Journal of Economic and Social Measurement , vol. 45, no. 1, pp. 65-81, 2020
We provide a new estimate of household-level housing wealth in Norway between 1993 and 2015 using an ensemble machine learning method on housing transaction data. The new housing wealth measure is an improvement over existing data sources for two reasons. First, the model outperforms previously applied regression models in out-of-sample prediction precision. Second, we extend the sample of estimated housing wealth by including cooperative units, non-id apartments, and cabins.