Retirement plans are built on an expected return. Once withdrawals begin, the order those returns arrive in matters separately from what they average to - and the plan that specifies only the average has not answered the question that decides whether the money lasts.

What follows isolates that mechanism on identical inputs, then sets out the two documented responses to it.

Identical Returns, Reordered, Cost Rs 5.25 Lakh

Two retirees start with ₹1 crore each and withdraw ₹5 lakh at the start of every year. Both see the same three annual returns - -30%, +10%, +20%. Only the order differs.

The Withdrawal, Not the Bad Year, Creates the Gap

Run the same three returns with no withdrawals and the gap vanishes entirely: 0.7 × 1.1 × 1.2 equals 1.2 × 1.1 × 0.7. Compounding alone is multiplication, and multiplication does not care about order.

A withdrawal breaks that commutativity. It locks in whatever the balance happens to be at that moment and removes it permanently, before any subsequent recovery can apply to it. The -30% year is not the problem by itself - retiree B lived through the identical year. The problem is which withdrawals had already happened when it arrived.

The Exposure Is Concentrated Around One Date

This is why the danger window in retirement-income research is not “whenever markets fall.” The commonly cited red zone runs roughly five years before retirement through five to ten years after, and industry work on the effect finds returns five years into retirement move the outcome about as much as returns ten years before it.

Outside that window a bad year is absorbed - by a portfolio not yet being drawn down, or by one that has built enough cushion to withstand it. The risk is not spread across the retirement; it is concentrated at its start.

Two Documented Responses Treat It as Scheduling

Guyton and Klinger’s 2006 guardrail rules cut spending by a fixed amount - commonly modelled around 10% - once a bad stretch pushes the current withdrawal rate more than 20% above the starting rate, with an increase under the mirror condition. The value is in replacing a discretionary decision with a mechanical one at the moment discretion is least reliable.

Pfau and Kitces’s 2013 work went further, testing a rising rather than declining equity glide path through retirement - starting around 20-40% equity and increasing - and found it reduced both the odds and the size of failure, because it under-exposes the account during precisely the years a crash does permanent rather than temporary damage.

What This Means for Allocators

Model the sequence, not only the average. A plan specified as an expected return has not been tested against the variable that decides the outcome.

Write the spending rule before the drawdown, not during it. That is the entire mechanism of the guardrail approach and the reason it works.

And treat the years either side of the retirement date as a distinct risk regime. The ₹1 crore figures above are illustrative arithmetic rather than a forecast - the direction holds across the published research.

Back to Writing