A mod­ern-day com­edy of er­rors

Exit polls of­ten go wrong in In­dia be­cause poll­sters do not sam­ple vot­ers in the poor­est parts of the coun­try or the core sup­port bases of dif­fer­ent political par­ties

Business Standard - - FRONT PAGE - The writer is Pro­fes­sor, In­dian Sta­tis­ti­cal In­sti­tute, Kolkata ATANU BISWAS

Exit polls of­ten go wrong in In­dia be­cause poll­sters do not sam­ple vot­ers in the poor­est parts of the coun­try or the core sup­port bases of dif­fer­ent political par­ties

The Elec­tion Com­mis­sion was fu­ri­ous with a Hindi daily af­ter it flashed the re­sult of the exit poll af­ter the first phase of the UP As­sem­bly elec­tion of 2017. Un­der­stand­ably so, be­cause it could have in­flu­enced the re­main­ing phases of the elec­tion. In the United States, in 1980, NBC an­nounced the vic­tory of Ronald Rea­gan at 8.15 pm Eastern Time, when it was 5.15 pm on the West Coast, where vot­ing was not yet over. It is pos­si­ble that many peo­ple did not come out to vote hear­ing the sur­vey re­sult.

How­ever, exit polls have ex­hib­ited a dis­as­trous mis­match with the fi­nal EVM/bal­lot counts on nu­mer­ous oc­ca­sions in the re­cent past, all over the world. Should we then treat exit polls se­ri­ously? Or should a statu­tory warn­ing be given that they are de­signed for pure en­ter­tain­ment pur­poses only, for they are a mod­ern-day com­edy of er­rors?

Amer­i­can poll­ster War­ren Mi­towsky is one of the pi­o­neers of exit polls. He or­gan­ised the first exit poll for CBS News in a lo­cal elec­tion in Ken­tucky in 1967. In exit polls, peo­ple are asked about their vot­ing pat­tern just af­ter the vote is cast, and a spe­cific pat­tern is fol­lowed — for ex­am­ple, ev­ery third or fifth voter. Mean­while, if some ov­er­en­thu­si­as­tic vot­ers want to opine, the sur­veyor needs to turn them away.

Sur­pris­ingly, exit polls have pro­vided hi­lar­i­ous re­sults in many im­por­tant elec­tions world­wide in the last few decades. It hap­pened in 1992 in Bri­tain. In 2000, the exit polls favoured Al Gore in the de­ci­sive state of Flor­ida. The rest is a his­tory of mon­u­men­tal blun­ders in US elec­tions. In the 2004 elec­tion, the dif­fer­ence in the vote shares of John Kerry and Ge­orge W Bush pre­dicted by the exit poll was 6.5 per cent more than the ac­tual! Political an­a­lysts and re­searchers are still strug­gling to ex­plain this elec­tion.

Al­though al­most all the exit polls in the 2014 Lok Sabha elec­tion pre­dicted an NDA vic­tory, most of them were nowhere close to the ac­tual tally of 336 seats. Al­most all the sur­veys in 2009 pre­dicted an NDA win. The rest is his­tory. In 2004, the exit polls pre­dicted an­other Va­j­payee vic­tory, and the ac­tual num­ber of seats the NDA got was 187 — a mas­sive er­ror!

Al­though most of the exit polls pre­dicted a Tri­namool vic­tory in the West Ben­gal As­sem­bly elec­tion of 2016, very few could pre­dict the party’s huge vic­tory. The As­sem­bly elec­tion of Bi­har in 2015 is a clas­sic ex­am­ple of er­ro­neous pre­dic­tion. Exit polls also failed mis­er­ably in pre­dict­ing the mas­sive AAP vic­tory in the Delhi As­sem­bly elec­tion of 2015.

In to­day’s world, in any elec­tion it is com­mon that about half a dozen exit polls would pre­dict the vic­tory of party A, an­other half a dozen would pre­dict party B’s vic­tory, and half a dozen would por­tray a close sce­nario. Exit polls might yield dis­as­trous re­sults for many rea­sons. The sam­ple size should not be very small. Again, quite of­ten the sam­ples are far from “ran­dom”.

Un­der­stand­ably, polling or­gan­i­sa­tions are in a hurry to pub­lish their re­sults as soon as pos­si­ble. Do the sur­vey­ors of all these agencies reach the poor­est parts and the re­mote cor­ners of the coun­try? Do they reach the core sup­port bases of dif­fer­ent par­ties in ru­ral In­dia in ap­pro­pri­ate pro­por­tions? Do they reach the sen­si­tive booths of the coun­try which com­prise nearly half of total booths? In many cases, it is likely that the sam­ples are not rep­re­sen­ta­tive ac­cord­ing to the so­cio-eco­nomic classes across re­li­gions and races.

Ma­jor er­rors in exit polls may oc­cur due to dif­fer­ences in en­thu­si­asm of the sup­port­ers of dif­fer­ent par­ties. John Kerry was ahead of Ge­orge W Bush in the exit polls for the 2004 US pres­i­den­tial elec­tion, but Mr Bush won. Many poll­sters like War­ren Mitof­sky be­lieved that it was mainly due to the over-en­thu­si­as­tic par­tic­i­pa­tion of Mr Kerry’s sup­port­ers in the exit polls.

Tra­di­tion­ally in the US, Repub­li­cans have been re­luc­tant to par­tic­i­pate in such stud­ies. This is called “non­re­sponse bias”. In Bri­tain, it is be­lieved that Con­ser­va­tives have a sim­i­lar “shy Tory fac­tor”, the phrase orig­i­nat­ing dur­ing the 1992 vic­tory of John Ma­jor. The “shy Trump vot­ers” story be­came very pop­u­lar af­ter the 2016 pres­i­den­tial elec­tion in the US, al­though many peo­ple didn't agree with this the­ory as far as Don­ald Trump's elec­tion is con­cerned. Ac­cord­ing to the “shy” the­ory, vot­ers are usu­ally re­luc­tant to ad­mit sup­port for par­tic­u­larly con­tro­ver­sial or po­lit­i­cally in­cor­rect can­di­dates.

Of course, “non-re­sponse bias” might also oc­cur for dif­fer­ent pat­terns of par­tic­i­pa­tion with re­spect to age, gen­der, race and caste. For ex­am­ple, in the US pres­i­den­tial elec­tion of 2004, 43 per cent among sixty-plus vot­ers and 55 per cent among younger vot­ers took part in the sur­vey.

Even one and a half decades back, a big dif­fer­ence be­tween the ac­tual re­sult and the exit poll re­sult was some­times looked upon by some peo­ple as ev­i­dence of al­le­ga­tions of elec­tion fraud, such was the cred­i­bil­ity of the exit polls — be it in Venezuela to take back Hugo Chávez, in the pres­i­den­tial elec­tion of Ukraine, or the US pres­i­den­tial bat­tle be­tween John Kerry and Ge­orge W Bush, all in 2004. Richard Chari­nen’s book, Prov­ing Elec­tion Fraud, might make for in­ter­est­ing read­ing in this con­text.

In the 2003 elec­tion in Ge­or­gia, al­le­ga­tions of fraud were strength­ened by the exit poll re­sult, which trig­gered the “Rose Rev­o­lu­tion” and the res­ig­na­tion of Ed­uard She­vard­nadze. To­day’s poll­sters may like to try and erad­i­cate the ridicu­lous present-day per­cep­tion of exit polls. We'd cer­tainly like to watch poll pre­dic­tions that are a lit­tle more se­ri­ous, in­stead of watch­ing com­edy shows!

Al­most all exit polls in the 2014 Lok Sabha elec­tion pre­dicted an NDA vic­tory, but most were nowhere close to the ac­tual tally of 336 seats


Vot­ers at a polling sta­tion in Jaipur show their iden­tity cards while they wait to cast their votes in Fri­day’s state As­sem­bly elec­tion in Ra­jasthan

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.