Chapter 2 of 3 · 11 min

What a loss costs to get back

Fall 20% and you need 25% to get back; fall 50% and you need 100%. Turn that into years, and then add withdrawals, which make it far worse.

By the end of this chapter you can
  • Compute the gain a loss needs, and why it grows faster than the loss
  • Turn a recovery into years at a steady return
  • Show what withdrawals do to the recovery
  • Find the largest fall a recovery horizon allows, and scale it to a mix
1

The intuition

A client's $100,000 falls 20% to $80,000. Getting back to $100,000 is a 25% gain, not 20%, because it is earned on the smaller balance. Fall 50% and the gain needed is 100%. The deeper the hole, the faster the required gain grows, which is the most useful thing to show a client when setting risk: a drawdown is not a number, it is years of waiting.

At a steady 6% a 20% fall takes about 3.8 years to recover. Add a retiree taking 4% of the balance out every year and the net growth is only 1.76% a year, so the same fall takes nearly thirteen years; take out more than the return and it never recovers at all. Money withdrawn during the recovery never takes part in it, which is why the same drawdown that is an inconvenience for a saver can be permanent for a retiree.

The key idea

Gain needed after a loss d = d ÷ (1 − d). Years to recover at return g = ln(1 ÷ (1 − d)) ÷ ln(1 + g). With withdrawals of w a year: replace (1 + g) with (1 + g)(1 − w). Largest loss recoverable in Y years = 1 − (1 + g)^−Y. A mixed portfolio's loss with bonds flat = equity weight × equity loss.

2

Why it works

  • The conventions here: a single drawdown followed by a constant annual return. Recovery means the pre-loss dollar value. Withdrawals are a fixed share of the current value at the end of each year, so the value grows by (1 + g)(1 − w) a year. Years are fractional.
  • Losses and gains are not symmetric. The gain is earned on what is left, so d ÷ (1 − d) rises faster than d. The arithmetic average of −50% and +50% is zero; the money is down 25%.
  • Years turn a percentage into a plan. Growth needed as a factor is 1 ÷ (1 − d); the years are its log over the log of the growth rate.
  • Withdrawals shrink the base every year. The net factor (1 + g)(1 − w) is what compounds; at or below one, the old value is never seen again, and the drawdown is a permanent cut to what the plan can pay.
  • Run it backwards for a limit. A client who must be whole within Y years can stand a fall of at most 1 − (1 + g)^−Y. Divide by a stress-test equity crash to turn that into an equity weight.
  • The geometric average is the one that describes what happened. After −50% and +50% it is √(0.5 × 1.5) − 1 ≈ −13.4% a year; the gap between it and the arithmetic average grows with volatility.
A 20% fall; 6% a year afterwards; a retiree withdrawing 4% of the balance; a 5-year horizon; a 60/40 mix in a 30% equity crash
Gain needed: 20% ÷ 80%25.00%
Years to recover: ln 1.25 ÷ ln 1.063.83
Net factor with withdrawals: 1.06 × 0.961.0176
Years with withdrawals: ln 1.25 ÷ ln 1.017612.8
Largest fall recoverable in 5 years: 1 − 1.06⁻⁵25.27%
60/40 in a 30% crash: loss 18%; gain needed 21.95%3.41 years to recover

Withdrawals more than tripled the recovery. Fall 50% instead of 20% and the saver needs a 100% gain and about 11.9 years at 6%.

3

The formulas

Gain needed = d ÷ (1 − d)

Earned on what is left, so it outgrows the loss.

Years to recover = ln(1 ÷ (1 − d)) ÷ ln(1 + g)

The log of the growth needed over the log of the growth rate.

With withdrawals: ln(1 ÷ (1 − d)) ÷ ln((1 + g)(1 − w))

The net factor is what compounds.

Largest loss recoverable in Y years = 1 − (1 + g)^−Y

The years formula run backwards.

Portfolio loss, bonds flat = equity weight × equity loss

Scale the crash to the mix.

4

Worked example

The growth needed as a factor is one over what is left. Take logs to turn the factor into years at the stated return.

Drawing the numbers…
5

See it move

Same client and the same return after the fall ("New company and numbers" changes it). Change the size of the fall, the withdrawal rate, the recovery horizon and the equity weight of the mix.

Drawing the numbers…
Try this
  • Raise the fall. The gain needed rises faster than the fall, and every recovery takes longer.
  • Raise the withdrawals. The net growth factor falls toward one, and nothing else moves; the slider stops a point below the return, because at the return the recovery never ends.
  • Lengthen the years the client can wait. The largest fall the horizon allows grows; nothing else moves.
  • Raise the equity weight of the mix. The mixed portfolio's loss and its recovery time rise toward the all-equity case.
6

Run it backwards

The client can wait a stated number of years and expects a stated return. What is the largest fall they can stand?

Drawing the numbers…

Set (1 − d)(1 + g)^Y = 1 and solve: d = 1 − (1 + g)^−Y. It is the years formula run backwards.

The follow-up turns the limit into an allocation by dividing by a stress-test equity fall: the same conversion the previous chapter used.

7

Traps

Saying a 20% fall needs a 20% gain.
It needs d ÷ (1 − d): 25%. The gain is earned on what is left.
Averaging returns arithmetically.
−50% then +50% is −25%, not zero. The geometric average describes the money.
Ignoring withdrawals in a retiree's recovery.
Compound (1 + g)(1 − w). At or below one, the old value never comes back.
Rounding years up before comparing.
The recipe's years are fractional; compare them as they are.
Applying the equity crash to the whole mix.
Bonds flat, the portfolio's loss is the equity weight times the equity loss; then the usual recovery formula.
8

Say it in the interview

The interviewer asks

How would you show a client what a big loss really costs?

Say yours out loud first, then compare.
9

Check yourself

4 fresh questions, with new numbers. Answer each one correctly to finish the chapter. Get one wrong and you will see the full working, then you can try it again with new numbers.

Answers within 1% are marked right. Type the number; $, %, x and M are fine. First tries count toward Learned: the topic is Learned once every chapter is done and 75% of first tries were right.

0 of 4
Drawing your questions…
Remember
  • Gain needed = d ÷ (1 − d); it outgrows the loss.
  • Years = ln(1 ÷ (1 − d)) ÷ ln(1 + g); with withdrawals, use (1 + g)(1 − w).
  • Withdrawals at or above the return make a drawdown permanent.
  • Largest recoverable fall = 1 − (1 + g)^−Y; divide by the crash for a weight.