What are Public Opinion Polls, and How do they Work?

The latest polls all claim to have key insights into Canadians' thoughts and opinions about top issues like voting intentions, U.S. relations, and more. But what do polls really tell us?

You may have seen the latest Ipsos poll showing the Federal Liberals shockingly taking the lead over the Conservatives (something not seen since 2021).

But looking at another poll — like Abacus Data’s latest federal poll — they claim that the Liberals are only polling at 28%.

Then you hear the caveat: “…but polls don’t predict elections”.

It’s incredibly confusing, so I’d like to walk through the following items to explore (1) what public opinion polls are, and (2) how polling works.

What is a “Public Opinion Poll”?

A public opinion poll is a method of measuring the views, beliefs, and attitudes of a targeted population through well-designed surveys.

As the Pew Research Centre puts it, “the general goal of a poll is to describe or explain something about a population of interest”, such as registered voters or the general public.

For example, Angus Reid Institute recently explored what Canadians think of U.S. tariff threats, and whether it will change their consumer behaviour.

How Polling Works

Polling is an effective tool to collect insights into a target population’s thoughts and opinions. It does this by effectively asking questions of a small number of people (the “sample”) who represent that larger group (the “population”). How does it do this?

A good poll has 3 key elements:

1. A set of people to ask (ie “sample”)

2. A set of questions (ie “questionnaire”)

3. a way of connecting the two (the “mode of interview”)

1) How do Pollsters Select a Sample?

The key to having a good sample is ensuring it accurately represents your target population. Now, you can “poll” your family to get a sense of “what Canadians think about Trump”… but that wouldn’t be an accurate view of what Canadians actually think about Trump.

A good poll needs a diverse sample that effectively represents the target population.

The best way to select your sample is through random sampling.

How does this work? Typically, random sampling gives everyone in your target population an equal chance of being included.

But random sampling isn’t perfect, and not all pollsters use it. Why? Well, not everyone in your sample can be contacted, and some don’t agree to participate…for starters.

Pollsters who do use random sampling can compensate for these drawbacks by using a technique known as “weighting”.

How Weighting Works

In essence, weighting boosts the voices of people belonging to groups less likely to participate in your poll.

For example: let’s say a pollster like Nanos wants to understand how young adults feel about affordability in Canada.

In trying to select an effective sample group, Nanos would look to include different members of that target population — such as university students, working professionals, recent graduates, etc.

A good pollster like Nanos may recognize that current university students may be less likely than recent graduates to participate in polls (eg they might have fewer financial means, unable to take time off, etc). And so Nanos could “weigh up” existing university student participants to match their actual share of the population.

Here’s a great graphic that gives more detail.

An Interesting Alternative: Opt-In Sampling

But what if the pollster doesn’t use random sampling?

An effective alternative is getting participants to volunteer to participate (or “opt in”), often in exchange for small rewards (eg gift cards).

Opt-in polls often make up a big slice of the market research world, and are used by many companies like Angus Reid Institute to poll about voting.

**It should be noted — opt-in surveys are generally less accurate than those from random sampling, as participants driven by financial or material incentives (eg $100 gift card) can often skew survey results. Op-in surveys are also more at risk for bogus respondents — where participants given insincere or ineligible responses just to get the survey reward.

2) How do Pollsters Design a Questionnaire

The absolute heart of any good poll is the set of questions asked (ie the “questionnaire”).

But this can be very challenging. Leading questions (eg “Why is Montreal better than Toronto?”) can introduce bias and may prompt participants to give answers that may not reflect their actual opinions.

So a good pollster knows how to ask questions that are simple, clear, and understandable for participants from a diverse number of backgrounds.

Let’s look at Abacus Data’s recent poll as an example.

In attempting to understand Ontario voters’ thoughts on the recent provincial election results, Abacus Data uses clear, direct questions that allow participants to respond directly about their opinions on issues of affordability, U.S. relations, healthcare, etc.

3) How to Pollsters Connect a Sample with their Questionnaire

This is the last piece to any effective poll. A good pollster needs to effectively connect sample with their questionnaire.

Now, there are many different ways that pollsters use to connect their sample with their questionnaire, and I will mention just a few of the top methods.

A. Issue Polling

Issue polling is focused on opinions and experiences, which includes what people know about an issue, how they think it affects them, and what they think should be done to address it.

Abacus Data is well-known for this approach, as shown below.

B. Market Research Surveys

Often, firms (such as Leger) will use this approach to understand consumer behaviour or preferences.

Have you been asked by a retailer like Costco or a restaurant you just visited about the quality of their products or services? That’s a market research survey!

C. Election Polls

A third kind of poll is of course — election polls. These attract a huge amount of attention, as we’ve seen recently.

Some well-established Canadian election pollsters include Abacus Data, Ipsos, Nanos, Angus Reid Institute, and Ekos (as pictured below).

Coming Up

Now that we have a general idea how polling works, in my follow-up post, I’d like to explore what polls can (and can’t) tell us about public opinion.

As always, appreciate you all for supporting this Substack. If the article resonated with you, feel free to share with your friends and colleagues!