Why Do Prices End in 99? The Effect Is Real, Narrow, and Not the One You Have Been Told

Search this and you get one answer, repeated everywhere with total confidence: we read left to right, so we anchor on the first digit and $4.99 registers as four-something rather than five. Some pages add an origin story about cash registers and dishonest shop assistants.

The first explanation is roughly half right, and the half that is wrong is the half that matters. The origin story we could not source at all.

The field experiments

Most of what circulates about pricing psychology comes from laboratory studies with hypothetical purchases. This question has better evidence than that.

Anderson and Simester (2003), in Quantitative Marketing and Economics 1(1), 93-110, ran three field experiments in which price endings were experimentally manipulated in a real retailer’s catalogues, with real customers spending real money.

Their two conclusions, verbatim from the abstract:

“First, use of a $9 price ending increased demand in all three experiments. Second, the increase in demand was stronger for new items than for items that the retailer had sold in previous years.”

And a third, weaker finding: “There is also some evidence that $9 price endings are less effective when retailers use ‘Sale’ cues.”

The authors’ own interpretation is the useful part. These results, they write, “suggest that $9-endings may be more effective when customers have limited information, which may in turn help to explain why retailers do not use $9 price endings on every item.”

That last clause deserves attention. If the nine-ending were a general trick that reliably lifted demand, retailers would apply it to everything. They do not, and the paper’s explanation is that it works by supplying a signal - this is a good deal - to customers who have no other way of judging. Where the customer already knows the item, or where a Sale sign is already making the claim, the ending has less to do.

The mechanism has a boundary condition nobody mentions

Thomas and Morwitz (2005), in the Journal of Consumer Research 32(1), 54-64, ran five experiments to establish when the left-digit effect actually operates. Their abstract states three findings, and the first is the one that corrects the folk explanation:

“First, this occurs only when the leftmost digits of the prices differ (e.g., $2.99 vs. $3.00).”

So $2.99 versus $3.00 produces the effect. $3.79 versus $3.80 does not. The popular account - that we skim the right-hand digits and therefore any nine-ending looks cheaper - predicts an effect in both cases. The experiments say it depends entirely on whether the leading digit rolls over.

Their second finding adds another condition: the effect “depends on the numerical and psychological distances between the target price and a competing product’s price. The closer the two prices being compared, the more likely is the left-digit effect.” It is a comparison effect, not a property of a price seen alone.

Third, it is not about money at all: “it also manifests with other multidigit numbers.” The same thing happens with distances, weights and ratings.

And a field experiment that found nothing

Balance requires the null.

Filippin (2009), in a Department of Economics working paper at the Università degli Studi di Milano, ran a field experiment removing the .99 ending from a product’s price. Descriptive statistics and a single-item multivariate model both suggested sales rose when the .99 was removed - a counterintuitive result.

But it did not survive proper controls. In a difference-in-differences model that also tracked daily sales of substitute products, the effect vanished. The author’s reading is that a different demand elasticity, not the price ending, drove the apparent change: “the treated item does not display significantly higher sales as compared to its substitutes when the . 99 ending price is removed.”

His framing of the field is that the results “tail behind other contributions showing how . 99 endings can be ineffective, casting doubts on their widespread use among retailers.”

This is an unrefereed working paper and we are labelling it as one. But it is a field experiment on real sales, and it found nothing, which is worth knowing when weighing three field experiments that found something.

The origin story we are not telling

There is a history attached to this topic that appears on almost every page about it: that nine-endings were introduced to force shop assistants to open the till and give change, preventing them from pocketing round-number cash payments. Names get attached - a Chicago newspaper proprietor, a department store, the invention of the cash register.

We went looking for a primary record of that and did not find one. No contemporary document, no company record, nothing but the story repeating itself across secondary sources.

So we are not telling it. It may well be true. It is the sort of thing that would be easy to verify if a record existed, and the absence of one after looking is itself worth reporting, given how confidently it is asserted.

What this actually tells you

The nine-ending is not a mind trick that makes you misread a price. On the evidence:

It is a signal. It communicates “this is a low price” to a customer who lacks a reference point. That is why the effect is largest on items the retailer has not sold before, and smaller when a Sale sign is already making the same claim.

It is conditional on the leading digit. $2.99 works because three becomes two. A nine-ending that leaves the first digit alone has no established effect.

It is comparative. It operates when there is a competing price nearby, and more strongly when the two are close.

It is not universal. One field experiment that controlled for substitute products found nothing at all.

Which makes it a narrower and more interesting design than the folk version. The interface is not exploiting a reading defect; it is supplying an inference, to people who need one, in a context where they have nothing better to go on.

That is the same logic as the neighbouring designs on this site, and worth reading against them. Scarcity and urgency cues supply a different inference and have their own regulatory record. And removing steps from the purchase itself turns out to have a smaller measured effect on spending than its reputation suggests. The composition of the choice set - adding a third, worse option to make one of the others look better - is a further variation, and it has a replication record of its own.

Limitations and disputes

• Filippin (2009) is a university working paper, not peer-reviewed, and is labelled as such.
• We have not printed specific demand figures from Anderson and Simester’s experiments. The abstract states the direction and the moderators; the exact magnitudes were not in the accessible record.
• The cash-register origin story is omitted because no primary source for it could be found.
• Anderson and Simester’s experiments were run with one retailer, in catalogues, and generalisation to other formats is an inference.
• Thomas and Morwitz’s five experiments are laboratory studies, so their boundary conditions are established under controlled comparison rather than in a shop.
• This says nothing about whether nine-endings are good or bad for shoppers, only about what has been measured.
• No claim is made about pricing law or regulation, which differs by jurisdiction.