Abstract
This Selected Issues paper for Peru shows that during the years of strong growth and high commodity prices, the Peruvian authorities have conducted a prudent fiscal policy, maintaining a broadly neutral fiscal stance. During 2004–08, while the revenue-to-GDP ratio increased 3.7 percentage points, the expenditure ratio rose only 0.9 percentage points. Expenditure control focused on current spending and coincided with increasing government investment aimed at enhancing public access to infrastructure and social services. Fiscal policy has also outperformed budgets approved by congress, owing to higher-than-anticipated revenue, as well as the need to limit inflation pressures.
II. Disentangling the Motives for Foreign Exchange Intervention in Peru4
6. Since the inception of its inflation-targeting framework (ITF) in 2003, Peru’s central bank has frequently intervened in the foreign exchange market. While ITFs are often associated with the implementation of flexible exchange rate regimes, Peru’s ITF has been designed to allow the central bank to respond to the risks posed by the still high financial dollarization, including those generated by drastic exchange rate swings on the economy’s balance sheets. In particular, intervention policy has sought both “…the moderation of excessive exchange rate volatility […] and maintaining a high level of central bank international reserves (NIR)” (Armas and Grippa, 2006, pp. 4), including against the backdrop of high and sustained capital inflows recorded in the last few years.
7. The central bank’s focus on exchange rate smoothing has been documented empirically. In particular, Humala and Rodriguez (2008) have emphasized the presence of a smoothing motive for the period 1994–2007, with their evidence suggesting that foreign exchange intervention in Peru has been induced both by a volatility-reducing goal as well as an intention to moderate the deviations of the exchange rate from its trend.5
Foreign Exchange Intervention and Nominal Exchange Rate
Citation: IMF Staff Country Reports 2009, 041; 10.5089/9781451831184.002.A002
Source: Banco Central de la Republica del Peru.8. This paper looks further into the central bank’s intervention policy from an empirical perspective. It seeks to differentiate between the authorities’ intention to smooth exchange rate movements from the explicit intention to accumulate reserves purely due to precautionary reasons. In particular, the paper studies the evolution in the relative weight of such motives in the intervention function since the inception of the inflation targeting regime and at detail over the past 18 months.
9. The findings suggest that while a smoothing motive for intervention predominated in the last few years, the importance of the precautionary motive has been increasing. Peru’s inflation targeting period has been characterized by large and significant parameters indicating a strong tendency to moderate exchange rate fluctuations, although this has declined over the last year. In contrast, the expected impact of the precautionary motive on the intervention function has gained strength—reflecting first the authorities’ intention to accumulate a reserve buffer in the context of large capital inflows, and later their use of such buffer as global uncertainties heightened.
10. The paper presents several estimations of the central bank’s implicit foreign exchange intervention function. The model is estimated with daily data between January 2001 to late September 2008, building over a widely known specification of the intervention function (see Appendix). This allows to differentiate empirically for the importance of the smoothing motive from the precautionary motive in the central bank’s intervention decisions:
The estimation includes three simultaneous definitions of the smoothing motive: first, limiting the speed of the exchange rate change from day to day; second, moderating the exchange rate deviation from a trend (given by the 90-day moving average), and third, reducing exchange rate volatility (measured as the GARCH-based conditional variance).6
The precautionary motive is measured as the gap between the level of the net international reserves relative to an implicit “optimal” reserve benchmark. Such a benchmark is assumed to equal the total of public external debt plus dollarized deposits in the previous quarter (Jeanne and Ranciere, 2006; Batini and Peschiera, 2007).7
A measure on inflation news is also introduced, to control econometrically for foreign exchange intervention aimed at supporting the inflation target.
11. The exercise relies on standard estimation techniques. In particular, the model is estimated through ordinary least-squares (Galati and Disyatat, 2007) to gauge the impact of the potential motives over the authorities’ intervention in the foreign exchange market (see Appendix for a description of the expected signs on estimated coefficients).8
A. A Look at the Motives for Intervention Over the Broader Horizon
12. A first estimation gauges the parameters of the intervention function over the inflation targeting period. Specifically, the model is estimated for each year of the full-fledged inflation targeting framework in 2004–08 as well as the transition year (2003) to test for structural changes in the parameters of the intervention function.9
13. The results suggest that:
The smoothing motive has generally dominated the intervention function over the whole inflation targeting period. The central bank has generally attempted to moderate large deviations of the exchange rate from its 90-day trend during the inflation targeting period, although in 2007 it has also focused on mitigating the acceleration of the exchange rate and exchange rate volatility (Figure 2, charts 1–3). Results also hint at a reduced importance of the smoothing motive in 2008, and greater tolerance for flexibility.
The precautionary motive has gained relevance in the last two years. The desire to accumulate a comfortable level of reserves—equivalent at least to the level of public external debt plus US-dollar denominated deposits in the banking system—became statistically significant in 2007 and 2008 (Figure 2, chart 4). These results suggest the authorities’ greater comfort in selling foreign exchange since mid-2008, as the central bank’s reserve buffer exceeded the estimated optimal level (Figure 3).10
The central bank’s intervention on any certain day is positively correlated to intervention in the previous day, generating “intervention episodes.” This “degree of persistence” was positive and statistically significant throughout 2003–08, explaining the clustering observed in the intervention data. The constant of the regression in the intervention function reflects the level of the central bank’s “autonomous” intervention in the market—that is, the intervention that cannot be attributed to the authorities’ smoothing or precautionary motives. However, such “autonomous intervention” has not been statistically significant during most of the period under analysis (Figure 4).
The “inflation news” have not generally had an impact on the level of intervention, confirming that Peru’s ITF enjoys a high level of credibility. Generally, if the central bank’s Taylor-type rule is regarded as credible and effective, there would be no need to use foreign exchange intervention to assist inflation reach its official target. This is consistent with estimating statistically insignificant coefficients for the impact of inflation news on the level of foreign-exchange intervention, as it occurs largely throughout the sample period (Figure 5). However, estimates also indicate that in 2007 there was a statistically significant effect of inflation surprises on foreign exchange intervention. This suggests that the central bank expected to raise its own policy rate ITF whenever inflation exceeded expectations, and was aware that such a policy decision would trigger further capital inflows. This effect, while still attesting to the credibility of the ITF, also hints at a weakening of monetary policy independence during 2007, as the high levels of global liquidity exacerbated the response of capital inflows to exchange rate differentials.
Intervention Response due to Smoothing and Precautionary Motives 1/ 2/ 3/
Citation: IMF Staff Country Reports 2009, 041; 10.5089/9781451831184.002.A002
Source: Author’s estimates1/ OLS estimation, based on White-heteroskedasticity-robust confidence intervals, at 95 percent of confidence.2/ Based on a measure of volatility equal to the conditional variance derived from a GARCH (1,1) model.3/ Based on optimal reserves defined as the total of public external debt plus dollarized deposits in the previous quarter (Jeanne and Ranciere, 2006).Foreign Exchange Intervention and Precautionary Motive
Citation: IMF Staff Country Reports 2009, 041; 10.5089/9781451831184.002.A002
Source: Banco Central de la Republica del Peru.Autonomy and Persistence of Intervention 1/
Citation: IMF Staff Country Reports 2009, 041; 10.5089/9781451831184.002.A002
Source: Author’s estimates1/ OLS estimation, based on White-heteroskedasticity-robust confidence intervals, at 95 percent of confidence.Impact of Inflation Surprises
Citation: IMF Staff Country Reports 2009, 041; 10.5089/9781451831184.002.A002
Source: Author’s estimates1/ OLS estimation, based on White-heteroskedasticity-robust confidence intervals, at 95 percent of confidence.B. Recent Evolution of the Foreign Exchange Intervention function
14. Recursive model’s estimates for the past two years confirm above’s results:11
The importance of the smoothing motive has diminished, but remains significant in the context of rising global uncertainties. The magnitude of all of the coefficients of the implicit intervention function linked to the smoothing motive declined in recent months, suggesting that intervention is responding less than before to exchange rate movements, and that the central bank is tolerating greater flexibility (Figure 6). However, some of these coefficients—particularly those linked to the deviation of the exchange rate from its trend, and to the exchange rate volatility—have become statistically significant since the Summer of 2008, suggesting that while the central bank has greater tolerance toward exchange rate fluctuations, it continues to moderate them, in light of the risks posed by the deterioration in the global environment.
The precautionary motive has remained largely significant in the past two years. Estimates indicate a tendency to build the reserve buffer throughout 2007, at the peak of the capital inflows. At the same time, there is evidence that the desaccumulation of reserves became more comfortable as the existing buffer rose relative to it benchmark level, and since the developments in global capital markets after mid-2008.
Intervention Response Due to Smoothing Motive 1/ 2/
Citation: IMF Staff Country Reports 2009, 041; 10.5089/9781451831184.002.A002
Precautionary: NIR Deviations from Benchmark
Citation: IMF Staff Country Reports 2009, 041; 10.5089/9781451831184.002.A002
Source: Author’s estimates1/ OLS estimation, based on White-heteroskedasticity-robust confidence intervals, at 95 percent of confidence.C. Conclusion
15. Peru’s high dollarization, its need to rebuild reserves, and the presence of large capital inflows have been important factors driving foreign exchange intervention. Evidence suggests that the authorities have frequently chosen to smooth the exchange rate movements to mitigate risks of sharp relative price adjustments that could stress the economy’s balance sheets. Moreover, they have also become focused about maintaining a robust reserve buffer against contingencies that might emerge from external shocks—particularly given the uncertain global environment.
16. The results also suggest that greater exchange rate flexibility has accompanied the intervention policy. In particular, estimates indicate that the authorities have balanced the heightened uncertainties in the international financial system by intervening somewhat less heavily to smooth the exchange rate path and moderate exchange rate volatility, thus allowing for a more rapid adjustment of the exchange rate to market conditions.
Technical Appendix
17. The paper presents a battery of estimations of the central bank’s foreign exchange intervention function. We estimate the model for daily data during the period January 2001 to late September 2008, building over a general specification of the intervention function widely used in the literature (Gersl and Holub, 2006):
Where:
It is the amount of central bank intervention as a share of the foreign exchange interbank market turnover;
Newst represents the surprise on monthly inflation relative to survey data;
Δln(et-1) and
represent, respectively, the central bank’s concern due to the one-day acceleration of the exchange rate and its deviation from a trend (given by its N-day moving average), where N=90 days; Volt reflects the degree of exchange rate volatility. For robustness, this is measured both as a simple N-day moving exchange-rate standard deviation and as a conditional variance derived from a simple GARCH model—both definitions are tested separately;
DevNIRt denotes the difference of NIR from its desired precautionary level. For robustness, we test separately for the gap of NIR relative to the daily level of dollarized deposits, and to an “optimal” reserve benchmark given by the sum of public external debt and US dollar-denominated deposits in the banking system (Jeanne and Ranciere, 2006; Batini and Peschiera, 2007), the N-day moving average of NIR, and the daily level of bank US-dollar denominated obligations subject to reserve requirements (TOSE).
Variable and Data Description
Variable and Data Description
| variable | name | Measure | Expected Coefficients | |
|---|---|---|---|---|
| OLS | PROBIT | |||
| It | Persistence | Lagged foreign exchange intervention (US$ million) as a share of the volume traded in the interbank foreign exchange market (US$ million) PROBIT estimated on a dummy that takes a value of 1 when the absolute value of intervention is different from 0. Source: BCRP. |
+,– | + |
| Δln(et-1) | (Lagged) Acceleration | One-day change in spot rate, in percent; an appreciation has a negative sign. PROBIT estimated on the absolute value Source: BCRP. |
– | + |
|
|
(Lagged) Deviation of exchange rate from reference level | Deviation of the log of the exchange rate from its N=90-day moving average, in percent. For an appreciating exchange rate this variable will exhibit a negative sign. PROBIT estimated on the absolute value Source: author’s calculations, with BCRP data on the exchange rate. |
– | + |
| Volt | Exchange rate volatility | N = 90-day moving standard deviation of the exchange rate. Conditional variance derived from a simple GARCH (1,1) model. Source: author’s calculations, with BCRP data on the exchange rate. |
+ | + |
| DevNIRt | Deviation of NIR from optimal reference level | ■ Difference of the central bank’s NIR to the sum of public external debt and dollarized deposits in the preceding quarter, as a share of the total external public debt plus dollarized deposits of the previous quarter. ■ Difference of the central bank’s NIR to their N-day moving average, as a share of the average ■ Difference of the central bank’s NIR relative to the daily total of banking system obligations in US dollars subject to reserve requirements (TOSE), as a share of the TOSE. PROBIT estimated on the absolute value Source: author’s calculations, based on BCRP data on daily net international reserves, daily TOSE and quarterly public external debt and dollarized deposits. |
– | + |
| Newst | Inflation News | Surprise in monthly Inflation relative to Bloomberg’s Survey expectation Source: BCRP and Bloomberg |
0 | 0 |
| Other Controls | Day of the Week (4), Post-Holiday | Dummy variables | +,– | +,– |
Variable and Data Description
| variable | name | Measure | Expected Coefficients | |
|---|---|---|---|---|
| OLS | PROBIT | |||
| It | Persistence | Lagged foreign exchange intervention (US$ million) as a share of the volume traded in the interbank foreign exchange market (US$ million) PROBIT estimated on a dummy that takes a value of 1 when the absolute value of intervention is different from 0. Source: BCRP. |
+,– | + |
| Δln(et-1) | (Lagged) Acceleration | One-day change in spot rate, in percent; an appreciation has a negative sign. PROBIT estimated on the absolute value Source: BCRP. |
– | + |
|
|
(Lagged) Deviation of exchange rate from reference level | Deviation of the log of the exchange rate from its N=90-day moving average, in percent. For an appreciating exchange rate this variable will exhibit a negative sign. PROBIT estimated on the absolute value Source: author’s calculations, with BCRP data on the exchange rate. |
– | + |
| Volt | Exchange rate volatility | N = 90-day moving standard deviation of the exchange rate. Conditional variance derived from a simple GARCH (1,1) model. Source: author’s calculations, with BCRP data on the exchange rate. |
+ | + |
| DevNIRt | Deviation of NIR from optimal reference level | ■ Difference of the central bank’s NIR to the sum of public external debt and dollarized deposits in the preceding quarter, as a share of the total external public debt plus dollarized deposits of the previous quarter. ■ Difference of the central bank’s NIR to their N-day moving average, as a share of the average ■ Difference of the central bank’s NIR relative to the daily total of banking system obligations in US dollars subject to reserve requirements (TOSE), as a share of the TOSE. PROBIT estimated on the absolute value Source: author’s calculations, based on BCRP data on daily net international reserves, daily TOSE and quarterly public external debt and dollarized deposits. |
– | + |
| Newst | Inflation News | Surprise in monthly Inflation relative to Bloomberg’s Survey expectation Source: BCRP and Bloomberg |
0 | 0 |
| Other Controls | Day of the Week (4), Post-Holiday | Dummy variables | +,– | +,– |
References
Arena, Marco and P. Tuesta (1999), “El objetivo de la intervención del banco central: el nivel del tipo de cambio, la reducción de la volatilidad cambiaria, o ambos? Un análisis de la experiencia peruana 1991–1998,” Banco Central de Reserva del Perú, Revista de Estudios Económicos 5, December.
Armas, Adrián and F. Grippa (2006), “Targeting Inflation in a Dollarised Economy: The Peruvian Experience” in “Financial Dollarization: The Policy Agenda,” International Monetary Fund, Washington DC.
Batini, Nicoletta and J. Peschiera (2007), “What is the Optimal Level of International Reserves for Peru?” International Monetary Fund, Draft.
Disyatat, Piti and G. Galati (2007), “The Effectiveness of Foreign Exchange Intervention in Emerging Market Countries: Evidence from the Czech Koruna” in Journal of International Money and Finance 26.
Gersl, Adam and T. Holub (2006), “Foreign Exchange Interventions Under Inflation Targeting: The Czech Experience,” Contemporary Economic Policy, Vol. 24, No. 4, October.
Guimaraes, Roberto and C. Karacadag (2006), “The Empirics of Foreign Exchange Intervention in Emerging Markets: Mexico and Turkey” in “Official Foreign Exchange Intervention” by S. Ishii, et al, International Monetary Fund, Occasional Paper 249.
Humala, Alberto and G. Rodríguez (2008), Foreign Exchange Intervention and Exchange Rate Volatility in Peru,” Banco Central de Reserva del Perú, draft working paper, April.
Jeanne, Olivier and R. Ranciere (2006), “The Optimal Level of International Reserves for Emerging Market Countries: Formulas and Applications,” International Monetary Fund Working Paper WP/06/229, June.
Kamil, Herman (2008), “Is Central Bank Intervention Effective Under Inflation Targeting Regimes? The Case of Colombia,” International Monetary Fund, Working Paper WP/08/08, April.
Rossini, Renzo and M. Vega (2007) “El mecanismo de transmisión de la política monetaria en un entorno de dolarización financiera: El caso del Perú entre 1996 y 2006,” Banco Central de Reserva del Perú, D.T. No. 2007–17, November.
This paper was prepared by Maria Gonzalez.
Arena y Tuesta (1999) have also shown that foreign exchange rate intervention was geared toward reducing exchange rate volatility during the 1990s, prior to the implementation of the inflation targeting regime.
Estimates based on the 90-day moving standard deviation do not alter the thrust of the results presented here.
Estimates based on alternative optimal reserve benchmarks measured as: (1) the 90-day moving average of NIR; and (2) the daily level of bank obligations in US dollars subject to reserve requirements (TOSE) are available upon request and do not alter our conclusions.
Estimates based on a PROBIT model allows us to verify for the robustness of our estimates, and are available.
The stages of implementation are described in Armas and Grippa (2006) and Rossini and Vega (2007).
The estimated regression coefficient for the precautionary motive in 2006 is significant but with an unexpected positive sign, reflecting an increased reduction in the level of reserves when these were still below the benchmark given by public external debt and dollarized deposits.
The estimation covers September 2006 to September 2008, with a rolling 126-day sample window.
