The Global Liquidity Scorecard aggregates eight central banks: the Fed, the ECB, the Bank of Japan, the People's Bank of China, the Bank of England, the Swiss National Bank, the Bank of Canada and the Reserve Bank of Australia. Those institutions report in eight different currencies. There is no arithmetic that turns eight currencies into one number without choosing a unit to express the answer in, which means every global liquidity aggregate ever constructed has a base currency inside it, whether or not the person quoting it was the one who chose it.
Most allocators never interrogate that choice. They should, because the numeraire is not a display preference. It is a modelling decision that can flip the sign of the thing you are measuring.
The same policy, two opposite readings
Take the cleanest possible example and do the arithmetic. Suppose one of the eight expands its balance sheet by 5 percent over some period, measured in its own currency. Unambiguous easing, locally. Now suppose its currency depreciates 10 percent against the dollar over the same period.
Translated into dollars, that balance sheet is now worth 1.05 multiplied by 0.90, which is 0.945. A contraction of about 5.5 percent. The central bank added liquidity. The dollar denominated aggregate records a withdrawal. Both statements are correct and they are describing the same policy.
Scale that up and the implication is uncomfortable. A dollar denominated global liquidity aggregate carries an embedded short dollar position. When the dollar strengthens broadly, the aggregate falls mechanically even if all eight institutions are expanding. When the dollar weakens, it rises even if all eight are flat. If your macro overlay is driven by that aggregate and you have not decomposed it, you are running a currency trade you did not authorise and cannot see in your risk report.
The right numeraire is the one your liabilities are in
There is no abstractly correct base currency. There is a correct one for a given book, and the test is simple: what unit do you owe money in.
If the fund's capital was subscribed in dollars, reported in dollars and its fee is struck in dollars, then dollar denominated liquidity is the relevant measure, because dollars are the unit in which the outcome will be judged. If the vehicle funds in euros, the same reasoning points at a euro numeraire, and a dollar based read will contain a EURUSD term that has nothing to do with the investment thesis and everything to do with translation.
This gets missed because the default is almost always dollars, and dollars feel neutral in a way no other currency does. They are not neutral. They are a choice that happens to be shared by enough people that nobody argues about it, which is a different property entirely.
The practical consequence for a multi currency shop is that you may need the same composite read twice, once in the base currency of each share class, and you should expect them to disagree during periods of large FX moves. That disagreement is real information about which of your clients is actually facing tighter conditions, and it is invisible if you only ever look at the aggregate in one unit.
Decomposing the move so the FX term is explicit
The technique here is borrowed straight from multinational financial reporting and it works exactly as well on central bank aggregates as it does on group revenue.
Any change in a translated aggregate splits into two parts. The local currency change, which is what the institution actually did, and the translation effect, which is what the exchange rate did to it. Compute both. Report both. Never report only the sum.
| Component | How it is computed | What it tells you |
|---|---|---|
| Local change | Restate the current period at the prior period exchange rate, then compare to prior period | What the central bank did |
| Translation effect | Current period at current rate, minus current period at prior rate | What FX did to the reported number |
| Reported change | The sum of the two | What a single translated aggregate shows you, undifferentiated |
Run that per country and you can answer the question that actually matters, which is whether a move in the composite is a policy event or an exchange rate event. If the local change column is near zero across all eight and the translation column carries the whole move, then nothing happened to global liquidity and something happened to the dollar. Those call for completely different responses and a single composite figure cannot distinguish them.
The tab strip on GLS carries a Countries view alongside the dashboard, and a per country breakdown is the right place to take this question, because the decomposition only exists before the eight are summed. Once they are collapsed into one score the FX term is unrecoverable.
Rules for citing an aggregate in a document
Four things I would insist on before a global liquidity figure appears in anything that leaves the desk.
- State the numeraire in the same sentence as the number. Not in a footnote. A liquidity figure quoted without its unit is as incomplete as a return quoted without a currency, and for the same reason.
- Fix it once, per book, in writing. The numeraire is chosen from the liability structure, not from the analysis, and it does not change because a different unit tells a more comfortable story. Retroactively switching base currency is one of the easier ways to produce whatever conclusion you already wanted, and it is very hard to spot from outside.
- Carry the decomposition wherever the aggregate goes. If the reported change is meaningfully different from the local currency change, say so in a clause. Reviewers who understand translation will ask, and the answer should already be on the page.
- Check whether the FX exposure you have just discovered is already in the book. This is the one people forget. If the macro overlay is implicitly short dollars through the composite, and the portfolio is also long a basket of non dollar assets, the overlay is not diversifying the position, it is doubling it. That concentration will not appear in any exposure report because nobody classified the liquidity read as an FX exposure.
The broader point is that a composite is a compression, and every compression discards something. With mixed frequency data what gets discarded is the vintage. With multi currency aggregation what gets discarded is the numeraire. Both are recoverable if you ask before the number reaches the page, and neither is recoverable afterward from a screenshot.