Alex Tabarrok wrote:
> Non-working women are likely to have husbands who earn more than the
> husbands of working women (all else equal) - this says the probability
> of a woman working increases with a *decrease* in *husband* income. But
> the finding is that the probability of a woman working increases with an
> *increase* in the *sister's husband's* income. Two different results.
It's hard to know how to interpret this result without knowing what
was actually estimated. Ray's point is quite correct if the model
looks like
y* = Xb + c*1(SH>H) + noise,
where y* is a latent variable governing employment, X contains whatever
else they controlled for aside from husbands' incomes, and SH and H are
sister's husband's income and own-husband's income respectively. This
model confounds changes in H and SH -- we'd expect c to be very large
simply because 1(SH>H) is highly (negatively) correlated with with H.
If the model looks like,
y* = Xb + dH + eSH + c*1(SH>H) + noise,
then a value of c implying an odds ratio of 16 to 25 is pretty startling.
Does anyone know the actual cite?
Chris Auld (403)220-4098
Economics, University of Calgary <mailto:[EMAIL PROTECTED]>
Calgary, Alberta, Canada <URL:http://jerry.ss.ucalgary.ca/>