Ebook Sovereign spreads, currency crises, and fundamentals: A non-linear analysis
In this paper, I use the framework developed in Boschi and Goenka (2007) to show that the pattern of sovereign bond spreads in emerging markets is affected by expectations on exchange rate dynamics and, more specifically, by the probability of a currency crisis occurring. This allows to reconcile the behaviour of economic fundamentals with that of sovereign spreads over periods of financial distress. Previous literature suggests that this relationship is non-linear, breaking down during crisis periods, thus casting doubts on the efficiency of international financial markets. The model of Boschi and Goenka (2007) can account for such non-linearity explaining it as the outcome of rational investors optimizing behaviour, allowing at the same time multiple equilibrium solutions. Further, I show how this model can be implemented empirically through the specification of a Vector Autoregression (VAR) model subject to Markov-switching (MS) regimes.
Given the growing market size of emerging countries’ bonds, understanding the dynamic behaviour and the determinants of sovereign spreads has important implications for research as well as for portfolio management. However, while the literature has devoted a lot of attention to emerging markets’ equities (e.g. Harvey (1995), Bekaert and Harvey (1995)), only recently research has leaned towards bond returns. The resurgence of capital flows to developing countries after the debt crisis of the mid-eighties has renewed the interest in sovereign bonds, one of the main instruments of the recent international financial integration.
Early studies on emerging markets’ spreads focused on bank loans. During the nineties these were the almost exclusive channel of access to international capital markets for developing countries. In a seminal paper, Edwards (1984) developed a framework commonly used in subsequent research on the valuation of sovereign risk premium. Emerging markets are modelled as small borrowers on perfectly competitive financial markets. Their fair value spreads are determined by the probability of default, which in turn depends on domestic macroeconomic fundamentals and on external shocks. Edwards (1984) applies this framework to estimate the determinants of primary yields on bank lending to emerging countries, while Edwards (1986) extends the analysis to estimate separately the default risk premia on international bank loans and bond markets. His results support the presumption that bonds are riskier than bank loans.
A subset of the literature on sovereign spreads is concerned on comparing the relative importance of domestic and foreign factors. Fernandez-Arias (1996), for example, finds that US government bond yields have a larger impact than domestic fundamentals. Using the same methodology as Edwards (1986), Min (1998) investigates the primary bond market finding that the terms of trade are the only external factor affecting spreads, while the role of US short-term interest rates and the oil price is insignificant. Kamin and Kleist 1999) study the influence of borrower’s creditworthiness, as measured by the credit ratings issued by the major rating agencies, on bond and loan launch yield spreads. They find that creditworthiness is an important determinant of emerging markets’ spreads, while they cannot identify any robust and statistically significant relationship between sovereign spreads and industrial countries’ interest rates. By contrast, Eichengreen and Mody (1998) find that the US government bond yield is negatively related to the issuance of sovereign bonds by emerging countries and, through the larger supply, to sovereign spreads.
Dungey et al. (2003) look at the impact of global risk aversion on emerging market debt in several crisis events. By decomposing global risk aversion in three factors, volatility, credit, and liquidity risks, they find that the Russian crisis was characterized by a sharp increase in global credit risk, while the relative size of global risk factors is mixed for the Brazilian crisis.
Jostova (2006) develops an error correction model to investigate the predictability of changes in Brady bond spreads as a function of a broad set of macroeconomic variables reflecting economic conditions and, in turn, the default probability of local governments. She finds evidence of return predictability in Brady bonds returns through a number of approaches to the evaluation of out-of-sample performance of the model. An important predictor is the deviation from the longrun equilibrium between spreads and the underlying economic fundamentals, while controlling for contagion effects from other emerging markets and the correlation with equity markets does not produce the expected results.
Ferrucci (2003) studies empirically the determinants of emerging markets’ sovereign spreads in order to disentangle the relative contribution of country specific fundamentals while controlling for global interest rates, market risk and liquidity in bond markets. One of the interesting findings of his paper is that while sovereign spreads reflect broadly market fundamentals, non-fundamental factors play a more important role. Ferrucci interprets this result as evidence against the efficient market hypothesis, according to which asset prices always reflect the information publicly available as processed by rational investors. By comparing the market-based spreads against their fundamental based counterparts he concludes that the generalized fall in emerging markets spreads between 1995 and 1997 was due to capital market imperfections as reflected in higher investors’ risk appetite resulting from lower global interest rates. The largest misalignments between actual and predicted spreads are reported for countries which are experiencing a financial crisis. Ferrucci comes to the conclusion that the break-down of the fundamental relationship during crisis periods represents evidence of market inefficiency. By contrast, in this paper I show that this can be the outcome of the optimization behaviour of a fully rational international investor with decreasing risk aversion in absence of market frictions.
Boschi and Goenka (2007) provides a theoretical framework to analyse the interaction between currency crises and investors’ beliefs in the context of a second generation model of currency crises a la Jeanne (1997). It shows that the devaluation probability of pegged currencies affects international investors’ wealth via portfolio returns. If investors are risk averse, a decrease in wealth will increase the risk premium they require on assets and this, in turn, will increase the government’s cost of defending the peg. For certain values of the fundamentals and parameters, multiple equilibria and self-fulfilling crises may arise as a consequence of inherent circularity investors’ formulate rational expectations on the government commitment to defend the peg and, conversely, the government’s behaviour is derived optimally conditioning on expected investors’ beliefs. This argument helps to understand that the relationship between sovereign spreads, as a proxy for the country-specific risk premium, and fundamentals weakens when the pressure exerted by investors on the foreign exchange market increases.
Another contribution of this paper is methodological in nature. I show that a second generation model of currency crises with multiple equilibria has in a MS-VAR model a natural empirical counterpart. Jeanne and Masson (2000) characterize the equilibria of a reduced form version of a class of “escape clause” or “second generation” models of currency crises. They identify the conditions under which multiple equilibria can arise and those under which an arbitrarily large number of equilibria, rather than only three as in most other models, can occur. Most importantly, they show how this class of models can be consistently brought to data through the use of Markov-switching regimes econometric techniques, in which the switch across regimes corresponds to jumps between different equilibria. They provide an illustration of this modelling strategy by applying it to the 1993 French franc crisis and find that allowing for sunspots to influence the devaluation probability improves the model performance.
The results of the estimation of the MS-VAR as well as several other univariate models of Brazilian sovereign spreads support the hypothesis of a non-linear relationship between spreads and their fundamental determinants due to expectations about currency devaluations. The rest of the paper is structured as follows. Section 2 sketches the theoretical framework. Section 3 describes the empirical methodology. Section 4 presents the results, and Section 5 concludes.
Download
PDF Ebook Sovereign spreads, currency crises, and fundamentals: A non-linear analysis
Posted in :
