Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Construction Management & Economics

ISSN: 0144-6193 (Print) 1466-433X (Online) Journal homepage: https://www.tandfonline.com/loi/rcme20

The divergence in aggregate and activity estimates


of US construction productivity

Paul M. Goodrum , Carl T. Haas & Robert W. Glover

To cite this article: Paul M. Goodrum , Carl T. Haas & Robert W. Glover (2002) The divergence
in aggregate and activity estimates of US construction productivity, Construction Management &
Economics, 20:5, 415-423, DOI: 10.1080/01446190210145868

To link to this article: https://doi.org/10.1080/01446190210145868

Published online: 21 Oct 2010.

Submit your article to this journal

Article views: 283

View related articles

Citing articles: 37 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=rcme20
Construction Management and Economics (2002) 20, 415–423

The divergence in aggregate and activity estimates of


US construction productivity
PAUL M. GOODRUM 1 *, CARL T. HAAS 2 and ROBERT W. GLOVER 3
1Departmentof Civil Engineering, The University of Kentucky, Lexington, KY 40506–0281, USA
2Departmentof Civil Engineering, and 3LBJ School of Public Affairs, The University of Texas at Austin, Austin,
TX 78712, USA

Received 8 August 2001; accepted 5 April 2002

Discrepancies exist between aggregate and activity productivity measurements in the US construction industry.
Multiple studies using aggregate industry measures suggest that construction productivity has declined over
the long term. A longstanding problem with the aggregate measures concerns the difŽ culty of controlling for
in ation so as to accurately measure real output. As an alternative, average activity productivity, measured
by individual work activities, indicates that construction productivity has increased over the same time period.
Activity measurement data have been collected for 200 construction activities over a 22-year time period
from commercial estimation manuals used by contractors and owners to estimate the cost and time require-
ments for construction. This paper examines the discrepancies between aggregate and activity measurements
and suggests possible reasons for their existence.

Keywords: Labour, productivity, trends, construction industry, USA

Introduction This paper examines the problems of using aggre-


gate level data to measure construction productivity.
According to multiple studies using aggregate data, It also reports changes in construction productivity
there has been a long term decline in construction using activity level data. By using activity level data,
productivity in the USA. Prior research suggested that the problems of output measurement associated with
construction real output (value added) per hour fell aggregate data are avoided. New problems introduced
by approximately 2.4–2.8% per year from 1968 to are noted. Activity level data reveal a very different
1980 (Stokes, 1980; Allen, 1985). This Ž nding was trend in construction productivity from that revealed
supported by the Business Roundtable’s Construction by the traditionally used aggregate level data.
Industry Cost Effectiveness Project (1983), which indi-
cated a similar decline. Recent research has indicated Problems with construction aggregate
that these declines are continuing. Teicholtz (2001) productivity measurements
used real output of construction per workhour as
measured by the US Department of Commerce and The difŽ culty in measuring productivity accurately is
the US Bureau of Labor Statistics, and found that the not unique to the construction industry. This problem
construction industry’s labour productivity declined by has stumped economists across a broad range of indus-
0.72% at an annual compound rate from 1964 to 2000. tries. In order to better accommodate the effects
Furthermore, construction’s multifactor productivity1 of rapidly declining relative prices, such as the decline
has declined by 0.4 to 0.9% at an annual compound in the cost of computer chips, the US Bureau of
rate according to government estimates on unpublished Economic Analysis (BEA) revised the method for
data (Gullickson and Harper, 1999). calculating the gross domestic product (GDP) in the
1990s. The revised methodology has been described
*Author for correspondence. e-mail: pgoodrum@engr.uky.edu in detail by others (Young, 1992). One criticism of the
Construction Management and Economics
ISSN 0144–6193 print/ISSN 1466-433X online © 2002 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/01446190210145868
416 Goodrum et al.

revised method is its inability to consider the apparent of a downtown ofŽ ce complex. Even when projects are
improved quality of goods across a range of industries of a similar type, varying ground conditions, weather,
(WSJ, 1995). In addition, the revisions signiŽ cantly and code requirements can make comparing one
reduced prior estimates of productivity growth in the project’s productivity to another very difŽ cult. Also,
overall US economy. Under the old GDP method, US over the past couple of decades, the Ž nal product in
productivity in the 1990s increased at 2.2% per year, construction has become more sophisticated. For
and under the revised method it increased by only example, advances in structural design, Ž re protection,
1.4% per year. The revised estimate of productivity HVAC systems, and information technology have
growth in the 1990s was nearly identical to the rate of dramatically changed the construction product and
productivity growth in the 1970s and 1980s. However, have signiŽ cantly raised the level of effort required to
many economists viewed this Ž nding as problematic, create the Ž nal construction product. RoseŽ elde and
because they had anecdotal evidence suggesting that Quinn Mills (1979) argued that any measure of
the average company used more advanced technology, construction productivity that does not account for
employed fewer people, and made more money than advances in design and quality would lead to low, if
in previous periods (Fleming, 1996). This same criti- not negative, estimates of construction productivity.
cism also exists in construction, where industry leaders US governmental agencies, like the Department of
have expressed doubts regarding declines in aggregate Commerce and the Bureau of Economic Analysis,
estimates of US construction productivity (e.g. T. primarily use the Census’ Value of Construction Put
Kennedy, Chairman of BE&K Inc., and D. McCarron, in Place to measure construction output. Although the
President of International Brotherhood and Carpenters Census’ Value of Construction is a useful measure of
and Joiners, in 1998). construction costs, there are problems with using it as
A multitude of scholars have examined the issue a measure of construction output. The Census’ Value
of measurement problems with construction produc- of Construction includes all the costs of construction
tivity. One of the more commonly cited problems is including material, labour, contractor’s proŽ t, archi-
the difŽ culty of measuring real output. Approximately tectural and engineering work, overhead, and miscel-
half of the construction industry’s output is adjusted laneous interest and taxes. The primary problem with
for in ation using input cost indices instead of output the Census’ Value of Construction is its use of both
price indices.2 Only 52.1% of construction output input cost and output price indices (Table 1).
was adjusted for in ation using output price indices Another problem is that the Census uses a number
according to 1995 measures of construction output in of proxy indices from other sectors of construction to
the Census’ Value of Construction Put in Place. The measure real output. An example is the US Census
remainder of construction output was adjusted for Bureau Single-Family Houses Under Construction
in ation using either input cost indices or a combina- index, which is designed to measure productivity in
tion of input cost and output price indices. The US the residential sector. This index is also used to
Census Bureau’s in ation indices, which convert the measure non-residential and military facilities’ real
value of construction to real output, are shown in output. As a result, more than half of the Census Value
Table 1. The input cost index de ates output and uses of Construction Put in Place is converted to real output
the assumption that a constant relationship exists using the US Census Bureau Single-Family Houses
between the input and output. An input cost index Under Construction Index. The problem is that there
also assumes that productivity is constant (Pieper, have been signiŽ cant breakthroughs in the means and
1989). Many have found evidence that input cost methods of non-residential and military construction,
indices in construction have overestimated construc- including modularization, prefabrication, informa-
tion in ation, thereby underestimating real output and tion technology, and construction automation. These
understating labour productivity (Dacy, 1965; Gordon, breakthroughs have not occurred in residential con-
1968; RoseŽ elde and Quinn Mills, 1979). In fact, Allen struction. Furthermore, signiŽ cant  uctuations in
(1985) acknowledged that half of the productivity new house prices have occurred as a result of interest
declines reported in his study were a result of the inac- rate changes. Therefore, the use of the Single-Family
curacies of adjusting for construction in ation. Houses Under Construction index in non-residential
Construction real outputs adjusted using input cost and military construction sectors will adversely effect
indices do not necessarily re ect innovation and design their output measurement.
improvements in construction (Dacy, 1965; Gordon, Although it is the preferred price index, the Census
1968; RoseŽ elde and Quinn Mills, 1979; Pieper, Single-Family Houses Under Construction index
1990). Construction projects are rarely similar. One still has a potential problem of ignoring changes in
project may involve the construction of a major airport the quality of construction. It is calculated using
runway while another project involves the construction a ‘hedonic’ or regression methodology. The Census
Estimates of US construction productivity 417

Table 1 Sources of cost and price indices used to adjust the value of construction put in place to constant dollars
Type of construction Name of index Type of index
Residential buildings US Census Bureau Single-Family Houses Price
Under Construction
Industrial buildings Unweighted average of:
OfŽ ce buildings • Turner Construction Company Turner: Cost
Hotels and motels • US Census Bureau Single-Family Houses Census: Price
Other commercial buildings Under Construction
Educational buildings
Hospital and institutional buildings
Religious buildings
Miscellaneous non-residential buildings
Farm non-residential
Railroads Unweighted average of:
• Federal Highway Administration Federal Highway: Price
Composite Index Bureau of Reclamation: Cost
• Bureau of Reclamation Composite Index
Telecommunications C.A. Turner Telephone Plant Cost
Electric light and power Weighted Average of:
• Handy–Whitman electric (weight 9) Handy–Whitman electric: Cost
• Handy–Whitman utility building (weight 1) Handy–Whitman utility: Cost
Gas and petroleum pipelines Handy–Whitman gas Cost
Military facilities Weighted average of:
• Federal Highway Administration Composite Federal Highway: Price
Index (weight 2) Census: Price
• US Census Bureau Single-Family Houses Turner: Cost
Under Construction (weight 1)
• Turner Construction Company (weight 1)
Highways and streets Federal Highway Administration Composite Index Price
Sewer systems and conservation and Bureau of Reclamation Cost
development
Water supply facilities Handy–Whitman Water Cost
All other private and public Unweighted average of:
• Federal Highway Administration Composite Federal Highway: Price
Index & Bureau of Reclamation: Cost
• Bureau of Reclamation
Source: Department of Commerce (1998).

annually collects data on the sales price, lot value, and measurements, with Ž replaces and central air condi-
house characteristics of approximately 9000 specula- tioning as the only exceptions (Pieper, 1990). The
tively built single-family homes. The index is calcu- Census index does not include measures for the quality
lated when the sales price less the estimated lot value of materials or design amenities (e.g. bathroom and
is regressed on ten house characteristics (e.g. square kitchen appliances, closet space, energy efŽ ciency
feet of  oor area, lot size, number of stories, number improvements, and patios). The Census index will
of bathrooms, presence or absence of a basement, overestimate the price when these omitted measures
garage, Ž replace, central air conditioning, metropolitan for quality have increased in cost relative to the
location, and geographic location), which were based included measures (Pieper, 1990). This appears to be
on a 1960 model home and revised in 1974 with the case. For example, in 1973 52% of all new houses
the addition of Ž replace and central air conditioning used electric powered heating, which is more expen-
variables. sive than gas and oil powered heating. By 1998, only
Unfortunately, most of the independent variables 26% of new houses were primarily heated by electricity
measure quality through size and location related (US Department of Commerce). Regarding household
418 Goodrum et al.

appliances, 65% of all new homes sold in 1973 had a summarized here. Examining construction productivity
dishwasher. By 1990, the last year for which the using activity level data offers three main advantages.
statistic was available from the US Census Bureau, First, the changes in physical output of construction
92% of all new homes included a dishwasher (US activities (e.g. the cubic metres of concrete placed) do
Department of Commerce). The absence of measures not rely on construction price indices to measure real
for quality leads the Census index to overestimate output for either labour or multifactor productivity.
construction in ation, thereby underestimating con- Second, for the measure of activity labour produc-
struction real output and construction productivity tivity, input is measured by the number of worker
(Pieper, 1990). This bias has a widespread effect on hours, which gives further independence in the use of
measuring output, since it accounts for more than half construction cost de ators. Third, input and output
of the Census Value of Construction Put in Place. are easier to compare over time at the activity level.
The measurement of productivity in construction is For example, when the activity involves the installa-
not a new problem. There are some who argue that tion of aluminium strip siding, 6-inch aluminium trays,
this problem of output measurement has always or 2-ply reinforced curing paper, the characteristics of
existed, and it does not explain all of the measured the Ž nal output tend to remain the same.
declines in construction productivity. Rees (1980)
argued that to attribute the slowdown in construction
Data source and activity selection criteria
productivity to errors in output measurement, one
must show that the errors increased from the 1970s, Data were collected for years 1976 and 1998 on 200
through the 1980s, and up to the 1990s. While this construction activities from Means’ Building Construc-
type of explanation is worthy of further research, it tion Cost Data, Richardson’s Process Plant Construction
is outside the scope of this paper. It is important Estimating Standards, and the Dodge Cost Guides (DCG).
to note, however, that the US Bureau of Labor Data were collected on 100 activities from Means, 50
Statistics does not maintain a productivity index for activities from Richardson, and 50 activities from
the construction industry, because it so difŽ cult to Dodge. These manuals provide construction cost data
adjust construction costs to measure construction real for the purpose of project estimating and budgeting
output (Rosenblum, 2000). purposes. The manuals give unit labour cost, unit
There are other problems with aggregate construc- equipment cost, physical output, and workhour
tion output measurements. Many estimates of con- requirement for each activity. These data are compiled
struction aggregate productivity use data from different on an annual basis from a variety of sources across the
US governmental agencies, such as the hours reported construction industry, including owner organizations,
by the Bureau of Labor Statistics and the dollar construction contractors, and construction trade orga-
amounts reported by the Department of Commerce. nizations.
Governmental agencies use different methodologies Although the estimation manuals provide one of
and analyses in creating their data series; this needs to the best sources of time-series data on construction
be addressed. Finally, there is also a need to under- productivity, there are weaknesses in the data.
stand how the growing trend of offsite production is Contractors who submit the information for the esti-
impacting construction productivity. One study indi- mation manuals know that they are not required to
cated that between 1984 and 1999 prefabrication as a construct a project using their estimations, and this
percentage of industrial project work almost doubled, tends to create in ated estimates of construction costs
from 14% to 27% (Haas et al., 1999). Many Ž rms who (Pieper, 1989). However, the annually published
construct prefabricated units are classiŽ ed as manu- manuals are sold in volume for commercial use to a
facturing Ž rms. Therefore, there is an issue of how multitude of construction contractors, owner compa-
their man-hours and output should be included in an nies and governmental agencies all of whom use the
aggregate measure of construction productivity. estimation manuals to predict project performance.
Furthermore, there is precedence for using estimation
manuals in research to estimate productivity (Thomas
Methodology and Yiakoumis, 1987) and, unlike other sources, esti-
mation manuals provide data going back to 1976.
There were three criteria used to select activities for
Activity level analysis of construction
inclusion in the study. First, the activity had to appear
productivity
in both the 1998 and 1976 estimation manuals. Some
In contrast to the aggregate measures described above, activities included in the 1976 manuals were phased
changes in the physical activity of individual construc- out of the 1998 manuals due to changes in method-
tion tasks have been estimated by the authors and are ology and materials. Likewise, several new activities
Estimates of US construction productivity 419

Input costs in multifactor productivity were de ated to


Sitework n=31
1990 dollars using the Construction Cost Index (CCI)3
Finishes n=23
from Engineering News Record. Note that Formulas
Concrete n=22 1 and 2 assume physical output measures do not
Metals n=21 change in quality. This is based on the assumption that
Doors and Windows n=17 changes in quality would be minimal, since the research
Moisture-therm n=17 examined construction activities that had not changed
in scope between 1976 and 1998.
Wood and Plastic n=17
Next, the percentage change in labour and multi-
Masonry n=17
factor productivity from 1976 to 1998 was measured
Electrical n=14 for each activity using the following formulas:
Mechanical n=14
Specialities n=7
% change in Labour Productivity, ’76–’98 =

0 10 20
Bars show counts
30
1 ––––––––––––––––––––––––––––––––––––––––––—
Labour Productivity, ’98 – Labour Productivity, ’76
Labour Productivity, ’76
2´ 100
(3)
Figure 1 Distribution of construction activities by division
% change in Multifactor Productivity, ’76–’98 =
(source: Means, Richardson, DCG)

were added to the 1998 manuals due to new advances


1 Multifactor
––––––––––––––––––––––––––––––––––––––––––—––– 2 ´ 100
Productivity, ’98 – Multifactor Productivity, ’76
Multifactor Productivity, ’76
(4)
in technology. Second, activities experiencing a diverse
range of technical change were selected. Third, activ- Note that there is a weakness in using just two points
ities from different divisions of the Construction in time to measure the change in productivity, since
SpeciŽ cation Institute (CSI) Masterformat (Figure 1) the results can be affected by the choice of the two
were selected to provide a wide range of activity types. years. Particularly, it is noted that 1976 was a year
of stag ation and excess capacity in the USA. It is
Productivity measurement expected that  uctuations in the change in productivity
would occur in a year-by-year analysis. However, by
One of the challenges in analysing productivity statis- examining the changes in productivity over a 22-year
tically is that each construction activity has a different time period, the research was designed to focus on the
unit of measurement. For example, a concrete place- long term trends in construction productivity.
ment activity may have a multifactor productivity
measured in cubic yards of concrete placed per unit
cost, while structural steel placement may be measured Results
in linear feet of steel placed per unit cost. Using rela-
tive instead of absolute values is one way to solve this As seen in Table 2, the mean change in labour produc-
issue. Thus, the percentage change in productivity tivity for the 200 activities was 30.9%, with a 95%
from 1976 to 1998 was used. conŽ dence interval of ± 9.2% and an annual comp-
Expected physical output data and crew formation ound rate of improvement of 1.2%. For the same activ-
data from the estimation manuals were used to calcu- ities, the mean change in multifactor productivity was
late each activity’s labour productivity (Formula 1 36.2%, with a 95% conŽ dence interval of ± 10.8%
below), and expected physical output data, labour and an annual compound rate of 1.4% (Table 3).
input cost data, and equipment input cost data from There are many speciŽ c examples of productivity
the estimation manuals were used to calculate each improvement in the sampled activities. For example,
activity’s multifactor productivity (Formula 2 below). wood framing of 2-inch ´ 6-inch beam and girders
(which beneŽ ted from an increased use of the pneu-
Labour
Expected Physical Output (Units) matic nail gun) experienced a 67.3% improvement in
Productivity, ———–––––––––––––––––––———
Workhour Requirements (Hrs) multifactor productivity and a 62.1% improvement
Year X =
in labour productivity between 1976 and 1998
(1)
(Means). Also, paint spraying of interior concrete walls
Multifactor experienced an 87.4% increase in multifactor produc-
Expected Physical Output (Units)
Productivity, ———–––––––––––––––––––——— tivity and a 76.1% increase in labour productivity
Year X = Labour Cost (1990$) + (Richardson). Concrete Ž nishing experienced a 52.0%
Equipment Cost (1990$) (2) increase in multifactor productivity and a 43.4%
420 Goodrum et al.

Table 2 Descriptive statistics of the % change in labour productivity from 1976 to 1998
95% ConŽ dence interval
of the mean
Mean ta df Sig. Min. Max. Std. dev. Lower Upper
+30.9% 6.7 199 .000 -50.0% 352.3% 65.9% +21.7% +40.1%
a
Test value = 0. Source: Means, Richardson, DCG.

Table 3 Descriptive statistics of the % change in multifactor productivity from 1976 to 1998

95% ConŽ dence interval


of the mean
Mean ta df Sig. Min. Max. Std. dev. Lower Upper
+36.2% 7.1 199 .00 -60.7% +353.7% 66.4% +25.4% +47.0%
a
Test Value = 0. Source: Means, Richardson, DCG.

increase in labour productivity (Dodge). Undoubtedly activities experienced the greatest improvement in
there are many factors that contributed to these labour and multifactor productivity. Electrical, mois-
improvements, from technological advances to work- ture and thermal protection, and woods and plastic
force related issues. activities experienced the smallest improvements in
The average change in labour and multifactor pro- labour and multifactor productivity. Further research
ductivity of the activities overall for each data source is required to determine the reasons for the differences
is shown in Table 4. All three manuals indicated that in productivity change by division.
labour and multifactor productivity increased from Although these results indicate that productivity, on
1976 to 1998. Means revealed a 0.8% compounded average, improved from 1976 to 1998, there were some
annual rate of improvement in labour productivity, exceptions. Labour productivity declined in 30 activi-
Richardson a 1.2% increase and Dodge a 1.8% increase. ties, was unchanged in 63 activities, and improved
For multifactor productivity, Means showed a 0.7% in 107 activities. There was a similar pattern in the
increase, Richardson a 0.7% increase and Dodge a changes for multifactor productivity: 57 activities expe-
2.9% increase. The different estimates of productivity rienced a decline in multifactor productivity, and 147
improvement are partially a re ection of the different experienced an improvement.
distribution of types of construction activity in different As illustrated in Figure 2, real wages in the US con-
divisions for each manual. struction industry, after being adjusted for in ation
Next, activities were grouped by CSI Masterformat using the Consumer Price Index,4 declined substan-
division, and the compounded annual rates of change tially over that couple of decades. There was also a
in labour and multifactor productivity were calculated decline in real wages in the total manufacturing indus-
for each division (Table 5). try, but the decline was more substantial in construc-
It is clear from this sample that different sectors of tion. Prior research suggested that this is a result of a
the construction industry experienced varying degrees decrease in unionization and prevailing wage agree-
of change in productivity. On average, sitework ments, deskilling of the workforce, increased safety, and

Table 4 Estimates of labour and multifactor productivity trends from 1976 to 1998 by data source

Data source Activity sample Labour productivity Multifactor productivity


size compounded annual compounded annual
rate rate
Means Building Construction
Cost Data 100 +0.8% +0.7%
Richardson Process Plant
Construction Estimating Standards 50 +1.2% +0.7%
Dodge Cost Guide 50 +1.8% +2.9%
Estimates of US construction productivity 421

Table 5 Compounded annual rate of change in labour and multifactor productivity for activities
by division from 1976 to 1998

Construction division Change in labour Change in multifactor


productivity 1976–1998 productivity 1976–1998
(compound annual rates) (compound annual rates)
Sitework +2.8% +2.4%
Doors & Windows +1.6% +1.8%
Metals +1.5% +1.0%
Finishes +1.2% +1.6%
Masonry +1.2% +0.8%
Concrete +1.1% +1.4%
Mechanical +1.0% +1.4%
Wood & Plastic +0.3% +0.4%
Moisture & Thermal Protection +0.2% +0.6%
Electrical +0.0% +0.8%
Source: Means, Richardson, DCG.

25.00 13.12% from 1976 to 1998, with a 95% conŽ dence


24.00
23.00 interval of 8.77% and an annual compounded rate of
Average Hourly Wage (1990$)

22.00
21.00 0.56%. The annual compounded rate of improvement
20.00
19.00 was 1.4% when no adjustment for real wage declines
18.00
17.00 was included. Although the Ž ndings indicate that real
16.00
15.00 wage declines affected multifactor productivity esti-
14.00
13.00 mates, further research is needed to fully understand
12.00
11.00 and quantify the factors behind these estimates.
10.00
1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998

Year Other trends supporting productivity


Total Manufacturing General Contractors improvement
Heavy Construction Special Trades
Thus far, the paper has presented evidence that
Figure 2 Real wage trends (according to average hourly construction productivity measured at the activity level
wages in 1990 Dollars) in construction from 1968 to 1999 and using a variety of data sources improved between
(source: Bureau of Labor Statistics)6 1976 and 1998. One would expect other industry
measures such as labour and capital utilization to have
increased numbers of migrant workers (Allmon et al., changed also in a manner to support such growth.
2000; Oppedahl, 2000). Much research is needed to fully understand how
Obviously, the real wage decline in the construction industry trends measured using other aggregate
industry has had a substantial effect on labour costs in measures impact construction productivity. Previous
multifactor productivity. This study examined this economic researchers have used macro changes in the
effect. To do this, labour costs as listed in the esti- capital/labour ratio, economies of scale, labour quality,
mation manuals were adjusted for in ation using the unionization, regional shifts, and changes in the mix
Engineering News Record’s CCI. Next, using historical of construction output to argue for various produc-
construction wage data from the US Bureau of Labor tivity trends in the past (Stokes, 1980; Allen, 1985).
Statistics, a real wage index was created to normalize Prior research found a positive relationship between
the real wage declines to 1990 levels (Goodrum, 2001). capital/labour ratio growth and productivity improve-
It should be noted that the real wage index is not meant ment in construction. Some believed that the decline
to represent an adjustment in labour quality. To adjust in capital/labour ratios5 is the biggest factor contribut-
for labour quality, an index would need to account for ing to aggregate estimates of productivity declines.
changes in education, experience and skills mix, among The capital/labour ratio increased in more than 70%
other traits. When this index was applied on top of the of the paper’s activities (Goodrum and Haas, 2002).
adjustment for in ation, the estimated improvements Furthermore, the increase in the capital/labour ratio
in multifactor productivity were not as great. With the and other technological measurements explained more
adjustment for real wage declines, overall multifactor than 36% of the variance in labour productivity, which
productivity in the 200 activities improved by only indicates that capital has replaced labour at the imme-
422 Goodrum et al.

diate point of construction (Goodrum and Haas, Goodrum, P. (2001) The Impact of Equipment Technology
2002). One anecdotal example is the increased used on Productivity in the U.S. Construction Industry.
of the pneumatic nail gun instead of the hand held Dissertation, The University of Texas.
hammer in carpentry and concrete trades over the past Goodrum, P. and Haas, C. (2002) Partial factor productivity
and equipment technology change at the activity level in the
couple of decades. It is important to note, however,
US construction industry. Journal of Construction Engineering
that capital/labour ratios explained only 10% of the and Management ASCE, in press.
improvements in multifactor productivity. This re ects Gordon, R. (1968) A new view of real investment in struc-
the labour saving bias of technical change on produc- tures, 1919–1966. Review of Economics and Statistics, 50,
tivity, and it suggests that rising capital/labour ratios 417–28.
alone do not contribute to productivity improvement. Gullickson, W. and Harper, M. (1999) Possible measurement
Relative prices, rates of technological progress, worker bias in aggregate productivity growth. Monthly Labor
skills mixes, improvements in management systems Review, February, 47–67.
and workforce relations also contribute to productivity. Haas, C., Tucker, R. and Eickmann, J. (1999) Prefabrication
and Preassembly Trends and Effects on the Construction
Workforce, Report No. 15, Center for Construction Industry
Conclusion Studies, The University of Texas at Austin.
Means Company (1976, 1998) Building Construction Cost
Doubts about changes in construction productivity are Data, Philip Waier, Sr. (ed.), Means Company, Inc.,
not new. Measurement of construction aggregate out- Kingston, MA.
Oppedahl, D. (2000) Understanding the (relative) rise and fall
put makes estimating productivity difŽ cult. These issues
of construction real wages. Chicago Federal Letter, 155, 1–4.
become less problematic when construction productiv- Pieper, P. (1989) Construction price statistics revisited. In
ity is measured at the activity level. This paper has pre- Jorgenson, D. and Landau, R. (eds), Technology and Capital
sented evidence that labour and multifactor productivity Formation , The MIT Press, Cambridge, MA.
increased from 1976 to 1998 in the US construction Pieper, P. (1990) The measurement of construction prices:
industry when using activity level data. What remains retrospect and prospect. In Berndt, E. and Triplet, J. (eds),
unclear is why these improvements are not re ected Fifty Years of Economic Measurement: The Jubilee of
when aggregate level data are used. One possible expla- the Conference on Research in Income and Wealth, The
nation for the discrepancy could be the problems with University of Chicago Press, Chicago, IL.
Rees, A. (1980) Improving productivity measurement. The
the aggregate output measurements, although other rea- American Economic Review, 70(2).
sons may exist that are yet to be identiŽ ed, like increas- Richardson (1976, 1998) Richardson Process Plant Construction
ing project complexity. Until government, academic, Estimating Standards, Richardson Engineering Services,
and industry leaders reach an agreement about the Mesa, AZ.
aggregate output measures, the debate over the increase RoseŽ elde, S. and Quinn Mills, D. (1979) Is construction
or decrease in productivity in the US construction technologically stagnant? In Lange, J. and Quinn Mills, D.
industry will continue. Considering the size of the con- (eds), The Construction Industry: Balance Wheel of the
struction industry and its impact on the overall US econ- Economy, Lexington Books, Lexington, MA.
omy, further research on this issue is warranted. Rosenblum, L. (2000) Re: US Construction Productivity,
Division of Productivity Research, US Bureau of Labor
Statistics; electronic correspondence, 31 October.
Stokes, H. (1981) An examination of the productivity decline
References in the construction industry. The Review of Economics and
Statistics, 63(4), 495–502.
Allen, S.G. (1985) Why construction industry productivity is Teicholz, P. (2001) Discussion: US Construction Labor
declining. The Review of Economics and Statistics, 117(4), Productivity Trends, 1970–1998. Journal of Construction
661–5. Engineering and Management ASCE, 127(5), 427–9.
Allmon, E., Haas, C., Borcherding, J. and Goodrum, P. Thomas, H. and Yiakoumis, I. (1987) Factor model of con-
(2000) US Construction Labor Productivity Trends, struction productivity. Journal of Construction Engine-
1970–1998. Journal of Construction Engineering and Manage- ering and Management ASCE, 113(4), 626–44.
ment ASCE, 126(2), 97–104. US Department of Commerce (1998) Value of Construction Put
Business Roundtable (1983) More Construction for the Money, in Place, Construction Report Series C30, US Department
The Business Roundtable, New York. of Commerce, Bureau of the Census, Washington, DC.
Dacy, D. (1965) Productivity and price trends in construction US Department of Commerce Characteristics of New Housing,
since 1947. Review of Economics and Statistics, 47, 406–11. Construction Report Series C25, US Department of
DCG (1976, 1998) Dodge Cost Guides, McGraw-Hill, New Commerce, Washington, DC.
York. WSJ (1995) New recipe for the GDP leaves sour taste. Wall
Fleming, M. (1996) The statistics corner: the GDP revisions Street Journal, 12 December, C1.
and understanding sluggish productivity growth. Business Young, A. (1992) Alternative measures of change in real out-
Economics , April, 62–5. put and prices. Survey of Current Business, 73, 31–41.
Estimates of US construction productivity 423

Notes allow a direct comparison with real wages in construc-


tion and manufacturing.
1 Multifactor productivity is the ratio of output to more 5 Declining capital/labour ratios were cited as a major
than one input factor. Many US governmental agencies, contributor in productivity declines from 1968 to 1978
e.g. the Bureau of Labor Statistics, use a measure of by Allen (1985). However, Pieper (1989) showed
multifactor productivity that may also be considered capital/labour ratios actually increased over this time
a total factor productivity estimate, since it uses a period, and were indicators of productivity growth in
combined measure of all inputs including labour, capital construction.
and materials. A detailed description of the methodology 6 The general contractors, heavy construction, and special
that the BLS uses in estimating multifactor productivity trades are classiŽ cations per the North American Industry
is given in the US Bureau of Labor Statistics Handbook ClassiŽ cation System (United States OfŽ ce of Manage-
of Methods, Bulletin 2497, April 1997. The present ment and Budget 1997). The general contracting sub-
paper also measures multifactor productivity, but it uses sector comprises establishments primarily responsible
a simpliŽ ed procedure that includes the input measures for the entire construction (i.e. new work, additions,
of labour and equipment only (see later sections). alterations and repair) of building projects. Builders,
2 Input cost indices estimate the change in the price developers, and general contractors, as well as land subdi-
of construction for the contractor, while output price viders and land developers are included in this subsector.
indices estimate the price that an owner pays to buy a Establishments identiŽ ed as construction management
constructed facility. Due to a lack of construction output Ž rms for building projects are also included. The heavy
price indexes, typically cost indexes have been used construction subsector group includes establishments
instead for many sectors of the construction industry. that engage in the construction of heavy engineering
3 The CCI is based on the price of ‘25 cwt of structural industrial projects (except buildings). For example,
steel shapes at the mill price prior to 1996 and the fabri- highway, powerplant, and pipeline contractors are
cated 20-city price from 1996, plus 1.128 tons of included in this group. Special trade contractors sub-
Portland cement at the 20-city price, plus 1088 board- sector engages in specialized construction activities such
ft of 2 ´ 4 lumber at the 20-city price, and 200 hours as plumbing, painting, and electrical work. Total manu-
of common labour from 20 cities throughout the USA’. facturing includes establishments engaged in the mechan-
4 Although the paper adjusts construction wages for in a- ical, physical, or chemical transformation of materials,
tion using the Engineering News Record CCI in its substances or components into new products. Construc-
analyses, construction wages in Figure 2 were adjusted tion establishments are not included in the total manu-
for in ation using the Consumer Price Index (CPI) to facturing subsector.

You might also like