Damian Loscherexplains the reasons for the adjusted figures in TNS mrbi opinion polls.
With a general election looming, opinion poll results are beginning to take on greater significance. As a natural consequence, opinion polling procedures are likely to come under the media microscope.
A healthy debate on the pros and cons of one technique over another is welcome. After all, opinion polling is constantly evolving. It is vital that ideas and insights are shared, for the benefit of all.
The techniques employed by polling companies need to reflect the changes in voter behaviour. Changes in methodology are inevitable and for this reason TNS mrbi has consistently invested in explorative polling studies to ensure any changes in technique are based on rigorous analysis.
Much of the science that underpins opinion polling has not changed. Each Irish Times/TNS mrbi opinion poll continues to canvass the opinions of a sample of 1,000 adults aged 18 years and upwards, interviewed at 100 points throughout the country. Quota controls are still used to ensure the profile of the poll sample matches that of the population.
The result is a representative sample of the Irish voting public. The techniques for identifying 100 sampling points and respondents, together with the tools for evaluating the statistical significance of the responses, have all remained unchanged since the first Irish Times/TNS mrbi opinion poll was conducted in 1982.
What has changed is how opinion polling companies view the reliability of voting intentions as expressed on the doorstep.
On this point, the 1997 general election was a critical juncture for opinion polling in Ireland; a sizeable proportion of the voting public did not do what they said they would do. What had been observed on the margins in previous years had become a more mainstream phenomenon.
When respondents are asked how they intend to vote, the answers they give are truthful. What they give is their best estimate of their voting intentions at the time of interview.
What has happened, over time, is that the likelihood of their intended behaviour translating into actual behaviour has reduced, arguably as a result of the growing disengagement between the public and politics.
Turnout in the general elections has been falling noticeably since the 1980s.
This has repercussions from a polling perspective as potential voters who choose to stay at home are not evenly distributed across the population.
Older citizens are more likely to vote, and they are more likely to vote for particular parties.
Unfortunately, the solution is not as simple as just asking respondents if they intend to vote. Many of us intend to vote but find we have other priorities on election day.
Recent TNS mrbi research has shown that awareness of the candidates who will be running in a respondent's constituency and which parties they actually represent may be the best predictor of voting intentions. But more research is needed.
The other fly in the polling ointment is the discrepancy between claimed and actual voting behaviour. This discrepancy is most apparent in the tendency for opinion polling companies to overstate the Fianna Fáil vote. Why? Quite simply because a growing segment of the electorate intend to vote Fianna Fáil at the time of interview, but change their minds on election day. But this change of mind is not as puzzling as it sounds.
Earlier this year TNS mrbi conducted a parallel study using both party support (where a respondent indicates which party they will give their first preference vote to) and simulated ballot paper (where a respondent chooses their preferred candidate from a ballot paper) techniques, to provide some insight into why a respondent claims support for Fianna Fáil but chooses a candidate from another party on election day.
The findings from this study were fascinating. Prior to election day, many respondents are unfamiliar with the candidates or are uncertain about the parties they are standing for. This is particularly true for smaller party candidates, making it difficult to translate affinity with these candidates into party support. I believe it would be fair to make the assumption that this disconnection has grown over the years and decades, not least because more and more people are setting up home in an area and a community that are unfamiliar to them.
There is no definitive answer to the question of how to adjust for these emerging contradictions between what respondents say and what they do. The TNS mrbi way is to apply as much science as possible to finding the answer.
In 1997 we conducted our first survey on election day, which asked the voting intentions of people who had yet to vote and the voting choice made by those who had voted. The survey allowed us to isolate and subsequently measure the difference between claimed and actual behaviour. The outcomes from this research were calibration factors which were subsequently incorporated into all Irish Times/TNS mrbi political polls in the form of an "adjustment".
The adjustment calculation process is relatively simple. Party support data from interviews conducted just prior to voting on the day of the election (pre) are compared with data from interviews conducted with respondents after they had voted (post). All interviews are conducted on election day so the timing of the research does not become an issue, and both pre and post samples are demographically matched to hold age, gender and social class variables constant.
The adjustment figure is the weighting which we would have had to apply to the level of support recorded for each party at the pre voting stage to accurately predict what each party achieved on election day.
Bearing in mind the tendency to overstate support for Fianna Fáil, it comes as no surprise that the weighting factor for Fianna Fáil produced by this pre/post research technique is less than one. Thus, the adjusted level of support for Fianna Fáil shown in today's poll is less than the level of support recorded at the time of interview. This adjusted level provides a more accurate measure of support for Fianna Fáil. Similarly, adjustments are made for other parties, except the weighting factors are greater than one, resulting in higher levels of support for these parties than indicated at the time of interview. TNS mrbi believes election day surveys are the most scientific way of calculating the necessary adjustments to voting intentions, and we will continue to conduct these surveys.