IIMB-WP N0. 579 WORKING PAPER NO: 579 Yield Curve and Economic Activity Abhishek Naresh Doctoral Student Indian Institute of Management Bangalore Bannerghatta Road, Bangalore – 5600 76 [email protected]Jong Kook Shin Newcastle University Business School, 5 Barrack Road, Newcastle Upon Type, NE12 8RP, England [email protected]Chetan Subramanian Professor Economics and Social Sciences Indian Institute of Management Bangalore Bannerghatta Road, Bangalore – 5600 76 [email protected]Year of Publication – December 2018
32
Embed
Indian Institute of Management Bangalore | IIM Bangalore
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
IIMB-WP N0. 579
WORKING PAPER NO: 579
Yield Curve and Economic Activity
Abhishek Naresh Doctoral Student
Indian Institute of Management Bangalore Bannerghatta Road, Bangalore – 5600 76
Abhishek Naresh∗ Jong Kook Shin † Chetan Subramanian‡
December 14, 2018
Abstract
In this paper, we document the fact that long-term interest rates in India exhibit
a high degree of co-movement with their US counterparts in the post-reform period.
Further, we also establish that the US yield spread does a better job at forecasting
Indian business cycles than the Indian yield spread. Do these empirical findings
suggest an effective decoupling of the long-term and short-term domestic interest
rates in India?
We investigate this issue by estimating a small open economy model augmented
to include both domestic and foreign yield curves. Our results suggest that the
higher persistence of the US relative to domestic shocks is able to explain the
aforementioned features of the data. The results therefore do not support the
decoupling hypothesis.
JEL classification: E43, E52, E58, F41
Keywords: Yield curve, Business cycle dating, Monetary policy spillovers
∗Department of Economics and Social Sciences, IIM, Bangalore, Bannergatta Road, India 560076.
E-mail: [email protected] Phone: (91)-99-58021101.†Newcastle University Business School, 5 Barrack Road, Newcastle upon Tyne, NE12 8RP, England,
UK; E-mail: [email protected]; Phone: +44-(0)191 208 1677‡Corresponding author: Department of Economics and Social Sciences, IIM, Bangalore, Bannergatta
Road, India 560076. E-mail: [email protected], Phone: (91)-80-26993345.
1
1 Introduction
In the last couple of decades, the Indian economy and financial markets have integrated
with the global markets to an extent unprecedented in history. To what extent does
this economic integration impact the transmission of monetary policy in an emerging
economy such as India? Does the increased economic integration translate into higher
global monetary policy spillovers for the Indian economy? Does, for instance, a mone-
tary contraction in the US lead to a recession in India? Our objective is to investigate
these issues both from an empirical and a theoretical standpoint in this paper. Specifi-
cally, we examine the linkages between the US and Indian interest rates along different
points on the yield curve and contrast the effectiveness of the US and Indian yield
spreads in predicting economic activity in India.
Modern central banks conduct monetary policy by influencing the short-term inter-
est rates. However, since the real economy is a function of the long-term real interest
rate, the effectiveness of such actions will crucially depend on the ability of policymaker
to influence the long-term nominal interest rate. Theoretically, this link between the
two rates is established through the expectations hypothesis, according to which the
long-term interest rates can be thought of as an average of expected future short-term
rates. Thus, an increase in the long-term rate can be interpreted as implying rising
future short-term rates. It follows that, in standard monetary business cycle models,
domestic monetary policy is effective to the extent that the long-term nominal interest
rates are not decoupled from the expected path of the short-term interest rates.
The yield spread, i.e., the difference between long-term and short-term interest
1
rates, therefore, becomes a critical indicator of monetary policy stance. It has over the
years emerged as an important input in forecasting the evolution of future real economic
activity, particularly in advanced economies such as the United States. For example,
monetary policy tightening, which usually means a rise in short-term interest rates,
is typically implemented to slow down economic activities and to reduce inflationary
pressures. However, once these pressures subside, if the policy is expected to ease, then
lower short-term interest rates will follow. Long-term rates which reflect expectations
of future short-term interest rates, therefore, should rise by less than the short-term
rate. Thus, monetary tightening both slows down the economy and flattens (or even
inverts) the yield curve. Consistent with this line of reasoning, the yield curve has
inverted in the US before each of the last six recessions.
Surprisingly and in stark contrast to large economies, there is growing evidence to
suggest a decoupling of the short-term and long-term interest rates in inflation target-
ing small open economies. Recent studies (Bernanke (2013), Kulish and Rees (2011),
Wright (2011), Dahlquist and Hasseltoft (2013), Swanson and Williams (2014)) have
highlighted the fact that long-term nominal rates have exhibited a strong co-movement
over the post-war period rates in a number of inflation targeting small open economies
(such as Australia, Canada, New Zealand, and the UK) with their US counterparts.
The interest rate correlation between the US and small open economies appears to be
stronger at longer maturities than at shorter maturities.
Figure 1 illustrates this phenomenon for India. The correlation between the Indian
and US 10-year government bond yields over the period from 2001 Q1 to 2014 Q3 is
significantly greater than the correlation at short-term maturities. Does this suggest
2
that the long-term nominal interest rate in India has somehow decoupled from short-
term domestic policy rate? If so, is the US yield spread a better predictor of Indian
business cycles? What are the underlying determinants of the Indian yield curve?
There is an emerging literature that has sought to address the apparent decoupling
of the long and short-term interest rates as exhibited in Figure 1. One strand of the
literature points to the existence of significant term premiums in domestic interest
rates which are correlated internationally. This creates a ‘wedge’between expectations
over the path of future policy rates and long-term interest rates (see Wright (2011),
Dahlquist and Hasseltoft (2013) and Jotikasthira et al. (2015)). On the other hand,
Kulish and Rees (2011), offer an alternative explanation that does not rely on the
decoupling hypothesis. They use the standard small open economy augmented to
include term structure of interest rates to show that the observed patterns of interest
rate correlations would emerge if foreign disturbances are more persistent than domestic
ones, a hypothesis for which they find support in Australian data.
In this paper, we estimate the Kulish-Rees model using Indian data and seek to
explain the pattern of the interest rates observed in Figure 1. We also use this frame-
work to analyze the effectiveness of both the domestic as well as foreign yield curve
in predicting economic activity in India. The rest of the paper is organized as follows.
Section 2 presents a chronology of business cycles in India and develops an empirical
model used to predict recessions in India. The model is set up in Section 3. In Section
4, we estimate the model and carry out a discussion of the results. Section 5 concludes.
3
2 Empirical Section
In this section, we compare the effectiveness of the US and Indian yield spreads in
predicting recessions in the post-reform period in India. There is a fairly large empirical
literature that has found the yield spread to be useful in predicting recessions in the
United States. Estrella and Mishkin (1998), for example, analyze several candidates
for recession predictors in the US context and conclude that the yield-curve slope is the
single most powerful predictor of recession. Surprisingly, there is very little work on
this issue in the context of emerging markets such as India. In carrying out this analysis
for India, we begin by laying out the methodology used to date business cycles. Next,
we specify the long-term and short-term interest rates which we will use to produce
the spread that constitutes our measure of steepness. Finally, we provide an empirical
model which uses the yield spread to predict the onset of a recession.
2.1 Defining Recessions
Our work in this section essentially follows Pandey et al. (2017). We use seasonally
adjusted quarterly GDP series from 1996 Q2 to 2014 Q3 from India to arrive at the
chronology of business cycles. There are broadly two approaches for measuring business
cycles. The classical approach identifies expansion and contraction based on the level of
output. By contrast, the “growth rate cycle”method identifies turning points based on
the growth rate of output. Following much of the literature on dating business cycles
in India, we use the growth cycle approach as this is deemed the most appropriate for
the post-reform period in India. Following Pandey et al. (2017), the trend is extracted
4
using the band-pass filter developed by Christiano and Fitzgerald (2003) and the cycles
are extracted using the NBER business cycle periodicity of 2-8 years. The Harding-
Pagan (2002) algorithm is then applied to the detrended data to date the periods of
peaks and troughs in a business cycle. Using this approach, the specific periods during
which the economy slipped into a downturn were (1)1999 Q4 to 2003 Q1 (2) 2007 Q2
to 2009 Q3 and (3) 2011 Q2 to 2012Q4.
2.2 Measuring the spread
In choosing the most appropriate rates to measure the yield spread, the literature has
largely focused on factors such as the availability of historical data and consistency
in the computation of rates over time.1 For the US, the Treasury rates easily meet
these criteria and are therefore the most popular candidates for constructing the yield
spreads. In terms of the maturity combination, the 10-year rate is the preferred choice
as the series has consistently computed data available over a long period. With regard to
the short-term rate, the literature has found that the three month Treasury rate, when
used together with the ten-year Treasury rate, is effective in predicting US recessions.
Since our objective is to examine monetary policy spillovers, we follow the literature in
measuring the US yield spread as the difference between ten-year and the three-month
Treasury rates. For comparison purposes, the Indian yield spread is constructed based
on the same combination of maturities using Indian treasury yields.
1Our discussion in this section largely follows from Duecker (1997).
5
2.3 Slope of the yield curve and recessions
In this section, we present a chronology of business cycles in the period 1996 Q2 to
2014 Q3 and examine the effectiveness of the US and Indian yield spreads in predicting
recessions in India. Are recessions in India preceded by a sharp decline in the slope
of the US and Indian yield curves? Figure 2 examines the link between the US yield
spread and recessions in India. The yield spread is measured as the difference between
the yields on 3-month and 10-year US Treasury securities for 1996 Q2 to 2014 Q3. The
shaded regions indicate recession periods for India using the methodology discussed
earlier.
It is clear from Figure 2 that two of the three Indian recessions since 1996 Q2 were
preceded by a flattening of the US yield curve or a large decline in the yield on 10-year
US Treasury securities relative to the yield on 3-month US Treasury securities. The
only occasion when the US yield curve flattened without a subsequent slow down in
economic activity was the recession beginning 2011 Q2. Figure 3 plots the movement of
Indian yield spread and Indian recessions. Unlike the case of the US yield spread, most
Indian recessions were not preceded by a flattening in the Indian yield curve. It would,
therefore, appear the US yield curve is more effective in predicting Indian business
cycles. In the next section, we examine this hypothesis formally by constructing a
model that translates the steepness of the US and Indian yield curves at the present
time into the likelihood of a future Indian recession.
6
2.4 Predictive power of the yield curve
Papers in the US context that examine the predictive power of the yield curve have
typically estimated probit models of the form
Prob(Rt) = Φ(c0 + c1SPRt−k) (1)
where Rt is a 0/1 indicator variable that equals 1 if and only if there is an NBER-dated
recession at time t. SPRt−k is a measure of the slope of the yield curve in period t−k,
and Φ denotes the standard normal cumulative distribution function. Intuitively, the
simple probit model uses the normal distribution to translate the term spread in period
t− k into a probability of a recession at some point t in the future.
Here, we follow the same methodology to estimate the probability of an Indian
recession 12 months ahead using both Indian and US yield spreads as the explanatory
variables. The model is estimated using data from 2001 Q1 to 2014 Q3. Table 1 presents
the results for the case where the Indian yield spread is the explanatory variable.
The insignificant coeffi cients suggest that Indian yield spread is not very effective in
predicting recessions. Figure 4 which plots the probability of a recession 12 months
ahead using the Indian yield spreads indicates that the estimated probability of a
recession did not increase significantly before any of the documented recessions. Next,
we examine the case where we use the US yield spread as the independent variable. This
model is estimated using data from 1996 Q2 to 2014 Q3. In contrast, Table 2 indicates
that the US yield spread is an effective predictor of Indian recessions. Furthermore,
when we convert the US yield spread into a probability of recession twelve months
7
ahead using the probit model, the estimated probability of recession rose close to 60
percent before both the 1999 Q4 to 2003 Q1 and 2007 Q2 to 2009 Q3 recessions. Figure
5 illustrates this point. The US yield curve, however, failed to predict the 2011-12
recession where the estimated 12-month ahead probability actually declined before the
onset of the recession. One could argue that the quantitative easing program pursued
by the Federal Reserve during this period where it aggressively intervened at the long-
end of the yield curve broke the conventional monetary policy transmission channel
rendering the yield curve ineffective as a recession predictor.
In the next section, we write down a simple model which we use to investigate
the pattern of interest rate correlations along different points on the yield curve. The
framework also enables us to contrast the effectiveness of the US and Indian yield
spreads in forecasting economic activity in India.
3 The model
In this section, we reproduce the model of Kulish and Rees (KR from now) (2011) who
extend Gali and Monacelli’s (GM from now) (2005) workhorse small open economy
framework to incorporate a term structure of interest rates. Our objective is to estimate
this model using Indian data and examine the conditions under which the workhorse
small open economy model is able to match the two empirical stylized facts pointed
out about the Indian yield curve.
In this setup, two asymmetric countries, a small open economy, and a rest-of-the-
world foreign economy coexist. The small open economy consists of households that
8
supply labor to firms and consume a composite good. Households have access to
complete asset markets. Firms hire labor, produce, and sell differentiated products in
monopolistically competitive goods markets. Households maximize the present value
of expected utility, and firms maximize profits.
In what follows, lower case letters denote log deviations from steady state, i.e.,
x ≡ ln (X) − ln(X̄), and the superscript ∗ is used to denote variables of the foreign
country. As in KR, we focus on the log-linear aggregate equations. We refer the reader
to KR and GM for details of the derivations.
3.1 Large Economy
The foreign economy block can be summarized by aggregate demand, aggregate supply,
monetary policy rule and an exogenous process for demand and supply shocks. The
aggregate demand for the foreign economy is given by
y∗t = Et y∗t+1 −
1
σ
(R∗1,t − Et π∗t+1
)+
1
σ
(1− ρ∗g
)g∗t (2)
where R∗1,tis the foreign short-term nominal interest rate; π∗t is the foreign inflation rate;
and σ is the intertemporal substitution; g∗t is the demand shock and ρ∗g its persistence.
The term structure of interest rates which follows from the expectations hypothesis is
given by
R∗m,t =1
mEt
m∑j=1
R∗1,t+j−1 m = 2, 3, 4 (3)
The expectational Phillips equation for the large open economy is given by
π∗t = βEtπ∗t+1 + λ(σ + ϕ)y∗t − λ(σ + ϕ)a∗t (4)
9
where λ ≡ (1−θ)(1−βθ)θ ; and θ is a measure of the degree of price stickiness: ϕ measures
the elasticity of labor supply and β is the discount factor and a∗t is the technology
shock. Monetary policy in the foreign economy is assumed to follow a Taylor rule of
the form
R∗1,t = ρ∗rR∗1,t + α∗ππ
∗t + α∗yy
∗t + ε∗r,t (5)
where ε∗r,t is an iid foreign monetary shock with mean zero and standard deviation σ∗εr .
The technology shocks a∗t and demand shock g∗t follow an AR(1) process given by
a∗t = ρ∗aa∗t−1 + ε∗a,t (6)
g∗t = ρ∗gg∗t−1 + ε∗g,t (7)
where ε∗a,t, ε∗g,t are zero mean iid disturbances with standard deviations given by σ
∗εa
and σ∗εg , respectively.
3.2 Small open economy
The consumption Euler equation for the small open economy is given by
ct = Etct+1 −1
σ(R1,t − Etπt+1) +
1
σ(1− ρg) gt (8)
It is easy to see by iterating equation (8) forward that consumption in the economy
is a function of the long-term real interest rates. Under the expectations hypothesis,
a change in the short-term policy rate impacts the long-term nominal interest rate
which in the presence of price stickiness influences the long-term real interest rate. The
10
market clearing condition in the goods market is given by
yt = ct +αω
σst (9)
where ω = στ+(1−α)(σι−1), τ is the intratemporal elasticity of substitution between
foreign and domestic goods, and ι is the elasticity of substitution across varieties of
foreign goods. The parameter α is the share of foreign goods in the consumption basket
and therefore serves as a measure of openness. The terms of trade is denoted by st,
which is defined is the ratio of foreign prices expressed in terms of domestic currency
(et+p∗t ) and domestic prices (pH,t). Given that the law of one price holds, we can write
the terms of trade as
st = et + p∗t − pH,t (10)
The log-linearization of the CPI around a steady state with PH = P ∗ yields
pt = pH,t + αst (11)
It follows that the CPI inflation, πt, in the economy is given by
πt = πH,t + α∆st (12)
The price inflation, πH,t, for domestically produced good is given by the expectational