Uruguay
Selected Issues
Author:
International Monetary Fund
Search for other papers by International Monetary Fund in
Current site
Google Scholar Close
Publication Date:
01 Feb 2008
eISBN:
9781451839418
Language:
English
Keywords:
ISCR; CR; inflation expectation; inflation process; inflation dynamics; inflation inertia; inflation coefficient; oil prices inflation; Inflation; Inflation targeting; Real effective exchange rates; Fiscal stance; Exchange rate flexibility
Download PDF (842.1 KB)

This paper examines the inflation expectations, monetary policy credibility, and dollarization. Country fundamentals have explained variation in sovereign spreads, but external factors play an important role. This paper assesses the role of and prospects for bank-lending from a cyclical and structural perspectives. A model calibrated for Uruguay, a financially dollarized economy, suggests that reserves are nearing optimal prudential levels. The results of a modified Merton framework, applied to the case of the Uruguayan banking system, appear to be promising for countries without equity markets.

Abstract

This paper examines the inflation expectations, monetary policy credibility, and dollarization. Country fundamentals have explained variation in sovereign spreads, but external factors play an important role. This paper assesses the role of and prospects for bank-lending from a cyclical and structural perspectives. A model calibrated for Uruguay, a financially dollarized economy, suggests that reserves are nearing optimal prudential levels. The results of a modified Merton framework, applied to the case of the Uruguayan banking system, appear to be promising for countries without equity markets.

I. The Inflation Process in Uruguay

By Gaston Gelos and Fernanda Rossi Iriondo

A. Introduction

1. Following chronic high inflation, Uruguay reduced inflation gradually through a disinflation program in the 1990s, reaching single digits by 1998. Inflation has since remained moderate, except for an outburst during 2002 financial crisis and devaluation. In that year, the country moved to a flexible exchange-rate regime and inflation began declining quickly to low single digits by 2005. It has since increased somewhat to around 6–8 percent.

2. With relatively high inflation vis-à-vis the region and other emerging markets, a relevant question is how costly it would be to reduce it further. Various aspects of the factors driving inflation in Uruguay remain to be understood better. First, what is the degree of inflation persistence and to what extent is inflation driven by backward-looking behavior and expectations? Understanding this is important for the conduct of monetary policy-with high backward-looking behavior, temporary inflationary shocks tend to persist and the output costs of reducing inflation quickly are higher. Second, what are the determinants of inflation expectations? This question is relevant given the monetary authorities' increased focus on the inflation objective, de-emphasizing monetary targets. In such a context, anchoring expectations is a precondition for a successful monetary policy.

3. This paper explores the dynamics of inflation in Uruguay, making use of survey data, focusing on the role of inflation expectations and their determinants. First, it assesses the role of inflation expectations in shaping the inflation process by estimating a New-Keynesian Phillips curve. Second, it explores the determinants of survey expectations. Third, it discusses the rationality of survey expectations.

4. The paper finds that expectations are more important than lagged inflation in shaping the inflation process, suggesting that inertia is relatively small. The importance of lagged inflation decreased with the adoption of a flexible exchange rate and has remained low since then. The results also suggest that, in addition to marginal cost proxies, fiscal variables are important in explaining expectations. There is some evidence that the central bank has gained credibility, with announced inflation targets playing an increasing role in forming expectations. Survey expectations over a 12-month horizon are neither unbiased nor efficient, but 1-month-ahead inflation expectations appear to be rational.

B. Inflation Dynamics and Inflation Expectations

5. To assess the importance of inflation expectations, a structural price-setting model is used nesting two types of price-setting behavior (see Gali and Gertler, 1999, Celasun, Gelos, and Prati, 2004a, b, and Celasun and McGettigan, 2005). Backward-looking price setters update the average new price based on the most recently observed inflation rate. Forward-looking price setters set prices based on their expectations, which may or not be rational. The resulting inflation rate in period t equals:

( 1 ) π t = α + δ π t − 1 + ( 1 − δ ) E t π t + 1 + φ m c t + ε t

where π is the 12-month (or monthly) CPI inflation rate, Etπt+1 is the expected inflation rate in 12 (or 1) month(s), πt-1 refers to inflation between 24 and 12 months ago (or between 2 and 1 months ago), and mc are marginal costs.

6. In the empirical work, annual inflation expectations from consensus forecasts is used. For comparison purposes, a shorter sample of monthly inflation expectations from a survey conducted since 2004 by the Uruguayan Central Bank is also used (Figure 1).1,2 Since a typical firm in a small open economy is likely to use imported intermediate goods and domestic labor as inputs in production, a combination of the real effective exchange rate and domestic real unit labor costs (both in deviations from a linear trend) is used as a proxy of real marginal costs. Data for the period 1998:01-2006:09 is used. To address potential endogeneity problems, the equation is estimated by GMM. The equation is also estimated on a monthly forecast horizon for inflation forecasts for 2004:01-2006:09 from a Central Bank of Uruguay survey.3

Figure 1.
Figure 1.

Actual and 12-months ahead expected inlfation

Citation: IMF Staff Country Reports 2008, 046; 10.5089/9781451839418.002.A001

7. The results indicate that inflation expectations drive the inflation process (Table 1). In the case of annual inflation, the coefficient on expected future inflation is close to one, while the coefficient on lagged inflation is insignificant in all but one case, where it is slightly negative. The real effective exchange rate enters the estimation with a statistically significant positive coefficient, in line with the notion that foreign prices are important in determining prices in Uruguay. Using monthly data, the role of expectations is even stronger. The coefficient on expected future inflation is larger than one, while the coefficient on lagged inflation is negative and smaller. Proxy marginal costs enter the estimation significantly with positive coefficients, but a much smaller fraction of the monthly variation in inflation can be explained compared to the case of annual data. Table 1 shows estimations with the coefficients on lagged and expected inflation constrained to sum one, as suggested by theory. The key results do not change when estimating the equations without restrictions.

Table 1.

CPI Inflation Regressions with Survey Data

article image
Notes: GMM estimates. Andrews bandwidth standard errors are given in parentheses. For 12-months forecast horizon the instruments for lagged inflation were inflation(-2) in equations 1 and 4; 12-month exchange rate change(-2) in equation 3 and lagged inflation in equation 2. Oil prices inflation(-1) was used as instrument for inflation forecast. The other regressors were used as instruments in the remaining equations. For 1-month forecast horizon, the equation 1 includes inflation forecast (-1) as instrument for inflation forecast. The other regressors were used as instruments in this and the remaining equations. *, **, *** indicate significance at the 10, 5 and 1 percent level, respectively.

8. Estimations allowing for the possibility of nonstationarity confirm the key results. With relatively short time series, the issue of possible nonstationarity is difficult to address given the low power of standard tests. Tests following Clemente, Montañés, and Reyes (1998) suggest the presence of a unit root even after allowing for two structural breaks (not shown). Thus, we also estimate a system using the FIML (full-information maximum-likelihood) method by Johansen and Swensen (1999), confirming a stationary relation between inflation, expected inflation and proxy variables of marginal costs. In line with the GMM results, the estimated coefficient on expected inflation was above one and significant. The proxies for marginal costs were positive but not statistically significant (not shown).

Recursive estimates

9. A recursive estimation shows that the importance of lagged inflation has declined. Indeed, it falls sharply after the floating of the exchange rate, stabilizing after mid-2003 (Figure 2).

Figure 2.
Figure 2.

Recursive coefficient equation (1)

(Recursive ordinary least squares estimates)

Citation: IMF Staff Country Reports 2008, 046; 10.5089/9781451839418.002.A001

Inflation persistence

10. The role of lagged inflation is limited, even after accounting for its impact on expectations and wage formation. If inflation expectations were largely driven by past inflation, the coefficient on lagged inflation would underestimate the overall role of past inflation in shaping the inflation process. To explore this issue, equation (1) is re-estimated, replacing expected inflation by the residual of a regression of expected inflation on lagged inflation, finding that the coefficient does not change substantially. Similarly, backward-looking wage indexation may induce inertia through the marginal cost variable, thus leading to underestimate the importance of lagged inflation. Therefore, the residuals of a regression of unit labor costs on lagged inflation as marginal cost proxy are also used, without altering the qualitative results. All these factors point to low inflation inertia.

C. Determinants of Inflation Expectations

11. What are the factors explaining inflation expectations? Since inflation expectations appear to be a key driving force of Uruguayan inflation dynamics, anchoring expectations is important for monetary policy. Following Celasun, Gelos, and Prati (2004), it is key to investigate the role of past inflation, fiscal outcomes, the exchange rate, monetary policy variables, and real wages on inflation expectations by estimating variants of the following general model:

( 2 ) π t e = α 0 + α 1 π t − 1 + α 2 p b t − 1 + α 3 m c t − 1 + α 4 m o n e t a r y var i a b l e s t − 1 + u t

Where pbt−1 is the ratio of the consolidated primary fiscal balance to GDP and the monetary variables include year-on-year M1 growth or alternatively exchange rate change (see Licandro and Vicente (2006), for a discussion of the link between fiscal policy and inflation objectives in Uruguay). It is also important to examine whether there has been a structural break in the post-crisis expectations formation using a dummy variable. Equation (2) is estimated by a generalized method of moments in order to address potential endogeneity problems, using both 12-months and 1-month-ahead horizons.

12. The results suggest that lagged inflation, the primary balance, and marginal costs proxies explain expected inflation (Table 2 and Figure 3). Monetary aggregates only become important after the adoption of a flexible exchange rate regime and do not play a role in shorter horizon forecasts. Changes in the nominal exchange rate affect expectations, but the effect is modest and not significant for 1-month-ahead expectations. Lagged inflation contributes 50–60 percent to explaining expected inflation; although for 1-month expectations the contribution is significantly smaller. For the 12-month ahead estimation, recursive coefficient estimates on lagged inflation stabilize around 0.6 after the adoption of a flexible exchange rate in 2002 (Figure 4).

Table 2.

Determinants of Inflation Expectations

article image
Notes for GMM estimates: Andrews bandwidth standard errors are given in parenthesis. Instruments include inflation(-2), inflation(-3), primary balance/GDP(-3), real wage gap(-2), real wage gap(-3), money growth rate (-2), money growth rate(-3), DC and DP. Estimate (5) and (6) excludes DC and DP as instruments, where DC and DP are dummy variables capturing two different effects of 2002 crisis. DC=1 in 2002:07-2003:2 period, thus it captures a transitory increase of inflationary expectations due to the crisis. DP=1 from 2003:3 onwards, thereby it captures some structural changes after the crisis. The estimation results showed that only lagged inflation coefficients have experienced a structural change due to the 2002 crisis. *, **, *** indicate significance at the 10, 5 and 1 percent level, respectively. Notes for C.E. (Cointegration Equation) estimates: standard errors are given in parenthesis.
Figure 3.
Figure 3.

Contributions to Total Explained Variation of Expected Inflation

Citation: IMF Staff Country Reports 2008, 046; 10.5089/9781451839418.002.A001

Figure 4.
Figure 4.

Determinants of Inflationary Expectations: recursive coefficients

(Recursive ordinary least squares estimates on an annual horizon)

Citation: IMF Staff Country Reports 2008, 046; 10.5089/9781451839418.002.A001

13. The primary balance plays a role in determining expected inflation. It enters with a negative coefficient in almost all the estimated equations (see Celasun, Gelos, and Prati, 2004a, b, and Cerisola and Gelos, forthcoming) and, for estimations with a 12-months ahead horizon, the coefficient increased slightly since 2003. The quantitative significance of this factor is, however, moderate: a one percent increase in the primary balance as a percent of GDP is estimated to yield a fall in inflation expectations by about 0.5 percentage points. For a shorter forecast horizon, it is also significant but less important. The model tracks inflation expectations quite well, including during the crisis period (Figure 5).

Figure 5.
Figure 5.

12-months ahead Expected Inflation: Actual Versus Fitted (GMM-based Model)

Citation: IMF Staff Country Reports 2008, 046; 10.5089/9781451839418.002.A001

Source: Author's calculations.

14. There is some evidence that announced inflation targets have gained credibility. Inference is limited given that the paper works with data starting January 2004, using the mid-point of the 12-month-ahead inflation target range. The coefficient on the inflation target is significant, and on lagged inflation insignificant. Recursive (OLS) estimates show that the coefficient has been approaching one during 2006 (Figure 6). These results suggest that forecasters are forward-looking and increasingly anchoring expectations around the central bank targets.1 The notion of a recent increase in central bank's credibility is consistent with the evolution of the dispersion of inflation forecasts. In particular, the coefficient of variation for survey expectations show no clear trend until early 2006 and a decline since then (Figure 7). However, this result needs to be taken with caution. As discussed in the next chapter (López-Mejía, Rebucci, and Saizar (2007), announced targets for inflation and indicative ceilings for M1, combined with intervention in the foreign exchange market partly to resist nominal appreciation, have created ambiguity about the objectives and instruments of monetary policy, thus undermining monetary policy credibility.

Figure 6.
Figure 6.

Recursive Estimates of Coefficient on Inflation Target

(Recursive OLS estimate)

Citation: IMF Staff Country Reports 2008, 046; 10.5089/9781451839418.002.A001

Figure 7.
Figure 7.

The Dispersion of Inflation Forecasts

(coefficient of variation)

Citation: IMF Staff Country Reports 2008, 046; 10.5089/9781451839418.002.A001

15. Estimations taking into account possible nonstationarity reinforce the results. A Johansen cointegration test finds a stable relationship between expected inflation, lagged inflation, primary balance, real effective exchange rate, and real wage gap; a vector error correction model suggests that the 12-months ahead inflation forecasts are determined by lagged inflation, primary balance and the marginal costs proxies. The estimated coefficients have the same signs and only slightly higher magnitudes than the GMM estimates.

Rationality

16. Are expectations rational? To explore this issue, the paper first examines unbiasedness, regressing actual inflation on both an intercept and expected inflation:

( 3 ) Π t = c + β Π t e + ε t

If expectations are unbiased, c should be zero and β one. This is the joint hypothesis tested in the paper (Table 3).

Table 3.

Tests of Unbiasedness of Inflation Forecasts

article image

17. The central bank of Uruguay inflation forecasts appear unbiased for 1-month ahead, but biased for 12-month ahead. However, the tests should be interpreted with caution given the potential nonstationarity of the series. Generally, the estimated coefficient of an I(1) variable is not normally distributed, invalidating the usual tests. In any case, however, the paper would reject unbiasedness for the 12-month ahead inflation expectations.2,3

18. Allowing for the possibility of nonstationarity, the results confirm that expectations do not use all available information. Johansen (1991, 1995) tests find a cointegration relation between actual and expected inflation (Appendix Table A1). The speed at which expectations are revised over time in light of new information is estimated through a bivariate VEC model with inflation and expected inflation (Appendix Table A2). If the parameter on the expectation error in this regression is significant, then consumers revise their expectations and adjust them toward the fully rational outcome. The estimated coefficients are negative and significant for both the 12-months ahead and 1-month ahead forecasts, implying low persistence in deviation between actual and expected inflation. However, regressions of the forecast errors on different variables indicate that 12-month-ahead forecasts fail to extract all possible information from all the relevant variables,4 except for past information on the real effective exchange rate gap. On the other hand, 1-month ahead forecasts seem efficient in the use of past information, except for the informational content of actual inflation (Appendix Table A3).

D. Conclusions

19. This paper examined the role of inflation expectations in shaping inflation dynamics, the degree of inflation persistence, and the rationality and factors driving inflation expectations. The key results are:

  • Inflation expectations are more important than past inflation in shaping the inflation process. As a result, inflation inertia is not high. However, recent changes in wage agreement mechanisms containing backward-looking elements may lead to an increase in inertia.

  • Moreover, the role of lagged inflation in shaping the inflation process has been declining since 2002.

  • Confirming previous studies' results, fiscal variables play a role in driving inflation expectations, with a higher primary balance associated with lower expected inflation.

  • The evidence suggests that the central bank has gained credibility since the adoption of a flexible exchange rate and the announcement of inflation targets. Forecasters increasingly take the central bank's targets into account when forming expectations.

  • 12-month-ahead inflation expectations are neither unbiased nor efficient, while 1-month ahead inflation surveys are.

20. For policy purposes, low inertia implies that responding to inflationary shocks entails more limited output losses. The results also underscore the role of consistent macroeconomic policies for anchoring inflation expectations, with fiscal credibility playing an important role. The increase in credibility of the central bank since the move to a flexible exchange rate regime bodes well for a gradual move to a full-fledged inflation targeting regime. As discussed in the next chapter, further increasing credibility will require increasing the clarity of objectives and procedures of monetary policy. Still, the evidence on the rationality of the forecasts is mixed. This suggests that further efforts in enhancing policy communications may help agents improve their forecasts and assist the authorities in anchoring expectations.

References

  • Bebby Mark, Stephan Hall, and Henry Brian, 2001, “Rational Expectations and Near Rational Alternatives: How Best to Form Expectations,” European Central Bank Working paper No. 86.

    • Search Google Scholar
    • Export Citation

  • Barkbu, Bergljot and Nicoletta Batini, 2005, “The New Keynesian Phillips Curve When Inflation is Non-Stationary: The Case for Canada.”

    • Search Google Scholar
    • Export Citation

  • Barkbu, Bergljot, Nicoletta Batini, and Roberto Garcia-Saltos, 2006, “Inflation Persistence in Latin America,” Prepared for the XI Meetings of LACEA, Mexico D.F., November 2006.

    • Search Google Scholar
    • Export Citation

  • Celasun, Oya, and Donal McGettigan, 2005, “Turkey's Inflation Process,” in Turkey at the Crossroads—From Crisis Resolution to EU Accession, IMF Occasional Paper No. 242 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Celasun, Oya, Gaston Gelos, and Alessandro Prati, 2004a, “Would Cold Turkey Work in Turkey?” Staff Papers, International Monetary Fund, Vol. 51, No. 3.

    • Search Google Scholar
    • Export Citation

  • Celasun, Oya, Gaston Gelos, and Alessandro Prati, 2004b, “Obstacles to Disinflation: What is the Role of Fiscal Expectations?” Economic Policy, Vol. 19 (October), No. 40, pp. 441–81.

    • Search Google Scholar
    • Export Citation

  • Cerisola, Martin, and Gaston Gelos, 2005, “What Drives Inflation Expectations in Brazil? An Empirical Analysis,” Applied Economics (forthcoming).

    • Search Google Scholar
    • Export Citation

  • Clemente, J., A. Montañés, and M. Reyes, 1998, “Testing for Unit Roots in Variables with a Double Change in the Mean,” Economics Letters, Vol. 59, 175–182

    • Search Google Scholar
    • Export Citation

  • Consensus Economics, various issues

  • De Carvalho, Fabia, and Maurício Bugarin, 2006, “Inflation Expectations in Latin America,” Economia, Vol. 6, No. 2, pp. 101–45.

  • Forsells, Magnus, and Geoff Kenny, 2002, “The Rationality of Consumers' Inflation Expectations: Survey-based Evidence for the Euro rea,” European Central Bank Working paper series, Working paper No. 163.

    • Search Google Scholar
    • Export Citation

  • Fildes, Robert, and Herman Stekler, 2002, “The State of Macroeconomic Forecasting,” Journal of Macroeconomics, No.24, pp. 435–68.

  • Galí, Jordi, and Mark Gertler, 1999, “Inflation Dynamics: A Structural Econometric Analysis,” Journal of Monetary Economics, Vol. 44, Issue 2, pp. 195–222.

    • Search Google Scholar
    • Export Citation

  • Gertler, Mark, 2004, Discussion of “Structural Inflation Modeling Papers,” presented at a conference on “Inflation Persistence in the Euro Area” organized by the European Central Bank, Frankfurt am Main, 10–11 December 2004.

    • Search Google Scholar
    • Export Citation

  • Licandro, Gerardo, and Leonardo Vicente, 2006, “Incentivos fiscales e inconsistencia temporal: Uruguay 1970–2005” (unpublished manuscript: Central Bank of Uruguay).

    • Search Google Scholar
    • Export Citation

  • Minella, A., P. Springer de Freitas, I. Goldfajn, and M.K. Muinhos, 2003, “Inflation Targeting in Brazil: Constructing Credibility under Exchange Rate Volatility,” Journal of International Money and Finance, Vol. 22, pp. 1015–40.

    • Search Google Scholar
    • Export Citation

  • Paquet, Alain, 1992, “Inflationary Expectations and Rationality,” Economics Letters, No 40 (September), pp. 303–08.

  • Roberts, J.M., 1995, “New Keynesian Economics and the Phillips Curve,” Journal of Money, Credit, and Banking, Vol. 27 (4), pp. 975–84.

    • Search Google Scholar
    • Export Citation

  • Roberts, J.M., 1997, “Is Inflation Sticky?” Journal of Monetary Economics, Vol. 39, pp. 173–96

  • Yeh, Andrew Jia-Yuh, 2006, “Are Inflation Expectations Rational in New Zealand?” Reserve Bank of New Zealand, Working Paper series.

Appendix: Co-integration Tests, Bivariate VEC Model, and Tests of Efficiency of Inflation Forecasts

Table A1.

Johansen Co-integration Test Results

article image
p-values for the null: none cointegrating equations are reported in parenthesis. 2 lags were used as indicated by Schwartz criterion.
Table A2.

Adjustment Coefficients in a Bivariate VEC on Actual and Expected Inflation

article image
Table A3.

Tests of Efficiency of Inflation Forecasts

article image
1

One year-ahead inflation expectations are calculated as a weighted average of this and next year's inflation forecast. Data are available on a monthly basis starting in 2001 and bimonthly prior to that. Data is interpolated for the missing months.

2

Roberts (1995) and Roberts (1997) pioneered the use of survey expectations in estimating Phillips curves and inflation dynamics in the U.S. Use in emerging markets is not widespread.

3

The monthly survey conducted by the Central Bank is available since 2004.

1

See Cerisola and Gelos (forthcoming) and Minella (2003) for similar results on the relevance of inflation targets for the case of Brazil.

2

See similar results for other Latin American countries in Carvalho and Bugarin (2006).

3

The results are not an artifact from the approximation for the 12-month forecasts as unbiasedness is also rejected for much shorter subperiods.

4

This result also holds for the shorter 12-month survey available from the central bank.

Same Series

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Staff Country Reports
Uruguay: Selected Issues
Author:
International Monetary Fund
Volume/Issue:
Volume 2008: Issue 046
Publisher:
International Monetary Fund
ISBN:
9781451839418
ISSN:
1934-7685
Pages:
78
DOI:
https://doi.org/10.5089/9781451839418.002
  • Figure 1.
    View in gallery
    Figure 1.

    Actual and 12-months ahead expected inlfation

  • Figure 2.
    View in gallery
    Figure 2.

    Recursive coefficient equation (1)

    (Recursive ordinary least squares estimates)

  • Figure 3.
    View in gallery
    Figure 3.

    Contributions to Total Explained Variation of Expected Inflation

  • Figure 4.
    View in gallery
    Figure 4.

    Determinants of Inflationary Expectations: recursive coefficients

    (Recursive ordinary least squares estimates on an annual horizon)

  • Figure 5.
    View in gallery
    Figure 5.

    12-months ahead Expected Inflation: Actual Versus Fitted (GMM-based Model)

  • Figure 6.
    View in gallery
    Figure 6.

    Recursive Estimates of Coefficient on Inflation Target

    (Recursive OLS estimate)

  • Figure 7.
    View in gallery
    Figure 7.

    The Dispersion of Inflation Forecasts

    (coefficient of variation)

Figure 1.

Actual and 12-months ahead expected inlfation
Figure 2.

Recursive coefficient equation (1)

(Recursive ordinary least squares estimates)
Figure 3.

Contributions to Total Explained Variation of Expected Inflation
Figure 4.

Determinants of Inflationary Expectations: recursive coefficients

(Recursive ordinary least squares estimates on an annual horizon)
Figure 5.

12-months ahead Expected Inflation: Actual Versus Fitted (GMM-based Model)
Figure 6.

Recursive Estimates of Coefficient on Inflation Target

(Recursive OLS estimate)
Figure 7.

The Dispersion of Inflation Forecasts

(coefficient of variation)