Libor Manipulation: This is meant to be a discussion on calculating the extent of a particular banks' role in Libor Manipulation.Compare/regress 1M Libor against Bank A's (Bank A is on the Libor Panel) CDS. Assuming an increase in the Bank A's CDS would mean an increase in the rate at which Bank A could borrow at thus increasing their 1M Libor posting for that day.Any other thoughts?

is this a homework assignment ?

No it is not. Thankfully I am done with school, but I was generally thinking how one would prove a banks involvement and extent of Libor manipulation.

QuoteOriginally posted by: nhutnyakLibor Manipulation: This is meant to be a discussion on calculating the extent of a particular banks' role in Libor Manipulation.Compare/regress 1M Libor against Bank A's (Bank A is on the Libor Panel) CDS. Assuming an increase in the Bank A's CDS would mean an increase in the rate at which Bank A could borrow at thus increasing their 1M Libor posting for that day.Any other thoughts?Where is the manipulation in that?

My specific example would not prove manipulation, however that Bank A's libor submission was accurate. However, if Bank A's CDS spread did not change, but their Libor submission (i.e. cost of borrowing based on Bank A's credit) increased or decreased significantly, then one would wonder why their cost of borrowing changed without a change in their credit quality. I know there are other factors affecting borrowing rates, but assume the predominant factor affecting borrowing is credit quality. What I am really wanting to discuss is the way to capture or analyze an individual bank's involvement in libor manipulation.

Well, it would be an interesting project to see what you could reasonably infer based upon the daily Libor panel submissions, the daily CDS spreads of the panel members (and otherless firm-specific explanatories: rates, equity returns, CDS indices, etc ). Do you have all the firm-specific data?

I do have most of the data, 1Y CDS spread back to 2007 and the individual 1M Libor Submissions. I have begun regressing the 1M Libor submission against the respective CDS spread and I am also looking into the 1M Libor submission versus the Actual Libor Rate looking at the var., distribution and R^2. Also trying to figure out a way to compare a specific bank's credit rating against actual Libor and submitted Libor. Any other thoughts?

The first step for me would be some visual data exploration. Perhaps you can post some charts.A scatter plot of the daily Actual Libor Rate vs the daily average CDS spread of the panel members would be useful to start,to see the aggregate relationship, (as well as the two individual time series here).Then, perhaps for each bank, the scatter plot of the difference of their submissions from the Actual Libor Rate vs.the difference of their CDS spread from the average.

I don't think you should directly run the regression like that. In your case, we are assuming that Libor is also risky. Therefore, Libor=true rf + risky rate. CDS spread is also affected by the change in rf and change in risk profile.It is known that nominal rf and inflation display some very persistent movement. If you directly run the reg, the risky rate could be buried under the true rf dynamics. In other words, if you want to study a and b but you regress x+a on x+b, what will happen? This is known as early as Fama and Bliss (1987). They study excess bond return and yld-rf. Another even messier issue is about inflation, people's inflation expectation, inflation risk and whether money is (super-)neutral (i.e., whether money/inflation is just a "veil" on real assets). An estimation of real rf using Kalman Filter is in "Estimation and Test of a Simple Model of Intertemporal Capital Asset Pricing" by Brennan, Wang and Xia.