Distributions, Correlations, and Model Outputs

Distributions, Correlations, and Model Outputs

Probability Distributions in \data\\❓\\-bs-toggle="modal" data-bs-target="#questionModal-123798" role="button" aria-label="Open Question" class="keyword-wrapper question-trigger">monte carlo simulation❓❓

Monte Carlo simulation relies heavily on probability distributions to represent the uncertainty associated with input variables. Selecting the appropriate distribution is crucial for the accuracy and reliability of the simulation results. A distribution describes the range of possible values a variable can take and the likelihood of each value occurring.

• Continuous vs. Discrete Distributions:

  • Continuous distributions represent variables that can take on any value within a given range (e.g., rental growth, cap rates). Examples include Normal, Lognormal, Uniform, Triangular, and Exponential distributions.
  • Discrete distributions represent variables that can only take on a finite number of distinct values (e.g., number of tenants, renovation stages). Examples include Bernoulli, Binomial, Poisson, and Discrete Uniform distributions. In practice, discrete data can often be treated as continuous when the number of observations is large.

• Common Distributions:

  • Normal Distribution: A symmetrical bell-shaped distribution defined by its mean (μ) and standard deviation❓ (σ). It is widely used due to the Central Limit Theorem, which states that the sum (or average) of a large number of independent, identically distributed random variables will be approximately normally distributed.
    • Probability Density Function (PDF):
      f(x) = (1 / (σ * sqrt(2 * π))) * e^(-((x - μ)^2) / (2 * σ^2))
    • Suitable for variables where values are clustered around a mean and deviations are equally likely in both directions (e.g., normally distributed error terms in regression models).
  • Lognormal Distribution: The logarithm of the variable is normally distributed. It is bounded by zero and skewed to the right, making it suitable for variables that cannot be negative and tend to have positive skewness (e.g., stock prices, real estate values).
    • If Y ~ N(μ, σ²), then X = e^Y follows a lognormal distribution.
    • PDF:
      f(x) = (1 / (x * σ * sqrt(2 * π))) * e^(-((ln(x) - μ)^2) / (2 * σ^2)), for x > 0
  • Uniform Distribution: All values within a given range are equally likely. Defined by a minimum (a) and maximum (b) value.
    • PDF:
      f(x) = 1 / (b - a), for a ≤ x ≤ b
    • Useful when there is no prior knowledge about the distribution of a variable, and all values within a range are considered equally plausible.
  • Triangular Distribution: Defined by a minimum (a), maximum (b), and most likely (mode, c) value. It is a simple distribution to use when limited data is available.
    • PDF: Piecewise function
      f(x) = { (2 * (x - a)) / ((b - a) * (c - a)), for a ≤ x ≤ c (2 * (b - x)) / ((b - a) * (b - c)), for c ≤ x ≤ b 0, otherwise }
    • Suitable for situations where you have estimates of the best-case, worst-case, and most likely scenarios.
  • Exponential Distribution: Describes the time until an event occurs in a Poisson process. Defined by a rate parameter (λ).
    • PDF:
      f(x) = λ * e^(-λ * x), for x ≥ 0
    • Useful for modeling the time between events, such as the time until a tenant vacates a property.
  • Bernoulli Distribution: A discrete distribution representing the probability of success or failure of a single trial. Defined by a probability of success (p).
    • Probability Mass Function (PMF):
      P(X = x) = { p, if x = 1 (success) 1 - p, if x = 0 (failure) }
    • Can be used to model binary events, such as whether or not a lease is renewed.
  • Binomial Distribution: A discrete distribution representing the number of successes in a fixed number of independent Bernoulli trials. Defined by the number of trials (n) and the probability of success (p).
    • PMF:
      P(X = k) = (n choose k) * p^k * (1 - p)^(n - k), for k = 0, 1, ..., n
    • Useful for modeling the number of successful leases signed out of a pool of potential tenants.
  • Poisson Distribution: A discrete distribution representing the number of events occurring in a fixed interval of time or space. Defined by a rate parameter (λ).
    • PMF:
      P(X = k) = (λ^k * e^(-λ)) / k!, for k = 0, 1, 2, ...
    • Suitable for modeling the number of repairs needed on a property in a given month.

• Considerations when Selecting a Distribution:

  1. Theoretical Consistency: Ensure the distribution aligns with underlying financial or economic theories (e.g., non-negativity of stock prices, interest rates).
  2. Data Fit: The distribution should adequately represent the historical data, if available. Statistical tests like the Kolmogorov-Smirnov test or Chi-squared test can be used to assess the goodness-of-fit.
  3. Continuity/Discreteness: Select a distribution appropriate for the nature of the variable.
  4. Skewness: Consider if the data is skewed. If so, symmetrical distributions like the Normal distribution may be inappropriate.
  5. Kurtosis: Kurtosis refers to the “tailedness” of a distribution. Distributions with high kurtosis have fatter tails, indicating a higher probability of extreme values.

Correlation Among Variables

Correlation describes the statistical relationship between two or more random variables. In Monte Carlo simulation, accurately modeling correlations is crucial because input variables are often interdependent. Ignoring correlations can lead to significantly biased results.

• Correlation Coefficient (ρ): A measure of the linear relationship between two variables. It ranges from -1 to +1.

  • ρ = +1: Perfect positive correlation. As one variable increases, the other increases proportionally.
  • ρ = -1: Perfect negative correlation. As one variable increases, the other decreases proportionally.
  • ρ = 0: No linear correlation. The variables are independent.

• Calculating the Correlation Coefficient:

  • For two random variables X and Y, the population correlation coefficient is:

    ρ(X, Y) = Cov(X, Y) / (σ_X * σ_Y)

    where:
    * Cov(X, Y) is the covariance between X and Y.
    * σ_X and σ_Y are the standard deviations of X and Y, respectively.
    * The sample correlation coefficient is calculated as:

    r = Σ[(x_i - x̄) * (y_i - ȳ)] / [sqrt(Σ(x_i - x̄)^2) * sqrt(Σ(y_i - ȳ)^2)]
    where:
    * x_i and y_i are the individual data points for variables X and Y, respectively.
    * x̄ and ȳ are the sample means of X and Y, respectively.

• Methods for Incorporating Correlation in Monte Carlo Simulations:

  1. Cholesky Decomposition: A common technique to generate correlated random variables from independent random variables. It involves decomposing the correlation matrix into a lower triangular matrix (L). The correlated variables are then obtained by multiplying L with a vector of independent random variables.
  2. Copulas: Functions that describe the dependence structure between random variables, independent of their marginal distributions. They allow you to model correlations between variables even if they have different distributions (e.g., correlating a normal variable with a lognormal variable). Common copulas include Gaussian copulas and t-copulas.

• Example: Correlation in Real Estate

  • Positive Correlation: Employment growth and office rental rates. As employment increases, demand for office space rises, leading to higher rental rates.
  • Negative Correlation: Vacancy rates and rental rates. As vacancy rates increase (more empty space), landlords may lower rental rates to attract tenants. Cap rates and rental growth may be negatively correlated. As rental growth increases, the required rate of return (cap rate) tends to decrease, causing prices to rise faster than rents.

Modeling Stochastic Growth

In financial modeling, particularly when projecting future cash flows or asset values, accounting for stochastic growth is essential. Simply applying a constant growth rate can be misleading when the variable is subject to random fluctuations.

• Stochastic Price Model (Geometric Brownian Motion): A commonly used model for simulating asset prices in continuous time. It assumes that price changes are random and follow a lognormal distribution.

  • Equation for price at time T:

    P_T = P_0 * e^((μ - 0.5 * σ^2) * T + σ * Z * sqrt(T))

    where:
    * P_T is the price at time T.
    * P_0 is the initial price.
    * μ is the expected rate of return (drift).
    * σ is the volatility (standard deviation of returns).
    * Z is a standard normal random variable (mean 0, standard deviation 1).
    * T is the time horizon.

• Explanation of the Formula:

  • The term (μ - 0.5 * σ^2) * T represents the expected growth of the asset, adjusted for the effect of volatility. The “0.5 * σ^2” term is a convexity adjustment known as Itô’s Lemma. It’s crucial because with lognormal distributions, the expected value of e^x is not e^E[x]. Higher volatility increases the expected terminal value.
  • The term σ * Z * sqrt(T) introduces randomness, causing the price to fluctuate around the expected growth path.
  • Without the volatility adjustment (-0.5 * σ²), the simulated prices would overestimate the expected future price.
  • If σ is zero, the formula simplifies to the standard exponential growth formula: P_T = P_0 * e^(μ * T).

Model Outputs and Analysis

The primary goal of Monte Carlo simulation is to generate a distribution of possible outcomes, providing insights into the range of potential results and the associated probabilities.

• Key Output Metrics:

  • Discounted Cash Flow (DCF): The present value of future cash flows, reflecting the time value of money.
  • Internal Rate of Return (IRR): The discount rate that makes the net present value (NPV) of all cash flows from a project equal to zero.
  • Net Present Value (NPV): The difference between the present value of cash inflows and the present value of cash outflows.
  • Profitability Index (PI): The ratio of the present value of cash inflows to the initial investment.
  • Equity Multiple: The total cash distributed to equity investors divided by the total equity invested.

• Analyzing Output Distributions:

  • Summary Statistics:
    • Mean: The average outcome.
    • Median: The middle value (50th percentile).
    • Standard Deviation: A measure of the dispersion or variability of the outcomes.
    • Percentiles: Values that divide the distribution into specified percentages (e.g., 10th percentile, 90th percentile).
    • Minimum and Maximum Values: The best-case and worst-case outcomes.
  • Visualizations:
    • Histograms: Show the frequency distribution of the outcomes.
    • Box Plots: Provide a summary of the distribution, including the median, quartiles, and outliers.
    • Cumulative Distribution Functions (CDFs): Show the probability that the outcome will be less than or equal to a given value.
  • Skewness and Kurtosis:
    • Skewness: Measures the asymmetry of the distribution. A positive skew indicates a long tail to the right (more upside potential), while a negative skew indicates a long tail to the left (more downside risk).
    • Kurtosis: Measures the “tailedness” of the distribution. High kurtosis indicates fat tails (higher probability of extreme events).

• Interpreting Results:

  1. Probability of Loss: Calculate the probability that the NPV or IRR will be below a certain threshold (e.g., zero or the cost of capital).
  2. Value at Risk (VaR): Estimate the maximum potential loss at a given confidence level (e.g., 95% VaR).
  3. Expected Shortfall (ES): Estimate the expected loss given that the loss exceeds a certain threshold (e.g., 95th percentile).
  4. Sensitivity Analysis: Examine how the output distributions change when the input assumptions (distributions, correlations) are varied.

Example: Using Multiple Regression to Determine Model Inputs

Multiple regression analysis can be used to establish a relationship between a random variable and other explanatory variables, thus predicting input parameters for your Monte Carlo Simulation.

• Model: St = Xt + εt
  • St: Office construction starts
  • Xt: Vector of explanatory variables (prices, vacancy rates, lagged construction starts, interest rates)
  • εt: Error term (mean of zero)
• Ordinary Least Squares (OLS) regression can be used to estimate the parameters of the model. However, if the error term and explanatory variables are correlated, this may produce inconsistent and biased results and another method such as two-stage least squares should be used.

Box-Whisker Plots to Describe Distributions

A box-whisker plot is a way to represent batches of data through five-number summaries. These summaries are the smallest observation (sample minimum), lower quartile, median, upper quartile and largest observation. The width of the box represents the interquartile range (IQR). The band near the middle is the 50th percentile, or the median. The cross is the mean. The lowest whisker represents data within 1.5 IQR of the lower quartile; the highest within 1.5 IQR of the upper quartile. Data beyond the whiskers are plotted as open squares. The solid outliers are the most extreme data points.

Practical Applications and Experiments

• Real Estate Investment Analysis: Simulating property cash flows, considering uncertainties in rental income, expenses, vacancy rates, and exit cap rates. Experimenting with different distribution assumptions and correlation structures to assess the impact on investment returns.
• Development Projects: Modeling project costs and completion times, accounting for risks related to construction delays, material price fluctuations, and permitting issues.
• Portfolio Optimization: Constructing real estate portfolios that balance risk and return, considering the correlations between different property types and locations.
• Risk Management: Identifying the key risk drivers in real estate investments and developing strategies to mitigate those risks.
• Valuation of Options: The Black-Scholes model, widely used for pricing options, relies on the lognormal distribution of stock prices. Monte Carlo simulations can be used to value more complex options where closed-form solutions are not available.
• Lease Structures: Simulating returns from upward-only adjusting leases, explicitly accounting for stochastic growth in rental rates.

By carefully selecting probability distributions, incorporating correlations, and analyzing output distributions, Monte Carlo simulation can provide valuable insights for making informed decisions in real estate investment analysis.