Someone always pastes me a global M2 chart with bitcoin overlaid on top, shifted forward by some number of weeks, and the fit looks uncanny. Liquidity goes up, bitcoin follows a couple of months later, the two lines hug each other for years. The implied trade is obvious. Watch the liquidity line, front-run the crowd. I used to nod along at these. Then I tried to rebuild one from scratch, and the whole thing came apart in ways that changed how I read every liquidity chart since.
The short version is that a lot of what looks like global liquidity in these charts is actually the dollar in disguise, and the famous bitcoin lag is mostly a number someone tuned until the picture looked good. Neither problem means global M2 is useless. It means you have to know what you are looking at before you size anything off it.
The FX problem hiding inside the sum
Global M2 is supposed to be the total broad money across major economies. The trouble starts the moment you add them up. Each country reports its money supply in its own currency. China reports M2 in yuan, the euro area in euros, Japan in yen, and so on. To get a single global number in dollars, you convert every one of those local figures at the current exchange rate and sum the result. That conversion step is where the chart quietly turns into something else.
Think about what happens on a day when nothing changes in the actual money supply anywhere. No central bank prints, no bank makes a loan, every local M2 sits still. Now the dollar weakens by a couple of percent against everything. When you reconvert all those unchanged local piles into dollars, the global total jumps. Not because there is more money, but because each foreign pile is now worth more dollars. Do the reverse on a strong-dollar day and global M2 drops even though nothing was destroyed.
So a big chunk of the wiggle in a dollar-denominated global M2 line is just the inverse of the dollar index. When people overlay that line against bitcoin and marvel at the correlation, part of what they are seeing is that bitcoin tends to rise when the dollar falls, which we already knew and did not need M2 to tell us. You have taken a dollar chart, flipped it upside down, dressed it up as liquidity, and rediscovered a relationship that was FX all along.
The tell is easy to check. Pull up the dollar index next to any dollar-denominated global M2 chart and eyeball the turning points. If the big swings in your liquidity line line up with dollar swings rather than with anything a central bank actually did, you are looking at an FX chart wearing a liquidity costume.
Constant-currency aggregation, which is the fix
The way out is to stop letting the exchange rate move under your feet while you measure. Pick a single reference date and freeze the FX rates as of that date. Convert every country's entire history using those frozen rates. Then sum. This is what people mean by constant-currency or fixed-FX aggregation, and it is the same trick multinational companies use when they report revenue growth ex-currency so a weak quarter for the dollar does not flatter the numbers.
Here is the workflow I actually use when I want a global M2 series I can reason about.
- Gather local-currency M2 for the economies that matter. In practice a handful of them dominate the total, so the US, euro area, China, and Japan get you most of the way, and adding the next several barely moves the shape.
- Choose a base date for your FX rates and write it down. Everything downstream depends on this choice, so it is not a footnote.
- Convert each country's full local-currency history to dollars using its rate on that base date only. Same rate for every point in that country's series.
- Sum across countries to get the global series.
- Optionally build a second version using live FX and keep both. The gap between them is your FX distortion, and seeing it plotted is more educational than any explanation.
When you do this, the chart calms down. A lot of the drama that used to look like liquidity flooding in and draining out turns out to have been the dollar sloshing around. What is left is closer to the thing you actually wanted, which is whether the world is creating broad money faster or slower.
One honest caveat. Constant currency is not automatically more correct, it is just cleaner for measuring money creation. If your thesis is genuinely about dollars available to chase assets, then live FX might be the more relevant lens, because a foreigner's buying power in dollars really does change when the dollar moves. The point is not that one is right and one is wrong. It is that they answer different questions, and most charts floating around do not tell you which one they are answering.
The lag is curve-fit, and here is how to catch it
The other move that deserves suspicion is the forward shift. You have seen it. Bitcoin overlaid on M2 but slid forward by ten weeks, or twelve, or whatever makes the lines kiss. The problem is that the lag is almost always chosen after the fact by sliding bitcoin back and forth until the overlap looks best on the history you already have. That is curve-fitting with one knob, and one knob is plenty to fool yourself over a chart with a few big shared moves.
Two things give it away. First, the optimal lag tends to drift. Refit it a year later and the magic number has changed, which is exactly what you would expect from noise rather than a stable lead-lag relationship. Second, the fit rests on a small number of large regimes rather than many independent episodes. When two series share three or four big up-and-down cycles, almost any smooth alignment will look convincing, and your eye will happily fill in the rest.
A cheap sanity check I run before trusting any lagged overlay:
- Ask what lag was used and whether it was picked in advance or discovered by tuning. If nobody can answer, treat it as tuned.
- Split the history. Fit the lag on the first half, then see whether the same lag holds up on the second half you did not fit on. Out-of-sample is the whole game.
- Count the independent events driving the fit. If it is really three big cycles, you have three data points, not three years of daily bars.
- Check whether the relationship survives on the constant-currency version. If the lovely fit only exists on the FX-contaminated series, you were trading the dollar with extra steps.
What to actually do with all this
None of this makes global liquidity irrelevant. Broad money conditions genuinely matter for risk assets, and a well-built constant-currency series is a reasonable thing to keep on your screen as slow-moving context. What changes is how much weight you let it carry. It is a backdrop, not a trigger. A liquidity line that has been rising for months tells you the wind is roughly at your back. It does not tell you to buy on Tuesday because a shifted overlay says bitcoin is due.
When someone hands you a liquidity-versus-bitcoin chart now, you have a short list to run before you take it seriously. Is it dollar-denominated, and if so how much of the wiggle is just the inverse dollar. Was the lag chosen in advance or tuned to fit. Does the relationship survive out of sample and on a constant-currency build. Most charts you see in the wild fail at least two of those, and knowing which two is usually enough to stop you from sizing a trade on a picture that was mostly FX and hindsight the whole time.