On Sat, 18 Dec 2004, Robert A. Book wrote: > I know there were lots of problems with the exit polls, but it is > possible for the positive correlation with exit polls and the negative > correlation with state per capita income to both be right?
Yes. The state correlation is calculated over aggregates whereas the exit polls are over individuals. There is no reason the two correlations should be the same or, as in this case, even of the same sign. In some literatures this dicrepancy is referred to as the "ecological fallacy," the fallacy occurring when one infers individual-level correlations from observed aggregate correlations. Cheers, M. Christopher Auld phone: 403.220.4098 Assistant Professor fax: 403.282.5262 Economics, University of Calgary email: [EMAIL PROTECTED] Calgary, Alberta, Canada web: http://jerry.ss.ucalgary.ca > > > > Recognizing the striking political split between rural America and > > Metropolis America, would anyone care to speculate/analyze the underlying > > causes? > > > > Is it that one party offers comparative economic advantages to Metropolis > > America and the other party offers comparative economic advantages to rural > > America? > > > > Is it that one party tends to appeal culturally to those people who live in > > or migrate to Metropolis America, while the other party tends to appeal > > culturally to those people who live or migrate to rural America? > > > > Adding one more item of information to fuel > > speculation/analysis--reportedly, among the most striking cultural divides > > are white birth rates in Red areas vs. Blue areas. > > > > Walt Warnick > > > > -----Original Message----- > > From: William Dickens [mailto:[EMAIL PROTECTED] > > Sent: Thursday, December 16, 2004 6:51 PM > > To: [EMAIL PROTECTED] > > Subject: Re: the answer is... > > > > > > >>> Cyril Morong <[EMAIL PROTECTED]> 12/16/04 04:39PM >>> > > >Could there be some collinearity with education or educational > > attainment? If > > >people with more education make more income (and were more likely to > > vote > > >for Kerry), maybe something else is going on. I actually don't know if > > Kerry > > >got more support from the best educated. > > > > This is an interesting question. Historically Democratic support is > > strongest at the two ends of the education distribution (Less than HS > > and advanced degrees) while Republican support is strongest with the > > college educated. Early on in the campaign there was at least one poll > > that suggested that this had changed radically and that Bush was pulling > > only from the least well educated. But at least one set of exit polls > > showed the historical relationship (see attached provided to me by Eric > > Crampton). > > > > That said, there is no reason why the relationship at the state level > > should be the same as the relationship at the individual level. The > > reason I guessed .75 as the correlation is because I've run a similar > > correlation with AFQT scores a while back and had some idea of how > > strong and close to linear the relationship is between state > > characteristics and Democratic vote share. And yes, there is a lot of > > multicolinearity here. All these characteristics line up fairly well > > (education, income, IQ etc.). All this shows, is something else that we > > know from the post election discussion. Republicans are increasingly the > > party of rural America and the ex-urbs while the Democrats are the party > > of Metropolis. Even in the South most cities voted democratic while in > > the bluest of blue states the rural areas voted Bush. More rural states > > vote Republican more and have lower income, education and test scores. - > > - Bill Dickens > > > > > > > > > > > > > > > > William T. Dickens > > The Brookings Institution > > 1775 Massachusetts Avenue, NW > > Washington, DC 20036 > > Phone: (202) 797-6113 > > FAX: (202) 797-6181 > > E-MAIL: [EMAIL PROTECTED] > > AOL IM: wtdickens > > >
