Polling through a pandemic: methods and madness

karen-cyphers-sm
karen-cyphers-sm
Share:

Polling through a pandemic: methods and madness

From managing canine “coworkers” to selecting the right virtual Zoom background, many Floridians have been navigating new waters in order to do our jobs amid COVID-19.

To gauge these changes, we’ve been surveying large, random samples of Florida voters weekly since the pandemic exploded in mid-March. Our goal was, and still is, to glean what people are thinking and feeling through these crazy times, and to see what trends may take shape.

Nobody is paying us to do this; rather, we are gathering data and sharing results through columns like this one and in graphics presented in Florida Politics’ evening email, Last Call.

As far as data trends, the results are largely as expected: feelings divide most along political lines; everyone is frustrated and worried, albeit sometimes about different things; and negative emotions increase when cases are spiking.

But through these months, it’s become clear that the practice of survey research itself may be shifting. A few things stand out:

  1. Response rates are way up. It could be boredom, being home, perhaps more tied to devices, perhaps more interested to share feelings and thoughts on a topic that’s actually of interest. While we would normally need to spend 2-3 days in the field in order to gather a minimum of 1,000 responses, we’re now able to open and close surveys in a much shorter period of time, or leave them in the field a little longer if we want to gather data from a larger sample.
  2. Representativeness is up, too. Where typically it’s more challenging to attain a sample that is equally representative of young and non-white voters, in the last few months our samples have been increasingly diverse. For example, voters ages 18-34 previously made up about 20% of our survey samples even though they make up about 30% of Florida’s population – yet in the past many weeks, the balance of young voters in our samples is quite close to their portion of the population at large and nearly matched to their portion of likely voters.
  3. But political representativeness is down. Typically our breakdowns of Rs, Ds, and NPAs fall close to their share of the general population, with Republicans responding to surveys just slightly less than Democrats. But not lately. During the pandemic, the sample has been skewing heavily Democratic, to the point that we have to use one of two scientifically valid methods to balance the proportions of our final responses. We know this method works, and as recently as the Aug. 18 primary we accurately predicted the outcomes of the three races on which we polled – each within less than 1% of the actual vote totals.

So what does this mean? Specifically, at least as it relates to polling methodology, the following questions feel relevant: Why are Republicans more reluctant to participate now than they were before? Or could it be that Republicans have remained about as reluctant as ever to participate, but Democrats are unusually eager to share their feelings on COVID-19 specifically?

In 2016, our polling accurately forecast Donald Trump’s Florida victory despite other polls showing him well behind Hillary Clinton. In part, this was because we adjusted for the assumption that right-leaning voters may be less likely to disclose their feelings to pollsters (or friends, even), thus leading us to oversample this population to find a balance that we expected would resemble the actual electorate. Based on what we’re seeing today, this gap in survey participation has widened – and may further complicate the already challenging task of estimating margins.

While Republicans may be less responsive to surveys about COVID-19 than others we have fielded, a different group has been piping up a whole lot more than before: conspiracy theorists.

Prior to COVID-19, we’d get an occasional email commenting about how we worded a question, offering a correction, or asking if we were polling for a specific client. We welcome these messages, and gladly respond to them. But recently, we’ve started to receive a new variety of response to survey invitations. These tend to be profanity-laden or just plain wrong,  accusing us of working for someone we do not (i.e., the Trump administration, or Democrats) or espousing any variety of conspiracy theories – such as the two most recent, one of which read, “It’s a fake pandemic to initiate a Psychological Operation to cover up Antify Jewish Child Molesters,” followed by one proclaiming “Another scumbag Trump Lie… As a hacker, just don’t try me lol.” 

To discourage such theories, I plan to adjust our survey invitation language to make it even more clear that our firm isn’t partisan, isn’t being paid to conduct these, and isn’t pushing an agenda.

In the meantime, we’ll continue to track how Floridians are navigating through the rest of this madness, while learning our own lessons accordingly.

Karen Cyphers, Ph.D., is a partner and Vice President of Research with Sachs Media.

Scroll to Top