At an SNS-CEPR lunchtime meeting in Stockholm on 12 April Carlo Favero, Professor of Economics at Università Bocconi, presented the results of his recent research with Francesco Giavazzi and Luigi Spaventa, which is reported in CEPR Discussion Paper No. 1330, ‘High Yields: The Spread of German Interest Rates’. Their research attempts to explain the behaviour of interest rates issued by the European ‘high yielders’ – Italy, Spain and Sweden – and in particular to decompose the yield spread (relative to Germany) into exchange rate risk and default risk.

The globalization of financial markets and the rapid growth of public debt in many countries has made the market for government bonds the most important segment of international financial markets. More than ever before, yield differentials between government bonds issued by different states reflect the assessment and the sentiments of market participants regarding countries’ short- and long-term financial prospects. Therefore, such differentials are immediately affected by news, carefully watched in the search for arbitrage opportunities, and viewed by government and monetary authorities as a signal of the credibility of their policies and as a measure of expectations. More recently yield differentials have acquired a specific policy relevance with the Maastricht Treaty, which stipulates that the long-term nominal interest rate on bonds issued by a member state should not exceed by more than two-hundred basis points the yield on the three best performing members in terms of price stability.

In Europe, yield spreads are measured in terms of differences of annualised interest rates with respect to Germany: German yields have been and are the lowest on all maturities and German bonds provide the yardstick of a risk-free asset. Favero’s research was designed to explore the relative importance of local and global factors in the determinations of spreads relative to German yields. In other words, do spreads for high yielders react to common factors independently of country-specific fundamentals, or are country-specific factors relevant in the determination of some portfolios among high yielders?

Favero explained that yield spreads can be decomposed into an exchange rate factor, reflecting expected depreciation (and foreign exchange risk premium, if any) and a ‘country’ or ‘default’ risk component, capturing fears about the future service of the debt. The correct decomposition requires the use of continuously compounded yields: the difference between annualised rates cannot be precisely assigned to the two determinants and, more importantly, is affected by the level of German interest rates. Given the spread, it must be decided which of the two components is measured independently and which is to be treated as a residual. The preferred measure for the exchange rate factor is the differential between fixed interest rates on swap contracts denominated in the currency under consideration and in Deutsche Marks. As a satisfactory candidate cannot be found for an independent measure of the default factor, Favero measured this by the difference between the bond spread and the swap spread.

Favero found that the total spreads on Italian and Spanish government bonds have a common trend, which is not shared by the Swedish spread. There is also a very close association between the Spanish and the Italian exchange rate factors: the average difference between the two tends to revert to zero. Thus, contrary to the conventional wisdom widely accepted in policy discussions about weaker currencies, membership of the ERM is not sufficient to stabilize expectations and therefore does not affect the exchange rate component of the spread.

He also considered the maturity structure of the two components of the spread for Spain and Italy. He conjectured that after the ERM crisis of September 1992 the average expected rate of depreciation is higher in the near future and then declines for more distant horizons: the analysis confirmed this for the high yielders. As for the default risk premium, the conjecture that it should be higher for longer maturities is fully borne out in the case of the 10-3 year differential, while the 5-year segment presents irregularities probably due to the structure of market demand. The effects of the exchange rate factor and of the default premium on the maturity structure of the total spread tend to work in opposite directions. On the whole, the net result is that high-yielders, unless there are sizeable domestic shocks to their default premium, have a maturity structure of their government bonds flatter and more stable than that of Germany.

Favero’s analysis also revealed how the total spread is affected by structural shocks to each economy by estimating two Vector Autoregressive models (VARS): one for Spain and one for Italy. He found that in the long-run total differentials are determined by exchange rate factors. There is evidence of uni-directional causality going from the exchange rate factor to the total yield differential. Such evidence holds for both Spain and Italy, although adjustment to the long-run equilibrium is faster in the Spanish case. This is evidence that in the long run the total spreads for these two countries are dominated by a common ‘international’ component. This finding does not extend to the spread on Swedish bonds, which, even in the long run, appear to be driven by different factors.

For the short-run the analysis suggests that the total yield differential is caused by the exchange rate factor. Moreover, Spanish and Italian exchange rate factors move together. The evidence from the long run seems to suggest two types of shocks: an international shock and a local shock. Favero interpreted the international shock as the structural shock to exchange rate factors and distinguished it from the local shock by imposing the assumption that the local shock has no permanent effect on the total yield differential. The international shocks, although, identified separately for Italy and Spain, are affected in a remarkably similar way by shocks to the dollar/mark exchange rate and to long-term German interest rates. In the short run the Spanish and Italian spreads behave differently – country-specific shocks linger longer in the memory in the Italian total spread than in the Spanish one.

Overall the analysis provides strong evidence in favour of the existence of a common trend for the Spanish and Italian spreads on Bunds. This trend is driven by international factors and is independent of country-specific shocks. Country-specific shocks are only relevant in explaining short-term cycles around the common stochastic trend.