Ordinary Monte Carlo treats your expected-return inputs as known and only randomizes the year-to-year noise around them. But the expected return itself is the least certain number in the whole plan.
With parameter uncertainty on, each simulated path first draws its own version of the expected returns. A constant per-asset offset is sized by the standard error you allow, and the path then lives its whole life under that market. Some paths inhabit a world where equities genuinely earn a point less than you assumed; others a point more.
What you'll see
- The fan widens, mostly in the tails: bad-market and bad-assumption paths now stack, which is exactly the risk a point-estimate plan hides.
- The median barely moves. This isn't pessimism; it's honesty about how precisely a mean return can be known.
A success rate quoted to one decimal place implies the inputs are exact. They aren't: even with decades of data, the standard error on an equity mean is on the order of whole percentage points. This switch prices that in; the offsets are drawn from their own seeded stream, so comparisons stay fair.