1
$\begingroup$

I am trying to find for each european bond in my database a proper Benchmark to compare them with the Bloomberg benchmarks for bonds.

What i have done so far is to extract a list of all government bonds in Europe and calculate a simple linear formula where I compare 4 fields with each goverment bond:

- Coupon
- Maturity Date
- Issue Date
- Volume

the formula used is:

f(bond,govBond) = abs(bondcoupon - govcoupon)*coefCoupon + abs (bondMaturityDate - gov MaturityDate)*coefMaturity + ...
Where the coef are used to give the same weight to each field

So, to find a proper benchmark, I choose the min value returned by this function (i.e the closest gov bond using these criterias).

The results of this approach is approximately a 25% match with Bloomberg list of Benchmark.

Is this a "convenient" way to benchmark bonds? I know that those 4 criterias are not sufficient (Maybe use Yield points from a yield curve or tenor, there are a lot of fields to choose) but I have chosen those 4 fields to simplify the computing (database contains at least 10000 bonds).

Improve this question
$\endgroup$
2
  • $\begingroup$ don't know why this was downvoted but seems like a perfectly good question to me. I don't have a good answer for you, but my personal experience was that this was quite difficult. I too tried to algorithmically come up with the benchmarks, but it never worked 100%. In the end, the traders just sent me the list of benchmarks on a daily basis and the list is what I used for analytics purposes... $\endgroup$
    – Helin
    Jul 9, 2014 at 22:07
  • $\begingroup$ Thank you for your comment. I believe also that this will not be as easy as I thought. Although, If i can have at least 50%, that would be great! $\endgroup$
    – Mehdi
    Jul 10, 2014 at 8:34

1 Answer 1

Reset to default
1
$\begingroup$

You represent your bond as a vector of 4 equivalently weighted parameter and try to find the optimal representation for the 1-norm.

Parameters are not equivalent, more than that they have non linear ties, it is not okay to represent a bond with a set of common parameters. If you have a family of bond with the same Coupon, Issue Date, Volume and different maturity date, the method will return the coupon with the nearest maturity date. But in fact you would probably be more interested to find the bond such that yield to maturity is the same.

You would have to find a better representation first.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Thank you for your answer. So what do you suggest? that I leave a "linear formula" and try to take a maximum of fields and try to make some switch cases on those fields to try to better approach my bond? Or to try to find another kind of formula? What i am doing right now is to pick random numbers for the coefs and plot the result in a graph so that I try to find which field is really important in the choice $\endgroup$
    – Mehdi
    Jul 10, 2014 at 8:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.