Who is likely to win? Depends on How you ask
The race is heating up. And so are the experts that are conducting, interpreting and concluding opinion polls. As the focus of US presidential election is moving away from on primaries to the final competition, we should expect results from hundreds of opinion polls predicting the winner.
A recent NY times article suggested that Clinton leads Trump by around 10 percentage points. If you believe these polls, then you may be hopeful or concerned depending on which side you are on. But there is a small issue. The result may not be accurate.
This article indicates that there is a significant difference in the results of polls conducted online vs over the phone. The lead comes down four percent when surveys are conducted online. One the reasons cited is the social desirability bias – a desire to project a positive image when one is worried of being judged based on their response. People may go at any lengths to avoid the discomfort and embarrassment of stating an unfavorable response even if that is their honest response. But in a situation of anonymity, I may go back to my preference.
The problem is not new and its not limited to presidential polls. The issue has been discussed extensively in market research. And it becomes much more pronounced in sensitive areas of financial and health care related decision making. Imagine talking to a individual undergoing financial hardships and being delinquent on their debts. Or a conversation around understanding why someone is not adhering to their antibiotics regimen.
Clearly, we need more sophisticated research methodologies to deal with such sensitive matters. In our work in the social sector, we have regularly innovated our research processes to mange these issues. For example, in one of our projects in Africa we used a gamification based research tool wherein the format incentivized true responses over socially desirable response. The research methodology was recognized by The Esomar Congress 2015 where Final Mile won the Best Case History award.
Innovative tools for learning voter’s preference also exist. The Iowa Electronics Market established back in 1988 is one of the early pioneers. Even changing the way the question is framed can have a significant impact. For example, instead of framing the survey question as which candidate are you likely to vote, a better question would be which candidate is likely to win. So while we monitor the election outcomes, it will also be interesting to study the prediction accuracies of the different research tools.
Image Credit: Indian Panorama