Do Pollsters Ask the Right Questions?

[vntd_hero_section images=”51423″ heading_dynamic=”IMPACT” subtitle=”” button1_label=”” height=”custom” height_custom=”400″ heading_font_size=”95px” tooltips=”%5B%7B%22tooltip_title%22%3A%22Hello%20there!%22%2C%22tooltip_url%22%3A%22%23%22%2C%22tooltip_placement%22%3A%22top%22%2C%22tooltip_depth%22%3A%220.42%22%7D%5D”]

Do Pollsters Ask the Right Questions?

[vc_single_image image=”61397″]

April 2018

Most pollsters attempting to predict election outcomes ask participants who they plan to vote for.

As the results of the 2016 U.S. presidential election showed, these “own-intention” polls have their defects. A new study suggests that a different approach might be more accurate.

Researchers from the Santa Fe Institute, working with academics from other institutions — including ones from Germany and England — instead asked participants who they think their friends and neighbors (called “social contacts” in the research) will support. Testing the system in both the most recent U.S. and French presidential elections, the researchers found that this “social-circle” method made for more accurate predictions.

The differences are not astounding — and they are freighted with impenetrable statistics — but they are significant nonetheless. In the U.S. elections, this new approach not only provided a closer tally in the popular vote but also was more accurate in forecasting the number of electoral votes and “were particularly useful for predicting election outcomes in a priori defined ‘swing states.’”

How the researchers account for the differences in the two models might be as instructive as the differences themselves. One possible explanation, they say, is that people are reluctant to report that they favor a “potentially embarrassing option [but] could nevertheless be willing to report that their social circle favours [sic] it.”

That the approach seems to have worked in two countries with different electoral systems also argues for the effectiveness of social-circle polling.