1. What is an opinion poll?
An opinion poll is a survey carried out to measure views on a certain topic within a specific group of people. For example, the topic may relate to who Kenyans support in the presidential race, in which case, the group of people interviewed will be registered voters.
2. How are interviewees for an opinion poll selected?
The group of people interviewed for an opinion poll is called a sample. As the name suggests, a sample is a group of people that represents the total population whose opinion is being surveyed. In a scientific opinion poll, everyone has an equal chance of being interviewed.
3. So how come I have never been interviewed for an opinion poll?
You have the same chance of being polled as anyone else living in Kenya. However, chances of this are very small and are estimated at about 1 in 14,000. This is because there are approximately 14 million registered voters in Kenya and, for practical and cost reasons, usually only between 1,000
and 2,000 people are interviewed for each survey carried out.
4. How can such a small group be representative of the entire population?
In order to ensure that the sample/survey group is representative of the population, the surveyors must ensure that the group reflects the characteristics of the whole. For instance, to get a general idea of who might win the Kenyan presidential election, only the views of registered voters in Kenya will be surveyed as these are the people who will be able to influence the election.
Furthermore, the survey will examine the characteristics of the registered voter population so that it is representative of ethnic groups, age groups, income brackets, rural and urban residents and other characteristics. If, for example, 30% of registered voters are male, the selected sample should reflect this proportion.
5. How are representative samples selected?
Polling companies normally select a representative sample using two methods:
- Random sampling involves unsystematically drawing the contact information of a certain number of
people from a list of the relevant population, for example, a list of registered voters. - Quota sampling where the company sets “quota controls,” for example, age and gender. If for instance, 52% of the population is female and 38% is aged between 25 and 44 years, the sample selected should ensure that the selection of people interviewed reflects these statistics.
6. Are opinion polls accurate?
Well-designed polls are usually accurate within 3 percentage points. This level of accuracy of plus or minus about 3 percent is referred to as the margin of error/sampling error. A well-designed poll is one which is representative of the population and contains balanced questions which are not worded in a way to lead people to certain answers.
National opinion polls are a fairly recent occurrence in Kenya. While there were electoral opinion polls conducted as early as 1992, many of them lacked methodologically sound practices.
One independent agency, the National Election Monitoring Unit, did accurately predict the Moi victory in 1992, but its lack of a representative sample caused it to inaccurately report that Moi and Kibaki were neck and neck with regard to numerical support. Several polls were conducted in the
run-up to the 1997 election, but many of them continued to suffer from methodological flaws.
It was only in the lead- up to the 2002 election that more reliable agencies began conducting polls, and some have accurately predicted results. For example, the International Republican Institute (IRI)’s exit polls accurately predicted the results of the 2002 election (within 2 percentage points) and
the 2005 referendum (within 1.5 percentage points). The Steadman Group was also able to accurately predict the results of the 2005 referendum.
7. What is a margin of error/sampling error?
Since the sample does not include all the members of the population, statistics on the sample will generally differ from statistics on the entire population. A margin of error, also known as a sampling error, thus occurs due to the differences between statistical characteristics of a population and a sample drawn from that population.
8. Do sampling errors mean opinion polls are not accurate?
No. Polls are generally a good guide to the state of opinion, even in cases where the margin of error is larger than 3 percent. For instance, if a poll determines that a population is divided 70% to 30% over an issue, a margin of error as big as 10% will not change the fact that one opinion is much more dominant than the other.
On the other hand, if opinions are much more closely divided, for example 55% to 45%, then polls may not be certain indicators of the views of the population on that matter.
9. How do I know if an opinion poll is reliable?
The following guidelines can help determine how reliable a poll is:
• Who conducted the poll? Was it an independent and reliable firm? Is the firm a member of the Marketing and Social Research Association (MSRA)? This Association upholds local and international best practices with regard to polling.
• Who paid for the poll and why was it done? A poll carried out for a respected media outlet, or for independent researchers, is probably reliable. A poll for a political party, pressure group or a company, may also be reliable if the firm used a scientific sampling/ question design and
impartial questions, and if it provided all the information about the questions asked and results obtained.
• Number of people sampled. Generally, the larger the sample, the more reliable the results. However, a scientific survey with a small sample is always more reliable than a non-scientific survey with a large sample. Once a sample gets to a size of between 500-700 interviewees, there are
fewer and fewer accuracy gains which come from increasing the sample size.
A scientifically–conducted poll of 1,000 people is not going to be much improved by adding more people to the sample.
• How the sample was chosen. Can you tell who was included and who was left out? If the poll claims to represent the public at large, is there evidence that the firm used random sampling to select interviewees?
• When the poll was conducted. Events can affect people’s opinions. For instance in political opinion polls, opinions can change very fast in response to political developments. The date of the poll should therefore always be disclosed.
• How were the interviews conducted? There are four main methods: face-to-face, by telephone, by email and by mail. Each method has its own strengths and weaknesses, and reputable companies use statistical techniques to address relevant weaknesses.
• What were the questions? Getting a copy of the full questionnaire (and not just the questions which were published), can help determine whether or not they are balanced.
10.How can I make the most of opinion polls?
• Never pay too much attention to just one poll. Look at a collection of different polls over time in order to get a more balanced picture. It is in the nature of polls to fluctuate.
• Make sure you understand the context behind the poll. The best polls ask the same set of questions over time. This allows them to interpret people’s opinions over the long term and to understand historical incidents that may have shaped those opinions.
• Remember that public opinion changes, sometimes drastically. There is no way a poll can accurately capture the impact of last-minute events that can drastically change people’s minds, for instance on a political matter. Opinion polls are therefore not necessarily accurate predictors of
an election outcome.
• Remember that the accuracy of polls depends on how truthful people are in their responses. People may lie to pollsters for a number of reasons, such as reluctance to admit to behaviour that is disapproved of. Furthermore, in the case of opinion polls on elections, there is no guarantee that the people who are polled are the same people
who will vote on Election Day.