Additive Hedonic Regression Models with Spatial Scaling Factors: An Application for Rents in Vienna

This paper is motivated by two common challenges in hedonic price modeling: nonlinear price functions, which require flexible modeling approaches, and the inherent spatial heterogeneity in real estate markets. We apply additive mixed regression models (AMM) to estimate hedonic price equations for rents in Vienna. Non-linear effects of continuous covariates as well as a smooth time trend are modeled non-parametrically through P-splines. Unobserved district-specific heterogeneity is modeled in two ways: First, by location specific intercepts with the postal code serving as a location variable. Second, in order to permit spatial variation in the nonlinear price gradients, we introduce multiplicative scaling factors for nonlinear covariates. This allows highly nonlinear implicit price functions to vary within a regularized framework, accounting for district-specific spatial heterogeneity, which leads to a considerable improvement of model quality and predictive power. Our findings provide insight into the spatially heterogeneous structure of price gradients in Vienna, showing substantial spatial variation. Accounting for spatial heterogeneity in a very general way, this approach permits higher accuracy in prediction and allows for location-specific nonlinear rent index construction.     

Spatial Dependence and House Price Index Construction

Accurate estimation of prevailing metropolitan housing prices is important for both business and research investigations of housing and mortgage markets. This is typically done by constructing quality-adjusted house price indices from hedonic price regressions for given metropolitan areas. A major limitation of currently available indices is their insensitivity to the geographic location of dwellings within the metropolitan area. Indices are constructed based on models that do not incorporate the underlying spatial structure in housing data sets. In this article, we argue that spatial structure, especially spatial dependence latent in housing data sets, will affect the precision and accuracy of resulting price estimates. We illustrate the importance of spatial dependence in both the specification and estimation of hedonic price models. Assessments are made on the importance of spatial dependence both on parameter estimates and on the accuracy of resulting indices.     

Anisotropic Autocorrelation in House Prices

This article examines anisotropic spatial autocorrelation in single-family house prices and in hedonic house-price equation residuals using a spherical semivariogram and transactions data for one county in the Philadelphia, Pennsylvania, MSA. Isotropic semivariograms model spatial relationships as a function of the distance separating properties in space. Anisotropic semivariograms model spatial relationships as a function of both the distance and the direction separating observations in space. The goals of this article are (1) to determine whether there is spatial autocorrelation in hedonic house-price equation residuals and (2) to empirically examine the validity of the isotropy assumption. We estimate the parameters of spherical semivariograms for house prices and for hedonic house-price equation residuals for 21 housing submarkets within Montgomery County, Pennsylvania. These housing submarkets are constructed by dividing the county into 21 groupings of economically similar adjacent census tracts. Census tracts are grouped according to 1990 census tract median house prices and according to characteristics of the housing stock. We fit the residuals of each submarket hedonic house price equation to both isotropic and anisotropic spherical semivariograms. We find evidence of spatial autocorrelation in the hedonic residuals in spite of a very elaborate hedonic specification. Additionally, we have determined that, in some submarkets, the spatial autocorrelation in the hedonic residuals is anisotropic rather than isotropic. The empirical results suggest that the spatial autocorrelation in Montgomery County single-family house-price equation residuals is anisotropic in submarkets where residents typically commute to a regional or local central business district.     

Appraisal Quality and Residential Mortgage Default: Evidence from Alaska

We empirically examine the effect of appraisal quality on subsequent mortgage loan performance using data from the high volatility housing market of Alaska in the 1980s. We develop measures of appraisal quality by computing the residual between a hedonic estimate of house value using available information from other appraisals compared to actual ex ante appraised value. We then estimate proportional hazard models of mortgage default and find that several measures of appraisal quality, particularly appraised value in excess of hedonic estimates, are significantly related to default risk. Using valuations subsequent to loan default, we are also able to evaluate how well house price indices perform in terms of estimating current loan-to-value and offer some additional evidence on the controversy over the role of net equity versus trigger events as determinants of mortgage default. We also show that defaults are related to ex ante measures of housing market conditions, with additional implications for underwriting policies and the current industry trend away from traditional appraisal and toward automated valuation.     

A State Space Model for Berlin House Prices: Estimation and Economic Interpretation

Hedonic regression has become the standard approach for modeling the behavior of house prices. Usually, the common price component is modeled via dummy variables. Based on an approximation for the present value, we deliver an economic interpretation of the common price component. This allows to include explanatory factors like inflation rates, mortgage rates and building permissions. The notional rents for houses are fitted with a flexible hedonic function. We use the EM algorithm to estimate our model with monthly data of single-family house sales from the four South-West districts of Berlin, Germany from 1982:8 to 1999:12. Emphasis is put on the interpretation of the results.     

Frequency of Transaction and House Price Modeling

This article examines the characteristics and price behavior of repeatedly transacted properties. Using data from four U.S. counties, we estimate hedonic price models of properties grouped by transaction frequency, and compare estimated standard deviations and estimated appreciation rates by group.For each of four counties studied, we find that estimated house price appreciation is systematically higher among properties that transact more frequently. One possible explanation for this result is that purchasers make property improvements that are not adequately reflected in the available data.We also find that estimated standard deviations of the disturbance term show a marked decrease as the frequency of transaction increases. Since frequently transacting properties are not found to be systematically more homogeneous than seldomly transacting properties, we do not attribute this to any increase in homogeneity for frequently transacting properties, but rather to the length of time elapsed between transactions of properties.The findings of this article suggest that repeat-sales price models may need to be adjusted to account for cross-sectional variation in transaction probabilities---that is, the selectivity of the subsample of properties that transacted (or transacted repeatedly) during any finite study period.     

Analysis of Spatial Autocorrelation in House Prices

This article examines spatial autocorrelation in transaction prices of single-family properties in Dallas, Texas. The empirical analysis is conducted using a semilog hedonic house price equation and a spherical autocorrelation function with data for over 5000 transactions of homes sold between 1991:4 and 1993:1. Properties are geocoded and assigned to separate housing submarkets within metropolitan Dallas. Hedonic and spherical autocorrelation parameters are estimated separately for each submarket using estimated generalized least squares (EGLS). We find strong evidence of spatial autocorrelation in transaction prices within submarkets. Results for spatially autocorrelated residuals are mixed. In four of eight submarkets, there is evidence of spatial autocorrelation in the hedonic residuals for single-family properties located within a 1200 meter radius. In two submarkets, the hedonic residuals are spatially autocorrelated throughout the submarket, while the hedonic residuals are spatially uncorrelated in the remaining two submarkets. Finally, we compare OLS and kriged EGLS predicted values for properties sold during 1993:1. Kriged EGLS predictions are more accurate than OLS in six of eight submarkets, while OLS has smaller prediction errors in submarkets where the residuals are spatially uncorrelated and the estimated semivariogram has a large variance.     

Price Indices Based on the Hedonic Repeat-Sales Method: Application to the Housing Market

Shiller (1993) proposes the hedonic repeated-measures (HRM) approach to measuring constant quality price indices for heterogeneous assets such as some bonds and real estate. We derive a mathematical relationship between the coefficients of the HRM model and those from the standard repeat-sales model, and we demonstrate how hedonic characteristics should be chosen for inclusion in the HRM model. Empirical estimates using Fairfax, Virginia, housing transactions data show that the HRM price index evaluated at the mean of the hedonic variable is virtually identical to the standard repeat sales index, just as predicted by our mathematical relationship. But the HRM allows estimation of different price paths for heterogeneous assets. We demonstrate that use of assessed value as the only hedonic characteristic allows parsimonious HRM estimates.     

Valuation in US Commercial Real Estate

We consider a log‐linearized version of a discounted rents model to price commercial real estate as an alternative to traditional hedonic models. First, we verify a key implication of the model, namely, that cap rates forecast commercial real estate returns. We do this using two different methodologies: time series regressions of 21 US metropolitan areas and mixed data sampling (MIDAS) regressions with aggregate REIT returns. Both approaches confirm that the cap rate is related to fluctuations in future returns. We also investigate the provenance of the predictability. Based on the model, we decompose fluctuations in the cap rate into three parts: (i) local state variables (demographic and local economic variables); (ii) growth in rents; and (iii) an orthogonal part. About 30% of the fluctuation in the cap rate is explained by the local state variables and the growth in rents. We use the cap rate decomposition into our predictive regression and find a positive relation between fluctuations in economic conditions and future returns. However, a larger and significant part of the cap rate predictability is due to the orthogonal part, which is unrelated to fundamentals. This implies that economic conditions, which are also used in hedonic pricing of real estate, cannot fully account for future movements in returns. We conclude that commercial real estate prices are better modelled as financial assets and that the discounted rent model might be more suitable than traditional hedonic models, at least at an aggregate level.     

Selectivity,Quality Adjustment and Mean Reversion in the Measurement of House Values*

This paper develops a model of price formation in the housing market which accounts for the non-random selection of those dwellings sold on the market from the stock of existing houses. The model we develop also accounts for changes in the quality of dwellings themselves and tests for mean reversion in individual house prices. The model is applied to a unique body of data representing all dwellings sold in Sweden's largest metropolitan area during the period 1982–1999. The analysis compares house price indices that account for selectivity, quality change and mean reversion with the conventional repeat sales models used to describe the course of metropolitan housing prices. We find that the repeat sales method yields systematically large biased estimates of the value of the housing stock. Our comparison suggests that the more general approach to the estimation of housing prices or housing wealth yields substantially improved estimates of the course of housing prices and housing wealth.     

Spatial Autocorrelations and Urban Housing Market Segmentation

This paper seeks to let data define urban housing market segments, replacing the conventional administrative or any pre-defined boundaries used in the previous housing submarket literature. We model housing transaction data using a conventional hedonic function. The hedonic residuals are used to estimate an isotropic semi-variogram, from which residual variance–covariance matrix is constructed. The correlations between hedonic residuals are used as identifier to assign housing units into clusters. Standard submarket identification tests are applied to each cluster to examine the segmentation of housing market. The results are compared with the prevailing structure of market segments. Weighted mean square test shows that the defined submarket structure can improve the precision of price prediction by 17.5%. This paper is experimental in the sense that it represents one of the first attempts at investigating market segmentation through house price spatial autocorrelations.     

The Hierarchical Trend Model for Property Valuation and Local Price Indices

This paper presents a hierarchical trend model (HTM) for selling prices of houses, addressing three main problems: the spatial and temporal dependence of selling prices and the dependency of price index changes on housing quality. In this model the general price trend, cluster-level price trends, and specific characteristics play a role. Every cluster, a combination of district and house type, has its own price development. The HTM is used for property valuation and for determining local price indices. Two applications are provided, one for the Breda region, and one for the Amsterdam region, lying respectively south and north in The Netherlands. For houses in these regions the accuracy of the valuation results are presented together with the price index results. Price indices based on the HTM are compared to a standard hedonic index and an index based on weighted median selling prices published by national brokerage organization. It is shown that, especially for small housing market segments the HTM produces price indices which are more accurate, detailed, and up-to-date.     

House Price Index Construction in the Nascent Housing Market: The Case of China

Most existing house price index construction methods are developed mainly based on transaction data from the secondary housing market, and are not necessarily suitable for the nascent housing markets where a predominant portion of housing transactions are new units. Using the booming market in China as an example, we evaluate and compare the performances of three most common house price measurement methods in the newly-built housing sector, including the simple average method without quality adjustment, the matching approach with the repeat sales modeling framework, and the hedonic modeling approach. Our analyses suggest that the simple average method fails to account for the substantial complex-level quality changes over time of sales during our sample period, and the matching model fails to control for the effect of developers’ pricing behaviors when adopted in the newly-built sector, hence both are downward biased. Based on this finding, we apply a hedonic method, which allows us to control for both quality changes over time of sales and developers’ pricing behaviors, to 35 major newly-built housing markets and provide the first multi-city constant-quality house price index in China. The new index reveals that the current Chinese housing market is facing a greater risk of mispricing than reported by the existing official metrics.     

The Value of Housing Characteristics: A Meta Analysis

This paper provides a meta regression analysis of the nine housing characteristics that are appear most often in hedonic pricing models for single-family housing: square footage, lot size, age, bedrooms, bathrooms, garage, swimming pool, fireplace, and air conditioning. Meta regression analysis is useful for comparing the estimated regression coefficients from different studies. The goal in this study is to determine if the estimated coefficients vary by geographical location, time, type of data, and model specification. The results show that the estimated coefficients for some characteristics vary significantly by geographical location. These include square footage, lot size, age, bathrooms, swimming pool, and air conditioning. Controlling for time shows that the effects of these housing characteristics on house price have not changed over time. Controlling for type of data produces differences in coefficients for bathrooms. Controlling for wealth as measured by median household income has no significant impact on the coefficients for the housing characteristics. If the study controlled for square footage, the coefficients for lot size decrease. Controlling for the size of the hedonic model affects the coefficient for square footage.     

Boom–Bust Cycles and the Forecasting Performance of Linear and Non-Linear Models of House Prices

The tremendous rise in house prices over the last decade has been both a national and a global phenomenon. The growth of secondary mortgage holdings and the increased impact of house prices on consumption and other components of economic activity imply ever-greater importance for accurate forecasts of home price changes. Given the boom–bust nature of housing markets, nonlinear techniques seem intuitively very well suited to forecasting prices, and better, for volatile markets, than linear models which impose symmetry of adjustment in both rising and falling price periods. Accordingly, Crawford and Fratantoni (Real Estate Economics 31:223–243, 2003) apply a Markov-switching model to U.S. home prices, and compare the performance with autoregressive-moving average (ARMA) and generalized autoregressive conditional heteroscedastic (GARCH) models. While the switching model shows great promise with excellent in-sample fit, its out-of-sample forecasts are generally inferior to more standard forecasting techniques. Since these results were published, some researchers have discovered that the Markov-switching model is particularly ill-suited for forecasting. We thus consider other non-linear models besides the Markov switching, and after evaluating alternatives, employ the generalized autoregressive (GAR) model. We find the GAR does a better job at out-of-sample forecasting than ARMA and GARCH models in many cases, especially in those markets traditionally associated with high home-price volatility.     

International house price cycles,monetary policy and credit

We evaluate three alternative predictors of house price corrections: anticipated tightenings of monetary policy, deviations of house prices from fundamentals, and rapid credit growth. A new cross-country measure of monetary policy expectations based on an international term structure model with time-varying risk premiums is constructed. House price overvaluation is estimated via an asset pricing model. The variables are incorporated into a panel logit regression model that estimates the likelihood of a large house price correction in 18 OECD countries. The results show that corrections are predicted by increases in the market’s forecast of higher policy rates. The estimated degree of house price overvaluation also contains significant information about subsequent price reversals. In contrast to the financial crisis literature, credit growth is less important. All of these variables help forecast recessions.     

Evaluating the Usefulness of the American Housing Survey for Creating House Price Indices

The American Housing Survey (AHS) is a valuable source of information on houses and occupants over time. The AHS has several advantages over sales data for use in the creation of price indices: it is readily available, has frequent observations over time and space, has data from the late 1970s through the mid-1990s, includes houses that do not sell, as well as those that do, and has information on the occupants. The drawbacks include: a time lag between the interview and the release of the data, data suppression issues, owner-stated house values, and a lack of neighborhood information. In this study, we use the metropolitan version of the AHS, which has been supplemented with the original survey data as well as Census tract data for three cities over 14 years to examine whether the AHS can be used to create indices. Indices are estimated using hedonic, repeat valuation, and hybrid techniques, overcoming some of the problems inherent in the estimation of indices. We find that the data-suppression issues and the owner-stated house values are not problematic. The biggest drawback of the AHS is its lack of objective information on neighborhood quality.     

A horse race of models and estimation methods for predicting bankruptcy

We use a comprehensive set of performance metrics to analyze the improvement in the classification power and prediction accuracy of various bankruptcy prediction models after adding governance variables and/or varying the estimation method used. In a sample covering bankruptcies of U.S. public firms in the period 2000 to 2015, we find that the addition of governance variables significantly improves the performance of all bankruptcy prediction models. We also find that the additional explanatory power provided by governance measures improves the further the firm is from bankruptcy, which suggests that governance variables may provide earlier and more accurate warning of the firm's bankruptcy potential. Our findings show that the performance of any bankruptcy prediction model is significantly affected by the estimation method used. We find that regardless of the bankruptcy model, hazard analysis provides the best classification and out-of-sample forecast accuracy among the parametric methods. Furthermore, non-parametric methods such as neural networks, data envelopment analysis or classification and regression trees appear to provide comparable and sometimes superior classification accuracy to hazard analysis. Lastly, we use the dynamic panel generalized methods of moments model to address concerns raised in prior studies about the susceptibility of similar studies to endogeneity issues and find that our findings continue to hold.     

Three New Real Estate Price Indices for Geneva,Switzerland

This paper develops constant-quality price indices for three categories of real estate-apartment buildings, vacant land, and condominiums—for the city of Geneva, Switzerland. We use both the hedonic and repeat sales models to estimate the price level and, in turn, the rate of price change. The general pattern of each series suggests that real estate prices in Geneva were fairly stable throughout the 1970s, increased sharply during the 1980s, but gave back some of these gains in the early 1990s. Interestingly, the sharp rise in prices in the second half of the 1980s is very similar to that found in some regions of the United States. We also consider the problem, implicit in the repeat sales method, of revisions in previously estimated price indices as additional data become available in later years.     

Off-balance sheet exposures and banking crises in OECD countries

Against the background of the acknowledged importance of off-balance-sheet exposures in the sub prime crisis, we seek to investigate whether this was a new phenomenon or common to earlier crises. Using a logit approach to predicting banking crises in 14 OECD countries we find a significant impact of a proxy for the ratio of banks’ off-balance-sheet activity to total (off and on balance sheet) activity, as well as capital and liquidity ratios, the current account balance and GDP growth. These results are robust to the exclusion of the most crisis prone countries in our model. For early warning purposes we show that real house price growth is a good proxy for off balance sheet activity prior to the sub-prime episode. Variables capturing off-balance sheet activity have been neglected in most early warning models to date. We consider it essential that regulators take into account the results for crisis prediction in regulating banks and their off-balance sheet exposures, and thus controlling their contribution to systemic risk.