Does Hil­lary have an edge over Trump?

Pakistan Observer - - OPINION - Fa­reed Zakaria

HIL­LARY Clin­ton has had a good run in re­cent days. She de­liv­ered her best speech yet, se­cured the Demo­cratic nom­i­na­tion, beat Bernie San­ders in Cal­i­for­nia and re­ceived Pres­i­dent Obama’s en­dorse­ment. But will this last? Can she win in Novem­ber? There are schol­ars who be­lieve that the out­come of a pres­i­den­tial elec­tion is de­ter­mined by a set of cold facts — chiefly, the state of the econ­omy and the num­ber of terms a party has been in power. And by that mea­sure, things ac­tu­ally look bad for Clin­ton.The most fa­mous elec­tion model, de­vised by Yale economist Ray Fair, is said to hold up well in analy­ses of elec­tions go­ing back a cen­tury. For 2016, it has a generic Repub­li­can can­di­date beat­ing the Demo­crat hand­ily. The logic be­hind this is sim­ple: The econ­omy re­mains slug­gish; specif­i­cally, in­come growth has been low. But Clin­ton’s big­gest chal­lenge is per­haps that it’s ex­tremely dif­fi­cult for a party to win a third con­sec­u­tive term in the White House. It last hap­pened 28 years ago, with Ge­orge H.W. Bush fol­low­ing Ron­ald Rea­gan. And be­fore that, it was four decades ear­lier, with Franklin Roo­sevelt in 1940.

Of course, there are other mod­els that look at sim­i­lar ob­jec­tive mea­sures but make dif­fer­ent as­sump­tions, es­pe­cially about how to as­sess the health of the econ­omy. De­pend­ing on what fac­tors they use — un­em­ploy­ment, in­fla­tion, home prices - the mod­els reach dif­fer­ent con­clu­sions, with some pre­dict­ing a Demo­crat win­ning. Some also look at the in­cum­bent pres­i­dent’s ap­proval rat­ing. (And while Pres­i­dent Obama’s has im­proved, it’s still not very high.) Van­der­bilt’s Larry Bar­tels, one of the fore­most schol­ars in the field, told me, “The mod­els that look at [eco­nomic] fun­da­men­tals are mixed but prob­a­bly av­er­age out to giv­ing the Repub­li­can can­di­date the edge.”

An­other way to pre­dict the out­come of a pres­i­den­tial elec­tion is to look at polls. There are mod­els that em­ploy sta­tis­ti­cal tech­niques to av­er­age and smooth out polling data. The logic of re­ly­ing on th­ese is also sim­ple. Elec­tions are de­ter­mined by many, many fac­tors. It’s im­pos­si­ble to know in ad­vance all of them or mea­sure them ac­cu­rately. Polls in­cor­po­rate how peo­ple feel about the econ­omy, the in­cum­bent, the state of the world, etc. The polls have Clin­ton win­ning, as do bet­ting mar­kets, which are re­ally just a re­flec­tion of the polls. So which is it? En­ter the blended mod­els, which use a mix of eco­nomic data and polls. Most of th­ese pre­dict Clin­ton win­ning, though some fa­vor the Repub­li­can. Why? The fun­da­men­tal fac­tors might fa­vor the Repub­li­cans, but only very slightly, and polling data re­flect the huge vari­able of 2016 - Don­ald Trump.

When think­ing about mod­els that use past pat­terns to pre­dict the fu­ture, it is worth keep­ing in mind that most of them were wrong about the Repub­li­can pri­maries. Even the es­timable Nate Sil­ver got it wrong, be­cause he ar­gued that polling data from early in the pri- mary sea­son were largely ir­rel­e­vant, and that past pat­terns showed that en­dorse­ments and fundrais­ing were cru­cial. (He has writ­ten some ex­tremely thought­ful posts on Five Thirty Eight ex­plain­ing why he was wrong and what to learn from the mis­take, and they are well worth read­ing.)

Hav­ing taken sta­tis­tics cour­ses when I was in grad­u­ate school, my own sense is one can make two er­rors in this area. One is to for­get that most po­lit­i­cal pat­terns - say, that early polls don’t mat­ter in pri­maries - are built on tiny amounts of data. The open pri­mary sys­tem has re­ally ex­isted only since 1972. The rules within it have changed along the way. When com­puter sci­en­tists talk about “big data” and the abil­ity to de­tect pat­terns, they are usu­ally talk­ing about mil­lions of data points. In pol­i­tics, we are look­ing some­times at just 30 or 40 elec­tions and gen­er­al­iz­ing from them. Any “pat­tern” that emerges from th­ese hand­ful of ex­am­ples is ex­tremely ten­ta­tive. — Cour­tesy: Khaleej Times.

Newspapers in English

Newspapers from Pakistan

© PressReader. All rights reserved.