What does ignoring correlation cost you?
Independent draws are the default in almost every spreadsheet model, and they are almost always wrong. Real cost lines move together, because one wage settlement, one bad winter or one recession reaches all of them in the same year. This template runs the same four lines twice from the same ranges, so the only difference between the halves is the admission that they move together.
Words on this sheet
- Percentile: The value a stated share of the results came in under. The 90th percentile is the figure nine runs in ten stayed below.
Finance Advanced Monte Carlo Pro engine
After you install, this is the model to open.
What Does Ignoring Correlation Cost You?
- In your spreadsheet, click the Sortia icon in the strip of icons down the right-hand edge. No strip? Click the arrow at the bottom-right to open it. You can also use Extensions, then Sortia, then Open Sortia.
- Click Start from a template and put that name in the search box.
- Pick the card with that name and click Load this template. It arrives on a new tab with real numbers already in it.
This one runs on a Pro engine, and every free install includes five full-quality runs on your own numbers, shared across all five Pro engines rather than five for each. After that, Pro is $199/year.
The answer
The averages agree. Everything that matters for setting a reserve does not:
- Median
- unchanged about $10.1M either way
- Spread
- 1.65x wider once correlated
- P95
- +$1.8M $12.5M to $14.3M
- Reserve breach
- 15% on a reserve sold as 5%
The middle is the trap. Both halves have a mean near $10.2M and a median near $10.1M, so anyone checking the average would conclude correlation does not matter here. The tail says otherwise: standard deviation goes from $1.35M to $2.23M, P95 moves from $12.5M to $14.3M, and P99 from $13.6M to $15.9M. Then the cell that turns it into money: hold $12.6M in reserve, which is exactly the P95 the independent run told you to hold, and the correlated run breaches it on about 15% of trials. A reserve signed off as a one-in-twenty risk is running closer to one in seven, and not one of the four cost lines changed. Prove the halves are otherwise identical by deleting the six correlation pairs and running again: the two totals then land within about a percent at P95 and the breach rate falls to about 5%, which is what the reserve was supposed to buy.
The model
Four sites of one business, each with an uncertain annual loss cost, written into the sheet twice. The upper block draws them independently. The lower block uses identical PERT ranges plus six correlation pairs at +0.6, which is what a single shared driver looks like. One click runs both.
| Four sites, upper block | drawn independently |
| Four sites, lower block | identical ranges, six pairs at +0.6 |
| Trials | 20,000 |
| Reserve under test | $12,600,000, the independent P95 |
| Breach flag | 1 on any trial where the correlated total exceeds the reserve |
Once it is in your sheet
- The model arrives with real numbers in it and runs as it stands, so you can press the button first and understand it second.
- Change the numbers to yours. The sheet marks which cells are inputs and which hold formulas, and most labels carry a note explaining the row.
- Press the run button at the bottom of the panel. It is labeled for the tool you are in, and the result lands on its own tab, with a written reading of it beside the figures.
Never used Google Sheets? Start here goes the whole way, in seven steps, and assumes nothing.
Next question
- Do your two holdings actually diversify each other?How Two Assets Move Together
- Will you run out of cash this quarter?Will the Cash Last Thirteen Weeks?
- What are the odds you breach the covenant?How Close Is the Covenant?
- Every line looks fine. Will the year still go over?Will the Budget Hold?
- How far could next year miss the revenue plan?What Will Next Year's Revenue Be, as a Range?
- How much of your risk is just the exchange rate?What Does the Exchange Rate Do to Profit?
Every model like this one, and the method behind them: Monte Carlo simulation.