Monte Carlo Simulation for Real Estate Investment

A standard DCF might assume 3% annual rent growth. But historical data shows rent growth is not a constant — it varies significantly. NCREIF data (1978–2023) shows that annual income return on commercial real estate ranged from -4.9% (2009) to +8.2% (1979), with a mean of approximately 7.2% and a standard deviation of approximately 2.8%. Monte Carlo simulation captures this variability by sampling from probability distributions rather than using fixed assumptions.

The process works as follows: (1) Identify key variables (rent growth, vacancy, operating expenses, exit cap rate, interest rates). (2) Define a probability distribution for each (e.g., normal distribution with mean 3% and standard deviation 2% for rent growth). (3) Randomly sample from each distribution. (4) Calculate the investment outcome (NPV, IRR) for that set of random draws. (5) Repeat 10,000+ times. The result is a distribution of possible outcomes rather than a single number.

The output of a Monte Carlo simulation is typically presented as a probability distribution of outcomes. Key statistics include the mean (expected) NPV or IRR, the standard deviation, the probability of a negative NPV (loss), and percentile values (10th, 25th, 50th, 75th, 90th). For example, a simulation might show: Mean IRR = 12.1%, Median IRR = 11.8% (positive skew from upside scenarios), P(IRR < 0) = 8%, 10th percentile IRR = 3.2%, 90th percentile IRR = 22.5%.

This information is far more useful than a single "expected IRR of 12%." It tells the investor there is an 8% chance of losing money, but a 10% chance of earning above 22.5%. The width of the distribution reveals total risk, while the downside tail (below the 10th percentile) reveals worst-case exposure. Institutional investors use these distributions to set risk limits — for example, requiring that the probability of negative NPV be below 15% for any approved investment.

Building an effective Monte Carlo model requires careful attention to correlations between variables. Rent growth and vacancy are negatively correlated (when rents rise, tenants leave, pushing up vacancy). Interest rates and cap rates are positively correlated (when interest rates rise, cap rates tend to expand). Ignoring these correlations produces unrealistic results — scenarios where rents surge while vacancy drops to zero, or interest rates spike while cap rates compress.

Software tools for Monte Carlo simulation range from Excel add-ins like @RISK and Crystal Ball (approximately $1,000–2,500/year) to Python libraries (NumPy, SciPy) that are free. For a basic real estate model, 10,000 simulations typically produce stable results. The primary limitation is "garbage in, garbage out" — if the input distributions are poorly calibrated (wrong means, too-narrow standard deviations, missing correlations), the output will be misleading regardless of how many simulations are run.