# Regression in hedge accounting

When using regression analysis for GAAP hedge accounting treatment, is it required to regress the percentage change in price of each variable over a given time period or can you regress the price levels of each variable of a given time period? In example below is it appropriate to regress 100, 110, 106 against 102, 108, 101 or regress 10% and -3.6% against 5.9% and -6.5%?

Example:

Price of A Price of B Percentage change A Percentage change B

100 102

110 108 (110/100)-1 or 10% (108/102)-1 or 5.9%

106 101 (106/110)-1 or -3.6% (101/108)-1 or -6.5%

I’d guess it is a regression of returns, but who knows what madness the OCC, SEC, Federal Reserve, CFPB, CFTC, or whoever came up with.

Actually not quite that straightforward. Hedging fuel exposure. Regressing the price of the fuel exposure against the price of the hedge instrument. I’m good with regressing the prices…but getting pushback that it should be regressing the percentage change in prices. Unfortunately they provide two different answers when considering if the hedge instrument is “highly effective”.

Price regression will give you spurious positive results, since the observations are auto correlated. For example, if both stocks start at \$100, and they just so happen to be \$200 halfway through the observation period, all observations after that point are more likely to be very high also. So, you will get a very positive slope and falsely small error term. This issue doesn’t exist with return regressions, since you can assume that the returns in each period are not serially correlated. There are some technical terms for this, but I forgot everything after not doing any math for years.

wouldn’t you take A and see if it’s 80-125 of B if that’s they hypo? …curious to know too…

The question is around whether you are taking the prices and regressing or whether you are taking the percentage change in prices and regressing. Whichever you do, yes you’d take the slope of the regression line and it needs to be within the 80-125 range. I just wasn’t sure if I could regress the prices and be good…sounds like I may need to accept it is more (?) valid to regress the percentage change in prices.