A. Motivation

1. The 2013 change in the central bank’s operational framework, thus far, has not brought back inflation within the target range yet. In July 2013, the central bank of Uruguay (BCU) switched from using the overnight interest rate as a monetary policy instrument to announcing a reference range for the growth of M1+ within its monetary policy framework. This change resulted in an immediate increase in short-term interest rates and a tightening of domestic financial conditions (Figure 1). However, inflation has remained well above the 3 to 7 percent target range, peaking at 11 percent in May 2016, raising the question whether monetary policy has been tight enough (while acknowledging the critical role of complementary policies for lowering inflation).

View Full Size

Uruguay: Short-Term Interest Rates and Financial Conditions

Citation: IMF Staff Country Reports 2017, 029; 10.5089/9781475573626.002.A003

• Download Figure
• Download figure as PowerPoint slide

2. The neutral interest rate provides a benchmark to assess the stance of monetary policy. The neutral interest rate differs from the long-term equilibrium interest rate (Bernhardsen and Gerdrup, 2007). The long-term equilibrium real interest rate is determined by structural factors such as domestic and global demographic trends, technology, and long-term saving preferences. The real neutral rate is affected, in addition, by medium-term disturbances, like changes in fiscal policy. Accordingly, we define the neutral real interest rate (RNIR) as the interest rate that is consistent with a closed output gap and stable inflation, for a given stance of other policies and market conditions. The gap between that neutral rate and the actual short-term real interest rate provides a measure of the degree to which monetary policy is stimulating or contracting the economy.

B. Neutral Interest Rate Estimates

3. There is no single best method to estimate the neutral interest rate. As written by Alan Blinder (1998), the neutral interest rate is “difficult to estimate and impossible to know with precision.” Because of this difficulty, this paper provides a range of estimates, using several dynamic models, following the approach of Magud and Tsounta (2012).

4. Dynamic approaches allow looking at fluctuations of the neutral interest rate over time. The models usually take the form of state-space systems in which the real neutral interest rate (RNIR) is an unobserved state variable, estimated using a Kalman filter (see for instance Laubach and Williams, 2003).

5. As a preliminary approximation the Hodrick-Prescott trend of the real one-month interest rate suggests an increase of the RNIR over the last ten years (Figure 2).² The end-of-period estimate is biased downward as the filter puts more weight on the most recent observations for which the real interest rate tended to decline. However, the overall trend remained upward, contrasting with the decline documented in the main advanced economies over the past 25 years (Holston, Laubach, and Williams, 2016). The rest of this section compares this pattern to that obtained with Kalman filter estimations.

View Full Size

Uruguay: HP-Filtered Real Interest Rate

Citation: IMF Staff Country Reports 2017, 029; 10.5089/9781475573626.002.A003

Sources: BEVSA, INE, and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

Implicit common stochastic trend

6. A first approach to the estimation of the RNIR is to assume the existence of a common stochastic trend that moves the short-term and long-term nominal interest rates jointly (Basdevant and others, 2004). In this model, a shift in the yield curve (i.e. an increase or decrease in both the short-term and long-term interest rates while the spread between the two remains constant) can be interpreted as a shift in the real neutral interest rate, whereas a steepening of the yield curve (i.e. an increase in the spread between the long-term and short-term interest rates) would be signaling that the real interest rate is below its neutral level. The model takes the form of the following system, with two signal and two state equations:

where r_t, R_t, and $r_t^{*}$ respectively denote the nominal short-term rate (one-month Treasury bill yield in peso), the nominal long-term rate (proxied by the 5-year Treasury bill in peso), and the real neutral interest rate, ${\pi }_t^e$ is the expected inflation at time t, α_t is the term premium, and ${\varepsilon }_t^1,{\varepsilon }_t^2,{\vartheta }_t^1$, and ${\vartheta }_t^2$ are i.i.d processes with zero means and constant variances. Inflation expectations are approximated by one-year ahead Consensus Forecasts.

7. This approach produces an estimate for the real neutral interest rate of 4.1 percent in September 2016, with movements that closely follow the changes in the short-term (and long-term) interest rates over time. The reason partially lies in the relatively short-maturity of our proxy for the long-term rate and thinness of the market for medium-term peso-denominated government securities, which implies a lot of co-movement of interest rates along the yield curve, picked up by our estimate of the real neutral interest rate.

Dynamic Taylor rule

8. Another approach is to use monetary policy rules to estimate the neutral real interest rate. The Taylor rule relates the nominal interest rate to the output gap and the deviation of actual inflation from the central bank’s target. When both gaps are closed, the interest rate suggested by the rule should be equal to the neutral interest rate. The neutral interest rate corresponds to the intercept in the rule. The formal model can be written as follows:

where r_t is the nominal policy rate, ${\overline{\pi }}_t$ is the inflation target of the central bank and ${\tilde{y}}_t$ denotes the output gap in period t, measured as the percentage deviation of real GDP from its potential level estimated with an Hodrick-Prescott filter. The RNIR is assumed to follow a random walk with trend growth g_t.

9. We consider two alternative specifications of the monetary policy rule which include a money growth target and an exchange rate target. The first alternative includes an additional term in the rule (first equation of the model above) capturing deviations of the growth rate for M1+ from the announced reference range, as suggested in Portillo and Ustyugova (2015), in order to capture the switch to a money targeting framework from July 2013. The second specification includes a term representing the change in the nominal exchange rate vis-à-vis the U.S. dollar. The inclusion of an exchange rate objective in the monetary policy rule is motivated by Florian and Montoro (2009) which describes a DSGE model for the partially dollarized Peruvian economy. In this model, some firms set prices in dollar, prices are rigid, and exchange rate fluctuations generate changes in relative prices that are inefficient. The central bank can therefore improve welfare by reducing exchange rate volatility.

10. The three rules produce estimates between 5 and 5½ percent (Figure 3). The estimates are highly correlated with each other and still strongly influenced by the movements of the short-term interest rate. The range of estimates lies above the HP-trend and the estimate derived from the common trend model. It is also higher than staff’s estimates of real potential growth (around 3 percent), which can, in principle, be considered a good proxy for the long-term equilibrium interest rate.

View Full Size

Uruguay: Real Neutral Interest Rate Estimates

Citation: IMF Staff Country Reports 2017, 029; 10.5089/9781475573626.002.A003

Sources: BCU, BEVSA, INE, and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

C. Discussion

11. The estimated level of the neutral real interest rate in Uruguay would imply a negative interest rate gap in 2016. This reflects the persistence of inflation well above the BCU’s target range and may suggest the existence of a moderately accommodative monetary policy. The widening of the interest rate gap over the last 18 months seems to indicate a gradual loosening of the monetary stance.

12. However, the results need to be considered with caution since they depend on the underlying model used in the estimation. There is uncertainty about the actual rule used by the BCU to conduct monetary policy. A Taylor rule, such as the ones considered in this paper, may also be a poor representation of the BCU’s reaction function under the new money targeting framework. The uncertainty around the estimation results is reflected in the large range for the interest gap (Figure 4).

View Full Size

Uruguay: Interest Rate Gap

Citation: IMF Staff Country Reports 2017, 029; 10.5089/9781475573626.002.A003

Sources: BEVSA, INE, and IMF staff calculations.

• Download Figure
• Download figure as PowerPoint slide

13. The high level of dollarization and high inflation persistence in Uruguay can explain a higher level of the RNIR. The estimates presented in this paper are within the range of previous estimates found in the literature for earlier years (e.g. Brum, Carballo and España, 2010; Magud and Tsounta, 2012). However, they are relatively high compared to estimates found for other countries in the region, with the possible exception of Brazil. This can be explained by a greater degree of dollarization of the economy and lower level of financial development. The low level of savings in peso can account for a higher equilibrium real interest rate (the bank peso loans to peso deposits ratio is close to 100 percent, as 80 percent of deposits are in dollar). These frictions also limit the effectiveness of monetary policy and tend to increase the inflation risk premium.³ Similarly, the high inflation persistence which results from the indexation of wages on past inflation makes inflation less responsive to interest rate changes and can result in a higher inflation risk premium and a structurally higher neutral interest rate.

Table 1.

Uruguay: Comparison of NRIR Estimates

Table 1.

Uruguay: Comparison of NRIR Estimates

In percent	Brum, Carballo, and España (2010)	Magud and Tsounta (2012)
Brazil	−	[4.5, 5.7]
Chile	−	[1.2, 2.9]
Colombia	−	[1.6,4.4]
Mexico	−	[1.3, 4.2]
Paraguay	−	[1.0, 3.8]
Peru	−	[1.0, 5.0]
Uruguay	[3.8, 7.0]	[1.3, 7.2]

View Table

Table 1.

Uruguay: Comparison of NRIR Estimates

In percent	Brum, Carballo, and España (2010)	Magud and Tsounta (2012)
Brazil	−	[4.5, 5.7]
Chile	−	[1.2, 2.9]
Colombia	−	[1.6,4.4]
Mexico	−	[1.3, 4.2]
Paraguay	−	[1.0, 3.8]
Peru	−	[1.0, 5.0]
Uruguay	[3.8, 7.0]	[1.3, 7.2]

14. The upward trend in the RNIR found above contrasts with the global decrease in real interest rates documented in the literature. Possible reasons for this rise include the high inflation volatility, resulting in an increase in the inflation risk premium, and the rise in public debt in the past five years not compensated by an increase in private savings. A successful fiscal adjustment and a durable decline in inflation within the BCU’s target range could therefore bring the RNIR more in line with global trends.