The Dollar Share Is a Signal, Not a Verdict
Central banks' official foreign-exchange reserve portfolios are the subject here. The question is whether a falling dollar share in global reserves means that central banks, broadly, are abandoning dollar assets. The evidence supports a more careful judgment: the aggregate share can decline even when countries move in different directions, when reserve portfolios change size, or when market prices alter the value of existing holdings. The practical skill is to separate those possibilities before turning a percentage into a claim about the international monetary system.
The headline is real—and still ambiguous
The headline is striking. The dollar's share of global official foreign-exchange reserves fell from 64 percent in 2015 to 56 percent in 2025, according to a recent analysis by the Federal Reserve Bank of New York. That is a substantial change in a widely watched statistic. It is reasonable to ask whether it signals a weakening of the dollar's role in official portfolios.
But the number is a global aggregate, not a survey in which every central bank gets one equal vote. A country with very large reserves has more influence on the reported share than a country with a small reserve portfolio. If one large reserve holder changes its allocation, the global percentage can move more than the behavior of many smaller holders combined. Conversely, if a country whose dollar share is below the average builds its reserves, that accumulation can pull down the global share even if the country raises the dollar portion of its own portfolio.
That distinction changes the question. Instead of asking only, “Did the dollar share fall?”, an analyst must ask, “Whose portfolios changed, in what way, and through which mechanism?” The first question describes an outcome. The second begins to explain it.
An illustrative comparison makes the weighting problem concrete. Suppose a very large reserve holder keeps most of its assets in dollars while a small holder shifts a large fraction of its much smaller portfolio into another currency. The small holder may have made the more dramatic percentage change within its own portfolio, yet the global share may barely move. Reverse the sizes and the same kind of country-level decision can become visible in the aggregate. The statistic is therefore sensitive not only to the direction of a decision but also to the amount of reserves exposed to it. That is why a falling global percentage cannot be translated directly into the proportion of central banks that are reducing dollar exposure.
A recent quarterly movement shows why the time horizon matters as well. The IMF's July 2026 data brief reported that the dollar share rose from 56.42 percent of official reserves in 2025's fourth quarter to 57.13 percent in the first quarter of 2026. The brief attributed around half of that increase to exchange-rate-driven valuation effects as the dollar appreciated modestly against major currencies. A rise in the share over one quarter does not, by itself, disprove a long-run decline; nor does a long-run decline prove that most reserve managers made a fresh decision to sell dollars. The two observations operate at different horizons and answer different questions.
What official reserve data can—and cannot—show
Official foreign-exchange reserves are assets held by public institutions for purposes connected to external financial stability and monetary management. The data considered here describe the currency composition of those official holdings. They do not provide a complete account of every use of the dollar: private borrowing, trade invoicing, payment networks, and other forms of international finance are outside this measurement.
That boundary is important because “the dollar's role” is a much larger idea than “the dollar's share of official reserves.” A reserve share can tell us something meaningful about how public reserve portfolios are composed. It cannot, on its own, settle whether the dollar remains central to global funding markets or whether private actors are changing their behavior. The strongest conclusion must stay attached to the subject actually measured: official reserve portfolios.
Nor does the statistic directly reveal intention. A reported share is a stock measured at a point in time. It reflects what assets are held and how they are valued, not a complete record of every purchase, sale, maturity, or currency-management operation that produced that position. The New York Fed's analysis helps by separating important sources of change across countries and over multi-year windows. The IMF's quarterly release adds a different kind of information by identifying valuation and reporting effects that can matter over short intervals. Neither source turns the aggregate into a transparent window on every central bank's strategy.
This stock-versus-flow distinction is easy to lose in a headline. A portfolio can end a quarter with a larger dollar share because the dollar appreciated against the currencies in which other assets are denominated, even if the reserve manager did not deliberately increase dollar purchases. It can also end with a smaller share after a change in bond prices, a reserve drawdown, or a rebalancing decision. The closing composition is the result that the dataset records; the route to that result has to be inferred from the available decomposition and explanatory notes. Good interpretation therefore asks not only what the portfolio contains, but what changed between the two observations and which parts of that change are directly visible.
Within the New York Fed's two multi-year windows, the aggregate dollar share declined by about three percentage points from 2015 to 2019 and by about two percentage points from 2019 to 2023. Those windows are useful precisely because they do not have identical evidence conditions. The earlier comparison has more complete country-level observations, while several large holders are missing from the later endpoint. The numbers establish a movement in the aggregate; they do not remove the need to inspect who is represented and how the movement was attributed.
Five questions turn a percentage into evidence
A disciplined reading of the headline starts with five checks.
First, how broad is the movement? If roughly equal numbers of countries increase and decrease their dollar shares, the evidence does not resemble a uniform shift in which most central banks reduce their allocations. The New York Fed reports that this directional balance appeared in both four-year windows it examined. Countries with changes smaller than half a percentage point in absolute value were classified as showing no change. That threshold belongs to the study's comparison; it is not a universal rule for every dataset. The broader lesson is to compare the number of countries moving in each direction rather than infer breadth from the global average.
Second, how much does each portfolio weigh? A change by a very large reserve holder can dominate the aggregate, while numerous small changes may barely register. This is why a global share should not be translated into a count of countries “leaving” or “staying with” the dollar. Breadth and weight are separate dimensions of the evidence.
Third, did countries change their preferences, or did the size of their reserve portfolios change? The New York Fed calls these the preferences channel and the reserve change channel. The first follows a change in a country's dollar share, weighted by the size of its starting reserves. The second follows the effect of adding to or drawing down reserves at a currency composition different from the relevant average. The same country can therefore affect the aggregate through both channels, potentially in opposite directions.
Fourth, what is the time horizon? A four-year change can reveal persistent portfolio patterns that a single quarter cannot. A quarterly change can be heavily affected by currency appreciation, interest rates, and the market value of bonds already held. Treating every movement as the same kind of event erases the mechanism that matters.
Fifth, who is missing from the data, and were earlier figures revised? Country-level disclosure is not always complete. If the missing countries hold large reserve portfolios, their absence can alter the interpretation of the observed sample. The IMF also notes that COFER data are revised when reporting is updated or corrected. A careful reader treats coverage and revisions as part of the measurement, not as afterthoughts.
These questions divide the problem into two related comparisons. The first is a cross-section: are many countries changing in the same direction, or are a few large portfolios carrying the result? The second is a decomposition through time: did the dollar share change because an existing portfolio was reallocated, because the reserve stock expanded or contracted, or because the market value of its assets moved? The cross-section tells us about breadth and concentration. The decomposition tells us about mechanism. A credible judgment about broad de-dollarization needs both, along with enough reporting coverage to know whether the largest portfolios are actually in view.
A useful discipline is to resist the temptation to let one kind of evidence stand in for another. A chart showing a declining global share cannot answer the breadth question. A list of countries that reduced their shares cannot answer the weighting question. A quarterly valuation explanation cannot settle a multi-year preference question. Each observation earns a narrower conclusion than the headline may invite, and the conclusions become stronger when the different tests point in the same direction.
These checks are not a way to talk away the decline. They are a way to identify what the decline means. A concentrated preference change by a few large holders is a real development. So is a broad-based shift across many countries. They have different implications, and the aggregate share alone cannot tell them apart.
The scope of the judgment
The New York Fed's conclusion is not that no country is moving away from dollar assets. It is that the decline after 2019 was not a broad-based phenomenon in the countries observed, and that a small number of large reserve holders made an outsized contribution. The analysis also finds that traditional reserve-management needs—dollar liquidity, currency management, and insurance against funding shocks—remain important. Those functions can coexist with diversification by individual countries.
That is a narrower claim than saying “nothing is changing,” but it is also a stronger claim than the headline alone can support. It says the available evidence does not show a systematic global shift in official portfolios. It leaves room for future change, and it specifies what would make that change more convincing: persistent preference movements across a broad set of countries, evidence that survives reserve-size decomposition, and results that are not mainly explained by exchange rates, bond prices, missing observations, or reporting revisions.
The decisive distinction, then, is not between a dollar share that fell and one that did not. It is between a number that describes the aggregate and an explanation of how the number got there. To make that distinction operational, the aggregate must be taken apart: one path follows changes in portfolio preference, while another follows changes in the size and composition of reserves.