Pricing of barrier options in practice 2003

earlyexercise · October 6th, 2003, 7:40 am

Can anybody tell me what's state of the art for pricing barrier options (BS, MC, FD) ? What is most common on trading floors (b/c sometimes not the most sophisticated model, but the one you can handle the best is used) ? Thanks a lot !

Charlie · October 6th, 2003, 9:31 am

Can I add to the question, please.What assumptions do these state of the are models make about the distribution? log-normal, implied, CEV, SRCEV, VG, GARCH?And what assumptions to they make about the vol and r/q? constant, term structure, stochastic?And what proportion are discretely observed? with what sort of frequency v. expiry? daily on a 1 year? higher/ lower freq?

Anton · October 6th, 2003, 10:14 am

I'm not an expert, but some rough ideas...The assumption about the distribution is the implied distribution given from the prices of European vanillas. BS gives wrong prices. All other models take too much time to calibrate and price touch and barrier options. I heard at the 2003 Mathfinance Workshop an excelent talk from J. Hakala on a time-discrete model used for pricing touch and barrier options, maybe the slides are to be found on the web. Also, one can have a look here Market prices of OT best, A.

granchio · October 6th, 2003, 12:18 pm

Quite a lot of variability here. Some houses are still using BlackScholes, even if aware that it is wrong (they fudge some correction in and normally loose the trade when in competition). Other use local vol (as Anton suggested). Others stochastic vols. Avellaneda is used as well.In at least one case they can do all of the above.Regarding frequency of barrier event sampling: in equities, most often it is continuous, though very common is the daily case (on the close) especially in some countries where regulators require it if the end products ends up with retail.

rhmari · November 18th, 2005, 7:39 am

hi everybody,i'am also looking for the state of art for pricing barrier optionsi've already priced them using a closed form but my model is not that strong i'am not looking for a very sophisticated model...