Chapter

Chapter 5. The Impact of the U.S. Term Premium on Emerging Markets

Author(s):
Luis I. Jacome H., Yan Carriere-Swallow, Hamid Faruqee, and Krishna Srinivasan
Published Date:
October 2016
Share
  • ShareShare
Show Summary Details
Author(s)
Alberto Naudon and Andrés Yany 

Undoubtedly, one important and ongoing challenge for central banks in Latin America—and of small open economies in general—is how to deal with the global financial cycle. The increasing complexity and interconnectedness of financial systems have made this challenge more relevant than ever, as local financial conditions become more dependent on what happens abroad. Of course, this reality is not new, as evidenced by an extensive literature that has tried to unravel the connection between fluctuations of financial conditions in advanced economies and the economic cycle in emerging market economies. However, the aggressive monetary policy pursued by the major advanced economies to combat the negative effects of the global financial crisis and its aftermath has given new impetus to the issue. Some scholars conclude that an open capital account is simply incompatible with an independent monetary policy (Rey 2013). In this context, this chapter will focus on the two following questions:

  • How relevant has the global financial cycle been for emerging market economies?

  • How have central banks reacted, and what are the main issues going forward?

This chapter presents new evidence on the importance of the global financial cycle in emerging market economies and on how central banks are addressing the challenges involved. In particular, the focus is the macroeconomic and financial effects of variations of the U.S. term premium on small and open economies. The motivation for analyzing the impact of changes in the U.S. term premium, rather than on the more common topic of changes in U.S. monetary policy, is twofold. First, there is already a substantial amount of work on the impact of U.S. monetary policy on the global economy (Mackowiak 2007; Canova 2005). Second, during the last decade we have learned that the term premium is often a more important determinant of medium- and long-term interest rates, that this premium can vary substantially over time, and that it often varies for reasons unrelated to monetary policy. The “Greenspan conundrum” during the second part of the 2000s and the “taper tantrum” of 2013 are good examples of this issue.

To conduct the analysis, we study the response of financial and macro variables in nine small and open economies to a change in the U.S. term premium since 2003. The sample includes three advanced economies (Australia, Canada, and Switzerland), and six emerging market economies, of which three are from Latin America (Chile, Colombia, and Mexico), two from Asia (Korea and Thailand), and one from Africa (South Africa).

We first decompose each country’s long-term interest rate into two elements: the expected path of future short-term rates, and a term premium. Separating both elements allows us to independently study the monetary policy response and the financial channel, since the first is more related to the expected path of the short-term rate, and the latter to the behavior of the term premium. Then we study the response of several variables to a shock to the U.S. term premium.1 We estimate a structural vector autoregression (SVAR) model with a particular identification scheme following the “agnostic” approach of Arias, Rubio-Ramirez, and Waggoner (2014), where the shock is identified imposing sign and zero restrictions directly on the impulse response functions (IRFs). This approach allows us to identify cross-country patterns and compare the differences in the impulse responses.

The analysis of the SVAR shows the large influence of global financial shocks on small open economies. In particular, our results indicate that a U.S. term premium shock has a clear pass-through to local term premiums, thus raising long-term yields. In some countries we also find inflationary effects on consumer prices through nominal depreciation of the exchange rate, and adverse effects on output. Finally, we show how central banks react to mitigate the effects of the U.S. term premium shock. We observe heterogeneity of monetary policy responses across countries, with some cutting local interest rates and others increasing them. This heterogeneity might reflect the existence of a degree of fear of floating, with some central banks worrying about potential adverse effects of local currency depreciation on inflation or financial stability.

The next section presents the theoretical framework for the computation of the term premium and reviews the evolution of long-term interest rates in the nine selected countries and its relation to the evolution of U.S. interest rates. The chapter then outlines the empirical methodology and results before presenting conclusions.

Decomposition of Yields

Theoretical Framework

A central piece of our analysis is the idea that long-term interest rates can be decomposed into the two elements in the following equation:

where Rtn is the n-period nominal interest rate, rt+s is the one-period nominal interest rate, and ρtn is the associated term premium.

The first element on the right-hand side of the equation is the expected future value of the short-term interest rate during the lifespan of the bond. As is common in the literature, we call this the risk-neutral rate. Since the short-term interest rate is typically used by central banks as a monetary policy instrument, we use this component as a measure of the monetary policy stance. It is important to note that, in contrast to most previous studies, we do not use the monetary policy rate, but the expected value of its future path to quantify the monetary policy. In our view, this is—at least conceptually—a better indicator of the monetary policy stance, since monetary policy is not only about setting the interest rate each month, but about guiding the market as to what the authorities will do in the future, a point that has been emphasized by the idea of forward guidance.

The second element on the right-hand side of equation (5.1) is the term premium, which captures the additional return that an investor requires to hold a long-term bond instead of rolling over a sequence of short-term bonds, since the first strategy involves interest rate risk. The term premium varies a great deal over time, and the mechanisms that determine it are not entirely clear. At times, it responds directly to changes in monetary policy. But often it moves independently, posing a challenge for monetary authorities to the extent that it causes financial conditions to deviate from the desired stance of monetary policy. A good example of this situation is the Greenspan conundrum of the mid-2000s, when a declining term premium offset a series of significant increases in the federal funds rate, leaving long-term interest rates unchanged (Bernanke 2013).

The variation of the term premium in the United States has been extensively studied (Ang and Piazzesi 2003; Wright 2011; Adrian, Crump, and Moench 2013; Joslin, Priebsch, and Singleton 2014). More recently, the large-scale asset purchase programs—or quantitative easing—pursued aggressively by the U.S. Federal Reserve have been associated with the reduction in long-term yields via a compression of the term premium. Recent estimates suggest that the first two quantitative-easing programs reduced 10-year yields by a total of about 100 basis points, attributing most of the drop in yields to a reduction in the term premium (Gagnon and others 2011).

Equation (5.1) also helps to explain how changes in the U.S. term premium can impact the long-term interest rate in other economies. First, the change in the long-term local rate depends on the reaction of monetary policy in each country. For example, if local authorities respond to an increase in the U.S. term premium by increasing their own policy rate—perhaps because they want to moderate the depreciation of their currency—then the expected path for the short-term rate will increase, and so will the long-term rate. Alternatively, term premiums in local economies could change—for instance, if the movement in the U.S. term premium generates portfolio reallocation that prompts some investors to sell domestic bonds—putting upward pressure on local long-term rates.

Evolution of Yields and the Term Premium in Small Open Economies

This section reviews the evolution of long-term interest rates in the nine small open economies and their relation with movements in U.S. long-term rates. For each country, we separate the movements of long-term interest rates into those associated with term premiums and those associated with changes in the risk-neutral interest rate. We use these two elements to study the behavior of the long-term interest rates in each economy and analyze their co-movement with corresponding series in the United States.

Figure 5.1 shows the evolution of the 10-year nominal interest rates in all nine selected countries since 2003, along with the 10-year interest rate in the United States. It is clear that long-term interest rates fell in most countries after the global financial crisis, although with varying intensity. It is also clear that, in general, the declines have coincided with the evolution of the interest rate in the United States. The falls have been substantial: in most cases the rates are now about half of their 2003–05 average. In the United States, the reduction has been almost 250 basis points. In advanced open economies, the fall has been of similar magnitude. In Latin American economies, changes have been somewhat larger, particularly in Colombia, where the 10-year rate fell more than 600 basis points during this period. In Mexico the fall was also large, at 350 basis points, while in Chile it was about 250 basis points.

Figure 5.1.Ten-Year Nominal Interest Rates in the United States and Selected Countries

(Percent)

Source: Bloomberg L.P.

Figure 5.2 shows the evolution of estimated term premiums in selected economies since 2003. Beyond the well-known downward trend of the term premium in the United States, it is evident that term premiums have fallen in almost all economies considered in our analysis. In most cases, the decline of the term premiums corresponds to more than two-thirds of the fall in long-term interest rates documented in Figure 5.1.

Figure 5.2.Term Premiums in the United States and Selected Countries

(Percent)

Source: Authors’ calculations based on the methodology explained in the text and in Appendix 5.1.

The importance of the term premium for the evolution of long-term interest rates in each economy is more formally analyzed in Table 5.1. For each economy, the variance of the long-term interest rate is decomposed into (1) its covariance with the term premium, and (2) its covariance with the risk-neutral rate. Results of this exercise are shown in panel 1. In all cases except Switzerland, the term premium accounts for a significant share of the variance of long-term interest rates. In most cases, the portion explained by the term premium is larger than 50 percent. In the case of Latin American economies, the importance of term premiums is among the highest in our sample. The case of Chile is particularly noticeable, with the term premium explaining all the variation of the long-term interest rate, since the covariance between the risk-neutral interest rate and the long-term interest rate is slightly negative. The results are stronger still when we analyze monthly differences of long-term rates (panel 2), indicating that the variation of term premiums is important in explaining movements at higher frequency, beyond the downward trend in interest rates that has been seen in recent years.

Table 5.1Variance Decomposition of Long-Term Interest Rates
AustraliaSwitzerlandCanadaChileColombiaMexicoKoreaThailandSouth AfricaUnited States
1. Levels
Term premium0.660.040.561.060.870.600.230.350.740.25
Risk-neutral rate0.340.960.44−0.060.130.400.770.650.260.75
2. First difference
Term premium0.540.060.610.991.430.690.450.510.980.82
Risk-neutral rate0.460.940.390.01−0.430.310.550.490.020.18
Source: Authors’ calculations.Note: The covariance decomposition is based on the fact that for each country VAR(R) = Cov(R, RNR) + Cov(R, TP), where R is the long-term interest rate, RNR is the risk-neutral rate, and TP is the term premium.
Source: Authors’ calculations.Note: The covariance decomposition is based on the fact that for each country VAR(R) = Cov(R, RNR) + Cov(R, TP), where R is the long-term interest rate, RNR is the risk-neutral rate, and TP is the term premium.

The relationship between the interest rates in small open economies and the interest rates in the United States is analyzed in Table 5.2. Panel 1 shows the correlations between long-term interest rates in each of these economies and their U.S. counterpart, as well as the correlation between the term premiums and between risk-neutral rates. Consistent with the visual examination of Figure 5.1, the correlation between the long-term interest rates is high, at 0.75 on average. The high correlation between interest rates of advanced economies is remarkable, with a coefficient of about 0.9. In emerging market economies, the correlations are somewhat lower, but still show levels that denote a high degree of co-movement. Among the economies of Latin America, Mexico shows the highest correlation, followed by Colombia and Chile. These rankings could be associated with the relative importance of foreign participants in each of these markets.

Table 5.2Correlations between Local Long-Term Interest Rates, Term Premiums, and Risk-Neutral Rates with Their U.S. Counterparts
AustraliaSwitzerlandCanadaChileColombiaMexicoKoreaThailandSouth Africa
1. Correlations (levels)
10-year yield0.890.940.910.620.700.820.830.670.49
Term premium0.300.830.670.640.550.340.550.470.64
Risk-neutral rate0.340.810.820.140.430.560.640.620.69
2. Share of covariance (levels)
Term premium0.380.300.340.740.560.320.41−0.181.10
Risk-neutral rate0.620.700.660.260.440.680.591.18−0.10
3. Correlations (first difference)
10-year yield0.520.590.500.210.100.310.310.270.32
Term premium0.310.420.630.200.130.260.020.220.28
Risk-neutral rate0.220.410.490.060.130.110.170.260.10
4. Share of covariance (first difference)
Term-premium0.720.760.790.692.020.790.580.560.92
Risk-neutral rate0.280.240.210.31−1.020.210.420.440.08
Source: Authors’ calculations.Note: Covariance decomposition is based on the fact that Cov(R, RUS) = Cov(R, RNRUS) + Cov(R, TPUS), where R is the long-term interest rate, RNR is the risk-neutral rate, and TP is the term premium.
Source: Authors’ calculations.Note: Covariance decomposition is based on the fact that Cov(R, RUS) = Cov(R, RNRUS) + Cov(R, TPUS), where R is the long-term interest rate, RNR is the risk-neutral rate, and TP is the term premium.

When we focus on cross-country correlations for each component of longterm interest rates, the results are somewhat mixed. In general, the correlations between the local components and their U.S. counterparts remain high for both the term premiums and the risk-neutral rates. Indeed, for both variables the coefficients are greater than 0.5 in most cases. Panel 2 in Table 5.2 shows the contributions of the U.S term premium and the risk-neutral rate to the co-movement between the local and the U.S. long-term interest rates. In general, the covariance between the local interest rate and the U.S. term premium accounts for more than a third of the covariance. In the Latin American countries, the share tends to be higher. Panels 3 and 4 repeat the exercise for monthly differences, and the results remain essentially unchanged.

The evidence presented above shows that long-term interest rates have fallen since the crisis, and that this decline is related to prospects of more dovish monetary policy and a reduction of term premiums. Additionally, a significant degree of correlation is observed between the evolution of long-term rates in the analyzed economies and what has happened in the United States. This correlation is due both to co-movement of monetary policies and the co-movement of term premiums.2

In the Latin American economies, these patterns tend to be even more significant, as most of the movement of long-term interest rates is associated with movements of the term premiums. Regarding the co-movement with long-term rates in the United States, the data show that a considerable part of this co-movement occurs through the co-movement with the U.S. term premium.

Impact of the U.S. Term Premium on Small and Open Economies

Empirical Model

We follow the approach of Fornero, Montero, and Yany (2015) to compute the IRFs to a 50 basis point U.S. term premium shock in small and open economies. The empirical model is a SVAR with two blocks: a first block composed of foreign variables and a second block composed of the domestic variables of a small and open economy. We assume block exogeneity for the foreign block, such that changes in domestic variables do not affect the responses of foreign variables. Using the notation of Arias, Rubio-Ramirez, and Waggoner (2014), the model can be written as:

where Yt=[yt*yt] denotes an n × 1 vector of endogenous variables of the small open economy, yt* denotes an n* × 1 vector of endogenous variables of the foreign block, and p is the lag length of the model. The zero blocks in the system reflect the block exogeneity assumption, and Al is the matrix of structural parameters to be estimated, with A0 invertible. The vector c is a vector of parameters and the model is defined for 1 ≤ tT, where T is the sample size. The vectors εt and εt* are Gaussian, with mean zero and variance-covariance matrix In+n., corresponding to the (n + n*) × (n + n*) identity matrix.

The model can be compactly written as:

where Yt=[yt*yt],Xt=[Yt1Ytp1]. The reduced-form model can be written as:

where B=A+A01,ut=ɛtA01 and E[utut]=Σ=(A0A0)1. Structural shocks and parameters are identified and estimated using an identification procedure that we describe briefly in the following subsection.

Identification Scheme

We use the “agnostic” procedure introduced in Uhlig (2005) and recently developed by Arias, Rubio-Ramirez, and Waggoner (2014). Under this approach, structural identification is achieved by imposing sign and zero restrictions directly on the IRFs of the model, and the estimation is done in a Bayesian framework. In this chapter we use the methodology of Fornero, Montero, and Yany (2015), which is an extension of Arias, Rubio-Ramirez, and Waggoner (2014) with the addition of a foreign exogenous block into the model.3 The identification algorithm can be briefly summarized as follows:

  • Draw (B; Σ) from the posterior of the reduced-form parameters;

  • Generate (A0*;A+*) by using a mapping between the reduced-form and the structural parameters;4

  • Draw an orthogonal matrix Q such that (A0*Q;A+*Q) satisfies the zero restrictions;5

  • Keep the draw if sign restrictions are satisfied;

  • Repeat the previous steps until the desired number of simulations is reached;

  • Compute the median for the full set of IRFs that satisfy the restrictions.

Data and Restrictions

As mentioned earlier, we independently estimate the SVAR model for nine economies: Australia, Canada, Chile, Colombia, Korea, Mexico, South Africa, Switzerland, and Thailand. All these countries are small open economies, such that foreign business cycles—characterized here by U.S. data—can be considered exogenous. We use official monthly data covering the period from January 2003 to March 2015. To facilitate the comparison across countries, for all reported countries we choose one lag in the VAR model and we add a constant as a deterministic trend.

The endogenous domestic block includes the following set of variables: (1) the output gap, measured as the Christiano-Fitzgerald filtered business cycle of the monthly index of production (in logs); (2) the annual Consumer Price Index (CPI) inflation rate; (3) the risk-neutral interest rate; (4) the 10-year term premium rate; and (5) the nominal exchange rate (in logs).

The exogenous foreign block includes (1) the U.S. output gap, measured as the Christiano-Fitzgerald filtered business cycle of the U.S. industrial production index; (2) the U.S. annual CPI inflation rate; (3) the U.S. risk-neutral interest rate; (4) the U.S. 10-year term premium rate; (5) a real commodity price index (in logs); (6) the CBOE Volatility Index (VIX, in logs); and (7) the 10-year–1-year yield spread. Details of variables, transformations, and sources are provided in Appendix 5.1.

In Table 5.3 we summarize the sign and zero restrictions imposed on the IRFs, which are quite standard and based on economic theory. We compute the IRFs of two different exercises (“Model 1” and “Model 2”). In both exercises, we choose a U.S. term premium shock that is positive for at least six months. The shock does not have contemporary effects on output gaps or annual inflation rates in either country. Also, in both identification schemes the shock has a positive impact on the interest rate spread for at least six months. However, in Model 2 we add a positive sign restriction for six months on the domestic term premium. We add this restriction to the domestic block for robustness and to help identify the shock’s transmission in the small open economy.

Table 5.3Sign and Zero Restrictions Imposed on the Impulse Response Functions in the Structural Vector Autoregression Model
Model 1Model 2
Variablet = 0t > 1t = 0t > 1
Foreign Block
U.S. output gap0?0?
U.S. annual inflation rate0?0?
U.S. risk-neutral rate????
U.S. term premium++5++5
U.S. 10-year versus 1-year spread++5++5
Log real commodity price????
Log VIX????
Domestic Block
Output gap0?0?
Annual inflation rate0?0?
Risk-neutral rate????
Term premium??++5
Log nominal exchange rate????
Source: Authors’ compilation.Note: +N indicates a positive sign restriction for N months, 0 indicates a zero restriction; ? indicates that no restriction was imposed on the variable.
Source: Authors’ compilation.Note: +N indicates a positive sign restriction for N months, 0 indicates a zero restriction; ? indicates that no restriction was imposed on the variable.

Results

Tables 5.4 to 5.7 show the median impulse responses of the foreign and domestic variables to an increase of 50 basis points in the U.S. term premium for the nine selected countries under both identification schemes (Models 1 and 2). We divide these countries into three categories: advanced economies (Australia, Canada, and Switzerland), Latin American countries (Chile, Colombia, and Mexico), and the others (Korea, South Africa, and Thailand) in order to identify regional cross-country patterns and differences across emerging market and advanced economies.

Table 5.4Monthly Impulse Response Functions of Foreign Variables to a U.S. Term Premium Shock of 50 Basis Points: Exercise with Model 1(Mean and standard deviation across countries)
Log VIX (%)U.S. Output Gap (basis points)U.S. Annual Inflation (basis points)U.S. Term Premium (basis points)U.S. RiskNeutral Rate (basis points)Log Real Commodity Price (%)
t = 04.4(1.6)0.0(0.0)0.0(0.0)50.0(0.0)−13.2(1.3)−0.1(0.4)
t = 16.2(1.4)0.0(1.1)4.8(0.6)46.2(0.5)−14.5(1.1)0.5(0.3)
t = 610.5(0.5)−16.5(3.2)6.6(2.2)30.8(0.9)−22.0(0.9)0.8(0.5)
t = 128.4(0.4)−41.5(3.8)−7.8(2.2)20.7(0.7)−26.6(1.3)−0.3(0.5)
t = 241.8(0.2)−52.9(3.3)−16.8(0.8)9.8(0.4)−19.9(1.4)−1.5(0.3)
t = 36−0.8(0.1)−34.9(2.1)−8.4(0.9)3.8(0.4)−7.9(1.0)−1.6(0.2)
Source : Authors’ calculations.Note: The variable Log Real Commodity Price in the exogenous block varies across countries, so the impulse responses of exogenous block variables are not the same in each estimate. The table shows the average estimate across countries. The corresponding standard deviation is in parentheses. VIX = Chicago Board Options Exchange Volatility Index.
Source : Authors’ calculations.Note: The variable Log Real Commodity Price in the exogenous block varies across countries, so the impulse responses of exogenous block variables are not the same in each estimate. The table shows the average estimate across countries. The corresponding standard deviation is in parentheses. VIX = Chicago Board Options Exchange Volatility Index.
Table 5.5Monthly Impulse Response Functions to a U.S. Term Premium Shock of 50 Basis Points: Exercise with Model 1
AustraliaSwitzerlandCanadaChileColombiaMexicoKoreaThailandSouth AfricaAdvanced EconomiesLatin America
1. Term Premium (basis points)
t = 07.3915.81531.611.5−0.2315.810.719.3
t = 111.112.417.42140.320.25.54.817.413.627.1
t = 614.112.216.429.469.629.711.412.316.414.242.9
t = 12410.49.222.273.517.38.49.39.27.937.7
t = 24−6.98.5−0.30.769.31.96.22−0.30.423.9
t =36−6.14.9−2.2−1.157.42.43.21.3−2.2−1.119.6
2. Log Nominal Exchange Rate (percent)
t = 01.50.50.20.90.90.91.1−0.20.20.70.9
t = 120.6−0.10.90.911.5−0.2−0.10.91
t = 60.90.3−0.61.30.90.320.3−0.60.20.9
t = 12−0.20.2−10.80.3−0.62.20.7−1−0.30.1
t = 24−0.60.4−1.20.10.501.80.7−1.2−0.50.2
t = 36−0.20.5−0.51.12.3110.4−0.5−0.11.4
3. Output Gap (basis points)
t = 000000000000
t = 14.7−0.68.7−3.612.3−2.4−0.3−36.28.74.32.1
t = 68.3−4.66.6−20.77.6−24.7−27.6−142.56.63.4−12.6
t = 120.3−14.7−29.3−35.6−72.7−45.5−47−150.2−29.3−14.6−51.3
t = 24−16.4−23.1−62.9−31.8−166.1−46.4−19.9−81.6−62.9−34.1−81.4
t = 36−17.1−15.8−55.7−21.2−53.9−24.211.8−48.6−55.7−29.6−33.1
4. Annual inflation (basis points)
t = 000000000000
t = 13.89.96.911.31.911.215.406.96.98.1
t = 624.419.414.242.120.430.741−2.114.219.431.1
t = 12185.8319.413.523.925.1−12.238.918.9
t = 24−8.6−6.1−11.3−40.8−31.62.6−2.4−14.1−11.3−8.7−23.3
t = 36−9.3−4−9.1−34−20.6−2.2−3.1−9.8−9.1−7.4−19
5. Risk-Neutral Rate (basis points)
t = 07.44.9−5.3−5.8−3.25.15.7−5.6−5.32.3−1.3
t = 113.29.8−1.8−6.60.210.26.5−5−1.871.3
t = 610.74.81.8−6.5−1.518.55.5−101.85.83.5
t = 121.5−8.6−3.2−3.9−21.412.4−2.1−12.2−3.2−3.4−4.3
t = 24−1−17.8−8.2−2−74−0.6−13−9.6−8.2−9−25.5
t = 36−0.1−9−6.1−2.9−64.7−0.9−11.8−4.6−6.1−5.1−22.9
Source: Authors’ calculations.Note: The last two columns show the average estimate across countries.
Source: Authors’ calculations.Note: The last two columns show the average estimate across countries.
Table 5.6Monthly Impulse Response Functions of Foreign Variables to a U.S. Term Premium Shock of 50 Basis Points: Exercise with Model 2(Mean and standard deviation across countries)
Log VIX (%)U.S. Output Gap (basis points)U.S. Annual Inflation (basis points)U.S. Term Premium (basis points)U.S. RiskNeutral Rate (basis points)Log Real Commodity Price (%)
t = 03.9(3.3)0.0(0.0)0.0(0.0)50.0(0.0)−10.1(3.9)0.6(0.9)
t = 16.0(2.8)0.8(2.3)5.9(1.6)46.2(0.4)−11.5(3.7)1.1(1.0)
t = 610.7(0.8)−13.1(7.2)9.8(5.7)31.1(0.9)−19.9(2.9)1.5(1.0)
t = 129.0(0.7)−39.2(7.4)−5.8(4.6)21.1(0.7)−25.6(2.3)0.4(0.9)
t = 242.5(0.5)−54.2(4.0)−15.9(1.0)10.7(0.7)−20.9(1.8)−1.0(0.5)
t = 36−0.3(0.3)−38.3(2.9)−8.9(0.9)4.6(0.5)−9.5(2.0)−1.1(0.3)
Source: Authors’ calculations.Note: The table shows the average estimate across countries. The corresponding standard deviation is in parentheses. The variable Log Real Commodity Price varies across countries, so its response is not the same in each estimate. VIX = Chicago Board Options Exchange Volatility Index.
Source: Authors’ calculations.Note: The table shows the average estimate across countries. The corresponding standard deviation is in parentheses. The variable Log Real Commodity Price varies across countries, so its response is not the same in each estimate. VIX = Chicago Board Options Exchange Volatility Index.
Table 5.7Monthly Impulse Response Functions to a U.S. Term Premium Shock of 50 Basis Points: Exercise with Model 2
AustraliaSwitzerlandCanadaChileColombiaMexicoKoreaThailandSouth AfricaAdvanced EconomiesLatin America
1. Term Premium (basis points)
t = 015.420.112.740.1105.626.218.828.151.216.157.3
t = 119.519.916.441.5109.534.82229.546.618.661.9
t = 621.617.21441.6106.740.218.225.133.817.662.8
t = 129.99.511.127.491.725.69.813.924.810.248.2
t = 24−5.2−1.69.41.3752.46.30.417.90.926.2
t = 36−6.5−3.25.2−0.6621.44.21.98.2−1.521
2. Log Nominal Exchange Rate (percent)
t = 00.30.30.91.11.91.81.7−0.330.51.6
t = 10.30.11.31.11.71.91.6−0.22.20.61.6
t = 6−0.3−0.91.21.51.10.61.80.6−0.201.1
t = 12−0.7−1.30.60.70.1−12.11.1−2−0.5−0.1
t = 24−0.8−1.50.5−0.10.1−0.71.90.4−1.8−0.6−0.3
t = 36−0.4−0.80.60.82.60.91.40.2−1.7−0.21.4
3. Output Gap (basis points)
t = 000000000000
t = 16.49.7−0.2−2.212.6−23−10.7−5.85.32.8
t = 617.29.6−6.1−15.27.7−25.6−16.6−18.3−28.76.9−11
t = 1210.9−25.5−16.4−29.6−85.3−52−42.5−17.2−35.8−10.3−55.6
t = 24−1.1−70.6−23.6−31.6−187−48.5−21.8−34−51.3−31.8−89
t = 36−6.9−58.4−16.3−22.4−45.9−22.18.8−30.5−61.3−27.2−30.1
4. Annual Inflation (basis points)
t = 000000000000
t = 16.77.38.27.7510.917.62.412.17.47.9
t = 633.517.717.732.223.534.648.312.743.12330.1
t = 1228.85.26.416.712.828.23312.319.913.519.3
t = 24−1.6−11.4−5.2−34.5−37.2−0.72.1−5.5−52.7−6.1−24.1
t = 36−7.6−10.5−3.9−34.8−23.2−4.3−1.1−10.5−57.7−7.3−20.8
5. Risk-Neutral Rate (basis points)
t = 015.1−5−2.3−9.5−48.813.810.7−6.7−92.6−14.8
t = 118.5−1.52.1−8−42.418.912−7.9−0.46.4−10.5
t = 612.82.40.9−2.9−25.926.513.1−10.64.15.4−0.8
t = 122.7−2.8−9.80−34.320.25.9−9.2−4.8−3.3−4.7
t = 24−0.8−9.1−17.2−0.6−85.90.9−9−8.1−12.3−9.1−28.5
t = 36−0.5−6.8−8.2−2.5−68.6−1.1−11.1−5.5−9.6−5.2−24.1
Source: Authors’ calculations.Note: The last two columns show the average estimate across countries.
Source: Authors’ calculations.Note: The last two columns show the average estimate across countries.

Table 5.4 shows the responses of the foreign variables in Model 1. The results suggest that the U.S. term premium shock is quite persistent, with a half-life of approximately 10 months. The estimated shock generates a persistent contraction of the U.S. output gap, with a decrease of 41 basis points on average after 12 months. In line with the contractionary effects of the shock on activity, the U.S risk-neutral rate—which represents the expected path of future short-term rates—decreases on average by 15 basis points on impact, and by 26 basis points after one year. Moreover, the shock causes a rise in volatility—measured here as the log of the VIX index—of 10 percent above its steady-state level after six months and 8 percent after one year. Finally, real commodity prices do not show important responses to the shock, decreasing by only 1.5 percent after one year.

Table 5.5 reports the impulse responses for the term premium in panel 1 and of the log nominal exchange rate in panel 2, under the identifying assumptions of Model 1. The SVAR estimates suggest a clear pass-through from U.S. to local term premiums, which increase in most cases between 10 and 30 basis points in the first six months. In most of the analyzed countries, the size of the response of local term premiums is less than 50 basis points, implying a pass-through coefficient that is smaller than 1. The magnitude of responses varies across country groups, with an average of 14 basis points in advanced economies, and 43 basis points in Latin American countries. In particular, Chile and Mexico have similar and strong responses, with a peak of almost 30 basis points after six months, and approximately 20 basis points after one year, respectively. In contrast, advanced economies have weaker and less persistent responses, with the local term premiums rising less than 20 basis points in Canada and only about 10 basis points in Australia and Switzerland. On the other hand, the nominal exchange rate depreciates between 1 and 2 percent in most countries. In particular, the exchange rate in Latin American countries depreciates 1 percent on impact and returns to its initial level after one year.

Domestic annual inflation (Table 5.5, panel 4) increases in most of the analyzed countries, in line with the nominal depreciation of local currencies. The results suggest a stronger and more persistent response in Latin American countries compared with advanced economies, with annual inflation increasing on average by 31 basis points after six months and 19 basis points after one year. In contrast, in advanced economies the response is on average 19 basis points after six months and only 9 basis points after one year. Regarding domestic activity, Table 5.5 (panel 3) shows the responses of the output gap in Model 1. The increase in the U.S. term premium is undoubtedly contractionary: after one year, the output gap decreases more than 30 basis points in most of the analyzed countries. The response is stronger in Latin American countries, where the output gap decreases 51 basis points on average after one year. In particular, in Chile and Mexico, the response is persistent after two years, with a decrease in the output gap of more than 30 basis points. In advanced economies, the responses are clearly weaker: the output gap diminishes only 15 basis points on average after one year. In the rest of the countries, the results are quite similar to those for Latin America, with the exception of Thailand.

How do central banks respond following a shock to the U.S. term premium? Table 5.5 (panel 5) reports the movement in risk-neutral interest rates. Across countries, responses are quite heterogeneous. On impact, some countries (Australia, Mexico, and Switzerland) show a rise in the expected monetary policy rate, while others show a reduction in it (Canada, Chile, Colombia, and South Africa). This heterogeneity might reflect the existence of monetary policy stability trade-offs. While some countries are expected to try to mitigate the contractionary effects on activity, others are expected to increase the policy rate to reduce inflation. For instance, Chile and Mexico show different responses of the risk-neutral rate, even though most of the other variables in the estimated model (output gap, annual inflation, and the term premium) react similarly.

So far we have discussed the effects of a U.S. term premium shock without imposing any sign restriction on the domestic term premium (Model 1). As mentioned earlier, we conduct a second exercise (Model 2) imposing a positive sign restriction on the domestic term premium for six months to facilitate the transmission of the shock to the domestic block. Results of this estimation are provided in Tables 5.6 and 5.7. In general, the responses are qualitatively comparable with the previous exercise: (1) contractionary effects on the output gap; (2) inflationary effects on local consumer prices; and (3) a persistent pass-through to domestic term premiums. However, with these additional restrictions, the pass-through to local financial conditions is clearly larger. Figure 5.3 displays the responses after one year of long-term interest rates and term premiums in Models 1 and 2. Long-term interest rates are, by definition, the sum of the term premium and the risk-neutral rate. According to the results reported in Figure 5.3, the U.S. term premium shock generates an increase in long-term interest rates and term premiums. This effect is larger in Latin American economies, which are situated in the upper-right corner of both panels of the figure. However, the impact on local term premiums and long-term interest rates are stronger when imposing sign and zero restrictions on the domestic block. For instance, in Chile and Mexico, the pass-through to long-term interest rates is almost twice as high in Model 2 as in Model 1. Finally, our results are broadly consistent with the recent results of Miyajima, Mohanty, and Yetman (2014), who analyzed the spillovers of unconventional monetary policy to Asian economies in the period from January 2003 to December 2007.

Figure 5.3.Responses of Domestic Term Premium and Implied 10-Year Interest Rate after One Year to a U.S. Term Premium Shock of 50 Basis Points

(Basis points)

Source: Authors’ calculations.

Note: Developed countries are indicated with an x, Latin American countries with circles, and the rest with triangles. Colombia is excluded from the figure because it is situated in the upper-right corner. S. Africa = South Africa; Swit. = Switzerland.

Conclusions

This chapter has studied the macroeconomic effects of the global financial cycle on small and open economies and the monetary policy implications of financial shocks. To tackle this issue, we conducted an estimation of a SVAR model with sign and zero restrictions for several countries following the methodology of Arias, Rubio-Ramirez, and Waggoner (2014). We included U.S. and domestic variables in each model and computed the impulse responses of the local variables to a U.S. term premium shock. Several findings emerge from this analysis:

  • A clear pass-through to local financial conditions, with a rise in domestic long-term interest rates and term premiums;

  • An increase in domestic inflation and contractionary impact on activity;

  • As a consequence, we observe heterogeneity across monetary policy responses that might reflect the existence of monetary policy trade-offs;

  • These effects are stronger in Latin American countries.

From a general perspective, the results indicate that the global financial cycle has significant effects on small open economies, especially in Latin American countries. In this context, central banks face monetary policy trade-offs that involve mitigating high inflation rates in countries with output deceleration. This issue will be crucial in the months ahead as the United States continues to normalize its monetary policy.

Appendix 5.1. Data and Term Premium Estimation

Data

We use official monthly data for each country on short- and long-term yields, production, the nominal exchange rate, and the annual inflation rate.

The output gap is measured as the Christiano-Fitzgerald filter (18–96; full sample and asymmetric; removing linear trends) of the monthly industrial production index. For Chile, we use the monthly economic activity index. We use the 12-month percentage change of the CPI as a proxy for annual inflation. Industrial production indexes, short- and long-term yields, CPIs, and the VIX are from Bloomberg L.P. and Organisation for Economic Co-operation and Development databases. The real commodity price index is computed with data taken from the IMF’s International Financial Statistics, and is deflated using the U.S. Producer Price Index.

Term Premium Estimation

Affine Model of the Term Structure

The decomposition of long-term yields into a risk-neutral rate and a term premium has been widely studied in the literature in the last two decades. A recent and popular tool has been the estimation of affine models of the term structure of yields (Ang and Piazzesi 2003; Adrian, Crump, and Moench 2013; Bauer, Rudebusch, and Wu 2014; Blake, Rule, and Rummel 2015). In this chapter, we decompose the 10-year yield for different countries following the approach of Ceballos and Romero (2015), who estimate affine models for several countries using the methodology of Adrian and others (2013), which we describe briefly in this appendix. First, we suppose that the term structure depends on its K principal components Xt, which evolve according to a VAR of order 1:

with vt ~ N(0, Σ). The short-term interest rate rt is assumed to depend on the principal components of the term structure and is modeled as an affine function:6

Second, we denote Ptn the price in period t of an n-period zero coupon bond. Using the assumption of no arbitrage (Ang and Piazzesi 2003), bond prices satisfy the following recursive dynamic:

where Mt+1 is the stochastic discount factor, which in turn is assumed to be exponentially affine:

The parameter λt is the market price of risk, which is also assumed to evolve according to an affine process with respect to the principal components of the term structure:

Using the assumptions above, one can deduce the following model-implied yield:7

where An and Bn satisfy the following recursive process:

with A1 = −δ0 and B1 = −δ1. With this, one can compute forecasts of bond prices Ptn and deduce the associated yield ytn, defined by:

Finally, to compute the risk-neutral rate, we set the market price of risk to zero (i.e., λ0 = λ1 = 0, so that λt = 0 ∀t):

Thus, the term premium can be computed as a residual of the model-implied yield ytn and the estimated risk-neutral rate.

Model Estimation

The parameters {μ, ϕ, Σ, δ0, δ1} are estimated independently by ordinary least squares (OLS) on the equations where they are defined. To estimate {λ,0, λ1}, first we denote rxt, the log excess holding return on a bond maturing in n periods:

It can be shown that rxt satisfies the following equation:

t = 1,…,T;∀n = 1, …,N, where et+1n1 is a return pricing error identically distributed with variance σ2 and

The system can be stacked and compactly rewritten as:

where rx is a N × T matrix of excess returns, β=[β1,β2,,βN],1T and 1N are T × 1 and N × 1 vectors of ones, X=[X0,X1,,XT1],B*=[vec(β1,β1),,vec(βn,βn)], and V and E are matrices of errors.

With this, the parameters can be estimated with the following three-step methodology:

  • 1. Estimate {μ, ϕ, Σ, δ0, δ1} independently by OLS on the equations where these parameters are defined.

  • 2. Estimate the following equation of excess returns by OLS:

    and deduce B^* and σ^2=tr(E^E^)/NT.

  • 3. Compute {λ0, λ1} using the following estimators:

References

    AdrianT.R. K.Crump and E.Moench.2013. “Pricing the Term Structure with Linear Regressions.Journal of Financial Economics110(1): 11038.

    AngA. and M.Piazzesi.2003. “A No-Arbitrage Vector Autoregression of Term Structure Dynamics with Macroeconomic and Latent Variables.Journal of Monetary Economics50(4): 74587.

    AriasJ. E.J. F.Rubio-Ramirez and D. F.Waggoner.2014. “Inference Based on SVARs Identified with Sign and Zero Restrictions: Theory and Applications.International Finance Discussion Paper 1100Board of Governors of the Federal Reserve System.

    BauerM. D.G. D.Rudebusch and J. C.Wu.2014. “Term Premia and Inflation Uncertainty: Empirical Evidence from an International Panel Dataset: Comment.American Economic Review104(1): 32337.

    BernankeB.2013. “The Past and Future of Monetary Policy.Speech at the Annual Monetary and Macroeconomics ConferenceSan Francisco Federal ReserveMarch 1.

    BlakeA. P.G. R.Rule and O. J.Rummel.2015. “Inflation Targeting and Term Premia Estimates for Latin America.Latin American Economic Review24(1): 121.

    CanovaF.2005The Transmission of U.S. Shocks to Latin Data.Journal of Applied Econometrics20: 22951.

    CeballosL.A.Naudon and D.Romero.2015. “Nominal Term Structure and Term Premia: Evidence from Chile.Working Paper 752Central Bank of ChileSantiago.

    CeballosL. and D.Romero.2015. “Decomposing Long-Term Interest Rates: An International Comparison.Working Paper 767Central Bank of ChileSantiago.

    ForneroJ.R.Montero and A.Yany.2015. “Reassessing the Effects of Foreign Monetary Policy on Output: New Evidence from Structural and Agnostic Identification Procedures.Central Bank of ChileSantiago. Unpublished.

    GagnonJ.M.RaskinJ.Remache and B.Sack.2011. “The Financial Market Effects of the Federal Reserve’s Large-Scale Asset Purchases.International Journal of Central Banking7(1): 343.

    JoslinS.M.Priebsch and K. J.Singleton.2014. “Risk Premiums in Dynamic Term Structure Models with Unspanned Macro Risks.Journal of Finance69(3): 1197233.

    MackowiakB.2007. “External Shocks, US Monetary Policy and Macroeconomic Fluctuations in Emerging Markets.Journal of Monetary Economics54(8): 251220.

    MiyajimaK.M.Mohanty and T.Chan.2012. “Emerging Market Local Currency Bonds: Diversification and Stability.BIS Working Paper 391Bank for International SettlementsBasel.

    MiyajimaK.M.Mohanty and J.Yetman.2014. “Spillovers of US Unconventional Monetary Policy to Asia: The Role of Long-Term Interest Rates.BIS Working Paper 478Bank for International SettlementsBasel.

    ReyH.2013. “Dilemma Not Trilemma: The Global Financial Cycle and Monetary Policy Independence.Proceedings from the Federal Reserve Bank of Kansas City Economic Policy SymposiumJackson Hole, WyomingAugust 22–24.

    TurnerP.2014. “The Global Long-Term Interest Rate, Financial Risks and Policy Choices in EMEs.BIS Working Paper 441Bank for International SettlementsBasel.

    UhligH.2005. “What Are the Effects of Monetary Policy on Output? Results from an Agnostic Identification Procedure.Journal of Monetary Economics52(2): 381419.

    WrightJ. H.2011. “Term Premia and Inflation Uncertainty: Empirical Evidence from an International Panel Dataset.American Economic Review101(4): 151434.

Ceballos, Naudon, and Romero (2015) conduct a similar analysis for the case of Chile. They show that most of the fall of long-term interest rates as well as their dynamics are related to the term premia, and that the the latter is driven primarily by uncertainty about nominal variables (expected inflation and the U.S. term premia).

Of course we are not the first to report the important correlation between rates in emerging market economies and the United States. See Turner (2014) and Miyajima, Mohanty, and Chan (2012).

A full description of the methodology is beyond the scope of this chapter. See Arias, Rubio-Ramirez, and Waggoner (2014) and Fornero, Montero, and Yany (2015) for a more detailed description.

The mapping between structural and reduced-form parameters can be implemented by using a function h() such that h(X)’h(X) = X, i.e., Cholesky decomposition: (A0*;A+*)=(h(Σ)1;Bh(Σ)1).

Using the QR decomposition (X = QR), which holds for any n × n random matrix on which each element is i.i.d. from a N(0,1).

We use the three-month yield as a proxy for the short-term interest rate.

See Ang and Piazzesi (2003) or Adrian, Crump, and Moench (2013) for a detailed computation of these equations.

    Other Resources Citing This Publication