Help with dummy variable


I have a set of OLS-fixed effects models comprising continuous and dummy variables, with the latter used as control parameters. My models aim to evaluate the effects of loan characteristics on the interest rate levied (dependent var), which has a range between 6 and 2000 basis points.

I am facing an issue with one dummy variable (IsSenior), which indicates whether the loan is considered to be senior or subordinate in quality. As expected, the sign is negative i.e., senior loans come with lower interest rates. But the beta value is in the range of (-570, -650) regardless of the other independent vars, although the 90th percentile of my dependent (interest rate) is only 550 basis points. If i presume this to be correct, then a senior loan should have a lower spread by at least 570 bps over a comparable subordinate loan, and in most cases the interest rate would be negative (550 - 570 = -20), which is not correct.

I am totally confused as to how I can explain this tendency using the other variables. Is there a way I can use other independent variables to explain this effect?

Any suggestion would be much appreciated.
Not sure what to suggest, but firstly can you check mean values of your interest rate for subordinate and senior loans. As you say your 90 percentile is 550 basis points and the max value is 2000, so the top 10% values are quite high. I am guessing that the difference in mean values of interest rate for subordinate and senior loans may be closer to the beta value range...
Another approach I would suggest is to look for outliers in your data.I mean if you are covering let us say 95% or more of your data between 6 and 700 bps, then you may want to get rid of outliers like 2000..