A stakeholder-centric hybrid framework for enhancing property valuation quality and decision-making in New Zealand : a thesis submitted in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Construction, School of Built Environment, College of Sciences, Auckland Campus, Massey University, New Zealand

Abstract

Property valuation is fundamental to mortgage lending, investment decisions, taxations, and housing policy. Recent advances in automated valuation models (AVMs), machine learning, and large property data sets have improved valuation accuracy by approximately 5-20%. However, existing approaches still face important challenges, including limited transparency, over-reliance on historical data, insufficient considerations of contextual and behavioural factors, and weak integration of stakeholder perspectives. As a result, property valuation quality (PVQ) continues to be assessed credibility. These limitations highlight the need for a more integrated socio-technical frameworks for residential property valuation (RPV). In New Zealand, RPV is becoming more complex due to changing market conditions, new technologies, and increasing stakeholder expectations. While AI and AVMs improve efficiency, concerns remain about transparency, valuation judgement (VJ), and trust. This study develops and evaluates a stakeholder-centric hybrid PVQ framework to support more transparent and informed valuation decisions in the residential property market.

A sequential mixed-methods design was adopted, beginning with a conceptual review of valuation theory, stakeholder decision-making, and PVQ, followed by two systematic literature reviews (SLRs). The first SLR synthesised 41 peer-reviewed studies examining digital transformation and smart real estate (SRE) technologies, identifying increasing reliance on AVMs, data-driven decision systems, and platform-based valuation environments. The second SLR reviewed 44 peer-reviewed studies AI applications in property valuation, analysing the performance of models such as random forest, support vector machines, decision trees, and gradient boosting. Both reviews highlighted limited stakeholder perspectives, weak VJ integration, and inadequate treatment of PVQ.

Building on these theoretical insights, semi-structured interviews were conducted with key residential property stakeholders, including property valuers (PVs), real estate agents (RAs), residential property investors (RPIs), and home buyers (HBs) (n= 34). The qualitative findings revealed that VJ is highly interpretive and context-dependent, shaped by professional experience, market uncertainty, behavioural considerations, and varying stakeholder priorities. Participants consistently emphasised that “𝘤𝘭𝘪𝘦𝘯𝘵𝘴 𝘥𝘰 𝘯𝘰𝘵 𝘫𝘶𝘴𝘵 𝘸𝘢𝘯𝘵 𝘢 𝘯𝘶𝘮𝘣𝘦𝘳, 𝘣𝘶𝘵 𝘢𝘯 𝘦𝘹𝘱𝘭𝘢𝘯𝘢𝘵𝘪𝘰𝘯 𝘰𝘧 𝘩𝘰𝘸 𝘵𝘩𝘢𝘵 𝘯𝘶𝘮𝘣𝘦𝘳 𝘸𝘢𝘴 𝘳𝘦𝘢𝘤𝘩𝘦𝘥” highlighting transparency and communicative clarity as foundational to trust. These findings informed the identification of strategic principles (SP’s) (SP1 – Trust), which reframes PVQ as a legitimacy-based outcome grounded in professional transparency and stakeholder confidence rather than accuracy alone. Stakeholders emphasised that AVMs should support, not replace, professional VJ, particularly in context-specific cases. This informed SP3 (Hybridisation) and operational principles (OP’s) (OP1 – Contextualisation), reinforcing the integration of technology with local expertise. Regulatory accountability and structured human oversight further shaped SP2 (Governance) and OP2-OP3 (Orchestration and Adaptation), defining the socio-technical structure underpinning perceived PVQ.

The qualitative results informed the development of a nationwide survey (n =112), which was analysed using exploratory factor analysis (EFA), cluster analysis, and partial least squares structural equation modelling (PLS-SEM). The structural model demonstrated moderate explanatory power, with the highest variance explained for VJ (R² = 56.9%), followed by Trust (R² = 35.4%) and PVQ (R² = 11.7%), providing empirical support for the proposed relationships. The findings revealed that RPV is inherently iterative and interpretive, shaped by the interaction of data, professional judgement, market context, and stakeholder expectations. PVQ is not determined solely by model accuracy and methodological compliance, but emerges from how valuation information is generated, interpreted, communicated, and trusted. While digital tools and AVMs enhance efficiency and consistency, they are most effective when used as supportive inputs within a hybrid decision-making process that retains human expertise and contextual awareness. Drawing on these insights, the thesis proposes a stakeholder-centric modular hybrid framework that enhances PVQ as a system level outcome. Framework evaluation findings confirm that the framework reflects contemporary valuation practice and offers practical value for improving transparency, communication, training, and quality assurance.

This research makes theoretical, methodological, and practical contributions. Theoretically, it redefines PVQ as more than just pricing accuracy by showing that it also depends on trust, stakeholder needs, professional judgement, and technology. Methodologically, it develops and tests a stakeholder-centred framework using structural modelling. Practically, it provides guidance on how AI and AVMs can be integrated into professional valuation in a transparent, accountable, and trustworthy way. Although the study is based on New Zealand’s residential property market and uses cross-sectional data, the framework can be adapted to other countries that are adopting digital technologies in property valuation.