Cost Modelling

M. Skitmore, V. Marston, M. Skitmore, V. Marston

• Publisher: Routledge
• Year: 2005
• Print ISBN: 9780419192305
• eBook ISBN: 9781135823634
• Edition: 1
• Topic: Architecture
• Subtopic: Architecture General

Cost models underlie all the techniques used in construction cost and price forecasting, yet until relatively recently industry has been unfamiliar with their characteristics and properties. An understanding of the various types of cost model is vital to enable effective cost control and the development of future forecasting techniques. This volume brings together more than 20 seminal contributions to building cost modelling and introduces the major landmarks in progress and thinking in this field: * strategies and directions * explorations in cost modelling * cost-product/process modelling * dealing with uncertainty The strong techniques bias of this book will appeal to construction professionals involved in estimating, as well as researchers and students of building economics.

Part 1: Strategies and directions

This Part contains papers which contribute to the delineation of the field-what cost modelling has been and should be trying to achieve and how this has been and should be operationalised. By their very nature, these papers are ambitious and highly speculative, motivated by a desire to bring some semblance of order to the chaos perceived by their authors. Collectively, these papers contain what is perhaps the most exhaustive coverage of the factors and issues surrounding cost modelling adumbrated up to their time of publication. Considering that the reference material used by most of the authors of the papers is likely to have been very similar, it is surprising and instructive how different are their conclusions.

The paradigm referred to in Brandon's Building cost research: need for a paradigm shift? (ch 1.1) is one of type adumbrated by Thomas Kuhn (1970) involving a radical change of direction in thinking and application pointing out the current inadequacies in models and theory and arguing against an incremental approach to cost model development—‘what if practice is using the wrong model?’. He maintains that cost modelling is a technology and therefore then the pressure for change is likely to be from other stakeholders, ie clients. He also ‘feels’ that there should not be several models and advocates the increased use of computer simulation for the development of production type models due to the potential benefits in explaining how costs are incurred in the building process and in occupation. The paper is inspirational and highly perceptive in several respects—the need for small bites of research, building cost is not an easy subject to study, the relevance of Mitroff's papers, the need for parallel research, the relevance of the distinction between scientific and technological (engineering?) research, that cost modelling is a ‘late starter’. On the other hand some major points are glossed over—who validates the models is examined but how the models are validated is completely overlooked, that increased pressure from client stakeholders was and is unlikely (ELSIE was not a product of client demand), that production models map site processes rather than contractors’ estimators’ processes, and why ‘simulation’ in preference to COCO (Beeston, ch 4.1) or lots of Wilderness (ch 3.1) charts for instance?

Raftery, in Models in building economics: a conceptual framework for the assessment of performance (ch 1.2), suggests that cost model development lacks rigour; and predicates a ‘consistent and conceptual framework within which the performance of models may be evaluated’. Although tentative and speculative, the paper is a very necessary and timely attempt to focus attention upon the need to improve rigour and consistency of model development and evaluation. A framework is proposed, based on a chain which leads from raw data, through a model and its output, to a decision maker [implementor]. Here the paper makes an important contribution to highlighting the significance of the context in which the model is placed. In developing the idea of a chain, a useful distinction is made between the ‘modelling environment’ and the ‘decision [implementation] environment’; and five points are identified where the chain may be tested: data; data/model interface; modelling technique; interpretation of output; and the implementation decision. The first four of these are commented upon in the paper—disappointingly it is commentary rather than discussion, but nevertheless some feel for the structure of the field does emerge. The fifth point, implementation testing, is wisely left for consideration elsewhere.

Beeston's A future for cost modelling (ch 1.3) develops Brandon's (ch 1.1) argument against the conventional use of several product models in prescribing a common basic model for all applications and the rejection of existing product models. Beeston advocates using existing skills more widely and developing models that quantify uncertainty more closely. ‘Realistic’ production models are recommended for use at an earlier stage than usual because of their enhanced explanatory powers resulting from the introduction of causality into the relationships. To forecast the cost effect of a design change, Beeston argues that all that is needed is to calculate the cost of replacing one component of the design with another with a separate allowance for general plant, labour and time requirements. The new data collection methods necessary to support these models are not considered to be a major problem due to the tendency to a common and economical method of planning and execution for a given design. Three types of realistic modelling approaches are proposed (1) simulation of construction processes (eg Marston and Skitmore's intelligent simulation model, ch 4.5), (2) attaching costs to activity networks (eg Bennett and Ormerod's CPS, ch 4.3), and (3) simulation of construction planning methods (eg Beeston's COCO ch 4.1). Beeston's advancement in the field is in his assumption that there is a common construction method for a design, with any residual alternatives being covered by parallel simulation. He overcomes the criticism that contractors do not use this model at the moment by assuming that they will follow the designers’ (tendering) model in due course. He also outlines some of the spin-offs that may follow from the use of realistic models. He is essentially proposing an integrated model for a fragmented procurement process—is this practicable or are the disintegrating forces of competition too great? To find answers to these difficult questions may mean taking a closer look at the economists view and considering the relationship between risk and incentive, and the economics of asymmetric information.

Skitmore and Patchell's Developments in contract price forecasting and bidding techniques (ch 1.4) section on Mathematical and topological features is perhaps the most relevant to the theme of this reader, and the models examined are summarised in their Table 1.4.1 in terms of mathematical symbols (all the models examined are in linear form); relevant contract type; general accuracy; deterministic/probabilistic structure; number and type of items (elements); derivation and deterministic/probabilistic nature of quantities (element values); and derivation, currency, weighting, quantity trended and deterministic/probabilistic nature of rates (functional values). This may be contrasted with Newton's (ch 1.5) ‘agenda’. The paper continues to describe the use of regression, CPS and ELSIE (not reprinted here), with some new (to this field) work on the accuracy of the regression forecasts. Also, the bidding model debiaser described in the final section of the paper is new and may be compared with Flanagan and Norman's (ch 5.5) CUSUM approach.

Newton's An agenda for cost modelling (ch 1.5) is concerned that there is no adequate way of comparing cost models or research into cost models; and nine ‘descriptive primitives’ are proposed as classificational features of all cost modelling research, the ones—which are relevant here being—data (design information), units of measurement (elements), approach (purpose), application (design stage), model (structure), technique (regression, network, etc), assumptions (eg element values and functional relationships) and uncertainty. Within each of these descriptive primitives is proposed a set of characteristics (simulation, generalisation, optimisation, etc) to represent the various internal ‘levels’. He forecasts a long term trend towards models for optimisation as the field becomes better understood and lodges a plea for explicating the assumptions inherent in the models. As Newton concedes, his work is only intended to provide a supporting framework—to give some order to the way in which we classify, talk and think about cost models—and, as a first attempt, is expected to be replaced by better future versions.

Fortune and Lees’ The relative performance of new and traditional cost models in strategic advice for clients (ch 1.6) seminal work is of great importance. Theirs has been a labour of love over several years of careful study investigating pre-tender estimating practice. It is a remarkable achievement in several respects. It is the first work to deal comprehensively with the practice of pre-tender estimating, it aspires to an utter level of rigour and it is fundamentally and relentlessly concerned with what counts in practice. The objectives of the paper are to establish and explain the incidence of competing forecasting models within the built environment, together with an evaluation of their use in terms of accuracy, reliability and value. An assessment of the degree of judgement involved in the particular forecasting model along with an approach to assess new models forms the remainder of the study's objectives. These objectives are clearly stated and combine to expeditiously orientate the new readers to this field, in terms of the state of its development and future direction. The paper begins by categorising the various currently used models under the headings of traditional models and newer models. The newer models comprise statistical models, knowledge based systems, life cycle costing techniques, resource/process based methods and risk analysis techniques. Many of the models selected, particularly the traditional models, are in fact approximate estimating methods. It is their selection and timing within an overall cost information system which qualify them as ‘strategic’ from the viewpoint of their user. Having grouped the pretender estimating methods into ‘traditional’ and newer models, they then proceed to find out, by postal questionnaire, how these are getting on ...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Series Preface
• Preface
• List of Sources and Acknowledgements
• Introduction to the Readings
• Part 1: Strategies and directions
• Part 2: Explorations in cost modelling
• Part 3: Cost-product modelling
• Part 4: Cost-process modelling
• Part 5: Dealing with uncertainty
• Selected Bibliography