Linear regression: raw prices VS changes

[Maxim Rastorguev]

dear all, in the course frm1 we usually regress prices or other raw data rather than their changes or returns. I wonder if this right or wrong? The thing is in practice I want to regress a bond yield ( YTM) over level of base rates. Though it seems logical that any bond YTM must be regressed well over the local rates, and it is really so (I actually got a good R squared and logical slope) but when I move to changes ( change in YTM over change in base rate) everything is complete broken. R squared goes to zero as well as slope falls like a knife. No relationship between changes at all! I guess if there is a relationship between raw prices (or rates of yield in my case) there should be also a nice dependence in their changes? Or what do I do wrong?

 

[ami44]

One possible explanation is a time lag between change in base rate and change in bond yield. Over what time interval do you calculate your changes? What base rates do you use?