I am having a bit of a problem with currency conversion issues. What I do: I sort European stocks based on their book-to-market ratio, each year I form a portfolio (equal-weight) with the 10 stocks that have the highest book-to-market ratio. I calculate the return on that porfolio in Euros. So basically I am constructing a factor-mimicking portfolio.
Then I download the Fama-French Factors from Kenneth French's website. I am planning to run a time-series regression of my portfolio returns on the factors to see whether my portfolio indeed has a exposure to the value factor (I hope so) and/or whether it has exposure to other factors as well. However, the FF factors are generally calculated in USD as far as I understand, so I converted the returns of the factors into € using:
return in € = (1+return in $)*(1+currency_return) - 1
where currency return is E/USD in t divided by E/USD in t-1, minus 1.
If I then run the time-series regression, I get no (significant) exposure to the value factor but significant exposures to the market (positiv), size (negativ, but that's because I filtered out a lot of small stocks I guess), momentum (negativ). So that's weird?!
If I run the time-series regressions with the original factor returns calculated in USD, everything works -- significant positive exposure to HML. No exposure to WML or SMB. But I am pretty sure it's technically incorrect because of the currency issue...