However, the drawback of these regressions is that they only test predictions of
how either foreign sales or currency derivatives affect exchange-rate exposure. Since
we observe an ex-post exposure (beta) that is determined by both foreign operations
and financial hedging, the regression that does not include both factors suffers from
an omitted variables problem. This problem could be more severe in this case,
because the two factors, foreign sales and currency derivative use, are positively
correlated. To include both variables in the regression, we must focus either on the
positive or the negative exposures. Because the presence of foreign sales creates a
positive exposure, we chose to perform our multivariate tests by using the sample
of positive exposures. In addition, foreign sales is an exposure factor for which firm level data is available, in contrast to imports and exports, for which only industry level data are available.