BRENNEN

Industries / Retail, hospitality and personal services / Retail and e-commerce / Pawn shops

Checking for stolen goods

This is the work of keeping stolen property out of pawnshops and second-hand stores, and of helping police find it when it gets in: identity checks, record books, holding periods, reports to police, and locks built into the goods themselves.

5

Milestones achieved, from 1989 to 2014.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1989

    Fake pawnshops may raise theft

    In a "storefront sting", police pose as pawnbrokers or fences, people who buy stolen goods, to catch thieves. Robert H. Langworthy, a criminologist at the University of Cincinnati, ran a time-series analysis of a 1985 to 1986 sting in Birmingham, Alabama, and found that it served the police department's own goals but not crime control and may have had a negative effect on the area; he wrote that police should give serious thought to dropping the tactic. A 2010 police guide by Michael Sutton for the Center for Problem-Oriented Policing cites this and later work in advising against storefront stings, which can encourage more theft nearby.

    LessonA buyer of stolen goods creates demand for theft, even when the buyer is the police.

    doi.orgapi.openalex.orgpopcenter.asu.edu

    Credit

    Robert H. Langworthy, in Justice Quarterly

  2. 2002

    Web auctions get lighter duties

    Japan's Pawnbroker Business Act of 1950 makes a pawnbroker confirm each customer's identity and tell the police at once if an item looks stolen. The shop must keep a ledger of pledges for three years and keep for six months each shinafure, a police notice describing stolen goods, and a theft victim can reclaim a stolen pledge free of charge for one year even if the shop took it in good faith. An amendment to the Secondhand Articles Dealer Act (Act No. 115 of 2002) brought online auction services for used goods under the second-hand dealers' law with lighter duties: they must notify the prefectural Public Safety Commission and report suspected stolen items to the police at once, but need only "endeavour" to confirm who their sellers are and to keep records. A service may apply for certification that its methods meet a national standard for preventing and finding stolen goods.

    laws.e-gov.go.jplaws.e-gov.go.jp

    Credit

    National Diet of Japan; National Public Safety Commission

  3. 2004

    Stolen goods blend into resale

    Simon M. Fass and Janice Francis, social scientists at the University of Texas at Dallas, analysed pawn transaction data from Texas in the Journal of Research in Crime and Delinquency in May 2004. They concluded that plans to cut theft by shrinking the market for stolen goods were unlikely to work, partly because similar steps already applied to pawnshops had shown limited effect, and mainly because "hot goods are invisible in the daily flow of secondhand merchandise through the general retail market". They advised police to keep their focus on catching thieves, for example through better tracking of pawn transactions.

    LessonPawn records help police catch thieves, while rules aimed at pawnshops had shown only limited effect on theft.

    doi.orgapi.crossref.org

    Credit

    Simon M. Fass and Janice Francis

  4. 2010

    One database, many police forces

    Maryland's Regional Automated Property Information Database (RAPID), run by the Maryland Department of State Police on software from Business Watch International (BWI), collects the transactions that pawnshops, precious-metal buyers, scrap yards and auto dismantlers must report electronically under state law. In 2010 to 2019 it helped recover $53,598,423 of property, leading to the arrests of 13,476 suspects and the closing of 18,707 criminal investigations, by the State Police's count of figures that agencies report themselves; its users grew from 1,559 in 2010 to 6,032 in 2019, in Maryland, five other states and Washington, DC. The same report says that as electronic reporting succeeded, more suspects turned to online auctions and other web sales. California's Assembly Bill 391 of 2012 likewise ordered a single statewide system, the California Pawn and Secondhand Dealer System (CAPSS), run by the state's Department of Justice.

    LessonWhen shops must report every item, some thieves move to channels that do not.

    mdsp.prod.maryland.govoag.ca.gov

    Credit

    Maryland Department of State Police; Business Watch International; California Department of Justice

  5. 2014

    Phone locks cut robberies

    Apple added Activation Lock to iPhones running iOS 7 on 18 September 2013, so that a wiped phone needs the owner's Apple ID and password before it works again. A report of 19 June 2014 by the Secure Our Smartphones initiative, co-chaired by New York Attorney General Eric Schneiderman, San Francisco District Attorney George Gascón and London Mayor Boris Johnson, found that in the first five months of 2014 robberies involving Apple products in New York City fell 19% and grand larcenies 29% from a year earlier, while those involving Samsung phones rose over 40%; in the six months after the lock arrived, iPhone robberies fell 38% in San Francisco and Apple thefts 24% in London. California's SB 962, by State Senator Mark Leno, signed on 25 August 2014, requires such a lock on smartphones made from 1 July 2015 and sold in the state, but does not apply to phones resold second-hand or held as collateral on a loan.

    LessonA phone that is useless to anyone but its owner is worth little to a thief or a fence.

    ag.ny.govleginfo.ca.gov

    Credit

    Apple; Office of the New York State Attorney General; California Legislature

Watch for

  1. Florida weighs one pawn database

    Florida had 1,138 licensed pawnshops on 5 October 2025, and its pawnbrokers and second-hand dealers report transactions to local police agencies. Second-hand dealers must hold most goods for 15 days and jewelry, gemstones, precious metals, antiques and art for 30 days, and police can order a 90-day hold on goods they believe stolen. A 2025 law had the Florida Department of Law Enforcement (FDLE) study a statewide pawn database, and FDLE's study of 4 November 2025 recommended collecting all pawn records through a single vendor into one database whose design FDLE controls, free and in real time for Florida police. Senate Bill 1582, which would have required dealers and pawnbrokers to share their data with FDLE as well, died in committee on 13 March 2026.

    Achieved whenFlorida passes a statewide sharing law and FDLE's database receives pawn transactions from across the state.

    flsenate.govflsenate.gov

    Claimed by

    Florida Department of Law Enforcement, November 2025

Open

  1. 01

    Database hit rates over time

    One dataset, built from the published annual reports of state pawn databases, of transactions reported, searches, cases, arrests and value recovered per year, starting with Maryland's RAPID, with a computed rate of cases per 100,000 reported transactions and a plain note on the limits of self-reported figures.

    The brief
    Credit · unclaimed
  2. 02

    Photo matches, measured

    A public benchmark of how well image search finds a stolen item from its owner's old photos. Photograph 500 owned or borrowed items (bikes, power tools, guitars, watches, rings) as owners do at home, photograph each again under pawn-counter conditions, hide them among 20,000 look-alike photos of other items, and score how often open image models put the right item first or in the top ten.

    The brief
    Credit · unclaimed

Open 01 · the brief

Database hit rates over time

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

One dataset, built from the published annual reports of state pawn databases, of transactions reported, searches, cases, arrests and value recovered per year, starting with Maryland's RAPID, with a computed rate of cases per 100,000 reported transactions and a plain note on the limits of self-reported figures.

Why it is still open

Fass and Francis wrote in 2004 that the research behind stolen-goods policy was thin. Maryland's 2019 RAPID report gives yearly counts of cases (1,942 in 2017, 1,438 in 2018, 1,271 in 2019), arrests and recoveries, and transaction counts by industry, but computes no rate and says its figures are self-reported by agencies. A search found program reports and news stories from Minneapolis, Tucson and Maryland, and no multi-year rate or study with a comparison group.

mdsp.prod.maryland.govapi.crossref.org

Built with what exists

Annual report PDFs from the Maryland State Police and other states; pdftotext, or a document-reading model for the tables (see the Note "The jobs agents repeat"); Python and pandas; public-records requests for missing years that ask only for aggregate counts.

  1. First step

    Find and download RAPID's annual reports, of which the 2019 report is Volume 10, and transcribe the cases, arrests, recovery and transaction tables from three of them.

  2. Done looks like

    At least ten years for Maryland and five for a second state, every figure traced to a page of a public report, published as CSV with a chart of the hit rate over time. No personal data is collected.

Open 02 · the brief

Photo matches, measured

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public benchmark of how well image search finds a stolen item from its owner's old photos. Photograph 500 owned or borrowed items (bikes, power tools, guitars, watches, rings) as owners do at home, photograph each again under pawn-counter conditions, hide them among 20,000 look-alike photos of other items, and score how often open image models put the right item first or in the top ten.

Why it is still open

LeadsOnline, the US service to which many pawnshops report, says police search its transaction images, and its own example matched pawned jewelry against victims' descriptions; 529 Detective, a bike registry's paid service, scans six marketplaces twice a day for listings "similar to" a stolen bike. Neither publishes how often a match is right, and a search found no published accuracy figure for matching pawned goods to theft reports.

leadsonline.comproject529.com

Built with what exists

A phone camera; Meta's open DINOv2 image model and OpenAI's open CLIP model; FAISS for similarity search; Label Studio for checking pairs; Python. Only items the team owns or borrows with consent are photographed.

  1. First step

    Photograph 50 bikes and tools twice, once at home and once on a counter under shop lighting, and run DINOv2 on them to see how often the pair comes first.

  2. Done looks like

    500 items with two photo sets each and 20,000 distractors, top-1 and top-10 accuracy published by category and model, with the photos released under an open licence where the owners agree.