Fraud Losses in Japan Rose 98% in a Year While Cases Cleared Stood Still — Most Reports From People in Their 30s, Most Money From Those in Their 70s

In confirmed figures published on 22 May 2026, the National Police Agency put losses from remote-contact fraud in 2025 at 142.31 billion yen[1]. That is 70.43 billion more than the year before, a rise of 98.0%, close to double in a single year. Reported cases came to 27,832[1], up 32.3%.

Another figure sits alongside it in the same document. Cases cleared numbered 6,575[1], down 0.0% year on year. Persons arrested came to 2,338, up 2.8%. Losses doubled; the number of cases cleared did not move.

The rise was concentrated in impersonation of a relative. Reported cases came to 14,489, up 114.6%[2], with losses of 113.81 billion yen, up 148.3%. Meanwhile the previously dominant methods fell: deposit-account fraud down 25.0%, refund fraud down 21.9%, cash-card theft by deception down 10.3%. The whole did not swell; one method jumped.

The statistics contain a second figure that unsettles an assumption: the share of older victims.

Across all remote-contact fraud, reported cases with victims aged 65 and over came to 14,273[2], up 3.9%. In absolute terms they rose. But as a share of reported cases excluding corporate victims they came to 51.3%[2], down 14.0 points. Older victims did not decline; everything else grew sharply.

Where it grew shows in the breakdown of impersonation fraud. The largest group by case count is those aged 80 and over, at 3,235, followed by people in their 70s at 2,453. So far, as before. Next, however, come people in their 30s at 2,284 cases[2], a rise of 329.3%, and those in their 20s at 1,759, up 361.7%. Together the two groups make up 27.9% of reported impersonation cases, a share 14.4 points higher than a year earlier.

Carrying much of that is fake-police fraud, in which the caller claims to be an officer. Its largest group by case count is people in their 30s, at 2,239[2], followed by those in their 20s at 1,718. It is concentrated among the young.

Sort the same method by money lost, though, and the order inverts. The largest is people in their 70s at 27.46 billion yen[2], then those in their 60s at 25.63 billion. The ages at the top by count and the ages at the top by amount are entirely different.

Two things follow.

The first is that the growth in harm has come loose from the capacity to catch it. While losses rose 98.0%, cases cleared moved 0.0%. In the same document, clearances for social-media investment and romance fraud rose 140.5% to 630, with persons arrested up 207.8% to 397. Investigative capacity is not flat across the board; it is flat in this one area. Where operations rotate disposable participants through short cycles, each cleared case reaches less far. The more the count is pursued, the further behind the growth in losses it falls.

The second is the question of which figure is taken as the basis. Measuring the same phenomenon by cases or by money sends the response to different places. By cases, warnings aimed at younger people take priority. By money, stopping the movement of older people's assets does. Fake-police fraud carries both at once: most cases in the 30s, most money in the 70s. Allocate resources on one of the two alone and the other falls through.

The awkward part is that the premise "this is a crime against the elderly" is built into how the countermeasures are made. Staff intervening at bank counters, neighbourhood watch, warnings passed to families. Each assumes a route that reaches an older person in person or by telephone. A method whose largest group is now in their 30s does not travel that route. The place where information asymmetry arises has moved from the counter to the phone screen.

As a matter of design, the two layers are better separated. The layer of counts is about closing the route of contact: giving people of every age a quick way to check whether a message claiming to be from the police is genuine. The layer of amounts is about stopping money just before it moves. Confirmation for large transfers and withdrawals is currently designed around older customers; if the threshold is an amount, age does not come into it.

Which indicator you cut the statistics with changes which problem you see. The same issue is taken up in designing outcome indicators（このサイトの記事）. How to turn public data into your own judgement is set out in reading official statistics in practice（このサイトの記事）. That losses doubled in a year when clearances did not move can be read now, without waiting for next year's figures.

Further Reading

• データリテラシー入門 : 日本の課題を読み解くスキル (Introduction to Data Literacy: Skills for Reading Japan's Problems)（外部サイト、新しいタブで開きます） (Akinori Tomohara, Iwanami Shoten). An introduction, built on Japanese social statistics, to how the same data yields different conclusions depending on how it is cut. It offers a way to spot the count-or-amount problem discussed above in other fields.

References

Reported and Cleared Cases of Remote-Contact Fraud and Social-Media Investment and Romance Fraud, 2025 (Confirmed Figures) — Organised Crime Department and Community Safety Planning Division, National Police Agency (2026). National Police Agency public information material

Reported and Cleared Cases of Remote-Contact Fraud and Social-Media Investment and Romance Fraud, 2024 — Organised Crime Department and Community Safety Planning Division, National Police Agency (2025). National Police Agency public information material

Reported and Cleared Cases of Remote-Contact Fraud — National Police Agency (2026). National Police Agency website, statistics

Remote-Contact Fraud: Current Situation — National Police Agency (2026). National Police Agency SOS47 anti-fraud pages