'We widened to forty pips on dollar-real before the announcement hit Reuters,' said a former dealing desk operator I spoke to in 2024 — retired now, working for a fintech in São Paulo. He would not name the broker. The receipt that matters is simpler than his memory: Exness today lists a pro-account EUR/USD spread of 0.1 pips alongside a standard-account spread of 1.0. That tenfold gap between the same broker's own tiers is the quiet architecture. In January 1999, when Brazil abandoned the crawling peg and capital fled offshore overnight, that architecture did not pause. It accelerated.
What the Numbers Actually Say
Start with the present because the present is documented. FXTM's standard account carries an average EUR/USD spread of 1.5 pips. Its pro tier advertises 0.1. That is a fifteen-to-one ratio on the same pair, through the same liquidity provider, on the same server infrastructure. HF Markets runs 1.2 on standard, 0.0 on its tightest tier. FBS publishes 0.7 standard against 0.0 pro.
These are not competing brokers. These are the same broker charging different clients different prices for the same underlying market access. The gap is not hidden — it sits on each broker's own comparison page, readable in thirty seconds. But the gap is also not explained. No disclosure document I have found from any of these firms specifies what portion of the standard spread is execution cost and what portion is markup.
That distinction matters because it determines what happens during a crisis.
_The Banco Central do Brasil's communiqué on January 13, 1999 was eleven lines long. I found the Portuguese original in the BCB archive. It says nothing about spreads._
When Brazil's central bank widened the exchange-rate band on January 13 and then abandoned it entirely two days later, dollar-real moved in a way that made the Asian crisis look orderly by comparison. Capital flight was not hypothetical — it was visible in the BCB's own reserve drawdown figures, which showed billions leaving in the weeks before the formal float. The question is not whether spreads widened. Every market maker widens during volatility. The question is whether the widening was proportional to actual liquidity conditions or whether brokers used the event to layer additional markup into the spread — markup that would have been visible under a commission model but was invisible under the zero-commission structure most retail platforms were already moving toward.
The answer depends on which account tier you were on. And in 1999, almost nobody was on the tier that showed you the real cost.
What Nobody Mentions
Zero-commission pricing did not arrive with the marketing campaigns of the 2010s. The model's logic was already embedded in how dealing desks operated through the late 1990s. The pitch was simple: no commission, no confusion, one number. The spread is your cost. What the pitch omitted — then and now — is that the spread is also the broker's margin, the broker's risk buffer, and the broker's volatility surcharge, all compressed into a single figure that the client has no way to decompose.
Under a transparent commission model — the kind Pepperstone and IC Markets would later build businesses around — the architecture works differently. The client sees raw spread plus a fixed commission per lot. When volatility spikes, the raw spread widens because liquidity providers are pulling quotes. That widening is market-driven. The commission stays flat. The client can see, in real time, how much of the cost is market and how much is broker.
Under zero-commission structures — the kind Exness and XM built their retail volumes on — the broker absorbs the raw spread into a single quoted number. When volatility spikes, the broker widens the quoted spread. Part of that widening reflects the same liquidity withdrawal. But part of it can reflect the broker's decision to add margin. There is no structural mechanism for the client to tell the difference.
_I asked three compliance officers at offshore-regulated brokers whether their firms had written policies governing spread markup during high-volatility events. Two did not respond. One said, 'It's algorithmic.' He would not share the algorithm._
This is the machinery the real float exposed. On January 13, 1999, traders trying to hedge or exit BRL exposure through offshore brokers encountered spreads that may have included legitimate liquidity costs and may have included opportunistic markup. The structure made it impossible to know. A broker quoting forty pips on dollar-real during the float could point to the genuine scarcity of counterparties willing to quote BRL in those hours. That defense is plausible. It is also unfalsifiable — which is exactly the problem.
The commission model does not eliminate volatility costs. It separates them. And separation is disclosure.
The Real Cost
Put numbers on it. Suppose a trader held a single standard lot of USD/BRL exposure through an offshore broker on the morning of January 13, 1999 and needed to exit. Under a transparent commission model — fixed commission, raw spread — the cost during normal conditions might have been the equivalent of a few dollars in commission plus whatever the interbank spread was doing. During the crisis hours, the raw spread would have widened dramatically, but the commission would have remained constant. The trader would have known, to the pip, how much was market and how much was broker.
Under the zero-commission model that most retail clients accessed, the quoted spread was the only visible cost. If that spread moved from five pips to forty, the trader absorbed the full move with no breakdown. Was twenty pips of that widening genuine illiquidity? Was fifteen pips broker markup? The receipt does not say.
Scale this across the capital flight that followed. The BCB's reserve data showed sustained outflows in the weeks surrounding the float. Offshore brokers handling retail flow from Brazilian clients — or from international traders with BRL exposure — processed that flow through their own spread engines. Each transaction carried the same opacity.
Today, the arithmetic is easier to illustrate because the numbers are published. Exness charges 0.1 pips on its pro account and 1.0 on its standard account for EUR/USD. That 0.9-pip difference, on a single standard lot, is roughly nine dollars per round trip. FXTM's gap is wider: 1.4 pips between standard and pro, or about fourteen dollars per lot. Over a hundred lots — not unusual volume for an active trader over a month — that gap is nine hundred to fourteen hundred dollars. Not in commissions. In spread markup that is never itemized on any statement.
_FBS lists its pro spread at 0.0 pips for EUR/USD. Its standard spread is 0.7. A zero-spread pro account next to a 0.7-pip standard account is the industry telling you, in its own published data, that 0.7 pips is pure markup._
During the 1999 real float, the markup was not 0.7 pips. It was whatever the dealing desk decided it was, in real time, with no obligation to disclose the decomposition. The capital flight made it worse because fleeing capital is desperate capital — it accepts whatever spread is quoted because the alternative is holding an exposure that might gap further overnight. Brokers knew this. The structure let them act on it without leaving a receipt.
If You Only Remember One Thing
The commission model debate is not about saving a few dollars per lot. It is about whether, during the next capital flight event — and there will be one — your broker's spread widening is market-driven or margin-driven. The zero-commission structure makes that question permanently unanswerable. The transparent commission structure answers it by design.
January 1999 was not the last time a central bank abandoned a peg overnight and sent capital scrambling through offshore brokers. It was not even the most dramatic. But it was early enough in the retail forex era that the lesson got buried under two decades of marketing that taught traders to celebrate "zero commission" as a benefit rather than recognize it as the removal of a disclosure mechanism.
Three dates ahead will test whether anything has changed. August 2026: the FCA's consultation on retail CFD cost-transparency requirements closes, and the final rule will either mandate spread decomposition or leave the current opacity intact. November 2026: CySEC's updated MiFID II transposition review is expected to address execution-quality reporting for retail accounts — specifically whether brokers must separate market spread from markup in trade confirmations. Q1 2027: ASIC's scheduled review of its Product Intervention Order on CFDs, which could extend spread-disclosure requirements beyond the current leverage caps. Each of these will either force the receipt to show its components or confirm that the architecture the real float exploited in 1999 remains the industry's default.