Why people fudge the numbers on surveys (and why our calculator doesn't care)

Ask a room full of people a personal "how many" question and you'll get a lot of answers ending in zero, some careful shading, and at least one person who says "oh, about ten?" with a face that means "I have no idea." That's not a character flaw. It's one of the best-studied habits in survey research.

This guide looks at why self-reported numbers wobble and why a joke calculator like ours is happy to take your best guess.

"Fudging" isn't always fibbing

When survey numbers come out wrong, it's tempting to assume people are lying. The research tells a kinder story, with at least three things going on:

• Answering for an audience. Nudging an answer toward what feels acceptable.
• Estimating instead of counting. Making a quick guess because the exact number isn't stored anywhere in your head.
• Rounding. Reaching for a tidy number like 10 or 20 because that's how fuzzy memories come out.

Only the first is about impressions. The other two are just memory.

Answering for an audience: social desirability bias

Survey researchers call the first habit social desirability bias. In a 2013 review in the journal Quality & Quantity, Ivar Krumpal sums it up in one line: "Due to self-presentation concerns, survey respondents underreport socially undesirable activities and overreport socially desirable ones" (abstract¹). In plain words, we tend to answer the way we'd like to be seen.

A major 2007 review in Psychological Bulletin by Roger Tourangeau and Ting Yan found that misreporting on sensitive topics "is quite common" and "largely situational" (abstract²). It depends on whether a person has anything embarrassing to report, and on how the survey is designed. Their description of what's happening is almost gentle: respondents "edit the information they report to avoid embarrassing themselves in the presence of an interviewer."

One clever experiment shows how much the setting matters. In a 2003 study, psychologists Michele Alexander and Terri Fisher gave 201 unmarried college students, ages 18 to 25, the same questionnaire about their attitudes and past experience under three conditions, according to Ohio State University's summary⁴:

• One group was hooked up to what they were told was a lie detector. It was an old model that didn't actually work.
• One group answered alone, with a promise of complete anonymity.
• One group was led to believe a researcher might read their answers, with the researcher sitting just outside an open door.

Women who thought their answers might be read reported an average of 2.6 past partners; with the "lie detector," 4.4; anonymous, 3.4. Men's answers moved much less, from 3.7 to 4.0. The study's abstract³ concludes that some differences in what people report "reflect responses influenced by normative expectations for men and women." People weren't simply honest or dishonest. They were answering for the room.

Estimating instead of counting

For a lot of "how many" questions, there's no exact number in your memory to look up. You have to build one on the spot.

In 1999, researchers Norman Brown and Robert Sinclair of the University of Alberta asked 1,787 university students to report their number of lifetime partners and then describe, in writing, how they came up with it (paper⁵). As the paper puts it, people "do not necessarily keep a running count," so they "must generate a suitable numerical response" using some strategy.

Among the students with the most experience to remember, the researchers found a few common approaches:

• Enumeration (39%): recalling each person and counting them up.
• Retrieving a tally (29%): a number they already kept track of. One student wrote, "I kept track in my diary."
• Rough approximation (17%): a quick guess, often with a shrug attached, like "Rough guess, give or take 1 or 2 partners."

The strategy made a big difference. Rough approximations produced the largest numbers on average and careful counting the smallest. Men were more likely to approximate and women more likely to count, and people who used the same strategy gave similar answers whatever their gender. When the question shrank to just the past 12 months, the gap between men's and women's answers was no longer statistically significant. The authors' conclusion is generous: the differences come from how people estimate, "not because they intentionally misrepresent" their histories.

A much bigger study came to a similar place. In a 2019 paper, Kirstin Mitchell and colleagues analyzed Natsal-3, a large national probability survey of the British population (abstract⁶). People who said they estimated reported far higher numbers than people who counted. Adjusting for counting versus estimating, along with a couple of other factors, helped explain a large share of the gap between men's and women's answers. The authors suggest that surveys could encourage "counting rather than estimating."

Why answers pile up at 5, 10 and 20

When people estimate, they don't pick random numbers. They pick round ones. Researchers call this "heaping."

In the Brown and Sinclair study, 79% of the people who used rough approximation gave a number divisible by 5. For people who counted one by one, it was 38% (paper). In Natsal-3, the full paper reports "a tendency to report partner numbers ending in 0 and 5, particularly with higher numbers of partners."

This isn't special to personal questions, either. A classic 1990 study by Janellen Huttenlocher, Larry Hedges and Norman Bradburn looked at how people report how long ago something happened (abstract⁷). People leaned on rounded values that, "although stated in days, actually represent larger temporal categories (e.g., 14, 21, 30, 60 days ago)." Two weeks, three weeks, a month, two months. Our memories file things in friendly bins, and the numbers come out shaped like the bins.

So if your answer to a big "how many" question is a suspiciously neat 20, congratulations. You're completely normal.

What researchers do about it

Because these habits are so predictable, good surveys are built around them. Tourangeau and Yan's review stresses that misreporting depends on "design features of the survey." Natsal-3, for example, had people answer the more sensitive questions by typing them into a computer themselves, rather than saying them out loud to an interviewer, according to the Mitchell paper. It also asked people who reported five or more partners how they worked out their number: whether they just knew it, counted, estimated, or remembered some and estimated the rest. Krumpal's review makes the same general point: surveys get better data when they "reduce respondents' discomfort."

Privacy, in other words, isn't just polite. It's how you get honest answers.

Why our calculator doesn't care

CountMyBits isn't a survey. It's a joke with a calculator attached, and we built it so there's nothing to fudge for:

• No name, no account. We never ask for your name or email, and there's nothing to sign up for.
• Your answers aren't stored. A scorecard is saved only if you choose to share it, and even then your individual answers aren't part of it.
• Shared pages don't hang around. Shared scorecards are automatically deleted 90 days after they're created. Want one gone sooner? Email the link to [email protected] and we'll remove it on request.
• Rough guesses are welcome. Our How it works page says it best: "Nobody is checking your math, least of all us."

The math itself is silly on purpose. It multiplies small guesses over the years and rounds generously, so the number on your card is a punchline, not a measurement. Estimate, round to the nearest ten, go with your gut. The calculator will be equally delighted.

"Is my number normal?"

The research gives a surprisingly comforting answer. The same person can produce different numbers depending on whether they count or guess, who they think is reading, and how far back they're asked to remember. A number that wobbles that much was never a good yardstick for comparing people. There's no right number, no wrong number, and no number that says anything about who you are.

The short version

• Survey numbers wobble because of social desirability bias (answering for an audience), estimating instead of counting, and rounding.
• People who estimate tend to give bigger, rounder numbers than people who count one by one.
• Self-reported counts pile up at numbers ending in 0 and 5.
• Privacy and good survey design help people answer more candidly.
• CountMyBits asks for no name or account, and shared pages auto-delete after 90 days.

Your number is your number. Round up, round down, or guess wildly. Take the quiz (18+). Curious how small guesses turn into big numbers? Read Fermi Estimation for Fun.