All of us are blind but some of us are more blind than the rest of us. At the height of the HIV epidemic, researchers wanted to know what measures men with a specific sexual orientation were taking to avoid contracting the virus. They asked respondents whether they always carried condoms around as a preventive measure. Many said they did. Things took a different turn when the researchers then asked the men to show the condoms they were carrying at that point. It turned out there were fewer people with condoms compared to the men who said they always had one with them. This phenomenon is called the social desirability bias — SDB. The sort of answers people gives to questions to feel good about themselves and/or to fit into a group or to be perceived in a certain way.
The ANAP/NOI polls have generated substantial conversation. The APC rejected it saying most of its voters were grassroots people who were not captured by the poll. The PDP called it a hatchet job. NNPP’s Rabiu Kwankwaso knocked the first poll. He has now gone from being a ‘dark horse’ to being part of a ‘four-horse race’. The Labour Party accepted the result. You would not need a survey to predict these reactions. Politicians, like most people, generally agree with results that favour them but find a way to knock the results that do not. But this is not about politicians and the predictable reactions, it’s about the scientific credibility of these polls.
A poll is a work of research, objectivity is a crucial element. Let’s put the ANAP/NOI polls to the same test one would put any such poll. How much care was been taken to ensure these polls weren’t biased? The better such care, the more trustworthy the poll.
Both polls suffered from Nonresponse Bias. Nonresponse is a situation where there is a substantial number of people who do not respond to a survey when compared to those who responded. In the September ANAP/NOI poll, 32% of the respondents were said to be undecided whilst 15% preferred not to reveal their choice. The poll could not show how a whopping 47% of the respondents would vote. In the December 2022 poll, undecided voters had reduced to 29% of respondents whilst the non-response had jumped to 23%, both accounting for 52% of respondents. Even if their sampling process was perfect, can 480 people predict the behaviour of 93.5 million potential voters?
The ANAP statement signed by its founder, Mr Atedo Peterside, suggested that respondents refusing to disclose who they were voting for was historically suggestive of fear of voter intimidation. He noted that the South-West’s 38% compared to the national range of 14–23% could be indicative of voter intimidation. In essence, the pollster was claiming that its respondents chose not to reveal their choice of candidate because of fear of intimidation. If your respondents do not feel safe enough to choose from your options, your poll is already biased.
INEC and the Nigerian government are responsible for the conditions around the election, but polling organisations are traditionally responsible for the conditions of their research process. If the organisation reflected on this instead of externalising the cause, it would see that nonresponse and the high number of supposedly undecided voters had something to do with its methodology. It can fix these for subsequent polls, but first it must accept responsibility.
Other sources of bias include;
- how were the questions and options worded? If you repeatedly placed a candidate at the top of your question whilst placing another repeatedly at the bottom, that is a biased process. Randomising the options would reveal a more credible result, ceteris paribus.
- what issues were top of the news when the polls were conducted? If a candidate was trending for the wrong reasons during the poll, the process is already biased against such candidate. The same applies, mutatis mutandis, if a candidate was enjoying a great press run.
- were the respondents primed before being asked their options? If I started out asking, for example, if a respondent had seen the 2nd Niger Bridge, before asking them to choose from a pool of candidates, I would be biasing the respondent towards the incumbent party. The reverse would apply if I asked them what they thought of banditry before asking them who they would vote for. From the released statement and the subject matters mentioned by Mr. Peterside, priming is suspected.
- How random was the polling sample? What measures were put in place to avoid a selection bias?
Mr. Peterside, has been seen to back LP candidate and 2-time winner of ANAP/NOI Presidential poll, Mr. Peter Obi, to run for president. It is not impossible for him to hold such a partisan position yet lead a non-partisan effort, especially if the platforms conducting such effort are seen to have a strong corporate governance structure to be able to reflect views other than that of their founder.
The NOI Poll website shows only one man on its board. ANAP on its part has 9 Advisory Board members, an impressive 5/4 male/female representation. What is not impressive about this board is self-evident; 5 Petersides and 4 others.
What is not self-evident is revealed by the fact that of the 4 Advisory Board members who aren’t ‘Peterside’, at least 2 return obvious google results that conclusively reflect the fact that ANAP is too in-house to be trusted with such matters as a presidential poll. You need diversity to conduct the sort of poll intended for the Nigerian public. I hope this will be addressed in the future constitution of these boards.
All of us are blinded by biases, but those who are aware of their biases are likely to see a lot better. It is my hope that ANAP/NOI does a better job of its subsequent polls. The path to a better Nigeria requires rigour, seriousness and scientific integrity at every level.