The monetary aggregates are nested measures of how much money exists, ordered from the narrowest and most liquid outward. They are counting exercises, and knowing exactly what each one counts is what stops you over-reading them.
| Aggregate | Counts | Who creates it |
|---|---|---|
| M0 (monetary base) | Physical currency plus commercial bank reserves held at the central bank | The central bank directly |
| M1 | Currency in circulation plus demand deposits — money you can spend immediately | Mostly commercial banks, through lending |
| M2 | M1 plus savings deposits and other near-money that converts to spending easily | Mostly commercial banks |
That mechanism explains why central bank action does not translate reliably into money growth. A central bank can make reserves abundant and cheap, and if banks are unwilling to lend or borrowers unwilling to borrow, M2 barely moves. Reserves are not lent out to the public — they circulate only between banks and the central bank.
Common belief
"M2 is going up, so asset prices must go up."
What is actually true
M2 growth is one input among several, and the relationship is regime-dependent. Money can grow while it is being held rather than deployed. It can grow while real yields rise, which pulls in the opposite direction. Periods exist with strong M2 growth and falling asset prices, and the reverse.
V = nominal GDP / M2
Velocity is how often each unit of money is spent. It is not measured directly — it is defined as the residual that makes the identity balance. So "velocity fell" often means "spending did not rise as much as money did", which is a description rather than an explanation.
Suppose M2 grows sharply because households receive support payments and deposit most of them. The money exists and is counted, but if it is held in savings rather than spent or invested, the effect on prices is muted and delayed. The aggregate moved; the behaviour did not.
In VILIQ, M2 is one of roughly twenty macro factors feeding the flow engine, normalised into a z-score and a momentum reading so it can be compared with factors measured in entirely different units. It is never read alone, and the engine reports how much of its expected input set was actually available.