How Bank FX Research Works: Sell-Side Strategists, Trade Ideas, and Why They Are Not Signals

Bank FX research is the analysis that strategists and economists at dealing banks write for institutional clients: macro views, exchange rate forecasts, positioning and flow commentary, and occasionally specific trade recommendations. It is not a signal service. It is written for professionals who decide for themselves whether, when and how much to trade; it is produced under conflict-of-interest rules because the bank is also a counterparty; and its forecasts are opinions about an asset that academic research has repeatedly found very hard to predict.

In short: sell-side FX research exists because it helps banks win and keep institutional trading relationships. In Europe, MiFID II forced asset managers to pay for it explicitly from January 2018, and the UK and EU have since partly relaxed those rules. In the US, FINRA’s research rules cover securities, not spot currencies, while the CFTC’s Regulation 23.605 governs research on derivatives by swap dealers. The useful way to read a bank forecast is against the forward rate, not against spot, and the useful way to judge a strategist’s calls is with the same statistics you would apply to any track record. When a signal seller forwards a screenshot of a bank’s trade idea, what reaches you has lost the context, sizing and risk framework that made it professional work.

This piece complements technical vs fundamental signals, which covers what retail “fundamental” signals claim, and how signal providers make money, which covers the incentives on the retail side.

The primary sources

MiFID II: Commission Delegated Directive (EU) 2017/593, Article 12 (minor non-monetary benefits), Article 13 (inducements in relation to research) and recital 28 (the definition of research), plus ESMA’s Q&A on investor protection topics (ESMA35-43-349), questions on macroeconomic analysis and FICC research.
The partial reversal: the FCA’s Policy Statement PS24/9, “Payment optionality for investment research” (July 2024), and the EU Listing Act, Directive (EU) 2024/2811.
US rules: FINRA Rule 2241 (equity research), FINRA Rule 2242 (debt research), and CFTC Regulation 17 CFR 23.605 (conflicts of interest for swap dealers, including research).
The FX Global Code (December 2024 version), Principles 19 to 22 on confidential information and market colour.
The forecasting evidence: Richard Meese and Kenneth Rogoff, “Empirical exchange rate models of the seventies: Do they fit out of sample?” (Journal of International Economics, 1983), and Barbara Rossi, “Exchange Rate Predictability” (Journal of Economic Literature, 2013).

Who writes it and what they produce

A large dealing bank’s FX research usually sits inside a global markets research department alongside rates, credit and commodities. The people involved fall into a few groups:

G10 FX strategists, who cover the major currencies and publish the house forecasts.
Emerging market strategists, often combining FX with local rates and sovereign credit.
Economists, who forecast growth, inflation and central bank decisions that feed the FX view.
Quantitative and flow strategists, who model positioning, valuation and seasonal patterns, and summarise the bank’s aggregated client flows.

Their output follows a recognisable rhythm. There is a morning note summarising overnight moves and the day’s events; event previews and reviews around central bank meetings and data releases; a forecast table, typically quarterly points over one to two years, revised as views change; longer thematic pieces on valuation, balance of payments or policy regimes; and, at some banks, a portfolio of trade recommendations with entry levels, targets, stop levels and a running performance record.

Separately, sales and trading desks produce desk commentary: short, fast notes from traders and salespeople about what they see in the market. Regulators treat these differently from research, as the CFTC definition below shows, and the FX Global Code restricts what they can say.

Market colour and its limits

Flow commentary is the part of bank output most often mistaken for inside information. The FX Global Code, the voluntary standard written by central banks and market participants through the Global Foreign Exchange Committee, permits sharing market colour but limits it sharply. Principle 22 says communications “should be restricted to information that is effectively aggregated and anonymised,” should not identify clients by name or code, should disclose flows “only by price range and not by exact rates relating to a single Client or flow,” and should be “restricted to sharing market views and levels of conviction.” Principle 19 requires firms to identify and protect confidential information, including details of their order books and clients’ orders for benchmark fixes. A bank note saying “real money buying on dips” is, by design, too aggregated to trade on mechanically.

How bank research is paid for

For decades research was “free” to institutional clients. It was paid for indirectly through trading: an asset manager that read a bank’s research and talked to its strategists sent that bank execution business, and the cost was buried in commissions or, in FX, in spreads.

MiFID II unbundling

MiFID II, which applied in the EU and UK from January 3, 2018, treated that arrangement as a potential inducement: something of value received by a portfolio manager that could bias its choice of broker. Article 13 of Delegated Directive 2017/593 says research received by an investment firm providing portfolio management is not an inducement only if it is paid for either by “direct payments by the investment firm out of its own resources” or from a separate research payment account funded by a specific research charge to clients, governed by a budget, and assessed regularly for quality. The research charge may “not be linked to the volume and/or value of transactions executed on behalf of the clients.” An executing firm must “identify separate charges for these services that only reflect the cost of executing the transaction.”

Recital 28 defines research broadly: material that “explicitly or implicitly recommends or suggests an investment strategy and provides a substantiated opinion as to the present or future value or price of such instruments or assets.” That plainly captures most FX strategy.

Two clarifications from ESMA mattered for currencies:

No FICC exemption. ESMA’s Q&A (question 9 in the research section) applies the same test to fixed income, currency and commodity (FICC) material as to equity research: depending on its nature it is either research, to be paid for under Article 13, or a minor non-monetary benefit, for example short-term market commentary with no substantive analysis, or material made openly available to all.
Macro research. Macroeconomic analysis is, as a starting point, likely to suggest an investment strategy and therefore be research, though material that is generic and openly available, for example published on a website, can be a minor non-monetary benefit.

Article 12(3)(a) of the same directive allows “information or documentation relating to a financial instrument or an investment service” that is “generic in nature” as a minor non-monetary benefit. The practical result was that many banks priced research subscriptions, put some macro and FX commentary on public websites, and kept fast desk commentary distinct from paid research.

The partial reversal

Policymakers later became concerned that unbundling had reduced research coverage, particularly of smaller companies. Two changes followed:

United Kingdom. The FCA’s PS24/9, published July 26, 2024 and applying from August 1, 2024, reintroduced the option of paying for research and execution together through joint payments, subject to guardrails: a written policy, research budgets, client disclosure, fair allocation of research costs across clients, and arrangements that identify the research cost separately within the joint charge. The existing two options remain.
European Union. The Listing Act, Directive (EU) 2024/2811, published in November 2024, amended MiFID II Article 24(9a) to let firms pay jointly or separately for research on issuers of any size, removing the earlier EUR 1 billion market-capitalisation threshold, with requirements to agree a remuneration methodology with providers and to assess research quality annually. ESMA delivered its technical advice on the implementing rules in April 2025.

For FX readers the significance is that research has a price. Institutions treat it as a service they buy and evaluate, not as a free tip.

The US rules: securities, derivatives, and spot FX

The US framework is organised by product, which leaves spot FX in an unusual position.

FINRA Rule 2241 governs equity research by broker-dealers: separation of research from investment banking, analyst compensation, disclosures, and quiet periods.
FINRA Rule 2242, which took effect in 2016, applies a similar framework to debt research, with a tiered approach that gives broad exemptions for research distributed only to institutional investors. Its definition of “debt security” excludes, among other things, US Treasury securities.
Spot currencies are not securities, so a bank’s spot FX commentary is outside both FINRA rules.
CFTC Regulation 23.605 fills much of the gap for swap dealers. It defines a research report as a written communication “that includes an analysis of the price or market for any derivative, and that provides information reasonably sufficient upon which to base a decision to enter into a derivatives transaction.” FX options and other FX derivatives fall within it.

Regulation 23.605’s substantive rules mirror the securities model. Non-research personnel may not direct a research analyst’s decision to publish or the views expressed. Analysts may not be supervised by, or have their pay influenced by, the business trading unit. Trading-desk staff may review a report before publication only for factual accuracy, formatting or conflicts, and only through or in the presence of legal or compliance staff. A swap dealer may not “directly or indirectly offer favorable research, or threaten to change research” to win business. Reports must disclose whether the analyst holds a financial interest in the derivatives covered and any other material conflicts known at publication.

The definition’s exclusions explain much of what clients see. It does not cover “commentaries on economic, political, or market conditions,” communications to fewer than 15 persons, or communications “generated by an employee of the business trading unit that is conveyed as a solicitation for entering into a derivatives transaction, and is conspicuously identified as such.” That is one reason desk commentary typically carries labels saying it is not research and may be written by people who trade the same products.

Why a bank trade idea is not a signal

A retail signal is an instruction: pair, direction, entry, stop, target, and an implied promise that following it is the product (see what are forex signals). A bank trade recommendation can look identical on a screenshot. It differs in almost every way that matters.

Audience. It is written for portfolio managers, treasurers and reserve managers who already hold currency exposure and run risk budgets. Many use it to hedge, to time an existing decision, or to stress-test their own view rather than to open a speculative position.
Sizing is left to the reader. Bank research rarely says how much to trade, because the bank does not know the client’s portfolio. A signal implicitly sizes for you; a research note assumes you have a risk framework.
Horizon. House forecasts are typically quarterly points one to two years out. A trade idea may run for weeks or months. Retail signals often run for hours.
The bank is a counterparty. The same institution makes prices to clients and may hold positions. The rules above manage that conflict through separation and disclosure; they do not remove it. Reading the disclosures is part of reading the research.
The forecast must beat the forward, not spot. This is the most important technical difference, and the one most often missed.

Worked example: reading a forecast against the forward

The numbers below are hypothetical and were computed in Python.

Suppose EUR/USD spot is 1.1500, the one-year US dollar interest rate is 3.75% and the one-year euro rate is 2.00%. Covered interest parity sets the one-year forward:

Forward = 1.1500 x (1 + 0.0375) / (1 + 0.0200) = 1.1697

The forward is 1.72% above spot, because holding euros earns less interest than holding dollars and the forward compensates for that. Now compare four possible bank forecasts for one year ahead:

One-year forecast	Change vs spot	Expected excess return from buying EUR forward (as % of spot)
1.1500 (no change)	0.00%	-1.72%
1.1697 (equal to the forward)	+1.72%	0.00%
1.2000	+4.35%	+2.63%
1.1200	-2.61%	-4.32%

Three points follow.

A forecast of euro strength can be a bearish trade. A strategist who forecasts EUR/USD rising from 1.15 to 1.1697 is forecasting exactly what the forward already prices, so there is no trade in it. A forecast of 1.16 is euro-bullish against spot but implies selling euros forward, with an expected excess return of -0.85% on a long position.
“No change” is a view. If the strategist expects spot to stay at 1.15, the implied position is to sell euros forward and collect the 1.72% interest differential, which is the carry trade in another form.
Only the gap between forecast and forward is information. Professionals translate a published forecast into this gap before doing anything with it. A signal channel that posts “bank X targets 1.20” and says “buy” has skipped the step.

How good are exchange rate forecasts?

The academic record sets a low bar. Meese and Rogoff (1983) found that a random walk performed as well as any of the standard structural exchange rate models at one- to twelve-month horizons for the dollar against the pound, the mark and the yen, even when the models were given the actual future values of their fundamentals. Barbara Rossi’s survey of the literature three decades later, “Exchange Rate Predictability” (Journal of Economic Literature, 2013), concludes that the random walk remains a hard benchmark to beat; some predictors work for some currencies in some periods, but the gains are small and unstable over time.

None of this means bank strategists are uninformed. It means that the value of their work lies mostly elsewhere: explaining what is priced, mapping scenarios around central bank decisions, and documenting how policy and flows interact. Clients pay for the framework, access to the strategist, and a fast read on events, not for a point forecast that beats the forward.

Judging a strategist’s trade record

Some banks publish the running performance of their recommendations. The statistics are the same as for any track record (see statistical significance of a track record). Computed in Python, against the null hypothesis that each call is a coin flip:

Record	Hit rate	One-sided p-value
24 of 40 calls right	60%	0.134
30 of 50	60%	0.101
66 of 120	55%	0.158
60 of 100	60%	0.028

A 60% hit rate over 40 calls, which is more than a year of active recommendations for most desks, is not distinguishable from luck at conventional levels. To show a true 55% hit rate is two standard errors above 50% takes roughly 400 independent calls; a true 60% needs about 100. And hit rate is not the whole story: a desk recommending trades with a 3% target and a 1.5% stop breaks even at a 33% hit rate before costs, so the payoff ratio matters as much as the count. Calls on the same theme, such as several dollar-bullish trades in one quarter, are also not independent, which makes the effective sample smaller still.

Professionals therefore treat a published record as weak evidence, look at how ideas are constructed and closed, and pay more attention to whether the reasoning held up than to the scorecard.

How bank research reaches retail traders

Bank research is mostly behind client portals. What reaches retail audiences comes through three channels:

Press coverage of house forecasts and big thematic calls, which strips out the conditions and probabilities attached to them.
Public versions of macro and FX commentary that banks publish openly, partly because generic, openly available material sits more comfortably under the MiFID II minor non-monetary benefit rules.
Signal sellers and social media accounts that repost screenshots of trade ideas or forecast tables, sometimes with the bank’s name as a selling point. The claim is usually that you are getting “what the banks are doing.” What you are actually getting is one line of a document written for a different reader, often days old, with the disclosures removed, and possibly shared in breach of the bank’s distribution terms.

There is also a timing problem. A bank trade idea is published to thousands of institutional clients at once. By the time a screenshot reaches a Telegram channel, the institutional readers who acted on it have already traded, and the entry level may no longer be available.

How professionals use sell-side research

Desks that consume bank research for a living tend to share some habits:

Translate forecasts into forward-relative terms before comparing them with their own view.
Read several banks and use the dispersion of forecasts as a measure of uncertainty, rather than any single call as a guide.
Separate research from desk commentary, and read both sets of disclosures.
Use strategists as sparring partners. The most valued service is often a conversation that challenges the client’s own thesis before an event.
Keep records of which research inputs influenced which decisions, which under MiFID II is also how they justify paying for it.
Size independently. Position size comes from the portfolio’s risk budget, never from the note.

Some of the same distinctions between view and implementation appear in the academic evidence on rule-based strategies, covered in do technical trading rules work in FX.

The short version

Bank FX research is written by strategists and economists at dealing banks for institutional clients, as part of a trading relationship. In Europe, MiFID II made asset managers pay for it explicitly from 2018, with no exemption for currencies; the UK in 2024 and the EU through the Listing Act have since allowed research and execution to be paid for jointly again under conditions. In the US, FINRA’s research rules cover equity and debt securities, while CFTC Regulation 23.605 separates swap dealers’ research analysts from their trading desks and requires conflict disclosures. A bank forecast only contains information to the extent it differs from the forward rate, and a published track record of 40 or even 100 calls is weak evidence of skill. A screenshot of a bank trade idea in a signal group has lost its audience, its sizing framework and its timing. This is an explanation of how the industry works, not a recommendation to trade on any forecast.