Adaptive Dynamic Model Averaging with an Application to House Price Forecasting

Abstract

Dynamic model averaging (DMA) combines the forecasts of a large number of dynamic linear models (DLMs) to predict the future value of a time series. The performance of DMA critically depends on the appropriate choice of two forgetting factors. The first of these controls the speed of adaptation of the coefficient vector of each DLM, while the second enables time variation in the model averaging stage. In this paper we develop a novel, adaptive dynamic model averaging (ADMA) methodology. The proposed methodology employs a stochastic optimisation algorithm that sequentially updates the forgetting factor of each DLM, and uses a state-of-the-art non-parametric model combination algorithm from the prediction with expert advice literature, which offers finite-time performance guarantees. An empirical application to quarterly UK house price data suggests that ADMA produces more accurate forecasts than the benchmark autoregressive model, as well as competing DMA specifications.

Predicting House Prices Using DMA Method: Evidence from Turkey

“…Another of the first studies that used the DMA method, this time among different countries, is the work of Risse and Kern (2016) that applied it to the six largest countries of the European Monetary Union between the years 1975 and 2015. Afterward, Yusupova et al (2019) developed the adaptive dynamic model averaging (ADMA) methodology, stating that DMA is important in macroeconomic time series prediction due to its ability to accommodate both the time variability in parameters and the specification of the optimal prediction model. In their work with the United Kingdom's regional house price indices from 1982 to 2017, the authors claimed that better prediction results could be obtained through the ADMA methodology.…”

“…While investigating the relationship between the real estate sector and the indicators of countries, inflation rates (Fry et al 2010;Aizenman and Jinjarak 2014;Risse and Kern 2016;Wei and Cao 2017;Yusupova et al 2019;Paul 2019) draw attention as the top macro variable that affects the housing prices. Home loan interest rates (Arsenault et al 2013 As a result of this literature review, the main purpose of this study is to predict the right housing prices, to find the right method to catch the right variables and the right clues, and to reach the most accurate results in the right country by using the available technological facilities.…”

“…This paper implements dynamic model averaging (DMA), a new technique, to predict the movement of housing prices. DMA is gaining increasing attention in macroeconomic time series forecasting due to its ability to accommodate time variation in both the parameters as well as the specification of the optimal forecasting model (Yusupova et al 2019). In addition to giving better results with macro variables, one of the advantages of the DMA method is that it allows the parameters and prediction model to change over time.…”

“…Moreover, the lack of an application of DMA to the Turkish housing market, which constitutes the sample of this study, and the advantages of the model are the main motivations for this study. It is also worth mentioning that, in the case of housing prices analysis, DMA has significant prediction gains compared to a linear autoregressive (AR) model, offering a series of predictions that compete with competitive dynamic and static prediction models (Yusupova et al 2019). The dynamic averaging scheme allows us to obtain the probability that each variable is included over time.…”

“…As seen from different DMA studies (Raftery et al 2010;Aye et al 2015;Bork and Moller 2015;Risse and Kern 2016;Wei and Cao 2017;Sousa 2018;Yusupova et al 2019), it is clear that the DMA model is generally applied for the analysis of datasets from developing countries. Turkey is considered a developing country, and the application of this method in Turkey is another contribution to the literature.…”