AROUND OZ 10/2019 SpotlightFotos: alzay, Floortje/iStock.com; privatPETER FLYNN is a
public-relations
consultant and
social commenta-
tor who lives in
Perth, Western
Australia.M
ore than two months after the Australian election,
the opinion pollsters and big bookmakers are still
trying to figure out how they got it all so wrong. Yes,
Australia had its own Trump and Brexit moment
when all the pollsters and bookmakers — right up to
election day — predicted a comfortable Labor win,
and, instead, the Conservative coalition government
was returned with a small but workable majority.
One bookmaker was so confident about Labor’s
chances that it paid out $A 1.5 million two days be-
fore the election (this is not an uncommon practice
with bookies, who rely on gamblers to reinvest their
“winnings” until they ultimately lose), but then had
to pay out another $A 4.5 million on the 7-to-1 odds
offered on the Conservatives.
While that’s probably small change for a corporate
bookmaker, pity the punter who lost a cool $A 1 mil-
lion on Labor (at odds to win just $A 250,000) with
a different bookmaker, who didn’t pay out early. No
doubt that punter was banking on the reliability of
Australia’s major opinion pollsters, which for three
years had collectively and consistently predicted a
Labor win. Not one of them got it right.
The reason is quite simple and the problem has
been getting worse for more than a decade. The
shrinking of the white pages, caused by the decline
in households with fixed phone lines, means that
pollsters can no longer find a random sample of vot-
ers who are representative of the country’s opinion,
a state or an individual electorate.
After decades of working successfully with market
researchers and pollsters, I can assure you that one
thing above all else underpins accuracy: randomness.
Landline phones will get you the right geographic,
but not a representative demographic, because peo-
ple with home phones are over 45 years of age.
Mobile phones — and everybody has at least one
if not two — enable pollsters to access younger andolder voters, but the problem is that the pollsters
have little or no idea where those voters live. Increas-
ing the size of the sample won’t help if the database
is already biased towards a subset of the population.
That’s why a properly representative random sam-
ple of just 500 people can be extrapolated to accurate-
ly reflect the views of hundreds of thousands, even
millions. That’s especially important in a country
where voting is obligatory.
A market researcher confided to me recently that,
as a guest speaker at a parliamentary luncheon back
in 2011, he told the gathered politicians that they
had seen the best of opinion polling. “It can only get
worse from here,” he said.
In our private conversation, he said his company
no longer does polling in important marginal elector-
ates because he has no confidence in getting a true
random sample, however small: “We’d just get egg all
over our face.” Wise man, my friend.AROUND OZ
Against all
expectations
Nachdem immer weniger Leute in Australien
Festnetzanschlüsse haben, ändert sich die
politische Meinungsforschung dort radikal.ADVANCEDbiased [(baIEst]
, voreingenommen
bookie [(bUki] ifml.
, Buchmacher
egg: get ~ all over one’s face
[eg] ifml.
, sich lächerlich machen,
dumm dastehen
electorate [i(lektErEt]
, Wählerschaft, Wahlkreis
extrapolate [Ik(strÄpEleIt]
, ableiten
figure out [(fIgEr )aUt]
, ergründen, verstehen
fixed phone line
[)fIkst (fEUn laIn]
, Festnetzanschluss
luncheon [(lVntSEn]
, (formales) Mittagessenmarginal [(mA:dZIn&l]
, hier: Rand-
opinion pollster
[E(pInjEn )pEUlstE]
, Meinungsforscher(in)
punter [(pVntE] Aus., UK ifml.
, Kunde, Kundin
random [(rÄndEm]
, zufällig
shrink [SrINk]
, schwinden
small change [)smO:l (tSeIndZ]
, Kleingeld
subset [(sVbset]
, Untermenge, Teilmenge
ultimately [(VltImEtli]
, letztlich
underpin [)VndE(pIn]
, untermauern, stützen63