A Brazilian company can spend twenty years building a brand. Advertising, winning customers, taking market share, turning that name into the most recognisable asset the business owns. Then one day it needs credit.
The bank looks at the property, the vehicles, the inventory, the receivables. The trademark can be pledged too: it is in the law, it has its own service at INPI and it has existed since 1997. In practice, it almost never is.
Among 1,138,069 trademark registrations granted in Brazil between 1978 and June 2013, we found 358 that were ever used as collateral for a loan. Three in every ten thousand.
In the United States, for the same generation of registrations, it was twelve in every hundred.
It is not that Brazilian trademarks do not circulate
The easy explanation would be that trademarks simply do not move in Brazil. That is not it. Against the study by Graham, Marco and Myers, which tracked what happened to every American registration issued between 1978 and 2013, the Brazilian trademark changes owner in 14.5% of cases against 21% of American ones. It is less, but it is not a chasm. Here too they are bought, sold, inherited and reorganised.
The chasm appears somewhere else: when the trademark has to stop being merely property and start working as capital.
Figure 1
The distance is not in transfers. It is in collateral.
Share of trademark registrations that received each type of recordal over their lifetime, among those granted from 1978 to 2013 in each country.
Source: TrademarkIQ, on the INPI recordal database, and Graham, Marco and Myers (2018) on the USPTO Trademark Assignment Dataset.
The mark shows up in collection, not in credit
INPI files everything that restricts a trademark under a single pair of recordal codes. We opened all 17,134 of them and read what each one actually said. What came out was not a credit market.
Judicial attachment and enforcement account for 59%. Tax listings and restrictions from the federal revenue service, 13%. Unavailability and blocking ordered by a judge, another 13%. Together, 85% of what INPI classifies as a lien involves coercive, collection or asset-restriction measures, rather than the voluntary granting of credit. Security contracted freely with a creditor is 3.8% of classified recordals. Within the comparable cohort, it reaches 358 registrations.
In the United States a trademark enters the financial system as an instrument for raising money. In Brazil it shows up mostly when someone is trying to collect money. By the time a Brazilian trademark reaches the world of debt, it is usually already too late.
Figure 2
What INPI files as a lien is mostly coercion
Distribution of recordals under codes 208 and 590, classified by the wording of the published complement. Full series, 1997 to June 2013.
Source: TrademarkIQ, classifying the 17,134 lien recordals published by INPI.
And when it is credit, the lender is rarely Brazilian
Among the consensual security recordals, reading the name of the creditor produces the most uncomfortable part of the study. Of the recordals that name who received the security, 302 point to a foreign institution and 52 to a Brazilian one.
Named creditors, full series
Thirty-five years, 1.1 million registrations
- United States Department of the Treasury140
- The Royal Bank of Scotland58
- Banco do Brasil and others39
- International Finance Corporation20
- The Bank of Nova Scotia19
- Banco BVA17
- BNDES Participações13
- Banca Intesa, UniCredit, Citicorp, HSBC USA, Crédit Lyonnais13
Itaú, Bradesco, Santander, Caixa Econômica, Safra and BTG do not appear once in the entire series.
The three Brazilian institutions on the list are two state banks and one bank that failed. Banco do Brasil appears in a single pledge deed drawn up in Bento Gonçalves, BNDES Participações in a contract with Contém 1g, and Banco BVA was placed into liquidation in 2013. No Brazilian private retail bank accepted a trademark as security on any registration in this cohort.
Property and capital are not the same thing
That is the distance these numbers measure. A Brazilian trademark is full property: the registration protects, excludes third parties, lasts ten years and renews indefinitely. What it almost never is, in practice, is mobilisable capital, something you take to a bank and turn into money.
This is not a failure of INPI, which grants more than 400,000 registrations a year and delivers what it promises. It is an absence on the other side of the counter. The legal instrument has existed since 1997, it sits in article 136 of the Brazilian Industrial Property Law, and the Brazilian credit market simply does not use it.
One necessary caveat: this does not measure the economic value of a trademark, it measures what gets recorded at INPI. A mark can generate revenue for a lifetime through licensing, franchising or royalties without ever appearing in these recordals, and a billion-real share acquisition may have the brand as its principal asset without producing a single entry, because the owner of record stays the same. What the data show is narrower and more specific: the right almost never becomes formal credit collateral.
The practical consequence is direct. For a Brazilian company, a trademark is an asset that protects the business rather than financing it. It sustains pricing, defends market share, sits on the balance sheet, and never enters the conversation with the bank. For anyone managing a portfolio, that changes the calculus of what to register, what to maintain and what to let lapse: an idle registration has nowhere to go, and the only way it becomes worth something again is by being used again.
How we got to these numbers
- Complete INPI recordal database, with data through Industrial Property Journal 2904, of 1 September 2026.
- Cohort. The 1,138,069 Brazilian registrations are selected by grant date, between 1978 and June 2013. The grant date defines who enters the study, not when the events occurred.
- Observation window. For each registration in that cohort we searched for transfers and lien recordals published through 30 June 2013. Later events enter none of the comparative indicators, even though the TrademarkIQ database runs to September 2026. Mean exposure for a cohort registration is 13.8 years and the median 12.2 years, and a quarter were granted in the final five years of the window, with little time to accumulate events. The American study has the same structure, with registrations issued between 1978 and 2013 and transactions tracked through 2013.
- Transfer is measured by recordal codes 565 and 235, transfer recorded. Name and address changes have distinct codes, 560 and 230, and were counted separately, preserving the same distinction used in the American study.
- Liens and restrictions are codes 590 and 208. The 17,134 recordals were classified by the wording of the published complement into five groups: consensual credit security, judicial attachment or enforcement, tax listing or restriction, unavailability or blocking, and cancellations and releases. Only the first group is comparable to the American security interest agreement, and it is the one that produces the 0.03% figure. Counting any lien or restriction recordal, the Brazilian share rises to 0.70%, but that figure is not comparable.
- Codes 590 and 208 only exist from June 1997, so the observation window for liens is shorter than for transfers. Repeating the calculation over the 675,925 registrations granted between June 1997 and June 2013, with grant and observation windows aligned, consensual security stands at 0.032%, effectively identical.
- Observation cannot run past June 2013 because INPI changed its recordal coding system on that date. From then on a transfer is recorded as a generic petition grant, indistinguishable from other grants without consulting the petition type, information the public database only exposes from 2020 onward. The limit is imposed by the source, and it coincides with the American study’s.
- Registrations granted near the end of the window have less exposure time, a limitation the American study shares. Neither study corrects for exposure time, so both proportions understate incidence over a full life of ten years or more.
- Creditors are read from the text of the recordal itself, where it names the party in whose favour the security was created. Of the 649 consensual security recordals, 354 name an identifiable creditor. A bank missing from the list means it does not appear named in those recordals, not that it has never accepted a trademark as security outside the INPI register.
- Graham, Stuart J. H., Alan C. Marco and Amanda F. Myers, Monetizing Marks: Insights from the USPTO Trademark Assignment Dataset, Journal of Economics & Management Strategy, vol. 27, no. 3, 2018.
- USPTO, Trademark Assignment Dataset, 2023 update, with more than 1.38 million transactions recorded between March 1952 and January 2024.
INPI does not publish values. None of the figures above are financial. The analysis of ownership transfers published in 2026, with seller reconstruction and the anatomy of who buys a trademark in Brazil, is the subject of a separate study. Material corrections will be published on this page with an update note.

