As a part of my studies on BI Norwegian Business School, I wrote a Master Thesis, together with a fellow student, Ketil Nereng. The topic: “Forecasting rental rates for Norwegian commercial real estate”. The study uses rental rate statistics from Eiendomsverdi Næring, a Norwegian firm solely working with gathering and producing real estate statistics. The goal of the study was to identify key determinants of changes in real rents and producing a forecasting model able to outperform a random walk and yield significant forecasts.
Why are these studies important?
Because too much of the future market rents are a function of today’s rent, conjecture and speculation on future rent levels. Of course a qualitative approach is an important aspect in establishing prognosis on rent levels, but a more quantitative approach is required. Rent levels in the Oslo market have showed a highly cyclical movement for the period in which data is available. That further implies that the success of both investments and the negotiation of lease contracts will be, to a certain point, dependent on the timing. In Norway, the Financial Supervisory has asked for a more methodological framework in valuations, both regarding rent levels and discount rates. (Financial Supervisory Report of 20th December 2010).
Further on, academic research within this field in Norway is a rarity, as opposed to the UK, US and Australia, where a larger part of the nations’ real estate portfolios are listed on stock exchanges and data is more readily available. We sought to create a starting point, on which further future research can be based upon.
What are the possibilities of such a model?
A better reference point when setting future rent levels, of course in combination with qualitative assessment. This could be an aid as an investment tool, when doing valuations and in analysis. Further on the identification of key determinants of changes in real rents gives a clear advice on which variables to follow when doing analysis and forecasts.
Academic references and methodology
The thesis is based on relevant literature of mainly US and UK based studies that apply a wide variety of econometric techniques to test several hypotheses. Chris Brooks and Sotiris Tsolacos have written a book on real estate modeling and forecasting. Several of the forecasting models that have been successful in other markets, are developed by these two, and have been tested by us in the Oslo market. The forecasting models examined are a classical linear regression model, an autoregressive moving average model and a vector autoregressive model.
What are the results?
As of today, the data series in the Norwegian (read Oslo) market are too short. The models do give significant results and identifies four key determinants of changes in real rental rates. Changes in previous period’s rents, employment rates, real interest rates and vacancy rates are all significant variables when regressed on change in real rents. However none of the three tested forecasting models are consistently able to outperform a random walk, but a clear trend of improvement in forecast accuracy is detected when gradually increasing the estimation sample.
So when can the models be used?
We saw a clear improvement in the results as the data series got longer. 3-5 years of quarterly rental rate statistics should improve the results significantly and a model that’s worth applying could be ready at hand. It would however require more research and modification of the models to get good enough results.
A copy of the study can be obtained by sending an e-mail to hs@newsec.no