Asset-Specific Forecasting: Methods and Techniques

Asset-Specific Forecasting: Methods and Techniques

Introduction to Asset-Specific Forecasting

Asset-specific forecasting focuses on predicting the future performance of individual real estate assets, rather than broad market trends. This granular approach considers the unique characteristics of each property, including its location, physical attributes, tenant mix, and lease terms, to derive more precise and actionable insights for investment decisions. This chapter will explore both informal and formal approaches to asset-specific real estate forecasting.

Informal Forecasting Approaches

• Rely on market experience and intuition.
• Reflect professional and entrepreneurial traditions of real estate.
• Place great emphasis on market sentiment and fundamental understanding.
• Qualitative; it often leverages conversations with local real estate professionals or on-the-ground analysis of tenant activity.

Formal Forecasting Approaches

These approaches can be quantitative, qualitative, or a combination of both.

Quantitative Approaches

• Subdivided into time-series/trend-based analysis and causal/structural analysis.
  • Time-series/trend-based: focus on identifying patterns in historic data without necessarily seeking explanatory theories.
  • Causal/structural: involve building, testing, and using models with strong theoretical underpinnings to identify causal relationships.

Time-Series/Trend-Based Models

These models do not generally involve strong theoretical frameworks. They are empirical approaches that seek patterns in historical data.

• Assume past patterns will repeat predictably in the future.
• Identify long-term trends, seasonal/cyclical patterns, momentum, and mean reversion.
• Limited data needs; quick and easy to develop.
• Provide a useful baseline for comparison with other models.
• Often successful for forecasting stable data series (e.g., income return).

Examples of Time-Series Models:

• Smoothing Models (Moving Averages, Exponential Smoothing): These models smooth out fluctuations in the data to identify underlying trends.

  • Moving Average: Calculates the average of a specific number of preceding data points to predict the next value.
    • Equation:
      • TRt = (TRt-1 + TRt-2 + ... + TRt-n) / n
        • Where TRt is the total return in period t, and n is the number of periods in the moving average.
  • Exponential Smoothing: Assigns exponentially decreasing weights to past observations.
    • Equation:
      • TRt+1 = αTRt + (1 - α)Ft
        • Where TRt+1 is the forecast for the next period, TRt is the actual return in the current period, Ft is the prior forecast for the current period, and α is the smoothing constant (0 < α < 1). A higher α gives more weight to recent data.

• Regression Models (Autoregressive Models, Partial Autocorrelation): Use historical data to predict future values based on relationships within the series itself.

  • Autoregressive (AR) Model: Predicts future values based on past values of the time series.

    • Equation: TRt = α + β1TRt-1 + β2TRt-2 + ... + βpTRt-p + εt
      • Where TRt is the total return in period t, α is a constant, β1, β2,... βp are coefficients, TRt-1, TRt-2,... TRt-p are the total returns in the previous p periods, and εt is the error term. p represents the order of the autoregressive model (i.e., how many previous periods are used).

  • Critical Properties of Time-Series Models:

    • Stationarity: The statistical properties of the time series (mean, variance, autocovariance) are constant over time. Non-stationary time series may require transformations (e.g., differencing) to achieve stationarity.
    • Autocorrelation: The correlation between a time series and its lagged values. Autocorrelation functions (ACF) and partial autocorrelation functions (PACF) are used to identify the order (p) of AR models.
    • Constant Mean
    • Constant Variance
    • Autocovariance is non-zero to lag q

Causal/Structural Models

Intrinsic to causal/structural models is a strong theoretical underpinning of the model. The model links the dependent variable (real estate return) with one or more fundamental independent variables (for example, demographic economic) that drive the performance of the dependent variable in a consistent and therefore predictable manner. Often, perhaps ideally, the dependent variable may lag by one or more periods any change in an independent variable.

• Link dependent variable (real estate return) to independent variables (e.g., demographic, economic) that drive its performance.
• Require a strong theoretical basis.
• Simplicity is an advantage (fewer variables are easier to manage and interpret).

Single Equation Models (Multiple Regression):
* Use a single equation to model the relationship between the dependent and independent variables.
* Equation: TRt = α + β1X1t + β2X2t + β3X3t + β4X4t + εt
* Where TRt is the total return in period t, α is a constant, β1, β2, β3, β4 are coefficients for variables X1, X2, X3, X4, X represents independent variables, and εt is an error term.
* Example: Predicting property value based on square footage, number of bedrooms, and location score.
* Critical Properties of Casual/Structural Models:
* Identify independent variables from theory
* Model diagnostics should be statistically significant and consistent with theory
* Error term ε should be minimised

Systems of Equations:
* Use a nested set of individual equations where the output of one equation feeds into another.
* Allow for more explicit modeling of the relationships between variables.
* Require careful attention to model stability (avoiding positive or negative feedback loops).

*   *Example:* A model might include:
    1.  An equation predicting employment growth based on economic indicators.
    2.  An equation predicting office space demand based on employment growth.
    3.  An equation predicting rental rates based on office space demand and supply.

Qualitative Approaches

Expert opinion or surveys of expert opinion and Delphic methods all provide a sense that a forecast is grounded in a ‘market view’ and hence a degree of reassurance that an investment decision is being rationally made. The consensus approach implicit in these methods suggest that even if the forecast turns out to be wrong, there would be few who could criticise a decision based upon it. Indeed, pure qualitative techniques can be well suited to the prediction of turning points and non-linear changes, for example, where the past is no longer a good guide to the future (but of course may/may not be correct).

• Rely on expert opinion, surveys, Delphi methods, and historical or geographical analogies.
• Provide a “market view” and reassurance for investment decisions.
• Well-suited for predicting turning points and non-linear changes.
• Can be used in combination with quantitative techniques to inform judgment.

Qualitative Modeling of Yields/Capitalization Rates:
* Estimate yields or cap rates based on market sentiment and risk appetite.

*   Equation: `K = RFR + RP – G + D`
    *   Where `K` is the yield or capitalization rate, `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.
    *   *Risk Premium (RP):* This is a key element, driven by market conditions and the perceived risk of real estate investment relative to other asset classes. Qualitative techniques can help estimate an appropriate risk premium based on market sentiment, surveys, and historical analysis.

Limitations of Qualitative Methods:
* May be susceptible to “anchoring” (disproportionate influence of current values or trends).
* Potential for expert opinion to overemphasize data confirming existing biases.

Practical Application: A Four-Column Approach to Asset-Specific Forecasting

This method explicitly presents the interconnection between strategic (market-level) forecasts and asset-level forecasts.

1. Benchmark Market: Forecast for the broader real estate market (e.g., national or regional).
2. Specific Sub-Market: Forecast for the sub-market where the asset is located (e.g., downtown office market).
3. Specific Asset: Forecast tailored to the individual asset, considering its unique characteristics.
4. Comment: Clear explanation of why the asset is expected to outperform or underperform the wider market.

Example:

Forecasting Income Return

• Emphasize understanding trends and forecasting real estate occupational markets (demand, supply, occupancy, rent).
• Lends itself to formal quantitative approaches:
  • Time-series/trend-based for supply.
  • Causal/structural for demand, occupancy, and rent.

Forecasting Capital Return

• Combine understanding of likely change in NOI with understanding of trends and forecasting of capital markets.
• Requires both quantitative and qualitative approaches.
• Qualitative aspects focus on forecasting capital market trends (e.g., risk premium over risk-free rate).

Modeling Demand, Supply, Occupancy, Rent, and NOI

• Demand: Use causal/structural econometric techniques (e.g., based on population, wealth, and price).
• Supply:
  • Short-term (2-3 years): monitor projects under construction or use causal/structural techniques (e.g., based on construction permits).
  • Long-term (over 3 years): use time-series/trend-based techniques.
• Occupancy, Rent, NOI: Forecast using a system of equations with common input variables and/or outputs from other equations.

Scenario Planning and Sensitivity Testing

Given the uncertainties inherent in forecasting, investors often run scenarios using different assumptions for key variables and conduct sensitivity tests to assess the impact of changes in these assumptions on the forecast results.

Backtesting and Model Validation

Where quantitative techniques are involved it is essential that the model is subjected to the range of relevant statistical diagnostic tests❓❓ and wherever possible back-tested against historical data.

Conclusion

Asset-specific forecasting requires a combination of quantitative and qualitative methods, tailored to the unique characteristics of each property. Understanding the strengths and limitations of each approach is crucial for developing robust and reliable forecasts that inform investment decisions.