Key takeaways
- CFA Institute's Elements of an Investment Policy Statement for Individual Investors lists 16 elements across four sections, and states the reason for the document in one sentence: an IPS "can offer an objective course of action to be followed during periods of market disruption when emotional or instinctive responses might otherwise motivate less prudent actions."
- CalPERS' Total Fund Investment Policy, effective 10 June 2024, runs to 120 pages, but the part that binds behaviour is a single table: 37% public equity with a range of plus or minus 7 percentage points, 28% income at plus or minus 6, and 17% private equity, 15% real assets and 8% private debt at plus or minus 5 each.
- Norway's sovereign fund operates one rebalancing sentence — trade when the benchmark equity share is more than 2 percentage points from its 70% strategic weight at month-end. In March 2020 that share fell to 65.8%, the rule triggered, and the fund bought equities into the crash.
- The evidence that written pre-commitment works is strong and comes from other fields. Gollwitzer and Sheeran's meta-analysis of 94 independent tests found an effect of d = 0.65 on goal attainment. A randomised commitment savings account in the Philippines raised average balances across everyone offered it by 81 percentage points over 12 months — but only 202 of the 710 offered, 28.4%, agreed to be bound.
- Panic is rarer than the folklore suggests. In 2020, the most volatile year since 2008, 10% of Vanguard defined-contribution participants traded and 90% didn't. By 2025 the figure was 5%.
The honest answer: the mechanism is well evidenced, the application to your portfolio isn't
How do you stop yourself making a panicked decision in the next crash? The most common answer is to write an investment policy statement. It's a short document that fixes, in advance and in calm conditions, what your portfolio is for, what it holds, and what has already been decided about what happens when it falls.
The mechanism is real and well documented. Deciding a rule in advance and deciding it in the moment are two different cognitive tasks. The first has better evidence behind it. What has almost no direct evidence behind it is the narrower claim that a retail investor who writes such a document ends up wealthier than one who doesn't. Nobody has run that trial.
The only definition of an IPS that carries legal weight comes from pension law
Personal-finance writing treats "investment policy statement" as a genre. Pension law treats it as a defined term. The US Department of Labor's interpretive bulletin on written statements of investment policy — 29 CFR 2509.94-2, issued in 1994 — defines it as "a written statement that provides the fiduciaries who are responsible for plan investments with guidelines or general instructions concerning various types or categories of investment management decisions ..."
The next sentence is the one that matters, because it separates an IPS from a to-do list: "A statement of investment policy is distinguished from directions as to the purchase or sale of a specific investment at a specific time ..."
An IPS is a rule about a class of decision, written before any instance of it arrives. That is why it can be written calmly. It never has to know what the market is doing on the day it gets applied.
The Department added that maintaining such a statement "is consistent with the fiduciary obligations set forth in ERISA section 404(a)(1)(A) and (B)". Once a named fiduciary issues one it becomes part of the "documents and instruments governing the plan", so a manager subject to it is required to comply.
CFA Institute's list runs to 16 elements, and only a few of them are about markets
The most detailed public specification for an individual's IPS is CFA Institute's Elements of an Investment Policy Statement for Individual Investors, published in May 2010. It sets out 16 elements under four headings: "Scope and Purpose", "Governance", "Investment, Return, and Risk Objectives", and "Risk Management". Each comes with sample wording drawn from fictitious clients.
Most of the 16 elements are about who decides what, who watches, what the money is for, and what's off-limits. Returns appear in one section of four.
The document is candid about why it exists. After listing governance, allocation, implementation, monitoring and reporting, its introduction adds: "Perhaps most importantly, the IPS serves as a policy guide that can offer an objective course of action to be followed during periods of market disruption when emotional or instinctive responses might otherwise motivate less prudent actions."
That is a behavioural claim, made by a standards body, inside a governance document. It's also unsourced. CFA Institute cites no evidence for it, and this piece could find none that tests it directly.
What the components are, and what each one is actually for
The table below sets out the elements that recur across CFA Institute's specification and the two published institutional policies discussed here, with the function each performs.
| Component | What it typically states | The purpose it serves |
|---|---|---|
| Scope | Which accounts and assets the document governs | Stops the policy being quietly evaded by moving money to an account it doesn't cover |
| Purpose of the assets | What the money is for, and by when | Turns an abstract loss into a concrete question about a specific obligation |
| Return or spending requirement | The real return needed, and the withdrawal rate it supports | Anchors the portfolio to a liability rather than to a benchmark |
| Risk tolerance | An absolute loss threshold. CFA Institute's sample wording names a 33% loss over any 12 months as intolerable | Fixes the number in calm conditions, so a crash tests a pre-existing threshold instead of setting one |
| Permitted asset classes | What may be held, and what may not | Rules out the instrument that would otherwise get discovered at the worst moment |
| Targets and ranges | A weight per asset class plus an allowable band around it | The operative clause: it defines the drift that requires action and, by implication, the drift that doesn't |
| Benchmarks | An index per asset class | Makes "this is doing badly" a measurable statement rather than a feeling |
| Rebalancing process | Trigger, timing, and minimum trade size | Moves the buy-low decision out of the moment when buying feels hardest |
| Constraints | Liquidity needs, tax treatment, leverage limits, exclusions | Records the reasons a holding exists that a future reader would otherwise forget |
| Governance | Who decides, who executes, who monitors | In institutions, creates accountability; for a household investor all three roles are one person |
| Review cadence | How often the document itself is revisited — CFA Institute's sample wording specifies no less frequently than annually | Separates a considered change from a reaction |
| Acceptance | Signature and date | Establishes when the decision was made, and by whom |
CalPERS runs 120 pages, and the part that constrains behaviour is one table
The largest US public pension fund publishes its policy in full. The CalPERS Total Fund Investment Policy, effective 10 June 2024, runs 120 pages. It covers derivatives, counterparty risk, securities lending, divestment and the role of board consultants. It sets out ten numbered investment beliefs in an appendix and is explicit that these "are not a checklist to be applied by rote to every decision."
Appendix 4 gives the strategic allocation: 37% total public equity, 28% income, 17% private equity, 15% real assets and 8% private debt, with total asset exposure of 105% funded by 5% strategic leverage. Beside each target sits a range — plus or minus 7 percentage points for public equity, 6 for income, 5 for each of the others.
Those bands are the pre-commitment. They state in advance how far the portfolio may wander before anyone is obliged to act. Equally, a 5-point move in public equity obliges nobody to do anything. The breach rule is stated too: "If an asset class allocation exceeds the policy range, staff shall return the asset allocation to within its policy range in a timely manner, with the exact time period primarily dependent on transaction costs and liquidity ..."
Norway's fund wrote one sentence, and in March 2020 it forced the fund to buy
The clearest published case of a policy clause overriding an instinct belongs to Norway's Government Pension Fund Global. Its benchmark, set by the Ministry of Finance, has been 70% equity and 30% fixed income since 1 May 2019. The rebalancing rule is a single sentence. Rebalancing takes place when the equity share in the actual benchmark index differs by more than 2 percentage points from the strategic equity weight on the last trading day of the month.
In March 2020 equity prices fell far enough that, in Norges Bank Investment Management's own account to the Ministry, "the equity share of the fund's benchmark index" fell "to 65.8 percent at the end of March. This triggered a rebalancing." The equity share "was subsequently adjusted gradually back to 70 percent", NBIM reported that the arrangement "functioned well and as intended". The share was back at 70% by the end of the second quarter.
No committee voted on whether the fall was over. A 2-percentage-point threshold agreed the year before meant the world's largest sovereign wealth fund bought equities in the worst month since 2008 without anybody having to be brave on the day.
The evidence that pre-commitment works is strong, and almost none of it is about investing
The behavioural literature behind this idea is substantial. Gollwitzer and Sheeran's 2006 meta-analysis of implementation intentions — plans of the form "if situation Y is encountered, then I will initiate goal-directed behaviour X" — pooled 94 independent tests. It found "a positive effect of medium-to-large magnitude (d = .65) on goal attainment". That is a large effect by the standards of applied psychology. The mechanism is specificity. Naming the trigger in advance means the behaviour no longer depends on noticing and deciding at the same time.
The nearest thing to a financial test is Ashraf, Karlan and Yin's randomised study of a commitment savings account at a Philippine bank. It was published in the Quarterly Journal of Economics in 2006. Clients offered an account that restricted access to their own money saw average balances rise by 81 percentage points relative to control after twelve months. That's an intent-to-treat effect, measured across everyone offered rather than the 28.4% who accepted. The authors concluded that the effect "represents a lasting change in savings, and not merely a short-term response to a new product."
Neither study is about selling equities in a crash. The inference from there to a document you wrote yourself and can edit at any time is a long one. It deserves to be labelled as an inference.
The case against: a document nobody enforces is not a commitment device
The serious objection to all of this is simple. A commitment device works because breaking it is costly. The Philippine savings account physically withheld the money. An implementation intention fires automatically. A one-page IPS in a drawer has neither property. The person bound by it is the person who can rewrite it, at any hour, with no counterparty and no penalty.
That objection has support inside the documents themselves. Even the legally weighty version carries an escape hatch. The Department of Labor's bulletin states that where "compliance with the guidelines in a given instance would be imprudent, then the investment manager's failure to follow the guidelines would not violate ERISA § 404(a)(1)(D)". CalPERS' policy provides that allocations "may temporarily deviate from policy ranges due to extreme market volatility or to accommodate contributions, distributions, or other short-term cash needs". Part of that is plumbing: cash moves in and out. The rest is a discretion to leave a breach open during a fall. Two of the most heavily governed pools of capital in the world each reserve the right to depart from their own written policy in exactly the conditions the policy exists to handle.
The take-up data is bleaker. In the Philippine experiment, 710 clients were offered the commitment account and 202 opened it — 28.4%. Seven in ten people shown a device that would genuinely have bound them declined to be bound. Self-reported adherence to a private document is weaker evidence than that, because nobody audits it.
And CFA Institute, whose specification is the source of most retail IPS templates, is dismissive of the format: "Templates that purport to offer convenience and ease in development of an IPS almost inevitably sacrifice consideration of factors that are highly relevant to the investor." The body that wrote the list does not regard a one-page version of it as the same thing.
Most investors did not panic in the last two crashes anyway
Fear of a panicked crash decision is far more widely felt than acted upon. Vanguard's How America Saves 2021, covering 4.7 million defined-contribution participants, reported that in 2020 — a year in which 11% of trading days moved the S&P 500 by more than 3%, against nearly 17% of days in 2008 — 10% of participants initiated a trade and 90% didn't. Across the whole book, a net 3% of assets moved to fixed income.
The chart shows that trading series over five years: 8%, 8%, 8%, 7%, then 10% in the crash year. Those count participant-directed trades; adding managed-account holders, whose advisers rebalance for them, lifts Vanguard's 2020 figure to 16%. Vanguard's preview of How America Saves 2026 puts 2025 at 5% of non-advised participants, in line with a record low set in 2024. That is despite what it describes as increased volatility in the spring.
Only 4% of participants holding a single target-date fund traded in 2020, against 10% overall. A target-date fund is a pre-commitment device that somebody else enforces. The rebalancing happens without the holder deciding anything. The lowest-trading group is the group whose policy was executed by a third party. That is evidence for the mechanism, and simultaneously evidence that self-enforcement is the weak link.
Aggregate flow data is noisier and cuts the same way. ICI's 2021 Fact Book records $834 billion of March 2020 inflows to government money market funds and $139 billion of outflows from prime money market funds, 17.6% of their February assets. Much of that was institutions building liquidity rather than households capitulating. ICI attributes part of the year's fund flows to "portfolio rebalancing" — the opposite of panic.
What the institutional version does that a household version cannot
The strongest reason to be careful with the analogy is that institutional policies exist largely for reasons unconnected to self-control. CalPERS states its purpose plainly: the policies exist so that the committee, staff, consultants, managers and beneficiaries "clearly understand the objectives and policies of the CalPERS investment program". They are "intended to ensure that the Committee is fulfilling its fiduciary responsibilities". The document is a control on agents. It tells an external manager what they may not do, and creates a record if they do it anyway.
A household investor has no agents. A large share of CFA Institute's 16 elements — governance, delegation, standard of care, proxy voting, reporting accountabilities — collapses to nothing when the investor, the trustee and the manager are the same person. What survives is a smaller behavioural core: purpose, risk threshold, targets and ranges, and a rebalancing trigger. That core is a real thing, but it isn't the document CFA Institute specified. Calling a one-page version "what institutions do" overstates the inheritance considerably.
What would change the conclusion
A randomised trial of written policies among retail investors. None exists that this piece could locate. If one showed that investors who wrote and dated a target allocation traded less in the following drawdown than a matched control, the case would move from inferential to direct. If it showed no difference, the practice would rest on a plausible story and an unsourced sentence in a 2010 standards document.
Evidence on whether the bands survive contact with a bear market. The interesting number isn't how many people write a policy but how many amend it during a fall. CalPERS and Norges Bank publish enough to check; individuals don't. A brokerage-level study of stated target allocations against realised trading in March 2020 would settle a great deal.
A crash that lasts. March 2020 recovered in months, which flatters every rule that says do nothing. The relevant test is a fall that takes years to come back, as several have. The record is set out in our piece on how long bear markets last. A pre-commitment that held for six weeks hasn't really been tested.
The limits are worth stating plainly. Two of the primary sources are single institutions, not samples. The Vanguard data covers one recordkeeper's participants, who are disproportionately auto-enrolled and defaulted into target-date funds, and it counts trades rather than the intentions behind them. A trade in a crash can be a rebalance, and many were. The commitment-device research comes from savings and health behaviour, over horizons far shorter than an investing life. The CFA Institute and Department of Labor documents are specifications and interpretations, not findings. Neither reports an outcome.
What the evidence does support is narrower than the usual claim and still useful. The number that matters in a fall is how far the portfolio has moved from the weights chosen when nothing was wrong. That figure exists whether or not anybody wrote it down, and LedgerTouch keeps it current where a spreadsheet computes it once a year. The value of having decided in advance isn't that a document holds power over anyone. It's that the number arrives with an answer already attached, written by a calmer author.