The 2015–19 Puzzle: China, Russia, and Switzerland Behind the Average
The official reserve portfolios in the 2015–19 comparison give the accounting model a demanding real-world test. The aggregate dollar share fell, but the relevant question is whether that decline represented a broad change in countries' preferences or a concentrated combination of portfolio and reserve-size effects. Following the reported contributors lets us make a more defensible judgment: the movement was material, yet its causes were not spread evenly across the countries in the sample.
A decline divided between two channels
The New York Fed analysis examines the reserve portfolios of seventy-nine countries for this period, with complete data at the beginning and end for seventy-six. That coverage is substantial enough to reveal patterns, but it is not the same as observing every reserve holder. The distinction matters from the outset: the results describe the countries and observations available to the analysis, not an unqualified census of all central banks.
Across the seventy-six countries with complete data, the aggregate dollar-share decline was divided almost evenly between the two channels. The preferences channel contributed 1.2 percentage points to the decline, while the reserve change channel contributed 1.5 percentage points. These are net contributions: positive and negative country-level effects have already been combined within each channel. A net number tells us how much a channel contributed overall, but not whether many countries made similarly sized moves or whether a few large contributions outweighed several smaller ones.
That distinction is the central finding of the period. The two channels were both material, so it would be incomplete to describe the decline as only a series of active decisions to reduce dollar allocations. At the same time, the fact that the channels were both important does not mean that responsibility was evenly distributed among countries. The New York Fed's underlying contributions show a high degree of concentration.
A useful way to read the result is to separate channel size from behavioral breadth. Channel size asks how much the global percentage changed through a particular mechanism after all country contributions were weighted and netted. Behavioral breadth asks how many countries moved in the same direction and by how much. A large channel can be generated by a small number of countries with large reserve stocks. A broad set of smaller movements can be real but have little effect on the global total. The aggregate cannot answer both questions at once.
The contribution presentation also requires care about units. A contribution measured in percentage points is a contribution to the change in the aggregate share, not a country's percentage-point change multiplied by an unweighted country count. The New York Fed presents the largest positive and negative contributors within each channel against the observed aggregate movement for the period. A country can therefore rank highly because it combines a meaningful portfolio change with a large starting reserve stock, while a country with a larger within-portfolio change can rank lower if its starting stock is small. The ranking is evidence about influence on the aggregate, not a league table of central-bank conviction.
This distinction helps explain why the period's almost-even channel split should not be read as a balanced division of responsibility among countries. The 1.2-point preference contribution and 1.5-point reserve-change contribution are net totals after positive and negative effects are combined. They reveal that both mechanisms mattered, but they do not imply that the same number of countries contributed equally to each one. The underlying cross-section is the part that tells us whether a channel was broad or concentrated.
China and Russia carried much of the preference effect
The preferences channel is the more direct measure of a change in an existing portfolio's currency mix. In the 2015–19 results, China and Russia dominated the downward contribution through this channel. Turkey, Peru, and Spain also added downward pressure, but each was much smaller—around negative 0.2 percentage point in the presentation. Sweden and Switzerland moved in the opposite direction by increasing their dollar portfolio shares.
The pattern matters for two reasons. First, the negative preference contribution was concentrated in large reserve holders rather than evenly distributed across the full country sample. That helps explain how the global share could fall even though the directional count of countries was roughly balanced between increases and decreases. Second, “increased” or “decreased” remains a description of portfolio allocation, not a complete explanation of the decision. The contribution tells us how a change in a country's dollar share affected the aggregate after weighting by its starting reserve level. It does not identify the central bank's motive.
The historical discipline is important here. Looking backward from a later aggregate statistic can create the impression that the contributors were destined to produce the outcome or that their choices all expressed the same intention. The evidence does not support either inference. China and Russia appear as large contributors to the measured preference effect in this period. That is enough to establish concentration; it is not enough to turn a contribution chart into a unified account of policy, geopolitics, or future behavior.
The smaller contributors also should not be treated as irrelevant. A country can move in the same direction as a large holder while making a much smaller contribution because its reserve portfolio is smaller. Conversely, Sweden and Switzerland's upward preference contributions offset part of the downward movement. The presence of positive contributions inside an overall negative channel is not noise. It is exactly what one should expect when countries are making independent reserve-management decisions rather than following a single global allocation rule.
There is an important difference between “concentrated” and “caused by only one country.” Concentration means that the weighted contributions are uneven: a few observations account for a large share of the net movement. It does not mean the other countries were motionless, or that their smaller positive and negative effects had no analytical value. The smaller effects establish that the result was assembled from many country portfolios, while the largest effects explain why the global statistic moved as far as it did. Both layers are needed to describe the case accurately.
A country-level contribution should therefore be read with three questions in mind. What was the country's own dollar-share change? How large was its reserve stock at the relevant starting point? And was the contribution being discussed generated through an allocation change or through a change in the amount of reserves? Those questions keep the case anchored to the measurement rather than inviting a retrospective story about intentions.
Switzerland shows why reserve growth is not preference change
The reserve change channel supplies the period's most counterintuitive case. Switzerland was the largest negative mechanical contributor through this channel because its dollar share at the start of the period was below the sample average and its reserves accumulated. That accumulation pulled down the cross-country aggregate: the expanding Swiss reserve stock carried less dollar weight than the average portfolio. Yet Switzerland simultaneously increased its own dollar allocation, producing a positive contribution through the preferences channel.
There is no contradiction between those statements. Imagine looking at Switzerland in two different ways. The first view compares the currency mix of its own portfolio at the beginning and end. On that measure, its dollar allocation increased. The second view asks what happened to the global ratio when Switzerland added to a reserve stock that began below the average dollar share. On that measure, the new stock exerted downward pressure. One view tracks the share within Switzerland; the other tracks the effect of changing Switzerland's weight in the global pool.
The reserve change channel therefore captures a mechanical interaction between portfolio size and starting composition. It does not label Switzerland a dollar avoider. In fact, the example demonstrates why that label would be misleading if it were based only on the global contribution. A country can raise the dollar share of its own holdings while making the global dollar share lower through the way its total reserve stock changes.
The same mechanism helps explain other contributions in the analysis. The Czech Republic and Russia also contributed negatively through reserve accumulation at below-average dollar shares. Hong Kong and China contributed positively through the reserve change channel; China's positive contribution reflected a reserve drawdown. These cases do not all describe the same reserve action. They show how the sign of a contribution depends on both the direction of the reserve-stock change and the currency composition associated with it.
A simple counterfactual clarifies the transmission path without inventing a motive. If Switzerland had kept its reserve stock at its initial size, the mechanical effect associated with its accumulation would not have entered the aggregate in the same way. That does not tell us what Switzerland would have done, and it does not erase the separate preference contribution. It only isolates the part of the reported result that comes from the change in portfolio size. Counterfactual reasoning is useful here because it asks which condition must be changed to remove an effect; it is not a claim about an unobserved alternative history.
China and Switzerland point in opposite directions at once
The New York Fed explicitly uses China and Switzerland to illustrate that a single country can contribute through opposing channels simultaneously. China contributed negatively through preferences as its dollar allocation fell, but its reserve drawdown contributed positively through the reserve change channel. Switzerland's own dollar allocation rose, creating an upward preference contribution, while its below-average-share reserve accumulation created a downward reserve change contribution.
This is more than an interesting exception. It exposes the danger of treating a country's net contribution as a one-word verdict. If an observer looks only at China's net effect, the opposing mechanisms disappear and the country may be described simply as having weakened the dollar. If the observer looks only at Switzerland's reserve accumulation, the increase in its own dollar allocation disappears and the country may be described as diversifying away from dollars. Both summaries discard information that the decomposition makes visible.
The channels can offset one another within a country, as in these examples, or reinforce one another if both the allocation change and the reserve-size effect push the aggregate in the same direction. The reported data establish the opposing contributions for China and Switzerland in this period. They do not establish that either country maintained the same pattern in every later period, nor do they identify a single strategic motive behind the measured movements.
The diagram is intentionally structural rather than chronological. A country contributes through two possible analytical paths, and each path can raise or lower the global dollar share. The paths meet at the aggregate outcome, but they do not become the same path merely because they are observed over the same four-year window. Keeping them separate is what preserves the explanation.
This also changes how a reader should approach a chart that ranks contributors. Start with the sign, but do not stop there. A negative preference contribution indicates downward pressure from a change in the country's dollar share after the stated weighting; it does not say that the country's entire reserve portfolio became less useful or that its total reserves shrank. A negative reserve-change contribution indicates downward pressure from the size and composition of new or withdrawn reserves relative to the average; it does not say that the country deliberately reduced its existing dollar allocation. The same label—“negative contributor”—can describe different economic events.
The 2015–19 case is consequently a test of explanation, not a morality play between currencies. China and Russia matter because their weighted preference contributions were large. Switzerland matters because its reserve accumulation demonstrates the mechanical channel in a way that runs against the simplest interpretation of its own allocation. The evidence becomes more informative when those cases are held together rather than collapsed into a single narrative about winners and losers.
What the 2015–19 case establishes
The period supports three bounded conclusions. First, the decline cannot be reduced to one mechanism: the preferences and reserve change channels made contributions of similar overall size. Second, the underlying effects were concentrated. China and Russia dominated the negative preference contribution, while Switzerland was especially important on the reserve change side, and other countries contributed in both directions. Third, the country cases show that an aggregate decline does not imply that every country's own dollar share fell.
Those conclusions are stronger than a bare chart of the global share, but narrower than a claim that central banks as a group had adopted a common anti-dollar strategy. The sample contains countries with incomplete coverage, and the contribution framework attributes movements within the available data rather than reading motives directly. The case is therefore best understood as evidence about how the aggregate was assembled in 2015–19, not as a final verdict on all official reserve behavior.
The next difficulty follows directly from that limitation. In a later comparison, several of the largest reserve holders are not observed at the endpoint. Once the portfolios capable of carrying the largest weights are missing, the analyst must distinguish what the complete sample shows from what the aggregate implies about an unobserved remainder.