How accurate is Numbeo? One price to the cent, one for a pass that no longer exists

By Skyler Bissell · September 1, 2026 · 9 min read

On 1 September 2026, Numbeo listed the monthly transit pass in Paris at EUR 90.80. The official price, set by Île-de-France Mobilités last November and in force since 1 January, is EUR 90.80. To the cent. The same table gives New York a monthly pass at $140.00, and New York retired its monthly pass in January. Both rows sat in the same database on the same day, and between them they hold most of the answer to the question this page is titled after.

What Numbeo says about its own data

Numbeo has been collecting prices since April 2009, when Mladen Adamović, a former Google engineer, launched it, and the model has stayed the same since: anyone can submit what things cost where they live. The site's methodology page describes what happens next. Submissions pass through what Numbeo calls "more than 30 sophisticated filters", entries classified as spam are discarded, and staff enter manually collected prices from supermarket sites, taxi tariffs and government institutions twice a year, each weighted three times as heavily as a user entry.

The most useful thing on that page is the data-retention rule, because it quietly concedes the point critics usually reach for. An item with more than 153 contributors in the past three months keeps only three months of data. More than 123 in six months, six months. Everywhere else, Numbeo uses entries up to 12 months old, stretching to 24 where contributions are scarce. In plain terms: London's prices are weeks old and a small city's can be two years old, by design rather than by accident, and Numbeo says so itself.

You can watch the gradient without leaving the site, because every city page prints its own sample size in the footer. On 1 September 2026, London reported 2,145 entries in the past 12 months from 411 contributors. Paris had 1,642 entries from 256 people, New York 1,178 from 198, and Ljubljana 845 entries from just 85 contributors, last updated 20 August. That one line is the fastest accuracy check Numbeo offers, and almost nobody reads it. What the prices are then rolled into, the indexes with their New York base of 100 and the rent exclusion, is a separate subject: see what a cost of living index measures.

What independent checks have found

Less than you would hope, in either direction. We went looking for a peer-reviewed study that tests Numbeo's price data against official statistics and could not find one to read: plenty of published research uses Numbeo as an input, which is a kind of endorsement, but nothing we found puts the price data itself on trial. We searched on 1 September 2026; if such a study exists, it is well hidden.

The most careful independent test we did find is from 2017. Ray Woodcock, assembling a list of low-cost cities, normalised city values from Numbeo, Expatistan, C2ER's index and UBS, benchmarked the US cities against MIT's Living Wage data, and took the median across sources because single sources "could offer values that seemed wildly mistaken", naming Numbeo as the most frequent offender. His examples: Numbeo rating Minneapolis above Fort Lauderdale and San Jose, and Wichita far above Reno. At country level he found the same pattern, with Numbeo departing from the other three sources' consensus most often. That is one analyst, nine years ago, on a younger and thinner dataset, and it deserves exactly that much weight: a documented city-level pattern of stray values the crowd had not yet corrected.

The loudest criticism on record concerns a different product. Numbeo's crime index was moved by a single determined contributor in Lund in 2017 ("Anyone, anywhere can change the data," he told The Local), and AFP's fact-checkers in 2022 quoted researchers calling its crime comparisons non-representative sampling. Those cases test the contribution model, and they test it on perception scores, where no receipt exists to check a claim against. A litre of petrol has a posted price; a feeling of safety has none. So the price data deserves its own trial, which is why we ran one.

Three prices against official sources, dated 1 September 2026

The rule for a fair check is to ask what the official number denominates before comparing anything. A national average is a different object from a city price, an asking rent from a sitting contract, a zoned travelcard from a flat pass. We picked items where an official price exists on the same basis Numbeo's contributors answer, and we dropped one item for failing that test.

Paris, monthly transit pass. Île-de-France Mobilités set the all-zones Navigo Mois at EUR 90.80 from 1 January 2026, up from EUR 88.80. One administered price for the whole region, so the basis is clean. Numbeo's value on 1 September: EUR 90.80, quoted range 90.00 to 95.00. Exact to the cent. This is the crowd at its best, because everyone pays the same posted number and knows it by heart.

London, petrol. The Department for Energy Security and Net Zero publishes a weekly average pump price; for the week of 24 August 2026 it stood at 161.63 pence per litre of unleaded. Numbeo's London figure in the same week: GBP 1.55, range 1.35 to 1.65. The official value sits inside Numbeo's range, with the mean running about 7 pence low. One caveat belongs in the record: the official series is a national average and Numbeo's is a London value, so a few pence of the gap may be geography rather than error.

London, transit pass: dropped. Numbeo lists GBP 200.00 with a range of 150 to 250, and that range spans several different products at once, because a London travelcard's price depends on zones. No official number shares that basis, so any verdict would be unfair in one direction or the other. This is the units question every index comparison has to survive, and here it could not.

New York, monthly pass. Numbeo listed $140.00, range 132.00 to 156.52. New York has had no monthly pass since 4 January 2026, when the MTA completed its move to OMNY, raised the base fare to $3.00 and retired the 30-day unlimited pass; local transit now caps at $35 in any seven days instead. Eight months on, the row is still live, averaging what contributors remember with what they compute for themselves under the weekly cap. Nobody lied, and every entry may have been sincere. The question the row answers quietly left the market.

Numbeo values and footer counts retrieved 1 September 2026; each official source is linked at the foot of the page.

When to trust a Numbeo number

Read the three rows together and a usable rule falls out.

Trust a Numbeo number more when:

• The city's footer shows entries in the thousands (London: 2,145 in the past 12 months).
• The item has one posted price everyone pays: fuel, a flat transit pass, a supermarket staple.
• The last update is days old, which high-contributor cities manage continuously.

Trust it less when:

• Contributors are few and the archive window is correspondingly long: a thin city can carry prices up to two years old.
• The item bundles several products under one name, the way a zoned travelcard does.
• Policy has moved the price recently. A crowd average adjusts at the speed of its slowest contributors, and New York's row is the proof.

Rent deserves its own sentence, since it is the Numbeo number people quote most. We deliberately left rent out of the spot-check: no official series shares Numbeo's basis, which blends new leases, long-held contracts and neighbourhood hearsay across districts. That is a comparability problem before it is an accuracy problem, and it means a Numbeo rent line is a starting bid, worth checking against local listings before it decides anything. The same modesty applies to any line where the footer count is thin: the number is a report from a small crowd, and the honest move is to widen the crowd before betting on it.

The question accuracy cannot settle

Now grant Numbeo a perfect day: every Paris price as exact as the Navigo line. The relocation question is still open, because the items that decide it never enter a price table. Income tax and social contributions take 36.0% of a $150,000 salary converted and earned in Paris, against 31.1% in New York. French childcare is subsidised, healthcare is folded into payroll instead of arriving as a premium bill, and the law grants 25 paid vacation days. None of that shows up on a till receipt, so no crowd can price it.

Our engine holds a household's net position constant across those systems and solves for the salary that keeps it. For a single renter on $150,000 in New York, the Paris answer is €101,864, which is $118,035 at the current rate. No price index, however accurate, can produce that figure, because most of what moves it was never a price. That boundary is the subject of what every cost-of-living calculator gets wrong. You can see the full ledger for this pair on the New York vs Paris comparison, or start from take-home pay by country to watch the tax layer alone reorder cities.

FAQ

Is Numbeo's rent data accurate?

Rent is Numbeo's hardest category to verify, which is why our spot-check skipped it. Contributors report a mix of new leases, long-held contracts and neighbourhood hearsay, while official rent series measure regulated stock or whole-market averages on a different basis. Treat a Numbeo rent line as a starting bid: useful for the order of magnitude, worth checking against local listing portals before it decides anything.

Is Numbeo more accurate than official statistics?

They measure differently. Official price statistics come from representative baskets collected by trained surveyors, published slowly and usually at country or region level. Numbeo is fast, city-level and worldwide, but unverified. For a single administered fare in a big city the two can match exactly; for the full cost-of-living picture, official sources are the anchor and Numbeo is the fresh estimate between anchor points.

Does Numbeo measure quality of life or just prices?

Both, in separate index families. Cost of living and rent are price data; the quality of life, crime, healthcare and pollution indexes are perception surveys scored by visitors. This page tested only the price data. The perception indexes have drawn the sharpest published criticism, and this page's verdicts do not transfer to them.

Is Numbeo accurate enough to plan a move on?

For shortlisting cities and getting a feel for daily prices, yes, with each city's footer counts as your confidence meter. For the decision itself, no price site computes taxes, childcare subsidies, healthcare structure or the value of statutory leave, and that is usually where a move is won or lost. cityparity vs Numbeo walks through where the two tools give different answers to the same question.

So: how accurate is Numbeo? Accurate enough that its Paris transit line matched the operator exactly, uneven enough that its New York line still sells a ghost, and candid enough to publish its entry counts at the foot of every page. Read that footer line the way you would read a label. Then, once the argument stops being about prices and starts being about a move, run the two cities you are weighing. The crowd can tell you what a litre costs. It cannot tell you what to earn.

Sources. Contribution model, filters and archiving windows: Numbeo, Motivation and Methodology; index definitions: Numbeo, About the Cost of Living Indexes; city values and footer counts from Numbeo's Paris, London, New York and Ljubljana pages, retrieved 1 September 2026. Navigo Mois price: Île-de-France Mobilités press release, 27 November 2025. UK petrol: Department for Energy Security and Net Zero, weekly road fuel prices, week of 24 August 2026. New York fare change: Gothamist, 2 January 2026. Independent comparison: Ray Woodcock, 229 Very Low-Cost Cities Around the World, June 2017. Equivalent salaries and tax rates are computed by cityparity's engine; per-field provenance is in data/_meta.json, and our own price lines split between official and crowdsourced sources as described in the methodology.

Equivalent salaries solve for equal net cash after tax, housing, childcare, healthcare and the cash value of statutory benefits, and they are the salary an offer has to clear rather than a salary any employer is obliged to pay. See the methodology.