Independent restaurant mortality in Latin America: the 90% first-year myth against the registries

Where did the 90% first-year failure rate come from?

Not from any official series: that 90% appears in no business demography register of any statistical institute in the region, and whoever repeats it never cites the table.

It is trade folklore circulating since the eighties, and it survives because it is convenient, both for the consultant selling fear and for the owner who needs an excuse. The data that does exist points elsewhere. Acodrés documented more than 2,000 restaurant closures in Colombia in a single year, a hard and painful figure, yet against a sector universe of tens of thousands of formal establishments it sits an enormous distance from nine out of ten. A development bank with an MSME portfolio in Central America asked us in June 2026 for a default probability on restaurants with twelve months of operation; had we handed over that 90%, the institution would have shut the credit window for the entire sector. Between year three and year five lies the real breaking point, not month eight.

The Latin American restaurant does not die of a heart attack: it dies of anemia

Business demography series from national institutes, the ILO Labour Overview and the MSME productivity studies from ECLAC and CAF describe a curve that slopes gently through the first twelve months and then collapses once the initial working capital is spent, rent has been reset twice and margin was never measured. That calendar difference is no academic nuance: it reorders public policy completely. ECLAC estimates MSMEs contribute roughly 25% of GDP in Latin America and the Caribbean, against 56% in the European Union, and much of that gap comes from firms that survive the launch and never reach scale. Surviving and consolidating are not the same thing. A deregistration may be a bankruptcy, but it may equally be the sale of a going concern, a change of corporate name, a relocation or a slide into informality. Multilateral lenders welcome this methodological caveat; the trade association tends to skip it, because the inflated figure helps when asking for relief.

Closure is not bankruptcy, and confusing them raises the cost of credit for everyone

The whole sector pays the bill: when a risk committee reads 90% mortality, the premium charged to the independent restaurant climbs and the collateral demanded jumps. The IFC's SME Finance Forum puts the MSME financing gap at close to USD 5.7 trillion in emerging markets, with women-led firms accounting for 34% of it, around USD 1.9 trillion. Every point of perceived mortality that is not in the data pushes that gap upward, and nobody pays for having invented it. The correct instrument is sustained technical assistance plus a revolving working-capital line, exactly the opposite of what most entrepreneurship programs in the region finance. Follow the counterfactual all the way: a fund hands a USD 6,000 launch subsidy to one hundred new restaurants; at twelve months the snapshot looks spectacular because nearly all are still open and the program reports success; at month forty-eight, having never touched those kitchens' cost accounting, the portfolio shows accumulated mortality that ate the grant and the follow-on credit.

If the break comes in year four, subsidizing the opening throws money away

The money arrived when the business did not need it and vanished when it did. Diego F. Parra insists at Masterestaurant on the same order: first the cost information system, then the capital, never the reverse. A healthy independent restaurant can run a food cost of 28% to 32% per dish and still run out of cash on a Tuesday in February, because the problem is not the margin but the weekly variance of the till combined with absent cost accounting. For a loan officer that distinction defines the product: were the risk one of structural profitability, the answer would be a partial portfolio guarantee; since it is volatility plus accounting blindness, what actually moves the default needle is assistance with information systems, at a fraction of the cost. I have argued this in committees where the risk manager was right to distrust and wrong on the diagnosis. An outlet measuring its prime cost weekly stops being an opaque risk and becomes a measurable one, which is a different class of client altogether.

Counting companies measures little; counting formal jobs measures what matters

The right impact metric for a sector program is sustained formal employment, not the number of live corporate registrations. ECLAC reported that in 2024 Brazil accounted for more than 60% of net regional job creation, and accommodation and food services weigh heavily in that arithmetic because they absorb labor-market entry by young people and women. The ILO documents that the NEET rate among young women doubles that of men, 28.1% against 13.1% in 2023, and the World Bank puts regional female labor participation at 52.1% versus 74.3% for men. A restaurant reaching year five with twelve people on payroll is worth more on that dashboard than three openings that close in year three. Change the indicator and the program design changes with it. Translate the curve to your own size before making any debt decision.

How to read these numbers in YOUR operation: three scenarios?

Small outlet, one to nine employees: your risk is not year one, it is the month rent goes up and you discover you never calculated your break-even;

measure food cost per dish weekly and hold a cash cushion of six to eight weeks of fixed costs. Mid-size operation, ten to forty-nine employees across two or three locations: here the danger is opening number three financed with the good location's cash flow, and the deciding metric is contribution margin per location, not total sales. Group with four or more locations: you already compete for capital, so the conversation with the bank changes if you arrive with standardized costs and documented variance. The regional gender gap in financial accounts, 66% of women against 74% of men according to Global Findex 2025, punishes the first scenario hardest.

Methodology: what holds these benchmarks up and where they break

These figures come from verifiable public sources, not from a proprietary sample: business demography from national statistical institutes, the ILO Labour Overview, MSME productivity studies from ECLAC and CAF, the World Bank's Global Findex 2025, the IFC's SME Finance Forum and the Acodrés trade report on those 2,000-plus Colombian closures. There are three limits, and they deserve saying out loud. First, most registers bundle accommodation with food services, so a clean restaurant-only figure almost never exists in isolation. Second, regional informality leaves a large share of the real universe outside the register altogether. Third, the series do not separate bankruptcy from transfer of ownership. Anyone offering you a restaurant mortality rate for Latin America with two decimal places is estimating, not measuring, and you should ask them for the table. Timing of the intervention. Were mortality concentrated in year one, the right policy would subsidize opening.

Three differences that reshape public policy

Since the real break happens between year three and five, the effective instrument is sustained technical assistance plus a revolving working-capital line, precisely the opposite of what most entrepreneurship programs in the region fund. Nature of the risk. The independent restaurant is not a low-profitability business: it is a high cash-volatility business with absent cost accounting. For a multilateral lender that distinction decides whether the right product is a partial portfolio guarantee or assistance in information systems, which costs a fraction and moves the default needle further. Impact metric. Counting live firms measures little. Counting formal employment sustained at twenty-four months, with Open Badges micro-credentials attached to each position, connects directly to SDG 8 and allows auditing without expensive surveys. Diego F. Parra and the Masterestaurant technical team built the operational indicator framework SATE Institute uses for that measurement, precisely because a restaurant's cash data is the one thing that does not lie.