US
Presidential Poling Data Reveal A Potential Validity Problem
Do you want to know how people are likely to vote?
Or, do you want to know who is likely to be the next president?
What is the purpose of national polls? Those interested in politics get a glimpse of how people might vote in an election. But how valid are those data?
What is not
obvious is that the national polls are not the most valid measure of who will
become president. That’s not because of problems of accuracy or methodology.
It’s because in the US, the popular vote does not determine who will be the
next president. Hence, national polls lack the level of predictive validity
available in other polls.
Validity refers to how well a measure measures what it is supposed to measure.
Pollsters assess validity in different ways.
The key question is what is the purpose of the poll?
If the purpose is to reveal the likely votes of Americans, then a national poll has a degree of construct validity. If the poll is to predict who will be the next president, then a more valid measure will assess the number of electors a candidate can count on to vote for them.
A poll reporting how voters are likely to vote may be less valid than a poll revealing how many electoral votes a candidate is likely to get.
Construct validity
consists of more than one way to assess the validity of a poll or survey. In addition to the content (questions or items),
we want to know how well a measure (poll, survey) predicts an outcome. That is, we want to
know predictive validity. Because US presidents are not elected by the voters,
national polling data will have lower predictive validity than polling data
focused on the assessment of how many electors’ votes a candidate is likely to
have.
[For those unfamiliar with US politics, electors are the people who represent a state and meet to vote for the next president each January following the national election. States have different rules for determine how their elector’s vote. The group of electors is referred to as the Electoral College. Following the vote of the electors, the US Congress must certify the election. For the process details, see Electoral College | USAGov]
Vital
Statistics
In summary,
the most important polling provides candidates with data that will predict how
many electors will likely vote for them. After removing states’ electors likely fully committed to a candidate from consideration, the focus should be on states that, based on
historical voting and polling data, could go to one candidate or another. These
so called “battleground states” are those vital to winning the national
election; hence, polling data from these states are likely more valid than are
data from national polls.
The
candidate with the most electoral votes wins. The candidate with the most votes
may or may not win.
Validity
Validity is not an all or nothing concept. Instead, validity is a multidimensional construct and each dimension has a range from low to high. Content validity is a judgment by experts about the questions or items pollsters present to the participants. Judges can offer a rating about the likelihood of a question or item to measure the intended outcome.
Predictive validity is a statistical concept, which relies on a formula yielding values that range from 0.0 to 1.00. Statistical measures are not perfect but pollsters can calculate a margin of error. A representative sample of 1,000+ voters is usually large enough to obtain a reasonable estimate of how people will vote. Because there is a range of values, polls should be considered as more or less valid instead of the describing them as either valid or invalid. This is true about most surveys and psychological tests too.
Data reported in terms of percentages of votes for the candidates often have a margin of error of +/- 3
percentage points, which is usually reasonable. The statisticians who work with
national polling companies will be among the best available. So many campaign
decisions will be based on those data.
Polling results from different organizations yield different results so some reporters use an average based on data from polls that have the best methodology, which includes the best sampling strategies.
As to the issue of predicting who will be the next president, we
will likely get a better idea (higher validity) by watching the polling data
from a half dozen or so of those battleground states, which are sometimes
referred to as key states or swing states.
Polling Examples
from 8 August 2024
NY Times report on three key states;
https://www.nytimes.com/2024/08/10/us/politics/harris-trump-battleground-polls.html
The
Washington Post
reports data from 7 battleground states
Post Author
Geoffrey W. Sutton PhD is Emeritus Professor of Psychology who publishes book and articles about clinical and social psychology including the psychology of religion. Website: www.suttong.com
Books available on AMAZON and the GOOGLE STORE
Connections
FACEBOOK Geoff W. Sutton
TWITTER @Geoff.W.Sutton
PINTEREST www.pinterest.com/GeoffWSutton
Read many published articles and book samples on:
Academia Geoff W Sutton
ResearchGate Geoffrey W Sutton
Comments
Post a Comment