PAPER – Construction Productivity – CONSTRUCTION PROJECT MANAGEMENT

5 A review of construction cost and price indices in Britain

Introduction

What determines the living standards of a society is the quantity of goods and services produced by the society. The importance of measuring this quantity is obvious for the understanding of economic progress. However, the statistical agencies of governments measure the monetary (nominal) value of the goods and services produced by their countries (usually called GDP at current prices). This value is a set of (not directly known) quantities multiplied by a set of (not directly known) prices. The conversion of the monetary value of output to the real output of the economy requires a price index because the changes in monetary value of the goods and services produced are the combination of two movements: monetary price level movements and quantity movements. Therefore the measure of the real outputs is only as accurate as are the price indices in measuring price changes, and hence the construction price and cost indices in construction productivity research in which the focus is on understanding the relationships between the changes in real inputs and outputs of the construction industry over time are of importance.

Our initial reasons for questioning the quality and accuracy of the existing British construction price and cost indices are threefold. First, Allen (1985, 1989), Dyer et al. (2012), Goodrum et al. (2002), Gordon (1968), Ive et al. (2004) and Pieper (1989, 1991) consider the biases in the published construction price and cost indices in, variously, the UK, US, France and Germany, as one of the probable main causes of inaccurate measures of productivity growth rates in the construction sector.

Second, the existing compilation method of British construction price indices was developed in the late 1960s and that for construction cost indices in the 1970s, and since then, there have been profound changes in construction technologies, changes in procurement routes and the associated contract documents, as well as the growing significance of mechanical and electrical services. For example, building projects procured via design and build, which has been gaining in popularity, are completely ignored in the existing building price indices.

Third, advances in the general economic theory of indexing have not been incorporated in the methods used for British construction price indexation since the late 1960s. Recent improvements such as hedonic price indices have been adopted by the UK, US and Germany statistical agencies, to name but a few, to compile other price indices.

Most research on construction price and cost indices is about forecasting and modelling using time series or other techniques, in which past values of price indices or other variables are used to forecast future values of the indices and thus construction price and cost inflation (Akintoye and Skitmore 1994; Akintoye et al. 1998; Fellows 1991; Hwang 2009, 2011; McCaffer et al. 1983 and Ng et al. 2004). Such models presume that the published indices do accurately measure construction inflation. To fill this ‘presumption’ gap, this chapter aims to scrutinize the compilation method used for construction price and cost indices with a view to indicating what is and what is not actually being measured, and identifying opportunities for improvements.

The construction sector is complex and its projects are heterogeneous, including infrastructure as well as buildings, and work to existing structures as well as new projects. Therefore, we require a number of price and cost indices measuring the inflation in each subsector of the construction industry (both new work and work to existing structures). As shown in Table 5.1, only about 50% of the output of the construction sector is believed to comprise projects of the kind covered by the tender price indices. The remaining 50% not covered consists mainly of repair and maintenance (40%) and new infrastructure work (10%). Buildings account for more than 80% of the new work output and infrastructure for the remainder. In comparison with new infrastructure and all repair and maintenance work, the output of the new building subsector is less diverse and easier to measure, and as a result, the price indices of new building work are relatively well developed. Moreover, new building work in Britain, as will be shown, has traditionally been the field of application of bills of quantities (BQs), in which the aggregate values of successful tenders are broken down into unit prices for specified quantities of elements in the finished building. The British method of compilation of tender price indices is based upon this fact. BQs, of course, only exist for new building and some major refurbishment projects and not for repair and maintenance projects. Moreover, it is noteworthy that the samples collected for compiling the British tender price indices (TPI) do not even represent the subsectors within new building work proportionally. We return to this later.

Methods used to compile tender price indices in Britain

BIS produces the most extensive public-sector TPI. The BCIS, the building information research arm of the Royal Institution of Chartered Surveyors (RICS), compiles its own building TPI drawing on its wide reach to private and public projects through the willingness of RICS members to supply data to their own chartered professional institute. Davis Langdon (DL, now part of AECOM), one of the largest quantity surveying practices, also publishes its own tender price index. However diverse the sources of information that these three organizations utilize and however differently the resulting indices signal market conditions, the TPI compilation methods behind their array of indices are very similar, and the origin of the method can be traced back to a joint task force of representatives of the RICS, University College London (UCL) and the then Ministry of Public Building and Works (MPBW).

Under the auspices of the MPBW and the RICS, Bowley and Corlett of UCL produced a report on trends in building prices (Bowley and Corlett 1970). The building price indices being published at that time by the government were input cost indices for labour and material cost, which would from time to time differ from the trends of tender prices in the building industry due primarily to changing productivity and/or market conditions (Fleming 1965). It was against this background that Bowley and Corlett reviewed several possible methods to compose a true tender price index for the building industry, and the method described in Chapter 5 of their report became the workhorse method used in the then Department of the Environment (DoE, now BIS), BCIS and DL ever since.

The method used in BIS is described first and then differences in the methods adopted in BCIS and DL later are highlighted. Mitchell (1971) made the first attempt to document the method used in the then DoE. The following description mainly relies on a manual produced for internal use by the then Quantity Surveyors Services Division (QSSD).

First of all, the data BIS collects for compiling their public nonhousing tender price index is the accepted BQs of building projects procured in a quarter of a year (known as the reference period). Under the traditional procurement route, the client of a building project employs quantity surveyors to quantify the building work as much as possible from the design, which facilitates the preparation of bids by construction firms based on a common framework. The bills of quantities comprise a number of bills, and each bill traditionally covers a separate trade or element. The bill items capture the quantity in suitable physical units, such as cubic metres, of, for example, in situ concrete to be contained in the finished building, as ‘taken off’ the drawings prepared by the design consultants. The construction firms compete by attaching different prices to each unit of measured work.

In addition to the measured work, there is typically a section called Prime Costs and Provisional Sums. Prime costs are usually allowed for specialist work (not designed by the architect) such as lifts, heating systems, air-conditioning systems and electricity supply systems, whereas provisional sums are for the work for which the design is not detailed enough to allow quantification, for example, landscaping. Therefore, works allowed in the Prime Costs and Provisional Sums section will be adjusted in the future according to the actual cost incurred, and the construction firms compete on the markup on these works, which is supposed to cover their profit and their overhead expense incurred because of these works. Traditionally, there is also a section of BQs called Preliminaries which covers the contractors’ general costs for executing the work as a whole. Therefore, the tender price is the summation of the bill items, prime costs and provisional sums, the preliminaries, and other adjustments such as commercial discounts.

BQs provide a rich source of information about the prices and quantities of various elements of the measured work of building projects at the reference period. To construct a price index, the prices at the base period (here, 1995) are also needed. When BIS analyse a BQ, they will reprice it by the rates in the Property Services Agency (PSA) Schedule of Rates of the base year, supplemented with some BQ rates they have collected at the base year. The BIS Public Sector Building (Non-Housing) TPI is a fixed-base, matched-item Paasche index.

To produce an index for each project (project index), from each trade of the project, the items are repriced in a descending order of value until the repriced items are equal to more than 25% of the value of the trade and all items with values greater than 1% of the measured work total are repriced. Therefore, it is a current-weight Paasche index. As only items that can be matched will be compared, so it is a matched-item index.

The sum of all items repriced at the rates of the Schedule of Rates is divided by the sum of the corresponding values at the bill rates with the allocated adjustments on measured work in the BQ to obtain a Schedule Factor.

Schedule Factor:

The adjustments on measured work are the adjustments made on the main summary of the BQs such as head office overhead, correction of arithmetic errors and commercial discount. These adjustments are allocated to the selected items pro rata according to their values.

With the Schedule Factor, the project index is computed by this formula:

Project Index:

The reason for deducting the preliminaries from the contract sum in the denominator is that the rates in the Schedule of Rates include allocated preliminaries.

Since location and function of the building are believed to be main cost drivers, and BIS want to reflect the general building price over time, independent of changes in these factors, each project index number is adjusted for these factors. The published index number is then the median value of these adjusted project index numbers in the quarter and is smoothed by use of a three-quarter moving average.

It is a fixed-base index because the Paasche index is a bilateral index. To construct a time series price index, BIS choose the same base year, say 1995, to compare all the subsequent BQ rates. Therefore, all the later year indices are compared against the 1995 Schedule of Rates. BIS have from time to time changed the base Schedule of Rates. In the past, the base Schedule of Rates was changed every five years, but rebasing has become less frequent, and the latest PSA Schedule of Rates produced by Carillion (one of the UK’s largest construction contractors) was rebased in 2005 prices.

The BCIS index is also a fixed-base, matched-item Paasche index. It matches comparable items and uses the current quantities in the BQs to weight the prices. BCIS use the same Schedule of Rates for the base prices as BIS. They only sample projects of more than £100,000. The difference between the BCIS and BIS methods lies in the way they adjust and aggregate project indices. From 1984, each BCIS project index has been adjusted for the size, location and contract type (firm price or fluctuating price) before aggregating into the published indices (whereas BIS adjust for location and function only). The other salient difference is that BCIS takes the geometric mean rather than the median of the adjusted project indices. The BQs, which cover both public and private sectors, are supplied by RICS members.

The Davis Langdon (DL) TPI is a chain-linked, matched-item Paasche index. The obvious difference is the application of the chain-linked system to join up the bilateral indices. In a chain-linked system, the reference period of the previous bilateral index becomes the base period of the succeeding bilateral index. For example, if 2012Q4 is the base period and 2013Q1 is the reference period in the first quarter, then in the second quarter, the base period is 2013Q1 and the reference period is 2013Q2. As DL publish a price book – Spon’s Architects’ and Builders’ Price Book – annually, the base prices are actually updated every year rather than quarterly. The index is, therefore, more accurately called an annually chain-linked index, and as such is similar to the Consumer Price Index (CPI) and Retail Price Index (RPI) compiled by the Office for National Statistics (ONS). Since the sample is confined to the projects in which DL is involved in Britain, and the sample size in number of projects is therefore smaller, more than 25% of items in terms of value are sampled in each project to reduce the sampling errors. The adjustment factors of the project indices are size, location and building function and the geometric mean is used to aggregate the project indices. The two advantages of the method used by DL are:

· it is chain-linked not fixed-base

· there are three adjustment factors not two, but its disadvantage is that DL draws on a smaller and less representative sample of BQs.

Compilation methods for construction cost indices in Britain

BCIS, BIS and DL publish general construction cost indices to reflect the inflation of the input prices paid by contractors in various subsectors of the construction industry. These cost indices are developed from the Price Adjustment Formulae indices, also known as NEDO indices (originally compiled for the Construction Committee of the National Economic Development Office).

Construction contracts with fluctuation provisions allow contractors to pass on to their clients the increase in input costs, such as wages and material prices, in the period between the date of tender and the work being carried out. NEDO’s Steering Group on Price Fluctuations Formulae published a report in 1969 which proposed a formula-based method to calculate input price (i.e. cost) fluctuations in building contracts. It suggested dividing the contract sum into trade-based work categories (such as brickwork and concrete) and adjusting by the published indices for each work category. By analysing 60 completed questionnaires, the report concluded that the administrative cost of agreeing the fluctuations by the recommended formula method would be 0.16% of the contract sum compared to 0.75% by the conventional method of auditing suppliers’ invoices and the wages set by the appropriate wage-fixing bodies.

The formulae methods of adjusting fluctuations in civil engineering contracts began in 1973 and in building and specialist engineering contracts in 1974. Each work category index is a weighted index of various labour wages, material prices and plant costs that reflect the cost of that particular work category. No productivity growth is assumed in the work category indices.

The sources of the labour wages are largely based on national labour agreements. At the time, quite a large proportion of the construction workforce had wages based at least in part on rates set by the national wage agreement bodies, which had representatives from employers and unions. This is no longer the case.

The construction material price indices are compiled by ONS as part of the whole-economy producer price indices. BIS publishes the material price indices in its Monthly Statistics of Building Materials and Components. The plant cost, including the cost of the operator, is a weighted average of depreciation, building labour cost and consumables such as tyres and fuel cost.

The weightings of labour wages, material prices and plant cost for the work categories have been revised infrequently since the first series was published in 1974. The second version, Series 2, was published in 1977, and the latest version, 1990 Series (also known as Series 3), was published in 1995.

The double-digit inflation that plagued the UK for the majority of the 1970s and 1980s made the fluctuation construction contracts very popular, and the Price Adjustment Formulae indices then had an important role to play. Inflation in the UK has come down under 5% since the mid-1990s and, consequently, fluctuation contracts have become exceptions rather than the norm.

BCIS publish nine building cost indices on a monthly basis:

· general building cost index

· general building cost, excluding mechanical and electrical (M&E), index

· steel-framed construction cost index

· concrete-framed construction cost index

· brick construction cost index

· mechanical and electrical engineering cost index

· basic labour cost index

· basic materials cost index

· basic plant cost index.

BCIS (1997) explained that it analysed 54 bills of quantities to work out different weightings for the various work category indices for each of the building cost indices. Since the weightings are fixed in the base year (1977), the BCIS Building Cost Indices are fixed-base matched-item Laspeyres indices. Figure 5.2 uses the BCIS Brick Construction Cost Index as an example to illustrate the relationship with the work categories indices and the underlying input indices.

BIS publish 35 resource cost indices for construction in the UK on a quarterly basis covering repair and maintenance as well as new work. Table 5.2 summarizes the availabilities of the cost indices in the different sectors and for the different inputs of the construction industry.

BIS (2012) reported that the weightings of the new building nonhousing index had been assessed in the 1970s, and the weightings of the rest were assessed by a panel of Chartered Quantity Surveyors in 1998.

Evaluation of the British TPI and opportunities for improvements

Before making any recommendations for improving the British TPI compilation method, it must be acknowledged that, having attempted to review the many different compilation methods documented for other countries, as summarized, for example, in OECD and EUROSTAT (2001) and EUROSTAT (1996), it is clear that the method adopted in the aforementioned three British organizations has led the world for a quarter of a century. Blessed with the availability of BQs, the British TPI does measure the output price of the building industry by making use of the contract prices as opposed to many other countries that use input prices such as labour wages and material prices as proxies for output prices. For instance, until recent years, the US agencies used input prices for deflating the output of the building industry, a procedure that has long been criticized in the United States (Gordon 1968; Pieper 1989, 1991). Although criticisms have finally led to the introduction in the US of a new building output price index by the Bureau of Labor Statistics and US Census Bureau, its prices are either deduced from the property price by attempting to remove the land value from the former or from questionnaires, and as such the validity is less than that of the contract price data obtained from BQs in Britain.

Despite its many advantages over systems in use elsewhere, the following opportunities for improvements in the British system have been identified.

Mechanical and electrical service items

Except for plumbing work, all mechanical and electrical (M&E) service items including comfort cooling, heating systems, lighting, electrical supply systems, lifts and fire-detection systems are not measured in the index because mechanical and electrical services are usually included as prime costs or provisional sums in BQs. In some nonresidential buildings such as offices and hospitals, mechanical and electrical services represent a significant portion, approximately 40%, of the total cost of the building. Leaving this out could result in significant measurement errors.

Figure 5.3 shows that during the 1980s, the building cost index and the mechanical and electrical cost index tracked each other closely, but during the 1990s, the mechanical and electrical cost index was consis-tently at a higher level than the non–mechanical-and-electrical building cost index. From around 2005 the trend reversed. This reflects the fact that the mechanical and electrical services and building work input markets are subject to different short-run cost drivers. If we look at the weightings of the cost indices, material prices have a higher weighting in the M&E cost index.

As previously noted, the building and M&E cost indices assume no productivity growth and are fixed-weight averages of the producer price indices of materials and of the wages in national labour agreements. The weightings were obtained from analyses of 54 bills of quantities for new building work (BCIS 1997).

BCIS Building Cost and M&E Cost Indices (1985 = 100) (BCIS Online)

It is generally observed that goods and services from a sector with higher technological progress and productivity growth have lower price inflation than those from a lower productivity growth sector. The personal computer is a typical example of the former, whereas haircutting services is a widely cited example of the latter.

Mechanical and electrical services, such as air conditioning and heating systems, are reckoned to have been subject to higher productivity growth in the past than the more traditional building trades, such as brickwork, being measured by the TPI. Anecdotal evidence in regard to ICT cabling also suggests that the quality of the cable has increased, say from cat 5 to cat 6, but the nominal prices have been stagnant or have even fallen. Another example is the significant drop in the supply and installation price for domestic solar PV since 2010 in the UK.

New elements and proprietary items

Since the method is to compare the prices of BQ items with the prices in the base schedule of rates, the price of new goods or proprietary items that cannot be matched will not be measured in the TPI. For new goods, frequently updating the base schedule of rates will alleviate part of the problem, and that is the reason ONS adopts an annually chain-linked system for compiling the RPI and CPI. In other (fixed-base) methods, the effect of introduction of new goods will not be measured, and ignoring this will often result in an upward bias of the price index because new goods can usually achieve the same outcome at a lower price than the old goods being replaced. Nordhaus (1997) demonstrated, for example, that ignoring the introduction of new goods substantially overestimates the true price of lighting over time.

Despite its importance, the appropriate method for estimating the price change of a new good when it is introduced is controversial (Bresnahan and Gordon 1996; Hausman 1999; Wolfson 1999).

The problems of proprietary items such as curtain walling and glazed internal partitions are also thorny because the design of the proprietary item is specific to each project, and this prevents them being matched or compared between projects over time.

Sample coverage

RICS (2006, 2012) revealed a clear overall shift of British procurement methods from lump-sum design-bid-build (traditional procurement) with BQ to lump-sum design and build over the period between 1985 and 2010. The share of workload procured under lump-sum design and build has increased from 8.0% to 39.2% by project value, whereas the share for traditional lump sum with BQ has dropped from 59.3% to 18.8%. This trend is unlikely to reverse because the design-and-build procurement route is widely adopted in private commercial projects and private finance initiative schemes and their variants. However, BIS, BCIS and DL only survey the BQs of the traditional procurement method for their TPI calculation. With the dwindling popularity of the design-bid-build with BQ method, continuing to rely on BQs for compiling TPI would make TPI prone to larger sampling errors or even biases. Emphasis needs to be placed on measuring the price movements in design and build contracts.

Sample size and distribution

BCIS aims to sample 80 projects in each quarter because it believes that if 80 projects are sampled, about 90% of the price indices of individual projects will cluster within a reasonable region (about ± 2.8 %) around the average. In the period between 1990 and 2012, the BCIS All-In TPI had an average quarterly sample size of 63, of which 36 were public-sector, nonhousing building projects. By contrast, BIS sampled 57 public nonhousing building projects on average in each quarter over the same period for its Pubsec TPI. Since the index compilation method adopted in BCIS and BIS is similar and both BCIS Public TPI and BIS Pubsec TPI measure the inflation of tender prices in the same domain, there is room for collaboration and specialization.

It would be advisable for BCIS to focus its effort and resources on collecting private-sector information, thereby increasing the sample size of private-sector projects. Currently two subindices of the BCIS All-In TPI, namely BCIS Private Commercial TPI and BCIS Private Industrial TPI, serve as data that BIS use to construct the construction output deflator because these two indices capture the tender price movement of the private sector, to which BIS has no access.

As mentioned in the introduction, Figure 5.4 shows that the distribution of BCIS samples over the period 1990 through 2012 is not aligned with the percentages of output of new building work. Of the 31% of all new output that is in the housing sector, private work accounts for 26.7%, while public work accounts for 4.6%. BCIS, however, note that the majority of their housing samples come from social housing projects. Therefore, the public housing sector is overrepresented, but the construction price movement in the private housing sector is hardly measured in the TPI. Since speculative builders in the private housing market may perform the dual role of developer and main contractor, the tender prices of the construction work, let alone bills of quantities, are generally not available. This problem is not specific to Britain; for example, the US Census Bureau estimates the value of construction work in the private housing market by applying a fixed ratio (currently 84.24%) to the average sales prices of houses.

Private commercial work also appears to be underrepresented in the samples. A significant portion of the private commercial work is major refurbishment of existing buildings which are not measured in the BCIS All-In TPI. However, using the same methodology, BCIS introduced a refurbishment TPI in 1991 with a sample size of around 14 per quarter, so that potentially the All-In TPI samples could be extended to include commercial major refurbishment.

New building work output distribution is compared with the BCIS Sample distribution, 1990 to 2012. (Source: Authors’ calculation, BCIS Online and ONS Statistical Bulletin: Output in the Construction Industry, various issues)

It is also noteworthy that since its sample size of private industrial projects has become too small, BCIS has adopted a different method to compile the TPI for the sector since 2010. In brief, it makes use of the trade price information collected in the BQs of other types of projects and reweights them using the historic BQs of private industrial projects.

Two possible ways to move forward

Having reviewed and assessed the existing TPI compilation methods, this section considers some ways to improve them. Even with suggested changes (see below) in the existing matched-item Paasche indices to improve coverage and samples and to update them to annually chain-linked, it is difficult to cater for the price movements of the diverse M&E items and the effect of quality changes on prices in the existing method. Against this background, we propose the employment of hedonic techniques as a supplement.

Improving the existing method

Regarding the sources of price information, the diminishing popularity of the traditional design-bid-build with BQ procurement route is a real challenge. Some design-and-build contractors produce full BQs for bidding or cost management purposes, and it is worthwhile pursuing the accessibility and pervasiveness of such information.

Alternatively, the feasibility of using cost plans in the BCIS Standard List of Building Elements format deserves further research. The majority of design-and-build projects in the PFI market and private sector include cost plans in the contract documents, and the rates in such cost plans are in principle comparable to schedules of rates such as those in the Approximate Estimates section of Spon’s Architects’ and Builders’ Price Book 2013.

This cost information may be less reliable than BQs for reflecting the true prices of various components of buildings, but it is still better than totally ignoring this growing sector. Also, the quantities measured in those cost plans are useful input information for a hedonic index, something discussed in the next section. There is potential for a circular relationship because the TPIs may be used in setting the cost plan rates, but measures are taken in practice that mitigate this concern: contractors market-test the significant cost elements before submitting a firm price bid, and professional QS firms working for clients ensure the prices in cost plans reflect market prices.

The current method only compares prices of items accounting for 25% of each measured trade by value. Mitchell (1971), Azzaro (1976) and BCIS (1983) show that the 25% rule was a practical compromise between stability of the index and the production cost given the computer technology of the early 1980s. Mitchell reports that the number of items to be compared for 25%, 50%, 75% and 100% of the trade value are 40, 98, 175 and more than 1,000, respectively. Mitchell found the project indices of 80 BQs using the 25% rule to be as stable as that using full repricing (100% rule). However, the 25% rule produced an aggregate index 4.4% higher than the full repricing index reported by Mitchell (1971), whereas the 25% rule underestimated the full repricing index by 1.1% in a separate study reported by Azzaro (1976). There is a case for repeating these studies with more recent data. If this shows discrepancies in estimated levels, then, with the advance of computer technology over the last quarter of the century, it is practical, at least for public-sector projects, to extend the sample items to far more than 25% by value.

DL has adopted the annually chain-linked system, which allows them to compare rates of new items more quickly than the base-linked system in BCIS and BIS with their less-frequent revisions of the base schedule of rates. As early as 1887, Alfred Marshall suggested that the chain-linked system would be a better measure of the price impact of invention of new commodities. The main difficulty in converting the current indices in BCIS and BIS to annually chain-linked indices is the need to update their base schedules of rates annually. RICS acquired a well-established building price book publisher in 2005 and merged it with BCIS, enabling, in theory, BCIS to convert its TPI to an annually chain-linked system by using their building price book published annually rather than the dated PSA Schedule of Rates for the base period rates. The methodology for compiling the PSA Schedules of Rates and the building price book is much the same. With the many annually published UK–based building price books such as BCIS Wessex, Griffiths, Hutchins and Laxton’s and the similarity of the methodology between these price books and that of the PSA Schedule of Rates, it is feasible that BIS could also switch to an annually chain-linked system for their TPI.

Hedonic price index

The hedonic regression technique has been gaining acceptance among statistical agencies such as ONS in the UK and US Census Bureau for compiling their price indices. Ball and Allen (2003) reported that the statistical agencies in Canada, Finland, France, Germany, Sweden and the US have used the hedonic technique to adjust for quality changes in electrical goods such as personal computers, dishwashers and TVs in their price indices. ONS have used the technique to adjust for quality improvements in personal computers, digital cameras, laptops and mobile telephone handsets since 2003 (Fenwick and Wingfield 2005). In real estate and construction statistics, the US Census Bureau has used the hedonic technique to produce their single-family house construction price deflator since 1968 (Musgrave 1969), and ONS have applied it to estimations of imputed rents for owner-occupiers’ houses (Richardson and Dolling 2005). In both cases, hedonic regression techniques are used to adjust the heterogeneities among buildings rather than adjust for improvement in quality over time. Meikle (2001) suggested hedonic methods as an area for further research on construction price indices.

What is a hedonic function? People value a good for its attributes or characteristics. Therefore, goods can be regarded as bundles of attributes, and their values are the sums of the values of the attributes within the bundles. Hedonic function refers to the relationship between the price of the good and the implicit prices of the various attributes embodied in the good. If quantities of attributes are measurable, regression techniques are commonly used to estimate the hedonic function of the good from historical data.

One promising application of hedonic price indices is to extend the coverage to projects that do not have BQs. Figure 5.3 suggests that the current BQ–based TPIs would be unrepresentative for the private housing, private commercial and private industrial sectors. Although hedonic indices have been applied to single-family housing in the US and lessons can be learnt from the relevant research (e.g. Somerville 1999; Dyer et al. 2012), it is probably not the most rewarding sector for the application of hedonic price indices in the UK because of the difficulty of separating the construction price from the total sale price.

A growing number of studies do, however, estimate the relationships between various attributes of buildings and their construction prices. Emsley et al. (2002) and Lowe et al. (2006a, 2006b and 2007) have identified some attributes driving construction price in the UK. Table 5.3 summarizes the significant price-driving attributes reported by some of the studies.

Table 5.3  Construction price-driving attributes from selected literature

To construct a hedonic price index, the dependent variable of the model is the price that the clients pay for the construction of the buildings. Hedonic functions of construction price for base period and reference period are estimated respectively. By inputting the attributes of a building built at base period into the hedonic function for the reference period, the reference period construction price of the building can be estimated. A Laspeyres index can be constructed by comparing the derived reference period prices to the base period prices. It is obvious that a Paasche index can be constructed by using the attributes of reference period buildings and hedonic functions; a Fisher ideal index can then be constructed from the Laspeyres and Paasche indices. Triplett (2004) provides a few other alternatives for the construction of a hedonic price index.

The price-driving attributes identified in the literature are mainly used to adjust the heterogeneities among buildings, such as the floor areas and functions of the buildings, rather than to adjust for aggregate average quality improvement over time. The BIS Tender Price Index (TPISH) published by BIS has also applied the hedonic technique to control the variations in project specifications.

Quality adjustment is more important in the hedonic price index for computers (Cole et al. 1986; Pakes 2003; Silver and Heravi 2004). The common attributes in the hedonic functions include the speed of the CPU and the memory of the hard disk. These are ‘vertical’ attributes since consumers prefer more of them (faster CPU and ‘larger’ hard disk) rather than less. They capture the quality improvement of computers over time.

A definitive solution for measuring inflation in the prices of mechanical and electrical services in buildings over time is elusive. However, the hedonic regression technique can shed light on this, and the following offers some suggestions for further research. Performance specifications for mechanical and electrical service systems are usually produced by professional engineers, appointed by clients. The first task is to translate these performance specifications into measurable input and output attributes of the systems. Prominent resource and input cost heterogeneities between systems such as underfloor heating systems versus traditional profiled surface radiator systems should be included in the hedonic model as dummy variables. Focus, however, needs to be given to the vertical (output, performance) attributes and capacities of the system, and energy efficiencies of the system could be two main attributes to be captured in a hedonic model. Capacities refer to the maximum power (kVA) of electricity generators, maximum loading of lifts and so on. The total or net area served by an M&E system is a good example of the capacity of a system. Energy efficiency is the unit of effective output of the system per energy input. The Seasonal Efficiency of Domestic Boilers in the UK (SEDBUK) for gas, LPG or oil boilers and luminaire-lumens per circuit-Watt of the lighting system are two examples of measures of energy efficiency of mechanical and electrical services. When Ohta (1975) produced a quality-adjusted price index for the US boiler and turbogenerator industries, he applied a hedonic technique to cater for the efficiency and capacity improvement of the boilers and turbogenerators over time. Berry et al. (1995) also found that the capacity variable (horsepower per unit weight) and efficiency variable (miles per dollar) played a key role in their hedonic model of automobile prices.

After measuring the attributes of mechanical and electrical service systems, the next task is to collect price information for the system. For building projects being procured via the traditional route, the sum can be found in the Prime Cost section of the BQ. The prime cost sums are fairly accurate since they usually reflect the fixed prices agreed between clients and nominated subcontractors. With the adoption of new standard forms of contract such as JCT 2005, nominating subcontractors has become less popular, and now the usual arrangements for procurement of M&E and other specialist trades in the traditional route are via Contractor’s Design Portion. Contract sum analysis of the M&E services is usually provided, which provides useful information for hedonic analysis. In design-and-build procurement, the mechanical and electrical service costs normally become part of the fixed price lump sum of the contract and can be discerned in the cost plan of the contract documents.

A hedonic index of mechanical and electrical services, if adequately developed, will be a significant supplement to the existing TPI method since it will capture the price movement of the most cost significant component of buildings unmeasured by the existing method. Perhaps, with the richness of tender price information, a hedonic index of non–M&E tender prices could also be developed. If so, its performance could then be monitored against the TPI compiled by the existing method. The challenge would be to develop performance measures for non–M&E elements of buildings as relevant and potentially precisely measurable on a continuous scale as the performance measures developed for M&E services. It is, however, encouraging to note that performance specification has grown in popularity in the US infrastructure construction sector (Guo et al. 2005), and performance-based contracting has been proposed in the UK (Gruneberg 2007).

A school of thought in the industry is that the rates in BQs are distorted by front-end loading strategies, opportunistic bidding behaviour where low rates are applied to small-quantity items, idiosyncratic methods for allocation of preliminaries, overhead and profit in BQs and so on. If these problems are important, one of the advantages of a hedonic TPI over the traditional TPI would be that it does not rely on the rates in BQs but on market prices of subcontracts. Moreover, factors such as locations, sizes and functions of buildings that BIS, BCIS and DL adjust for in the TPI compilation process could, with a hedonic index, be explicitly modelled.

Evaluation of the UK construction cost indices and opportunities for improvements

Weightings

The construction cost indices reviewed earlier are fixed-base, matched-item Laspeyres indices. Since the method uses the weighting fixed in the base year, it does not allow for substituting cheaper inputs for more expensive inputs in the reference year and would tend to overstate inflation or understate deflation. This problem can be alleviated by updating the weighting more frequently. However, the last update of weightings of the Price Adjustment Formulae was in 1995.

Quality of the input and productivity growth

It is important to emphasize that construction cost indices are not intended to reflect reduction in cost due to productivity growth in the construction sector. In brief, productivity growth means requiring less input to produce the same output. When contractors find a way to use fewer man hours or less material to produce the same output, their ‘input cost’ should drop, but the current compilation method used for the construction cost index would not capture it. A related but different issue is the quality of the input. Both the quantity and quality of the input are multidimensional, and the input price can only be based on one dimension of the quantity. For example, a wage is usually based on time, while concrete and steel are based on volume and weight. The other dimensions of the quality of input, such as the skill level of the labour and the strength and durability of steel, are assumed to be constant, which may not be the case and would bias the cost indices.

Material price

A common criticism of the construction cost index is that the material prices are ‘list prices’ and discounts are ignored. However, ONS reports that it does attempt to collect real transaction prices when compiling their producer price indices, and it is outside the scope here to verify this. The UK Statistics Authority (2011b) reviewed methods of the Monthly Statistics of Building Materials and Components produced by BIS and confirmed their status as National Statistics.

Labour wage

The labour wage component of construction cost indices based on the national labour agreements is more concerning. There could be a variable time lag between market conditions and the wages reflected in the national labour agreement. In addition, given the change in unionization in the UK construction industry over time, it is likely that such national wage agreements will deviate from market wages.

Table 5.4 and Figure 5.5 compare the labour cost indices based on national wage agreements and the survey-based labour cost collected by ONS.

The national labour wage agreement indices (BCIS Labour Cost Index covering buildings and Civil Engineering Labour Cost covering civil engineering) show a higher growth (more than 70% between 2000 and 2012) than other measures of construction labour cost or earning indices (circa 50%) reported by various ONS surveys.

While a detailed examination of various labour earnings indices is outside the scope here, it is worthwhile to compare these indices.

The Monthly Wages and Salaries Survey is an employer-based survey. Its average weekly earnings of regular pay and total pay (including bonuses) in the construction industry shows a growth of 50% and 48%, respectively, between 2000 and 2012. This suggests bonuses shrank slightly compared to regular pay.

One reason that may explain the difference between the ONS indices and the national wage agreement–based indices is a drop in working hours of construction workers over time. This appears to be part of the explanation. The two hourly indices collected from the Labour Force Survey and Annual Survey of Hours and Earnings showed a higher growth (56% and 55%) compared to the weekly earnings indices (49% and 45%) between 2000 and 2012. This is consistent with the 5.2% drop in average weekly hours of work in the construction industry collected by the Labour Force Survey. The experimental index of labour costs per hour published by ONS also reports a 55% increase in labour costs per hour in construction between 2000 and 2012.

Another possible reason is the change in composition of the construction labour force. If the proportion of low-skill construction workers increases over time, the average earnings growth would be lower than the rate of increase in the hourly rates in the national labour agreement. This would require a significant change in the composition to explain the difference, and if such composition change occurs, one would then question the fixed weighting in the construction cost index, which would overstate the labour cost inflation by not allowing for substitution. However, Franklin and Mistry’s (2013) data suggests that construction labour quality has marginally improved (by around 2%) between 2000 and 2012.

There is no good measure of the labour cost holding the quality and composition of the labour force in construction constant, but a comparison of the hourly rate of a few occupations in the construction industry between the 2000 and 2012 Annual Survey of Hours and Earnings provides an intriguing result, shown in Table 5.5.

Generally the growth of the hourly pay of the generic ‘skilled construction and building trades’ is in line with the ONS hourly earnings statistics (Median Hourly Earning excluding Overtime and Construction Average Gross Hourly Earnings of all Employees in Table 5.4), while the hourly rates for specific trades displayed a slower growth. This seems to suggest that the wages of the traditional trades covered by the national wage agreements (as reflected in BCIS Labour Cost Index and Civil Engineering Labour Index in Table 5.4) would grow more slowly, not faster, than the ONS hourly earnings statistics.

Recommendations for construction cost indices

With regard to the CCI, a detailed study of the labour cost indices is recommended. ONS’s household- and employer-based surveys report a lower growth than the national wage agreement–based indices. Focus should be given to analysing changes in the composition and skill levels of the labour force.

The Price Adjustment Formula–based weightings were last updated in 1995 and could benefit from updating.

Concluding remarks

In this chapter, the compilation methods used to produce the three best-known tender price indices for new buildings in Britain and the two sets of construction cost indices in Britain were surveyed. Having reviewed the compilation methods and the sources of data, we consider that the TPIs published in Britain tend to overstate the inflation of contract prices. The reason is that TPIs only measure the inflation of the traditional trade items such as those relating to the structure and the internal finishes works under the conventional BQ procurement route, but M&E services items and proprietary items such as curtain walls, which are subject to higher productivity growth, are not measured in the indices. Moreover, quality of building work such as energy efficiency and safety driven by building regulations tends to improve over time, and the lack of measurement of quality will tend to overstate the prices over time. In theory, the current expenditure-weighted nature of TPI will tend to understate inflation, but the effect will be limited by new items not being matched to items in the dated schedule of rates.

Measuring the price movement of M&E items and broadening the sample base to design and build contracts are two areas well worth pursuing to restore the representative nature of the indices. For design and build, acquiring access to contract price information such as contract cost plans and the effectiveness of using such cost information to produce TPI deserve further study. Because of the diversity of the M&E items used to achieve comparable performance, it is difficult to stretch the current matched-item index method to measure the price movement of the M&E items. Therefore, there is a need to depart from the presently adopted method, and a hedonic index is an appealing alternative. Although the indices may become less consistent than the existing pure item-matching method, this is a trade-off for improving representativeness.

The CCIs, with an infrequent revision of the base basket, suffer the general base-basket-index shortcoming of overstating inflation. The CCIs are also not designed to reflect productivity growth. Looking at the components, the reliability of the labour cost components must be questioned, and an in-depth study with specific focus on the change in the composition and skill levels of the construction labour force is recommended. These three factors – base-basket weights, no reflection of productivity growth and the labour cost components of the CCIs – all tend to give the indices an upward bias.