Forecasting local currency bond risk premia of emerging markets: The role of cross‐country macrofinancial linkages

• Published date: 01 September 2020
• DOI: http://doi.org/10.1002/for.2669
• Date: 01 September 2020
• Author: I. Ethem Güney,M. Yilmaz,Oguzhan Cepni,Rangan Gupta

Received: 5 September 2019 Accepted: 12 January 2020

DOI: 10.1002/for.2669

RESEARCH ARTICLE

Forecasting local currency bond risk premia of emerging

markets: The role of cross-country macrofinancial linkages

Oguzhan Cepni1Rangan Gupta2I. Ethem Güney1M. Yilmaz1

1Central Bank of the Republic of Turkey,

Ankara, Turkey

2Department of Economics, University of

Pretoria, Pretoria, South Africa

Correspondence

Oguzhan Cepni, Central Bank of the

Republic of Turkey,Haci Bayram Mah.

Istiklal Cad. No. 10 06050 Ulus, Altndag,

Ankara, Turkey.

Email: Oguzhan.Cepni@tcmb.gov.tr

Abstract

In this paper, we forecast local currency debt of five major emerging market

countries (Brazil, Indonesia, Mexico, South Africa, and Turkey) overthe period

January 2010 to January 2019 (with an in-sample period: March 2005 to Decem-

ber 2009). We exploit information from a large set of economic and financial

time series to assess the importance not only of “own-country” factors (derived

from principal component and partial least squares approaches), but also cre-

ate “global” predictors by combining the country-specific variables across the

five emerging economies. We find that, while information on own-country fac-

tors can outperform the historical average model, global factors tend to produce

not only greater statistical and economic gains, but also enhance market timing

ability of investors, especially when we use the target variable (bond premium)

approach under the partial least squares method to extract our factors. Our

results have important implications not only for fund managers but also for

policymakers.

KEYWORDS

bond risk premia, emerging markets, factor extraction methods, out-of-sample forecasting

JEL CLASSIFICATION

C22C53C55G12

1INTRODUCTION

Local currency sovereign debt of emerging markets is a

government bond denominated in the domestic currency

of an emerging market issuer. Emerging countries have

started to rely increasingly on their domestic debt markets

for financing their expenditures, and hence have issued

bonds in domestic currencies. Further, these local cur-

rency bonds of emerging markets have caught the atten-

tion of global investors aiming to diversify their portfo-

lios with the relatively higher yields, on average, derived

from these bonds in comparison to their developed market

counterparts. Naturally, the massive size of the local debt

markets of emerging economies, which as of in 2017 stood

at USD 21.9 trillion(International Monetary Fund, 2018),

is hardly surprising. Understandably, it is of paramount

importance to investors, as well as policymakers aiming to

finance their expenditures, to determine the factors that

drive the future path of local currency bonds of emerging

markets.

Theoretically, the yield on a long-term nominal govern-

ment bond can be expressed as the sum of expectations

of future short-term rates over the maturity of the bond

and a maturity-specific term premium. Since long-term

bonds have a greater duration to maturity than short-term

debt, investors typically demand a risk premium. In this

regard, although a large number of studies have inves-

tigated the determinants of bond premia for advanced

Journal of Forecasting. 2020;39:966–985.

wileyonlinelibrary.com/journal/for© 2020 John Wiley & Sons, Ltd.

966

economies (see; Cepni, Demirer, Gupta, & Pierdzioch,

2019; Cepni, Gupta, & Wohar,2019; Cochrane & Piazzesi,

2005; Gargano, Pettenuzzo, & Timmermann, 2019; Ghy-

sels, Horan, & Moench, 2018; Laborda & Olmo, 2014;

Ludvigson and Ng, 2009, 2011; Zhu, 2015),1the corre-

sponding literature on emerging market local currency

bonds is scarce, with the existing studies primarily dealing

with in-sample predictability (see, e.g., Akgiray, Baronyan,

Sener,& Yilmaz, 2016; Cepni, Gul, & Gupta, 2019; Cepni &

Güney,2019; Gadanecz, Miyajima, & Shu, 2018; Miyajima,

Mohanty,& Chan, 2015). Miyajima et al. (2015) show that,

while resilient to global risk aversion shocks, forecasts of

the domestic short-term interest rate, output growth, and

the fiscal balance explain a large part of the local currency

bond yields of emerging markets. Akgiray et al. (2016)

looked at predictability of local currency excess bond

returns in the emerging markets of Brazil, Mexico, South

Africa, and Turkey. As in Ludvigson and Ng (2009, 2011),

using a dynamic factor approach based on a large panel of

economic and financial time series, these authors detected

strong predictable variation in bond premia derived from

macroeconomic activity. Gadanecz et al. (2018), based on

a large sample of emerging countries, found that when the

volatility and expected depreciation of the exchange rate

increased, investors required a larger yield compensation

for holding local currency bonds. While Cepni and Güney

(2019) highlight the role of credit ratings besides macroe-

conomic and financial (including exchange rate volatility)

variables, Cepni et al. (2019) document the importance of

uncertainty related to economic policies in driving local

currency bond premia of emerging countries.

Realizing that real-time forecasts are more important

for fund managers than information of in-sample pre-

dictability (Welch & Goyal, 2008), and also the fact that

the ultimate test of any predictive model (in terms of

the econometric methodologies and predictors used) is

in its out-of-sample performance (Campbell, 2008), we

build on the in-sample-based evidence of Akgiray et al.

(2016) in several directions when conducting a full-fledged

out-of-sample forecasting exercise. First, we not only

extract common factors from a large data set using prin-

cipal component analysis, but we also rely on partial least

squares. The latter approach also constructs a set of latent

factors from a large set of variables, but unlike principal

components analysis it estimates factors that are specifi-

cally valuable for forecasting a given target, which in our

case is the local currency bond premia of emerging mar-

kets. Second, when extractingfactors, we not only consider

“own-country” variables but also create factors by combin-

1In this regard, important earlier studies are those of Keim and Stam-

baugh (1986), Fama and Bliss (1987), Fama and French (1989), and

Campbell and Shiller (1991).

ing the country-specific variables across the five emerging

economies. This resulting set of “Global” predictors, and

its global subsets of “Macroeconomic” and “Financial” fac-

tors, are used individually to specify new factors, which

are then combined with the own-country factor model.

Note that the motivation for doing this emanates from the

widespread evidence of spillovers (comovements) across

sovereign bond markets of emerging countries due to com-

mon underlying cross-country factors (Bunda, Hamann,

& Lall, 2009; Subramaniam & Prasanna, 2017; Subra-

maniam, Prasanna, & Bhaduri, 2016). Third, given that

forecasts for which conventional prediction error statis-

tics outperform the benchmark models might not result

in profitable investment strategies, we employ a direc-

tional accuracy test that analyzes market timing ability

by constructing the hit ratio with the assumption related

to the probability of independence between predictions

and realizations. Fourth, since directional predictive abil-

ity does not ensure economic significance, we also analyze

the economic value of active trading strategies formed on

the local currency bond risk premium forecasts by uti-

lizing the setting for a mean–variance investor aiming to

optimally allocate wealth across risky and risk-free instru-

ments using a utility-based metric. Finally, as a minor

addition to the work of Akgiray et al. (2016), we also

include Indonesia in our analysis besides Brazil, Mexico,

South Africa, and Turkey, as analyzed by these authors.

Note that we select these five major emerging sovereign

bond markets by notional amount outstanding.2These

countries, as pointed out by Akgiray et al. (2016), share

three essential features: (a) they belong to the J. P. Morgan

Government Bond Index—Emerging Markets (GBI-EM),

which is an investable indexfor emerging market local cur-

rency bonds; (b) they have large and liquid local currency

bond markets in which search and trading costs are low;

and (c) they offer long-term local bonds. Moreover, as of

the first quarter of 2019, these five economies comprise

26.7% of the total market size of local currency bonds of

(18)3emerging markets in the GBI-EM index (Debt Secu-

rities Database, Bank of International Settlements), with

Mexico (10.3%) leading the pack and followed by Turkey

(5.7%), Indonesia (4.7%), Brazil (4.5%), and South Africa

(1.5%).

To the best of our knowledge, this is the first paper to

incorporate the role of local and global factors in forecast-

2Based on the Debt Securities Database of the Bank of International Set-

tlements, the values in billions of USD in the first quarter of 2019 stood

at 119, 123, 271, 38, and 151 for Brazil, Indonesia, Mexico, South Africa,

and Turkey,respectively (availableat https://www.bis.org/statistics/full&

urluscore;data&urluscore;sets.htm).

3The remaining 13 countries are Chile, China,Colombia, Hungary, India,

Malaysia, Nigeria, Peru,Philippines, Poland, Romania, Russia, and Thai-

land.

CEPNI ET AL 967