The user is not presentMandate parameters and authority
Once the agreement was settled, the villager went back to his fields. He has only a few evenings a year to look at the accounts, yet the scribe has work to do every day: making deposits, adjusting holdings, deciding after a storm whether to sell, deciding when he falls ill whether to touch the house money. If it asks him about everything, he is worn down and the scribe is of no use; if it never asks, he loses control of his own property. The agreement cannot cover every case, and who decides the cases it leaves out must be made clear in advance. The agreement must also state goals, boundaries, rules, exceptions and term; authority must be graded by how large a matter is and whether it can be undone; and when the scribe must come back and ask him must be written in as well.
1. He cannot be present all the time
Every day the scribe may meet something that calls for a decision. If it asks him about each one, he will be interrupted more than a hundred times a year and will soon stop reading, stop answering, or nod without thinking. If it asks about none, every decision it makes for him rests only on its own judgment. Between the two lies a sensible position: the scribe handles small, routine matters itself under the agreement, and brings the few important ones back to him.
There is one new constraint: the user cannot be present all the time, so continuing delegation depends on an agreement made in advance.
Today's counterpart. In a discretionary account, the client gives written authorization in advance, and the fiduciary decides purchases and sales on its own within the agreed scope. In a non-discretionary account, every trade requires the client's confirmation.
How much to ask
The scribe makes 200 decisions a year; 20 of them matter and depend on what he cares about. He can answer only 12 requests a year with real care; beyond that he nods without thinking. Drag the share of decisions referred to him and see how many bad decisions result.
Illustrative: the scribe ranks matters by importance and refers the most important first; an important matter decided by the scribe alone goes against his wishes 40% of the time; the probability that he answers with care is 12 ÷ number of requests (capped at 1), and a careless answer is the same as leaving the decision to the scribe. Vertical axis: number of bad decisions per year.
2. The agreement cannot be complete
However detailed the agreement, it cannot foresee everything: the fleet changes its route, the village adopts a new rule, a child is born into his family, or something happens that nobody expected. Someone has to decide the cases the agreement does not cover. Work through who should hold that decision. Give it to the scribe, and it can adapt to circumstances, but it may not know what he cares about. Keep it with him, and he knows what he cares about, but he may not be there in time. The decision can go to only one side first, so it must be settled in advance which uncovered cases the scribe handles in the spirit of the agreement and which must wait for his return.
Today's counterpart. FINRA's rule on discretionary accounts requires that, before a fiduciary exercises discretion, the client has given prior written authorization and the firm has accepted the account in writing. Where the authority covers only the time and price of execution, it is generally valid only for that day.
Who decides the cases the agreement leaves out
Eight things that might happen over ten years. The agreement states only goals, boundaries and everyday rules. Tick "State the residual right to decide" to see how the uncovered cases are handled.
Covered by the agreementNot covered; the scribe acts in the spirit of the agreement and tells himNot covered; waits for his returnNot covered; the scribe decides on its own
Illustrative: the eight events are invented examples.
3. What the agreement should contain
That the agreement cannot be complete is no reason to leave it unwritten. Work out what the villager must write down at the least: what he wants to achieve (goals, namely the three accounts); what must not be crossed (boundaries: the share of ship shares in each account, the maximum loss, the fee ceiling, what may not be bought); how things are done in the ordinary course (rules: when to deposit, when to adjust, how far a deviation may go before holdings are brought back); when exceptions apply (which situations require stopping to ask him); and how long the agreement lasts and how it can be revoked. Goals and boundaries tell the scribe where to go and where it must not go; rules mean most everyday matters need no request; exceptions and term let him keep final control.
Today's counterpart. The elements of an investment policy statement that the CFA Institute sets out for individual investors include scope and purpose; governance (who drafts, implements and monitors the investment policy); return and risk objectives and constraints; and risk management and rebalancing.
The five parts of an agreement
Select a part of the agreement to see what the villager's agreement says and what happens without it.
Illustrative: the contents of the agreement are invented examples.
4. Grading authority
Not everything is worth asking him about, and not everything can go unasked. Sort matters by two questions: how large is this, and can a mistake be undone? Size can be measured by the amount involved as a share of the household's property; for example, no more than 2% counts as small, and once trust is established the threshold can be raised. Small and reversible matters, such as scheduled regular deposits or bringing holdings back to the agreed proportions, the scribe handles on its own and simply records afterwards. Large and reversible matters, such as a sizeable adjustment, it carries out and then promptly tells him about, and he can reverse them. Small and irreversible matters, such as selling a few ship shares to pay a small fee, it also carries out first within the scope of the agreement and promptly tells him about. Large and irreversible matters, such as moving grain out of the granary, drawing early on the house or retirement money, or changing the agreement itself, require his own confirmation in advance. Anything outside the agreement's boundaries, such as buying a product leveraged with borrowed money, is forbidden whatever its size. This gives four levels: act on its own; act, then tell; confirm first; and forbidden.
Today's counterpart. SAE International's standard for levels of driving automation defines six levels, from 0 to 5, and specifies that the level depends on the features engaged at the time; the same vehicle can operate at different levels in different situations.
Grading by size and reversibility
Two questions place every matter in one of four levels. Drag "Trust established" to see how the boundary of acting on its own moves; select a cell to see examples.
Illustrative: "small" means no more than 2% of the household's property; as trust grows, the threshold for "small" is raised one step for each year the agreement is kept, to 5%, 10% and 25% in turn. Large and irreversible matters always require confirmation in advance, however high the trust.
5. Letting go and keeping control
The agreement says "ask him about the big things", but he knows far less than the scribe: the scribe watches the fleet and the accounts every day, while he looks only a few times a year. When the scribe proposes a plan, he can usually only nod. The formal right to decide stays with him, but more and more of the real decision passes to the scribe. Letting go has benefits: the scribe will look for opportunities and deal with problems more actively. The cost is that he loses control. The less time he spends understanding the situation, the more pronounced the shift becomes.
Today's counterpart. The choice between a discretionary and a non-discretionary account is, at bottom, a trade-off between initiative and control.
He decides on paper; who decides in fact?
Each year the scribe puts 10 important proposals to him, 2 of which in fact go against his wishes. Drag the time he spends each year understanding the situation, and see how many he catches and who holds the real decision.
Illustrative: the probability that he catches a proposal that goes against his wishes is time ÷ (time + 10 hours); proposals he does not catch, he approves all the same (the distinction between formal and real authority drawn by Aghion and Tirole, 1997).
6. When to come back and ask him
The key to letting go while keeping control is to write into the agreement when the scribe must come back and ask him, rather than leaving it wholly to the scribe's judgment. There can be three kinds of trigger. The first is quantitative: an account has fallen by more than the agreed amount, or has drifted too far from its target. The second is an event: he changes jobs, a child is born, someone falls ill. The third is time: a scheduled review once a year, even if nothing has happened. Once a trigger fires, the scribe must set out the situation, the options and its recommendation, and he decides.
Today's counterpart. Among the elements of an investment policy statement listed by the CFA Institute, the governance section requires stating how the investment policy is reviewed and updated.
When to come back and ask him
Eight events over ten years, laid out on a timeline. Tick the triggers written into the agreement and see which events make the scribe stop and ask him.
Illustrative: the eight events are invented examples. A "quiet drift" is a deviation that builds up over several years without any single large fall, for example when his income quietly declines and the agreed deposit rate gradually exceeds what he can afford.
7. Ten years in his absence: Civilization Simulator, version 21
The simulator unlocks "mandate parameters and authority". Over ten years the villager meets all kinds of events: storms, a change of job, a new child, illness, a new village rule. You can switch on or off the agreement's exception clauses, graded authority and scheduled reviews, and set how often he can respond to requests. Then see, over the ten years, which matters the scribe handled well on its own, which it should have asked about and did not, which it asked about to no effect, and how many times he was interrupted along the way.
Civilization Simulator · Chapter 21: ten years in his absence
Over ten years the scribe handles about twenty routine matters a year and from time to time meets storms, changes in his family, new situations the agreement does not cover, and quiet drifts. Switch the three rules on or off, set how many requests he can answer with care each year, and see the results.
Product: what to promiseProduct: mandate and authorityProduct: process records and verifiable evidence · Ch. 22
Illustrative model: about twenty routine matters a year; a 30% chance each year of a storm (a quantitative trigger, about which the scribe always asks); a 35% chance each year of a change in his family; a 30% chance each year of a new situation the agreement does not cover, half of them irreversible; a 20% chance each year of a quiet drift, which only a scheduled review can detect. Without graded authority, routine matters are also referred one by one; the probability that he answers with care is "requests he can answer with care per year ÷ requests per year" (capped at 1).
8. What this step leaves open
The agreement and the grading of authority let the scribe act while he is away, but he still cannot see what it has done:
He cannot read the accounts. Every action of the scribe is recorded, but he looks only a few times a year. Can he understand what he sees?
He has no way to check. The scribe says it kept to the agreement. Why should he believe it?
A nod is not consent. Does a nod given without understanding the situation count as genuine confirmation?
Agents make mistakes. Who bears the consequences of an error made within the scope of authority?
What this step leaves open
Four problems that remain once the agreement and authority let the scribe act while he is away. Select one to see the assumption it hides.
Ten years in his absenceAgreement and authority
9. Implications for products
For a wealth-management product built around an agent, the second part that has to grow is "mandate parameters and authority". Three implications follow. First, make the agreement part of the product: goals, boundaries, rules, exceptions and term, written in words the user understands and enforced by the system, rather than a mere disclaimer. Second, grade authority by the size of a matter and whether it can be undone: the agent handles small, reversible matters on its own and records them afterwards; large, irreversible ones require confirmation in advance; and the boundaries between levels can be adjusted as trust grows. Third, write into the agreement when to come back and ask the user: quantitative, event and time triggers; once a trigger fires, set out the situation, options and recommendation and let the user decide, so that the real decision does not pass entirely to the agent without anyone noticing.
Chapter summary: the user is not present
Four conclusions, each tied to one mechanism in this chapter.
Incomplete contractsUncovered cases need an owner
Settle in advance who holds the residual right to decide: reversible matters are handled in the spirit of the agreement; irreversible ones wait for his return.
The agreementGoals, boundaries, rules, exceptions and term
Most everyday matters need no request, and final control is kept.
Rerun: least privilegeFour levels by size and reversibility
Act on its own; act, then tell; confirm first; forbidden. The boundaries shift as trust grows.
Formal and real authorityWrite "when to come back and ask" into the agreement
Quantitative, event and time triggers keep the real decision from passing away unnoticed.
References
Aghion, P., & Tirole, J. (1997). Formal and real authority in organizations. Journal of Political Economy, 105(1), 1–29.
CFA Institute (2010). Elements of an Investment Policy Statement for Individual Investors. Charlottesville: CFA Institute.
FINRA. Rule 3260: Discretionary Accounts. Financial Industry Regulatory Authority Rulebook.
Grossman, S. J., & Hart, O. D. (1986). The costs and benefits of ownership: A theory of vertical and lateral integration. Journal of Political Economy, 94(4), 691–719.
Hart, O., & Moore, J. (1990). Property rights and the nature of the firm. Journal of Political Economy, 98(6), 1119–1158.
Nobel Prize Outreach. The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2016. NobelPrize.org.
SAE International (2021). Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles (J3016_202104).
Smith, R. E. (2012). A contemporary look at Saltzer and Schroeder's 1975 design principles. IEEE Security & Privacy, 10(6), 20–25.
Saltzer, J. H., & Schroeder, M. D. (1975). The protection of information in computer systems. Proceedings of the IEEE, 63(9), 1278–1308.