The polls are all wrong. Election polling data still relying on dated information gathered from surveys completed via landlines.
According to a recent government study, nearly 43 percent of households have ditched landlines for cell phones.
This has turned out to be a huge problem for pollsters during election time.
Years ago, when home phones were the norm and caller ID wasn’t ubiquitous, people would answer their phones without knowing who was on the other end (weird, right?). Thanks to smartphones and their infectious popularity, everyone has the option of ignoring unknown numbers and letting the call go to voicemail.
Ignore a random number, and they don’t leave a message? Probably a telemarketer – bullet dodged.
This is perfect for the individual who doesn’t want to be bothered by an untimely sales pitch, but when pollsters are trying to gather election data, this lack of answering makes things pretty tricky.
Research has shown that between the late 1990s and 2012, the response rate of polling calls has dropped from almost 40 percent to just 9 and is continuing to decrease.
When it comes to telemarketing, many people take advantage of the National Do Not Call Registry. Enacted in 2003, this registry makes it illegal for telemarketers to call individuals unless they already have done business with them in the past. Luckily for pollsters, they are exempt from these laws and can still freely make their calls.
Still, many survey researchers understand that people who don’t want to be called deserve respect. This is why many organizations maintain internal do-not-call lists to reduce irritating a significant portion of the population.
In the past, pollsters and survey takers would have access to thousands of resident’s phone numbers from public directories such as the white pages. Now that cell phones have all but wiped out landlines, where are they getting their numbers?
This is where things get a bit creative.
Survey companies create and sell lists of potential cell phone numbers in the U.S.
According to the Pew Research Center,
“They start with the fact that certain area codes and exchanges are dedicated only to cellphones. For area code and exchange combinations that include both landlines and cell phones, additional work is done to identify the specific blocks of numbers assigned to cell phones. Once the relevant area codes and exchanges (and, if necessary, specific blocks) are defined, the sample vendors append all possible last four digits to each. For instance, if they know the 202 area code and 555 exchange within that area code are only used for cellphones, then they can add every number from 202-555-0000 to 202-555-9999 to their list. They then repeat that with every area code and exchange combination known to be used only for cellphones to create a (very long) list of all possible cell phone numbers in the United States. From that list, telephone sample vendors draw a random sample of phone numbers to be used for a particular poll.” (source)
A huge question that arises from this landline-cellphone transition is whether or not the actual data that gets collected from phone calls ends up biased due to demographic differences. According to an article in the Washington Post, two dynamics have to be at play for response bias to exist in polling calls.
- Certain groups of people will be more likely to answer a pollster’s calls.
- Those who do answer must have differing political views than those who don’t answer.
“An elderly white woman is 21 times more likely to answer a phone poll than a young Hispanic male,” says David Shor, a data scientist for Civis Analytics. “So polling samples are often inherently misrepresentative.” (source)
What’s the solution?
Surely analysts aren’t taking this one-sided data and reporting as fact. With an incomplete pool of respondents, adjustments have to be made and organizations like Civis work to find solutions.
Using what you already know about a population is extremely beneficial in ensuring you’re making the right phone calls. This, along with incorporating online surveys, is key to gaining accurate results.
Instead of randomly dialing hundreds of thousands of numbers to reach your target sample size, Civis and other analytics firms peruse their databases for individuals who meet specific criteria. From there, they can determine whether a call or online survey is more likely to get a response.
It’s not a perfect system, but as times change and people have more control over who they communicate with and how, polling tactics have to be adapted.
The key is determining who uses what methods of communication, using this information to reach an equal number of people from varying demographics, and trying to reduce as much bias as possible to produce accurate statistics.
It’s not an easy task, but as technology and the population changes, so must the methods of gathering unbiased, useful data.