A. Introduction

1. This section analyzes several issues regarding fiscal sustainability (intertemporal solvency) and fiscal adjustment in Brazil during the 1990s and searches for econometric evidence of a monetary dominant regime during some subperiods, i.e., a regime where adjustments to the deterioration in public sector net worth are carried out via fiscal policy, rather than price level adjustments. The paper finds that for the 1990s as a whole, we cannot reject the hypothesis that fiscal policy was sustainable but there is little evidence supporting the presence of a monetary dominant regime. In particular, we find that:

• While for the 1990s as a whole, the operational deficit measured at constant prices was stationary—i.e., it was not growing boundlessly, implying an unsustainable path for real net public debt—the analysis of different subperiods shows significant heterogeneity.

• In the years that preceded the inception of the Real Plan of July 1994, econometric evidence points to the presence of a Fiscal Dominant (FD), or Non-Ricardian regime, rather than a Monetary Dominant (MD) regime, since there is no statistical evidence that the primary surplus actively responded to changes in real government indebtedness. During this period, inflation was an important source of fiscal finance, with monetary policy subordinated and highly constrained by fiscal financing requirements.

• A perceptible regime break seems to have taken place during the first couple of years following the implementation of the real—The presence of a MD regime cannot be ruled out, possibly helped by a substantial increase in tax revenues.

• During 1995–96, as the ex-post implicit real interest rates on public debt fell, the government borrowed and spent and, the regime seems to have shifted back to FD again around the end of 1996 until early 1999, with the fiscal deficit and net public debt entering an unsustainable path.

• The fiscal adjustment effort envisaged in the three-year Fiscal Stabilization Plan announced at the end of 1998, helped to bring public debt back to a sustainable path, but evidence does not yet point to the unequivocal presence of a MD regime. We may need more data of strong fiscal performance for the expected regime shift to be statistically robust and perceived as being permanent. The main findings are summarized in the Table below.

		Fiscal Policy
		Sustainable	Unsustainable
Regime	Monetary Dominance	Cannot rule out a monetary dominant regime from August 1994 to end-1996, and potentially 1999 to 2000 (weak evidence)
	Fiscal Dominance	1991 to July 1994	End-1996 end-1998 (weak fiscal dominance)

View Table

		Fiscal Policy
		Sustainable	Unsustainable
Regime	Monetary Dominance	Cannot rule out a monetary dominant regime from August 1994 to end-1996, and potentially 1999 to 2000 (weak evidence)
	Fiscal Dominance	1991 to July 1994	End-1996 end-1998 (weak fiscal dominance)

B. The Sustainability of Fiscal Policy in Brazil

2. In recent years, a sizable literature devoted to assessing the sustainability of a country’s fiscal policy has emerged. Several techniques from this literature are applied to Brazil.

3. The government, like all other economic agents, faces an intertemporal budget constraint: the present discounted value of its net liabilities must, by definition, equal the present discounted value of future primary surpluses (tax revenues minus noninterest expenditures). If this constraint in real terms can be satisfied without a change in either policy or the price level, current fiscal policy is said to be sustainable. That is, the equilibrium price level and the level of seignorage can be determined independently by monetary policy (equilibrium in the money market) irrespective of budget financing considerations.

4. The Brazilian government satisfied its intertemporal budget constraint in fairly different ways during the post-war period. Upward adjustments to the primary surplus contributed to limit the real value of Brazilian debt between the mid-1970s and the mid-1980s. Between 1985 and 1994—a period encompassing six stabilization plans—the nominal balance (PSBR) followed the ups and downs of inflation, even if we can clearly identify two distinct subperiods: The first period spans the entire decade of the 1980s, and is characterized by substantial operational deficits, while the second period (1990–94) is characterized by a less acute fiscal disequilibrium, with operational balances around zero (i.e., the primary surplus was able to cover the real interest bill).

Brazil: Operational deficit, real interest payments, and primary surplus

Sources: BCB; and Fund staff estimates.

Brazil: Operational deficit, real interest payments, and primary surplus

Average (% GDP)	1985-89	1990-94	1995-98	1999
Operational deficit	5.2%	0.0%	5.1%	3.9%
Real interest payments	5.8%	2.8%	4.9%	7.1%
Primary surplus	0.7%	2.8%	-0.2%	3.2%

Sources: BCB; and Fund staff estimates.

View Table

Brazil: Operational deficit, real interest payments, and primary surplus

Average (% GDP)	1985-89	1990-94	1995-98	1999
Operational deficit	5.2%	0.0%	5.1%	3.9%
Real interest payments	5.8%	2.8%	4.9%	7.1%
Primary surplus	0.7%	2.8%	-0.2%	3.2%

Sources: BCB; and Fund staff estimates.

5. The reverse Tanzi effect on revenues² and the high levels of inflation were behind the decline of the operational deficit observed during the early 1990s. During this period, real government debt was reduced with the help of inflation (which also led to sizable primary surpluses) and under indexation of liabilities (systematic overshooting of inflation in tandem with a reduction of the operational fiscal deficit). At the time, many argued that, despite the low operational deficits, fiscal adjustment would be needed, as the ex-inflation structural deficit remained high.

6. The Real Plan of 1994 successfully conquered inflation, and initially boosted the primary surplus (through higher tax collections). From 1995 until late 1998, the primary deficit and real interest payments rose, causing a buildup of real public debt (see Table above). During the period 1994–98, the states and municipalities posted the worst primary balances of all three levels of government, although the largest fiscal deterioration came from the central government—discretionary current and capital expenditures (OCCs) grew by 1.0 percentage point of GDP (accumulated real growth of 78 percent) and social security benefits for private sector workers expanded by 1.3 percentage point of GDP (see Giambiagi and Além 1999). In all, nonfinancial expenditure of the central government increased by 2.8 percentage points of GDP from 1994 to 1998.

7. During the high inflation years the OCCs had been the main variable of adjustment. However, during 1994–98, OCCs increased in real terms from R$17 billion in 1994 to R$31 billion in 1998 (at December 1998 prices). The fact that the bulk of the increase in OCCs happened after 1995, and not immediately after July 1994 (Real Plan), seems to indicate that the real increase in discretionary spending was not so much the inevitable consequence of the decrease in the inflation rate but instead a conscious political decision (see Giambiagi and Além 1999).

8. From 1990–94 the operational deficit was smaller than the revenues from seignorage, which favored the monetization of debt and released the pressure for tighter fiscal execution. Consequently, net public debt to GDP was on a declining trend since the operational deficit was basically zero and GDP was growing at a moderate pace (average of 1.3 percent). The opposite happened during 1995–98 (see table and figure below) leading the ratio of net public debt to GDP to increase at a fast rate despite the moderate expansion of real GDP.

Fiscal Deficit and Real Seigniorage

(% GDP)

Source: Giambiagi and Além (1999).

Fiscal Deficit and Real Seigniorage

(% GDP)

	1981/84	1985/89	1990/94	1995/98
Operacional deficit	5.0	5.1	0.0	5.1
Real seigniorage	1.8	2.8	3.3	0.6

Source: Giambiagi and Além (1999).

View Table

Fiscal Deficit and Real Seigniorage

(% GDP)

	1981/84	1985/89	1990/94	1995/98
Operacional deficit	5.0	5.1	0.0	5.1
Real seigniorage	1.8	2.8	3.3	0.6

Source: Giambiagi and Além (1999).

Operational Deficit and Seigniorage Financing, 1985/1998

(% GDP)

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Operational Deficit and Seigniorage Financing, 1985/1998

(% GDP)

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

Operational Deficit and Seigniorage Financing, 1985/1998

(% GDP)

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

9. The sharp deterioration of the fiscal stance during 1995–98 eventually contributed to the abandonment of Brazil’s crawling-peg exchange rate regime in early 1999. In response, taxes were raised, expenditures cut, and the primary surplus increased to over 3 percent of GDP. Through the end of the third quarter of 2000, key fiscal targets were met under a three-year (1999–01) Stand-by arrangement from the Fund. However, at the center of the recent fiscal policy debate in Brazil has been the search for an answer to the question whether the recent fiscal adjustment effort undertaken in the wake of the Russian moratorium of August 1998, and the January 1999 floatation of the real, represents a short-term, transitory adjustment effort or whether it represents a deeper and structural change in fiscal regime.

10. The analysis of the whole period 1991–00 and the recent evolution of the fiscal stance should help shed some light and provide valuable inputs into some of these questions. However, as seen above, with the possible exception of the last two years and the initial period of the Real Plan, fiscal policy does not seem to have actively responded to changes in government indebtedness which leads us to suspect that, evidence of a monetary dominant regime may be meager. In the sections that follow, we will present a stationarity test for the real operational deficit, and two tests designed to distinguish between fiscal and monetary dominant regimes (a simple single equation test, and a richer vector autoregression system (VAR)). Finally, to assess which, if any, fiscal policy variable responded to changes in public indebtedness—extra revenues or cuts in expenditure—a more disaggregated VAR system is used.

Empirical results

12. During the period under study, there have been dramatic changes in policy, and parameters may vary over time. Accordingly, estimates are presented for both the entire 1991–00 period and selected subsamples, namely the pre-real period (1991:1–1994:6), the post-real period (1994:7–2000), and the post-Tequila period (1995:4–2000). However, since these subperiods are to some degree arbitrary, estimates for rolling 24-month windows, beginning with the period 1991:1–1993:1 and ending with the period 1998–00, are also presented.

13. Table 1.1 presents the stationarity tests mentioned above: ADF, Zt, and Zα. For the entire sample, as well as all three subperiods, most results suggest that the operational deficit was stationary rendering fiscal policy sustainable: with the exception of the ADF statistics for the pre-Real period, for all three statistics and all periods, the null hypothesis of the nonstationarity of ODEF is rejected at the 90 percent confidence level or better.

Table 1.1:

Stationarity Tests, Real Operational Deficit (ODEF)

$\text{(1)\hspace{0.17em}\hspace{0.17em}}\mathrm{\Delta }ODE{F}_t=a_0+a_{1}ODE{F}_{t-1}+\underset{j=2}{\overset{j}{\mathrm{\Sigma }}}{a}_j\mathrm{\Delta }ODE{F}_{t-j+1}+{\xi }_t$

Notes: Null hypothesis is H₀: a₁=0 (i.e. non stationary). ADF(x), Zt(x), and Zα(x) are the Augmented Dickey Fuller (Equation (5)) test, Phillip’s Zt and Zα tests (See Phillips (1987) and Phillips and Perron (1988), respectively, where x is the number of augmented terms included in the test. The 95 and 99 percent critical values for the ADF(x) and Zt tests are -3.00 and -3.75, respectively. The 90 and 95 percent critical values for the Zα test are -8.0 and -13.6, respectively.

^*Represents failure to reject the null hypothesis of nonstationary process. Critical values are from Fuller (1976, pp.371—73).

Table 1.1:

Stationarity Tests, Real Operational Deficit (ODEF)

$\text{(1)\hspace{0.17em}\hspace{0.17em}}\mathrm{\Delta }ODE{F}_t=a_0+a_{1}ODE{F}_{t-1}+\underset{j=2}{\overset{j}{\mathrm{\Sigma }}}{a}_j\mathrm{\Delta }ODE{F}_{t-j+1}+{\xi }_t$

	ADF(1)	ADF(2)	Zt(1)	Zt(1)	Zα(1)	Zα(1)
All sample	-8.55	-5.81	-9.69	-9.68	-97.67	-90.49
Pre-real	-2.39*	-1.72*	-3.93	-3.91	-32.28	-31.95
Post-real	-7.14	-4.66	-7.58	-7.55	-62.15	-56.19
Post-Tequila	-6.71	-4.25	-7.09	-7.09	-52.89	-47.50

Notes: Null hypothesis is H₀: a₁=0 (i.e. non stationary). ADF(x), Zt(x), and Zα(x) are the Augmented Dickey Fuller (Equation (5)) test, Phillip’s Zt and Zα tests (See Phillips (1987) and Phillips and Perron (1988), respectively, where x is the number of augmented terms included in the test. The 95 and 99 percent critical values for the ADF(x) and Zt tests are -3.00 and -3.75, respectively. The 90 and 95 percent critical values for the Zα test are -8.0 and -13.6, respectively.

^*Represents failure to reject the null hypothesis of nonstationary process. Critical values are from Fuller (1976, pp.371—73).

View Table

Table 1.1:

Stationarity Tests, Real Operational Deficit (ODEF)

$\text{(1)\hspace{0.17em}\hspace{0.17em}}\mathrm{\Delta }ODE{F}_t=a_0+a_{1}ODE{F}_{t-1}+\underset{j=2}{\overset{j}{\mathrm{\Sigma }}}{a}_j\mathrm{\Delta }ODE{F}_{t-j+1}+{\xi }_t$

	ADF(1)	ADF(2)	Zt(1)	Zt(1)	Zα(1)	Zα(1)
All sample	-8.55	-5.81	-9.69	-9.68	-97.67	-90.49
Pre-real	-2.39*	-1.72*	-3.93	-3.91	-32.28	-31.95
Post-real	-7.14	-4.66	-7.58	-7.55	-62.15	-56.19
Post-Tequila	-6.71	-4.25	-7.09	-7.09	-52.89	-47.50

Notes: Null hypothesis is H₀: a₁=0 (i.e. non stationary). ADF(x), Zt(x), and Zα(x) are the Augmented Dickey Fuller (Equation (5)) test, Phillip’s Zt and Zα tests (See Phillips (1987) and Phillips and Perron (1988), respectively, where x is the number of augmented terms included in the test. The 95 and 99 percent critical values for the ADF(x) and Zt tests are -3.00 and -3.75, respectively. The 90 and 95 percent critical values for the Zα test are -8.0 and -13.6, respectively.

^*Represents failure to reject the null hypothesis of nonstationary process. Critical values are from Fuller (1976, pp.371—73).

14. However, ADF statistics from rolling two-year samples (Figure 1.1) reveal a somewhat different picture. During the pre-real period, it is only in 1992 that it is not possible to reject the null hypothesis of nonstationarity³. For windows covering most of 1994 through 1996, we reject nonstationarity at the 95 percent level. Importantly, for 24-month rolling windows starting in early 1995 (capturing already part of 1997) through early 1997 (that is, ending in early 1999), the ADF statistic is higher in absolute value than the critical value, once again suggesting a nonstationary process for the operational deficit. Finally, for rolling windows starting from 1997:3 onward, we observe something akin to a regime shift, since the null hypothesis of nonstationarity is now rejected at the 90 percent level or better.

ADF Statistic, Two-Year Rolling Samples

Brazil: Estimates of ADF (2) Statistic, ODEF

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

ADF Statistic, Two-Year Rolling Samples

Brazil: Estimates of ADF (2) Statistic, ODEF

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

ADF Statistic, Two-Year Rolling Samples

Brazil: Estimates of ADF (2) Statistic, ODEF

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

15. It can be clearly visualized in Figure 1.1 that there is a progressive deterioration of the fiscal stance since mid-95 until early 1997 with the ADF statistic line increasing continuously until it drops abruptly in early 1997 (for the 24-month windows covering mid-1997 to mid-1999) since by mid-1999 there are already the first solid indications that the government has started to consistently post sizable primary surpluses.

C. The Adjustment Variable: Price Level (Fiscal Dominance) Versus Primary Surplus (Monetary Dominance)

16. Interconnected with the fiscal sustainability issues is the government’s choice of adjustment mechanism, that is, what variables limit government indebtedness according to present value relations (A2) and (A3) (see Appendix I). If the primary deficit is the policy adjustment variable of choice (PDEF – G-T^* ; through changes in G and/or T^*) it should respond to variations in the real value of government debt, thus guaranteeing balance of the government’s intertemporal budget, according to equations (A2) and (A3), without requiring price level adjustments (i.e. resorting to the inflation tax). That is, the government adjusts the primary surplus upward to ward off or to limit debt accumulation, not forcing the central bank to inflate away the debt in order to satisfy the intertemporal budget constraint. In this case, monetary policy variables (domestic interest rates, the exchange rate, and the domestic price level) are not determined by fiscal considerations, but simply by money market equilibrium relations. Such a regime has been called in the recent literature Monetary Dominant or Ricardian (see Canzoneri, Cumby, and Diba (1998), CCD).

17. On the other hand, if the level of the primary surplus is chosen independently (possibly following an exogenous political process) of the level of expected real interest payments or the level of real net public debt, equations (A2) and (A3) must be satisfied, i.e., intertemporal equilibrium must be achieved, by changes in the current price level in order to tailor the real level of net public debt to the expected path of current and future primary deficits. In this case, monetary policy is subordinated to fiscal needs and we live under a Fiscal Dominant or non-Ricardian regime. The price level is then determined by fiscal needs, not the traditional equilibrium in the money market, since recurrent adjustments in the price level, rather than the primary surplus, guarantee intertemporal balance as defined by (A3) at any point in time by debasing the stock of nominal public debt—this has led to the so-called fiscal theory of the price level (FTPL) pioneered by recent research of Woodford (1995, 1996), Cochrane (1998), CCD (1998), Sims (1994, 1998), and others.⁴ Of course, the FTPL fails to hold under an MD regime. For the FTPL to hold, as pointed out by several authors, a necessary but not sufficient condition is the prevalence of a FD regime since, under a MD regime, the price level is determined in the monetary sector of the economy.

Single-equation test: The response of ΔPDEF to ΔRIP

20. Even if the government’s intertemporal budget constraint (A2) holds through adjustments in PDEF, it may be unrealistic to expect that the PDEF and RIP/(1+r) move (minus) one-to-one over the short run. Instead, consider the testable reaction function (see Appendix I):

$\begin{array}{cc}\left(2\right)& \mathrm{\Delta }PDE{F}_t={\beta }_0+{\beta }_1\mathrm{\Delta }RI{P}_t+erro{r}_t\end{array}$

21. This is a backward-looking test since today’s real interest payments apply to yesterday’s outstanding debt stock and since equation (2) is specified in first differences, β₁ reflects a short-term relationship between PDEF and RIP, not a cointegrating relationship.⁵

22. Unfortunately, while equation (2) can in some cases rule out a MD regime, it cannot distinguish between MD and FD regimes. For example, under a MD regime, the primary deficit should decline when real interest payments go up (β₁ < 0). Yet, under a FD regime, if the FTPL also holds, the price level jumps and real interest payments fall in anticipation of future primary deficits, implying again β₁ < 0. In either case—MD regime or FD and FTPL—β₁ would be negative. However, β₁ will be zero or positive only under a FD regime. Therefore, if β₁ > 0 or β₁ = 0, a MD regime is ruled out.

23. Table 1.2 presents estimates of equation (2) for both the entire period and the same subsamples presented above. In all cases, there appears to be no statistically significant reaction of ΔPDEF to ΔRIP. Instead, the null hypothesis of H₀: β₁ = 0 is never rejected. The results do not favor a MD regime: i. e. for neither the entire period nor the subperiods was there evidence that the deterioration of the public sector’s net worth was arrested or limited by regular adjustments to the primary deficit during the 1990s, which corroborates evidence presented earlier in the section showing that inflation seems to have been the adjustment variable of choice during the 1990–94 period, while the stochastic process driving the primary deficit, at least from 1996 until 1998, seems that have been chosen independently of the deterioration of the public sector net worth witnessed during the period.

Table 1.2:

Response of ΔPDEF to ΔRIP (β₁)

$\mathit{\Delta }PDE{F}_t={\beta }_0\text{+}{\beta }_l\mathrm{\Delta }RI{P}_t+erro{r}_t$

Table 1.2:

Response of ΔPDEF to ΔRIP (β₁)

$\mathit{\Delta }PDE{F}_t={\beta }_0\text{+}{\beta }_l\mathrm{\Delta }RI{P}_t+erro{r}_t$

(Standard errors in parenthesis)	β0	β1
All sample	-18.0	0.00
	(233.8)	(0.02)
Pre-real	-10.3	-0.08
	(334.3)	(0.08)
Post-real	-0.11	0.00
	(316.5)	(0.02)
Post-Tequila	-79.9	0.00
	(332.4)	(0.03)

View Table

Table 1.2:

Response of ΔPDEF to ΔRIP (β₁)

$\mathit{\Delta }PDE{F}_t={\beta }_0\text{+}{\beta }_l\mathrm{\Delta }RI{P}_t+erro{r}_t$

(Standard errors in parenthesis)	β0	β1
All sample	-18.0	0.00
	(233.8)	(0.02)
Pre-real	-10.3	-0.08
	(334.3)	(0.08)
Post-real	-0.11	0.00
	(316.5)	(0.02)
Post-Tequila	-79.9	0.00
	(332.4)	(0.03)

24. However, estimates of β₁ from rolling 24-month windows (see Figure 1.2) show evidence of periods where the presence of a MD regime cannot be ruled out. For most of the sample period, β₁ is not significantly different from zero, i.e., the zero line stays inside the two-standard bands. However, from 1995:4–1997:4 to 1995:11–1997:11, a period that includes the beginning of the Real Plan (July 1994), β₁ is negative and statistically different from zero (the zero line stays outside the two-standard upper band). Thus, a MD regime cannot be ruled out during this period. This result supports the evidence portrayed by the sustainability tests and heuristic evidence that points to a relatively conservative fiscal policy in the earlier years of the real followed by the loosening of the fiscal stance after 1996. However, for subsequent periods, estimates of β₁ are again not different from zero, once more ruling out a MD regime, despite the fact that for the most recent period (1999 and part of 2000), the standard errors of the parameter estimates narrow significantly (see Figure 1.2).

Response of ΔPDEF to ΔRIP (β₁, Equation (2)) Two-Year Rolling Samples

(Dotted Lines Indicate Two-Standard Errors).

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Response of ΔPDEF to ΔRIP (β₁, Equation (2)) Two-Year Rolling Samples

(Dotted Lines Indicate Two-Standard Errors).

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

Response of ΔPDEF to ΔRIP (β₁, Equation (2)) Two-Year Rolling Samples

(Dotted Lines Indicate Two-Standard Errors).

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

25. As noted before, after 1998Q4 fiscal policy responded consistently and forcefully to the increase in real interest payments and the deterioration of the public sector net worth due to high real interest payments and the debt stock effect of the floatation of the real. However, one cannot yet unequivocally point to the presence of a MD regime since these adjustments are recent and as such may have not have been perceived, in a statistical sense, as sufficiently permanent (possibly due to lack of enough observations) and hence not statistically related to the sharp increase in real interest payments. There is, however, a perceived break with the sharp deterioration observed from 1995 until end-1998.

Vector autoregression analysis

26. The single equation tests performed above have some limitations. Such a framework does not allow one to discriminate between ex-post backward-looking adjustments—where the primary deficit is reduced in response to increases in real interest payments (consistent with a MD regime)—and ex-ante forward-looking adjustments of the level of real interest payments (through contemporaneous price increases) in anticipation of future primary deficits (consistent with a FD regime and the FTPL). A modification of CCD’s (1998) VAR will allow a richer analysis of the causality and direction of fiscal adjustment. Let,

$\begin{array}{cc}\left(3\right)& \mathrm{\Delta }{X}_t=a_0+a_1\mathrm{\Delta }{X}_{t-\mathit{1}}+a_2\mathrm{\Delta }{X}_{t-\mathit{2}}\text{+\hspace{0.17em}}\end{array}\mathrm{\dots .}+v_t$

where X = [RIP, PDEF] is a dim(2) vector of the real interest payments and the real primary deficit, a_t is a vector of coefficients, and v_t = (v_PDEF, v_RIP) is a vector of error terms.⁶ We will assume a richer error covariance structure by positing that each element of the error vector v_t is in turn composed of “own” error terms w_t = (W_PDEFt, W_RIPt) and contemporaneous correlations with “other” errors:

$\begin{array}{cc}\left(4\right)& v_t=B{w}_t\end{array}$

where B is a 2 × 2 matrix whose diagonal elements equal one and whose nonzero offdiagonal elements reflect contemporaneous correlations among the error terms. Also, (4) yields impulse response functions (IRF) that summarize the effects of current innovations w_t on values of X.

27. The VAR system (3) estimates relationships of time-series causality between variables (Maddala, 1992) that run in both directions. These timeseries relationships have economic interpretations that depend on the direction and the sign, as summarized in Table 1.3 below.

Table 1.3:

Summary of Interpretation, System (3), X = [RIP, PDEF]

^1/One standard innovation.

Table 1.3:

Summary of Interpretation, System (3), X = [RIP, PDEF]

	Current Primary Deficit (ΔPDEFt)1/ → Future Real Interest Payments (ΔRIPt+j)
Positive	Government pays down future debt, consistent with MD regime.
Zero	Primary deficit exogenous, consistent with FD regime.
Negative	Government anticipates future interest bill, consistent with MD regime.
	Current real interest payments (ΔRIPt)1/ → future primary deficit (ΔPDEFt+i)
Positive	Unstable policy, consistent with FD regime; or interest rates anticipate future primary deficits.
Zero	Primary deficit exogenous, consistent with FD regime.
Negative	Primary deficit reduction offsets increase in public debt, consistent with MD regime; or price level increases in anticipation of future primary deficits, consistent with FD regime and FTPL.

^1/One standard innovation.

View Table

Table 1.3:

Summary of Interpretation, System (3), X = [RIP, PDEF]

	Current Primary Deficit (ΔPDEFt)1/ → Future Real Interest Payments (ΔRIPt+j)
Positive	Government pays down future debt, consistent with MD regime.
Zero	Primary deficit exogenous, consistent with FD regime.
Negative	Government anticipates future interest bill, consistent with MD regime.
	Current real interest payments (ΔRIPt)1/ → future primary deficit (ΔPDEFt+i)
Positive	Unstable policy, consistent with FD regime; or interest rates anticipate future primary deficits.
Zero	Primary deficit exogenous, consistent with FD regime.
Negative	Primary deficit reduction offsets increase in public debt, consistent with MD regime; or price level increases in anticipation of future primary deficits, consistent with FD regime and FTPL.

^1/One standard innovation.

28. Consider first relationships in system (3) that run from the current primary deficit (ΔPDEF_t) to future real interest payments (ΔRIP_t+i). Under a MD regime current innovations to the primary deficit W_PDEFt should be positively related to future government debt and hence interest payments. For example, when the government reduces the primary deficit, it pays down the debt and hence reduces future interest payments (either through the stock of liabilities, the interest rate, or both). However, a negative relationship between W_PDEFt and future RIP is also consistent with a MD regime if the policy makers respond to lower expected future interest payments by borrowing more and hence running higher primary deficits today. By contrast, under a FD regime, W_PDEFt would be uncorrelated with future RIP.⁷

29. Consider next the causality direction running from current real interest payments (ΔRIP_t) to future primary deficits (ΔPDEF_t+i). Like equation (2) a negative relationship may either indicate that primary deficits compensate for changes in real interest payments to help limit debt accumulation (consistent with a MD regime; more plausible under a low inflation environment) or that the price level (and hence real interest payments) anticipate future primary deficits (consistent with a FD regime and the FTPL, more plausible under a high inflation environment). By contrast, a positive relationship indicates that primary deficits respond to real interest payments in an unstable fashion (consistent with a FD regime) or that current interest rates (and hence interest payments) increase in anticipation of perceived future primary deficits (reflecting higher risk). The FTPL interpretation in the literature ignores the possibility of a risk premium on interest rates: when private agents perceive a gap between the value of government debt today and the present value of primary surpluses in the future, they may require higher interest rates in the initial period to hold government debt. This allows for market feedback to government policies. The absence of a relationship suggests that the primary deficit is exogenous, consistent with a FD regime.

30. Exclusion (Granger causality) tests and impulse response functions (IRF’s) were estimated for system (3), with two and four lags, for both the entire 1991–00 period and the selected subsamples. Like the estimates of equation (2), we found little evidence for either the entire sample or the selected subsamples that current ΔRIP helps explain future ΔPDEF, Rather, the null hypothesis that current ΔRIP does not explain future ΔPDEF was never rejected. In addition, all of the corresponding IRF’s were insignificant.

31. However, there was weak evidence that current ΔPDEF helped explain future ΔRIP with the correponding IRF’s negative, particularly for the post-Real period, but the evidence was not very robust since it was highly sensitive to the number of lags used (see regression results in Appendix I, Table A.1.1): i.e., positive current values of W_PDEF imply future decreases in RIP.

32. The rolling 24-month samples were also applied separately to both equations in the VAR system (3). F-statistics for the exclusion of past ΔRIP on current ΔPDEF and past ΔPDEF on current ΔRIP are presented in Figures 1.3 and 1.4, respectively. Under the 24-month window microsscope, we found (see figures) that during certain periods, there were now statistically significant relationships between ΔRIP and ΔPDEF and in both directions, which is not conflicting (see interpretation below).

F-Statistics for Causality Tests of ΔRIP_t on ΔPDEF_t+i, Rolling Two-Year Samples.

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

F-Statistics for Causality Tests of ΔRIP_t on ΔPDEF_t+i, Rolling Two-Year Samples.

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

F-Statistics for Causality Tests of ΔRIP_t on ΔPDEF_t+i, Rolling Two-Year Samples.

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

F-Statistics for Causality Tests of ΔPDEF, on ΔRIP_t+i Rolling Two-Year Samples.

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

F-Statistics for Causality Tests of ΔPDEF, on ΔRIP_t+i Rolling Two-Year Samples.

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

F-Statistics for Causality Tests of ΔPDEF, on ΔRIP_t+i Rolling Two-Year Samples.

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

33. Figure 1.3 renders evidence that confirms the results of the single equation analysis of the previous section: there was a significant negative relationship (especially for the two-lag model) between current ΔRIP on future ΔPDEF during the first years of the Real Plan, from (approximately) the third quarter of 1994 through early 1996. This being the low inflation period, we take this result as supporting the backward-looking interpretation that during the initial months of the Real Plan, government debt was kept under control by reductions in the primary deficit since the real interest bill increased sharply in 1995 (4.8 percent of GDP) from the level observed in 1993 (2.7 percent of GDP)—rejecting the forward-looking alternative of adjustments of the price level in anticipation of future primary deficits (consistent with the FTPL).

34. Likewise, Figure 1.4 (forward-looking tests) also shows a statistically significant negative relationship between past ΔPDEF on current ΔRIP, primarily during 1995. Indeed, between 1996 and 1997, both the primary surplus and the implicit real interest rates on domestic debt fell (see figure below). This finding may represent an implicit rational intertemporal decision by the government: with the implicit real interest rate on public debt falling during 1996–97, the government was not as compelled as it might have otherwise been to either reduce the primary deficit or to quickly implement needed reforms. That is, the government seems to have anticipated the spending of the stabilization dividends in the belief that the Real Plan had brought about a new paradigm, a new era of lower debt costs (possibly due to lower risk premiums) and stricter budget constraints in the near future through the deepening of the structural reform process.

35. The evidence shown above—causality running in both directions—is reconciled by realizing that the primary surplus increased in 1994 (the first year of the Real Plan) in response to a growing real interest bill since 1993 (backward looking adjustment), and fell in 1995 in anticipation of declining future interest payments (forward looking adjustment; in fact real interest payments dropped to 3.4 percent of GDP in 1997 from over 4.8 percent of GDP in 1995). Both findings suggest that from 1994 until 1996 the presence of a FD regime (as specified in Table 1.3) was unlikely.

Brazil: Real Interest Rates

(SELIC; IGP-C)

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Brazil: Real Interest Rates

(SELIC; IGP-C)

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

Brazil: Real Interest Rates

(SELIC; IGP-C)

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

• Download Figure
• Download figure as PowerPoint slide

36. In conclusion, regarding the adjustment between PDEF and RIP, both backward- and forward-looking tests reveal some temporal relationships between these variables during certain periods (primarily the post-Real Plan) but little relationship for the 1991–00 period as a whole. Backward-looking tests reveal that, immediately after the Real Plan, ΔPDEF decreased in response to ΔRIP, in a stabilizing fashion, helping to restrain the debt.

D. Adjustment of Discretionary Expenditures and Revenue

37. The econometric tests performed above (particularly the rolling 24-month windows) seem to point that apart from the initial period of the Real Plan, Brazil lived under a fiscal dominant regime throughout the 1990s, in the sense that fiscal execution, in particular the primary deficit did not adjust to changes in the level of public indebtedness. At this stage, it is valuable to seek evidence whether responses of the PDEF to RIP, if present, occurred through policy decisions involving increases in taxes, cuts in expenditures, or a combination of both. A peculiarity of the Brazilian public finances is that the authorities are constrained in their ability to quickly implement such changes. On the expenditure side, changes in public wages, certain social entitlements, such as pension benefits, and constitutional transfers to the subnational governments are legally mandated, and many times constitutionally predetermined. These three categories represent around three-fourths of the total nonfinancial expenditure of the federal government, and probably an even higher share at the subnational level. For this reason, the room for discretionary expenditure cuts in the short-run is basically circumscribed to a few categories in the so-called other current and capital expenditure item. Thus, in order to seek evidence on how the discretionary elements of the federal budget adjusted, consider the following expression for the real operational deficit:

where G^D is discretionary federal spending, T^F is federal tax revenue, and G^ND contains all other nondiscretionary/mandatory elements of the primary expenditure (such as public sector wages, social security entitlements, and certain constitutionally mandated intragovernmental transfers). Accordingly, a modified VAR system (5) uses the vector X = [RIP, G^ND, G^D, T^F]. Revenue and expenditure data limitations for the subnational governments and the state-owned enterprises will limit this exercise to the accounts of the central government.

38. Exclusion (Granger causality) tests and impulse response functions (IRF’s) (not reported here) were computed for the modified VAR system (5), estimated with two and four lags, for both the entire period and the selected subsamples. Once again, we found weak and unstable evidence (sensitive to the number of lags included) for the causality running from past past ΔRIP to current elements of ΔPDEF.

39. However, relationships between lagged elements of ΔPDEF and current ΔRIP were stronger. For the entire sample, the null hypotheses that past ΔG^ND, ΔG^D and ΔT do not explain current ΔRIP can each be rejected at the 90 percent confidence level or better, and the corresponding IRF’s are significant. Note, however, that significant relationships were concentrated mainly in the immediate post-real period.

40. We found that (IRFs) current innovations in T^F are associated with lower real interest payments in the future, but only after the Real Plan (and not before). This result thus suggests that the Real Plan represented a break in the fiscal regime, since the authorities attempted to limit indebtedness by allowing the tax burden to rise during this period.⁸ Interesting enough, the analysis reveals no relationship between current innovations to discretionary spending G^D and future RIP, as the stochastic process guiding expenditures seems to have been set independently of the evolution of net public debt. Further, current innovations to nondiscretionary expenditure items (G^ND, such as wages, pension benefits, and transfers to subnational governments) are negatively related to future RIP, perhaps reflecting the ex-ante rational intertemporal strategy mentioned above. Thus, the important information content of this last test is that, possibly in anticipation of lower interest payments, the government seems to have loosened the fiscal stance in 1995 by increasing mainly the portion of the primary deficit that contains wages and entitlements, not discretionary spending (OCCs). For instance, expenditure with benefits by the social security system for private sector workers (INSS) increased from 4.1 percent of GDP in 1994 to 5.3 percent of GDP in 1995, with contribution revenues basically held constant, while the discretionary spending component was kept basically constant at around 3.3 percent of GDP from 1994 until 1996. Discretionary spending as a percentage of GDP did however accelerate very rapidly in 1997 and 1998.

E. Sustainability Tests: Conclusions and Policy Implications

41. It is borne out in the data that substantial fiscal adjustment took place at the end of 1998, and especially during 1999 and 2000 (see Figure 1.5). During this period, Brazil’s fiscal policy was sustainable in the sense that its operational deficit appears to be stationary. But a MD regime as defined—a statistically significant relationship between PDEF and RIP—is still found to be absent.

Brazil: Nominal and Primary Deficit (% of GDP: 12-months); 1997M1-2000M9

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

Source: Central Bank, and Fund staff estimates(-) Surplus, (+) Deficit.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Brazil: Nominal and Primary Deficit (% of GDP: 12-months); 1997M1-2000M9

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

Source: Central Bank, and Fund staff estimates(-) Surplus, (+) Deficit.

• Download Figure
• Download figure as PowerPoint slide

Brazil: Nominal and Primary Deficit (% of GDP: 12-months); 1997M1-2000M9

Citation: IMF Staff Country Reports 2001, 010; 10.5089/9781451805901.002.A001

Source: Central Bank, and Fund staff estimates(-) Surplus, (+) Deficit.

• Download Figure
• Download figure as PowerPoint slide

42. Active responses of fiscal policy to variations in real interest payments (backward- and forward-looking) seem to have been concentrated in the period that followed the implementation of the Real Plan. The government allowed the primary surplus to increase in 1994 offsetting the observed increase in real interest payments, but relaxed the fiscal stance in 1995 and 1996 mainly through increases in the nondiscretionary entitlement component, possibly in anticipation of expected lower interest payments in the near future. However, subsequent increases in discretionary spending in 1997 and 1998 shifted the debt dynamics into an unsustainable path until, late in 1998, a strong fiscal adjustment program at all levels of government, in place until this day, placed net pubic debt on a sustainable and declining path again.

43. In June 1999, the central bank announced the adoption of an inflation targeting regime as the main anchor for inflation expectations. It is well known that such a program requires that the monetary authority not be dominated by fiscal financing requirements. Nonetheless, it should be stressed that to empirically find that the presence of a MD regime cannot be confirmed does not necessarily imply that the monetary authority will be unable to pursue an independent monetary policy in an inflation targeting framework. We suspect that more observation showing consistently strong fiscal results, as envisaged under the three-year Fiscal Stabilization Plan, may consolidate the fiscal adjustments in an econometric sense, showing then a perceptible regime switch. However, what an independent monetary policy requires above all is credibility: that is, markets must believe that future adjustments to the primary surplus, if required, will in fact occur.