Product × Finance × AI: Financial Products in the Age of AIPart Three Deriving a product中文
Chapter 20

One person and one sum of moneyWhat a product promises

A thirty-year-old villager brings his savings to a scribe. He wants to build a new house in five years, he wants to stop working the fields after sixty, and he wants to keep some grain aside for emergencies. The three have different deadlines and can bear different ups and downs: the grain for the house must be there in five years, while the grain for retirement can wait thirty. He has a harvest every year, and his future labour is itself a sizeable part of his wealth. The scribe must first find out what he wants and how much he can bear before deciding how to arrange things. In the end it must answer one question: what can it promise him? Time, systematic risk and irreversible losses do not disappear because of the scribe; it can promise only the part it can actually deliver.

1. What he wants

This villager cannot say "I want such-and-such an annual return". What he can name are a few concrete things: a house in five years, retirement in thirty, and a little kept at hand for emergencies. Each has an amount and a deadline, and being late or short on either has a cost. The first thing a product must do is not recommend ship shares but write these things down one by one: how much is needed, when, and what happens if it comes a little late or a little short.

There is one new constraint: the user has his own goals, deadlines and capacity to bear risk, and the product must make arrangements for this one person and this one sum of money.

Today's counterpart. Financial planning usually begins by setting out the user's goals: education, buying a home, retirement and an emergency reserve, each with an amount and a deadline.

Three goals, three deadlines

This villager's three goals laid out over his life. Pick a goal to see how much it needs, when, and what happens if it comes late or short.

Illustrative: amounts are measured in years of income.

2. How much he can bear

Bearing risk has two meanings. One is capacity: how steady his harvest is, how many mouths he has to feed, how long until each goal falls due. The steadier the harvest and the longer the horizon, the larger the swings he can bear. The other is willingness: faced with the same swings, some people sleep soundly and some do not. With little capacity but high willingness, he will take risks he should not; with ample capacity but low willingness, he will miss compensation he could have earned. The scribe must ask about the two separately and go by whichever is lower.

Today's counterpart. The suitability rule of the Financial Industry Regulatory Authority in the United States lists what a broker must learn about a customer before making a recommendation, including age, other investments, financial situation and needs, tax status, investment objectives, investment experience, investment time horizon, liquidity needs and risk tolerance.

Capacity and willingness: take the lower

The horizontal axis is capacity to bear risk (how steady the harvest is, how long the horizon); the vertical axis is willingness to bear risk. Drag the two sliders to see the risk level the scribe adopts.

Illustrative: each has five levels, and the risk level adopted is the lower of the two; the number in each cell is the level adopted.

3. How much risk to take

Suppose there are only two places to put grain: the granary, with almost no risk and a very low return, and ship shares, with a higher return and larger swings. Each extra portion in ship shares brings more expected return and more swings. The more this villager dislikes swings, or the larger the swings of ship shares, the lower the share he puts in ship shares; the higher the extra return of ship shares, the higher that share. Followed through to the end, the share depends only on these three things, and not on how much grain he has or how old he is.

Today's counterpart. The result of a risk-assessment questionnaire usually corresponds to a range for the split between stocks and bonds.

How much to put in ship shares

Drag the three sliders and see how the optimal share in ship shares changes.

Share = extra return ÷ (aversion coefficient × volatility²). Above 100% it means borrowing to buy; the figure caps it at 100% (Merton, 1969). The values of the aversion coefficient are illustrative.

4. A person is an asset too

The previous step left something out: this villager's harvests over the next thirty years. As long as his harvest is steady, it is like a bond that pays interest every year: predictable, and far larger than his current savings. Count it as part of his wealth, and he in fact holds a great deal of "safe assets", so his savings can bear more swings. As he grows older, fewer future harvests remain, and the share of his savings bearing swings should fall accordingly. If he can also work a few more years and stop later, he has room to manoeuvre in bad years and can bear even more. Conversely, if his harvest itself rises and falls with the fleet, for example because he works on the ships, his future income is less like a bond and more like ship shares, and his savings should be kept safer.

Today's counterpart. Target-date funds gradually lower their share of stocks as the target date approaches.

Counting future harvests as wealth

Drag his age to see the present value of his future harvests and his savings, and the share of savings that should go into ship shares; then switch whether his harvest rises and falls with the fleet.

Illustrative: annual income is 1, he works until 65, and future income is discounted at 3%; from age 25 he saves 12% of income each year, growing at 5%. Present value and savings are in years of income. The Merton ratio is taken as 50%: of future income and savings combined, half in total bears the risk of ship shares. With a steady harvest, future income is treated as a risk-free asset; when it moves with the fleet, 60% of it is treated as bearing ship-share risk (following the reasoning of Bodie, Merton and Samuelson, 1992).

5. One account for each goal

Combining the three goals and computing a single share creates a dilemma: set by the retirement horizon, the house money takes too much risk; set by the house horizon, the retirement money is too conservative. The more natural approach is the one people already use in their heads: keep a separate account for each goal. The emergency grain goes in the granary, where it can be drawn at any time. The house grain is needed in five years, so most of it goes somewhere safe. The retirement grain has thirty years, so more of it can go into ship shares. The risk of each account is measured by the standard this goal cares about most: how likely it is to fall far short when it comes due.

Today's counterpart. Some retirement accounts and education savings accounts receive separate tax treatment, with their uses and withdrawal conditions also set separately.

One share, or three accounts

Switch between two arrangements and see how likely the house and retirement goals are to fall far short when they come due.

Illustrative: ship shares return 6% a year on average with 16% volatility; safe holdings return 2% with 3% volatility. The house money is needed in five years, the retirement money in thirty. "Falling far short" means having less than 80% of the expected amount calculated at 2% when the goal comes due. Probabilities are computed from two thousand runs of years.

6. What can be promised

At this point it is possible to say what the scribe can promise him. It cannot promise that the grain for the house will certainly be enough, or that the grain for retirement will certainly be plentiful: time will not grow shorter, and storms that strike the whole village will still come. What it can promise is the part it actually controls: to hold the agreed allocation, to deposit and rebalance on time, to keep fees within the agreed ceiling and the risk of each account within the agreed limit, and to tell him truthfully how likely each goal is to be met and what the worst run of years would look like.

Such promises can be checked. The agreed allocation is written down, every step of execution is recorded, and the limits on fees and risk can be compared item by item. Whether the result is good depends on the run of years, but not on whether the scribe kept its word. What is promised is a process; what is delivered is a distribution of outcomes.

Today's counterpart. Asset management products usually promise an investment scope, investment proportions, fee rates and information disclosure, but not returns.

What can be promised, and what cannot

Pick an item to see why it can or cannot be promised.

Can be promised, and checked

Cannot be promised

7. One household's agreement: Civilization Simulator, version 20

From this version on, the simulator looks at a single villager. You can set his age, whether his harvest is steady, how strongly he dislikes swings, and how much he can save each year, then see the allocation the scribe gives each of the three accounts and, across two thousand runs of years, how likely each goal is to be met on time.

Civilization Simulator · Chapter 20: one household's agreement

Set this villager's age, whether his harvest is steady, how strongly he dislikes swings, and how much he can save each year. The scribe gives an allocation and annual deposits for each of the three accounts, and calculates across two thousand runs of years how likely each goal is to be met on time.

What did not changeProduct: what to promiseProduct: mandate and authority · Ch. 21
Illustrative model: income is 1 a year, and savings equal 1.5 years of income. The emergency account holds half a year's income in the granary. The house account starts with 0.6 years of income and needs 1.2 years of income in five years. The retirement account starts with 0.4 years of income and needs eight years of income at age 60. Ship shares return 6% a year on average with 16% volatility; safe holdings return 2% with 3% volatility. The house account's share in ship shares is the smaller of the Merton ratio and "horizon ÷ 25". The retirement account uses the share obtained after counting future income as wealth, capped at 100%; when the harvest moves with the fleet, it is reduced on the basis that 60% of future income bears ship-share risk. In the first five years, deposits first cover what the house account needs, calculated at 2%, and the rest go to the retirement account; after year five, everything goes to the retirement account. "Met on time" means having at least the required amount when due.

8. What this step leaves open

Once the product knows what to promise, new questions follow:

  1. He cannot always be present. Once the agreement is set, the scribe must keep acting on his behalf. How detailed must the agreement be, and how much authority should it grant?
  2. He cannot see the process. How can he understand and check whether the scribe acts according to the agreement?
  3. Agreements go out of date. His harvest, family and goals will change. Who updates the agreement, and when?
  4. Agents make mistakes. If the limits of the promise are breached, who bears the consequences?

What this step leaves open

Four problems that remain unsolved once the product knows what to promise. Pick a problem to see the assumption it hides.

One household's agreementPromise the process, not the outcome

9. Implications for products

The first part an agent-centred wealth product must grow is "what to promise". Three implications follow. First, start from the user's goals, not from the product: first write down how much is needed, when, and what happens if it comes late or short, then decide how to arrange things. Second, assess capacity and willingness to bear risk separately and go by the lower; count future labour income as part of the user's wealth, so that the portfolio changes with age and occupation. Third, promise the process, not the outcome: the allocation, execution, fees and risk limits are all written clearly and can be checked, and the user is told truthfully how likely each goal is to be met.

Chapter summary: one person and one sum of money

Four conclusions, each tied to one mechanism in this chapter.

GoalsFirst write down what he wants

The amount and deadline of each goal, and what happens if it comes late or short.

Merton ratio and human capitalHow much risk to take

It depends on compensation, swings and aversion to them; future harvests are part of his wealth too.

Accounts by goalOne account for each goal

Each account chooses its allocation by its own horizon; its risk is the chance of falling far short when due.

PromisesPromise the process, not the outcome

Allocation, execution, fees and risk limits can be checked; the outcome is a distribution.

References