Introduction to Monte Carlo Simulations

Long-term investing math has a flaw: it assumes consistent returns. Reality is volatile. Markets don't return exactly 8% every year. They return 15% one year, -5% the next, 12% the year after. The sequence and magnitude of returns matter enormously.

This is where Monte Carlo simulations enter. Named after the casino (where probability plays out in randomness), Monte Carlo simulations model investment outcomes by running thousands of scenarios, each with different randomly-sequenced returns drawn from historical distributions. The result is a picture of possible futures, not a single false certainty.

Quick definition: Monte Carlo simulation is a computational technique that models investment returns using random sequences drawn from historical return distributions, allowing investors to estimate the probability of different outcomes (success, failure, specific wealth targets) across thousands of scenarios.

Key Takeaways

1. Monte Carlo generates thousands of plausible futures, not just a single "average case"
2. Success probability is more meaningful than average outcome, because averages hide downside risks
3. Asset allocation (stock/bond split) is the primary risk lever, more impactful than stock picking
4. Withdrawal strategies can be stress-tested against all 10,000 simulated scenarios
5. Correlation between assets matters, because diversification works only if returns don't all crash together
6. Safe withdrawal rates depend on success probability, typically 90-95% across historical scenarios

The Limitation of Deterministic Models

A deterministic projection looks like this:

Your $500,000 portfolio at 7% annual return:

• Year 5: $701,277
• Year 10: $982,006
• Year 20: $1,926,689
• Year 30: $3,777,998

This assumes a perfectly smooth 7% year after year. But markets don't work that way. In reality:

• The S&P 500 has achieved positive returns in only 67 of the past 100 years (67% of the time)
• Worst year: -43% (2008)
• Best year: +54% (1954)
• Average return disguises this massive range

A deterministic projection is useful for understanding the "expected" outcome, but it's dangerously incomplete. It doesn't tell you the probability you'll actually achieve that outcome or what could go wrong.

How Monte Carlo Works: A Simplified Example

1. Historical data: Analyze 100 years of annual S&P 500 returns. Calculate mean (9.6%) and volatility (standard deviation: 18%).

2. Random sampling: For each "year" in the simulation, randomly draw a return from a distribution centered at 9.6% with standard deviation of 18%. The return might be -5%, 14%, 22%, -10%, etc.

3. Sequence: Build a 30-year sequence of these random returns. Year 1 might be -5%, Year 2 might be 14%, Year 3 might be 22%, etc.

4. Portfolio value: Starting with $500,000, apply these returns sequentially, calculating the ending portfolio value after 30 years.

5. Repeat 10,000 times: Generate 10,000 different 30-year sequences, each unique due to randomness.

6. Analyze outcomes: Out of 10,000 scenarios, how many ended with portfolio values > $1M? How many dropped below $300,000 at some point? This gives probability distributions.

Simplified Example Output

Percentile	10-Year Ending Value	30-Year Ending Value
10th (worst 10%)	$650,000	$1,200,000
25th	$750,000	$2,000,000
50th (median)	$900,000	$3,400,000
75th	$1,100,000	$5,500,000
90th (best 10%)	$1,350,000	$8,200,000

Notice: The "median" (50th percentile) is lower than the "mean" (average). This is because extreme negative years drag down averages more than extreme positive years compensate, due to volatility drag.

A naive deterministic model might project $982,000 for 10 years and $3,777,000 for 30 years. The Monte Carlo median is lower (showing realistic outcome), with wide ranges (showing uncertainty).

Asset Allocation and the Efficient Frontier

One of Monte Carlo's most powerful uses is comparing different asset allocations. Consider three portfolios:

Portfolio A (100% Stocks):

• Average return: 9.5% (nominal)
• Volatility (standard deviation): 18%
• 10th percentile (bad scenario) 30-year outcome: $1.2M
• 90th percentile (good scenario) 30-year outcome: $9.5M

Portfolio B (60% Stocks / 40% Bonds):

• Average return: 7.2%
• Volatility: 11%
• 10th percentile 30-year outcome: $1.8M
• 90th percentile 30-year outcome: $6.8M

Portfolio C (40% Stocks / 60% Bonds):

• Average return: 5.5%
• Volatility: 8%
• 10th percentile 30-year outcome: $2.0M
• 90th percentile 30-year outcome: $5.5M

A deterministic model would say: 100% stocks is best (9.5% return). Monte Carlo reveals the tradeoff: while best-case scenarios are better with 100% stocks, worst-case scenarios are much worse ($1.2M vs. $2.0M). For many investors, Portfolio B (60/40) offers the best risk-adjusted outcome.

Monte Carlo and Withdrawal Strategies

For retirees, Monte Carlo's greatest value is stress-testing withdrawal plans. Instead of assuming 7% average returns, test: "If I withdraw $30,000 annually from my $500,000 portfolio, what's the probability I'll run out of money before 30 years?"

Running 10,000 Monte Carlo scenarios:

• Scenario 1: Sequence is -20%, 15%, 8%, 5%, 12%... → Money lasts 32 years
• Scenario 2: Sequence is 14%, 9%, 22%, -5%, 11%... → Money lasts 28 years
• Scenario 3: Sequence is -30%, -5%, 8%, 12%, 10%... → Money lasts 24 years
• ...repeat 10,000 times

Out of 10,000 scenarios, 9,200 succeed (money lasts 30 years). 800 fail (money runs out early). Success rate: 92%.

This is far more informative than saying "7% returns support 6% withdrawals." It accounts for sequence risk. If the early years bring crashes, withdrawals might deplete the portfolio. If early years bring gains, you're fine even at higher withdrawal rates.

The "4% rule" (withdraw 4% of portfolio annually, adjusted for inflation) was historically safe because it succeeded in 95% of historical scenarios. But in low-yield, high-valuation environments, the success rate might drop to 85%. Monte Carlo reveals this.

Modeling Correlations and Diversification

A critical feature of Monte Carlo: modeling correlation between asset classes. Stocks and bonds aren't independent. In many scenarios, they rise together (inflation surprises). In others, they fall together (a crash).

Historical correlation (1990-2023): U.S. Stocks and Bonds = 0.15 (weak positive).

But in extreme scenarios (like 2020, COVID), correlation spiked. The Monte Carlo model can incorporate this:

Scenario	Stocks	Bonds	Combined 60/40
Normal Bull	+15%	+4%	+10.8%
Normal Bear	-12%	+3%	-6.6%
Crisis (High Correlation)	-25%	-8%	-18.8%
Stagflation	+3%	-5%	+0.8%

Diversification works best when correlations are low. A true Monte Carlo model captures scenarios where correlations rise (reducing diversification benefit) and scenarios where they remain low (maximizing it).

The Importance of Asset Class Selection

Which asset classes do you include in the simulation? A 60/40 portfolio (stocks/bonds) is fundamentally different from a diversified portfolio including real estate, commodities, and alternatives:

Asset Mix	30-Year Average	10th Percentile	90th Percentile
60/40 (US)	$3.4M	$1.8M	$6.8M
50/30/20 (Stocks/Bonds/Real Estate)	$3.5M	$2.1M	$6.6M
Global 60/40	$3.6M	$2.0M	$7.2M

The addition of real estate or international exposure shifts outcomes: lower worst-case (due to diversification), similar average, similar best-case. This is the true value of diversification—protecting against bad outcomes, not enhancing good ones.

Stress Testing and Black Swan Events

Monte Carlo can incorporate "black swan" events: rare, extreme scenarios. The 2008 financial crisis (-37% annually), COVID crash (-34%), or 1987 crash (-22%) are outside normal distributions.

A standard Monte Carlo model (using historical mean and standard deviation) would predict such crashes occurring roughly once per century. But by incorporating historical extreme returns, you can stress-test: "If a -35% crash occurred in year 5, what's the impact?"

Run 10,000 scenarios, with a 1% probability each year of a -35% return (instead of relying on the distribution to generate it naturally). This forces the model to test genuinely catastrophic scenarios.

Outcome: Withdrawal rates that seemed "safe" at 4% might drop to 2.5% if stress-tested against frequent severe crashes.

Real-World Application: Retirement Planning

A 65-year-old with $1,000,000 wants to retire. Monte Carlo addresses these questions:

1. Can I withdraw $50,000 annually? Run 10,000 scenarios. If 95% succeed, yes.
2. What if I live to 100? Extend simulations to 35 years (age 65 to 100). Success rate might drop from 95% to 87%.
3. What if I need $60,000 annually? Test this withdrawal rate. Success rate might drop to 75%.
4. What asset allocation is best? Test multiple allocations (60/40, 50/50, 70/30) against the same scenarios. The 50/50 might offer the best balance of success rate and upside.
5. What if I work 3 more years? Push retirement to 68, allowing $1,000,000 to grow. Run simulations. Success rate jumps to 97%.

Each question is answered probabilistically, not deterministically.

Common Mistakes

1. Assuming Monte Carlo is predictive: It's not. It's a range of possibilities based on historical distributions. If historical mean is 8% but true mean is 5% (lower growth environment), the simulations will overestimate outcomes.

2. Ignoring correlation scenarios: A model assuming fixed correlation between assets will miss scenarios where correlation spikes (crashes) and diversification fails. Use dynamic correlation.

3. Not stress-testing against real historical events: Just because 2008 was a -37% year doesn't mean the model will generate it naturally. Explicitly test harsh scenarios.

4. Over-optimizing based on median outcomes: The median (50th percentile) is not the important number. The 10th percentile (bad outcomes) and success rates (% of scenarios where plan works) matter more for planning.

5. Using inappropriate return distributions: Historical distributions assume stocks return 9-10% on average. In a low-growth environment or very high valuation era, 7% may be more realistic. Adjust your model's assumptions.

6. Ignoring fees and taxes: A $1M portfolio shrinks with 0.5% annual fees and 20% annual taxes. Monte Carlo should model these drags.

FAQ

Q: What software can I use to run Monte Carlo simulations? A: Free tools: Morningstar's retirement calculator, FIREcalc, cFIREsim. Premium: Morningstar Retirement Planner, Personal Capital, Vanguard retirement tools. Or build your own in Excel using the RAND() function and historical return distributions.

Q: Is 10,000 scenarios enough? A: Generally, yes. 10,000 scenarios provide good statistical stability. Running 1,000,000 scenarios provides only marginally more insight. 1,000 scenarios is adequate for rough estimates.

Q: Should I use historical returns or forward-looking return estimates? A: Ideally, forward-looking. If valuations are very high, forward returns might be lower than historical (7% instead of 10%). But historical returns are more objective and widely available. Use historical as the base case, and run sensitivity analyses with lower assumptions.

Q: How do I know if my simulation is reliable? A: Backtest it. Run simulations using historical data, then check if the predicted distribution matches actual historical outcomes. If simulations predict a 10% chance of a 30%+ loss, did that occur roughly 1% of the time? If not, adjust the model.

Q: Can Monte Carlo account for behavioral mistakes? A: Not directly. But you can model "sequence of returns risk" (the danger of withdrawing during downturns), which captures the behavioral risk of panic. More advanced models include probability of panic-selling or panic-pausing contributions.

Q: What success rate should I target? A: 90-95% for most investors. A 90% success rate means roughly 1 in 10 simulated futures fails. Over a 40-year retirement, that's acceptable—some scenarios will be bad. A 70% success rate is too risky. A 99% success rate means an overly conservative plan.

Related Concepts

1. Volatility and Standard Deviation: The input to Monte Carlo; determines the range of outcomes
2. Asset Allocation: The primary lever in Monte Carlo; different allocations generate different distributions
3. Sequence of Returns Risk: The impact of return order; Monte Carlo's strength
4. Safe Withdrawal Rates: Determined by Monte Carlo testing across scenarios
5. Distribution Analysis: Understanding percentiles, confidence intervals, tail risks
6. Backtesting: Validating Monte Carlo models against historical data

Summary

Monte Carlo simulations transform investment planning from a single deterministic projection ("you'll have $3.8M") into a probabilistic landscape ("90% of scenarios end with $2M–$7M, with a median of $3.4M").

For long-term investors, Monte Carlo's power is threefold:

1. It accounts for sequence risk: Order matters. Early crashes impact outcomes differently than late crashes.
2. It provides probabilities: Success rates (% of scenarios where your plan works) are more useful than average outcomes.
3. It enables stress-testing: You can test withdrawal rates, asset allocations, and life events against thousands of scenarios.

The simulations aren't predictions. They're a map of the territory you'll likely traverse. Some paths (scenarios) lead to success; others lead to running out of money. By understanding the distribution—not just the average—you build more robust, resilient plans.

Next

Having modeled the distribution of possible outcomes, we now address the final piece of the math: Setting Realistic Return Expectations, where we examine how to choose appropriate return assumptions for your own planning based on current valuations, asset class yields, and economic conditions.