In a press conference delivered a few days ago, Edi Rama published the results of what he considered to be the results of the opinion polls that the Socialist Party is currently holding. He called on the media to carry out their own opinion polls in order to compare them. According to Rama, 70% of Albanians believe that the DP and the Tent are in a race not to allow the implementation of the judicial reform, while 68% of voters do not want the postponement of the election date. However, Rama also showed a certain level of skepticism about the polls in general when he said before the French presidential election that “let us hope this time the polls do not get it wrong”.
Two months ago, when I presented an opinion poll in a TV show, members of the panel were also skeptic. They mentioned the “failures” of the opinion polls prior to the elections in Holland and USA and prior to the referendum on Brexit.
In fact, in the latter case, the polls really failed. According to them, 52% of the people would vote against Brexit, while 52% of them voted in favor of Britain’s leaving the EU. But sometimes, the most serious polls fail too. Meanwhile, for the other two cases, the calculations were not that simple. Below I have carried out an analysis of the polls in the USA and Holland and of the polls in Albania.
Polls in the USA
At the moment when Donald Trump was announced a winner by American media (when he secured 270 delegates in the Electoral College), only 93% of the votes on a national level were counted. And up to that point of ballot counting, he was ahead on a national level and also in terms of the votes in the Electoral College. This led to Los Angeles Times (a left wing newspaper) to declare that its pre election opinion poll had been accurate, despite the fact that analysts had mocked it for several months in a row.
But why was the Californian newspaper mocked?
The US portal, Real Clear Politics reported national opinion polls on the race between Clinton and Trump before the US election and constantly calculated their average. The average of the recent polls prior to the elections showed that Clinton was 3.2 points ahead on a national level, while Los Angeles Times showed Trump being ahead by 3 points on a national level (other media, which included right wing and left wing media such as Bloomberg, IBD, Economist, ABC, Washington Post, Fox News, Monmouth, NBC News, Wall Street Journal and CBS News, offered an advantage to Clinton on a national level, ranging from 1 to 6%).
As the ballot counting continued, on a national level, Clinton received almost 3 million votes more than Trump (2.1% ahead). Meanwhile, the latter suddenly received a few votes more in Pennsylvania and Wisconsin, two states which sealed the winner of the Electoral College.
In other words, on a national level, the majority of polls conducted by the main media were very close to the final result (if we bear in mind a margin of error of about +/-3%).
Meanwhile, local opinion polls missed the final results in only a few states.
So, a day before the elections, the average of local polls in Wisconsin showed Clinton being ahead by 6.5%, but Trump won there by 1%. In Pennsylvania, the average of local polls showed Clinton being ahead by 2%, but Trump won there by 1.2%. In Michigan, the average of local polls showed Clinton being ahead by 3.5%, but Trump won there by 0.3%. In Nevada, the average of local polls showed Trump being ahead by 1%, but Clinton won there by 2.5%.
Meanwhile, in other states, the results of local polls were quite close to the final results. Thus, in Florida, the average of local polls showed Trump being ahead by 0.3% and in fact, he won there by 1.3%. In North Carolina, the average of local polls showed Trump being ahead of 1 point and in fact, Trump won there by 3.7%. In New Hampshire, the average of local polls showed Clinton being ahead by 0.6% and in fact, it won there by 0.3%.
In the end, it can be said that the majority of national opinion polls and opinion polls were quite close to the final results in the USA. The only ones that were mistaken were the ones that didn’t predict Trump’s victory in those few states that sealed the winner in the Electoral College.
Polls in Holland
A day prior to the elections in Holland (on 14 March), four national opinion polls were published: TNS NIPO, Peil, I&O Research and Ipsos. They projected that Prime Minister Rutte’s party was going to win 27-29 seats in parliament, while his party actually won 33 seats, enabling this party to form a government.
The result was off the error margin for polls, but the opinion poll had started 4-5 days before their publication. The 4-5 days prior to the elections in Holland were filled with developments for Prime Minister Rutte, who had a strong clash with Erdogan, as he didn’t allow two Turkish ministers to hold a campaign for a Turkish referendum. If the opinion poll were held a day before the elections (and not 4-5 days prior to the elections), the polls could have been more accurate. As the exit polls showed, those voters who hadn’t made up their mind yet and voters of several other small parties had made up their mind to vote for Prime Minister Rutte’s party in the last days and their decision was mainly affected by his strong stance against Erdogan.
Opinion polls in Albania
The latest opinion polls which have been published prior to the elections in Albania were the Pioepoli and IPS Marketing polls. This happened a week prior to the 2015 local government elections. These opinion polls predicted a victory by Erion Veliaj with 53-54% and Veliaj won by 53.6% of the votes.
An opinion poll that I published in a national TV network three months prior to 2015 local government elections (when the candidates had not yet been announced) showed that the candidates for mayors run by the majority would obtain 800 thousand votes and those of the opposition would obtain 650 thousand votes if the elections were to be held in March. Three months later, after the candidates were announced and the campaign had taken place, the result was: 830 thousand to 670 thousand.
A week prior to the 2013 elections, Pioepoli and IPS Marketing predicted a victory of the left wing coalition by 52%-54%. A month before the elections, I published an opinion poll which predicted a victory of the opposition by 53.5% of the votes. In the elections, the coalition obtained 57.6% of the votes on a national level.
Two months prior to the 2011 elections, I predicted in a national TV network that Rama would win in Tirana by 52.5% of the votes. In fact, Rama received 50% of the votes, but lost the election, after 80 ballot papers were found by CEC and these ballot papers had been inserted in the “wrong boxes”. Gani Bobi company marked a spectacular failure that year (2011), as it predicted a victory by Basha with 58% of the votes, while he barely managed to get as many votes as Rama did.
Note: The latest opinion polls before the presidential election in France were very accurate in predicting the winner in both the first round and the second one which was held yesterday.
The republication of this article is strictly forbidden without a written permission from the Albanian Free Press newsroom
Note: The views expressed in this article are the author's own and do not necessarily reflect Albanian Free Press’ editorial policy.
