عوامل مؤثر بر میزان بازده مستقل یورو در منطقه : یک رویکرد بیزین

We analyze the determinants of sovereign yields spreads of EMU member states applying Bayesian Model Averaging (BMA) to annual panel data from 1999 to 2009. BMA is well-suited in cases of small samples and high model uncertainty. This seems to be the case in modeling sovereign yield spreads in the Eurozone since the literature reports heterogeneous results with respect to significant explanatory variables. We are testing a number of variables reported to be significant in the literature and find that the most likely country specific drivers of yield spreads are fiscal variables such as budget balance and government debt, as well as external sector variables, such as terms of trade, trade balance and openness. Global financing conditions, indicated by the US interest rate, and market sentiments, indicated by corporate bond spreads, are likely to influence sovereign yield spreads.

The paper aims to provide answers to the question of what drives sovereign yield spreads of EMU countries’ government bonds, which is an important issue in the current political debate about the further development and even the survival of the Euro and the Eurozone. We analyze potential determinants of sovereign yield spreads of EMU member states (to German bond yields) observed on secondary bond markets. These yield spreads result from several reasons. Apart from default risk, the yield spreads are influenced by liquidity risk and market sentiments toward investments in risky bonds. We examine the relation between spreads and several variables related to these causes using Bayesian Model Averaging (BMA). We particularly focus on the default risk component and test a variety of variables related to this issue, however, we also control for the other issues. Identifying the determinants of sovereign yield spreads is an important research question because it helps to understand which factors determine countries’ capital costs. For example, we can analyze whether interest rates paid for borrowing capital depend on countries’ fiscal discipline, (certain drivers of the) competitiveness of the economy, global financing conditions or even market sentiments. A broad and interesting literature exists that deals with determinants of sovereign bond yield spreads. However, the results of these studies are rather heterogeneous, i.e. different papers report different variables as the main drivers of spreads.1 This may be due to differences in econometric models, country samples, observation periods and variables considered. Edwards, 1986, Eichengreen and Mody, 1998 and Kamin and von Kleist, 1999, and Min (1998), e.g., consider primary (market) spreads of several developing countries, i.e. spreads observed when bonds are issued. Arora and Cerisola (2001) perform individual regressions for time series of secondary market spreads of observed countries, whereas Cantor and Packer (1996) consider cross-section data. Examples of papers that employ panel data on developing countries’ secondary market yield spreads are: Baldacci et al., 2008, Dailami et al., 2005 and Hilscher and Nosbusch, 2010, and Rowland and Torres (2004). While the studies mentioned so far exemplify the bulk literature where the determinants of yield spreads for developing countries are estimated, several papers consider the spreads for developed countries. Bernoth et al. (2006) provide a detailed literature review of OECD countries in general and Gale and Orszag (2002) of the US in particular. Examples for other papers dealing with the US are Gale and Orszag, 2004 and Laubach, 2003, and Poterba and Rueben (1999). OECD countries are considered, e.g., in Alesina et al. (1992), and Ardagna et al. (2004). A number of papers analyze yield spreads of EU government bonds. They report some similar but also some very different results. Bernoth et al. (2006) report debt to GDP, deficit to GDP, debt service to revenues, US corporate bond spreads, a EMU dummy, the short-term US rate, as well as liquidity and maturity of the issue as significant. Schuknecht et al. (2009) identify debt to GDP, fiscal balance to GDP, US corporate bond spreads, region dummies, liquidity and maturity of the issue and the short-term US rate as significant determinants of sovereign yield spreads. Schuknecht et al. (2009) confirm these results, but in addition, they find crises and turmoil dummies to be significant. Whereas these papers consider spreads observed for single bond issues, others, such as Codogno et al., 2003 and Lemmen and Goodhart, 1999, and Manganelli and Wolswijk (2007), use data for a benchmark curve that is related to a fixed maturity, usually a time to maturity of 10 years is considered. Lemmen and Goodhart (1999) report the debt to GDP ratio, capacity to acquire taxes, inflation, and inflation variability to be significant spread drivers. Codogno et al. (2003) only find debt to GDP ratio, US swap spread and US corporate bond spreads significant. Manganelli and Wolswijk (2007) only report ratings and short-term interest rates as significant spread drivers. The literature reviewed so far provides interesting insights into drivers of yield spreads. However, the results are rather heterogeneous, not only for different samples, but even for rather similar samples, such as the Eurozone countries, which a comparison of results from Lemmen and Goodhart (1999) with those of Bernoth et al. (2006), for example, shows. To some extent, these differences may be a result of different observation periods. 2 Another reason could be, however, that papers in the literature also differ considerably with respect to tested variables. The fact that there is no consensus about the key determinants of sovereign yield spreads may be seen as indication for high uncertainty about the “true” empirical model. One appropriate approach to deal with this model uncertainty is Bayesian Model Averaging (BMA). It explicitly accounts for the high model uncertainty by considering (approximately) the entire model space, i.e. any possible combination of regressors out of a given set of potential determinants. In classical statistics, by contrast, the conclusions are based on just one model (or a small sample of models). Often only a small set of potentially explaining variables is even tested (see the discussion above) or smaller models out of a comprehensive sample are selected using heuristics. However, model selection is problematic given the size of the potential model space and, what is more, information from almost all possible models is neglected. Even testing the full model does not solve this issue due to the multi-co-linearity problem, which in particular implies that we may reject variables by mistake. This is particularly an issue for small samples as it is the case in the EMU sample. By considering the entire model space BMA is supposed to provide more solid information about the determinants of spreads than classical regressions. BMA is applied successfully to various other topics. Examples are economic growth (see, e.g., Eicher et al., 2007, Fernandez et al., 2001b, Masanjala and Papageorgiou, 2008 and Sala-i-Martin et al., 2004), monetary policy (see, e.g., Hineline, 2007, and Milani, 2008), the relationship between consumption and wealth (see, e.g., Koop et al., 2008), and pricing of stocks (see, e.g., Avramov, 2002, and Cremers, 2002) or hedge funds (see Vrontos et al., 2008). We contribute to the literature by applying BMA to the issue of sovereign yield spreads in the member states of the Eurozone. We test a number of hypotheses on causes of these spreads and include related variables reported to be significant in the literature in our analysis. In the next section we overview the BMA approach. In Section 3, we formulate the hypothesis with respect to causation of yield spreads tested in our empirical analysis and discuss the variables used to indicate these hypotheses. Section 4 deals with the setting of our empirical analysis and discusses the results. Section 5 concludes.

Considerable default risk of several EMU member states is an important issue because it causes financial turmoil in the Eurozone, imposes high costs for all member states, and is even a threat for the future existence of the EMU. Especially the market perception of default risk is important, and high interest rates for government bonds are a major problem for a number of EMU member states. An interesting strand of the empirical literature analyzes this issue by regressing several potentially explaining variables on government bond yield spreads. These papers provide interesting insights into the determinants of sovereign yield spreads in general and into the nature and causes of country default risk in particular. However, the results are heterogeneous, even in papers that focus on EU countries, which points to some degree of model uncertainty. We contribute to this literature by tackling the issue with Bayesian Model Averaging (BMA), which has been successfully applied in other cases, but not in regards to the issue of EMU sovereign yield spreads. BMA is an elegant approach to test a variety of potentially explaining variables and to deal with model uncertainty. It uses information from the entire model space and not only a single model (or some selected models), as it is the case in classical regressions, to identify the best models and assess the quality of potential regressors. In this sense, BMA provides additional and broader information. Therefore our paper complements the existing literature. We apply BMA in several settings and consider different time spans in order to carefully analyze the issue and provide robust results. We confirm some important findings of the literature, disprove others and provide mixed evidence for some. By considering 10 EMU member countries in the observation period 1999–2009, we find that the most important country specific drivers of sovereign yield spreads in the Eurozone are budget balance to GDP, terms of trade, trade balance and countries’ openness. Budget balance and terms of trade are also significant in lagged estimations, i.e. they could be used as early warning signals. In order to discern whether our findings are driven by the crisis period or also hold true in tranquil periods, we run separate estimations for the pre-crisis period 1999–2007. Budget balance to GDP and the trade balance are very likely to influence spreads in the pre-crisis period as well as in the entire sample period, whereas openness and terms of trade play no role in the pre-crisis period. In addition, the level of debt to GDP is a very likely driver of yield spreads in the pre-crisis period, but not in the entire period where the change of debt indicated by the budget balance is important only. The importance of these fiscal variables indicates that financial markets watch fiscal variables indeed closely, and thus impose fiscal discipline by increasing interest yields when the fiscal situation deteriorates and the default risk increases. In contrast to these spread drivers that are related to default risk, indicators for market depth and liquidity yield rather low inclusion probabilities for both the pre-crisis sample and the entire sample. In contrast, global financing conditions, indicated by the US interest rate, and global market sentiments, indicated by spreads of US BBB corporate bond spreads, are likely to drive spreads at least in some periods. In regards to country specific drivers of sovereign yield spreads that are supposed to be influenced by the market perception of default risk, we can conclude that fiscal variables, such as budget balance and debt level (at least in the pre-crisis period), are found to be important determinants of sovereign yield spreads in EMU countries – i.e. they decisively influence the market perception of default risk and the borrowing costs of countries. As a result, avoiding defaults – and perhaps the survival of the EMU – crucially depends on the successful budget consolidation of the member states and the reduction of debt to GDP. The success seems to be partly dependent on – hopefully – favorable conditions in the external sector, since in addition to fiscal variables, variables related to the external sector, as terms of trade, trade balance and openness, are also found to be important drivers of sovereign yield spreads.