ThisDay's Voodoo 'Trends Analysis'
Trends analysis - also known as trends extrapolation - is a systematized pattern of information gathering with a view to spotting a pattern and predicting a future outcome. In project management studies, it denotes a mathematical technique that uses historical results to predict future outcome by tracking variances in cost and schedule performance.
Trends analysis often begets the twin questions of legitimacy and acceptance: when does the 'trend baby' gain acceptance as a bona fide trend? There are of course different ways this could happen such as when it gets enough confirmation in the various media, surveys or questionnaires or from other similar trends studies. But trends analysis could also be used to manipulate reality, mislead, frame a discussion of create a bandwagon effect. In a very influential book, How to Lie with Statistics (1954), the American writer, Darrell Huff, outlines common errors, both intentional and unintentional, associated with the interpretation of statistics, and how these errors could lead to inaccurate conclusions. The funny business of lying cleverly with figures, which Darrell Huff described in the 1950s, is routinely employed by rogue analysts, including in trends studies.
This brings me to the recent 'trends analysis' of how PDP delegates may vote, which ThisDay newspaper has been publishing for sometime. The third of such 'updates' was published on Tuesday, September 7, 2010.
Let me quickly state my two possible biases in this commentary: ThisDay is my favourite newspaper, and I lean some 100 percent towards Jonathan's candidacy in the 2011 elections. Having said this, while I commended the efforts of my favourite paper in venturing into the murky waters of trends studies when I read the first instalment, I became apprehensive when it published a second 'update.' There were simply a few things that didn't seem to add up quite well. When it published the third 'update', I chuckled after a closer study of the figures and quickly remembered some of the tricks in Huff's best selling book on how to lie with statistics which I read years ago as an undergraduate student in the USA.
In the said third 'update', the paper claims that out of a possible 3,483 statutory delegates, its trends extrapolation was that Jonathan would get 1,606 votes, IBB 901 and Atiku, whose waiver to contest for president had not even been granted by PDP when the update was published, would get 63 votes while the undecided were 913. I had assumed that the paper's aim after the publication of the first 'trends analysis' and the second 'update' was to create PR gimmick that would bolster the chances of Jonathan winning the PDP primaries, which was fine with me. Closer analysis of the figures however show that the key beneficiaries are in fact Atiku and Babangida - not Jonathan - and this has set me wondering the paper's motive.
For instance, if we accept the figures from ThisDay , of the 3,483 delegates, the total number of those likely to vote for Jonathan is 1,606 compared to 1,833 who are not for him. If you factor in the fact that a presidential candidate's favourability rating with potential delegates tends to be higher before the candidate's formal declaration (partly because his opponents have not begun subjecting him aggressively to scrutiny) and the suspicion that Babangida will withdraw from the race at a point, then ThisDay's 'trends analysis' is literally handing over the nomination to Atiku, who had not even secured a waiver to run from his party when the third 'update' was published. With the suspicion that the fear of the EFCC is making many Governors to say one thing while meaning another, it may be safe to assume that nearly all the 913 delegates who claimed they were 'undecided' may actually be solidly against the president while the tally of the delegates projected to vote for the president should be discounted by some 50 per cent because of the 'fear of EFCC factor'. This raises a very important question of ThisDay's motives in these 'updates'? I suspect at least two motives:
One, contrary to initial impression, the paper's trends analysis may actually not be aimed at creating a momentum for President Jonathan but may simply be a new business for the paper in which candidates who 'subscribe' will be favourably rewarded in the next 'update' while those who refuse to 'subscribe' will be 'punished' with fewer allocation of delegates. If this assumption is correct, then the owners of the paper have borrowed heavily from Huff's book on how to lie with statistics because the figures are presented in such a way that depending on the premises and assumptions, (every projection is based on some assumptions) each of the three candidates could either come top or bottom of the ranking. This clever ploy of course ensures that they are panicked enough to want to 'subscribe' to the new business. In fact, a senior staff in the campaign team of one of the candidates told me their principal would have been top of the league but for the fact that he bluntly refused to see the paper's 'emissary'.
Two, there are several methodological issues in the 'trends analysis' that it defies logic, and raises a crucial question of how a highly regarded newspaper like ThisDay will publish such - on its front page. For instance in Bayelsa state where the State Governor was for a long time in a cat-and-mouse relationship with the President and where it is still not certain where his loyalty lies, 58 of the available delegates were awarded to Jonathan. What is the paper trying to insinuate here? Is it deliberately and mischievously trying to mislead the President by giving him a false sense of security that will lead to his humiliation at the primaries? Is ThisDay actually working for the President's opponents?
Again, though the paper rightly admits that its 'trends analysis' is not a poll, it claims that estimation of the number of delegates for each candidate was based on 'political intelligence, interviews with key party figures and emerging alliances as well as traditional working relationships.' There are several questions here: how does the paper define 'political intelligence' so we can assess the likely reliability of the information from that source? It claimed to have interviewed 'party figures' - which figures were interviewed and with what sort of questionnaires so we can assess likely biases? What does it exactly mean by 'traditional working relationship' in an environment in which notions of political friendship and alliances are actually very dynamic and fluid? What methods were used in the interview - phone calls, questionnaires, face-to-face interviews? What methods, if any, were used to control responses to ensure the information given were reliable? Above all, did ThisDay really carry out this 'trends analysis' or are the figures just contrived to serve a purpose?
There are of course reasons why Nduka Obaigbena is regarded as a shrewd business man. It will certainly not surprise any one if other newspapers and companies jump into the bandwagon of the apparently lucrative business of 'trends analyses. After all, this is the season of the long knives.
Dr Peter Briggs, an expert in trends analysis, is the CEO of the Abuja-based marketing company, Cleos International. He can be reached at