The Missing Giants: What 2019–23 Reveals—and Leaves Uncertain

The official reserve portfolios in the 2019–23 comparison pose a different problem from the earlier country cases: several of the largest holders are not observed at the endpoint. The question is not whether missing data are inconvenient, but how much of the aggregate movement can be attributed to the countries we can see and how much is inferred about those we cannot. The useful judgment is to distinguish a directly observed contribution from an assumption-dependent residual, without treating the missing evidence as either harmless or fatal.

The missing observations are large enough to matter

The New York Fed analysis has complete dollar-share data for sixty-two countries in the 2019–23 comparison. China, Russia, Mexico, and Morocco reported usable information in 2019 but did not disclose their 2023 dollar allocations in the country-level data used for the comparison. Those four countries are not a random handful of small observations. Their combined reserves were roughly equal in size to the reserves held by the remaining sixty-two-country sample combined.

That relative size changes the meaning of “the sample.” If a small country is absent, the missing observation may have little effect on a reserve-weighted aggregate. If countries holding a large share of the relevant reserve stock are absent, the observed countries may describe only part of the movement that matters most to the global percentage. The problem is therefore not simply that the number of countries falls from a larger total to sixty-two. It is that the missingness is concentrated among portfolios capable of carrying substantial weight.

This also explains why complete data can be both useful and insufficient. The sixty-two countries still provide observed evidence about their own dollar shares, reserve changes, and contributions. Their results are not made unreliable merely because other countries are missing. But they cannot, by themselves, be treated as representative of the entire set of official reserve portfolios when the missing holders are unusually large.

A careful analyst should separate two questions. What did the countries with complete data do? And what must have happened in the unobserved part of the reserve pool for the overall COFER movement to look as it did? The first can be answered from observed country-level changes. The second can be approached only through assumptions that connect the observed sample to the aggregate.

What the complete sample actually shows

For the sixty-two countries with complete data, the preferences channel was slightly positive at 0.3 percentage point. In other words, taken together and using the analysis's specified weighting, these countries increased their dollar allocations enough to make a small upward contribution through the preference path. Their reserve change channel contributed a modest negative 0.5 percentage point. The two effects largely offset in the complete sample, so the countries that can be observed directly contributed relatively little to the much larger decline in the aggregate dollar share.

This result is easy to misread in two opposite ways.