Site Overlay

Land price risk: a regime-based approach

Estimated reading time: 6 minutes

In many residential real estate projects the development of the historical land price risk plays an important role. Land price is defined as the difference between the market value and construction costs1 of the property; it is also known as the residual value. It represents the additional value over the construction cost of the property. It is the add-on that people are willing, or able, to pay to live in a certain area.

As property prices and construction costs fluctuate due to market movements, also the residual land value can become quite volatile. This represents an important risk factor for landowners andland developersas their results largely depend on changes in these residual values.

Key questions for land and project developers are therefore:

  1. How much do land prices change historically ?
  2. What is the land price risk in the current environment?
  3. How reliable is the risk estimate?

To answer these questions, we start with a historical overview of land prices since 1914. Next we create context-dependent risk estimates using a regime-based risk model. We measure the reliability of the estimation by performing out-of-sample backtests and by comparing the performance to a base risk model. Finally we discuss the implications for land owners.

Land prices in historical perspective since 1914

Land prices can be obtained as the difference between the market prices and the construction prices, therefore we can reconstruct the historical residential land price from these two components.

The graph below shows both new house prices, construction costs and the derived land prices since 1914 for an average house in the Netherlands:

Historical land prices 1914-2023
Figure 1: Derived average new house prices and construction costs from 1914 to 2023 for The Netherlands. The reconstructed datasets combine real new house values and construction price inflation indices from CBS and KEF and take into account tax and profit %.

We observe that average new home prices have increased significantly over the last century, from 11k to 501k, or 3,6% (CAGR) per year. Construction costs have increased to 308k. By the end of 2023 the average residential residual land value in the Netherlands was therefore 193k, or 38,5% of the total market price.

Land price movements are not constant over time.  In addition volatility increases with longer projection horizons as used in development projects. The graph below presents the 4-year look-back percentual changes of land prices.

residual land value or land price -historical changes
Figure 2. Historical overview of new housing price, construction costs and land price changes over 4 years. Each point represents a change of the respective variable over the last 4 years. Source: calculations by Asset Mechanics

We clearly observe alternating cycles: 6 periods with increasing land prices and 6 periods with decreasing land prices.  The dashed line shows that in 20% of the years land prices decreased with more than 42% over a 4-year horizon. However, the graph also shows 4 periods with land price decreases up to -100%: both World Wars, the ‘70s and the ‘80s. In these periods construction costs increased much more than property prices, which led to falling land prices.

Regime-based land price risk estimation

As land price risk depends substantially on the (macro-economic) context, we want to take this into account when we make risk estimations.  A regime-based risk model2 based onreference class forecasting principles helps us to achieve that. We first identify the macro-economic and other variables that best explain land price movements. Next, for every observed year we select reference classes (years) whereby the (macro-economic) context is most comparable.

When we apply this model over the last 75 years we obtain the following back-test results for a 4-year ahead forecast:

Regime-based land price risk backtest
Figure 3. Backtest of 4-year ahead forecasts of land price changes: expected, upper and lower tails (VaR50, VaR90 and VaR10) versus realised. Base risk estimates are presented as dashed lines, regime-based estimates as solid lines. The black line presents the realised land price changes. Breach frequency for the lower tail is 5.3% for base and 10.7% for regime-based model, where 10% is expected.

We observe that the regime-based model is much more accurate and adaptable than a standard historical risk model. We see that the realised breach frequency is now much more in line with the expected breach frequency. Furthermore, the regime-based model adapts better to changing regimes as can be seen by lower model costs. The regime-model is on average less conservative and more agile. Therefore we become more resilient in changing environments, which enables better decision-making.

Implications for landowners

From the historical graph we saw that land price risk can sometimes be very high. Accurate risk measurement therefore becomes imperative for landowners. However, it is not a trivial task. Depending on the selected method (e.g. base or regime), the resulting risk estimates can differ substantially.

The application of the base method is relatively straightforward; however, it leads to overestimation of risk in most of the years. This implies that buffers would tend to be large even in low-risk periods that do not require it. Furthermore, the base method is stable, i.e. risk levels do not vary substantially over the years. This simplifies risk reporting, but comes at the cost of not being able to leverage risk measurements for strategic decision-making.

Employing a regime-based method, on the other hand, requires more careful consideration of the (macro-economic) context within which one operates. The dynamic assessment of the risk environment enables optimal decision making and buffer allocations. It allows one to use risk metrics to guide when a project should be continued, put on hold or stopped entirely.

So, what does the future hold? Although we do not know the exact direction of land prices, we now have a much clearer and robust view on the risks. This helps us a lot to be more agile in our decision making and to substantiate our decisions with data. Reach out for a free consultation on risk mapping and measurement within your organization. Or see how this model can be used to obtain a complete overview of the risks with the land development risk tool.

  1. Construction costs including VAT and profit margin for the contractor / project developer ↩︎
  2. We apply an ‘optimized (multivariable) similarity-based risk model’. The Regime-Simple model selects the most relevant features from the observed variable itself and captures the business-cycle risk. The Regime-Macro model also includes other relevant macro-economic and business variables. The models are optimized out-of-sample to obtain the most robust forecast of the risk. ↩︎