Rethinking Risk Budgeting in Real Estate Redevelopment

Abstract

Real estate redevelopment projects are developed under conditions of partial information. Existing buildings may contain hidden technical conditions, incomplete documentation and complex interfaces between old and new systems, while tenant requirements, design decisions and execution constraints may continue to evolve during project development. At the same time, redevelopment projects are investment-driven, meaning that cost estimates, risk buffers and contractual commitments strongly influence whether a project proceeds and how financial exposure is accepted. In current practice, redevelopment costs are often estimated in considerable component-level detail, while uncertainty is frequently treated through aggregated project-level contingencies. This creates a mismatch between detailed cost estimating and comparatively general risk budgeting.
This thesis examines whether risk budgeting in real estate redevelopment can be improved through component-level differentiation of cost deviations and an explicit distinction between uncertainty reducibility and probabilistic representability. The research connects component-level cost analysis with uncertainty theory. The epistemic–aleatory distinction is used to assess whether uncertainty is knowledge-related and potentially reducible, or whether it reflects inherent or externally driven variability. Knight’s distinction between measurable risk and fundamental uncertainty is used to assess whether an uncertainty mechanism is sufficiently recurrent, comparable and stable to support probabilistic treatment. Reference Class Forecasting is used as a theoretical benchmark for evaluating the conditions under which historical cost deviations can support probabilistic buffering.
A multiple case study was conducted within the CBRE / Turner & Townsend redevelopment context. Six completed real estate redevelopment projects were analysed using a mixed-method design. The quantitative analysis compared the contractually fixed Aannemingsovereenkomst (AOK) baseline with approved post-contract additional and reduced works. Cost data were decomposed using an Element-Based Breakdown Structure for building-related components and an additional Category Z for non-elemental project-process costs. This analysis identified where realised cost deviations concentrated and selected dominant deviations for qualitative follow-up. Semi-structured interviews and project documentation were then used to reconstruct the underlying mechanisms and classify twenty-one dominant component-level deviations according to causal origin, reducibility and probabilistic representability.
The findings show that redevelopment cost deviations were strongly concentrated rather than evenly distributed across project budgets. Across the six cases, four of the twenty Level 2 component categories, D50 Electrical, B20 Exterior Enclosure, D40 Fire Protection and D30 HVAC, accounted for 52.32% of the total absolute cost deviation. However, component recurrence alone did not imply causal comparability. Similar component categories reflected different uncertainty mechanisms depending on project context, baseline maturity, stakeholder conditions and technical interfaces. The qualitative classification showed that epistemic uncertainty was prevalent: ten deviations were classified as Epistemic Knightian and six as Epistemic Risk, together representing 76.2% of the analysed dominant deviations. These deviations were commonly linked to incomplete technical elaboration, insufficient verification, unresolved existing-building information, coordination issues, evolving stakeholder requirements and deliberately postponed scope decisions.
The study concludes that redevelopment risk budgeting can be improved meaningfully, but conditionally. The improvement does not primarily lie in calculating a more accurate aggregated contingency percentage. Instead, risk budgeting should move from aggregated financial buffering towards component-level uncertainty diagnosis and differentiated risk treatment. Potentially reducible uncertainty should first be addressed through investigation, technical verification, coordination or scope clarification. Probabilistic buffering and Reference Class Forecasting are more defensible where remaining uncertainty is recurrent, comparable and stable. Where comparability is weak, explicit assumptions, scenario reasoning and managerial judgement remain necessary. The contribution of this research is therefore diagnostic rather than predictive: it reframes redevelopment cost deviations as indicators of underlying uncertainty and provides a basis for deciding which uncertainties should be reduced, which should be buffered and which require project-specific judgement.