rounding

[wrongsaidfred]

Hi David,

I know this may an unanswerable question, but do you know how much we are supposed to round while performing computations on the exam? 99% of the time it does not seem to matter, but for some types of problems, like calculating the effective convexity, it seems makes a huge difference because the numbers we are dealing with are so small and so close to one another.

Thanks in advance for any help you can provide.

Mike

 

[David Harper CFA FRM]

Hi Mike,

My first preference is to avoid rounding altogether: I format the TI BII+ to 4 decimals and I use STO and RCL (I like STO and RCL specifically so I don't need to re-enter interim values--a technique the reduces my error--but it has the side benefit of not imposing rounding).

GARP does not tend to exploit fine precision difference (thankfully). I agree with you that small numbers (e.g., variance) can require more decimals, but I think you only need to go out more decimals if it's a small number; e.g., there have been questions, occasionally, with numbers like 0.00050, so in that case you have no choice. Now, if it's like convexity = 118.09, GARP will never insist on fine precision and will not put two answers near together; GARP will not test you on your decimal precision per se.

So, if rounding must be an issue (I am not clear why you must round, but only why you might format to 3 or 4 decimals), then ASIDE from small numbers (0.0000x), i think 3 decimals covers all other situations.

In all the post-exam feedback we've gotten over several years, I don't think rounding has EVER come up even once; i.e., nobody sincerely suggested that two different, correct answers were too close to distinguish, hope that helps, David

 

[wrongsaidfred]

Ceratinly does. The reason I asked was because of some problem in Schwesser. When calculating convexity, the prices of the bonds after the up and down moves were so close that it really threw things off when the answer key rounded to 3 places instead of carrying everything out. I was unsure if that was a regular convention because I also use the STO and RCL features when doing most problems.

Thanks again.

Mike