The analysis in our original paper that I commented on in the blog (Spreads and Ratings July 26th) used data to the end of 2006. However, as many readers have commented emerging markets spreads have risen. In this Postscript I present some additional results including an analysis of what might happen to EM spreads if indices of risk continue to rise1.
Is the Risk Effect on Spreads Stable?
First off some people sent us comments that perhaps it is not just that indices of risk such as the US high yield or VIX have fallen through the end of 2006 but that also the relationship between spreads and these indices have changed. One view is that markets have got more globalized and hence these indices now matter more for EM risk. But it might also go the other way. Perhaps the advent of new instruments and new investors means that risk is now spread more efficiently and that changes in one market affect others less.
To see if the relationship has changed we conducted our favorite specification panel regression of spreads against ratings with fixed effects with the US Treasury and the VIX index – in our favorite version, the US High Yield was not significant while the VIX was highly significant – until January 2004. We then added one month of new data and ran the regression again. We then added another month of data and did the regression again and carried on with this recursive estimation until the end of the sample – now June 2007. We graph the coefficient on the VIX also indicating plus and minus 2 standard errors below. The graph shows some evidence that the coefficient on the VIX has increased. There is if anything more of an effect from this general index of risk to EM spreads now than before. But the differences are not (quite) statistically significant2.
As we all know the VIX has been on the rise, especially last week. So what are the implications of the level of the VIX for EM spreads? We use our empirical model to compute the relation between the spread and rating using the results across our whole sample for different levels of the VIX. In Chart 2 we plot the results as an exponential regression (trend line) across the single country estimates grouping them according to rating. The lowest level of the VIX in the graph corresponds to the sample minimum 10.91 at Nov 2006. The next lowest level of the VIX is the sample mean – which is about 21, we are just a little higher than that level today. So the VIX has risen from around 11 to about 21. The next level of the VIX (labeled High) is one standard deviation (about 7) higher than the mean; this gives a value of almost 28. And the highest value is the sample maximum is 44.3, in August 1998.
The graph shows that the level of the VIX has a significant impact on the curve. To give an example if the VIX moves up from today’s level to one standard deviation higher than the mean; 28 then Peru a BB+ risk would rise from a spread of 333 basis points to 539 basis points 3. A higher rated risk such as Chile (S&P rating of A) is of course less affected (the move for a typical A would be from 103 to 155 bpts). Lower rated sovereigns would though be affected much more – a typical B rated sovereign would move from 503 to 752). Finally, if the VIX would reach the sample max of 44.3, then Peru as a BB+ risk would be expected to reach 931 bpts.
This is to be understood as a pure risk effect which we argue affects the relationship between spreads and ratings. Of course if the increase in risk aversion is related to events that also weaken fundamentals then we would also expect some downgrades to occur. Indeed the main point of our paper was to argue that most of the improvement in the fundamentals that drive credit quality (and hence credit quality itself) has been due to global factors or put another way, the significant improvement in country fundamentals has been mostly endogenous to such developments as unprecedented world growth and export prices.