Bond risk premia in emerging markets: evidence from Brazil, China, Mexico, and Russia

ABSTRACT

We employ an affine term structure model with no-arbitrage restrictions and unspanned risk factors to analyse the global and domestic determinants of bond risk premia in four major emerging markets (Brazil, China, Mexico, and Russia). Among the risk factors, we select national inflation and economic growth, and the country-specific nominal exchange rate against the US dollar as the variables related to the domestic economy. We include a measure of worldwide economic activity, the Market Volatility Index (VIX), and an aggregate price index of commodities in the group of global factors. Our model captures (long-term) movements of realized risk premia and indicates that global economic and financial factors play a relevant role in explaining country-specific bond risk premia dynamics. In contrast, domestic variables carry little explanatory power to rationalize risk premia developments in these four economies. We also provide evidence of heterogeneous responses of country-specific risk premia to global and domestic shocks.

KEYWORDS:

• Risk premia
• term structure of interest rates
• international finance
• emerging markets

JEL CODES:

• E43
• E44
• G15

Acknowledgments

We thank the editor Mark Taylor and one anonymous referee. We are also grateful to Bertrand Candelon, Anh Le, Sydney Ludvigson and the participants of the UCLouvain Finance PhD workshop for helpful comments over the development of this article. The authors are responsible for the remaining errors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ This growth rate is the simple average of the yearly growth rates of South Africa, Indonesia, Brazil, Russia, India, China, Korea, Turkey, and Mexico from 2000 to 2019. If we weight the yearly growth of those countries by their relative size, in terms of GDP, the average growth rate is around 6%. Source: 2020 IMF world’s economic outlook and Organization for Economic Cooperation and Development.

² In February 2020, S&P Global estimates that the top 20 EMs are expected to issue a combined $1.62 trillion of government debt in 2020, up 4% from 2019 and a historical high; see https://www.spglobal.com/en/.

³ According to the World Bank, in 2018, the trade-to-GDP ratios of Brazil, China, Mexico, and Russia were 29%, 38%, 80%, and 51%, respectively. Furthermore, Brazil and Russia are considered as export-commodity-dependent countries while China and Mexico are not. A country is considered to be export-commodity-dependent when more than 60% of its total merchandise exports are composed of commodities (UNCTAD 2019). Although there are interest rate databases for other emerging economies, their coverage in terms of time series length and maturity spectrum is rather limited. For this reason, we do not include India and possibly other EM countries in our sample.

⁴ The Cochrane and Piazzesi (2005) factor is obtained by regressing 1-year excess returns across maturities at each time t on five forward rates. They show that a tent-shaped linear combination of forward rates predicts annual excess returns on one- to five-year maturity bonds with $R^2$ up to 0.44.

⁵ Dai and Singleton (2000) provide a complete characterization of the affine class of term structure models, specifying the conditions for admissibility and identification of such models.

⁶ Other studies in this line of research include Bekaert, Cho, and Moreno (2010), Dewachter and Lyrio (2006) and Rudebusch and Wu (2008), among others. In these papers, the factors either have a clear macroeconomic interpretation or are structural in nature. Gürkaynak and Wright (2012) and Piazzesi (2010) provide a detailed survey of the literature.

⁷ This methodology has been used in different contexts by Bauer and Rudebusch (2020) and Dewachter et al. (2015), among others.

⁸ See Ang and Piazzesi (2003) for a detailed derivation.

⁹ See Bauer and Rudebusch (2020).

¹⁰ Since the Bloomberg data are for par yields, we bootstrap the original data to obtain zero-coupon yields. The 1-month yield for Russia, Mexico and China are obtained by estimating and fitting a Nelson-Siegel model.

¹¹ These are derivative instruments with the underlying values linked to Brazil’s interbank lending rate. For a detailed explanation regarding the use of such instruments, see Vicente and Tabak (2008).

¹² The Killian index is based on variations in global shipping freight rates and is tailored to capture fluctuations in the demand for industrial commodities in the global economy.

¹³ Traditionally, term structure works resort to three principal components to account for roughly all of the cross-sectional variation in bond yields, see e.g. Litterman and Scheinkman (1991). As the inclusion of six unspanned factors renders the model dimension to be fairly large, we limit our framework to contain only two spanned factors. Yet, we show that the first two PCs explain at least 95% of yield variations in all countries (see Table 3). Previous multi-country term structure models, as the one of Jotikasthira, Le, and Lundblad (2015), also make use of two principal components to fit bond yield cross-sections.

Additional information

Funding

This work was supported by the National Council for Scientific and Technological Development (Conselho Nacional de Desenvolvimento Científico e Tecnológico - CNPq) [308310/2016-0]; Fonds De La Recherche Scientifique - FNRS [EQP U.N006.18, PDR T.0138.15].