Debt Sustainability and Procyclical Fiscal Policies in Latin America

Enrique Alberola and José Manuel Montero

II REUNIÓN REDIMA CENTROAMERICAII REUNIÓN REDIMA CENTROAMERICA10 de Noviembre de 200510 de Noviembre de 2005, Santiago de ChileSantiago de Chile

RED DE DIÁLOGO MACROECONÓMICO (R E D I M A)

2

OUTLINE OF THE PRESENTATION

I. MotivationII. Procyclicality of Fiscal PolicyIII.Debt Sustainability and Fiscal StanceIV.Conclusions

3

I. Motivation

Empirical evidence shows that Fiscal policy is procyclical in Latin America

–During expansions (contractions)

•cyclically-adjusted revenues decrease (increase)

•cyclically-adjusted expenditures increase (decrease)

–Gavin and Perotti (1997), Alberola and Molina (2003)

–Widespread phenomenon:

• Talvi and Vegh (2000) show that FP is procyclical in a sample of 20 industrial countries and 36 developing countries. Exception: G-7 countries.

•Kaminsky, Reinhart and Vegh (2004) for a sample of 104 countries.

Destabilizing role of fiscal policy

– introduces an additional source of volatility.

4

I. MotivationPUBLIC SECTOR CREDITWORTHINESS:

–Procyclical FP determined by changes in financing conditions

Procyclical fiscal policy

Loose

Fiscal policyTight

Expansion

ActivityContraction

Loosening

FinancingConstraintsTightening

Improvement

Debt sustainabilit

yDeterioration

Voracity effects

5

I. Motivation

Does the stance of fiscal policy depend on creditworthiness, as measured by debt sustainability perceptions?

Empirical strategy1) Test the procyclicality of FP in LA

•Compute output gap and structural primary balance (SPB)

•Are they correlated?

2) Test how perceptions of credit worthiness, embedded in debt sustainability impinges on fiscal stance

•Derive indicator of debt sustainability : current threshold balance (CTB)

•Estimate relationship between fiscal stance and CTB

6

II. Procyclicality of Fiscal Policy

Three steps:

Derive the output gap

Estimate the structural balance–Fiscal Stance= change of SPB as percentage of GDP

SPB associated with cyclically-adjusted revenuespublic spending

contractionary fiscal stance

Is Fiscal policy procyclical?

Test link SPB changes and output gap

7

II. Procyclicality of Fiscal Policy: Output Gap

OUTPUT GAP

– Production function: OECD, IMF (EC shifting towards it)•Problems: data availability (labour stock, capacity use, K stock), data homogeneity, crises and volatility

– Modified Hodrick-Prescott Filter (Kaiser and Maravall, 1999)

•Pre-adjust the series by removing outliers ==> tackle sharp drops in activity•To overcome accuracy problems in both ends of the series we added forecasts and backcasts (actually, original series for backcasts and Consensus Forecasts for forecasts)

8

II. Procyclicality of FP: Structural Primary Balance

STRUCTURAL PRIMARY BALANCE

Simplified scheme differs from OECD

Cycle sensitive revenues, considered as a whole (no data) Expenditure not depending on cycle (no unemployment benefits) Account for the importance of commodity-related taxes:

OIL: Colombia, Ecuador, Mexico and Venezuela; COPPER: Chile

9

II. Procyclicality of FP: Structural Primary Balance

1.- Revenue elasticities wrt GDP and commodity prices

2.- Structural component of public revenues:

where both Y* and P* are estimated by applying the Modified H-P filter

3.- Structual Primary Balance (SPB, henceforth):

tcommttt pYT logloglog

TtS Tt

Y *tYt

p*tptcomm

tStt GTSPB

10

II. Procyclicality of FP: Structural Primary Balance

Sample:Period: 1980-2004

Countries: Argentina, Brazil, Chile, Colombia, Ecuador, Mexico, Peru, Uruguay and Venezuela

Sources: IMF’s GFS and IFS complemented with national statistics

Caveat: fiscal data is for the central government, except for the public debt, which is for the consolidated government

11

II. Procyclicality of FP: Structural Primary Balance

Table 1: Elasticities of fiscal revenues with respect to real GDP and commodity priceGDP Commodity GDP Commodity

Argentina 1.538 Mexico 0.647 0.109[0.256]*** [0.116]*** [0.042]**

Brazil 1.723 Peru 1.595[0.228]*** [0.208]***

Chile 0.7 1 Uruguay 1.510[0.067]***

Colombia 1.833 0.195 Venezuela 0.153 0.134[0.080]*** [0.039]*** [0.199] [0.064]*

Ecuador 0.522 0.077[0.296]* [0.029]**

Note: Estimated by Dynamic OLS through 1980-2004 with annual real data. Revenues adjusted by outliers.*, **, *** denote statistical significance at 10%, 5% and 1% respectively.

12

II. Procyclicality of FP: Structural Primary Balance

Table 2A: Slope coefficients of change in SPB on output gap SPB calculated using the estimated elasticity of government revenues to GDP

Correlation OLS Correlation OLSArgentina -0.216 -0.090 -0.268 -0.077**Brazil -0.153 -0.259 -0.186 -0.235*Chile 0.214 0.259 0.290 0.243Colombia -0.123 -0.122 -0.193 -0.095Ecuador 0.008 0.003 -0.109 -0.062Mexico -0.140 -0.090 -0.017 0.008Peru -0.386 -0.248*** -0.410 -0.194***Uruguay -0.531 -0.327*** -0.517 -0.212***Venezuela -0.302 -0.231 -0.283 -0.177OLS estimation, robust standard errors. Pairwise correlations.*, **, *** denote statistical significance at 10%, 5% and 1%, respectively, for OLS estimates

lambda=6.7 lambda=100

13

II. Procyclicality of FP: SPB vs Output Gap

y = 0.21x + 0.0017R2 = 0.1212

y = -0.1608x + 0.0003R2 = 0.0403

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

-0.15 -0.1 -0.05 0 0.05 0.1 0.15

Latin America USA

DSPB

OUTPUT GAP

14

II. Procyclicality of FP: SPB vs Output Gap

In a more formal test, procyclicality of FP in LA is confirmed

Table 3: Panel data estimation of procyclicality of fiscal policy in LA Dependent variable: change in structural primary balanceLambda=6.7 and SPB calculated with the estimated revenue elasticity to GDP

FE RE FE FE1981-2004 1981-2004 1981-1990 1991-2004

constant 0.0007 0.0007 0.004 -0.001[0.002] [0.002] [0.004] [0.002]

output gap -0.143 -0.141 -0.107 -0.181[0.053]*** [0.051]*** [0.094] [0.059]***

R2 0.035 0.035 0.018 0.073Observations 209 209 84 125No. Of countries 9 9 9 9Standard errors in brackets. *, **, *** denote statistical significance at 10%, 5% and 1% respectivelyFE: fixed-effects estimator; RE: random effects estimator

15

III. Debt sustainability and Fiscal Stance

Why is fiscal policy procyclical? We focus on sustainability concerns

impinge on perception of credit worthines are reflected on financial indicators

Financial indicators coupled with debt level determine debt dynamicsDebt dynamics signal debt sustainability and therefore concernsThe level of debt is binding through the cycle if

High enoughFinancing conditions are volatile and

closely related to the cycle affect debt sustainability concerns

Putting these intuitions into testing:1.- Derive an indicator of fiscal sustainability: current threshold balance2.- CTB v. Fiscal Stance

16

III. Debt sustainability and Fiscal Stance: CTB

1. CURRENT THRESHOLD BALANCE (CTB) Starting point: government’s budget constraint (%GDP)

Simplifying assumptions due to data availability: contingent liabilities, types of debt, seignorage

Hence, current threshold balance = debt-stabilising primary balance (D=0)

Note: no distinction internal/external debt, equal cost We use a measure of implicit real interest rate derived from dividing

interest payments over the stock of debt, both as %GDP Overcome problems linked to that simplification

1

*

1 )1()1(

)()1()(

tt

ttttt

t

ttttt D

ggei

Dggi

PBD

1)1()(

tt

tttt D

ggi

CTB

17

III. Debt sustainability and Fiscal Stance: CTB

Caution: valuation effects, contingent liabilities, government definition

BRAZIL

-0.06

-0.03

0

0.03

0.06

0.09

1991 1993 1995 1997 1999 2001 2003

CTB PB D(debt)

18

III. Debt sustainability and Fiscal Stance: SPB vs CTB

2. DOES THE FISCAL STANCE DEPEND ON DEBT SUSTAINABILITY CONCERNS?

Empirical analysis framed within the following regression:

CTB expected positive sign– Higher required primary balance triggers fiscal contraction (SPB)

FP reaction to sustainability concerns is expected to be a function of the sustainability problem itself

– Larger gap, more impact on changes in SPB expected negative sign– More evident impact of CTB on SPB

Controls: inflation, terms of trade, output gap, years-in-default dummies

Econometric issues: endogeneity and fixed effectsIV estimation methods

itititititiit uCONTROLSCTBPBCTBSPB 4113210 )(

11 itit CTBPB

19

III. Debt sustainability and Fiscal Stance: SPB vs CTBBaseline results: GMM difference estimator

Table 4B: Panel data estimation of financial restrictions effects on fiscal policy in LA Dependent variable: D(SPB) =change in structural primary balanceSample:1981-2004

(1) (2) (3) (4)D(CTB) 0.239 0.442 0.434 0.450

[0.108]* [0.143]** [0.134]** [0.171]**PB(-1)-CTB(-1) -0.416 -0.385 -0.368

[0.102]*** [0.109]*** [0.114]**GAP -0.094 -0.089

[0.074] [0.073]D(inflation) 0.0004

[0.0002]*Dlog(TOT) 0.016

[0.024]Observations 158 158 158 158No. Of countries 9 9 9 9AR(1) (p-value) 0.03 0.02 0.02 0.02AR(2) (p-value) 0.21 0.07 0.07 0.19Sargan-Hansen Test (p-value) 0.99 0.99 0.99 1.00Standard errors in brackets. *, **, *** denote statistical significance at 10%, 5% and 1% respectivelyGMM: GMM difference estimator (See Arellano and Bond, 1991). PB(-2)-CTB(-2) used as only instrument.All regressions include dummies that account for the periods in which any country was declared to be in default by Standard and Poor's. These dummies turned out to be negative, though non-significant.

Panel regression, GMM difference estimator

20

III. Debt sustainability and Fiscal Stance: SPB vs CTB

Table 4C: Panel data estimation of financial restrictions effects on fiscal policy in LA Dependent variable: D(SPB) =change in structural primary balanceSample:1981-2004

(1) (2) (3) (4)D(CTB) 0.370 0.536 0.510 0.504

[0.131]*** [0.129]*** [0.181]*** [0.185]***PB(-1)-CTB(-1) -0.408 -0.397 -0.359

[0.140]*** [0.185]** [0.195]*GAP -0.067 -0.080

[0.125] [0.120]D(inflation) 0.0005

[0.0003]Dlog(TOT) 0.017

[0.016]Observations 158 158 158 158No. Of countries 9 9 9 9Sargan-Hansen Test (p-value) 0.20 0.50 0.23 0.36Robust standard errors in brackets. *, **, *** denote statistical significance at 10%, 5% and 1% respectivelyIV: IV estimator for the differenced equation. D(CTB) and PB(-1)-CTB(-1) instrumented with SPB(-2), CTB(-2), PB(-2)-CTB(-2) and GAP(-1). Sargan-Hansen Tests of overidentification restrictions.All regressions include dummies that account for the periods in which any country was declared to be in default by Standard and Poor's. These dummies turned out to be negative, though non-significant.

Panel regression, IV estimation

Robustness: estimation method: “simple” IV

21

III. Debt sustainability and Fiscal Stance: SPB vs CTBRobustness: shorter sample: 1991-2004

Table 6B: Panel data estimation of financial restrictions effects on fiscal policy in LA Dependent variable: D(SPB) =change in structural primary balanceSample:1991-2004

(1) (2) (3) (4)D(CTB) 0.238 0.407 0.389 0.375

[0.108]* [0.157]** [0.148]** [0.179]*PB(-1)-CTB(-1) -0.361 -0.330 -0.259

[0.122]** [0.129]** [0.140]*GAP -0.086 -0.130

[0.052] [0.071]*D(inflation) 0.0007

[0.0003]**Dlog(TOT) -0.023

[0.035]Observations 123 123 123 123No. Of countries 9 9 9 9AR(1) (p-value) 0.04 0.02 0.02 0.01AR(2) (p-value) 0.96 0.65 0.64 0.73Sargan-Hansen Test (p-value) 0.89 0.81 0.89 1.00Standard errors in brackets. *, **, *** denote statistical significance at 10%, 5% and 1% respectivelyGMM: GMM difference estimator (See Arellano and Bond, 1991). PB(-2)-CTB(-2) used as only instrument.All regressions include dummies that account for the periods in which any country was declared to be in default by Standard and Poor's. These dummies turned out to be negative, though non-significant.

Panel regression, GMM difference estimator

22

IV. Conclusions

We have presented robust evidence on

Fiscal policy in Latin America is procyclical

This procyclicality is shown to be related to the evolution of fiscal sustainability

–Deterioration of fiscal sustainability indicator, as measured by the CTB, leads to a fiscal tightening

–The tightening is larger the worse is the level of debt sustainability, as measured by the ECM term

–Once we control for debt sustainability fiscal policy is neutralExtensions

–Normative consequences. Is fiscal policy adequate?•Long term values of debt dynamics determinants

–Why not doing it for CA + R.D?