Accurate land valuation is an important input to planning, development, and public investment decisions. The Welsh Government sought to explore whether alternative land valuation methods could offer a more robust foundation for future policy development in Wales. 

Alma Economics was commissioned to produce a real-world proof of concept for land valuation, with the aim of identifying the practical challenges, data requirements, and limitations of different approaches.  

Two statistical methods were applied, conventional regression and machine learning, both using a hedonic pricing model in which property or land price is modelled as a function of its characteristics. Data was drawn from HM Land Registry, the Ministry of Housing, Communities and Local Government (MHCLG), and CoStar. 

Machine learning methods produced more precise estimates than conventional regression. However, the limited number of historical land-only transactions in Wales makes it difficult to apply these methods directly for land valuation in a policy context.

The approach is not currently recommended as a standalone tool, but could be used alongside other valuation methods or where more suitable data becomes available.

Read the report here