Monte Carlo Simulation — Asymmetrica

What it is

A Monte Carlo simulation does not predict the future. It admits that the future is unknowable and instead asks: given what we know about how returns behave, what range of outcomes is plausible?

The method works by running the same scenario thousands of times — each run with a slightly different sequence of randomly generated returns drawn from a realistic distribution. The result is not one answer but a spread of answers: a 10th-percentile outcome (the unlucky retiree), a 50th-percentile outcome (the median retiree), and a 90th-percentile outcome (the lucky one).

In retirement planning, a typical simulation runs 10,000 scenarios and asks: in how many did the portfolio survive 30 years without hitting zero? If 8,200 out of 10,000 did, the “success rate” is 82%.

Why it matters more than a spreadsheet projection

Most retirement calculators assume a fixed return every year — say, 10% equity returns compounding steadily. That is not how markets work. Markets lurch. They crash 40% in one year and recover 60% the next. A fixed-return projection misses all the ways a bad sequence can destroy a plan even when the long-run average looks fine.

Monte Carlo captures this by sampling from the full distribution of historical returns, including the tails — the crashes and the booms — not just the average.

A retiree who earns exactly 10% every year for 30 years is in a different universe from one who earns an average of 10% but saw −38% in year 1. The first lives comfortably; the second may run out of money. Monte Carlo shows you both.

How a basic simulation is built

For each of 10,000 trials:
  1. Start with corpus = ₹2 crore
  2. For each year (1 to 30):
       a. Draw a random return from a normal distribution
          (mean = 8%, std dev = 18% — calibrated to Nifty 50 history)
       b. Apply return: corpus = corpus × (1 + return)
       c. Subtract annual withdrawal (e.g. ₹6 lakh, inflation-adjusted)
       d. If corpus < 0: mark as FAILURE, stop
  3. Record: did the corpus survive all 30 years?

Success rate = (trials that survived) / 10,000

The output is typically shown as a fan chart — the middle band is the median outcome, the outer bands show the 10th and 90th percentiles.

The India calibration problem

Most Monte Carlo tools you find online (cFIREsim, FICalc) are calibrated to US markets: S&P 500 returns, US inflation (CPI ~2-3%), and US bond yields. Plugging in these numbers for an Indian retirement plan gives dangerously optimistic results.

India-specific inputs that change the simulation materially:

Input	US assumption	India reality
Equity mean return	7–10% real	6–8% real (Nifty 50, since 1990)
Equity volatility (std dev)	~15%	~22–24%
Inflation	2–3%	5–7% (CPI, long-run)
Bond/debt return	3–5% real	1–3% real
Safe withdrawal rate	4%	~2.5–3.5% (India-calibrated)

The higher volatility and higher inflation both reduce the safe withdrawal rate significantly compared to US numbers.

Reading the output: what “82% success” actually means

An 82% success rate means that in 1,800 out of 10,000 simulated retirements, the money ran out before 30 years. Whether that is acceptable depends on your risk tolerance and what “failure” means in practice — it does not mean starvation; it means corpus hit zero and spending had to be cut.

Most financial planners target 85–95% success. The tradeoff: pushing success from 80% to 95% often requires reducing withdrawal by 15–20%, which means working longer or spending less in retirement.

Key takeaways

Monte Carlo shows a range of futures, not a single prediction
A plan that looks fine on average can still fail 20–30% of the time due to sequence risk
India-calibrated simulations require higher inflation (6–7%) and higher equity volatility (22–24%)
The output — “X% success rate” — is a probability, not a guarantee
Use Monte Carlo to stress-test your plan, not to find false precision