The Journal of Real Estate Finance and Economics

Published by Springer Nature
Online ISSN: 1573-045X
Print ISSN: 0895-5638
Learn more about this page
Recent publications
Results for Going-In Cap Rate and State Governor
Results for Going-In Cap Rate, State Governor, and Political Attitudes
We investigate the impact of governmental restrictions on the short-term risk perception, as proxied by the going-in cap rate, of investors in regional and neighborhood shopping centers. We use the COVID-19 pandemic as a natural experiment and proxy for the length and severity of COVID-19 restrictions with the political affiliation of state governors. Using a sample of 40 metropolitan statistical areas (MSAs) across 27 states over the period of 2018 to 2021, we find that for states with Republican governors, which proxy for shorter and fewer COVID-19 restrictions, investors in regional malls required a lower going-in cap rate in the pandemic period than for states with Democratic governors. This effect does not exist for neighborhood shopping centers, whose tenants were not as affected by COVID-19 restrictions. Robustness checks suggest that our findings can be explained with mask mandates as one type of governmental restrictions, and that COVID-19 related restrictions do not impact the long-term risk perception of retail real estate investors. We furthermore find that the political attitudes of an MSA have an impact on investor risk perception.
Distribution of Appraisal Errors. Notes: The density plot shows the distribution of the raw residuals (appraisal errors) for all property types and for each property type individually. The dotted horizontal line marks the null point on the x-axis
Bootstrap Distribution of Model Performance. Notes: The density plot shows the bootstrap distribution of the model performance for all five models using 1,000 random bootstrap samples. A performance improvement occurs whenever the ratio σAppraisalσBoosting>1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{\upsigma }_{\mathrm{Appraisal }}}{{\upsigma }_{\mathrm{Boosting }}} >1$$\end{document}, as indicated by the dotted horizontal line. The area to the right of the dotted line can be interpreted as the confidence interval for which the null hypothesis σAppraisalσBoosting≤1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{\upsigma }_{\mathrm{Appraisal }}}{{\upsigma }_{\mathrm{Boosting }}} \le 1$$\end{document} can be rejected. The null hypothesis can be rejected at a 5% level of significance for all models and at a 1% level of significance for all models except for the retail model. The respective ratios measured by 10-fold cross-validation are presented in Table 6
Comparison of Residual Variation. Notes: The boxplots show the distribution of the raw appraisal errors (solid line) in comparison to the boosted appraisal errors (dashed line). The box of each boxplot represents 50% of the data within the 25th and 75th percentile. The bold line within the box indicates the median of each distribution. The whiskers indicate the 1.5 interquartile range (IQR). The dotted horizontal line marks the null point on the y-axis
Relative Permutation Feature Importance. Notes: The bar chart shows the relative permutation feature importance of both components Xa and Xb (indicated by the linetype) and the various feature clusters described in the "Explanatory Variables" section (indicated by the color) for each of the five models. The relative importance on the y-axis indicates the relative contribution of each component and cluster to the reduction of the prediction error. The order of groupings is arbitrary
In this article, we examine the accuracy and bias of market valuations in the U.S. commercial real estate sector using properties included in the NCREIF Property Index (NPI) between 1997 and 2021 and assess the potential of machine learning algorithms (i.e., boosting trees) to shrink the deviations between market values and subsequent transaction prices. Under consideration of 50 covariates, we find that these deviations exhibit structured variation that boosting trees can capture and further explain, thereby increasing appraisal accuracy and eliminating structural bias. The understanding of the models is greatest for apartments and industrial properties, followed by office and retail buildings. This study is the first in the literature to extend the application of machine learning in the context of property pricing and valuation from residential use types and commercial multifamily to office, retail, and industrial assets. In addition, this article contributes to the existing literature by providing an indication of the room for improvement in state-of-the-art valuation practices in the U.S. commercial real estate sector that can be exploited by using the guidance of supervised machine learning methods. The contributions of this study are, thus, timely and important to many parties in the real estate sector, including authorities, banks, insurers and pension and sovereign wealth funds.
Propensity Score Matching. Note: Propensity score matching is a statistical analysis of the observational data of the agents (such as experience, leader, age, educational level and the substitutes variable of either Log of Total Commission Periods 2 & 3 (TC) or Total Transactions Periods 2 & 3 (TC), this statistical matching technique attempts to estimate the effect of adopters signing up for the PAR and the control group. This technique accounts for the covariates that predict receiving the treatment. Figure 1 has two sets of propensity score matching charts. Chart 1, on the left, the variable used TC, while Chart 2, the right chart, substitute TC with TT. Chart 2 uses TT as a variable, is part of the sensitivity test. From the charts, both TC and TT are more closely aligned in the matched status versus the unmatched status. These are exceptionally important as the objective is to reduce biasness in the results between treated and control groups. Essentially, both adopters and non-adopters performance either in TC or TT are more closely aligned.
Adopter versus Non-Adopter Total Commissions Earned. Note: Fig. 2 illustrates the differences in performance after those agents adopt the PAR. The figure shows the results of the standard treated and control groups totalling 718 agents or 359 on each group. Periods 1, 2 and 3, denote the time before the launch of PAR; for Periods 4 and 5, it is the duration of the author studies of the effect of PAR on the performances of the agents who has adopted this technology disruption in OT. The chart ‘Commission’ refers to average amount of commission earned by each of the agents in either the control or treated group. As for ‘Transactions’ chart (on the right), it refers to the average number of transactions per agent in either the control or treated group
Sensitivity Test Charts of Adopters versus Non-Adopters. Note: Fig. 3 is derived from the sensitivity test that the author has conducted. There are 688 agents selected for this test. It includes 344 agents who are adopters of PAR and 344 agents who are not adopters. From the charts, ‘Commission’ and ‘Transactions’, it shows the performances of agents before and after the adoption of the PAR. The dollar value on the ‘Commission’ chart in Fig. 3 refers to the average commission earned for each period for each agent under either the treated and control groups. For the next chart, ‘Transactions’, it refers to the average number of transactions per period for either the treated or the control groups
The real estate agency industry has seen the emergence and growth of disruptive technology and innovation to the extent that real estate agents view search portals such as Zillow and Purplebricks as serious competition. In response to these threats, a major real estate agency in Singapore, OrangeTee, launched a property agent review and rating program called Property Agent Review (PAR) to provide better information on their agents for prospective clients. The PAR program provides a natural experiment to test the effect of informative reviews and ratings on agent performance in terms of commissions and transactions. This is done via a difference-in-difference approach, carefully controlling for observed agent characteristics and market conditions. This paper also analyses the informativeness of reviews.
Abnormal returns (AR) around REIT conversion announcement day. This graph shows the abnormal returns (AR) around the REIT conversion announcement day. We use both the Fama-French three-factor and the CRSP-Ziman total return REIT index as benchmarks. This approach uses the two-step framework where the model parameters are estimated using data from outside the event window. The estimation period ends 46 days before the event date and has a minimum estimation length of 3 days and a maximum estimation length of 255 days. The AR for the event window (−10, +10) around conversion events are reported.
Abnormal returns (AR) around de-REIT conversion announcement day. This graph shows the abnormal returns (AR) around the de-REIT conversion announcement day. We use both Fama-French three factors and CRSP Ziman REIT index as our benchmark. This approach uses the two-step framework where the model parameters are estimated using data from outside the event window. The estimation period ends 46 days before the event date. It has a minimum estimation length of 3 days and a maximum estimation length of 255 days. The AR for the event window (−10, +10) around conversion events are reported.
Sub-sample Abnormal returns (AR) around REIT conversion announcement day. This graph shows the Fama-French three-factor abnormal returns (AR) around the REIT conversion announcement day. This approach uses the two-step framework where the model parameters are estimated using data from outside the event window. The estimation period ends 46 days before the event date. It has a minimum estimation length of 3 days and a maximum estimation length of 255 days. The sample is divided by prior conversion dividend and tax ratio, respectively. The AR for the event window (−10, +10) around conversion events are reported
Sub-sample Abnormal returns (AR) around de-REIT conversion announcement day. This graph shows the Fama-French three-factor abnormal returns (AR) around the de-REIT conversion announcement day. This approach uses a two-step framework where the model parameters are estimated using data from outside the event window. The estimation period ends 46 days before the event date. It has a minimum estimation length of 3 days and a maximum estimation length of 255 days. The sample is divided by pre-conversion dividend adjusted operating cash flow and pretax income, respectively. The AR for the event window (−10, +10) around conversion events are reported.
We explore the determinants and value implications of publicly traded real estate companies converting to real estate investment trusts (REITs), which we term REITing, and publicly-traded REITs giving up their REIT status, termed de-REITing. Non-REIT real estate firms that pay relatively high dividends and have high income tax ratios are more likely to convert to a REIT; while REITs that have lower pretax incomes, dividend adjusted operating cash flows, and higher leverage ratios are significantly more likely to de-REIT. REITing generates significant positive abnormal returns (ARs) around the REITing announcement. These positive ARs are concentrated in firms with higher income tax liabilities and firms paying larger dividends pre REIT-conversion. De-REITing announcements generate significant negative ARs, which are mitigated when the de-REITing firm has low potential tax liabilities, or when the firm is cash flow constrained with respect to its dividend payment. Based on these results, we argue that the degree to which REITing (de-REITing) decisions are value generating (destroying) depends on the magnitude of potential tax and dividend implications. We also examine the longer run valuation effects of REITing and de-REITing decisions and find no evidence of a reversion of the short-run announcement effects.
Despite having abundant literature blaming a faulty financial system and exuberant price expectations as the primary causes of housing bubbles, there is a lack of research that looks at the impact of house price instability on the economy. This study aims to fill this gap by thoroughly examining the connection between house prices and economic output, and the effect of house price volatility on economic stability. Drawing from long-spanning quarterly data from 17 OECD countries from 1970 to 2019, the study develops and tests economic growth and volatility models to uncover significant insights. The empirical results show that house price returns have a significant asymmetric impact on economic growth, with negative returns having twice the effect of positive ones. Moreover, the results indicate that house price volatility significantly contributes to economic instability. In light of these findings, the paper concludes with valuable policy recommendations to enhance the housing market and improve overall economic stability. This study provides a compelling argument for the importance of closely monitoring and regulating the real estate market in order to maintain a healthy and stable economy.
We exploit the local variations crime across time to measure the effect of local crime on house prices. We have access to a very granular micro-dataset consisting of real estate listings for sale in Buenos Aires City, Mexico City and São Paulo. We follow an IV procedure and instrument local crime with the homicide rate. We find that our IV estimates for local crime are greater than their corresponding pseudo-panel linear model coefficient estimates. Second, we show that the direct crime elasticity to house prices is robust to varying sizes of the regular polygons. However, the spillover effects of crime on house prices fade away when the size of the polygons is sufficiently large.
The impulse response function is based on a VECM model, in which the endogenous variables are non-housing consumption, housing consumption, and housing mortgage leverage. Non-housing consumption is measured by China’s total retail sales of consumer goods (TRSCG) (The National Statistics Bureau of China classifies TRSCG into food, clothing, household utilities, medical services, transportation, education, and miscellaneous services. TRSCG is an appropriate measure of non-housing consumption (Yang et al., 2018)), and housing consumption is measured by China’s residential sales. Following Wei and Chen (2017), housing mortgage leverage is measured by the ratio of housing mortgages divided by the sum of down payments plus housing mortgages. The housing mortgage and down payment indicators are two components in real estate enterprises’ sources of funds. All raw samples are at a monthly frequency, with a sample period from Jan. 2007 to Dec. 2018. This sample period was chosen to ensure the integrity of the overall samples since samples for calculating housing mortgage leverage before 2007 are missing. We first take the logarithms of the raw samples and then normalize the logarithms into the [0,1] interval. Moreover, we set lag 5 in the VECM model by AIC. The data source is the RESSET database (See the RESSET database website:
This figure depicts the characteristics of the housing-mortgage-leverage factor loadings. The top half corresponds to the estimated results based on the SHSE, and the bottom half corresponds to the results based on the SZSE. According to the order from left to right, the corresponding models employed are LCAPM, Epstein-Zin-LCAPM, and LCAPM-BM
This study detects the linkage between housing mortgage leverage and stock asset pricing in China’s A-share stock market. We deduce relevant asset pricing models in which a significant pricing factor—termed the housing-mortgage-leverage factor and measured by the growth rate of housing mortgage leverage—is included. Based on these models, corresponding empirical tests on the role of housing mortgage leverage in stock asset pricing are conducted in the Chinese A-share stock market. Congruently, two significant results are presented. First, the housing mortgage-leverage factor positively correlates with excess stock returns, and the price of the housing mortgage-leverage risk is positive, giving rise to the premium associated with fluctuations in housing mortgage leverage. Second, the housing mortgage-leverage factor accounts for variations in cross-sectional stock returns and explains the size effect to some extent. On further reflection, an excessively rapid increase in housing mortgage leverage can somewhat result in dampening stock investments, in which smaller (larger) stocks suffer higher (lower) degrees of suppression.
No. of buyers from Mainland China in the housing market of Hong Kong (1995Q1-2015Q2)
The number is compiled based on raw data from EPRC
This study examines how heterogeneous traders on both sides of transactions behave in the housing market under information asymmetry. Two types of buyers, namely, informed and uninformed buyers, correspond to local and non-local buyers in the empirical tests. Non-local housing buyers in Hong Kong pay a 2.8% premium over local buyers in the second-hand market for housing units with similar observable attributes. We distinguish real estate developers from sellers in the second-hand market. The former has an incentive and ability to reduce information asymmetry by providing a quality guarantee on the building structure and signaling with brand name, none of which can be fulfilled by sellers in the second-hand market. Empirical results show that developers’ efforts to reduce information asymmetry allow them to fetch a higher price than sellers in the second-hand market, holding property characteristics constant. Such efforts are particularly valued by uninformed buyers as non-local buyers prefer to purchase in the first-hand market rather than in the second-hand market, especially when the problem of information asymmetry is serious.
Location Map of the Treatment and Control Group Roads. The figure shows the geographical map of the area in Bangalore (India) that contains the treatment group roads (those selected under the Tender S.U.R.E project) and control group roads (adjacent roads not selected for improvement project). Many of the roads selected for improvements under the first phase of the Tender S.U.R.E project connect key intersections and the control group roads have been identified as feeder roads to the same key intersections
Common Trends in Property Values on Treatment and Control Group Roads. This figure depicts common trends in property values on Treatment group roads (those impacted by the Tender S.U.R.E project) and on control group roads (those that were not impacted by the project but are adjacent to treatment group roads) during the period July 1, 2011 to December 31, 2016. Trends represent predicted values drawn from the hedonic models described in Tables 4 and 5 for a representative multi-storey apartment and an office property respectively. Median values in the data are used for all other continuous hedonic variables. Vertical lines represent the start of the different phases of the project. Panel A: Trends in Sale and Rental Value for a Representative Multi-storey Apartment. Panel B: Trends in Sale and Rental Value for a Representative Office Property
Prior studies show infrastructural improvements impact property values positively. The effect is often reflected soon after the announcement and continues until the project is complete. These studies, however, are primarily set in developed countries. Emerging markets pose unique risks where uncertainty around implementation and funding could dampen these positive effects significantly. We utilize a quasi-natural experiment around an inner-city road redesign and improvement project in Bangalore, the fastest growing city in India. We exploit the difference-in-difference regression approach to examine the project’s impact on residential and commercial property values around two sets of geographically proximate roads, one of which was chosen for the redesign. Unlike in developed countries, we find that property values are unaffected by the announcement when uncertainty is the highest but start reflecting the positive value of the infrastructure once construction starts and show significant gains upon completion. Our findings carry important policy implications for structuring value capture strategies in emerging markets.
Location of properties (purchased by foreign and domestic investors. Note: The figures show the locations of properties bought by foreign investors (stars) and domestic investors (dots). (a) Los Angeles and Paris. (b) Toronto and London. (c)Tokyo and Sydney
Marginal effect of ForeignBuyer (baseline). Note: The figure shows the marginal effect of ForeignBuyer conditional on the level of CUMINV. The estimated coefficients are taken from Table 3, column (2)
Difference between time-variant seller effects and buyer effects: The cases of equity funds and pension funds. Note: The figure shows the difference between (i) coefficients on time-variant seller investor type dummies and (ii) coefficients on time-variant buyer investor type dummies. Specifically, the figure shows the differences for equity funds and pension funds. The estimated coefficients are taken from Table 3, column (4)
Average and standard deviation of the difference between time-variant seller effects and buyer effects. Note: The figure shows the averages and standard deviations of the differences between (i) coefficients on time-variant seller investor type dummies and (ii) coefficients on time-variant buyer investor type dummies. We calculate the differences for all the types of investors. The estimated coefficients are taken from Table 3, column (4)
In this paper, we examine the role of international capital flows in real estate prices by quantifying the extent foreign buyers overpay for their realty investments as well as the spillover effect of such behavior on property prices domestic buyers pay. Using a unique dataset accounting for about 30,000 realty investment transactions in Australia, Canada, France, Hong Kong, Japan, the Netherlands, the United Kingdom, and the United States, we find the following. First, foreign investors pay significantly higher prices than domestic investors, even after taking a wide variety of controls into account. Second, this paying over the odds becomes smaller the larger the buyers’ exposure to realty investments in the host countries. These results indicate that foreign investors are overcharged when they are less informed about the property market and that the extent to which they are overcharged decreases the more investment experience they have. Third, we did not find any significant spillover effects from overpaying by foreign investors to real estate prices in host countries. This finding is consistent with a group of extant studies employing aggregate-level data to examine the link between international capital flows and real estate prices.
Information Shocks
The efficiency of the real estate market is a major concern for homeowners, investors, lenders, policymakers, and researchers. Modern academic literature has mostly moved beyond an early emphasis on formal tests of informational efficiency. The Grossman and Stiglitz (The American Economic Review 70:393–408, 1980) paradox holds that perfect informational efficiency is impossible and the joint hypothesis problem implies that market efficiency is not even testable. Instead, researchers now commonly examine the speed, accuracy, and persistence of price movements in response to new information, as the allocative efficiency of a market ultimately depends on its degree of informational (and operational) efficiency. This special issue is devoted to exploring these issues.
Imputed distribution of borrower heterogeneity parameters (κ and δ) and their relation with ridge strength parameters (ρ). The κ parameters represent borrower-specific intercepts (or baseline probabilities of default). With a strong penalty, these parameters are all equal to the common, naïve model intercept. As the penalty strength is relaxed, these parameters begin to vary, and eventually suffer from model overfitting. The linear trend parameters (δ) follow a similar relation; however, they converge to 0 with a strong penalty, since the naïve and fixed effects models assume no dynamic heterogeneity
Ridge traces for nonlinear dynamics parameters (α). As with the linear trend model, a strong ridge penalty results in trivial parameter magnitudes and yields the same results as the naïve model. As the penalty strength is relaxed, the average parameter value begins to vary and the smoothing penalty has a non-trivial impact, which overfits for a weak penalty (small ρ2)
Ridge traces of model R² for each penalty function. In all models, a strong penalty results in the same fit as the naïve model. Although all of the models suffer from severe overfitting with weak penalties, only the nonlinear model can achieve a perfect in-sample fit since it has a unique parameter for every observation
Ridge trace of the mean AUC for Model 1 from three random partitions for 10-fold cross-validation. The × marks the optimal value for Δ1 that maximizes the out-of-sample predictive performance. (Δ1∗=0.137\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${\Delta }_{1}^{*}=0.137$\end{document})
Ridge trace of the mean AUCs for Models 2 and 3 from three partitions of 10-fold cross-validation. The ×’s mark the optimal value for Δ2 that maximizes the out-of-sample predictive performance in each model (Δ2∗=0.02\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}${\Delta }_{2}^{*}=0.02$\end{document} and 0.009 respectively for the linear and smooth dynamics imputations)
The determinants of mortgage default have been an area of rising interest since the 2008 recession. There are two distinguishing features of mortgage default analysis. First, predictor variables are often only recorded at origination. However, many important variables such as credit scores vary over time. Second, there are omitted variables (such as borrower’s income and job security). If omitted variables are correlated with included regressors or if only origination values are used in a dynamic model, then biases may be present in econometric models for default risk. Our focus is to develop a ridge regression model to impute the dynamics of time-varying predictors and to capture unobserved borrower heterogeneity. The model is evaluated using cross-validation, and the relevant parameters are tuned to maximize out-of-sample predictive performance. After allowing for imputed dynamics and borrower heterogeneity, we find that the loan-to-value ratio becomes a larger signal of default risk and that credit scores as well as full documentation become smaller signals of default risk. These changes primarily are driven by imputing static variables, rather than dynamics, and may pertain to either omitted liquidity factors or strategic factors.
Geographical distribution of HOS estates and the shares of S2P transactions. Note: The blue dots in this figure plots the geographical distribution of the HOS estates in the full sample. The darkness of the blue dots indicates the percentages of S2P transactions in total transactions in each estate
Capital constraints are a major obstacle that holds back cash-poor households from purchasing a home. A workaround is to compromise the housing size and quality by buying a starter home one can marginally afford first. This study aims to investigate how capital constraints distort the pricing of starter homes. In Hong Kong, the government builds subsidized starter homes, which can be resold either to any households at full market prices through the privatized submarket or to households of limited affordability at lower prices through the affordable submarket. The subsidy in the latter case comes from the equity contribution of the government. If there were no capital constraints, the price gap between the two submarkets should simply be the government’s equity. However, our empirical analysis reveals a much smaller price gap, indicating that households with limited affordability are willing to pay a starter home premium in order to relax their capital constraints. Our estimation shows that the premium is in the range of 4.5% to 6.8%, and enlarges when the housing market becomes more unaffordable. The pricing of starter homes is based not only on their quality but also on their ability to relax capital constraints.
Housing price and homeownership rate
This paper examines whether there is a housing disparity between homeowners and renters. Using data from the Chinese Urban Household Survey, we find that homeowners on average have much higher housing quality than renters after controlling for household characteristics, regional factors, location, and time fixed effects, and such disparity increases significantly over time. We also find that the disparity is mainly concentrated on the people who are disabled, divorced, and the people with low education or rural hukou. Furthermore, we investigate possible factors contributing to this increasing disparity. Our results suggest that the disparity is largely driven by increasingly home improvement for owner-occupied houses but not rental houses. In addition, we also find that the 1994-1998 housing reform has contributed to inequality in homeownership between state employees and other groups, especially the disadvantaged groups such as people with rural hukou who are completely excluded from the reform. With housing prices soaring in the past two decades, people who benefited from the housing reform in the 1990s have enjoyed enormous housing appreciation. However, the disadvantaged people who were excluded from the housing reform now live in bad housing conditions. Given the evidence that housing quality is key to both mental and physical health, our findings have important policy implications for the Chinese government.
Housing sentiments and returns. Notes. The blue line represents hedonic HPI returns (in percentage) and the red line represents housing sentiments
Sentiments and future multi-horizon returns. Notes. This figure depicts the coefficient of sentiment βb\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{b}$$\end{document} for horizon b\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$b$$\end{document} ranged from 1 to 12 months in regression (5). The dots on the curve indicates the coefficient βb\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{b}$$\end{document} is significant and no dots means insignificant
This paper examines the impacts of local housing sentiments on the housing price dynamics of China. With a massive second-hand transaction dataset, we construct monthly local housing sentiment indices for 18 major cities in China from January 2016 to October 2020. We create three sentiment proxies representing the local housing market liquidity and speculative behaviors from the transaction dataset and then use partial least squares (PLS) to extract a recursive look-ahead-bias-free local housing sentiment index for each city considered. The local housing sentiments are shown to have robust predictive powers for future housing returns with a salient short-run underreaction and long-run overreaction pattern. Further analysis shows that local housing sentiment impacts are asymmetric, and housing returns in cities with relatively inelastic housing supply are more sensitive to local housing sentiments. We also document a significant feedback effect between housing returns and market sentiments, indicating the existence of a pricing-sentiment spiral which could potentially enhance the ongoing market fever of Chinese housing markets. The main estimation results are robust to alternative sentiment extraction methods and alternative sentiment proxies, and consistent for the sample period before COVID-19.
This study examines the relationship between corporate real estate (CRE) holdings and stock returns before and after the Global Financial Crisis (GFC). We find that (1) the United States and the United Kingdom show a negative relationship before the GFC and positive after the GFC. (2) Firms that pay positive tax or have positive R&D investments are not systematically different from the full sample. This finding cannot support the "scarce capital" theory or the tax incentive explanation, but it is consistent with the “empire building” theory. After the GFC, financial constraints tightened, and both CRE holding and stock returns dropped. (3) European (excluding the United Kingdom) sample shows a positive relationship in the pre-crisis period. This finding is compatible with the "illiquidity premium" theory. However, the association becomes inconclusive in the post-crisis period. (3) The Japanese sample shows a negative association between CRE and stock returns in the pre-crisis period, like the United States and the United Kingdom. However, the relationship becomes statistically insignificant in the post-crisis period, consistent with the theory of financial constraint tightening after the GFC.
Employing a large dataset (at most, the order of n = 106), this study attempts enhance the literature on the comparison between regression and machine learning-based rent price prediction models by adding new empirical evidence and considering the spatial dependence of the observations. The regression-based approach incorporates the nearest neighbor Gaussian processes (NNGP) model, enabling the application of kriging to large datasets. In contrast, the machine learning-based approach utilizes typical models: extreme gradient boosting (XGBoost), random forest (RF), and deep neural network (DNN). The out-of-sample prediction accuracy of these models was compared using Japanese apartment rent data, with a varying order of sample sizes (i.e., n = 104, 105, 106). The results showed that, as the sample size increased, XGBoost and RF outperformed NNGP with higher out-of-sample prediction accuracy. XGBoost achieved the highest prediction accuracy for all sample sizes and error measures in both logarithmic and real scales and for all price bands if the distribution of rents is similar in training and test data. A comparison of several methods to account for the spatial dependence in RF showed that simply adding spatial coordinates to the explanatory variables may be sufficient.
Kernel Density of Rent per Square Foot
Quantile Coefficient Estimates
This paper examines the principle-agency problem between landlords and real estate agents using novel data on rental contracts. Real estate agents are found to obtain higher contract rents by approximately 1% more for themselves (and family members) than for other landlords, which is economically small. The results suggest that the principle-agency program with real estate agents is less of a concern in the rental market than the ownership market. The reason potentially relates to the commission structure, the relatively low effort associated with finding a tenant, the landlord’s ability to evaluate an agent’s performance, and reputation concerns from repeated interactions.
The relationships between the spatial models. Notes: Adapted from ``How to Interpret the Coefficients of Spatial Models: Spillovers, Direct and Indirect Effects", by Golgher, A. B., & Voss, P. R. (2016). Spatial Demography, 4(3), p. 176. Copyright 2016 by Springer International Publishing AG
Map of flooded roads on 30 September 2017. Notes: The study area consists of a large urban area south of HCMC. House projects (white dots) are located across the three districts District 7, Binh Chanh, and Nha Be. The central business district, District 1, is a red bold dot in the north. Sections of flooded roads are shown as pink lines
Difference-in-differences estimation for the flood event on 30 September 2017
House prices recovery after the flood event on 30 September 2017. Notes: **5% level of significance; *10% level of significance
The impact of climate change is devastating in developing countries where flood protection and insurance schemes are limited. Certain parts of Ho Chi Minh City, Vietnam are under the constant threat of inundation due to sea-level rise. We integrate the hedonic property model in a difference-in-differences framework and spatial econometric analysis into a single analytical framework to estimate the economic effect of pluvial flooding. We find prices for affected houses were discounted by 9% after a large flood event on 30 September 2017. This research contributes to the existing literature as follows. First, we study the economic impact of pluvial floods, which has received less attention in existing studies where large and irregular floods are their focus. Second, the inclusion of legal status as a control variable accounts for the unique character of the Vietnamese housing market. Third, we also identify the recovery of house prices after the flood event.
Monthly housing returns over time. Notes: Monthly housing returns in New York (left-hand side) and Los Angles (right-hand side). The sample spans from April 1996 to January 2018. New York and Los Angeles home price data obtained from Zillow. We use the monthly seasonally adjusted Consumer Price Index, from the St Louis Fed’s database FRED, to obtain the series in real terms. CPI is equal to 100 in April 1996. Logarithmic returns calculated as rt = [ln (Pt)-ln (Pt-1)] × 100, where Pt is the real housing price at month t. Real housing returns are in percentage terms
Dynamic correlations of New York with Los Angeles, Chicago, Dallas, and Industrial Production. Notes: The figures represent the time series of the DCC-GARCH estimates of the dynamic conditional correlations of returns between the housing market of New York with Los Angeles (first quadrant), Chicago (second quadrant), Dallas (third quadrant), and Industrial Production (fourth quadrant)
Contagion occurs when cross-market correlation increases because of a shock to one market. Identifying shocks as episodes of house price exuberance, we provide evidence for contagion effects among the largest metropolitan markets in the US. We find that changes in income, interest rates, and unemployment also create contagion effects. These empirical findings are consistent with a model in which shocks to house prices and economic variables relax household down payment constraints and increase household mobility and housing demand. These effects are explored in an equilibrium framework in which house prices and household choices are determined endogenously, and we account for this endogeneity in our empirical study. Our results are robust to various empirical specifications, and we discuss the implications of these findings for households and investors.
This paper examines market and institutional ownership reactions to REIT security issuances. Specifically, we examine the short- and long-term market performance of issuing REITs compared to their non-issuing counterparts. We additionally examine changes in systematic and idiosyncratic risk curtailing from security issuances as well as institutional ownership shifts stemming from the sale of new securities. We find that equity issuances are met with negative market reaction in the event window, while there is no significant immediate market reaction to debt issuances. Buy-and-hold abnormal returns are generally higher for equity issuers in comparison to debt-issuing and non-issuing firms. We also find that equity issuances increase systematic risk exposure while both types of security issuances reduce idiosyncratic risk. Finally, we document that serial issuers of both equity and debt see increases in institutional investment following issuances.
Online businesses have been surging worldwide during the past decade, especially during the recent COVID-19 epidemic. However, the market share of online real estate transactions is still limited, mainly due to the information-asymmetry problem. In this study, we manually collect data on online judicial housing auctions in China, which is currently the largest online real estate market globally, and investigate how information disclosure facilitates real estate transactions. The empirical results suggest that disclosing better quality information online can attract more potential buyers. In particular, providing more comprehensive information such as professional appraisal reports or videos of the property can help to convert buyers’ initial interests into completed transactions and higher sales proceeds. The positive effects of information are particularly strong when combined with offline services, in a more mature online market, and for low-value properties. We also provide preliminary analysis of factors affecting online-information-disclosure quality from both the macro and micro perspectives. We also provide preliminary analysis of factors affecting online-information-disclosure quality from both the macro and micro perspectives.
This paper empirically tests housing market efficiency in the spatial dimension by using the spatial autoregressive conditional heteroskedastic (ARCH) and spatial quantile regression models. The tests were conducted in terms of both housing returns and squared returns (volatility). The sale price data used is from Cook County residential MLS for the years 2010–2016. The main findings are that housing returns are not spatially correlated but squared returns are spatially correlated, and the spatial dependence of squared returns seems to be stronger for higher squared return quantiles.
Local Economic Conditions. This figure illustrates the means and medians of the CBSA-level economic condition variables (GDP Level and GDP Growth) in the sample. Variables have been winsorized at the 1% and 99% tails of the distributions to avoid the influence of extreme observations
CRE Performance and Local Economic Conditions. This figure plots the CBSA-level total return of CRE on the vertical axis against the CBSA-level economic condition variables (log GDP level and GDP growth) on the horizontal axis in the same. The slope, t-statistic, p-value, r-squared, and number of observations of the univariate regression are reported. t-statistics are calculated using standard errors clustered at the CBSA-level. The linear fit and 95% confidence interval are reported. Variables have been winsorized at the 1% and 99% tails of the distributions to avoid the influence of extreme observations
Total Returns Across CRE Portfolios Sorted by Local Economic Conditions. This figure illustrates the mean and median of total return across four CRE portfolios sorted by CBSA-level economic condition variables (log GDP level and GDP growth). High GDP Level (GDP Growth) indicates a CBSA’s log GDP level (GDP growth) is greater than the 70th percentile at a given year, while Low GDP Level (GDP Growth) indicates a CBSA’s log GDP level (GDP growth) is less than the 30th percentage
Cumulative Returns of CRE Portfolios Sorted by Local Economic Conditions. This figure illustrates one-year to three-year cumulative returns of CRE portfolios sorted by CBSA-level economic condition variables (log GDP level and GDP growth). Each year, CBSAs are sorted based on their previous-year log GDP level and GDP growth, and then placed into different groups. For instance, if a CBSA’s log GDP level (GDP growth) is greater than the 70th percentile of the variable in year t, the CBSA is included in the high GDP level (growth) portfolio in year t. If a CBSA’s log GDP level (GDP growth) is less than the 30th percentile in year t, it is included in the low GDP level (growth) portfolio in year t. These portfolios are rebalanced each year. The one-year to five-year cumulative returns within each portfolio are calculated
The Interaction between Local Economic Conditions and Land Supply Elasticity. This figure illustrates the mean and median of total return across four CRE portfolios sorted by CBSA-level economic condition variables (log GDP level and GDP growth) and land supply elasticity (LSE). High GDP level, GDP growth or LSE indicates a CBSA’s log GDP level, GDP growth or LSE is greater than the 70th percentile at a given year, while Low LSE indicates a CBSA’s LSE is less than the 30th percentage
Local economy should be an important determinant of commercial real estate (CRE) performance. This paper empirically examines how the economic conditions of a metropolitan area drive the performance of CRE in the area. This paper shows that areas with better economic conditions provide a higher total return on commercial properties than those with worse economic conditions. Further analysis indicates that both the income return and capital appreciation of CRE are significantly affected by the size of the economy (proxied as GDP level), while the capital return (but not income return) is significantly affected by the growth of the economy (proxied as GDP growth). The results are largely consistent in the Fama–MacBeth regression, the portfolio analysis, and the propensity score matching model, providing solid evidence on the important effects of local economy on CRE.
Market-by-market walkability premiums using Walk Score before (circle) and after (triangle) including SLD location efficiency controls. New York City is excluded because the national scaling of Walk Score gives every complex in that market a score in the high 90s (out of 100), so Walk Score in NYC instruments for fixed market effects
This paper examines U.S. residential consumer willingness to pay for location efficiency, a normative advancement of new urbanism. Drawing on a national sample of multi-family housing data joined to measures of urban form and spatial structure, empirical models suggest three contributions to the literature. First, renters are willing to pay for greater location efficiency and for individual attributes of more efficient locations. Second, renters’ tastes and preferences for location efficiency are spatially heterogeneous. Third, location efficiency data appears to provide a meaningful level of control for locational quality. These contributions extend prior research efforts related to bid rent and urban amenities.
Panel A depicts the relative frequency of buildings’ construction year. Panel B depicts the relative frequency of total mortgage original balance that is defined as the sum across all loan components on the same building. Panel C presents the earliest mortgage origination year that is associated with a building.
Survivor functions.This figure shows Kaplan-Meier time-to-default over a 20-year period for two mortgage groups: mortgages with energy efficient (EE = 1) and non-energy efficient (EE = 0) buildings. The Log-rank test for equality of survivor functions gives a p-value of 0.0001
We investigate the relationship between building energy efficiency and the probability of mortgage default. To this end, we construct a novel panel data set by combining Dutch loan-level mortgage information with provisional building energy ratings provided by the Netherlands Enterprise Agency. Using the logit regression and the extended Cox model, we find that building energy efficiency is associated with a lower probability of mortgage default. There are three possible channels that might drive the results: (i) personal borrower characteristics captured by the choice of an energy-efficient building, (ii) improvements in building performance that could help to free-up the borrower’s disposable income, and (iii) improvements in dwelling value that lower the loan-to-value ratio. We address all three channels. In particular, we find that the default rate is lower for borrowers with less disposable income. The results hold for a battery of robustness checks. This suggests that the energy efficiency ratings complement borrowers’ credit information and that a lender using information from both sources can make superior lending decisions than a lender using only traditional credit information. These aspects are not only crucial for shaping future energy policy, but also have implications for the risk management of European financial institutions.
Using home purchase loan application data, we study buyer responses to the uncommon occurrence of the appraised value coming in below the contract price (i.e. a low appraisal), which sharply raises the probability of downward price renegotiation. We propose that two mechanisms drive the higher renegotiation rates. First, a liquidity channel, visible for financially constrained borrowers for whom a low appraisal impacts financing costs. Second, for financially unconstrained borrowers, we identify a news channel whereby the information content of the low appraisal alone induces borrower renegotiation. Importantly, we show that low appraisals result in lower renegotiated prices through these channels without a substantially lower likelihood of a loan application leading to loan origination or notably longer times from contract signing to sale.
Sales of suppliers to customers that undergo M&As (relative to the sales amount of the year before M&As). This figure shows the average sales of a supplier firm to an acquirer (the customer) around the year that M&A becomes effective. Sales are scaled by the sales amount of the year before M&A effective date. The horizontal axis shows the year relative to the M&A effective year, and the vertical axis shows the normalized sales amount. The sample includes all firms in our final sample, where their primary customers go through at least an M&A deal
Probability of losing customers following M&As. This figure depicts the likelihood of a customer relation ending in years following M&A activities in the customer’s industry. The striped columns show the likelihood of a relation ending when the customer’s industry experiences a low level of M&A activities. The solid columns show the likelihood of a relation ending when the customer’s industry experiences a high level of M&A activities. The horizontal axis shows the number of years after M&As occur. See Fig. 1 for sample descriptions
We present new empirical evidence that higher customer concentration leads to lower corporate real estate holdings at the supplier firm level. Further evidence shows that this effect is causal and more pronounced when the likelihood/impact of losing primary customers is higher or when suppliers have less bargaining power. Finally, we show that firms with a concentrated customer base tend to choose capitalized leasing in lieu of holding real estate.
Evolution of Equity REITs performance (1993 to 2018). This graph illustrates the evolution of the equity REITs in (1) annual total return on the FTSE-NAREIT All Equity REIT Index, (2) FFO adjustment (%), (3) NI-based ROA. Data for annual total return on the FTSE-NAREIT All Equity REIT Index is from NAREIT website, (, FFO adjustment (%), is defined as the difference between FFO and NI scaled by total assets. NI-based ROA is defined as the NI divided by total assets
This paper investigates the non-GAAP performance measures of the REIT industry, specifically the difference (FFO adjustment) between concurrent FFO and Net Income (NI). Using the U.S. Equity REIT data from 1993 to 2018, we first find evidence that both NI and FFO are associated with REIT market-adjusted stock returns, suggesting both contain information that is valuable to investors. Second, we document a significant and positive relationship between contemporaneous FFO adjustment and NI, indicating a possible “selective” and “intentional” inclusion and/or omission of “good” vs. “bad” news in FFO reporting. Third, we find direct evidence that more powerful CEOs are indeed associated with higher FFO adjustments, suggesting CEOs’ involvement in hiding subpar performance. Finally, we focus our attention on a more recent period, when the National Association of Real Estate Investment Trusts (NAREIT) provided additional clarifications and guidelines, and the U.S. Securities and Exchange Commission (SEC) increased scrutiny on FFO reporting. Our results show a diminished “manipulation” for the majority of the REITs, suggesting these guidelines and scrutiny have achieved their intended purposes. While non-GAAP performance measures might supply additional information to investors, our results indicate that providing continuous guidance and monitoring is essential.
Opinion dispersion and returns. Reprinted from Miller (1977): “Risk, Uncertainty and Divergence of Opinion”, Journal of Finance. The ABC curve represents the demand curve for a stock. The supply curve is represented by the vertical line N. The price is determined by the intersection of the demand and supply curves. The FBJ (DBE) curve is the demand curve when divergence of opinion among market participants are relatively high (low)
REIT market performance and NAV dispersion. This figure portrays the relationship between REIT market returns and NAV dispersion. Panel A shows a line plot of the level of the NAREIT All REIT index with a bar graph of the average NAV dispersion over time. Panel B presents a scatter plot of the NAREIT All REIT index returns and NAV dispersion. NAV dispersion is measured as the standard deviation of analysts’ NAV estimates divided by the mean estimate. Analyst NAV estimate data are from S&P Global. Panel A. REIT Market Index and NAV Dispersion. Panel B. Scatter Plot of REIT Market Index Returns and NAV Dispersion
Value of $1 investment in portfolios based on NAV dispersion. This figure shows the value of an initial $1 investment in portfolios of high, medium, and low NAV dispersion. Portfolios are rebalanced at the end of each month. NAV dispersion is measured as the standard deviation of analysts’ NAV estimates divided by the mean estimate. Analyst NAV estimate data are from S&P Global
NAV dispersion: marginal R² contribution analysis. This bar graph shows the marginal R² contribution for the explanatory variables used to explain NAV dispersion in column 1 of Table 6. The marginal R² contribution refers to the decrease in the model’s R² when the variable in question is removed. NAV dispersion is measured as the standard deviation of analysts’ NAV estimates divided by the mean estimate. Analyst NAV estimate data are from S&P Global
In this paper we explore the drivers and implications of divergence in investor opinion of firm value. We use dispersion in analyst estimates of Net Asset Value in REITs as a measure of divergence. We find that divergence in opinion of value is positively associated with portfolio geographic diversity, the presence of international buildings, and firm leverage. Portfolio concentration in tertiary versus gateway markets has no effect on dispersion of value estimates. We find that greater divergence in analyst opinion of value predicts lower stock returns and higher return volatility. Consistent with theoretical predictions from Miller (1977), we find that firms for which investors have the highest disagreement on valuation, pessimistic views of investors are not fully incorporated into prices, resulting in lower future returns.
Shows the Industry size, defined as the sum of funds’ assets under management (AUM) divided by the total market value of REIT stocks over time. It represents the fraction of the total REIT stock market capitalization owned by REMFs included in the sample at the end of the month
Shows the industry size breakdown. The top panel of Fig. 2 shows the assets under management (AUM) of the REMFs in the sample at the end of the month in $ million, over time. The bottom panel shows the total market capitalization of the REIT industry at the end of the month in $ million, over time
Shows the size of the average REMF in the sample, measured by the fund’s assets under management in millions at the end of each month
Shows the number of individual REMFs over time for each month in our sample
This paper investigates the role of scale in Real Estate Mutual Fund (REMF) performance. We test the impact of both fund-level and industry-level economies of scale on fund performance. We provide consistent evidence that industry size erodes REMF performance. After controlling for endogeneity concerns, we document an insignificant relation between fund size and performance. Taken together, these findings suggest that the rapid growth of the REMF industry over the past few decades has materially impacted active managers’ ability to consistently outperform their passive benchmarks. As more capital flows into the industry, competition for alpha increases, and investment opportunities dwindle. This effect is stronger in funds who are especially active in seeking out those increasingly elusive opportunities. Specifically, the effect of industry size on alpha is particularly negative for funds with higher turnover ratios, expense ratios, and volatility of returns.
This study examined contagion involving the aggregate and regional housing markets of the United States (US) with other asset markets using multichannel tests during the 2007–2008 global financial crisis based on a unique high-frequency, i.e., daily data set. To arrive at bias free results several contagion tests: the Forbes and Rigobon (FR) correlation test for contagion, the Fry, Martin and Tang coskewness (CS) test for contagion, the Hsiao cokurtosis (CK) test for contagion and the Hsiao covolatility (CV) test for contagion were employed. At the country level, the linear (correlation) channel indicates that contagion is present from (to) average housing returns to (from) the S&P500, with the correlation contagion also running from average housing returns to REITs. Moreover, the coskewness, cokurtosis and covolatility channels are strongly active with contagion running only from average housing returns to the S&P500, bond returns and REITs. At the Metropolitan Statistical Area (MSA) level, our results indicate that the linear (correlation) channel of contagion is relatively inactive, but the coskewness, cokurtosis and covolatility channels are strongly active with contagion running mostly from housing returns to the S&P500. Our results have important implications for investor and policymakers, given the possibility of differential results based on tests and whether we rely on regional or aggregate data.
Infill development activity Melbourne 2005–2014, distance to CBD. Source: Authors’ calculation
Home sales price effects of infill development in Melbourne, by distance and development type
Infill investments are argued to mitigate environmental footprints, regenerate places and accommodating population growth, but frequently generate local opposition. However, there is a dearth of knowledge around how different types of infill affect different segments of local property markets, how persistent effects are and how far they reach. Using detailed geocoded infill development activity and sales data, we test the price level and trajectory impacts of five infill types, distinguished by the net scale of additional dwellings, on property prices within five concentric 100-meter rings. Using an adjusted interrupted time-series estimation strategy with locality, property and neighborhood characteristic controls we find that developments that generate a net increase in dwellings of four or less, typically result in an appreciation in the average sales prices of proximate dwellings. Moderate and large-scale developments generate negative price effects, but these effects predominantly affect apartments and townhouses, not the dominant detached house submarket. Over time, amenity effects and local market potential may even have a further positive expectation effect on detached house prices. Infill type differentiation shows that urban densification may result in positive affordability outcome in the apartment submarket, but has the opposite effects in the detached house submarket. Divergent price trajectories also contribute to widening wealth disparities.
The Renovation Class of House transactions by share of sales between 2014-Q1 and 2019-Q2. N = 10,350. The figure shows the quarterly results of a classification of sales by renovation level based on real estate listings in Oslo, Norway. Notes: R-1 ∼\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim$$\end{document} unmaintained, R0 ∼\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim$$\end{document} neutral, R1 ∼\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim$$\end{document} partially renovated, R2 ∼\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sim$$\end{document} fully renovated
Renovation Premiums in a housing market (left). Separate hedonic House price Indexes by Renovation class (right) (a) Random Forest Renovation premium (b) Random Forest Index (c) Linear Renovation premium (d) Linear model Index (e) Linear Rolling window Renovation premium (f) Linear Rolling window Index. The figure displays results for the temporal variation in the renovation premiums. Notes: The renovation premium in period t\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t$$\end{document} (figures on the left) is defined as the difference between the estimates of the HPI level of fully renovated (R2) and neutral (R0) units at t\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$t$$\end{document}, for out-of-sample HPI predictions (right-hand side). Similarly for R-1. The confidence interval for the renovation premium is twice the standard error of the differences in the average predictions. The random forest uses the jackknife median standard errors of the predictions
Renovation premium, House price growth and Housing investment growth. The figures display results for the renovation premiums compared to the housing market cycle. Notes: House price and residential investment growth are official estimates from Eiendomsverdi ASA and Statistics Norway, respectively. The premium for unmaintained dwellings is shifted up by 0.07 percentage points for ease of interpretation
Quarterly House Price Growth, with and without Renovation information (a) City total (b) Central (c) East. The figure displays results for the quarterly absolute renovation omission bias (%) of the linear classical hedonic model for the City total, the Central strata, and the East strata. Notes: HPG omitted omit renovation and HPG renovation includes renovation in the hedonic regression
Strata: Random Forest Renovation premiums (a) Central (b) East. The figure displays our results for the temporal variation in the renovation premiums in two urban strata, the most central urban area around the CBD (Central) and a less affluent eastern suburb (East). Note: The methodology is identical to Fig. 2. The out-of-sample dataset for 2014–2015 for the Central region is of size Ncentral,O∈(Q114,Q415)=493\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N}_{central,O\in (Q114,Q415)}=493$$\end{document}, whereas the estimation dataset is of size Ncentral,T=3,212\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N}_{central, T}=\mathrm{3,212}$$\end{document}. Similarly for the East, Neast,O∈(Q114,Q415)=172\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N}_{east,O\in (Q114,Q415)}=172$$\end{document} and Neast,T=1,098\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${N}_{east,T}=\mathrm{1,098}$$\end{document}
We rely on novel textual analysis of real estate listings and identify renovated dwellings in a dataset of Norwegian transactions to estimate the renovation premium in an urban housing market. The renovation premium is estimated in a hedonic framework by classical regression approaches and a random forest algorithm. The strength of the latter is that it allows for a more complex interplay between the renovation premium and explanatory variables. We estimate a significant positive renovation premium of 5–7 percent for renovated dwellings and a negative premium of 9–10 percent for unmaintained/neglected dwellings. These averages mask significant variations in these premiums over time, particularly, a counter-cyclical effect. Omitting renovation information also has implications for estimated short-term house price growth. Unmaintained dwellings tend to transact more in the fourth quarter, indicating that parts of the seasonal price variation reported in the literature are due to compositional variation with respect to renovation. This composition effect bias price movement estimates downward, if uncontrolled for, as unmaintained dwellings transact at significantly lower prices.
This article reviews research on the economics of information in real estate. It covers equity investment in private and public markets and intermediation by brokers. The review shows how, by examining the nature and extent of information frictions in these important markets, research has improved our understanding of potential market failures and corrections which can improve market functioning. © 2022, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
In this paper we analyze market segmentation by firm size in the commercial real estate transaction process. Using novel micro-level data, we look at the probability distribution of investors acquiring a specific bundle of real estate characteristics, distinguishing between investors based on the size of their real estate portfolio. We find evidence of market segmentation by investor size: institutional investors segment across property characteristics based on the size of their real estate portfolio. The probability that a large (small) seller will sell a property to a similar-sized buyer is higher, keeping all else equal. We explore potential drivers of this market segmentation and find that it is mainly driven by investor preferences. During the Global Financial Crisis (GFC), large investors were less likely to buy the ‘average’ property, as compared to the period before or after the crisis, indicating time-varying investor preferences.
The Coronavirus Aid, Relief, and Economic Security (CARES) Act was passed in response to both the global pandemic’s immediate negative and expected long-lasting impacts on the economy. Under the Act, mortgage borrowers are allowed to cease making payments if their income was negatively impacted by Covid-19. Importantly, borrowers were not required to demonstrate proof of impaction, either currently or retrospectively. Exploring the economic implications of this policy, this study uses an experimental design to first identify strategic forbearance incidence, and then to quantify where the forborne mortgage payment dollars were spent. Our results suggest strategic mortgage forbearance can be significantly reduced, saving taxpayers billions of dollars in potential losses, simply by requiring a 1-page attestation with lender recourse for borrowers wishing to engage in COVID-19 related mortgage payment cessation programs. Additionally, we demonstrate the use of these forborne mortgage payments range from enhancing the financial safety net for distressed borrowers by increasing precautionary savings, to buying necessities, to equity investing and debt consolidation.
Daily returns of Covid-19 Risk Factors (CRFs). Note: The Covid-19 Risk Factor (CRF) represents a portfolio long on stocks with high exposure to Covid-19 cases and short on stocks with low exposure to Covid-19 cases in the respective regions. The exposure is based on a Factor model with respect to Covid-19 reported infections. The blue line is based on US data for Covid-19 cases and FTSE-EPRA-NAREIT real estate index. The orange line is based on aggregated Covid-19 cases for Japan, Singapore and Hong Kong as well as on the FTSE-EPRA-NAREIT developed Asia index. We also use the Fama French factors for the respective regions. See the text for explanation of the calculation of the CRFs
Cumulative daily changes of global stock market returns and Covid-19 cases. Note: Left axis is Covid-19 confirmed cases cumulative growth rate since 23 January 2020. Right axis is MSCI World stock market index cumulative daily returns since 23 January 2020
Distribution of daily returns in the cross-section. a Asia Developed. b US. Note: The box plots show the minimum, first quartile, median, third quartile, and maximum of average cross-sectional returns. The box plots are based on around 200 observations for each region. Asia Developed includes Singapore, Hong Kong and Japan. (P) stays for the pre-pandemic period. (C) for the Covid period. The pre-Covid period is from 1 November 2019 until 22 January 2020. The Covid period is from 24 January 2020 until 21 April 2020
Box plots of cross-sectional betas and alphas before and during the pandemic. a Market factor (MF) betas by sector. b Real Estate Factor (REF) betas by sector. Note: A box plot shows the minimum, first quartile, median, third quartile, and maximum of our data. The coefficients are estimated using the factor model in Eq. (2) for two periods using listed real estate companies in Hong Kong, Japan, Singapore, and US. The pre-Covid (P) period is from 1 November 2019 until 22 January 2020. The Covid-19 (C) period is from 24 January 2020 until 21 April 2020
This paper uses the global systemic shock associated with the outbreak of the novel coronavirus Covid-19 to assess the risk-return relationship in the cross-section of real estate equities in the US and in selected Asian countries. I construct regional Covid-19 Risk Factors (CRFs) to assess how the risk exposure of stocks to the pandemic affects their performance. I find substantial differences between stocks in Asia and the US as a result of the pandemic. During the early stages of the pandemic, the sensitivity of Asian real estate companies to the market becomes negative, while it remains positive and increases in the US. Real estate sectors experience strong divergence in performance in the US while little sectoral difference is observed in Asia. The most affected sectors in the US are retail and hotels, while in Asia it is office. A Fama–MacBeth regression shows evidence for a low-risk effect during the Covid period: while insignificant prior to the pandemic, the return-risk relationship becomes significantly negative during the Covid period, with valuation effects driving the results in both regions. Firms in the US perform significantly worse if their exposure to the pandemic is higher, which is not the case in Asia. The results point towards strong divergence of expectations between US and Asian real estate companies in the onset of Covid-19, which may be associated with the level of prior experience to similar pandemics.
We focus on the housing market and examine why nonlocal home buyers pay 12% more for houses than local home buyers. We established a database on the residential housing market for Lafayette and West Lafayette, Indiana, that includes house transactions from 2000 to 2020. The dataset contains highly detailed information on individual buyers and house characteristics. We explain the price differential controlling for arguments such as imperfect information on prices, wealth effects, heterogeneous buyer preferences, and differential search and travel costs across buyers, among others. We estimate a housing demand model that returns heterogeneous marginal willingness to pay parameters for housing attributes. Our results show that nonlocal home buyers are willing to pay more for specific housing attributes, especially for house size, school quality, and house age. We also find that arguments such as gratification, reward, and imperfect price information explain the price differential to a large extent. Search and travel cost arguments have an adverse effect on nonlocal buyers’ house spending.
Trends of operating lease decision and operating lease intensity. This figure presents the trends of the usage of operating lease in the sample period. Panel a presents the trend of the percentages of REITs that opt for operating lease and operating lease on income-generating properties. Panel b presents the mean operating lease intensity and mean operating lease intensity on income-generating properties
Extant REIT research largely overlooks operating leases as an alternative source of financing. In this study, we hand-collect lease information of 334 unique REITs over the period of 1993 to 2018, and we document that an increasing number of REITs have been including operating leases in their capital structure to finance income-generating investment properties. We examine the determinants of the operating lease decision and find that REITs which adopt operating leases tend to be larger and have more growth opportunities as measured by Tobin’s Q. But they also have higher leverage, report lower funds from operations, and higher risk. We further find that operating lease intensity for REITs is negatively affected by credit ratings, but not by growth opportunities. Lastly, we examine the market effect related to operating lease decision and find that REITs with operating leases are associated with lower shareholder returns. Overall, our findings imply that operating leases are employed as an alternative financing source by REITs that are highly levered and cannot rely much on their internal funding. As a result, the market does not view the use of operating leases in the REIT sector favorably.
Average housing price for street name fluency groups. This figure presents the mean housing price for each fluency group using our six fluency measures. Higher ranked groups are more fluent. Bars represent 5% and 95% confidence intervals. Englishness Group measures how often a combination of letters appears in English media. Words Group is the number of words in the street name. MS Word indicates whether a street name passes the MS Word spell check. CommonName Group is the number of suburbs that share the same street name. Syllables Group is the number of syllables in a street name. Letters Group is the number of letters in a street name. See Section 2.2 for more details on fluency measures
Fluency Heatmaps across Sydney suburbs using six fluency measures. This figure illustrates the fluency score of street names in various suburbs across Sydney using heat maps, whereby greener shades represent higher fluency scores, and browner shades correspond to lower fluency scores. The results for our six different fluency measures are presented in Panels A to F. Englishness Group measures how often a combination of letters appears in English media. Words Group is the number of words in the street name. MS Word indicates whether a street name passes the MS Word spell check. CommonName Group is the number of suburbs that share the same street name. Syllables Group is the number of syllables in a street name. Letters Group is the number of letters in a street name. See Section 2.2 for more details on fluency measures
Google trends index for australian region search of ‘Brock’. The figure reports the Google Trends Index for the search term ‘Brock’ in the Australian region from 2014 January to 2018 August. The index spans from 0 to 100 and hits the maximum 100 in September 2006 when legendary Australian racecar driver Peter Brock passed away
This paper examines whether and how street name fluency affects housing prices using a rich sample of housing transactions in Sydney, Australia. We find street names with longer words are preferred, i.e., homes on street names with more letters are priced with a 0.6% premium. Homes with unique street names are sold 1.6% (or A$10,835) higher than those with more common names, implying disfluency and uniqueness preference. Moreover, homes with less fluent street names are valued more conditional on the street name is rare or the home is in the luxury price range. This is consistent with the consumption context effect in the psychology literature that in the context of special occasion high-end goods, lower fluency and grater uniqueness makes the products feel more desirable and valuable. While we show disfluency preference on aggregate, we also find evidence of fluency preference by non-English speaking buyers and for new developments. Preferences for royal names or popular words proxied by Google Trends are also documented. Overall, our findings shed light on understanding how name fluency affects the investment decision of special occasion goods such as real estate.
In 2009, the Federal Reserve subjected nonbank mortgage-originating subsidiaries of bank holding companies (BHCs), but not independent nonbank (INB) mortgage originators, to consumer compliance supervision. We examine the effects of this regulatory change on the pricing and performance of nonbank originations using a sample of conventional, first-lien, amortizing mortgages originated between 2000 and 2015. We find that subsidiary nonbank (SNB) loans, which had a higher probability of default than INB mortgages prior to the policy change, had a lower probability of default following the change. In addition, we identify small but statistically significant decreases in loan interest rates and loan-to-value ratios for SNB mortgages relative to INB mortgages. When we split our sample into prime and subprime mortgages, we find those effects hold for prime mortgages. For subprime mortgages, after the policy change SNB originations had higher interest rates and lower LTV ratios than INB mortgages, with only weakly significant differences in probabilities of default. The findings are robust to several potential confounding effects, including those due to firm entries and exits. Our findings are consistent with BHCs reducing risk shifting in mortgage lending across subsidiaries following their heightened regulatory scrutiny.
a Random Partitioning of Folds. b Spatial Partitioning of Folds. Notes: This figure depicts the conceptual difference between random partitioning and spatial partitioning of folds using k-means clustering during cross-validation
a Spatial Distribution of Apartment Rents. b Mean Apartment Rents on ZIP-Code Level. Notes: The upper map depicts the absolute monthly asking rent in Euro per month of each individual observation in our sample of 9256 listings between January 2019 and March 2020 in Frankfurt. The bottom map shows the respective mean asking rents in Euro per month aggregated on a ZIP-code level
a Semi-variogram of the log rent. b Distribution of Neighbors within the Spatial Autocorrelation Range. Notes: The empirical Matérn semi-variogram model suggest a spatial autocorrelation range of 0.58 km, which is the distance up to which spatial autocorrelation persists in the data. The histogram presents the distribution of neighbors within the spatial autocorrelation range
a Distribution of the Absolute Percentage Error for Models including Spatial Controls. b Distribution of the Absolute Percentage Error for Models excluding Spatial Controls. Notes: The density plots present the expected distribution of the absolute percentage error resulting from the respective resampling strategies in comparison to the true out-of-sample distribution of the absolute percentage error from the holdout sample. The line type represents the resampling strategy used in the inner loop for model selection (non-parametric models only) and the line color represents the resampling strategy applied in the outer loop for model assessment. The true out-of-sample distribution is represented in black. The shaded areas depict the interquartile range, which is the area between the first quartile and the third quartile of the true absolute percentage error with the middle line representing the median
Data-driven machine learning algorithms have initiated a paradigm shift in hedonic house price and rent modeling through their ability to capture highly complex and non-monotonic relationships. Their superior accuracy compared to parametric model alternatives has been demonstrated repeatedly in the literature. However, the statistical independence of the data implicitly assumed by resampling-based error estimates is unlikely to hold in a real estate context as price-formation processes in property markets are inherently spatial, which leads to spatial dependence structures in the data. When performing conventional cross-validation techniques for model selection and model assessment, spatial dependence between training and test data may lead to undetected overfitting and overoptimistic perception of predictive power. This study sheds light on the bias in cross-validation errors of tree-based algorithms induced by spatial autocorrelation and proposes a bias-reduced spatial cross-validation strategy. The findings confirm that error estimates from non-spatial resampling methods are overly optimistic, whereas spatially conscious techniques are more dependable and can increase generalizability. As accurate and unbiased error estimates are crucial to automated valuation methods, our results prove helpful for applications including, but not limited to, mass appraisal, credit risk management, portfolio allocation and investment decision making.
House price index for Baltimore MSA from 2004-2016Q1. U.S. Federal Housing Finance Agency via the Federal Reserve Economic Data website. This figure shows the temporal movement of the housing price index for the Baltimore MSA between 2004 and Q1 2016
Map of study area with observations. This figure represents the counties in our final sample of SFH and MFH homes in the Baltimore MSA. Starting at the top left, the counties are Carroll, Baltimore, and Harford. Baltimore City is in the interior of Baltimore County while Anne Arundel is at the bottom of the figure. Figure created by authors using ArcGIS
Plot of yearly coefficients for dependent variables with 95 percent confidence intervals. This figure plots the movement in ln(price), ln(DOM), and DOP measures across years for all homes included in the full sample. It is a graphical display of the results from Table 3 (Year variable). Figure created by Stata 16
Plot of SFH vs. MFH coefficients for dependent variables with 95 percent confidence intervals. This figure displays the divergence in ln(price), ln(DOM), and DOP measures across years for SFH (compared to MFH). It is a graphical display of the results from Table 4 (SFH X Year variable). Figure created by Stata 16
Real estate research has primarily focused on examining aggregate shocks and dynamics to describe housing market trends; however, heterogeneity across housing types has been largely ignored especially on an intra-city level. In this paper, we address this heterogeneity through exploring how two different market types – single-family homes (SFH) and multi-family homes (MFH) within in a single metropolitan area – responded to the housing market bust and recovery between 2008 and 2016. Results from a robust series of specifications provide evidence that market dynamics of price, liquidity, and degree of overpricing deviated substantially both during the housing market bust and recovery period, and showed significant variation across housing types. These outcomes are novel in demonstrating that SFH and MFH are differentially affected by market shocks, and that a time-constant control for housing type may not accurately capture housing market dynamics in any single locale, potentially leading to erroneous conclusions about policy implementation.
Relation between estimated degrees-of-freedom ν~\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$\tilde \nu $\end{document} and median absolute errors med |e|
Automated valuation models (AVMs) are widely used by financial institutions to estimate the property value for a residential mortgage. The distribution of pricing errors obtained from AVMs generally show fat tails (Pender 2016; Demiroglu and James Management Science, 64 (4), 1747–1760 2018). The extreme events on the tails are usually known as “black swans” (Taleb 2010) in finance and their existence complicates financial risk management, assessment, and regulation. We show via theory, Monte Carlo experiments, and an empirical example that a direct relation exists between non-normality of the pricing errors and goodness-of-fit of the house pricing models. Specifically, we provide an empirical example using US housing prices where we demonstrate an almost perfect linear relation between the estimated degrees-of-freedom for a Student’s t distribution and the goodness-of-fit of sophisticated evaluation models with spatial and spatialtemporal dependence.
The impact of immigrants on housing prices and rents has been well documented in the literature. There has been less research, however, on other mechanisms by which global economic and financial developments may impact on local housing markets. We investigate whether the foreign-born—that is the stock of previous immigrants—act as a conduit for economic changes abroad to influence local housing markets. Examining disaggregated regions in Australia from 2006-16, we construct a measure of the average performance of the motherland economies of the foreign-born for each region. We find evidence that house prices and rental growth tend to rise when motherland economies are performing poorly. This effect is economically meaningful, robust and appears to represent a distinct channel from the immigration effect.
Housing policy, as well as academic research, are increasingly concerned with the role of bias in subjective dwelling valuations as a proximate measure of households’ house price expectations and their relationship with housing demand. This paper contributes to this area of study by exploring the possibility of simultaneous relationships between households’ price expectations and incentive to maximise the size of housing services demanded also accounting for the supply side factors and regional perspective. The empirical estimation takes the form of a system of a two simultaneous equations model applying two stage least squares estimation technique. Cross sectional estimations utilise data extracted from the Israeli Longitudinal Panel Survey (LPS) data. Applying the best available proxy for households’ price expectations, calculated as the ratio between subjective dwelling valuations (LPS) and the estimated market value of the same properties, research has identified the interrelated factors that simultaneously influence householders’ price expectations and housing demand. Results offer conceptual and empirical advantages, highlighting the imperfect nature of the housing market, as reflected by the inseparability of bias in subjective valuations and housing decisions.
This paper examines the operational efficiency of equity Real Estate Investment Trusts (REITs) with respect to external advisement and management. We employ data envelopment analysis (DEA), a non-parametric statistical procedure that tests whether decision-making units are operating on their efficient frontier, to measure the relative performance of REITs before, during, and after the 2008–2010 financial crisis. Annual observations of both advising and management status of each REIT allow us to parse efficiency by these groups in various combinations. Our evidence suggests the inefficiency of externally-advised REITs has diminished in recent years, and the structure is no longer strictly inferior. External management of property operations, however, remains less efficient than self-management. General and administrative expenses, external advisory fees and property management fees are the main sources of inefficiency over the study period. In a difference-in-difference specification we find industry-wide operational efficiency was higher in the post-crisis than the pre-crisis period, indicating efficiency gains following the crisis.
Top-cited authors
Harry H. Kelejian
  • University of Maryland, College Park
C. F. Sirmans
  • Florida State University
David Geltner
  • Massachusetts Institute of Technology
John M. Clapp
  • University of Connecticut
John L. Glascock
  • University of Connecticut