Evaluating Spatial Model Accuracy in Mass Real Estate Appraisal: A Comparison of Geographically Weighted Regression and the Spatial Lag Model

摘要: AbstractGeographically weighted regression (GWR) has been shown to greatly increase the performance of ordinary least squares-based appraisal models, specifically regarding industry standard measurements equity, namely price-related differential and coefficient dispersion (COD; Borst McCluskey, 2008; Lockwood Rossini, 2011; McCluskey et al., 2013; Moore, 2009; Moore Myers, 2010). Additional spatial such as lag models (SLMs), have improve multiple real estate that suffer from heterogeneity (Wilhelmsson, 2002). This research is performed using arms-length residential sales 2010 2012 in Norfolk, Virginia, compares GWR SLM by extrapolating each model's aggregate subaggregate levels. Findings indicate achieves a lower COD than SLM.(ProQuest: ... denotes formulae omitted.)IntroductionAd valorem property taxes are prominent source government revenue jurisdictions around world. Taxing authorities held accountable ensure these valuations fair equitable. In roles, optimization accuracy mass valuation approaches critical.Because their precision time- cost-saving advantages, methods employ regression-based known automated (AVMs), becoming increasingly practice received attention academic community. AVMs used host industries-both public private- including loan origination, fraud detection, portfolio (Downie Robson, 2007), promoted advanced organizations International Association Assessing Officers (IAAO). Statistical standards equity established give additional benchmarks which modelers may test various methodologies.Academic expanded geographically specific dummy variables distance coefficients, and, although this approach squares (OLS)-based they often still biased coefficients t-scores (Berry Bednarz, 1975; Fotheringham, Brunsdon, Charlton, 2002; McMillen Redfearn, Some researchers (Fotheringham, 2002) (GWR), locally technique, improved model employing weighting function allowing for fluctuate across geographic space (Huang, Wu, Barry, 2010; LeSage, 2004). Similarly, (SLM)-a autoregressive (SAR) model- addresses an autocorrelation weights matrix (Anselin, 1988).Because markets behave differently space, free consideration produce inaccurate, misleading results (Anselin Griffith, 1988; Ball, 1973; Berry 1975). prominently demonstrated throughout literature more accurate alternative analysis (MRA) (for example, SAR sufficiently predictive (Borst 2007; Conway ah, Quintos, Wilhelmsson, Descriptions findings summarized exhibit 1.Despite popularity both housing research, our knowledge, study simultaneously industry-accepted IAAO extrapolates levels yet be published. …