Asset-Specific Forecasting: Models and Techniques

Asset-Specific Forecasting: Models and Techniques

The Importance of Asset-Specific Forecasting

Asset-specific forecasting goes beyond broad market❓ trends to analyze the unique characteristics and potential performance of individual properties. This level of detail is crucial for:

• Informed Investment Decisions: Understanding the specific drivers of value for a particular asset allows for more accurate valuation and risk assessment.
• strategic❓ Asset Management: Tailoring management strategies to the specific needs and opportunities presented by each asset can optimize returns.
• Identifying Opportunities and Risks: Asset-specific analysis can reveal hidden potential or vulnerabilities that might be missed at a market-wide level.
• Performance Benchmarking: Comparing an asset’s actual performance against its forecast allows for evaluation of management effectiveness and identification of areas for improvement.

As noted in the provided document, explicit presentation of the interconnection between strategic and asset-level forecasts is essential, and the ability to make reasoned adjustments based on benchmark market changes adds further value. The commentary explaining why an individual asset might outperform or underperform the wider market is perhaps the most critical element.

Forecasting Approaches: A Conceptual Framework

There are two main approaches to forecasting: informal and formal.

• Informal Approaches: Rely heavily on market experience, intuition, and sentiment. These approaches are often rooted in the professional and entrepreneurial aspects of real estate.

• Formal Approaches: Can be quantitative, qualitative, or a combination of both.

  • Quantitative: Subdivided into time-series/trend-based analysis and causal/structural analysis.
  • Qualitative: Include expert opinions, Delphi methods, and historical or geographical analogies.

The most appropriate forecasting technique varies based on the specific metric of interest:

• Income Return Forecasting: Emphasizes understanding trends and forecasting real estate occupational markets (NOI). Formal quantitative approaches – perhaps time-series/trend-based for supply and causal/structural for demand, occupancy and rent – are often used.

• Capital Return Forecasting: Requires a combination of understanding likely changes in NOI and forecasting capital markets trends. This necessitates both quantitative and qualitative approaches (e.g., forecasting risk premium over a risk-free rate).

Quantitative Modelling Approaches

Quantitative models are broadly classified into two categories: time series/trend-based and causal/structural. Both rely on the crucial assumption that the past is a good guide to the future, albeit in different ways.

1. Time Series/Trend-Based Models

• Description: These models are empirical and focus on identifying patterns in historical data without necessarily seeking underlying theoretical explanations. They assume that observed patterns will repeat in a predictable manner.

• Examples: Identifying long-term trends, seasonal patterns, momentum, and mean reversion.

• Advantages:

  • Limited data needs.
  • Quick and easy to develop.
  • Useful baseline for comparison with other models.
  • Can be effective for forecasting non-volatile data series (e.g., income return).

• Disadvantages:

  • Lack of explanatory power.
  • Accuracy may decrease over longer forecast horizons.
  • Vulnerable to unforeseen events or structural changes.

• Examples of Time-Series Models:

  • Smoothing Models (e.g., Moving Averages, Exponential Smoothing): These models assign weights to past observations to smooth out fluctuations and identify underlying trends. A moving average calculates the average of a set of past values over a specified period. Exponential smoothing assigns exponentially decreasing weights to older observations.

  • Regression Models (e.g., Autoregressive Models): These models use past values of the variable being forecast to predict future values.

    • Autoregressive Model (AR(q)): This model predicts the total return in period t based on a linear combination of the total returns in the previous q periods, plus an error term:

      • TR_t = α + β₁TR_t-1 + β₂TR_t-2 + … + β_qTR_t-q + ε_t

        • TR_t: Total return in period t.
        • α: Constant term.
        • β_i: Coefficient for the i-th lagged total return.
        • TR_t-i: Total return in period t-i.
        • ε_t: Error term.
        • q: Number of lagged periods.

      • Critical Properties: stationarity❓❓ is essential for reliable autoregressive models. Stationarity means that the statistical properties of the time series (mean, variance, autocovariance) do not change over time.

• Equation and Critical Properties (Generalized for Regression):

  *  TR<sub>t</sub> = α + ρ<sub>t</sub> + ß<sub>1</sub>ρ<sub>t-1</sub> +
  ß<sub>2</sub>ρ<sub>t-2</sub> + … + ß<sub>q</sub>ρ<sub>t-q</sub>
  
  *  Constant mean
  *  Constant variance
  * Autocovariance is non-zero to lag q
  

2. Causal/Structural Models

• Description: These models are based on a theoretical understanding of the relationships between the dependent variable (real estate return) and independent variables (e.g., demographic, economic factors).

• Key Principle: The model links the dependent variable with fundamental independent variables that drive its performance in a consistent and predictable manner. Ideally, the dependent variable lags behind changes in the independent variables.

• Advantages:

  • Provide insights into the underlying drivers of real estate performance.
  • Can be used to simulate the effects of different scenarios.
  • Potentially more accurate than trend-based models when relationships are well-defined.

• Disadvantages:

  • Require more data and expertise to develop and maintain.
  • Dependence on forecasts of independent variables introduces uncertainty.
  • Model complexity can make them difficult to interpret and explain.
  • The need for forecasting of independent variables transfers the problem of forecasting.

• Example: Multiple Regression Model:

  • TR_t = α + β₁X_1t + β₂X_2t + β₃X_3t + β₄X_4t + ε_t

    • TR_t: Total return in period t.
    • α: Constant term.
    • β_i: Coefficient for variable X_i.
    • X_it: Independent variable i in period t.
    • ε_t: Error term.

  • Critical Properties:

    • Independent variables should be identified based on solid theory.
    • Model diagnostics (e.g., R-squared, t-statistics) should be statistically significant and consistent with the theory.
    • The error term should be minimized.

• Modelling Techniques within Causal/Structural Models:

  • Single Equation: Typically, a multiple regression-based equation with strong theoretical underpinnings and rigorously tested model specification.

  • Systems of Equations: A nested set of individual equations where the results from one equation feed into another as an input. Each equation has strong theoretical underpinnings and a rigorously tested model specification.

    • Simpler, but potential interaction between variables is not necessarily made explicit.
    • nested system of equations should be stable.

Practical Considerations for Quantitative Models

• Data Availability: The lack of good-quality, long-term data sets can be a significant limitation, particularly for causal/structural models.
• Model Simplicity: Simpler models are often more robust and easier to explain and implement. A balance must be struck between explanatory power and suitability for investment decisions.
• Scenario Analysis and Sensitivity Testing: Running scenarios and conducting sensitivity tests on key assumptions is crucial to assess the robustness of the model and understand the potential range of outcomes.

Qualitative Approaches

Qualitative approaches remain valuable, especially considering the difficulties in quantitative modeling of real estate returns. They serve as an alternative, a check on quantitative results, or a combination of both.

• Methods:

  • Expert Opinion: Gathering insights from experienced market participants.
  • Surveys of Expert Opinion: Systematically collecting and analyzing opinions from a panel of experts.
  • Delphi Methods: A structured communication technique used to develop a consensus among a panel of experts.
  • Historical or Geographical Analogy: Drawing parallels with past events or similar markets to inform forecasts.

• Advantages:

  • Can capture market sentiment and non-linear changes.
  • Useful for predicting turning points.
  • Provide a sense of market grounding and rationality.
  • May inform judgements regarding a model’s utility, variable usage, input adjustments, or results.

• Disadvantages:

  • Subject to biases and subjective interpretations.
  • May overemphasize current trends (‘anchoring’).

• Qualitative Modelling of Yields/Capitalization Rates:

  • Yields and cap rates are often difficult to forecast quantitatively and reflect market sentiment and risk appetite.

  • A qualitative model to estimate yield or cap rate can be based on the Gordon Growth Model:

    • K = RFR + RP – G + D

      • K: Yield or capitalization rate.
      • RFR: Risk-free rate (e.g., long-term government bond rate).
      • RP: Risk premium for real estate investment compared to the RFR.
      • G: Long-term average rental growth rate.
      • D: Long-term average depreciation rate (or annual average maintenance investment).

    • The risk premium (RP) is influenced by the availability of equity and debt, the supply and demand of investment properties, and overall market sentiment.

Forecasting in Practice: A Pragmatic Approach

Organizations often adopt a pragmatic approach, focusing on forecasting systems that demonstrate a significant correlation with actual results over time.

• Statistical Diagnostic Tests and Back-Testing: Quantitative models must be subjected to relevant statistical diagnostic tests and back-tested against historical data.

• Conceptual Framework: A common framework involves:

  1. Modeling occupational markets and forecasting NOI using causal/structural quantitative techniques to derive the income return.
  2. Using qualitative techniques to forecast yield or cap rate, serving as an input to capital value calculation and capital return forecasting.
  3. Summing the income and capital return to calculate the total return.

Key Components of Real Estate Forecasting

1. Demand Forecasting

• Approach: Best suited to causal/structural econometric techniques.
• Theoretical Basis: Demand is often a function of population size and wealth in a defined trade area, inversely related to past real estate prices.
• Model Components: Demographic, economic, and price variables.
• Key Generic Variables: Demographic and economic factors interact through employment, which significantly influences real estate demand.

2. Supply Forecasting

• Short-Term (2-3 years):
  • Monitoring projects under construction and building a granular database.
  • Simple causal/structural techniques (e.g., new supply depends on construction permits issued in previous periods).
• Long-Term (3+ years):
  • Projecting completions using time-series/trend-based techniques (e.g., new supply depends on supply in the previous period).

3. Occupancy, Rent, and NOI

• Forecasting Methodology: Use forecasts of demand and supply as inputs to predict occupancy rates, rent (gross, net, net effective), and NOI.
• System of Equations: These factors are often forecast using a system of equations, with common input variables or outputs feeding into other equations. The result is a forecast of the income return.

4. Total Return

• In this framework, the total return is the sum of the income and capital return.

The Future of Forecasting

The future of real estate forecasting will be shaped by:

• Improved Data Quality and Availability: The increasing sophistication of the industry and the wider availability of data will enhance forecasting accuracy.
• Integration of New Technologies: Machine learning and artificial intelligence may play a more significant role in identifying patterns and improving forecasting accuracy.
• Enhanced Scenario Planning: Developing more sophisticated scenario planning tools will enable investors to better assess risks and opportunities in a dynamic market environment.