Automated Valuation Models Research Articles

Race, profit, and algorithms: Neighborhood-level analysis of iBuyers’ profit margin

ABSTRACT iBuyers are firms that use automated valuation models (AVMs), streamline home buying processes, and provide all-cash offers to purchase homes. Although the previous literature has explored the roles and limitations of iBuyers in the housing market, empirical research on the racial implications of these algorithmic home buying processes remains understudied. Using a spatial lag model, this study shows the spatial clustering of iBuyer profit margins, that iBuyers gain more profits when they resell to individuals than institutions, and that some iBuyers have a statistically significant correlation between their profit margins and the proportion of marginalized racial groups within a census tract, while controlling for individual housing characteristics, neighborhood housing quality and demand, and neighborhood amenities and socioeconomic factors. These findings suggest that the more adeptly iBuyers can forecast housing values, the greater the potential to maximize profits from homeowners in communities of color. Consequently, this research contributes to the understanding of how technological mechanisms operate within a purportedly race-neutral framework and advocates for the development and deployment of algorithmic systems guided by the principles of antisubordination, rather than relying solely on notions of “fairness” and anticlassification.

Towards a Better Uncertainty Quantification in Automated Valuation Models

AbstractThis study introduces a novel framework for quantifying prediction uncertainty in automated valuation models (AVMs), crucial tools in modern real estate finance. While non-linear AVMs excel in predictive performance, their limited methods for assessing prediction uncertainty reduces reliability and practical utility. We address this gap by proposing an approach for quantifying the uncertainty associated with predicted house prices and by introducing a model-specific AVM uncertainty estimate (AVMU) for AVM comparisons. Using a dataset of 51,747 historical apartment transactions in Oslo, Norway, we train three AVMs (XGBoost, random forest, support vector machine) to predict sales prices. Thereafter, we develop three base uncertainty estimators (direct loss estimation, bootstrap ensemble, quantile regression) and three meta estimators (average regressor, voting regressor, stacked generalization) for uncertainty quantification. Conformal calibration aligns the outputted uncertainty estimates from the six estimators with standard deviations of corresponding prediction errors. Having strong positive correlations with observed absolute prediction errors, the calibrated uncertainty estimators are shown to effectively capture prediction uncertainty. While the direct loss estimation excels among base estimators, the voting regressor and stacked generalization meta estimators consistently outperform it. Furthermore, by using the AVMU estimate from the stacked generalization meta estimator we can successfully identify the best-performing AVM for three separate apartment portfolios without knowing true sales prices. This alignment of the mean estimated prediction uncertainty with observed deviations underlines the utility of pre-factual AVMU estimates for model comparisons. In conclusion, our framework helps bridge prediction accuracy and uncertainty for AVMs, enhancing their reliability and supporting informed decision making for stakeholders.

Mapping the landscape: A systematic literature review on automated valuation models and strategic applications in real estate

In the rapidly evolving real estate industry, integrating automated valuation models (AVMs) has become critical for improving property assessment accuracy and transparency. Although there is some research on the subject, no thorough qualitative systematic review has been done in this field. This paper aims to provide an up-to-date and systematic understanding of the strategic applications of AVMs across various real estate subsectors (i.e., real estate development, real estate investment, land administration, and taxation), shedding light on their broad contributions to value enhancement, decision-making, and market insights. The systematic review is based on 97 papers selected out of 652 search results with an application of the PRISMA-based method. The findings highlight the transformative role of AVMs approaches in streamlining valuation processes, enhancing market efficiency, and supporting data-driven decision-making in the real estate industry, along with developing an original conceptual framework. Key areas of future research, including data integration, ethical implications, and the development of hybrid AVMs approaches are identified to advance the field and address emerging challenges. Ultimately, stakeholders can create new avenues for real estate valuation efficiency, accuracy, and transparency by judiciously utilizing AVMs approaches, leading to more educated real estate investment decisions.

Real Estate Automated Valuation Model with Explainable Artificial Intelligence Based on Shapley Values

Real Estate Automated Valuation Model with Explainable Artificial Intelligence Based on Shapley Values

From Administrative Price to Market Value of Real Estate. The Evolution of the Valuation System in Poland

Abstract The article applies the method of historical research, using a temporal perspective dating back to 1985, to trace the evolution of market valuation principles for real estate in Poland. These principles evolved similarly to those in many other post-socialist countries, influenced by political and socio-economic transformations and the resulting list of objectives for which these values became essential. The changing legal regulations allowing for the emergence and development of a free real estate market played a decisive role in this process. It was also a period of preparing real estate valuation professionals to meet these requirements. Today, the methodology of valuation, under increasing pressure from various real estate market entities and the rapid advancement of intelligent data collection and processing technologies, is undergoing further evolution. In many countries, including Poland, lively discussions and disputes are ongoing regarding the legal authorization of statistical tools and automated valuation models in valuation practice. These possibilities are being considered particularly in the context of mass property valuations for tax purposes. The methodology involves the analysis of Polish legal provisions, foreign literature, and documents proving the gradual marketization of valuation principles.

Scalable multimodal assessment of the micro-neighborhood using orthogonal visual inputs

AbstractThe features of the micro-location and in particular the micro-neighborhood that residents perceive on a daily basis have a considerable influence on the quality of living and also on housing prices. For automated valuation models (AVMs), the use of micro-neighborhood information would be beneficial, as incorporating additional spatial effects into the price estimate could potentially reduce the empirical error. However, measuring related features is difficult, as they must first be defined and then collected, which is extremely challenging at such a small spatial level. In this study, we investigate the extent to which the quality of micro-neighborhoods can be assessed holistically using multiple data modalities. We design a scalable approach using alternative data (images and text), with the potential to expand coverage to other urban regions. To achieve this, we propose a multimodal deep learning architecture that integrates both textual and visual inputs and fuses this information. In addition, we introduce a training strategy that enables a targeted fusion of orthogonal visual representations of the residential area within the model architecture. In our experiments, we test and compare different unimodal models with our multimodal architectures. The results demonstrate that the multimodal model with targeted fusion of the orthogonal visual inputs achieves the best performance and also improves the prediction accuracy for underrepresented location quality classes.

Unlocking visual data to enhance the accuracy of AI-enabled mass valuation of urban houses: An Australian city case study

Despite earlier praises of hedonic price models (HPMs) in predicting the prices of residential properties, stakeholders in the property market have raised concerns regarding inaccuracy in automated valuation models (AVMs) using conventional HPMs. Furthermore, despite significant advancements in integrating artificial intelligence and image data into AVMs, research in the Australian context is absent. This paper investigates how image data capturing visual features of properties could enhance valuation accuracy using the data of actual sales of 34,399 properties from 2018 to 2020 across 128 urban suburbs in Brisbane, the third largest city in Australia. We develop a convolutional neural network model to extract visual features from large-scale data of 320,000 street-view and aerial-view images. We develop fusion models to integrate these additional visual features in HPMs. Using several experimental designs, our preferred fusion models generate a reduction of 28.35% in root-mean-square errors. Less predictive errors of AI-enabled AVMs through the use of visual data would enhance business confidence for AVMs’ end-users for investment, reporting, risk management, tax estimation, and urban planning purposes.

Towards a Better Uncertainty Quantification in Automated Valuation Models

Towards a Better Uncertainty Quantification in Automated Valuation Models

Assessing the Explanatory Power of Dwelling Condition in Automated Valuation Models

This study investigates the role of dwelling-condition attributes in automated valuation models (AVMs) by utilizing detailed dwelling-condition assessment reports in Norway. A hedonic linear regression model, a gradient boosted decision tree, and a support vector machine were trained and evaluated to predict the sale price of dwellings in three urban regions. The study aims to evaluate the explanatory power of condition attributes in AVMs through a comparison of predictive performance with and without these attributes. The results indicate that the inclusion of condition attributes significantly improves the accuracy of all models across all regions. Furthermore, the models show consistent results regarding which condition attributes are important and their relationship to the price. The study finds that the condition of the bathroom has a high impact on the price, while the condition of doors, roof, and exterior extensions has a low impact. The study concludes that dwelling condition holds explanatory power in both linear and nonlinear AVMs, which can benefit researchers, practitioners, and homeowners looking to renovate. The findings highlight the importance of including detailed dwelling-condition attributes in AVMs and provide insights into the valuation of different aspects of a dwelling.

Leveraging supplementary modalities in automated real estate valuation using comparative judgments and deep learning

PurposeIn this study the authors aim to outline new ways of information extraction for automated valuation models, which in turn would help to increase transparency in valuation procedures and thus contribute to more reliable statements about the value of real estate.Design/methodology/approachThe authors hypothesize that empirical error in the interpretation and qualitative assessment of visual content can be minimized by collating the assessments of multiple individuals and through use of repeated trials. Motivated by this problem, the authors developed an experimental approach for semi-automatic extraction of qualitative real estate metadata based on Comparative Judgments and Deep Learning. The authors evaluate the feasibility of our approach with the help of Hedonic Models.FindingsThe results show that the collated assessments of qualitative features of interior images show a notable effect on the price models and thus over potential for further research within this paradigm.Originality/valueTo the best of the authors’ knowledge, this is the first approach that combines and collates the subjective ratings of visual features and deep learning for real estate use cases.

Automated valuation models: improving model performance by choosing the optimal spatial training level

ABSTRACT The academic community has discussed using Automated Valuation Models (AVMs) in the context of traditional real estate valuations and their performance for several decades. Most studies focus on finding the best method for estimating property values. One aspect that has not yet to be studied scientifically is the appropriate choice of the spatial training level. The published research on AVMs usually deals with a manually defined region and fails to test the methods used on different spatial levels. Our research aims to investigate the impact of training AVM algorithms at different spatial levels regarding valuation accuracy. We use a dataset with 1.2 million residential properties from Germany and test four methods: Ordinary Least Square, Generalised Additive Models, eXtreme Gradient Boosting and Deep Neural Network. Our results show that the right choice of spatial training level can significantly impact the model performance, and that this impact varies across the different methods.

Assessing the accuracy of individual property values estimated by automated valuation models

PurposeThe intent of this paper is to identify uncertainty surrounding automated valuation models (AVMs) valuations and the criteria by which a valuer could judge the accuracy of an AVM estimate of value when being assisted by such AVM as a valuation tool.Design/methodology/approachEuropean law and European Valuation Standards allow valuers to use AVMs as one tool among others in reaching an estimation of Market Value, but only insofar as the valuer is able to satisfy him/herself and the client of the relevance of the AVM report, its inputs and outputs. To enable this, it thus becomes essential that AVMs be more transparent and their accuracy verified.FindingsThis paper recommends minimum reporting requirements thereby enabling an assessment of AVM valuations. At the outset, a distinction needs to be made between two groups of AVM users: banks and valuers. Banks will require considerably more information, including details of the types of models employed and “Bulk” accuracy test results.Practical implicationsThis paper addresses the minimum information needed by valuers in order to gauge the usefulness and accuracy of the AVMs they propose to use as one of their valuation tools.Originality/valueThis paper provides guidance on minimum information requirements for AVMs. Indeed, it may be that the AVM vendors' industry will recognise that providing more transparency in their reports along the lines suggested would facilitate a wider and more supportive acceptance of AVMs.

Machine learning approach to residential valuation: a convolutional neural network model for geographic variation

Geographic location and neighbourhood attributes are major contributors to residential property values. Automated valuation models (AVM) often use hedonic pricing with location and neighbourhood attributes in the form of numeric and categorical variables. This paper proposed a novel approach to automated property valuation using a machine learning model with a convolutional neural network (CNN), fully connected neural network layers with numeric and categorical variables. In this study we compare the results of a fused model, which treat geographical data as an input with the performance of the baseline neural network model with only numerically or categorically represented data. Furthermore, the residential valuation by the proposed fused model was tested with actual sold price data in Greater Sydney, Australia. The study found that the fused model produced valuations with a significantly lower mean absolute percentage error (MAPE) (8.71%) than the MAPE of the baseline model (11.59%). The results show that the fused model with CNN significantly improves the accuracy for residential valuation, reducing spatial information loss by data manipulation and distance calibration.

Estimating Pluvial Depth–Damage Functions for Areas within the United States Using Historical Claims Data

Flooding has been the most costly natural disaster over the last 2 decades within the US. Therefore, recent research has focused on more accurately predicting economic losses from flooding to aid decision makers and mitigate economic exposure. For this, depth–damage functions have commonly been employed to predict the relative or absolute damage to buildings caused by different magnitudes of flooding. Although depth–damage functions, such as those adopted by the US Army Corps of Engineers, are widely available for fluvial and coastal flooding, less work has been done to develop functions for pluvial-induced flooding. Here, we use a database containing 13.5 million claims to develop pluvial depth–damage functions. For this, recently released flood hazard data are utilized to identify claims within the database that are likely related to pluvial flooding. We employed two types of regression models to fit the depth–damage functions. Secondarily, we developed an automated valuation model (AVM) to estimate building values across the state of New Jersey. These building values were then combined with flood hazard layers in order to apply the depth–damage functions and compute an aggregate annualized loss for New Jersey. The results indicated moderate agreement between the observed damage within the state of New Jersey and that computed by applying the study-developed depth–damage curves to buildings within the state using pluvial flood hazard layers. It is anticipated that the depth–damage functions developed by this research will aid future work in more accurately quantifying the economic risks associated with flooding across the US.

Apply GIS, Fuzzy Theory and AHP to Determine Land Value for Land Valuation: A Case Study of Dong Anh District, Hanoi

The land valuation process is an integrated assessment of land price determinants that help managers formulate land-use policies, taxes, and the real estate market. In this study, ten factors affecting land value are selected and divided into 03 groups of factors, including i) Accessibility; ii) Environment; and iii) Socio-culture. The study then determined the weights of the elements by AHP and performed the spatial analysis by GIS, Fuzzy, to determine land value. The results reflect the distribution of land value zones in the Dong Anh district and show a strong correlation with the actual land prices traded in the market. This land value information is assigned to the land parcels to support automated valuation models and urban planning. This study can be implemented in other areas or on a larger scale.
 

Adverse selection in iBuyer business models—don’t buy lemons!

The rise of instant buyer (iBuyer) businesses in the past years has made automated valuation models (AVMs) an important part of the property market. Although iBuyer services are in demand, large actors within the segment have reported dissatisfying profits over time. The business model is subject to adverse selection as homeowners based on their superior knowledge of their home are more likely to accept overpriced bids than underpriced bids, making the iBuyer purchase more overpriced dwellings. In this paper, we use a dataset consisting of 84,905 apartment transactions from Oslo, the Norwegian capital. We use 80% of the dataset to train three different AVMs similar to those used by iBuyers. Next, we construct some simple purchasing rules from the predictive accuracies found in the training dataset. Finally, taking the remaining 20% of the data in a test dataset, we introduce an adverse selection indicator based on accepted probability distributions and calculate the average expected resale profits per apartment for a hypothetical iBuyer. We find that adverse selection has a large negative impact on average profits for the hypothetical iBuyer. Furthermore, the simple purchasing rules are able to improve the profit by 1 percentage point per apartment when adverse selection is present.

Testing machine learning systems in real estate

AbstractUncertainty about the inner workings of machine learning (ML) models holds back the application of ML‐enabled systems in real estate markets. How do ML models arrive at their estimates? Given the lack of model transparency, how can practitioners guarantee that ML systems do not run afoul of the law? This article first advocates a dedicated software testing framework for applied ML systems, as commonly found in computer science. Second, it demonstrates how system testing can verify that applied ML models indeed perform as intended. Two system‐testing procedures developed for ML image classifiers used in automated valuation models (AVMs) illustrate the approach.

Modern Technologies in the Real Estate Market—Opponents vs. Proponents of Their Use: Does New Category of Value Solve the Problem?

The scientific literature and practical studies show that one of the main challenges facing the ongoing development of sustainable real estate markets is the understanding of the specifics of real estate as an object of increasing importance in the global economy. Misinterpretation of the principles of modern valuation models causes conflict between opponents and proponents of their use. To reconcile the two sides, the authors of this study propose the possibility of extending the methodology with new solutions—hybrid automated valuation models—and introducing a new value called the rough value. The study mostly draws on monographic and deductive reasoning methods, which includes an analysis of the basic principles of real estate appraisal (as described by internationally recognized professional standards) and valuation methods. The introduction of proposed solutions to valuation practice should be preceded by the development of unified standards and the principles of their application.

A FRAMEWORK TO INTEGRATE BIM WITH ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING-BASED PROPERTY VALUATION METHODS

Abstract. Property valuation is of extreme importance since variations in the real estate market enormously influence people’s life. The main goal of Automated Valuation Models (AVMs) is to calculate the market value of a large number of properties with an acceptable accuracy. The Hedonic Price Model (HPM) is the most widely used AVM for the valuation purposes. Despite its simplicity, ease of use and straightforwardness, HPM lacks the capability to address the non-linear relationships between different value-related factors. Hence, researchers have developed other state-of-the-art property valuation methods based on the advancements in computer science including Artificial Intelligence (AI), Machine Learning (ML), computer vision and deep learning. Design, development, and validation of such advanced AVMs require establishment of a database including data on the different influential factors. Two types of factors are used in the literature, including textual and visual features. Reliable data sources are required for the implementation of AVMs since the accuracy of the provided valuations is definitely linked to the reliability of the used real estate databases. Building Information Modelling (BIM) provides precise information on different components of properties. Although some scholars have tried to use BIM for property valuation, BIM benefits in different valuation procedures have not been fully investigated. Hence, this paper provides a framework that consider BIM capabilities to be integrated with different stages and processes in property valuation, especially in relation to advanced AVMs based on AI and ML.

Adoption of Automated Valuation Models in Malawi; Valuers' Perception

Valuation profession is historically known for being more traditional and extensively uses traditional methods of valuation. Despite this, in the recent past, there have been strides to incorporate automated valuation methods (AVMs) in the valuation profession especially for rating, mortgages and revaluation purposes. This new approach, however, attracted controversy when it was piloted in one of the cities in Malawi. This study sought to explore property valuers’ perceptions on the adoption of AVM in rating valuation against the use of traditional methods of valuation. The researchers used a descriptive study design with qualitative data collection and analysis methods. Semi-structured interview guides were used to collect data on valuers' perceptions on AVM. A total of 20 in-depth interviews were conducted with professional valuers. The valuers were asked how they perceive AVM approach to rating valuation compared to the use of traditional methods of valuation in terms of its accuracy, reliability and easiness. Data was manually analysed using content analysis. The general perception of the valuers towards AVM was mixed. One school of thought gave AVM the benefit of doubt for future incorporation into the profession if modified to suit the local environment. On the other hand, AVM was considered as posing threat to the valuation profession since anyone may be deemed qualified to carry out property valuation. However, the valuers are of the perception that the models will never replace traditional methods of valuation, hence suggestions to incorporate AVM as a supplement to traditional methods of valuation so that the former is used as a verification and auditing tool for the latter.