Most liquidity overlays run outside the policy document. The framework exists, the desk uses it, the committee has seen the dashboard, and nothing binding has been written down. That arrangement works fine until the overlay is at its extreme setting during a bad quarter and somebody asks under what authority the allocation moved. At that point an undocumented process is indistinguishable from discretion, and discretion is the thing an investment policy statement exists to bound.
What a clause has to name before it can bind anything
A regime clause is only enforceable if every term in it resolves to something a third party could check without asking you what you meant.
Start from what the source actually publishes. The Regime tab carries a COMPOSITE score on a stated 0 to 100 scale, currently 85. It carries a REGIME field with a sub-label of "Risk classification", currently RISK-ON. It carries a POLICY field, currently EASING. It carries three component reads, LIQUIDITY on net flows, FUNDING on the SOFR and IORB relationship, and MARKETS on asset momentum, all three currently NEUTRAL. And it carries a refresh stamp, 08:19 AM on the capture in front of me. Coverage is stated as eight central banks, being the Fed, ECB, Bank of Japan, People's Bank of China, Bank of England, Swiss National Bank, Bank of Canada and Reserve Bank of Australia.
Now the drafting problem. The module does not publish the rule that produces the word in the REGIME field. That is not a criticism of the vendor, it is a constraint on you. A clause that says "when the regime is classified risk-off" delegates the definition of your allocation trigger to an undisclosed and revisable third-party rule. If the vendor changes the classification methodology, your policy changes with it, silently, and nobody at your firm approved that.
So the clause defines the state on the numeric composite, which is published on a stated scale, and treats the vendor's word as corroborating evidence rather than as the trigger.
Draft language for the trigger
Short, and deliberately mechanical.
The Liquidity Overlay applies to the Overlay Sleeve only, defined as up to twenty five percent of total portfolio market value. The remaining Core Allocation is not subject to this clause under any regime state. The Liquidity State is determined by the composite liquidity score published by the Global Liquidity Scorecard on its stated 0 to 100 scale, observed at the close of business on the last business day of each week. A change in Liquidity State is recognised only when the composite has been within the band of the new state at two consecutive weekly observations. No intra-week observation constitutes a state change.
Three things are doing work there. The sleeve cap means the worst case is bounded before any judgement is exercised. The fixed observation time removes the ability to choose a favourable reading. The two-observation persistence requirement removes the ability to trade a wobble and, more importantly, removes the argument about whether a given wobble counts.
Draft language for the bands and the ranges
Bands are your decision, not the vendor's. Set them from your own tolerance and your own reconstruction of the composite's history, and state in the clause that they are internally set.
| Liquidity State | Composite band | Overlay Sleeve risk asset range |
|---|---|---|
| Expansionary | Upper band | 75 to 100 percent of sleeve |
| Neutral | Middle band | 40 to 75 percent of sleeve |
| Contractionary | Lower band | 0 to 40 percent of sleeve |
Ranges rather than points, and overlapping at the edges on purpose. A policy that specifies a single target allocation per state forces a trade every time the state changes and turns a slow macro read into a turnover obligation. Overlapping ranges mean a state change permits a reallocation without compelling one, and the person implementing it can wait for a decent execution window.
Fill in the numeric band edges yourselves and record the date they were set. Do not let them be revised in the same meeting that discusses current positioning, because band edges chosen while looking at today's reading are not bands, they are a position with a policy citation attached.
Who may invoke it, and what evidence is required
The clause needs an owner, an alternate and an evidence standard, or it will be invoked by whoever happens to be looking at the screen.
The Chief Investment Officer, or in their absence the named alternate, may implement a reallocation within the range prescribed for the prevailing Liquidity State. Implementation requires a Liquidity State Record filed before the first order is placed, containing the two qualifying weekly observations with their composite values and refresh timestamps, the regime and policy fields as displayed, the three component reads, the resulting state and target sleeve allocation, and the signature of the invoking officer. Movement outside the prescribed range requires prior approval of the Investment Committee.
The record is the part that matters in a review twelve months later. A screenshot in a folder proves somebody opened a page. A transcribed record with the composite value, the refresh stamp and the two qualifying observations proves what was visible at the moment the decision was made, before any underlying series were revised.
Add one line to the record template that most firms omit. The effective as-of date of the reading, which is governed by the slowest weighted input in the composite rather than by the refresh time on the header. A composite blending monthly money supply with daily credit spreads is not as current as its timestamp suggests, and a reviewer who understands macro data will look for that acknowledgement.
The failures the clause has to survive
Draft for the days the source does not behave, because those are the days the clause is tested.
Vendor methodology change. Include a provision that any disclosed change to the composite construction suspends state changes until the Investment Committee has reviewed the change. Without it, a vendor revision silently rewrites your allocation policy.
Source unavailability. Specify what happens if the reading cannot be obtained at the scheduled observation. The safe default is that the prevailing state persists, and that a stated number of consecutive missed observations forces the sleeve to the neutral range rather than leaving it at an extreme setting on stale information.
Whipsaw. The two-observation rule handles most of it, but add a minimum dwell time. A state that has been in force for fewer than a stated number of weeks does not permit a second reallocation. This prevents the sleeve from being traded around a boundary and it caps the turnover the clause can generate.
Annual review. The clause should require the committee to review the band edges and the sleeve cap once a year and to record the review even when nothing changes. An unreviewed mapping drifts out of alignment with the portfolio it governs, and the moment somebody discovers that during a drawdown is the worst possible moment to renegotiate it.