Statistics Definitions > Undercoverage definition
What is Undercoverage?
When some members of your population aren’t represented in a sample, it’s called undercoverage. In other words, some members of a population have zero chance of being included in the survey or experiment. It’s a type of selection bias that often occurs in convenience sampling, where you collect a sample that’s easy to obtain, like shoppers at your local mall or residents at a town meeting.
At first glance, collecting data at a mall or a town meeting might seem like a convenient way of getting the data you need. However, by only asking shoppers at a mall about their opinions, you risk undercoverage of several populations. For example:
- People who only shop locally.
- People who are housebound.
- People who don’t like the mall.
The same goes for voters at a town meeting. If your goal is to poll voters about some local issue, then the town meeting is probably a guaranteed way to get undercoverage of several key groups, like:
- People who work evenings (and therefore miss the meetings).
- People who do not understand local government workings.
- People who don’t have transportation to get to meetings.
The above two examples are extreme cases of detectable undercoverage: it’s pretty obvious that you’re excluding certain groups. However, undercoverage is sometimes difficult to detect.
Even the US Census has problems with undercoverage. So much so, that they have developed complex techniques to deal with nonresponse rates, which tend to be higher in males than in females, and higher in blacks than in non-blacks.
The Literary Digest Poll
Perhaps the most famous case of undercoverage occurred in 1936, when a Literary Digest survey predicted that Alfred Landon would win the presidential election over Franklin D. Roosevelt. Roosevelt actually won by a landslide (62% of the vote). The main reason for undercoverage in this example was that the poll was heavily skewed towards wealthier voters (for example, people who had magazine subscriptions, club memberships and telephones. However, this was the time of the Great Depression, when the votes of the poor were extremely important to the election outcome. And this was the group that was severely undercovered. The magazine went out of business two years after their disastrous poll.
References
Dodge, Y. (2008). The Concise Encyclopedia of Statistics. Springer.
Levine, D. (2014). Even You Can Learn Statistics and Analytics: An Easy to Understand Guide to Statistics and Analytics 3rd Edition. Pearson FT Press
Vogt, W.P. (2005). Dictionary of Statistics & Methodology: A Nontechnical Guide for the Social Sciences. SAGE.