Daniel Felsenstein (p), Michael Beenstock , Dai Xieer

Andreas Gohs

We argue that housing affordability is as much about incomes as it is about house prices. Consequently, a comprehensive analysis of housing affordability should be conducted in spatial general equilibrium in which regional incomes and house prices are determined through the specification of labor, capital and product markets in addition to housing markets. A spatial econometric model for Israel is used to study the effects on housing affordability of income generating shocks in labor and capital markets, as well as supply shocks in housing markets. Particular attention is paid to the effects on affordability of planning delays in tendering land for housing construction and the issue of building permits. Spatiotemporal impulse responses for housing affordability show that region-specific shocks, such as accelerated planning permission and the provision of regional investment grants, percolate across the economy as a whole.

Andreas Gohs (p)

Daniel Felsenstein

This study evaluates the accuracy of macroeconomic time series projections for the national and sub-national levels of Germany. Therefore the forecasting performances of several competing time series- and spatio-temporal-methods are compared. For this purpose, employment and real estate market data for Germany are used. They are provided in different aggregation levels of German regions and time domains. As usual, for smaller geographical units, time series are just available in higher periodicity.
The forecasting performances of the following time series methods are reported and compared: Traditional methods like deterministic, linear or non-linear trends, naive forecasts and exponential smoothing (double and Holt-Winters), and procedures of modern time series analysis, Autoregressive Integrated Moving Average (ARIMA) and for panel data. Also advanced procedures are employed, like spatio-temporal procedures, which incorporate the spatial association between neighbour regions.
If spatial dependencies exist between Germany’s labour and real estate markets, knowledge about spillover-effects should possibly improve the efficiency to forecast numbers at the regional and the aggregated national levels.
Not just the accuracy of regional forecasts is of interest. Also pooling strategies to increase forecasting accuracies for national level data are scrutinized. Vice versa, data for the national level are incorporated in regional forecasts.
For time series which are available in quarterly periodicity, forecasts are also done and evaluated in the higher annual periodicity. Since the applicable variant of a time series method might depend on the frequency of the time series to be on hand, this enriches the method choice. Further, comparisons of time series procedures are done for several favoured forecasting horizons.
The performances of the forecasting-procedures are evaluated with several loss functions, Root-Mean-Squared-Percentage-Error (RMSPE), Theil’s inequality coefficient, etc.