Getting Real About the R in RPIE

Adam Probolsky is president of Probolsky Research

It can be painful watching communicators operate with one hand tied behind their back – having to create campaigns from assumptions and gut feelings when the real answers are available from a poll.

If you have any doubts about the utility or efficacy of polling, talk to any peer who has had access polling data in the past. They will tell you that they never want to go back to guessing, and that knowing made their job much easier.

So why do so many public sector communication professionals launch large-scale outreach campaigns without statistically valid research? Budget is a major factor. But there is also a lack of understanding about how to engage in the research process, how to get statistically valid data, and how to get a representative response to a poll. While there are plenty of self-directed survey tools available to you, the polling process requires you to hire a pollster.

Let’s demystify this process and give more PIOs the tools to get the most out of the research part of the researching, planning, implementing, and evaluating (RPIE) process.

Step 1 – Develop the poll questionnaire with your pollster: Both science and art go into this process, but the most important thing is accuracy – make sure everything is either factual or a message that your audience might hear from opponents. You should test the opposition’s message, even if it’s not truthful. Also, don’t forget anything. Pollsters can analyze and reanalyze data in 100 different ways. But we cannot get you data on a question that we didn’t ask. A typical poll might have 15 – 45 questions, sometimes more.

Step 2 – Define the target population: Generally, this means the residents (or voters) in your political subdivision – city, county, or special district. But sometimes you’ll want to include unincorporated county islands or residents beyond your service area, like when you are considering annexation. Your GIS team is often important here, helping to map the area if you are using a geography that isn’t already pre-defined by census or voter boundaries.

Then the pollster develops the sample – the list that includes your targeted population. We often use consumer data purchased from vendors. These lists include virtually every adult in the U.S. regardless of age, immigration status and even includes the unhoused. If the poll is related to an election, we can use voter registration lists sourced from county registrars of voters and then have them enhanced by a data vendor. These lists include names, addresses, phone numbers (land lines and mobiles) and emails. They also have lots of other demographic information too. To achieve representative results, your pollster should use a stratified random sample. Your pollster needs to pursue, and get, the right mix of all demographics in the community to make sure the poll is truly accurate.

Next, we figure out the appropriate sample size – how many respondents you want to include in your poll to get meaningful results. This is typically 300 or 400 for just about any city or county and as many as 1,200 for a statewide poll. The sample size determines the margin of error (which is simply a formulaic function of math) and would typically be 3 to 6%. Hint: There is very little reason to conduct a poll of more than 400 completed respondents in virtually any city, county, or special district. This would yield a +/-5% margin of error.

Step 3 – Conduct the survey: Polls can be conducted in numerous formats, but they are most accurate when you use multiple modes like phone and online. Sometimes mail is involved. Since most of us communicate mostly online, think text, email and social; it is hard for a poll to be accurate if it does not include online via email and text at a minimum.

Once the poll is done, the analysis is conducted, and reporting is developed. Understanding the topline results doesn’t require a background in statistics. But there is value in using advanced software for diving into the numbers and finding the results that really drive behaviors and opinions. The results should be reported in different ways for each audience within your agency and possibly externally too.

If you get pushback internally for wanting to conduct a poll with a pollster, just explain that it’s kind of like designing a bridge, you would never do that without an engineer.

Return to CAPIO Communicator