Wednesday, October 8, 2008

On Turkish Treasury benchmark rates (I)

It turns out that the most popular question I get is also my least favorite one, many times I've felt the urge to declare that if I (or anyone else for that matter) knew where the interest rates or the dollar were heading, I would have made my millions and stay retired for the rest of my life. Anyway, I just couldn't say no to my friend who asked about the Turkish benchmark this morning; besides I need to be able to have a good grasp of the behavior of Turkish interest rates for a consulting project I am about to start for a bank (more on that later), and the benchmark rates seemed like a good place to start, so here's my first cut at Tbill rates:

I modeled benchmark rates as function of 2-year Turkish Credit Default Swap spreads (a proxy for default risk), CBT policy rates, inflation expectations and real US 2-year government bond rates (a proxy for risk free rates). Since I was more interested in long-term relationships, I used the autoregressive distributive lag (ARDL) model for my estimations. The results are in the table below:

CDS turns out to have the most pronounced effect on the benchmark. A 10bp increase in CDS increases benchmark rates bu 15.1bp. Similarly, 10 bp rise in inflation expectations, the policy rate and real US rates increase the benchmark by 7.1bp, 4.9bp and 4.4bp, respectively.

The model does a pretty good fit for time time period (January 1 2003, October 7 2008). Of more relevance to my friend's question, by plugging in all the variables, I get that the current level of the benchmark (around 20.90% compound) is 70bp lower than implied by the model. However, given the volatile environment in markets right now (CDS tightened while I was writing this blog entry after the joint Central Bank rate cut announcement), Turkish Tbills and bonds are likely to go where the global headwinds are to take them in the short-run.

In fact, as interesting as the preliminary results are, the short-run is really what my friend was asking. Besides, it would be interesting to see the interaction between stocks, Tbills and the FX market. So, the next step would be to use vector autoregressions (VARs) or more likely a vector error correction model (VECM) and also look at the direction of causality between the different markets. But I still haven't had the time for my daily run (too much newswatching), so that will have to wait for now...

No comments: