Do sale signs fool your brain? A research idea for high school students
Can a crossed-out "was" price make you buy? See how researchers test sale signs, and how to run a small survey experiment at your school.
"I always buy things I don't need when there's a sale." People say it with a small laugh, often while holding a bag they didn't plan to carry home. It sounds like a confession about willpower. An economist hears something else: a claim about cause and effect that you can test. Does the word "sale," or a crossed-out "was" price, change what you choose? If it does, by how much, and why? Here's how researchers test it, and how you could run a small version at your school.
How a "was" price might change what you pay
Picture a jacket tagged "was $50, now $40." You didn't come in for a jacket. You leave having spent $40, and it feels as if you saved $10. In the simplest model of buying, the old price shouldn't matter at all: you compare $40 with what the jacket is worth to you, and you decide. Yet the $50 seems to do some of the work.
In a 1974 paper in Science, psychologists Amos Tversky and Daniel Kahneman described a mental shortcut they called adjustment from an anchor. When a relevant number is in front of you, you start from it and adjust your estimate from there. Shortcuts like this usually work, they wrote, but they lead to "systematic and predictable errors." A "was" price may be that kind of anchor, pulling your sense of the jacket's value toward $50. Their paper was about estimates in general, not price tags, so whether a "was" price acts this way is what you'd test.
A second explanation is less about being fooled. In a 1998 paper built on a theoretical model, researchers Eric Anderson and Duncan Simester argued that sale signs tell shoppers who don't know market prices which products are relatively cheap. On that view, a sign carries information, and following it can be sensible.
Regulators take "was" prices seriously too. The U.S. Federal Trade Commission's Guides Against Deceptive Pricing treat a "was" price as legitimate when the item was really offered at that price, on a regular basis, for a reasonably substantial period. If a store inflates a price just to advertise a big cut later, the guides call the bargain a false one.
A single number on a tag can become the starting point for what the item seems to be worth.
Have a question you want to investigate?
Explore our faculty-guided programs, or tell admissions what you are curious about.
Most of this research is a controlled experiment: keep everything the same (the product, price, photo and page), change one thing, such as the sign, and compare what people choose.
It also needs random assignment. Shoppers who walk over to the sale rack may already differ from shoppers who don't. When chance decides who sees which version, the groups start out alike, so a difference at the end points to the sign.
The idea works at different scales:
Field experiments: real customers get different versions of a real catalog or store page, and researchers count orders.
Survey experiments: people see one version of an ad and say what they'd do.
Store data: sales from weeks with and without signs. It's real buying, but the signs weren't placed at random, so other causes are harder to rule out.
A sign is one of the easiest things in a store to change, which makes it a clean thing to test.
What a 2001 catalog study measured
Anderson and Simester tested their model in "Are Sale Signs Less Effective When More Products Have Them?", published in Marketing Science in 2001. They used four kinds of data: past orders from a women's clothing catalog, a field test that mailed different versions of the catalog to randomly selected groups of customers, survey answers about sample catalog pages, and grocery store scanner data for frozen juice, toothpaste and tuna.
Sale signs raised demand, in some cases by more than 50%. Signs also worked less well when more items carried them, and total category sales were highest when some products had signs but not all. In the survey, a sign made people think the item was less likely to be sold at a lower price later, though that effect was weaker when many items had signs. The authors conclude that signs lose power when they're everywhere partly because shoppers believe them less.
The study has limits. It drew on one catalog and three grocery categories, and it came out in 2001, before phone shopping was normal. The "more than 50%" figure describes some cases, not a typical one. And it tested "SALE" signs, not crossed-out "was" prices, which are a related but separate cue.
A survey experiment you can run at school
You can't mail out catalogs, but you can run a survey experiment. Plan on five to eight weeks in total, much of it waiting for approval.
1. Collect 30 real sale ads (2 to 3 hours over a week). Screenshot ads from shopping sites and apps you already use. Pick items a student might buy, roughly $10 to $60 (or the same range in won), and keep only ads that show both a sale price and a "was" price. In a Google Sheets file, record each item's category, sale price and "was" price.
Online stores are an easy place to collect real ads: save the item, its sale price and any 'was' price.
2. Make versions with and without "was" (4 to 6 hours). Rebuild every ad in one plain template in Google Slides, with the same photo, name and sale price. Version A adds the crossed-out "was" price. Version B shows the sale price alone. Rebuilding strips out logos and design differences, so the "was" price is the only change.
Next, make two surveys in Google Forms. Students in Korea can also use Naver Form, a Korean-language survey service. Form 1 shows ads 1 to 15 with the "was" price and ads 16 to 30 without it; Form 2 does the reverse. Nobody sees both versions of one item, yet every item gets tested both ways. For every ad, ask two things: "Would you buy this at this price?" (yes or no) and "What is the most you would pay for it?"
3. Survey classmates, with school approval (1 to 3 weeks for approval, about 1 week to collect). Show your teacher or school the forms before anyone answers them. Open each form with a consent page: taking part is voluntary, people can stop at any time, and no names or email addresses are collected. Classmates under 18 need a parent's permission. Planning to enter a science fair? Read its rules early. ISEF, for example, requires approval from an IRB (an ethics review board) before you recruit anyone, and for minors the survey must be attached to the parental permission form. To assign forms at random, shuffle printed slips carrying the two links or QR codes and hand them out.
4. Compare how many would buy (3 to 4 hours). Export the answers to Google Sheets. Give each student two scores: how many "was" ads they'd buy and how many plain ads they'd buy. Show the two averages in one bar chart. To check whether the gap could come from chance alone, run a paired-samples t-test in jamovi, a free statistics program (T-Tests, then Paired Samples T-Test). The free textbook Learning Statistics with jamovi explains it. Don't pool all the answers into one chi-square test: each student gave 30 answers, and that test assumes independent answers. Score the price question the same way: divide each answer by the item's sale price, then average it for each version.
Know the limits. A few dozen students can only reveal fairly large differences. Saying "I'd buy it" isn't paying, and one school doesn't speak for every shopper. Seeing both versions in one form may hint at what you're testing. And a higher price answer fits both explanations above, anchor or information, so the survey shows whether the cue works, not why.
After about a week of writing, you'll have a poster with one chart and one table, or a short paper of four to six pages: introduction, method, results, limits.
From "sales make me spend" to a question you can answer
Too big
Workable
Do sales make people buy things they don't need?
At my school, does a crossed-out "was" price raise the share of students who say they'd buy a $10 to $60 item?
How do anchors affect what we pay?
Do students say they'd pay more for an item when its "was" price is 60% above the sale price than when it is 20% above?
Do stores run too many sales?
In a set of 10 ads, does a "was" price raise buying intent less when 8 ads carry one than when only 2 do?
In each row on the right, you know who you'll survey, the one thing you'll change and what you'll measure. The third echoes the 2001 finding on a survey scale.
A survey is one way to write a paper on sale signs. A literature review and a research proposal are two others, each a full paper in its own right. For a review, you read published studies and compare them: did studies of real orders (field tests and store data) find smaller effects from "was" prices than surveys that asked what people would buy? Google Scholar and IDEAS/RePEc, a free index of economics papers, are good places to look. For a proposal, you design a study only a real store could run, such as an eight-week test with an online store that shows a "was" price to half its visitors at random. You'd state the question, the method, the data you'd need (anonymized order records from the store) and the ethics approval it would require, since real shoppers would take part without signing up. Under the FTC guides, each "was" price would also have to be genuine.
Your sale habit was the question. Now research it.
Anderson, E. T., & Simester, D. I. (2001). Are sale signs less effective when more products have them? Marketing Science, 20(2), 121-142. Kellogg School of Management; abstract also on IDEAS/RePEc
Anderson, E. T., & Simester, D. I. (1998). The role of sale signs. Marketing Science, 17(2), 139-155. Kellogg School of Management
Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185(4157), 1124-1131. link
U.S. Federal Trade Commission. Guides Against Deceptive Pricing, 16 CFR § 233.1, Former price comparisons. Legal Information Institute, Cornell Law School. link
Society for Science. ISEF International Rules: Human Participants. link
Navarro, D. J., & Foxcroft, D. R. The paired-samples t-test. Learning Statistics with jamovi. link