Forecasting Methods: Quantitative, Qualitative, and Practical Application

Forecasting Methods: Quantitative, Qualitative, and Practical Application

Quantitative Forecasting Methods

Quantitative forecasting relies on numerical data and statistical techniques to predict future outcomes. These methods are particularly useful when historical data is available and when patterns can be identified and extrapolated.

Time Series/Trend-Based Models

These models analyze historical data patterns over time without necessarily seeking underlying causes. They are useful for short-term forecasting and identifying trends, seasonality, and cyclical patterns.

• Smoothing Techniques: These methods average past data to smooth out random fluctuations and identify underlying trends.

  • Moving Average: Calculates the average of a specific number of past data points, moving the window forward over time.

    • Formula: MA_t = ( Y_t-1 + Y_t-2 + … + Y_t-n ) / n
      • Where: MA_t is the moving average at time t, Y_t-i is the value at time t-i, and n is the number of periods in the moving average.
      • Exponential Smoothing: Assigns exponentially decreasing weights to past data points, giving more weight to recent observations. This allows the model to be more responsive to recent changes in the data.
    • Formula: S_t = α * Y_t + (1 - α) * S_t-1
      • Where: S_t is the smoothed value at time t, Y_t is the actual value at time t, S_t-1 is the smoothed value at time t-1, and α is the smoothing constant (0 < α < 1).
      • A higher alpha gives more weight to recent values and vice versa
    • Regression Models: Use statistical techniques to find the relationship between a dependent variable (e.g., total return) and its past values.
      • Autoregressive (AR) Models: Predict future values based on a linear combination of past values of the same variable.
    • Formula: TR_t = α + β₁TR_t-1 + β₂TR_t-2 + … + β_pTR_t-p + ε_t
      • Where: TR_t is the total return at time t, α is a constant, β_i are coefficients, TR_t-i are past values of total return, p is the order of the model (number of past periods used), and ε_t is the error term.
      • Partial Autocorrelation Function (PACF): Helps determine the appropriate order (p) for an AR model by measuring the correlation between a variable and its past values, controlling for the effects of intervening lags.
      • Experiment Example: Collect historical quarterly total return data for a specific real estate asset class (e.g., apartments). Use an AR model to forecast total returns for the next four quarters. Vary the order (p) of the model and compare the accuracy of the forecasts using metrics like Mean Absolute Error (MAE) or Root Mean Squared Error (RMSE).

Causal/Structural Models

These models aim to explain the relationships between real estate variables and other economic, demographic, and financial factors. They are based on economic theory and seek to identify the drivers of real estate performance.

• Single Equation Models: Use a single equation to relate a dependent variable (e.g., real estate return) to one or more independent variables.
  • Multiple Regression: A statistical technique that estimates the relationship between a dependent variable and several independent variables.
    • Formula: TR_t = α + β₁X_1t + β₂X_2t + … + β_kX_kt + ε_t
      • Where: TR_t is the total return at time t, α is a constant, β_i are coefficients, X_it are independent variables (e.g., GDP growth, population growth, interest rates), k is the number of independent variables, and ε_t is the error term.
      • To implement, the model diagnostics❓ must be statistically significant and consistent with the theory
• Systems of Equations Models: Consist of a set of interrelated equations that describe the relationships between different variables in the real estate market.
  • Simultaneous Equation Models: Capture the feedback effects between variables, where a change in one variable affects another, which in turn affects the first variable.
  • Experiment Example: Develop a multiple regression model to forecast office rent growth. Independent variables could include employment growth in office-using sectors, vacancy rates, and construction costs. Collect historical data for these variables and estimate the model coefficients. Evaluate the model’s performance using statistical tests (e.g., t-tests, F-tests) and compare its forecast accuracy to a simple time series model.
• Model Diagnostics: Rigorous tests that need to be conducted on the model in order to validate it, such as evaluating error terms and minimzation.

Qualitative Forecasting Methods

Qualitative forecasting relies on expert opinion, market surveys, and other subjective assessments to predict future outcomes. These methods are particularly useful when historical data is limited or when significant changes in the market are expected.

• Expert Opinion: Gathering insights from real estate professionals, economists, and other experts who have knowledge of the market.
• Delphi Method: A structured process for collecting and synthesizing expert opinions, involving multiple rounds of questionnaires and feedback.
• Surveys: Collecting information from market participants (e.g., investors, developers, tenants) about their expectations for the future.
• Historical Analogy: Comparing the current market situation to similar situations in the past to identify potential future outcomes.
• Geographical Analogy: Comparing current market situation to similar dynamics in different geographical locations to identify potential future outcomes.
• Yield/Cap Rate Estimation: Qualitative analysis of yields or cap rates for the estimation of real estate returns.
  • Formula: K = RFR + RP – G + D
    • Where: K is the yield or capitalisation rate of an asset, RFR is the risk❓-free rate, RP is the risk premium, G is the long-term average rental growth rate and D is the long-term average depreciation rate of the property (or the annual average amount of investment required to maintain the quality of the asset)

Practical Application: Integrating Quantitative and Qualitative Methods

In practice, the most effective forecasting approach often involves combining quantitative and qualitative methods.

1. Start with Quantitative Analysis: Use time series and causal models to establish a baseline forecast based on historical data and economic relationships.
2. Incorporate Qualitative Insights: Adjust the quantitative forecast based on expert opinion, market surveys, and other qualitative assessments. Consider factors such as changes in government policy, technological innovations, and shifts in investor sentiment.
3. Scenario Planning: Develop multiple scenarios based on different assumptions about key variables. This can help to assess the range of potential outcomes and to identify the factors that are most likely to drive real estate performance.
4. Sensitivity Analysis: Test the sensitivity of the forecast to changes in key assumptions. This can help to identify the most critical variables and to assess the potential impact of uncertainty.
5. Backtesting and Model Validation: Regularly compare the model’s forecasts to actual outcomes and make adjustments as needed. Conduct statistical diagnostic tests to validate models and ensure that it has a correlation to actual results over time.

Asset-Specific Forecasting: A Four-Column Approach

A structured approach to asset-specific forecasting involves considering different spatial units and explicitly presenting the interconnection between the strategic forecast and the asset-level forecast:

• Benchmark Market: The overall real estate market (e.g., national or regional).
• Specific Sub-Market: A more narrowly defined geographic area or property type (e.g., downtown office market).
• Specific Asset: The individual property being analyzed.
• Comment: A detailed explanation of why the asset is expected to perform differently from the benchmark market and sub-market.

Real Estate Demand and Supply Model

• Demand Factors: D_t = f( Pop_t, Inc_t, P_t-1 )
  • Where:
    • D_t is demand for real estate at time t.
    • Pop_t is population at time t.
    • Inc_t is average income at time t.
    • P_t-1 is price of real estate in the previous period.
• Short-Term Supply Factors S_t = f ( CP_t-3)
  • Where:
    • S_t is new supply at time t.
    • CP_t-3 is construction permits issued three periods prior.
• Long-Term Supply Factors S_t = f ( S_t-1)
  • Where:
    • S_t is new supply at time t.
    • S_t-1 is new supply in the previous period.