1.Theoutputfromtheregressiongivesusameasureofhow
strong the relationship is between the multiple and the
variable being used. Thus, if we are contending that
higher-growth companies have higher P/E ratios, the
regression shouldyield cluesto both howgrowthand P/E
ratiosare related(through thecoefficient on growth asan
independent variable) and how strong the relationship is
(through the t-statistics and R-squared).
2.Iftherelationshipbetweenamultipleandthefundamental
weareusingtoexplainitisnonlinear,theregressioncanbe
modified to allow for the relationship.
- Unlikethe modified multiple approach,where we were
able to control for differences on only one variable, a
regression can be extended to allow for more than one
variable and even for cross effects across these variables.
Ingeneral,regressionsseemparticularlysuitedtoourtaskin
relativevaluation,whichistomakesenseofvoluminousand
sometimes contradictorydata. Therearetwokey questions
that we face when running sector regressions:
1.Thefirstrelatestohowwedefinethesector.Ifwedefine
sectorstoonarrowly,weruntheriskofhavingsmallsample
sizes, which undercut the usefulness of the regression.
Definingsectorsbroadlyentailsfewerrisks.Whiletheremay
be largedifferencesacross firmswhenwe dothis, wecan
control for those differences in the regression.
2.Thesecondinvolvestheindependentvariablesthatweuse
in theregression. Whereasthefocusin statistics classes is
increasingtheexplanatorypoweroftheregression(through