Mexico: Selected Economic Indicators

article image
Source: IMF staff estimates.Note: EFF = extended Fund facility.

Including transfers.

Overall fiscal balance adjusted for the inflation component of interest payments on the public sector debt denominated in domestic currency.

Change from 1988 to average rate of EFF years.

Pesos per U.S. dollar.

Percent a year.

Expressed in U.S. dollars.

Increase means appreciation.

2. Elimination of fiscal imbalances. During 1989–92, the public sector borrowing requirement fell relative to GDP by almost 14 percentage points, while the operational balance improved by 6 percentage points. The adjustment stemmed largely from lower interest payments associated with the large fall in inflation and the completion of the debt-reduction package with commercial banks, but was also supported by tight control over noninterest expenditures. The private sector is likely to have revised downward its expectations about the future tax burden in light of this expenditure-driven improvement in public finances and reacted by reducing its saving rate.

3. Large fall in inflation. Although the data on public and private saving already exclude the inflation component of interest payments on domestic public debt, they are still subject to an inflation bias insofar as they fail to count the proceeds from the inflation tax as a fiscal revenue. Since inflation-tax proceeds fell sharply during the arrangement the series may overstate the actual decline in the rate of private saving.20 Also, the fact that inflation did not converge as fast as expected to the path set for the nominal exchange rate put upward pressure on the real value of the peso, which may have contributed to lower private saving.

4. Increased access to domestic and foreign borrowing. The financial liberalization of 1988–89 improved access to bank credit for previously rationed households and firms. In addition, the reform of the foreign investment code in 1989, the completion of the bank financing package, and large privatizations in 1990–92 set the stage for Mexico’s reentry into the international capital market and prompted a surge in private capital inflows—from –0.2 percent of GDP during 1986–88 to 5.7 percent in 1989–92—that fueled a sustained expansion of aggregate spending and a real appreciation. The easing of borrowing constraints and possible redistributive and wealth effects from the real appreciation are likely to have depressed the rate of private saving.

The depth and strength of the structural changes make it especially difficult to assess the role played by the other possible determinants of private saving. Overall, however, the marginal effect of these determinants is likely to have been of a second order of magnitude, although some of them may have forestalled an even larger decline in the rate of private saving. For instance, aggregate output grew at an average rate of 3.5 percent in 1989–92, more than twice the rate of growth in 1987–88 and also higher than envisaged in the program. The external terms of trade improved by more than 10 percent during the arrangement; however, the largest improvement in the terms of trade, which occurred in 1990, coincided with the largest fall in the private saving rate. The real interest rate doubled from 10 to 20 percent in 1989, before falling sharply in the next three years; it could be argued that lower interest rates reinforced the decline in private saving—through a substitution effect—but this would not explain the large fall in the rate of private saving in 1989. Finally, broad money (M4) increased by more than 66 percent in real terms during 1989–92, reflecting the combined effects of the financial reform, increased confidence in the adjustment program, and re-intermediation of capital flows. The fact that financial savings rose sharply while the private saving rate fell indicates how differently the two can respond to a more liberal economic environment, and flies in the face of the strength of the positive relation between these variables obtained in the cross-program regressions.

Tunisia: Rising Private Saving in an Environment of Gradual Adjustment and Reform

During the high-growth decade of the 1970s the national saving rate in Tunisia fluctuated widely around 21 percent of GDP, reflecting the swings in international prices of oil, gas, arid phosphates (Chart 6-3). National saving fell steadily during 1981–86, influenced by a declining trend in petroleum production, lower agricultural output, deteriorating terms of trade, large public sector imbalances, and a rising external debt burden. After 1983 a declining rate of private saving accounted for most of this fall.21 In 1986, driven by a severe drop in cereal production and a large fall in the terms of trade, the private saving rate bottomed out at 7.7 percent of GDP. In 1987, however, both private and national saving rates rebounded strongly. The abrupt rise in private saving was largely induced by exogenous factors—a tripling of cereal output, a large inflow of tourists following the reopening of the border with Libya, and higher petroleum prices—but may also have been influenced by the lifting of price controls on agriculture and manufacturing, the rise in public enterprise tariffs, and the frequent exchange rate adjustments undertaken during the 1986–87 stand-by arrangement.

The extended arrangement approved in July 1988 had relatively modest saving objectives. At that time, both national and private saving rates were projected to increase by a cumulative 1.5 percentage points during 1988–91. Saving rates were expected to be affected adversely by the 1988 drought and to rise thereafter at a decreasing rate. The saving objectives in the one-year-ahead projections in each annual segment of the arrangement were generally more modest. Taken together, these projections envisaged that the rate of national saving would fall by 3.3 percentage points during 1988–91, or by 1.5 percentage points if the program’s extension to 1992 is included.22 The bulk of the projected fall was to stem from a lower rate of central government saving, with only a small cumulative decline in the rate of private saving (Chart 6-4). The projected change in national saving was expected to be mirrored by fluctuations in domestic investment so that the external current account would vary around 2 percent of GDP in the short and medium term.

Table 6-5.

Tunisia: Selected Economic Indicators

article image
Source: IMF staff estimates.Note: EFF = extended Fund facility.

Central government saving.

Including transfers.

Including net lending.

Change from 1987 to average rate of EFF years.

Dinars per U.S. dollar.

Percent a year.

Expressed in SDRs.

Increase means appreciation.

As it turned out, private saving increased relative to GDP by 3.5 percentage points during the arrangement as extended to 1992 (Table 6-5). The increase more than offset a larger-than-expected deterioration in central government saving; the national saving rate therefore rose by almost 2 percentage points. Private saving rates rose strongly in 1988–89, despite the adverse effects of the long-lasting drought on agricultural income, and rebounded from a fall in 1991 caused by the sharp drop in exports, tourism receipts, and workers’ remittances brought about by the Middle East crisis. Notwithstanding the strength of private and national saving, the external current account worsened markedly in 1989–92, driven by a large increase in investment by private and public enterprises.23

At first glance, the contrast between the behavior of private saving in Mexico and Tunisia may seem somewhat perverse. Both economies underwent a considerable transformation, although Mexico’s was more dramatic as its financial situation improved from a much worse starting position, structural reforms were farther reaching, and external developments were more favorable. Still, private saving rates fell sharply in Mexico and rose in Tunisia. The regression estimates reported above shed some light on this contrasting behavior. The regressions account for a larger fraction of the observed variation in saving rates in Tunisia than in Mexico, suggesting that the dampening effect on private saving of those factors not included in the regressions were relatively less important in Tunisia. More generally, the changes in private saving appear to have been more related to quantifiable determinants of saving in Tunisia than in Mexico. The main factors that seem to account for the behavior of private saving in Tunisia include:

1. Slow progress in reducing the fiscal deficit. The steady decline in petroleum (nontax) revenues throughout the arrangement was not accompanied by a commensurate reduction in current spending. In part this was due to the bulge in consumer subsidies induced by the 1988–89 drought and to the cost of servicing government debt at increasingly market-determined interest rates, but there were also difficulties in cutting other current expenditures. To the extent that they supplemented private incomes, higher consumer subsidies and interest payments may have had a positive effect on private saving. Moreover, the inability to achieve a lasting reduction in central government spending in spite of declining revenues may have signaled an increase in the future tax burden and raised the desired rate of private saving.

2. Strong growth. Despite the output disruptions stemming from the 1988–89 drought and the 1990–91 Middle East crisis, average GDP growth during 1989–92 at around 6 percent was considerably higher than the 3.6 percent average during 1985–87 and also the 3.5 percent average rate in Mexico. Growth was broadly based but continued to be strongly influenced by climate-sensitive developments in agriculture. Both the pick-up in growth and the high variability of income are likely to have positively affected private saving.

3. Rising real interest rates. Following a temporary fall in 1988, real interest rates trended upward during the arrangement as a result of the gradual liberalization of interest rates, the introduction of new financial instruments, and the government’s increased recourse to nonbank financing. Not only did these reforms have the expected favorable effect on financial deepening, as in Mexico, but they also appear to have attracted larger inflows of workers’ remittances and private transfers—which jointly increased from 2 percent of GDP in 1986–87 to 3 percent in 1988–92—that were translated into a higher recorded rate of private saving.

4. Low inflation. Inflation was low and stable before and during the extended arrangement; this suggests that—unlike Mexico—the inflation bias in the measures of private and public saving was small, and, more generally, that inflation was not a major determinant of the changes in private saving.

5. Public enterprise performance. By construction, the private saving data in Tunisia include savings of public sector entities other than the central government. The financial position of public enterprises is believed to have improved markedly with the early reforms to the price and exchange systems and, after 1989, the introduction of three-year performance contracts that required firms to set well-defined economic and financial targets. To the extent that the saving rates of public enterprises rose, the measured changes in “private” saving overstate the actual rise.

6. Gradual structural reforms. Despite the ambitiousness of the initial objectives, the actual pace of liberalization of the price, trade, exchange, and financial systems initiated in 1986–87 slowed during the extended arrangement. Price controls and import barriers were phased out selectively, foreign exchange restrictions were eased gradually, the privatization of large and medium-sized public enterprises was delayed, complex investment incentives were only partially simplified, and the deregulation of the domestic financial system was slower than expected. Although progress was made in all these areas, the impact on expected private wealth, and therefore on private saving, is likely to have been weaker than in Mexico.

7. Limited access to external financing. Contrary to Mexico, Tunisia did not experience a surge in private capital inflows. Aside from the increase in private transfers noted before, external borrowing and foreign investment remained at about their prere-form levels until 1992, when the first syndicated loan since 1986 and a large influx of foreign investment for the construction of a new gas pipeline were secured. Without significant private capital inflows or vigorous deregulation of the domestic financial market, the borrowing constraints faced by Tunisian households and firms are likely to have remained tight. Reinforced by the policy of preserving external competitiveness through continuous and small devaluations, the lack of large-scale external financing placed a curb on the expansion of private consumption and forestalled a drop in private saving.

Concluding Observations

  • On average, and regardless of the length of IMF involvement, national and public saving rates increased while private saving rates fell during the programs under review. Declines in private saving tended to be slightly more frequent but larger than rises in private saving and provided an important offset to the increases in public saving achieved. Private saving rates were envisaged to decline in about half of the programs and, on average, fell almost in line with projections. Nonetheless, deviations from program projections for private saving rates in individual arrangements tended to account for a substantial fraction of the deviations from program targets of national saving.

  • Among a number of quantifiable factors tested, changes in output growth and in the rates of public and foreign saving had the most important and systematic effects on private saving. The dominance of these influences underscores the difficulty of significantly increasing private saving during programs: output growth depends on the interplay of exogenous factors, policy variables, and expectations subject to great uncertainty and unpredictable lags even in the best of circumstances; increases in public saving, while systematically contributing to higher national saving, tend to elicit partially offsetting declines in private saving rates regardless of whether they are driven by higher public revenues or lower current expenditures; and changes in foreign saving depend as much on the overall strength of the countries’ adjustment efforts as on their access to international capital markets and the behavior of external terms of trade. Contrary to the findings of other studies, however, the latter were found to have had only a weak independent effect on private saving.

  • Apart from the three major influences noted above, a range of other policies implemented in programs—such as raising interest rates, devaluing the domestic currency, reducing inflation through tight financial policies, and liberalizing the trade and financial systems—were often expected to lead to rapid increases in private saving rates. Consistent with theory and other empirical evidence, however, these policies appear to have had weak and unpredictable effects, at least in the short-term period under review. Overall, the experience reviewed does not reveal clear channels for raising private saving rates in the short term during macroeconomic adjustment.

  • The inverse relationship between public and private saving rates seems to be part of a wider pattern of adverse short-term responses of private saving to adjustment and reform programs. Although the actual response depends on the economy’s initial conditions, it is not uncommon to see private saving rates fall as strong programs, including both financial and structural measures, are implemented, especially when the programs succeed in attracting large inflows of foreign capital. This is a natural short-term response that reflects domestic residents’ attempts to raise and smooth their consumption paths as economic prospects in general, and access to foreign credit in particular, improve. While this process operates as a somewhat vexing damper on external adjustment, a body of evidence on longer-term experiences suggests that maintaining stable financial policies and market-oriented reforms considerably increase the likelihood of achieving sustained growth and rising private saving over the medium term.

Bibliography

  • Aghevli, Bijan, and others, The Role of National Saving in the World Economy: Recent Trends and Prospects, IMF Occasional Paper No. 67 (Washington: International Monetary Fund, 1990).

    • Search Google Scholar
    • Export Citation

  • Arrau, Patricio, and Daniel Oks, Private Saving in Mexico, 1980–90, Policy Research Working Paper No. WPS 861 (Washington: World Bank, February 1992).

    • Search Google Scholar
    • Export Citation

  • Bosworth, Barry, Saving and Investment in a Global Economy (Washington: Brookings Institution, 1993).

  • Corbo, Vittorio, and Klaus Schmidt-Hebbel, “Public Policies and Saving in Developing Countries,” Journal of Development Economics, Vol. 36 (July 1991), pp. 89–115.

    • Search Google Scholar
    • Export Citation

  • Deaton, Angus, “Saving in Developing Countries: Theory and Review,” Proceedings of the World Bank Annual Conference on Development Economics 1989 (Washington: World Bank, 1989).

    • Search Google Scholar
    • Export Citation

  • Easterly, William, and Klaus Schmidt-Hebbel, “Fiscal Deficits and Macroeconomic Performance in Developing Countries,” World Bank Research Observer, Vol. 8 (July 1993), pp. 211–37.

    • Search Google Scholar
    • Export Citation

  • El-Erian, Mohamed A., “Mexico’s External Debt Policies, 1982–90,” in Mexico: The Strategy to Achieve Sustained Economic Growth, ed. by Claudio Loser and Eliot Kalter, IMF Occasional Paper No. 99 (Washington: International Monetary Fund, 1992).

    • Search Google Scholar
    • Export Citation

  • Hamann, A. Javier, Private Saving, Public Saving, and the Inflation Tax: Another Look at an Old Issue, IMF Working Paper No. WP/93/37 (Washington: International Monetary Fund, April 1993).

    • Search Google Scholar
    • Export Citation

  • Haque, Nadeem U., “Fiscal Policy and Private Sector Behavior in Developing Economies,” Staff Papers, International Monetary Fund, Vol. 35 (June 1988), pp. 316–35.

    • Search Google Scholar
    • Export Citation

  • Haque, Nadeem U., and Peter Montiel, “Consumption in Developing Countries: Tests for Liquidity Constraints and Finite Horizons,” Review of Economics and Statistics, Vol. 71 (August 1989), pp. 408–15.

    • Search Google Scholar
    • Export Citation

  • Hill, Peter, “Inflation, Holding Gains, and Saving,” OECD Economic Studies, No. 2 (Spring 1984), pp. 151–64.

  • International Monetary Fund, Mexico: Recent Economic Developments (unpublished; Washington: International Monetary Fund, 1993).

  • Krugman, Paul, and Lance Taylor, “Contractionary Effects of Devaluation,” Journal of International Economics, Vol. 8 (November 1978), pp. 445–56.

    • Search Google Scholar
    • Export Citation

  • Lizondo, J. Saúl, and Peter Montiel, Contractionary Devaluation in Developing Countries: An Analytical Overview, IMF Working Paper No. WP/88/51 (Washington: International Monetary Fund, June 1988).

    • Search Google Scholar
    • Export Citation

  • Loser, Claudio, and Eliot Kalter, eds., Mexico: The Strategy to Achieve Sustained Economic Growth, IMF Occasional Paper No. 99 (Washington: International Monetary Fund, 1992).

    • Search Google Scholar
    • Export Citation

  • Marfán, Manuel, and Barry Bosworth, “Saving, Investment, and Economic Growth,” in The Chilean Economy: Policy Lessons and Challenges, ed. by B. Bosworth and others (Washington: Brookings Institution, 1994).

    • Search Google Scholar
    • Export Citation

  • Nsouli, Saleh M., and others, The Path to Convertibility and Growth: The Tunisian Experience, IMF Occasional Paper No. 109 (Washington: International Monetary Fund, 1993).

    • Search Google Scholar
    • Export Citation

  • Rossi, Nicola, “Government Spending, the Real Interest Rate, and the Behavior of Liquidity-Constrained Consumers in Developing Countries,” Staff Papers, International Monetary Fund, Vol. 35 (March 1988), pp. 104–40.

    • Search Google Scholar
    • Export Citation

  • Schmidt-Hebbel, Klaus, Steven B. Webb, and Giancarlo Corsetti, “Household Saving in Developing Countries: First Cross-Country Evidence,” World Bank Economic Review, Vol. 6 (September 1992), pp. 529–47.

    • Search Google Scholar
    • Export Citation

  • World Bank, The East Asian Miracle: Economic Growth and Public Policy (New York: Oxford University Press for the World Bank, September 1993).

    • Search Google Scholar
    • Export Citation

Appendix Tables

Table 6-A1.

Saving Rates: Total, Public, and Private—Full Sample

(In percentage points of GDP)

article image
article image
Source: IMF staff estimates.

Ratios at constant prices.

Fiscal year ending June 30.

Fiscal year ending September 30.

Fiscal year beginning April 1.

Ratios to GNP.

Table 6-A2.

Saving Rates: Private, Public, and National, Changes by Program Year

(In percentage points of GDP)

article image
Source: Appendix Table 6-A1.

Ratios at constant prices.

Domestic saving.

Fiscal year ending June 30.

Fiscal year ending September 30.

Fiscal year beginning April 1.

Ratios to GNP.

Table 6-A3.

Cumulative Changes in Saving Rates During IMF Involvement

(Saving rates in percent of GDP; changes in percentage points of GDP)

article image
Source: Appendix Table 6-A1.Note: EFF = extended Fund facility.

Where the period of involvement covers more than one consecutive arrangement, the pre-arrangement year is defined as that preceding the first arrangement.

Ratios at constant prices.

Domestic saving.

Fiscal year ending June 30.

Fiscal year ending September 30.

Fiscal year beginning April 1.

Ratios to GNP.