Stress-testing a retirement plan

Most retirement projections show a single line reaching a comfortable conclusion. The valuable work is establishing what range of outcomes is possible and which assumptions the result depends on.

The problem with single projections

A projection using an average return produces a smooth curve that will not occur.

Actual returns arrive in an order, and as discussed elsewhere on this site, the order matters enormously once withdrawals begin.

A plan that works with a constant six per cent return may fail with an average six per cent return delivered in a poor sequence.

Monte Carlo analysis and its limits

The standard tool, which runs many simulations using randomised returns and reports the proportion in which the plan succeeded.

It is a genuine improvement on a single line, and it has limitations worth understanding.

Results depend heavily on the assumed return and volatility inputs, which are generally derived from historical data that may not repeat.

Many implementations assume returns are independent between years, which does not match historical behaviour.

And the headline output — a success probability — invites false precision. A plan reported as having an eighty-five per cent success rate is not meaningfully different from one at eighty-two.

The useful interpretation is directional: a plan in the nineties has substantial margin, one in the fifties does not, and small differences within the range are noise.

The scenarios worth testing specifically

More useful than a probability figure, because each identifies a specific vulnerability.

Poor returns in the first five years. A decline of thirty per cent early, followed by normal returns. This tests sequence risk directly.

Higher inflation. Run the plan at two percentage points above the assumed rate. Plans that work at two per cent and fail at four are common.

Living to ninety-five or a hundred. Extending the horizon reveals whether the plan depends on dying on schedule.

Long-term care for one spouse for three years. At realistic local rates.

Early forced retirement. Plans assuming work until sixty-seven should be tested against stopping at sixty-two, since a substantial share of retirees leave earlier than planned.

Death of one spouse early. Which reduces Social Security income, may end a pension, and changes the tax filing status to less favourable single thresholds.

This scenario is frequently omitted and is one of the more likely to occur.

Reduced Social Security benefits, for those with a long horizon.

A significant unplanned expense — supporting an adult child, a major home repair, a health event.

What to do with the results

The point is not to build a plan that survives every scenario, which would require saving an implausible amount.

The point is to identify which scenarios cause failure and what the response would be.

For each vulnerability, the useful question is: what would we do?

Reduce spending by how much. Sell the house. Take a part-time job. Move somewhere cheaper. Draw on home equity.

A plan with identified responses to identified risks is robust. A plan with a high success probability and no contingencies is fragile in a way the number conceals.

The flexibility test

The single most informative question.

How much could spending be reduced without genuine hardship?

A household where every dollar is committed to fixed costs has no capacity to respond. One where a meaningful share is discretionary can absorb a bad decade.

Modelling consistently shows that willingness to adjust spending is worth more than almost any other change to a plan.

How often

Annually, briefly, and after any significant change.

The annual review compares actual spending and portfolio value against the plan, adjusts assumptions, and checks whether the withdrawal rate has drifted.

Small adjustments made early are considerably less painful than large ones made late, and the only way to make them early is to look.

General information only, not financial advice. Projections rely on assumptions that may not hold — consult a qualified adviser about your own situation.