Public opinion polls are a professional, mathematical, and sophisticated enterprise. Polls in respected publications are generally conducted utilizing state-of-the-art techniques. If people use a critical eye and understand the nuance of drawing inferences about a population from a random sample, then they’re much more likely to learn something valuable when they come across polls about politics.
The Concepts of Liberalism and Conservatism
Liberalism is generally characterized by the following general values: reduce economic inequalities, champion the rights of minorities or those who are disadvantaged, tolerate a broad range of people and behaviors, and support government intervention in the economy to achieve social goals.
In contrast, conservatism today is generally characterized by the following general values: support for the social and economic status quo, distrust of government for solving social problems, emphasis on personal liberty and self-reliance, and strong faith in the free-market and private enterprise.
These liberal and conservative ideologies are the most common representations of political ideas and values today. A shorthand way to think about this, liberalism most often values equality, whereas conservatism most often values liberty.
The Importance of Sampling in Taking Public Opinion Polls
There was a magical golden age in the United States between the 1940s and the 1990s where an entire industry of polling was born, and it shined. Taking public opinion polls during that time was wonderful because most Americans had land-line telephones in their homes and telephone numbers were organized in geographically systematic ways.
This was important for polling because it meant that if a pollster wanted to poll a random sample of Americans, the pollster could use a technique called random-digit-dialing or RDD. The RDD technique was incredibly successful for a period of time because it produced truly random—or close to random—samples of United States households.
But how the sample is drawn is really important. The wrong sampling technique can make the data meaningless. The sample needs to be representative of the population the pollsters want to say something about.
This is a transcript from the video series Understanding the US Government. Watch it now, on Wondrium.
The Difference Between an Election Forecast and Surveying
It’s important to recognize that an election forecast is not the same thing as surveying Americans about their preferred pizza topping. A surveyor seeks opinions or attitudes on a specific topic or question at a particular point in time, but an election forecaster seeks to know what a voter will do at a future date, or more precisely, on Election Day.
Predicting what people will do in the future is a particularly difficult task for a pollster. One of the key things an election forecaster must attempt to do is guess which voters will actually turn out to vote.
Different polling organizations use various models of forecasting voter turnout, and to some extent, the differences between election forecasts in 2016 came down to the different assumptions that polling organizations used about voter turnout.
Learn more about the fundamentals of elections and voting.
Polls Taken Prior to the 2016 Election
Of the polls that were taken in 2016, many of them were very close to accurate. The reason is that most of the polls took random national samples and asked Americans if they were likely to vote and who they were likely to vote for.
Most of these polls predicted that, nationally, more Americans would vote for Hillary Clinton over Donald Trump, which they did. While Trump was elected president because he won more electoral college votes, it is also true that Clinton won more votes overall than Trump.
Thinking of the polls state by state, rather than nationally, some states had pre-election polls that were decidedly off the mark, and it turned out that these really mattered. In a typical election, a few local state polls that don’t quite get it right do not make a difference in anyone’s expectations.
But in 2016, a few local polls getting it wrong mattered a lot for people’s expectations. Not only did some state polls wildly over-predict Democratic turnout, but those polls also underappreciated how close the election would be.
The Accuracy of the 2016 Polls
The problem with the polls in 2016, then, wasn’t that they were wrong—in fact, most of them were right—but in how the polls were interpreted. On election eve, most forecasters had the probability of Trump winning the election somewhere between 1% and 30%. By all accounts, it certainly looked like Clinton had a better chance at winning the election than Trump, but events that have a 30% chance of happening, happen three out of every 10 times.
If the weather forecast says there’s a 30% chance of rain, do people take an umbrella? They might. So, the polls in 2016 were mostly accurate, and pollsters have developed newer and more sophisticated methods of drawing random samples of Americans in order to measure their political attitudes and forecast elections.
Learn more about how American democracy works.
Three Tips for Meaningful Polling
First, notice the size of the sample. If the poll does not report any descriptive statistics about its sample—like how many people were questioned—it’s probably not a poll worth listening to. However, it takes a surprisingly small number of respondents to draw reasonable inferences about very large populations.
Second, the poll should report the margin of error. The margin of error tells how precise the estimate is from the sample. The pollsters can make an estimate more precise by increasing the sample size if they have the resources to do so. When a poll reports the margin of error, one should think of it like bookends around the estimate.
Finally, the interpretation of the poll should match the numbers that are presented. If a writer says that candidate A is ahead of candidate B, but the polls say that A’s support is 35% and B’s support is 32%, and the poll has a margin of error of 4%, then one really cannot tell which candidate is ahead because the estimates are within the margin of error.
Common Questions about Ideologies and How Public Opinion Polls Are Measured
How the sample is drawn is really important. The wrong sampling technique can make the data meaningless. The sample needs to be representative of the population the pollsters want to say something about.
An election forecast means asking people about who they’re going to vote for in the future, and a survey asks people about their opinions or attitudes at any one moment.
Most of these polls predicted that, nationally, more Americans would vote for Hillary Clinton over Donald Trump, which they did.
However, although Clinton won more votes overall than Trump, Trump was elected president because he won more electoral college votes.