How a Global Dollar Share Is Built
For central banks' official foreign-exchange reserve portfolios, the difficult question is not how to calculate a percentage but how to explain its movement. A global dollar share combines many countries with very different reserve sizes, and the same change in that share can arise from an active allocation decision, a change in the amount of reserves, or a change in the market value of existing assets. The useful judgment is to trace those paths separately before deciding what the aggregate says about central-bank behavior.
A global share is a weighted relationship
At its simplest, the dollar share of official reserves compares the value of dollar-denominated reserve assets with the value of all included foreign-exchange reserve assets. The global figure is formed by adding across countries before—or, equivalently for this purpose, by taking a reserve-weighted combination of—country-level holdings. It is therefore not the ordinary average of each country's percentage. A country holding $10 billion in reserves does not receive the same weight as one holding $1 trillion.
That difference creates two distinct facts that can both be true. A majority of countries might increase their dollar shares while the global share falls, if the countries reducing their shares hold much larger portfolios. Or a majority might reduce their shares while the global figure barely moves, if their portfolios are small relative to the countries that remain stable. The direction of the typical country and the direction of the aggregate are related, but neither substitutes for the other.
Consider an intentionally simple illustration. Imagine a large reserve holder with 70 percent of its portfolio in dollars and a small reserve holder with 20 percent. If the small holder cuts its dollar share sharply, the change may be important for that country's own liquidity choices but nearly invisible in the global total. If the large holder makes a modest reduction, the worldwide share can move materially even though the percentage-point change in its own portfolio is smaller. The arithmetic is not a statement about any particular country; it shows why portfolio size is part of the meaning of the statistic.
This is also why a global decline cannot be translated into a count of central banks “leaving” the dollar. The aggregate weights decisions by exposure, not by the number of decision makers. A country-level map and a global time series can appear to tell different stories without either being calculated incorrectly. They are answering different questions: one asks how countries changed their own portfolios, while the other asks how the combined stock of reserves changed.
The New York Fed analysis makes this distinction visible in two four-year comparisons. It reports a decline of roughly three percentage points in the dollar share from 2015 to 2019 and roughly two percentage points from 2019 to 2023. It also finds roughly equal numbers of countries increasing and decreasing their dollar shares in both periods, among countries with usable observations. The coexistence of those findings is not a contradiction. It is evidence that the weights attached to countries—and the mechanisms behind their movements—matter.
The first path: changing an existing preference
The New York Fed calls the first path the preferences channel. It captures a change in a country's dollar share, weighted by the reserve level at the beginning of the period. In ordinary language, it asks: holding the starting size of the portfolio in view, did the country shift the mix of assets it already held toward or away from dollars?
This channel is the closest of the two to the everyday idea of reallocation. If a reserve manager sells some dollar assets and purchases assets denominated in another currency, or otherwise changes the currency mix of the portfolio, the country's dollar share can fall. Because a large portfolio contains more assets whose allocation can affect the global total, the same percentage-point change has a larger aggregate contribution when it occurs in a large reserve stock.
The weighting by the starting reserve level matters. Suppose two countries each reduce their dollar share by five percentage points. If one began with ten times the reserves of the other, the larger country's preference contribution will be much larger under the decomposition. The comparison is not saying that the larger country made a more dramatic percentage change. It is saying that the decision was applied to a larger base and therefore had a greater effect on the aggregate stock.
The channel still does not provide a complete account of motive. It identifies a change in the observed currency allocation and measures its contribution under a specified weighting. It does not tell us which policy, risk, or liquidity consideration produced the change. An allocation result is evidence about what the portfolio did, not a direct transcript of why the reserve manager did it.
The second path: changing the size of reserves
The reserve change channel captures a different event. It follows what happens when a country adds to or draws down foreign-exchange reserves whose currency composition differs from the relevant average. The key comparison is between the country's starting dollar share and the average dollar share for the sample. The amount of the reserve stock changes, and that change enters the global ratio with a sign determined partly by the difference between those shares.
The counterintuitive case is a country that starts below the average dollar share and then accumulates reserves. If the newly accumulated reserves retain a composition below the average, the country has added more weight to the non-dollar side than the average portfolio would have added. The global dollar share is pulled down mechanically. Yet the country's own dollar share can rise at the same time if the new allocation is more dollar-heavy than its old allocation. “The country's dollar share increased” and “the country's reserve accumulation lowered the global dollar share” can therefore both be accurate.
The reverse can happen when a below-average country draws down reserves, or when a country with an above-average dollar share expands its reserve stock. What matters is not simply whether reserves rose or fell. It is the interaction between the change in portfolio size and the currency composition associated with that change. This is why the reserve change channel should not be treated as a second name for preference change. It describes the effect of a changing denominator and stock composition, not necessarily a decision to alter the mix of an existing portfolio.
A verbal accounting identity helps keep the paths distinct. The preferences channel changes the share applied to the starting stock. The reserve change channel changes the stock to which the relevant share applies. When both occur in the same period, the global movement is the combined result. Looking only at the ending percentage hides which part came from a change in the mix and which part came from a change in the amount being mixed.
When the two channels meet
The New York Fed's country examples show why channel labels are not merely technical bookkeeping. In the 2015–19 period, Switzerland was the largest negative contributor through the reserve change channel because it accumulated reserves while starting with a dollar share below the sample average. Yet Switzerland simultaneously increased its own dollar allocation, which pushed through the preferences channel in the opposite direction. Reserve growth lowered the aggregate dollar share; the change in Switzerland's own allocation raised it.
China provides the converse pattern in the same period. Its movement away from dollars dominated the negative preferences contribution, while its reserve drawdown contributed positively through the reserve change channel. A drawdown reduces the amount of the portfolio being counted. When the reserves being drawn down have a dollar share below the relevant average, that reduction can lift the remaining aggregate share even while the country's own preference contribution is pulling it down.
The point is not that China or Switzerland followed a single, easily summarized strategy. The point is that a single country's contribution can have more than one sign because the decomposition is tracking different aspects of its reserve position. The aggregate number records their net effect. Without the two channels, an observer might describe both cases simply as evidence that the country “moved away from the dollar” or “supported the dollar,” statements too coarse to explain what happened.
The channels can also reinforce one another. If a large country lowers its dollar share and expands reserves at a composition below the relevant average, both paths push the global share downward. If it raises its dollar share while expanding reserves at an above-average dollar composition, both can push upward. Those combinations are logical consequences of the definitions; they should not be attributed to a particular country or period unless the evidence reports them. The important analytical habit is to ask whether the channels point in the same direction or offset one another.
What the decomposition can prove
A decomposition is valuable because it turns an aggregate movement into a more informative accounting of contributions. It can show whether a reported decline is concentrated in a few large reserve holders, whether reserve accumulation is doing mechanical work, and whether country-level changes are directionally broad. It can also reveal why the same global result could have arisen through very different combinations of country decisions.
But the method has clear stopping points. It does not observe every transaction. It does not automatically identify the reason for a reallocation. It applies to the countries, dates, and definitions for which the underlying data are available. The New York Fed's contribution figures are therefore not a universal decomposition of every central bank's behavior, and they should not be extended silently to private holdings or to every dimension of the dollar's international role.
Nor should the two channels be confused with the valuation effects highlighted in quarterly reserve releases. The New York Fed's preferences channel concerns a change in a country's dollar share over the studied period, while the reserve change channel concerns the size and composition of the reserve stock. A later change in exchange rates or bond prices can alter the reported market value of those holdings without representing a comparable preference change. The time horizon and the measurement question must remain attached to the mechanism being discussed.
The diagram's two arrows are not two stages that always occur in sequence. They are two paths that can operate at the same time. A reserve manager can change the composition of an existing portfolio while the total reserve stock also changes. Their contributions may reinforce one another or cancel out. The ending global share is the result after those effects are combined, which is why it cannot by itself identify the path.
A better question than “Are countries selling dollars?”
Once the mechanism is clear, the question becomes more precise. Rather than asking whether “countries” are selling dollars, ask which countries changed their own shares, how large their starting portfolios were, and whether the reserve stock expanded or contracted at a composition above or below the relevant average. Then ask whether the observed data cover the large holders whose decisions could dominate the result.
That sequence does not make the aggregate less important. It makes the aggregate more interpretable. A broad, persistent decline in country-level dollar preferences would be meaningful evidence of a different kind from a decline driven by reserve growth in a few below-average portfolios. An offsetting reserve drawdown would tell a different story again. The global percentage is the starting observation; the channel decomposition supplies the explanation.
The next empirical question is therefore not whether the model is elegant, but what it reveals when applied to a specific period. The 2015–19 evidence provides a useful test because it combines a measurable aggregate decline with concentrated contributions from large reserve holders—and with countries whose preferences and reserve-size effects did not point in the same direction.